predicting native pasture growth in the victoria river ... · caz smith who are the current...

Predicting native pasture growth in the Victoria River

District of the Northern Territory

A thesis submitted by

Michael D. Cobiac B. App. Sc. (Ag.)

to the

School of Agriculture, Food and Wine

Faculty of Sciences

The University of Adelaide, Australia

as fulfilment of the requirements of

Doctor of Philosophy

December 2006

ii

Declaration of Originality

This work contains no material which has been accepted for the award of any other degree

or diploma in any university or other tertiary institution and, to the best of my knowledge

and belief, contains no material previously published or written by another person, except

where due reference has been made in the text.

I give consent to this copy of my thesis, when deposited in the University Library, being

made available in all forms of media, now or hereafter known.

Signed:

Michael Cobiac

Date:

iii

Acknowledgments

The successful completion of this study has involved considerable assistance from many

people throughout the planning, research, analysis, and reporting phases. The efforts of the

following people are gratefully acknowledged.

Funding: The fieldwork and GRASP model calibration reported in this study were funded

by Meat and Livestock Australia’s (MLA’s) North Australia Program. In addition, MLA’s

provision of a Junior Research Fellowship allowed me to concentrate full time on

finalising this thesis. Their contribution has been central to the completion of this work.

My principal supervisor Dr Bill Bellotti (The University of Adelaide) for encouraging me

down the post-graduate path, always being available for advice when needed during the

preparation of this thesis, and his quiet insistence that I produced work of a high standard.

The role of supervisor is crucial to a student’s candidature, and I have been very fortunate

having Bill’s guidance throughout my study.

My co-supervisor Ken Day (Qld Dept. of Natural Resources and Mines) for a great many

things. He provided advice on field data collection, collaborated with me during calibration

of the GRASP model, and reviewed my work many times. Mostly, however, I thank Ken

for his undying faith, patience and friendship throughout the course of this study. This

work would never have been completed without his commitment, and for that I am

indebted.

Dr Greg McKeon (Qld Dept. of Natural Resources and Mines) for his repeated efforts to

educate me about the GRASP model, his endless enthusiasm for the quest to improve

landscape use by the northern Australian pastoral industry, and leading me to think in ways

I never knew were possible. It has been a privilege to work with a scientist of his calibre.

Former NT Dept. of Primary Industry and Fisheries (now DPIFM) staff: Tom Stockwell

for the job; Rodd Dyer for being a great friend and demonstrating work ethic like I’d never

seen before; and Anne Lyon, Bruno Hogan, Linda Cafe and the staff of Victoria River

Research Station (Kidman Springs) for being good mates and for assistance with collection

of data in the field, sometimes under very difficult conditions.

iv

Neil MacDonald, Robyn Cowley, Trudi Oxley, Kieren McCosker, Annemarie Huey and

Caz Smith who are the current generation of bright minds at Katherine Research Station

furthering our understanding of sustainable pastoralism in the semi-arid tropics of the NT.

It is a great motivating factor to know the outcomes of this work will (indeed already do)

play a significant role in ongoing pastoral research.

The NT Cattlemans Association and the producers of the Victoria River District for

allowing me to conduct field research on their properties, for access to confidential

property data, and for their hospitality and open-minded discussions over many years. This

thesis is ultimately for them, the people who manage the landscape.

Shafiq, Ali, Eun-Young, Yasmine, Bagarath, Juan, Vahid and Bandara: the international

postgraduate students at The University of Adelaide’s School of Agriculture, Food and

Wine for befriending me, broadening my mind, and who present their theses in English - to

them a second language. As challenging as things got sometimes, nothing in my study was

as hard as that.

The many other people with whom I interacted during this study and who provided

encouragement when the path forward seemed long, arduous and unclear.

Most importantly, I thank Cath for putting up with my long absences, understanding my

need to see this study through to completion, and supporting me when it was needed most.

Her faith in me carried me through many moments of doubt.

v

Abstract

Pastoralism is the major economic activity in the Victoria River District (VRD), and is

dependent on sustainable pasture use. Analysing grazing practices for sustainability

requires knowledge of annual pasture production, but little quantitative data is available. A

study was undertaken to develop the capacity for predicting native pasture growth in the

VRD using systems modelling. Twenty one field sites were studied for two years using a

standard methodology, and the Grass Production (GRASP) model was calibrated using this

field data. End of growing season total standing dry matter (TSDM) was well predicted

(mean = 2513kg/ha, r2(1:1) = 0.966, RMSE = 132kg/ha, and 98% of predictions within

measurement variance).

Developing generic parameters for common soil and pasture types allowed extrapolation of

the model. Predictive skill declined when using generic parameters (r2(1:1) = -0.265,

RMSE = 807kg/ha and 64% of predictions within measurement variance). However,

observation and prediction means were very similar, indicating that generic parameters are

suitable for broad scale applications, but site-specific parameters are necessary if a high

degree of accuracy is required. Parameters controlling plant water uptake largely determine

pasture growth in low rainfall years, while nitrogen uptake and dilution parameters limit

growth in high rainfall years. Pasture growth is constrained by nitrogen supply in 91% of

seasons in the northern VRD, and in 25% of seasons in the drier south.

Example applications of the model were demonstrated. Current and expected future levels

of pasture utilisation in the district were calculated, showing a current average of 16%,

rising to an expected 20% in the next decade. These levels are within the safe utilisation

rates recommended for the region. Economic analysis shows positive returns ($4.54

million per year) from pasture augmentation with introduced legumes if past problems with

establishment and persistence can be overcome.

Model performance would be improved by accounting for simultaneous wetting of the

entire profile in cracking clay soils, calculating growth of perennial and annual pasture

species separately, and simulating variation in nitrogen uptake and dilution between years.

Incorporation of these processes must be balanced against the increased complexity of the

model and the additional data required for calibration.

vi

Table of Contents

Declaration of Originality ...................................................................................................ii

Acknowledgments .............................................................................................................. iii

Abstract.................................................................................................................................v

Table of Contents ................................................................................................................vi

List of Figures................................................................................................................... viii

List of Tables .....................................................................................................................xiv

List of Plates ................................................................................................................... xviii

1.0 Introduction................................................................................................................1

2.0 Review of literature....................................................................................................5

2.1 A brief description of the Victoria River District ................................................................ 5 2.2 The pastoral industry............................................................................................................ 9 2.3 Current understanding of factors influencing pasture growth............................................ 11 2.4 A modelling approach to assessing pasture growth ........................................................... 20 2.5 Testing the performance of a systems model..................................................................... 28 2.6 The role of systems modelling in grazing land management............................................. 33 2.7 Conclusions........................................................................................................................ 34 2.8 Outline of the study ahead ................................................................................................. 34

3.0 A field study of native pasture growth in the VRD...............................................36

3.1 Introduction........................................................................................................................ 36 3.2 Rationale for obtaining data to calibrate the GRASP pasture growth model..................... 37 3.3 Methods of data collection................................................................................................. 38 3.4 Field results ........................................................................................................................ 50 3.5 Discussion of field study results ...................................................................................... 105 3.6 Conclusions...................................................................................................................... 117

4.0 Analysis of field measurements using a systems modelling approach ..............119

4.1 Introduction...................................................................................................................... 119 4.2 An overview of the GRASP pasture growth model ......................................................... 121 4.3 Method for deriving model parameters and calibrating GRASP ..................................... 127 4.4 Results of calibration: Final model parameters................................................................ 133 4.5 Comparing model outputs with field data........................................................................ 143 4.6 Discussion of the calibration procedure and modelling results........................................ 156 4.7 Conclusions...................................................................................................................... 169

5.0 Testing the performance of GRASP for application to the wider landscape ...170

5.1 Introduction...................................................................................................................... 170

vii

5.2 Generic parameters suitable for extrapolation across the landscape ................................171 5.3 Testing model performance using independent data ........................................................177 5.4 Discussion of independent validation of GRASP.............................................................186 5.5 Conclusions ......................................................................................................................189

6.0 Determining parameters most influential on predictions of long-term pasture

growth ...............................................................................................................................190

6.1 Introduction ......................................................................................................................190 6.2 Modelling year-to-year variability in pasture growth across a climate gradient ..............191 6.3 Sensitivity of model predictions to changes in value of influential parameters ...............200 6.4 The effect of trees on predictions of pasture growth ........................................................205 6.5 Discussion.........................................................................................................................208 6.6 Conclusions ......................................................................................................................213

7.0 Implications for analysing grazing practices in the VRD: examples of model

application ........................................................................................................................214

7.1 Introduction ......................................................................................................................214 7.2 Pasture utilisation in the VRD ..........................................................................................215 7.3 Potential benefits of alleviating the nitrogen limitation to pasture growth.......................228 7.4 Discussion.........................................................................................................................241 7.5 Conclusions ......................................................................................................................248

8.0 Integrating discussion and final conclusions of study ........................................249

8.1 Introduction ......................................................................................................................249 8.2 The capacity to predict pasture growth in the VRD .........................................................249 8.3 Limitations of the systems modelling approach used in this study ..................................252 8.4 Recommendations for future work ...................................................................................256 8.5 Final conclusion................................................................................................................257

Appendices........................................................................................................................258

References.........................................................................................................................311

viii

List of Figures

Chapter 1

Figure 1.1 The location of the Victoria River District of the Northern Territory. ............................................ 2

Figure 1.2 Illustration of the structure of this thesis. ........................................................................................ 4

Chapter 2

Figure 2.1 Seasonal rainfall (July to June, mm) at Victoria River Downs Station for the period 1900/01 to

2003/04. The horizontal dashed line represents the median value (639mm) (Source: DataDrill 2005). ........... 6

Figure 2.2 Land tenure in the Victoria River District in 2004 (data sourced from NT Dept of Lands). ........... 8

Figure 2.3 Total cattle population and annual turnoff (number of animals sold) for the Victoria River

District 1883 – 2004 (S. Murti pers.comm., based on Australian Bureau of Statistics and NT Office of

Resource Development data). Dashed lines represent interpolation across periods where no data is available.

Early data has been left as isolated points as no basis for interpolation is available. ...................................... 10

Figure 2.4 Annual trends in LI, TI, MI and GI values at Katherine, NT for: a) tropical grasses, and b) tropical

legumes; c) the relationship between mean daily temperature (0F) and fractional dry matter production in

three groups of pastures (Fitzpatrick and Nix 1970)........................................................................................ 23

Figure 2.5 a) Thermal response curves of tropical grasses and legumes including the solid line used in the

study of McCown (1981a); b) moisture index function of McCown et al. (1974). ......................................... 25

Figure 2.6 Structure of the water balance model and pasture sub-model in GRASP (Littleboy and McKeon

1997)................................................................................................................................................................ 27

Figure 2.7 Examples of GRASP model output for individual seasons, and over 15 years from Johnston

(1996). ............................................................................................................................................................. 28

Figure 2.8 Examples of two approaches to model testing from Mitchell and Sheehy (1997). ....................... 32

Chapter 3

Figure 3.1 The structure of Chapter 3. ............................................................................................................ 37

Figure 3.2 The five main locations of the study sites within the Victoria River District (shaded area). Dashed

lines represent approximate average annual rainfall isohyets (Source: adapted from BoM data). .................. 40

Figure 3.3 Design and layout of study site showing cells (dashed line squares) and an example pattern of

quadrat placement (solid line squares) for each pasture measurement (H1 to H8). Quadrat location for each

pasture measurement was randomised for each site. ....................................................................................... 42

Figure 3.4 Time-series of management and data collection from each site over the study period.................. 51

Figure 3.5 Maximum and minimum daily temperatures (0C) at Mt Sanford (17012’S, 130036’E) over the

study period (upper graph) compared to the 7-day moving averages of 47-year values (1957/58 – 2003/04);

ix

and measured daily rainfall averaged across Sites 1-6 (lower graph). Temperature data derived from

DataDrill (2005)............................................................................................................................................... 54

Figure 3.6 Total monthly measured rainfall at Mt Sanford averaged across Sites 1-6, and 47-year median

values (1957/58 – 2003/04) derived from DataDrill (2005). ........................................................................... 54

Figure 3.7 Maximum and minimum daily temperatures (0C) at Kidman Springs (16006’S, 131000’E) over the


and measured daily rainfall averaged across Sites 7-12 for 1993/94, Sites 7-14 for 1994/95 and Sites 13-14

for 1995/96 (lower graph). Temperature data derived from DataDrill (2005)................................................. 55

Figure 3.8 Total monthly measured rainfall at Kidman Springs averaged across Sites 7-12 for 1993/94, Sites

7-14 for 1994/95 and Sites 13-14 for 1995/96, and 47-year median values (1957/58 – 2003/04) derived from

DataDrill (2005)............................................................................................................................................... 55

Figure 3.9 Maximum and minimum daily temperatures (0C) at Victoria River Downs (16024’S, 131006’E)

over the study period (upper graph) compared to the 7-day moving averages of 47-year values (1957/58 –

2003/04); and measured daily rainfall at the station homestead (lower graph). Temperature data derived from

DataDrill (2005)............................................................................................................................................... 56

Figure 3.10 Total monthly measured rainfall at Victoria River Downs homestead, and 47-year median values

(1957/58 – 2003/04) derived from DataDrill (2005). ...................................................................................... 56

Figure 3.11 Maximum and minimum daily temperatures (0C) at Rosewood (16030’S, 129000’E) over the


and measured daily rainfall at the station homestead (lower graph). Temperature data derived from DataDrill

(2005)............................................................................................................................................................... 57

Figure 3.12 Total monthly measured rainfall at Rosewood homestead, and 47-year median values (1957/58 –

2003/04) derived from DataDrill (2005).......................................................................................................... 57

Figure 3.13 Maximum and minimum daily temperatures (0C) at Auvergne (15024’S, 130000’E) over the


and measured daily rainfall at the station homestead (lower graph). Temperature data derived from DataDrill

(2005)............................................................................................................................................................... 58

Figure 3.14 Total monthly measured rainfall at Auvergne homestead, and 47-year median values (1957/58 –

2003/04) derived from DataDrill (2005).......................................................................................................... 58

Figure 3.15 Field-measured soil water contents of sites located on red earths overlying basalt. .................... 63

Figure 3.16 Field-measured soil water contents of sites located on red earth overlying limestone. ............... 65

Figure 3.17 Field-measured soil water contents of sites located on cracking clays overlying basalt.............. 67

Figure 3.18 Field-measured soil water contents of sites located on cracking clays of alluvial origin. ........... 71

Figure 3.19 The observed phases of plant growth during the study period for: a) the annual short grass

Brachyachne convergens; b) the annual mid-height grass Iseilema vaginiflorum; c) the perennial tussock

grass Astrebla pectinata; and d) the perennial tuft grass Chrysopogon fallax................................................. 80

Figure 3.20 Pasture composition of three sites dominated by barley Mitchell grass. Error bars indicate the

standard error of the site mean for total standing dry matter at each sampling time (harvest). ....................... 82

x

Figure 3.21 Pasture nitrogen contents (dashed lines) and nitrogen uptake (vertical bars) of three sites

dominated by barley Mitchell grass................................................................................................................. 83

Figure 3.22 Plots of total standing dry matter (TSDM) against: a) total plant cover; and b) plant height for

sites dominated by barley Mitchell grass......................................................................................................... 84

Figure 3.23 Pasture composition of three sites dominated by ribbon grass. Error bars indicate the standard

error of the site mean for total standing dry matter at each sampling time (harvest)....................................... 87


dominated by ribbon grass............................................................................................................................... 88


sites dominated by ribbon grass....................................................................................................................... 90

Figure 3.26 Pasture composition of three sites dominated by other perennial grass species. Error bars

indicate the standard error of the site mean for total standing dry matter at each sampling time (harvest). .... 92


dominated by other perennial grass species..................................................................................................... 93


sites dominated by other perennial grass species............................................................................................. 95

Figure 3.29 Pasture composition of sites dominated by annual short grass species. Error bars indicate the

standard error of the site mean for total standing dry matter at each sampling time (harvest). ....................... 97

Figure 3.30 Pasture nitrogen contents (dashed lines) and nitrogen uptake (vertical bars) of sites dominated by

annual short grass species. ............................................................................................................................... 98


sites dominated by annual short grass species. ................................................................................................ 99

Figure 3.32 Pasture composition of sites dominated by forb species. Error bars indicate the standard error of

the site mean for total standing dry matter at each sampling time (harvest).................................................. 102


forb species. ................................................................................................................................................... 103


sites dominated by forb species. .................................................................................................................... 104

Figure 3.35 Relationships between perennial grass basal area (PGBA) and end of growing season standing

dry matter (SDM) of a) perennial grasses; b) annual grasses and forbs; and c) total pasture........................ 115

Chapter 4

Figure 4.1 The structure of Chapter 4. .......................................................................................................... 120

Figure 4.2 The four phases of systems analysis (bolded text in box) and an indication of where the

components of this study fit within this framework (after Grant et al. 1997)................................................ 120

Figure 4.3 Relationship between accumulated pasture transpiration and nitrogen uptake (from Littleboy and

McKeon 1997). The user-defined coefficients for stored plant N reserves, N uptake per 100mm of

xi

transpiration, and maximum N available for uptake shown in this figure are realistic for the semi-arid tropics,

but are examples only. ................................................................................................................................... 124

Figure 4.4 Flow of dry matter through the biomass pools (bolded) in the pasture system (modified from

Littleboy and McKeon 1997). Processes that transfer dry matter from one biomass pool to another are

presented in italics. ........................................................................................................................................ 125

Figure 4.5 Typical daily biomass accumulation curve for native pasture in the Victoria River District,

including approximate times of field measurements and the parameters in GRASP that are calibrated from

data collected at these times........................................................................................................................... 133

Figure 4.6 Examples of time-series plots of prediction curves (lines) with observed values (points, including

95% confidence limits in the TSDM plots). Data presented are from sites of different soil type and pasture

species composition: Site 4 (forbs on basalt clay); Site 8 (ribbon grass on alluvial clay); Site 1 (annual short

grasses on basalt red earth); Site 14 (other perennial grasses on limestone red earth); Site 5 (barley Mitchell

grass on basalt clay); Site 12 (other perennial grasses on limestone red earth); Site 3 (barley Mitchell grass on

basalt clay); and Site 15 (forbs on alluvial clay)............................................................................................ 148

Figure 4.7 Observed vs. predicted data for all sites. Model predictions are the results of using individual site

parameter sets. Variables presented are total standing dry matter (TSDM), nitrogen uptake (N uptake), plant

available water content in the 0-50cm layer of the soil (PAWC), and green plant cover. Associated statistics

are presented in Table 4.7. ............................................................................................................................. 151

Figure 4.8 Deviation of predictions (points) from observed values (line of zero deviation) of total standing

dry matter for all sites and years. Predictions are generated by GRASP using the individual parameter sets

presented in Table 4.1 to Table 4.6. Dashed lines indicate the envelope of acceptable precision, equal to the

average magnitude of measurement variance (±35% of observation values). ............................................... 155

Figure 4.9 Deviation of prediction values (points) from their corresponding observations of TSDM (x on line

of zero deviation and including 95% confidence limits) for Site 4 and Site 8. .............................................. 155

Figure 4.10 Deviation of predictions (points) from observed values of total standing dry matter (x on line of

zero deviation and including 95% confidence limits) for all observations less than 500kg/ha...................... 164

Chapter 5


Figure 5.2 Illustration of the procedure for assembling data for the independent validation of GRASP, using

the annual short grasses group as an example. S01Y1 refers to Site 1, Year 1; S01Y2 refers to Site 1, Year 2;

and so on. ....................................................................................................................................................... 180

Figure 5.3 Observed vs. predicted TSDM for independent model validation using four approaches to

developing generic parameter sets. ................................................................................................................ 183


dry matter (TSDM) during independent validation. Predictions are generated by GRASP using four

approaches to developing generic group parameter sets (Table 5.2). Dashed lines indicate the envelope of

xii

acceptable precision, equal to the average magnitude of measurement variance (±35% of observation values).

....................................................................................................................................................................... 184

Chapter 6


Figure 6.2 Time-series of simulated seasonal pasture growth (SSPG, July to June) over a 45-year period

(1959/60 to 2003/04) using the Regional VRD parameter set at three locations in the VRD. ....................... 198

Figure 6.3 Probability distribution of simulated seasonal pasture growth (SSPG, July to June) over a 45-year

period (1959/60 to 2003/04) using the Regional VRD parameter set at three locations in the VRD. The

horizontal dashed line represents the median (50th percentile) value............................................................. 198

Figure 6.4 Probability distributions of simulated seasonal pasture growth (SSPG, July to June) over a 45-

year period (1959/60 to 2003/04) using a) five Species group parameter sets; and b) three Soil group

parameter sets at Victoria River Downs. Horizontal dashed lines represents median (50th percentile) values.

....................................................................................................................................................................... 198

Figure 6.4...................................................................................................................................................... 199

Figure 6.5 a) Time-series of seasonal rainfall (July to June); and b) relationship between seasonal rainfall

and simulated seasonal pasture growth (SSPG) over a 45-year period (1959/60 to 2003/04) using Regional

VRD parameters at three locations in the VRD. The plotted regression is for seasons when SSPG was less

than the upper limit of 3345kg/ha.................................................................................................................. 199

Figure 6.6 a) Time-series of cumulative seasonal transpiration (July to June); and b) relationship between

seasonal transpiration and simulated seasonal pasture growth (SSPG) over a 45-year period (1959/60 to

2003/04) using Regional VRD parameters at three locations in the VRD. The plotted regression is for seasons

when SSPG was less than the upper limit of 3345kg/ha................................................................................ 199

Figure 6.7 a) Time-series of nitrogen uptake (July to June); and b) relationship between N uptake and

simulated seasonal pasture growth (SSPG) over a 45-year period (1959/60 to 2003/04) using Regional VRD

parameters at three locations in the VRD. The plotted regression is for seasons when N uptake was less than

the upper limit of 24kg/ha/year...................................................................................................................... 199

Figure 6.8 Sensitivity of simulated seasonal pasture growth results (SSPG) using a standard parameter set at

Victoria River Downs to changes in values of transpiration-use-efficiency, green yield at 50% green cover,

and soil water index at which growth stops. The horizontal dashed lines represent the median (50th percentile)

value. ............................................................................................................................................................. 203

Figure 6.9 Sensitivity of simulated seasonal pasture growth (SSPG) using a standard parameter set at

Victoria River Downs to changes in values of maximum N uptake, N content at which growth stops, and N

uptake per 100mm of transpiration. The horizontal dashed lines represent the median (50th percentile) value.

....................................................................................................................................................................... 204

Figure 6.10 Time-series of simulated seasonal pasture growth (SSPG, July to June) in the absence and the

presence of trees over 45 years (1959/60 to 2003/04) at a) Auvergne (mean annual rainfall = 900mm); and b)

xiii

Inverway (mean annual rainfall = 577mm). Tree basal area (TBA) was set at 6.0m2/ha at Auvergne and

2.0m2/ha at Inverway. .................................................................................................................................... 207

Figure 6.11 Probability distribution of simulated seasonal pasture growth (SSPG) in the absence and the

presence of trees over 45 years (1959/60 to 2003/04) at two locations in the VRD. ..................................... 207

Figure 6.12 Presentation yield of tallgrass pasture at the end of growing season in lightly grazed native

pasture woodlands at Manbulloo, near Katherine NT (McIvor et al. 1994) and GRASP predictions of

seasonal pasture growth (July to June) using Regional VRD parameters and tree basal area of 7.0m2/ha..... 207

Chapter 7


Figure 7.2 Relationship between intake at long-term stocking rate and median simulated seasonal pasture

growth (SSPG) across 10 properties in the VRD using stocking rate data provided in the 1997 survey (Smith

1998) and pasture growth predicted using the Soil x Species approach to developing generic parameters. The

dashed line represents a similar relationship from three regions in Queensland (Hall et al. 1998). .............. 223

Figure 7.3 Effect of climate variability on annual pasture utilisation over a 45 year period (1959/60 to

2003/2004) when a constant stocking rate is maintained (district average = 9.2AE/km2, Table 7.4)............ 227

Figure 7.4 Effect of climate variability on annual stocking rate over a 45 year period (1959/60 to 2003/2004)

when a constant utilisation rate is maintained (district average = 16.3%, Table 7.4). ................................... 227

Figure 7.5 a) Time-series; and b) probability distribution of simulated seasonal pasture growth (SSPG) in the

absence of trees over 45-year period (1959/60 to 2003/04) using Regional VRD parameters at Auvergne

when maximum nitrogen supply (MaxN) is limited (24kg/ha, the observed district average) and theoretically

unlimited (96kg/ha)........................................................................................................................................ 233

Figure 7.6 Relationship between simulated seasonal pasture growth (SSPG, July to June) and a) seasonal

transpiration (July to June); and b) total nitrogen uptake in the absence of trees over 45 years (1959/60 to

2003/04) at Auvergne when maximum nitrogen supply (MaxN) is limited (24kg/ha, the observed district

average) and theoretically unlimited (96kg/ha). ............................................................................................ 234

Appendices

Figure 9.1 Pasture composition results not presented in Chapter 3. Error bars indicate the standard error of

the site mean for total standing dry matter at each sampling time (harvest). (Figure continued overleaf) .... 280

Figure 9.2 Pasture nitrogen contents (dashed lines) and nitrogen uptake (vertical bars) of sites not presented

in Chapter 3. (Figure continued overleaf) ...................................................................................................... 282

xiv

List of Tables

Chapter 2

Table 2.1 Climate data at Victoria River Downs Station for the period 1900/01 to 2003/04 (source: DataDrill

2005).................................................................................................................................................................. 6

Table 2.2 Summary of existing pasture production data for the VRD. ........................................................... 19

Chapter 3

Table 3.1 Matrix of land systems by parent material and soil type (adapted from Stewart et al. 1970). ........ 39

Table 3.2 General description of the 21 study sites at 5 locations in the Victoria River District (VRD)........ 44

Table 3.3 Measured seasonal rainfall (July - June) over the study period, the 47-year (1957/58 – 2003/04)

median value, and rainfall percentiles. Median data calculated from DataDrill (2005). ................................. 53

Table 3.4 Monthly values for radiation, evaporation and vapour pressure deficit at Victoria River Downs

during the study period (source: DataDrill 2005). ........................................................................................... 60

Table 3.5 Soil description for sites located on red earths overlying basalt. .................................................... 62

Table 3.6 Physical properties of soils at sites located on red earths overlying basalt. .................................... 62

Table 3.7 Soil description for sites located on red earths overlying limestone. .............................................. 64

Table 3.8 Physical properties of soils at sites located on red earths overlying limestone. .............................. 64

Table 3.9 Soil description for sites located on cracking clays overlying basalt. ............................................. 66

Table 3.10 Physical properties of soils at sites located on cracking clays overlying basalt. ........................... 67

Table 3.11 Soil description for sites located on cracking clays of alluvial origin........................................... 69

Table 3.12 Physical properties of soils at sites located on cracking clays of alluvial origin........................... 70

Table 3.13 Example to demonstrate extrapolating soil moisture values to missing data points. .................... 73

Table 3.14 Estimated dates of initiation of pasture growth for each study location, based on criteria of

receiving both 50mm of rain within 14 days and 75mm within 28 days of the initiation date. ....................... 77

Table 3.15 Main pasture and tree species present during the study period for sites dominated by barley

Mitchell grass (Astrebla pectinata).................................................................................................................. 81

Table 3.16 Results for N and P status of pasture at sites dominated by barley Mitchell grass. ...................... 84

Table 3.17 Some important pasture variables at sites dominated by barley Mitchell grass. ........................... 84

Table 3.18 Main pasture and tree species present during the study period for sites dominated by ribbon grass

(Chrysopogon fallax). ...................................................................................................................................... 86

Table 3.19 Results for N and P status of pasture at sites dominated by ribbon grass. .................................... 89

Table 3.20 Some important pasture variables at sites dominated by ribbon grass. ......................................... 89

xv

Table 3.21 Main pasture and tree species present during the study period for sites dominated by other

perennial grass species..................................................................................................................................... 91

Table 3.22 Results for N and P status of pasture at sites dominated by other perennial grass species............ 94

Table 3.23 Some important pasture variables at sites dominated by other perennial grass species. ............... 94

Table 3.24 Main pasture and tree species present during the study period for sites dominated by annual short

grass species..................................................................................................................................................... 96

Table 3.25 Results for N and P status of pasture at sites dominated by annual short grass species. ............... 99

Table 3.26 Some important pasture variables at sites dominated by annual short grass species..................... 99

Table 3.27 Main pasture and tree species present during the study period for sites dominated by forb species.

....................................................................................................................................................................... 101

Table 3.28 Results for N and P status of pasture at sites dominated by forb species. ................................... 104

Table 3.29 Some important pasture variables at sites dominated by forb species......................................... 104

Table 3.30 Variability in rainfall at the five study locations in the VRD. Rainfall data from DataDrill (2005).

....................................................................................................................................................................... 106

Table 3.31 Summary of results for important soil variables. ........................................................................ 109

Table 3.32 Summary of soil chemistry results for each of the major soil types in this study. ...................... 110

Table 3.33 Summary of field results for important pasture variables. .......................................................... 112

Table 3.34 Variance in field measurements of total standing dry matter (TSDM) for 21 sites over two years.

Values represent 95% confidence limits expressed as a proportion of the harvest mean. ............................. 113

Chapter 4

Table 4.1 Site-by-year calibrated GRASP soil parameters. (Table continued overleaf) ............................... 135

Table 4.2 Site-by-year calibrated GRASP sward structure parameters. (Table continued below)................ 138

Table 4.3 Site-by-year calibrated GRASP plant growth parameters. (Table continued below).................... 139

Table 4.4 Site-by-year calibrated GRASP nitrogen parameters. (Table continued below)........................... 140

Table 4.5 Tree parameters for Site 19. .......................................................................................................... 141

Table 4.6 Site-by-year calibrated GRASP detachment parameters. (Table continued below)...................... 142

Table 4.7 Results of statistical comparison of predictions and observed values for all sites, using individual

site parameter sets. Data presented is total standing dry matter (TSDM), plant available water content of the

0-50cm layer of the soil (PAWC), nitrogen uptake (N uptake), and green plant cover (Cover).................... 153

Table 4.8 Summary of the number of predictions of TSDM that fell outside the envelopes of acceptable

precision......................................................................................................................................................... 156

Table 4.9 Summary of parameter values derived during calibration of GRASP to the study sites. .............. 158

Table 4.10 Summary of model calibration results for total standing dry matter. .......................................... 162

xvi

Chapter 5

Table 5.1 Matrix of study sites (numbers in table) as they relate to soil and species groups for development

of generic parameter sets. .............................................................................................................................. 172

Table 5.2 Generic parameter values for Soil, Species, and Regional VRD groups. (Table continued overleaf)

....................................................................................................................................................................... 175

Table 5.3 Source of parameters used to calculate partial estimates for independent model validation. This

table is based upon the matrix presented in Table 5.1. .................................................................................. 181

Table 5.4 Results of comparing model predictions with observed values of TSDM for independent validation

using four approaches to developing generic parameter sets......................................................................... 185

Table 5.5 Summary of predictions of TSDM that fell outside the envelopes of acceptable precision (95%

confidence limits of the corresponding observation) using four approaches to developing generic parameter

sets. ................................................................................................................................................................ 185

Table 5.6 Summary of results of independent validation of GRASP (including comparative results from

calibration in Chapter 4) for end of wet season total standing dry matter. .................................................... 188

Chapter 6

Table 6.1 Summary of year-to-year variation in predictions of pasture growth over 45 years at three locations

in the VRD..................................................................................................................................................... 209

Chapter 7

Table 7.1 Some estimates of long-term safe levels of pasture utilisation across northern Australia............. 216

Table 7.2 Stocking rate and burning frequency for each land system in the 1997 survey; and estimations of

tree basal area, closest associated generic pasture and soil parameters, and estimated soil depth for each land

system. ........................................................................................................................................................... 221

Table 7.3 Median simulated seasonal pasture growth (SSPG) calculated using the Site x Species approach to

developing generic parameters; and long term utilisation rates for specific land systems on 10 properties in

the VRD, using stocking rate data provided in the 1997 survey (Smith 1998).............................................. 224

Table 7.4 Current and expected future stocking rates (SR) and levels of pasture utilisation (Util) on 22

properties in the VRD using property carrying capacity and grazing area data from the 2004 survey (Oxley

2006). Pasture growth used to determine utilisation calculated from the Regional VRD parameter set at

Victoria River Downs. ................................................................................................................................... 226

Table 7.5 Comparison of simulated key nitrogen and pasture variables under nitrogen-limited and unlimited

conditions, and in the presence and absence of trees at Auvergne over a 45 year period (1959/60 to 2003/04).

Values across each row are not necessarily from the same season................................................................ 234

Table 7.6 Year-to-year costs of augmenting native pasture with legumes in the northern Ord-Victoria Area

(adapted from Oxley and Walker (2003). Pastures require renovation and re-establishment after 10 years. All

values are $/ha. .............................................................................................................................................. 238

xvii

Table 7.7 Projected benefit of augmenting native pastures with legumes in the northern Ord-Victoria Area.

....................................................................................................................................................................... 239

Table 7.8 Sensitivity of benefit-cost analysis to changes in input values. Base values are those used in

calculation of costs (Table 7.6) and projected benefits (Table 7.7). Percentage change for net regional benefit

is relative to the base value of $4.54million. ................................................................................................. 240

Table 7.9 Predicted percentiles of: utilisation at four constant stocking rates; and stocking rates at four

constant levels of utilisation. Calculations based on pasture growth predictions over a 45 year period

(1959/60 to 2003/2004) at Victoria River Downs using the Regional VRD parameter set. ........................... 244

Chapter 8

Table 8.1 Summary of the current suitability of GRASP for application to analysis of grazing practices in the

VRD when using parameters developed in this study. .................................................................................. 251

Appendices

Table 9.1 Study sites classified using: 1 Perry (1970); 2 Stewart et al. (1970); and 3 DIPE (unpublished). .. 268

Table 9.2 Study sites classified according to: 1 Northcote (1979); 2 Northcote et al. (1975); 3 Stace et al.

(1968); 4 Isbell (1996); 5 McDonald et al. (1990). ......................................................................................... 269

Table 9.3 Study sites classified according to Wilson et al. (1990)................................................................ 270

Table 9.4 Study sites classified according to Tothill and Gillies (1992)....................................................... 271

Table 9.5 Laboratory analysis results of soil samples from sites located on red earths overlying basalt...... 272

Table 9.6 Laboratory analysis results of soil samples from sites located on red earths overlying limestone.

....................................................................................................................................................................... 273

Table 9.7 Laboratory analysis results of soil samples from sites located on cracking clays overlying basalt.

....................................................................................................................................................................... 273

Table 9.8 Laboratory analysis results of soil samples from sites located on cracking clays of alluvial origin.

....................................................................................................................................................................... 274

Table 9.9 Amount of soil nutrients present in surface layers of study sites. (Table continued overleaf) ...... 275

Table 9.10 Plant species nomenclature for all individual species presented in this study............................. 278

Table 9.11 Complete list of plant species present at Site 1, July 1995.......................................................... 279

Table 9.12 Descriptive statistics for total standing dry matter data collected at each harvest for all sites. Eight

1m2 quadrats were cut at each harvest. All units are kg/ha. (Table continued overleaf)................................ 284

Table 9.13 Observed and predicted total standing dry matter (TSDM) values from GRASP model calibration.

(Table continued overleaf)............................................................................................................................. 290

Table 9.14 Summary of results from parameter sensitivity analysis. Data presented is end of growing season

pasture TSDM. All units are kg/ha. ............................................................................................................... 297

xviii

List of Plates

Appendices

Plate 1 Northern Victoria River District landscape showing escarpment country, woodland plains and

alluvial flats. .................................................................................................................................................. 258

Plate 2 Central Victoria River District landscape showing cattle grazing Chrysopogon fallax and Iseilema

spp. pasture on cracking clay soil. ................................................................................................................. 258

Plate 3 Southern Victoria River District landscape showing open grassy plains with scattered trees and

occasional rocky basalt outcrops. .................................................................................................................. 259

Plate 4 Land showing evidence of past heavy grazing: bare and scalded ground, dead trees, and annual

pasture species. .............................................................................................................................................. 259

Plate 5 Fencing a study site to exclude cattle from grazing. ......................................................................... 260

Plate 6 Burning to remove carryover material at commencement of the study period (Site 16)................... 260

Plate 7 Marking out sampling cells at site establishment (Site 2). ................................................................ 260

Plate 8 a) Bulk density sampling in soil profile pit; and b) hand-augering soil moisture cores during field

measurements. ............................................................................................................................................... 261

Plate 9 a) Measuring perennial grass basal area using a 5-point frame; and b) pasture sampling during field

measurements. ............................................................................................................................................... 261

Plate 10 Red earth overlying basalt (Site 19). ............................................................................................... 262

Plate 11 Red earth overlying limestone (Site 11).......................................................................................... 262

Plate 12 Cracking clay overlying basalt (Site 4). .......................................................................................... 263

Plate 13 Alluvial cracking clay (Site 7). ....................................................................................................... 263

Plate 14 Barley Mitchell grass pasture (Site 18). .......................................................................................... 264

Plate 15 Overhead view of 1m2 quadrat in a) barley Mitchell grass pasture (Site 18); and b) ribbon grass

pasture (Site 8)............................................................................................................................................... 264

Plate 16 Ribbon grass pasture (Site 8). ......................................................................................................... 264

Plate 17 White grass pasture (Site 14). ......................................................................................................... 265

Plate 18 Overhead view of 1m2 quadrat in a) white grass pasture (Site 14); and b) annual short grass pasture

(Site 1). .......................................................................................................................................................... 265

Plate 19 Annual short grass pasture (Site 1). ................................................................................................ 265

Plate 20 Forb pasture (Site 4)........................................................................................................................ 266

Plate 21 a) Overhead view of 1m2 quadrat in a forb pasture (Site 4); and b) an example of the abundant,

diverse forbs present at Site 15 on 2 May 1995............................................................................................. 266

Plate 22 Inundated conditions at Site 15 on 9 March 1995........................................................................... 266

xix

Plate 23 Annual short grass pasture early in the wet season with no burning at site establishment (Site 13, 12

Jan 1995)........................................................................................................................................................ 267

Plate 24 Overhead view of 1m2 quadrats in an annual short grass pasture (Site 13) during early wet season a)

with no burning at site establishment; and b) after burning to remove carryover material............................ 267

Plate 25 Annual short grass pasture early in the wet season after site burnt to remove carryover material (Site

13, 9 Jan 1996). Fire most likely destroyed much of the seed bank, resulting in poor germination. ............. 267

Chapter 1 Introduction

1

1.0 Introduction

The production of beef cattle from native pastures (pastoralism) is the main economic use

of tropical woodlands across northern Australia today, as it has been for more than a

century (MacLeod et al. 2004). In the Victoria River District of the Northern Territory

(Figure 1.1), attempts to develop land for crop production and replace native pastures with

exotic species have been limited by overriding climatic constraints and low fertility soils

(Bauer 1985). Consequently, pastoralism is the dominant land use in the Victoria River

District (VRD) and this is unlikely to change in the foreseeable future.

For much of its history the pastoral industry in the VRD had little control over herds, and

cattle populations fluctuated depending upon prevailing seasonal conditions. Occasional

periods of high grazing pressure occurred, particularly on favoured pasture types and close

to stock drinking water. Some periods of high grazing pressure resulted in land degradation

(Condon 1986; Tothill and Gillies 1992). Land degradation reduces the productivity and

profitability of pastoral enterprises (MacLeod et al. 2004). Improvements to property

infrastructure over the past few decades have increased control over the timing, intensity

and duration of grazing. Control of grazing allows for development of improved grazing

practices that minimise the risk of future degradation episodes.

A crucial component of long-term sustainable use of native pastures is balancing cattle

demand for feed with the amount of pasture available (MLA 2004b). Climate has long

been recognised as a major determinant of pasture production. Therefore, achieving the

balance between forage demand and pasture production requires knowledge of the

relationships between climate and pasture growth (McKeon et al. 1990). Despite the long

history of grazing in the region, little quantitative data on long-term pasture production is

available and this is a severe limitation to analysis of grazing practices. A means of

predicting pasture growth using climate data would benefit the development, evaluation

and improvement of management practices for grazed native pastures in the VRD.

Aim of this study

This study aims to develop the capacity for predicting native pasture growth in the Victoria

River District of the Northern Territory and demonstrate application of that capacity to

analysis of grazing practices.


2

Figure 1.1 The location of the Victoria River District of the Northern Territory.

Research approach

To achieve this aim, Chapter 2 first describes the features of the VRD and its pastoral

industry. The current knowledge the factors affecting pasture production are then

reviewed, revealing a need to quantify the long-term productivity of native pastures in the

region. The development of systems modelling as a means of quantifying pasture

productivity for application to grazing management is discussed and the GRASP pasture

growth model (Littleboy and McKeon 1997) is briefly described.

Chapter 3 documents a field study that collected the necessary data to establish

relationships between climate, soil water supply and pasture growth for commonly grazed

native pasture communities throughout the region. This data is used in Chapter 4 to adapt

(calibrate) GRASP for simulating the pasture system at the study locations. When a model

is calibrated from field data, the process involves closely fitting model output to the field

observations. Chapter 4 briefly describes the calibration process and then concentrates on

evaluating the degree of fit between model predictions and measured data. The input

parameter values derived for each location form the basis for extending the use of the

model beyond the boundaries of this study.

At this stage of the thesis, results of model calibration are specific to the individual site

during the period of field measurements. On their own, they are of limited value for

application to analysis of grazing practices. Before GRASP can be applied for grazing

Katherine

Victoria River Downs Station

Northern TerritoryInverway Station

Auvergne Station

Darwin

Alice Springs

Victoria River

District


3

management purposes, the capability of the model to replicate events (i.e. pasture growth)

at times and places for which field data has not been collected is tested. Two steps are

necessary to test the model in this manner: 1) individual site parameters are summarised to

form generic parameters representing land types of interest; and 2) simulation results when

using generic parameters are compared with independent measured data to evaluate the

ability of GRASP to simulate growth of native pastures across the region. The process of

summarising individual site parameters to form generic parameter sets and testing model

with independent data is described in Chapter 5.

McKeon et al. (1990) points out that some individual model parameters have greater

influence on predictions of pasture growth than others. Using generic values for these

parameters may compromise accurate simulation of the pasture system in some instances.

Identifying the parameters which have most influence on model behaviour in the VRD will

provide future model users with guidance when selecting appropriate parameter values and

interpreting the results of their simulation studies. Chapter 6 documents the process and

outcomes of identifying these influential parameters and their effect on predictions of

pasture growth.

The purpose of developing a capacity to predict native pasture growth is to facilitate the

development and analysis of sustainable grazing practices. GRASP has been used as a tool

for assisting grazing management since the early 1990’s (e.g. McKeon et al. 1994;

Johnston et al. 1996a; Day et al. 1997b; Hall et al. 2001) and has many applications. This

thesis will demonstrate two applications of a well-tested model to grazing land

management in the VRD: 1) calculating current and expected future levels of pasture

utilisation (the proportion of annual pasture growth consumed by grazing cattle) in the

VRD; and 2) the effect of improving the nitrogen supply (a known constraint to pasture

growth) in the northern high rainfall zone of the district. These applications are presented

in Chapter 7. An integrating discussion of the findings of this study leads to the final study

conclusions in Chapter 8.

The approach to structuring this thesis is based upon Evans and Gruba (2002), although

numerous departures from their recommendations occur. An illustrative representation of

the structure is presented in Figure 1.2.


4

Figure 1.2 Illustration of the structure of this thesis.

Background and reason for study Chapter 2

Calibration of GRASP to each individual site

Chapter 4

Assessing the fit between model results and field data

Field study methods

Climate results Soil results Pasture results

Chapter 3

Identifying major influences on pasture growth

Chapter 6

Final discussion and conclusions of study

Chapter 8

Generic parameters for testing model with independent data

VRD as a region Pasture types Soil types Chapter 5

Evaluating the capability of GRASP for extrapolation to the wider landscape of the VRD

Applications of model to management of grazing land

Current and future levels of pasture utilisation

Chapter 7 Alleviating the nitrogen limit to

pasture growth

Chapter 2 Review of Literature

5

2.0 Review of literature

2.1 A brief description of the Victoria River District

Physical descriptions of the region are given in CSIRO (1970) and Muchow (1985), and a

relevant summary of northern Australia is given by Shaw and Norman (1970). Kraatz

(2000) provides a concise overview of the natural environment of the VRD and similar

information is accessible online at the Tropical Savannas Cooperative Research Centre

website (TSCRC 2006). These descriptions are summarised below to provide a basic

outline of the region. Photographic examples of landscape in the northern, central and

southern VRD are shown in Plate 1, Plate 2, and Plate 3 respectively (Appendix 1).

Location

The Victoria River District (VRD) is a region of about 126 000km2 lying between the

latitudes 150S and 190S and longitudes 1290E and 1320E in the northwest of the Northern

Territory (NT) of Australia. The boundaries of the district are the Joseph Bonaparte Gulf

coastline and the Fitzmaurice and Flora Rivers to the north, the Sturt Plateau to the east,

the Tanami Desert to the south, and the Western Australian border and Kimberley region

to the west. Figure 1.1 and Figure 3.2 show the location of the VRD within the NT.

Climate

Climate in the VRD has been described by Slatyer (1960), Slatyer (1970) and Williams et

al. (1985) and more generally by Tothill et al. (1985), Partridge (1994) and Colls and

Whittaker (2001). The region can be classified as having a semi-arid tropical (monsoonal)

climate with hot, wet and humid summers during which 95% of annual rainfall occurs; and

sunny, warm and dry winters that are virtually rainless. Short transitional periods merge

these two distinct seasons. A climate summary for Victoria River Downs Station in the

centre of the district is shown in Figure 2.1 and Table 2.1. The data highlights the generally

high temperatures of the region, strong seasonal and year-to-year variability of rainfall,

seasonal variation in humidity, and high evaporative demand throughout the year.

Chapter 2 Review of Literature ------------------------------------------------------------------------------------------------------------------------------------------

NOTE: This figure is included on page 6 in the print copy of the thesis held in the University of Adelaide Library.

Figure 2.1 Seasonal rainfall (July to June, mm) at Victoria River Downs Station for the period 1900/01 to 2003/04. The horizontal dashed line represents the median value (639mm) (Source: DataDrill 2005).

Table 2.1 Climate data at Victoria River Downs Station for the period 1900/01 to 2003/04 (source: DataDrill 2005).

Topo

The

on A

altitu

(Pat

____

NOTE: This table is included on page 6 in the print copy of the thesisheld in the University of Adelaide Library.

graphy

major topographic feature of the region is the Victoria River Plateau. It is formed

delaidean sediments and is a large dissected plateau rarely exceeding 350m in

de. It is a collection of structural plateaux and benches, mesas and buttes

erson 1970). This

_____________________________________________________________ 6


7

plateau forms the catchment area of the Victoria River, the major watercourse from which

the region takes its name. Much of the area consists of rocks and their resistant nature is

visible across the landscape (Williams et al. 1985).

Geology

Many geologic periods are represented in the VRD. Traves et al. (1970) and Paterson

(1970) give summaries of the geology and geomorphology of the district, including maps.

Isbell (1983) and Johnson (2004) describe the geology of northern Australia in more

general terms, and Kraatz (2000) provides a brief history of geologic activity in the region.

To summarise, the oldest rocks are volcanics, sandstones, and siltstones in the north-

eastern part of the district formed during the Archaean-Lower Proterozoic period (3500 to

1000 million years ago (Mya)). The youngest soils are the Quaternary (<1.8 Mya) alluvia

found on floodplains associated with the coast and major watercourses and drainage

systems of the districts north. Adelaidean (550+ Mya) sediments (sandstone, siltstone,

dolomite, and shale) make up a large proportion of the district’s centre. Lower Cambrian

(545 to 490 Mya) volcanics are the other major geologic group present in the region,

occurring in the western, southern and eastern parts. The desert lands of the very south are

generally Tertiary (65 to 1.8 Mya) laterite.

Soils

Steep tablelands and hills with rock outcrops dominate much of the district and these lands

generally have shallow, immature skeletal soils, with deeper soils restricted to the gentle

slopes and plains (Kraatz 2000). On erosional and alluvial plains there are distinct

relationships between soils and climate, drainage, and parent materials. Stewart (1970b)

outlines these relationships for the major soil groups. Generally speaking, better-drained

sites have medium textured red and yellow earths in higher rainfall areas, and coarse sandy

desert soils in the south. Gentle slopes and poorly drained sites usually have cracking clay

soils. Floodplain alluvia almost invariably consist of cracking clay soils. The most striking

feature of soils in this region is the universal presence of very low levels of nitrogen and

phosphorus (Perry 1960; Isbell 1983; Williams et al. 1985).

Vegetation

In their report on the vegetation of the NT, Wilson et al. (1990) give a summary of

vegetation in the semi-arid zone, within which the VRD lies. They state that the most


8

widespread communities in this zone are woodlands and open woodlands. Throughout the

VRD the overstorey component is dominated by the Eucalyptus and Corymbia genera on

most soils, except the clays where Lysiphyllum, Terminalia or Melaleuca occur. The

ground layer is dominated by grasses, particularly long-lived perennials like Sehima,

Heteropogon, Themeda and Sorghum on medium textured soils, Plectrachne on drier soils

and Dichanthium, Chrysopogon and Astrebla on the clays. Other broad descriptions of

vegetation of the region are given by Andrew et al. (1985), Mott et al. (1985) and Kraatz

(2000). More detailed information on individual species can be found in Petheram and Kok

(1983), Vallance et al. (1993), Wheaton (1994) and Moore (2005).

Land Use

About 65% (83 000km2) of the VRD is currently pastoral lease land, and some aboriginal

freehold land is also used for pastoralism. Individual private leaseholders own many of the

30 stations in the district and Aboriginal interests and Australian or international

companies have the rest (Oxley 2006). The remaining land is not used for commercial

production. Figure 2.2 shows proportions of each land tenure type in the district in 2004.

The economic importance of cattle production in the region is described in Section 2.2.

The focus of this study is the grazing lands of the district and although the outcomes of this

work will have some application to non-grazing lands, they are not considered further.

Figure 2.2 Land tenure in the Victoria River District in 2004 (data sourced from NT Dept of Lands).

Pastoral lease65%

Aboriginal freehold14%

National parks11%

Other (mainly defence)

10%


9

2.2 The pastoral industry

2.2.1 History

Pastoralism (beef cattle production from native pastures) has been the dominant land use in

the district as it has been since European settlement in the late 1800’s (Perry 1970; Winter

et al. 1985; Kraatz 2000). Traditionally, properties covered huge tracts of land and inputs

from management were very low. Few fences meant cattle movements were largely

uncontrolled and large herds tended to congregate around the natural watering points

(rivers, waterholes and springs). Considerable feral horse and donkey populations were

also present (TSCRC 2006).

Implementation of the Brucellosis and Tuberculosis Eradication Campaign (BTEC) during

the 1970’s lead to substantial changes in the industry. Fences were constructed, bores

drilled to draw cattle away from natural waters, and aerial mustering introduced to improve

the control over stock numbers and movements needed to comply with the campaign’s

testing requirements. Condon (1986) states: “the (BTEC) program has changed the

concept of the cattle industry from a hunting operation of semi-wild cattle to one in which

increasing attention will be paid to the management of the land and animal resource”.

Cattle producer surveys document characteristics of the pastoral industry at this time (Hill

1976; Robertson 1980; Robertson 1982), and a northern Australian context is provided by

O'Rourke et al. (1992).

Today cattle stations in the district are still very large, averaging 3275km2 (327 500ha) of

which 2227km2 (222 700ha) are grazed. Paddocks typically exceed 100km2 (10 000ha).

Herds average 21 500 head per property, giving a district population of over 500 000 head

(Oxley 2006). Annual turnoff (number of animals sold per year) from the VRD over the

past 20 years has averaged about 90 000 head (Figure 2.3), a contribution of about 25%

(over $50m) to the $200m NT beef cattle industry (DPIF and ORD 2001; S. Murti

pers.comm.). Live export of growing cattle is the primary turnoff market (Bortolussi et al.

2005a), with Indonesia and the Philippines the main destinations. Domestic abattoirs and

feedlots take the balance of marketed animals.


10

Figure 2.3 Total cattle population and annual turnoff (number of animals sold) for the Victoria River

District 1883 – 2004 (S. Murti pers.comm., based on Australian Bureau of Statistics and NT Office of

Resource Development data). Dashed lines represent interpolation across periods where no data is available.

Early data has been left as isolated points as no basis for interpolation is available.

2.2.2 Grazing management

The low level of grazing management practiced for much of the pastoral industry’s history

is evident by the conspicuous lack of reference to it in the producer surveys of the 1970’s

(Hill 1976; Robertson 1980). Due to improved livestock control and marketing

opportunities, grazing management in the region today is much more sophisticated.

Current knowledge of sustainable grazing management in the VRD has recently been

synthesized into an education package known as Grazing Land Management (GLM) (MLA

2004a). The main elements of GLM are outlined below.

Sustainable management of native pastures for grazing aims to achieve two desirable

outcomes: 1) optimising animal production; and 2) maintaining healthy and productive

land. These outcomes are achieved by attention to three basic principles of grazing land

management:

• improving land condition (conversion of rainfall into useful pastures);

• improving the evenness of pasture use (consumption of pasture by cattle); and

• improving diet quality (converting pasture into beef).

These three principles are directly affected by pasture utilisation; the balance between

pasture supply and the feed requirements of grazing animals. Consequently, sustainable

0

100

200

300

400

500

600

1880 1900 1920 1940 1960 1980 2000

Num

ber o

f Cat

tle (x

1000

) Total cattle populationAnnual turnoff


11

grazing management in the VRD relies heavily on knowledge of the productivity of the

native pasture communities used for grazing.

2.3 Current understanding of factors influencing pasture growth

Above-ground grass growth at the individual plant level is a function of energy (radiation

and temperature), water (soil moisture), plant–available nutrients (soil fertility), and gas

(atmospheric CO2) (Fitzpatrick and Nix 1970). At the landscape level, total pasture

biomass production is also a result of species composition, plant density, competition for

water and nutrients with other vegetation groups (e.g. shrubs, trees and exotic weeds),

impacts of fire and grazing defoliation (native and introduced), and pests and diseases.

Some of these factors are of little relevance to this study and, therefore, the following

review is restricted to the most influential factors driving plant growth in the VRD.

Literature relevant to this study uses several expressions relating to the amount of pasture

present in the field. ‘Yield’, ‘biomass’, and ‘total standing dry matter’ refer to the actual

amount of pasture present at any point in time and are often used interchangeably. ‘Pasture

growth’ is used to refer to the amount of pasture dry matter that has accumulated since the

beginning of the growing season in the absence of grazing and excluding any carryover

material from previous seasons. ‘Pasture productivity’ and ‘pasture production’ refer to

total pasture amount present at the end of the growing season in the absence of grazing.

These terms will be applied throughout this thesis in a manner that most clearly expresses

the intended meaning.

2.3.1 The influence of climate

Climate patterns for the northern part of the Northern Territory have been analysed by

several authors (Slatyer 1960; Slatyer 1970; Taylor and Tulloch 1985), each noting the

high level of rainfall variability. The importance of climate as a factor influencing pasture

growth has been well researched and documented. Many attempts have been made to

quantify the relationships between climate variables and pasture growth for tropical

Australia, mostly in a modelling framework. Such frameworks use a soil water balance

approach, combined with defined response functions to other variables (Slatyer 1960;

Fitzpatrick and Nix 1970; McAlpine 1970; McCown et al. 1974; McCown 1981a; McKeon

et al. 1982a; Rickert and McKeon 1982; Littleboy and McKeon 1997). These approaches


12

are reviewed in Section 2.4. The following paragraphs summarise the influence of climate

variables on pasture growth in the study area.

Rainfall

Where grass production is highly seasonal, as in the semi-arid tropics of the VRD, grass

production is directly related to rainfall. The amount, distribution, and variability of

rainfall are major factors determining the time of initiation of growth and the period for

which growth will be sustained. Slatyer (1960) states that in relation to plant growth,

climatic factors other than rainfall are of secondary importance at Katherine. In the VRD

rainfall patterns are highly variable with frequent dry spells of varying duration within the

wet season, and this places a limitation on the period during which there is sufficient soil

moisture available for plant growth (Slatyer 1970). Much of the precipitation occurs in

high intensity storms and part of this rainfall becomes surface runoff or infiltrates beyond

the root zone of pasture plants (Williams et al. 1985). Johnson and Tothill (1985) point out

that while annual rainfall totals provide a useful frame of reference, seasonality of the

rainfall pattern, effective rainfall during the growing season, and the duration and effect of

the dry season are of greater importance. Mott et al. (1985) state that the major climatic

constraint to grass production is soil water and hence the highly seasonal nature of rainfall

in the VRD is reflected by the correspondingly seasonal plant growth patterns.

Temperature

Different groups of plants have fundamentally different responses to temperature

(Fitzpatrick and Nix 1970; Sweeney and Hopkinson 1975). The normal extremes of

temperature and when they occur in the annual cycle are most important to pasture growth.

Tropical grasses have high temperature optimums and, while northern Australia

experiences relatively high ambient temperatures, some authors suggest they may not be

sufficient for maximum growth rates. For example, Doley and Trivett (1974) report an

optimum temperature for photosynthesis of 400C for a native perennial grass. However

Christie (1975b) found reduced growth rate of two tropical grasses at this temperature and

report temperature optimums of closer to 300C. Tropical legumes have lower thermal

optimums (Whiteman 1968; Sweeney and Hopkinson 1975) and are likely to grow

unrestricted by temperature in this region, although Fitzpatrick and Nix (1970) suggest that

the high temperatures during the growing season may in fact surpass optimums and

sometimes reduce the growth rate of tropical legumes. Ivory and Whiteman (1978) report


13

that growth of subtropical grasses ceases at day temperatures of 12.50C and night

temperatures of 100C. These conditions are rare in the Victoria River District and never

occur during the growing season. Thus, temperature is not a factor in the cessation of

growth in the environment of the VRD.

Slatyer (1960) reports short periods of very high temperatures occurring on occasions

during the wet season and these may have a direct effect on plant development. This is

particularly true when they are associated with periods of low humidity, and occur during

critical periods of plant development such as flowering time.

The occurrence of frost is rare, with the northern parts of the VRD virtually frost free and

southern parts having occasional light frosts in some years, particularly in topographically

suited localities (Slatyer 1970). However, frosts never occur during the period of plant

growth, and are therefore not considered any further in this study.

Radiation and day length

The influence of solar radiation (sunlight) on plant growth depends on the species under

consideration. Photosynthetic rate is a result of interaction between plant structure and

physiological properties, and the physical properties of the incident light (Fitzpatrick and

Nix 1970; Ludlow 1985). Fitzpatrick and Nix (1970) suggest that light is rarely a limiting

factor to plant growth in northern Australia, except for some minor effect during high

rainfall periods when cloud cover reduces incident light. Plant response to photoperiod is

also species-specific. Fitzpatrick and Nix (1970) suggest that some perennial grass species

in the monsoonal tropics are short-day plants; that is, their flowering is initiated by

reducing day-length. For all species, day length during the natural growing season always

exceeds 12 hours and plants are passing through their seed setting and maturing phases at a

time of decreasing day length (Slatyer 1960).

Fitzpatrick and Nix (1970) show the average daily radiation received in the VRD varies

only slightly throughout the year and is always high relative to most of Australia. Values

range from 550cal/cm2/day (23MJ/m2/day) during January, down to 425cal/cm2/day

(17.8MJ/m2/day) in July. Sunshine hours reach a peak in August (300hrs/month). Values

decrease from September to November as cloud cover increases, reach their minimum in

February while monsoon cloud is present, and then steadily increase again until August.


14

Day length varies little throughout the year from 11.2 hours per day in June to 13.0 hours

per day in December (Williams et al. 1985). Neither Fitzpatrick and Nix (1970) nor

McCown (1981a) consider solar radiation as a limiting factor in plant growth in the

Australian tropics.

In summary, the amount and distribution of rainfall is the most influential climate variable

on pasture growth in the VRD. Temperature has occasional effects for short periods and

sunlight is nearly always non-limiting.

2.3.2 Limitations to pasture production other than climate

Climate is not the only determinant of pasture production. In their natural state, most

northern Australian soils have low levels of many nutrients important for plant growth

(Williams and Andrew 1970; Schmidt and Lamble 2002). Mott et al. (1985) suggest that

nutrients, either alone or in combination with soil water supply, are important in

determining plant production in semi-arid rangelands. Species composition of the pasture

sward and the presence of other vegetation classes (e.g. trees) also impacts upon pasture

production. These are discussed below.

Nutrients

Mott et al. (1985) state that in some soil/climate systems (e.g. the monsoonal tropics of the

northern VRD), water will always be secondary to nutrients as the major determinant of

seasonal pasture production. In other climates and soils, nutrients and water will interact so

that in some seasons nutrients will be the main controlling factor, whereas in other seasons

it will be water and temperature. Schmidt and Lamble (2002) report a strong link existing

between soil nutrient availability and pasture productivity for northern Australian

savannas, and McIvor et al. (1994) and Cook and Andrew (1991) also support the view

that nutrients are the main limiter of plant growth in the tallgrass systems of the Northern

Territory. Christie (1981) suggests that on some land types in semi-arid Queensland

nitrogen is the major limiting nutrient for biomass production, while on others it is

phosphorus.


15

Nitrogen

Nitrogen (N) is quantitatively the most important nutrient that pasture plants take up from

the soil, however N levels are low in most semi-arid areas of Australia (Williams and

Raupach 1983; Schmidt and Lamble 2002). Early agricultural research at Katherine on red

clay soils revealed that nitrogen and phosphorus (P) were the primary limiting nutrients for

plant growth and no others had been identified at that time (Norman 1962). Results in

Norman (1962) demonstrate a marked interaction between N and P in pasture growth

responses. Cook and Andrew (1991) report nitrogen limitations to native pasture growth in

the monsoonal tropics near Darwin, while Friedel et al. (1980) found positive pasture

growth responses by native pasture species to additional N supply under field conditions in

central Australia, but not all species responded equally. Christie (1981) concluded that

nitrogen is the major limiting nutrient for biomass production for the Mitchell grassland

systems in Queensland (and, by inference, those in the NT including the southern VRD).

Herbaceous legumes are a common component of native pastures in the VRD. Similarly, a

number of woody leguminous tree and shrub species occupy some land types. No data is

available on the contribution of these species to soil nitrogen status, although it is

reasonable to assume the major contribution is from decomposition of roots and leaf litter

(Mott et al. 1985; Abbadie et al. 1992). Lawrie (1981) reports nitrogen fixation of less than

1kg/ha in native legumes on a number of soil types in temperate Victoria.

Wetselaar and Hutton (1963) recorded nitrogen contents in rainfall equivalent to adding

around 1.5kg/ha to the soil in a typical wet season, with much of this occurring early in the

season. However, they point out that the origin of this nitrogen is likely to be terrestrial -

from dust, smoke, and volatilisation of nitrogen from the soil surface. They conclude that

such an addition of N is not true accession but rather a cycling of previously lost nutrients.

Phosphorus

Phosphorus is widely deficient in soils across Australia, and present only at low levels in

the ancient, heavily leached soils of tropical Australia (Isbell 1983). McIvor (1984) states

that the ability of a species to grow in low P soils is related to its capacity to absorb P at

low levels and/or a low internal requirement of P for optimum growth. A significant

consequence of phosphorus deficiency in semi-arid grasses is reduced root growth

(Christie 1975a). Experimental work suggested that the native grasses Chrysopogon fallax


16

and Heteropogon contortus (both widespread in the VRD) had low requirements for

phosphorus, as did other native grasses in the study (McIvor 1984). Norman (1962) found

mixed responses of native grasses to varying levels of P supply, but general trends were

that additional P supply resulted in increased pasture growth. Christie and Moorby (1975)

report similar findings. Friedel et al. (1980) found that P is potentially even more deficient

in degraded soils than healthy soils in central Australia.

Trees

Johnson and Tothill (1985) define savanna vegetation (native pasture lands) as “a

continuous graminoid stratum more or less interrupted by trees or shrubs”, referring to the

almost universal presence of trees across the landscape of northern Australia. Other authors

also refer to the widespread presence of trees on the savanna lands of northern Australia

(Perry 1970; Mott et al. 1985; Burrows et al. 1990; McKeon et al. 1990; Jackson and Ash

1998) and overseas (Walker et al. 1981). Where pasture and tree root distributions overlap,

they likely compete for water and nutrients (Schmidt and Lamble 2002).

Scanlan (2002) reviewed tree-grass dynamics in Queensland and concluded that pasture

production response to tree density depends upon tree species, rainfall, soil type, climatic

history, fire, and grazing, and will change over time. The relationship between pasture

production and increasing tree density can vary from linear decrease, to exponential

decrease to initial stimulation followed by a decrease.

No published work is available on the competitive relationships between trees and pasture

in the VRD. Eyles et al. (1984) and Winter et al. (1989b) report increased grass production

in the first years following killing of trees in tropical tall-grass pastures near Katherine, but

the effect declined after several years. More recent work at Katherine also found an

increase in pasture production following removal of trees (Cafe et al. 1999). A similar

trend has been observed at Kidman Springs in the central VRD (L.M. Cafe pers.comm.)

and in other semi-arid environments (e.g. south-west Queensland, Beale 1973). Day et al.

(2003) reports considerable competition between trees and pasture for soil nitrogen in the

semi-arid region of Zimbabwe. In the Katherine Region competition between trees and

grasses is greater for nutrients than for soil water (Winter et al. 1989b).


17

Species composition

Species composition is primarily determined by climate, soil type and grazing history.

Pasture species composition has been identified by McIvor et al. (1995a) as being of

significance to seasonal biomass production. While they reviewed previous studies on this

topic and the variable trends in pasture production reported in them, their own research on

the monsoonal tall-grasses near Katherine showed that as perennial grass content declined

and annual grasses and forbs increased, total pasture production decreased. This decrease

was associated with an increase in the nitrogen and phosphorus uptake by pasture (Ash and

McIvor 1995), and improved animal liveweight gain (Ash et al. 1995). The increased risk

of further undesirable changes caused by grazing annual species dominated pastures more

than offsets any improvement in animal nutrition.

To summarise, rainfall patterns and soil type influence soil water supply. Soil water and

nutrient availability are the major influences on seasonal pasture production, and species

composition is also an important factor. Scholes (1993) provides this useful synopsis of

pasture growth in the semi-arid tropics: “it is a convenient simplification to consider that

water controls growth duration, while nutrients and temperature control growth rate”.

2.3.3 Previous studies of pasture production in the VRD

The earliest assessment of pasture production in the VRD was in terms of estimated cattle

carrying capacity (number of grazing animals per unit area) for a particular land type.

Early estimations of carrying capacity for pasture lands in the region are reported by Perry

(1960) and Perry (1970).

Some formal assessments of the production of native pasture communities were carried out

in the 1950’s and 1960’s at Katherine Research Station, just outside the VRD. While much

research during this period centred on finding productive introduced species, growth rate

and nutritional studies were made on several native pasture communities (Norman and

Wetselaar 1960; Smith 1960; Norman 1962; Norman 1963a; Norman 1969). These native

pasture studies were restricted to deep soils with agricultural potential close to the

township of Katherine. Potential for beef cattle production from these pastures have also

been experimentally recorded (Norman 1960; Norman 1965; Norman 1967).


18

The land units of Kidman Springs in the centre of the district (Figure 3.2) are described by

Forster and Laity (1972). They identify a number of discreet land units and give an

ambiguous grazing value for each unit in words (no grazing, little, rough, productive

grazing etc.) and did not attempt to quantify this with pasture production or recommended

stocking rates. Foran et al. (1985) give data on pasture species composition and standing

biomass under grazed and ungrazed conditions on two pasture communities at Kidman

Springs. Foran et al. (1985) also explore the effect of pasture condition on pasture

production and species composition for a common land type (calcareous red earth), noting

the predominance of arid short grasses on the heavily utilised areas of this land type. Bastin

and Andison (1990) and Bastin et al. (2003) provide an update on those studies, both

noting that prevailing seasonal conditions and grazing history influenced changes in

species composition and hence overall pasture productivity. These results and others from

the studies discussed in the following paragraphs are summarised in Table 2.2.

A description of lands in the western VRD is given by Robinson (1971) and Aldrick et al.

(1978) and, while they make comment on the impact of grazing on species composition,

they make no attempt to estimate pasture production or livestock carrying capacities.

Tothill and Gillies (1992) describe the land types across northern Australia (including the

VRD) using local pasture units (LPU’s). They classify land into one of three pasture

conditions within each LPU. Their classification system, while not clearly defined,

indicates pasture condition, or the health of the pasture communities, is very high in most

of the VRD. However some landscapes, particularly watercourse frontage country, short

grass pastures on red earths, and perennial grasses on cracking clays have experienced

some periods of heavy utilisation in the past and are in a state of ‘modified pristine’ or

slightly deteriorating condition.

Pasture condition was investigated by McIvor et al. (1995a) and Ash and McIvor (1995)

on tropical tall-grass pasture communities in the north-east of the region. These studies

highlighted the importance of both soil and vegetation factors in determining pasture

productivity. Ash et al. (1995) take this further and detail the animal production associated

with differences in pasture condition, highlighting the fact that often animal production can

be enhanced by species composition change from perennial to annual based pasture

swards. Annual species based pastures are, however, at greater risk of fluctuating yields.


19

Animal production from such pastures in the long term is therefore less favourable than

from pastures with high perennial grass content.

Since 1993, a grazing trial aimed at determining the relationship between stocking rate and

pasture productivity has been conducted at Mt Sanford in the southern part of the region.

Detailed measurements of species composition and standing pasture yield have been

recorded for the main pasture types present. Results include standing dry matter of various

pasture components under grazing (perennials and annuals, palatable and unpalatable

species, plant cover and bare ground), and pasture utilisation and animal liveweight gains

at different stocking rates (MacDonald et al. 1997; R.N. MacDonald, unpublished data.).

Table 2.2 Summary of existing pasture production data for the VRD.

Location Soil type Pasture species

Grazing intensity

TSDM

(kg/ha)

Years Reference

Katherine Tippera red

earth

Themeda,

Chrysopogon,

Sorghum

Ungrazed 1050 to 1550 1957/58 to

1967/68

Norman 1969

Kidman Springs

Cracking clay Chrysopogon,

Dichanthium,

Iseilema

Ungrazed 800 to 2100 Foran et al. 1985

Calcareous

red earth

Brachyachne,

Sporobolus,

Aristida

Ungrazed 150 to 1991 1973/74 to

1978/79

Calcareous

red earth

Enneapogon Ungrazed 600 to 1590 1973/74 to

1978/79

Calcareous Heteropogon Ungrazed 900 1989 Bastin et al. 2003

red earth 1500 to 2000 1994

1250 1999

1700 to 3000 2002

Manbulloo Red earth Chrysopogon,

Sorghum,

Sehima

Low 1490 to 2240 1981/82 to

1988/89

McIvor et al. 1994

Mt Sanford Cracking clay Astrebla Low 2410 1993/94 to

1998/99

R.N. MacDonald

unpublished data

Various Red earth Brachyachne,

Enneapogon

Low 900 1997/98 to

1998/99

Dyer et al. 2001a

Cracking clay Astrebla Low 1539

Various Sorghum Low 1250

Various Chrysopogon,

Dichanthium

Low 2319


20

In summary, some isolated measures of native pasture production have been made in and

around the VRD (Table 2.2), but these measures provide little value for estimating

production at other times and locations. Knowledge of the annual variation in pasture

production due to the influences of climate, soil characteristics, species composition and

plant density does not exist for most of the region. Determining the relationships between

such factors is a complex task, and too large and time-consuming to consider by field

experimentation alone. Another approach is required.

2.4 A modelling approach to assessing pasture growth

One approach to understanding the inter-relationships between factors that govern pasture

productivity, and the range of likely productivity from different land types, is to use

systems modelling. Mathematical modelling of the plant growth system is a well-

developed field and models can calculate plant growth as a product of many influential

factors. A systems modelling approach is currently the only option for providing objective

estimates of pasture productivity where no measurements have been taken. In the

following section previous relevant attempts to model pasture growth are discussed.

McCown et al. (1974) and Grant et al. (1997) give concise explanations of why models of

pasture growth may be used in preference to field studies in order to quantify the long-term

productivity of pasture communities. McCown et al. (1974) state that in the Australian

tropics, short-term trials are insufficient to obtain a quantitative estimate of the likely year-

to-year variation in pasture yield. Furthermore, longer-term trials have serious drawbacks

including difficulty in maintaining original botanical composition and fertility, and

limitation of the scope of research by committing finite research resources to these long-

term trials. McKeon et al. (1990) agree, stating that the development and testing of whole

farm systems is expensive in terms of time and resources if traditional methods of

experimentation are used. Modelling is seen as a practical method of evaluating many

options over longer time periods and for a greater range of environments than can be

achieved with field experimentation (Grant et al. 1997). This view is supported by Alcock

(2006) who states “grazing systems models … are essential tools for analysing the impact

of climate-driven production variability”.


21

2.4.1 Development of systems modelling of pasture growth

The approach of using climatic records to estimate plant growth by means of a soil water

balance has been used for many years. Slatyer (1960) details a method for estimating crop

growing season characteristics at Katherine using rainfall, soil water storage capacity and

evapotranspiration. McAlpine (1970) also uses a water balance approach on a weekly time-

step to provide probabilistic estimates of pasture production in central Australia. Johnston

(1996) explains the basis for modelling plant growth: the soil water balance describes

changes in soil moisture in terms of the difference between the input of rainfall and the

output of soil evaporation, plant transpiration, runoff and drainage. Plant growth is

assumed to be proportional to transpiration.

In 1970, Fitzpatrick and Nix introduced the concept of the Growth Index as a simple

empirical model of pasture production for Australian grasslands. They argued that the three

most important climatic variables influencing pasture growth were precipitation, incident

radiation, and temperature. The index of Fitzpatrick and Nix (1970) is described as:

GI = LI * TI * MI (Equation 1)

where: GI is Growth Index

LI is Light (radiation) Index

TI is Temperature Index, and

MI is Moisture Index

All indices are expressed as fractional dry matter production, that is, the relative

production compared to potential production if the factor was non-limiting. As such, values

for each index range from zero to one. Estimates of potential dry matter production are

required, upon which the index is applied.

LI is calculated from a general relationship between fractional dry matter production and

total daily solar radiation. This relationship was derived from both theoretical and

experimental data on responses of different plant groups to light regimes. Similarly, TI

uses defined relationships between plant production and thermal regimes as ascertained by

controlled environment experiments. Three distinctive thermal response curves were


22

developed for: 1) tropical grasses; 2) tropical legumes; and 3) temperate grasses and

legumes (Figure 2.4c). MI is determined by water balance accounting. MI is defined as the

ratio of estimated actual to potential evapotranspiration and is assumed to be a linear

relationship between unity (where water is non-limiting), and zero (where the available soil

water storage is exhausted).

GI is the product of these three indices (LI, TI and MI) and therefore, in any given time-

step, GI can never exceed the value of the most limiting factor. Also, if any factor equals

zero, so does GI. Fitzpatrick and Nix (1970) state that the highest GI values for tropical

grasses in Australia occur at inland places such as Katherine (Figure 2.4a) because of the

highly favourable combination of light, thermal and moisture regimes over short periods.

Tropical legumes may experience periods of supra-optimal temperatures, when

temperatures are so high that they restrict growth (Figure 2.4b). However, temperature

response is variable between species, and these individual responses are not well

understood.

A Growth Index modelling approach, while simple and requiring only basic data inputs on

a weekly time-step, takes no account of more complex environment-vegetation interactions

that may influence plant growth at a particular location (e.g. available nutrients). The

responses of plant growth to climatic conditions were developed at a broad scale and

responses generalized into three broad pasture groups and as such do not provide a suitable

framework for investigation of individual pasture types except if pasture types are

described in such general terms.


Figutropicprodu

Inco

Rose

Kath

and

calc

are d

yield

appr

envi

____

NOTE: These figures are included on page 23 in the print copy of thethesis held in the University of Adelaide Library.

re 2.4 Annual trends in LI, TI, MI and GI values at Katherine, NT for: a) tropical grasses, and b) al legumes; c) the relationship between mean daily temperature (0F) and fractional dry matter ction in three groups of pastures (Fitzpatrick and Nix 1970).

rporation of nutrients into a simulation model of pasture growth is attempted by

et al. (1972) for an introduced annual tropical legume (Stylosanthes humilis) at

erine. Rose et al. (1972) used an assumed ratio between evapotranspiration (ET)

potential (pan) evaporation to calculate actual ET. Pasture biomass is then

ulated as a function of ET on a weekly time step. Both variables (ET and growth)

ependent on phosphate nutrition. Rose et al. (1972) report over-estimations of

compared to measured results in above and below rainfall seasons. Hence, their

oach is limited in terms of accurately simulating plant growth from

ronmental variables.

_________________________________________________________________ 23


24

McCown et al. (1974) proposed a model of pasture growth also based on rainfall. They

used a weekly time-step with inputs of rainfall and pan evaporation (assumed to be

potential ET) to simulate a single soil water store of known water holding capacity. In the

model run-off and drainage are assumed to be nil until the soil water store is full. Actual

ET is calculated from a defined function (Figure 2.5b). When relating actual accumulated

ET to pasture yield from field experiments in north-eastern Queensland, strong

relationships were revealed. ET as a predictor of plant growth improved the accuracy of

growth predictions compared to using rainfall. McCown et al. (1974) concluded that ET

could be used with some confidence to calculate pasture growth. McCown’s approach was

the most successful at the time in terms of calculating pasture yield on the basis of climate

(principally water supply).

McCown (1981a) extended the Growth Index approach by defining a ‘green season’. The

green season was defined as a period of sustained green growth during the true ‘wet’

season and the extension of growth after rains cease through using stored soil moisture.

The green season began with an eight-week period in which GI was 0.1 or more in three of

the first four and in six of the eight weeks. The green season was considered to have ceased

at the first of two consecutive weeks when GI was less than 0.1. The GI was a product of a

moisture index and a temperature index modified from Fitzpatrick and Nix (1970) with one

function representing all species (Figure 2.5a).


Figuthe st 2.4. By cet almodserieSoci1982 ____


re 2.5 a) Thermal response curves of tropical grasses and legumes including the solid line used in udy of McCown (1981a); b) moisture index function of McCown et al. (1974).

2 The GRASP model

ombining the successful approaches of Fitzpatrick and Nix (1970) and McCown . (1974) an integrated water balance, pasture growth and animal performance el was developed which eventually became GRASs Production (GRASP). A s of papers introducing these integrated modules were presented at the Australian ety of Animal Production conference in 1982. These papers (Hendricksen et al. ; McKeon et al.

_________________________________________________________________ 25


26

1982a; McKeon et al. 1982b; Rickert and McKeon 1982; Scattini and Powell 1982) and

later publications (McKeon and Rickert 1984; Rickert and McKeon 1985) present the

rationale behind the integrated model and provide some examples of its application.

On-going research and development of GRASP have improved its performance and

incorporated many processes left out of the original structure. Littleboy and McKeon

(1997) provide the most recent comprehensive description of the model, and diagrammatic

representations of the soil water balance and flow of dry matter through the pasture system

are reproduced here (Figure 2.6). GRASP has been adapted from its original FORTRAN®

language to the Microsoft Windows® operating system as a user-friendly general research

and extension tool (Timmers et al. 1999). A brief description of the processes in GRASP is

given here and in more detail in Chapter 4 (Section 4.2). The following paragraphs have

been taken from Littleboy and McKeon (1997).

First, the soil water balance is determined:

“Rainfall is partitioned into infiltration and runoff using functions that relate runoff to

surface cover and rainfall intensity. Soil water is updated on a daily basis by any rainfall

exceeding the predicted runoff volume. Movement of water down the profile occurs when a

soil profile layer is above its user-defined field capacity. Any movement of water from the

bottom profile layer is assumed lost as deep drainage. There is no estimation of upward

movement of soil water caused by either capillary rise or water table fluctuations. The soil

also loses water through evapotranspiration (ET). Potential ET is expressed as a function

of pan evaporation. Transpiration from grass and trees, and soil evaporation are

calculated separately. Each is estimated using the concept of a potential rate adjusted by a

soil water supply index”. GRASP uses relationships developed by Scanlan et al. (1996b)

and Owens et al. (2003) to calculate runoff. These relationships apply to mid-slope

locations where surface cover, rainfall intensity and soil moisture are the main influences

on runoff, rather than topographic gradient (McIvor et al. 1995b; Scanlan et al. 1996b).

Pasture growth and death are then calculated:

“Green biomass is calculated under both water-limiting and radiation-limiting conditions

for each day with the most limiting factor estimating growth. Under water-limited

conditions, pasture growth is determined from the product of transpiration and

Chapter 2 Review of Literature ------------------------------------------------------------------------------------------------------------------------------------------ transpiration efficiency. Pasture growth under radiation-limiting conditions is

determined from intercepted solar radiation and radiation use efficiency. Indices for

temperature, nitrogen and soil water are applied to growth. Growth is partitioned

into green stem or green leaf. Green biomass can be reduced by either animal intake

or senescence.

Senescence can occur as a result of a background death rate, water stress or frost”.

Processes of detachment and litter breakdown are also defined:

“A pool containing dead biomass is maintained by GRASP. Dead biomass can either

be eaten by animals or detached from the plant to form litter. A background rate of

detachment, and detachment from trampling of animals, is simulated. Any detached

biomass is added to the litter pool. The rate of breakdown of the litter is determined

from the soil water content and temperature with litter breakdown accelerated under

wetter and hotter conditions”.

FiguMcK

Usin

wate

deep

deta

deta

(Litt

matt

____


re 2.6 Structure of the water balance model and pasture sub-model in GRASP (Littleboy and eon 1997).

g daily climate inputs, GRASP simulates a number of variables including the

r balance (runoff, infiltration, evaporation from the soil, plant transpiration and

drainage), nitrogen uptake, pasture growth (green growth, death and

chment), total standing dry matter (TSDM - the net result of growth and

chment) and animal intake (diet selection, pasture utilisation and liveweight gain)

leboy and McKeon 1997). A typical accumulating daily pasture standing dry

er curve is shown below (Figure 2.7).

_________________________________________________________________ 27

Chapter 2 Review of Literature ------------------------------------------------------------------------------------------------------------------------------------------ GRASP is also capable of calculating pasture growth over many years. Figure 2.7

shows standing dry matter over a 15-year period.

Figu(1996 GRA

in a

mod

a mo

GRA

grow

field

fram

2.5 The

mod

past

majo

and,

worl

usua

suita

____


re 2.7 Examples of GRASP model output for individual seasons, and over 15 years from Johnston ).

SP is not the only systems model designed for simulating biophysical processes

rangeland environment. Rickert et al. (2000) identify other readily available

els that may be utilised in a similar manner. However, the intended application of

del is the primary factor in choice of the appropriate modelling framework.

SP has been constructed specifically to simulate year-to-year variation in pasture

th on tropical tussock grasslands with a constant set of parameters derived from

measurements (Day et al. 1997a). As such, it is an appropriate modelling

ework for the VRD.

Testing the performance of a systems model

issue of evaluating models for their ability to represent the real system is as old as

elling itself. With respect to models of agricultural production systems (including

ure growth) Jones and Carberry (1994) state that “development … involves two

r tasks: first, the fitting of functional relationships which make up the models

secondly, the testing of these relationships and completed models against real-

d data”. Both the model fitting (calibration) and model testing (validation) tasks

lly involve evaluating the level of agreement between model predictions and

ble field data. The methods of

____________________________________________________________

28


29

evaluating model results against field data (a step common to both tasks in model

development) are the subject of this review.

Many measures of model performance using comparison of model predictions and field

observations exist. Power (1993) and more particularly Rykiel (1996) review a number of

these measures. Those relevant to this study are:

• visualisation technique, where time-series plots and other displays form the basis for

comparisons between system and model. Most often, model performance is determined

subjectively by a statement about visual goodness-of-fit;

• sensitivity analysis – the same relationships that occur in the system should also occur

in the model. Those parameters that cause significant changes in the models behaviour

should also be estimated with the greatest accuracy.

• statistical validation, which includes a variety of tests performed during both model

calibration and testing. Two common cases are: the model produces output with the

same statistical properties as the measured data from the real system; and the error

associated with critical output variables falls within specified limits.

The merits and shortcomings of these measures are briefly discussed here.

2.5.1 Time-series plots

Time-series plots of model predictions and corresponding field observations are often used

in the initial stages of model calibration, and during testing (e.g. Farre et al. 2004). Mayer

et al. (1994) discuss issues with using time-series data for independent model testing. They

argue that time-series data are not independent because they are autocorrelated; that is, the

absolute values of many variables (e.g. soil water and plant growth) are influenced by the

preceding measurements. Accordingly, time-series evaluation of model performance is a

useful visual appraisal but is generally inadequate for objectively evaluating model

performance.

A second form of time-series plots is useful when testing a model. Simulations may be run

over many years and single values representing each year can be plotted to provide a long-

term view of system dynamics (e.g. Figure 2.7b). The influence that one measured value

(e.g. maximum TSDM) has on the same variable the following year is minimal, thus much


30

reducing the incidence of autocorrelation. Time-series plots of this kind are useful tools in

such tasks as testing the sensitivity of a model to changes in individual parameter values.

2.5.2 Regression analysis

Regressions of model predictions against measured data have been commonly used to

evaluate the performance of systems models in the past, including GRASP (e.g. Johnston

1996). An ideal model would result in all predictions equalling all corresponding field

measurements, and regression of the two datasets would produce the function y = x (i.e.

unity slope and zero intercept, or the one-is-to-one line). In most cases however, this does

not occur. Mitchell and Sheehy (1997) emphasise that variation between model and system

occurs because “the model is a simplification of the real system: certain factors were

omitted or assumed constant and they produce variability in the observations which is not

reproduced by the model predictions”.

Given such variations, the task of interpreting the fit between model and system using

regression analysis can take two forms. The first form involves calculating the regression

function from the plot of model results and field measurements in the form of y = ax + b,

and determining if the slope and intercept are significantly different to one and zero

respectively. This was the approach of Johnston (1996) to validating GRASP for native

pastures in south west Queensland. This measure remains a popular performance indicator

in modelling studies (e.g. Robertson et al. 2002; Hill et al. 2006).

Alternatively, the regression coefficient can be calculated for the perfect fit function y = x

to give the statistic r2(1:1) (e.g. Asseng et al. 1998). Mayer and Butler (1993) use the term

modelling efficiency (EF) to describe this statistic. This approach was used to assess

GRASP by McKeon et al. (1990) when evaluating a range of livestock management

scenarios, and by Day et al. (1997a) in a comprehensive study of pasture productivity

across Queensland’s grazing lands. An example of this approach to model testing is

illustrated in Figure 2.8.

Debate continues over the suitability of regression analysis in model testing. Wallach and

Goffinet (1989), Harrison (1990), Mayer et al. (1994) and Mitchell (1997) provide diverse

views on the issue. In particular, Harrison (1990) and Mitchell (1997) dispute the


31

regression approach as a relevant measure of model performance because: 1) there is no

intention to make predictions from calculated regressions; 2) regressions from highly

scattered samples of points are more likely to have slopes not significantly different from

one; and 3) assumptions of the independence of the observations from each other is suspect

if they are values from a time-series or autocorrelated in any way. Despite these concerns,

regression analysis continues to be used for evaluating systems model performance.

2.5.3 Deviation of model results from field measurements

Descriptive statistics based upon the absolute differences between model predictions and

corresponding field measurements are also common measures of model performance.

Wallach and Goffinet (1989) and Jones and Carberry (1995) use prediction variance, or

Mean Squared Error of Prediction (MSEP), while Mayer and Butler (1993) use the related

statistic Root Mean Squared Error (RMSE) to evaluate models. The measures are based

upon the absolute difference (deviation) between model predictions and field

measurements. Day et al. (1993) used RMSE to evaluate GRASP performance. Mitchell

(1997) suggests limitations of this statistic for describing model performance because

single-figure summaries such as MSEP and RMSE give no indication of bias for over- or

under-prediction by the model and cannot be used to compute the precision of the

predictions.

Mitchell and Sheehy (1997) present an alternative method of assessing model performance

based upon prediction deviations. This method plots deviations (as model prediction minus

field measurement) against the field measurements (Figure 2.8). The two criteria for

adequacy of model performance are the envelope of acceptable precision around the line of

zero deviation, and the proportion of points that must lie within it. The envelope of

acceptable precision is dependent upon the application of the model. Mitchell and Sheehy

(1997) suggest acceptable precision be comparable to the precision of the field

measurements plus some factor due to models not being expected to perform as well as

measurements. Precision of measurements is suggested as equating 95% confidence limits.

Rykiel (1996) also suggests tests of model adequacy based upon not only how closely

predictions match measurements, but also how often. A proposal offered by Rykiel (1996)

is that model outputs fall within the 95% confidence limits 75% of the time for the most

important variables.


Figu 2.5. Inter

• the

accu

• on

the s

• Ry

perf

ther

be a

abili

Bion

data

____


re 2.8 Examples of two approaches to model testing from Mitchell and Sheehy (1997).

4 Interpreting model testing results

pretation of model testing results is complicated by several facts:

ability of a model to predict independent or future data does not mean it is an

rate cause and effect representation of the real system;

the other hand, a model may accurately simulate the fundamental behaviour of

ystem without quantitative accuracy; and

kiel (1996) points out that “data are not an infallible standard for judging model

ormance … we cannot assume that data accurately represent the real system and

efore constitute the best test for the model … it can be argued that the model may

better representation of reality than data that are limited by our technological

ties for measurement and subjectively biased by our perceptions of the system”.

dini et al. (1991) also recognise the likely existence of random errors in field

.

_________________________________________________________________ 32


33

The diversity of viewpoints presented here indicates that consensus on a standard approach

to systems model testing has not been reached and it is the intended use of the model that

dictates the method (or methods) of testing that are most appropriate in each case. This

study does not advance the debate on appropriate measures of model testing, but instead

recognises that using several measures is an appropriate approach when evaluating model

performance.

2.6 The role of systems modelling in grazing land management

A means of objectively estimating daily and seasonal pasture growth has many

implications for the management of pastoral land. Some of these implications are:

• Quantifying the year-to-year variability in pasture growth as a consequence of the

prevailing climate (Day et al. 2003);

• establishing safe levels of utilisation and carrying capacity for native pastures (e.g.

McKeon et al. 1994; Scanlan et al. 1994; Johnston et al. 1996b; Day et al. 1997b;

Novelly and Baird 2001);

• predicting the impact of climate change on native pasture growth (e.g. Conroy et al.

1998; Hall et al. 1998; McKeon et al. 1998);

• incorporating seasonal climate forecasting information into pasture management (e.g.

papers in Hammer et al. 2000);

• determining the risks associated with the variability of commencement, duration, and

cessation of the pasture growing season (e.g. McCown 1981b);

• regional monitoring of seasonal conditions for indications of drought and land

degradation (e.g. McKeon et al. 1996; Day et al. 1997a; Hall et al. 2001);

• reconstructing historical degradation episodes (e.g. McKeon et al. 2004)

• evaluating the likely outcomes of various grazing management options (e.g. McKeon et

al. 1990);

• development of appropriate grazing policies; and

• extrapolating the results of grazing trials to the wider pastoral landscape.

Using systems modelling for these and other applications raises the issue of scale. For

example, predicting pasture growth at the regional scale to investigate the potential

response of native pastures to climate change only requires a model that shows overall


34

trends satisfactorily. On the other hand, making ‘keep or sell livestock’ decisions at the

property or paddock level based on the risk of feed shortage according to forecast seasonal

conditions would require great confidence in model results. Rykiel (1996) recognises this

and offers: “our notion of what is acceptable depends not only on the objectives of the

model but also on the ecological scale at which the model is framed. Changes in scale may

permit aggregations that at face value are conceptually erroneous to produce operational

results that are acceptable.”

2.7 Conclusions

Examination of the literature has revealed a distinct lack of pasture production data for

grazing land in the VRD. Existing data is limited to a few land systems and a few years,

and is insufficient to understand the spatial and temporal variability that characterises the

region. This limits the development and analysis of sustainable grazing practices for the

region’s $50million-per-year pastoral industry. Overcoming this lack of data by traditional

field experimentation is not feasible, and another approach is required.

By quantifying relationships between climate, soil water supply, pasture growth and

nutrient uptake, a systems modelling framework can be used to extrapolate these

relationships across the landscape and over time. Predictions of pasture production can be

made where no field measurements are available, and these can then be used in analysis

and development of grazing practices that optimise both landscape use and enterprise

profitability. The GRASP model is a suitable framework for this task as it has been

specifically designed to simulate year-to-year variability in pasture growth in tropical

tussock grasslands.

2.8 Outline of the study ahead

This thesis takes the following steps from here:

• a field study to collect the data required to calibrate the GRASP model is described and

results presented in Chapter 3;

• results from the field study are used to derive model parameter values for individual

years at each site, a process described in Chapter 4;

• individual site parameters are the basis for developing more generic values so the

model can be applied across time and space. Assessing model results against


35

independent data provides measures of model performance when applying the model

where no field data are available. These steps are described in Chapter 5;

• identifying the parameters most influential when modelling pasture growth helps

understand the major relationships governing pasture productivity in the region. The

sensitivity of model calculations to variation in the values of several influential

parameters is demonstrated and conclusions drawn about the most important factors

affecting pasture growth in the VRD; a process described in Chapter 6;

• examples of application of a tested pasture growth model to developing and analysing

grazing land management in the VRD are presented in Chapter 7; and

• finally, the thesis is summarised, the original aims addressed, and major conclusions

are presented. (Chapter 8). Supporting data is presented in the appendices.

Chapter 3 A field study of pasture growth in the VRD

36

3.0 A field study of native pasture growth in the VRD

3.1 Introduction

Review of current literature in Chapter 2 revealed that development and analysis of

sustainable grazing practices in the VRD is hindered by a distinct lack of pasture

production data. Overcoming this paucity of data is best addressed using a simulation

modelling approach, and the GRASP model is a suitable framework for providing

predictions of pasture growth on tropical grasslands where no measurements are available.

Applying GRASP to the grazing lands of the VRD requires development of input

parameter values derived from local data (a process known as calibration). No data suitable

for this purpose has been identified in the literature.

This chapter documents a field study aimed at collecting the data necessary to calibrate

GRASP for the commonly grazed land types in the VRD. A standard method exists for

undertaking such a study (Day and Philp 1997), the rationale of which is briefly outlined in

Section 3.2. Pasture productivity varies from land type to land type, so a series of field

sites were selected that represented a range of soils and pastures in the VRD. These sites

were studied over a two-year period, with climate, soil and pasture measurements

recorded. Details of the site selection process and the measurement techniques used are

described in Section 3.3. Results of the field study presented in Section 3.4 are 1)

comprehensive; 2) the first of their kind to be collected in the region; and 3) form the

foundation for developing the capability to simulate the pasture system in later chapters.

Results are extensively discussed in an ecological context in Section 3.5 and conclusions

drawn in Section 3.6 as to the suitability of the data for use in calibrating the GRASP

model. The structure of this chapter is illustrated in Figure 3.1.


37

Figure 3.1 The structure of Chapter 3.

3.2 Rationale for obtaining data to calibrate the GRASP pasture growth

model

The methodology employed in this study is termed ‘Swiftsynd’ and is documented by Day

and Philp (1997). This methodology is an abbreviated version of the methods used in the

GUNSYNpD project (McKeon et al. 1988; McKeon and Johnston 1990). The

GUNSYNpD project sought to collect detailed biophysical data from native pasture

communities across Queensland. From this information, relationships in the pasture growth

model GRASP (McKeon et al. 1982b; Littleboy and McKeon 1997) were quantified and

tested across these pasture communities. Swiftsynd is the methodology now employed to

collect a reduced set of field data for calibrating GRASP for a new location and/or pasture

type. This methodology is applicable to all northern Australian native pasture communities

(Day and Philp 1997).

GRASP is built upon the concept that plant growth is directly related to plant water use

(transpiration). This relationship depends upon pasture species composition and soil

fertility. Another important component of the model is the ability of the pasture to grow

with little or no photosynthetic green material: i.e. from seed, re-shooting after the winter

dormancy or following defoliation (burning or grazing). Plant growth is sometimes limited

by climate or nutrient supply. Collecting field data capable of quantifying these and other

processes is the major purpose for employing the methodology used in this study. Once

Method of site establishment and data collection

(Section 3.3)

Climate results (Section 3.4.1)

Soil results (Section 3.4.2)

Pasture results (Section 3.4.3)

Discussion and conclusions (Section 3.5 and Section 3.6)

Rationale for collecting field data (Section 3.2)


38

quantified, the relationships can be used to extrapolate pasture growth from the study sites

to the wider landscape using simulation modelling.

3.3 Methods of data collection

3.3.1 A framework for structuring a study of pasture growth in the VRD

A number of resource surveys have been conducted in the Victoria River District and land

types have been classified using a range of criteria (Appendix 2). CSIR (now CSIRO)

conducted a field study of the Ord-Victoria area in 1948 and 1952. The region was

described using the land systems method of Christian and Stewart (1953). They define a

land system as “an area, or group of areas, throughout which there is a recurring pattern

of topography, soils, and vegetation”. A change in this pattern determines the boundary of

a land system. Perry (1960), and more particularly CSIRO (1970), report the field study in

detail. The land system classifications of CSIRO (1970) are still widely used to describe

the land resources of the region. Land systems form the framework used to establish sites

in this study.

Land systems in the VRD have been categorised according to parent material, soil type and

dominant topography in a manner similar to Stewart et al. (1970). Table 3.1 shows the

results of classification. The land systems selected for this study are highlighted in bold

and underlined. These land systems are large, widespread across the region, and commonly

used for grazing by the regions pastoral industry.


39

Table 3.1 Matrix of land systems by parent material and soil type (adapted from Stewart et al. 1970).

Parent Material Skeletal Soils Red Earths Yellow Earths Podzols Calcareous Cracking Clays Saline Rocky Gravelly Sandy Loamy Sandy Loamy Sandy Loamy Desert Soils Brown Grey Clays

Volcanic Granite Macphee 4

Basalt Napier 2 Frayne 4 Willeroo 4 Willeroo 4

Antrim 2 Wave Hill 4 Wave Hill 4

Sedimentary Limestone Tanmurra2 Dinnabung 4 Dinnabung 4 Nelson 4 Argyle 4 Argyle 4

Humbert 2 Jindara 4 Gordon 4

Headley 2 Wriggley 4 Montejinni 4

Sandstone Pinkerton 1 Winnecke 3 Cockatoo 4 Cockatoo 4 Hawk 4

Weaber 1 Buchanan 4

Brocks Creek 1

Wickham 1

Elder 1

Laterite Mullaman 1 Franklin 3 Wingate 4 Matheson 4 Matheson 4 Birrimba 4 Matheson 4

Coolindie 4 Barry 4

Redsan 4 Geebee 4

Shale/ Cockburn 3

Siltstone

Alluvial Angallari 4 Inverway 4 Inverway 4 Carpentaria 4

Dillinya 4

Ivanhoe 4 Ivanhoe 4

Legune 4

Topography: 1 Rugged hilly country; 2 Hilly country; 3 Low hilly to undulating country; 4 Plains (flat to gently undulating country).

Bold and underlined land systems indicate those chosen for this study. Shaded area represents parent material and soil type combinations not present in the VRD.


40

3.3.2 Site selection and establishment

Site selection

Resources available to this study (labour, time and finances) were insufficient for

comprehensive study of all land types and locations used for grazing in the district.

Therefore, a targeted sampling approach was used, focussing on the major land types and

climate zones as a compromise of what is theoretically desirable and practically

achievable. Consequently, five locations were chosen to represent the main grazing lands

in the VRD (Mt Sanford, Kidman Springs, Victoria River Downs, Rosewood and

Auvergne cattle stations - Figure 3.2) These locations have all been grazed by cattle for

many years.

Figure 3.2 The five main locations of the study sites within the Victoria River District (shaded area). Dashed

lines represent approximate average annual rainfall isohyets (Source: adapted from BoM data).

Northern Territory


41

Site establishment

In total, twenty-one sites were established throughout the VRD (Table 3.2). Sites

represented a variety of soil types and species composition combinations. Not all

combinations of soil type and pasture composition within the VRD were sampled due to

difficulty in locating particular examples, or lack of access to such sites during the pasture

growing season. Site replication took two forms. Each site was measured for two

successive growing seasons and individual years are considered replicates unless they

displayed contrasting results. Additionally, sites were grouped according to common

functional attributes such as soil type or pasture species composition, with sites within each

group also being replicates.

Sites are located on flat or gently sloping land, and the nearest trees are at least twice their

height away from the sites in most cases. These two factors are important considerations in

site location to minimise the influence of surface water movement, and tree transpiration

on the water balance. This approach makes the calculation of pasture transpiration simpler

and more reliable. The effect of trees on the soil water balance was the focus of a separate

study (Cafe et al. 1999). An over-riding constraint was that sites were located in an area

able to be accessed throughout the growing season. Each site was of uniform soil and

pasture type, and representative of the wider landscape.

Site establishment occurred late in the dry season (September to November), and each site

consisted of a 30m by 30m area fenced to exclude grazing animals (Plate 5 in Appendix 1).

All carryover pasture was removed by raking or burning (Plate 6). Native grazers, mainly

wallabies, are present in only very low populations and considered unlikely to impact on

the pasture being measured. No evidence of their grazing was detected during the data

collection period. Within each exclosed site, eight 5m by 5m ‘cells’ are marked with pegs

in each corner (Plate 7). Plant data is collected from within each of these cells (4

measurements per year for two years) at each measurement time. Figure 3.3 shows the

typical layout of an exclosure.


42

Figure 3.3 Design and layout of study site showing cells (dashed line squares) and an example pattern of

quadrat placement (solid line squares) for each pasture measurement (H1 to H8). Quadrat location for each

pasture measurement was randomised for each site.

3.3.3 Site characteristics

Mt Sanford

Sites 1 and 2 were located on gentle slopes adjacent to rocky basalt hills. They are typical

of the small areas of red earth that occur within the larger cracking clay plains in the

southern VRD. Widely scattered trees dot the landscape. Sites 3 to 6 were located on open,

gently undulating cracking clay plains, the major landscape feature of the area. Site 3 was

located on the lower part of the plains about 100m from a minor watercourse. Site 4 was

mid-plain and Site 5 was on an open gentle crest in the plain and Site 6 was mid-plain

amongst scattered trees.

Kidman Springs

Sites 7, 8, 9 and 10 were located on a flat alluvial cracking clay plain. Alluvial plains are

important grazing lands and occur throughout the region. Sites 7 and 8 were located about

H1 H4 H7

H2

H3

H5

H6

H8

H5

H1 H2

H3

H4

H6H7

H8

H1

H2

H3

H4

H5

H6

H7

H8

H2

H1

H3

H4

H5

H6

H7

H8

H8

H1

H2

H3 H4

H5

H6

H7

H7

H1 H2

H3

H4

H5

H6

H8

H1

H2

H3 H4

H5

H6 H7

H8 H4

H1 H2

H3

H5

H6

H7

H8

5.0m

5.0m

2.0m

2.0m

30.0m

30.0

m

Standing Ungrazed Pasture

Gate

Cell1m2 Quadrat

Barbed Wire Fence


43

500m apart on a treeless part of the plain and were very similar apart from diversity of

species composition. Sites 9 and 10 were also close together, but set amongst scattered

trees and appeared to differ only in their dominant pasture species. Sites 11, 12, 13 and 14

were located on red earth soils. While they are classified into two land systems (Table 3.2),

the sites all have individual characteristics that make them different to each other. The

landscape in this area contains a complex of soil types within the red earth classification,

and these are reflected by subtle differences in woody and pasture species composition.

Site 11 was in an area where the underlying limestone is occasionally exposed and Site 12

was amongst denser trees on flat, deeper soil. Site 13 was located on the margin of a large

scald with some surface erosion evident nearby, and Site 14 was in gently undulating open

woodland.

Victoria River Downs

Site 15 was in the middle of a treeless cracking clay plain about 1km from a watering point

(a billabong). Grass tussocks were somewhat pedestalled and the immediate surrounding

area showed signs of preferred grazing. Site 16 was closer to the edge of the same treeless

cracking clay plain about 2km from the billabong. There were visible differences in soil

colour and species dominance between the two sites.

Rosewood

Sites 17 and 18 were both located on undulating cracking clay plains that lie between

rocky basalt hills and represent important grazing country in this area. Site 17 was on an

upper part of these plains while Site 18 was in a lower lying area. Site 19 was located on a

moderate slope adjacent to a rocky basalt hill, the common landscape feature in this area.

Auvergne

Sites 20 and 21 were both located on flat and deep alluvial cracking clay plains that fringe

the major watercourses of the region. Site 20 was in open woodland about 5 km from the

West Baines River, and Site 21 about 10km from the river on a natural clearing within

sparse open woodland.


44

Table 3.2 General description of the 21 study sites at 5 locations in the Victoria River District (VRD).

Site Location Latitude Longitude Position in landscape Land system 1 Grazing impact 2 Soil description 3

1 Mt Sanford 170 10’ 50’’S 1300 37’ 25’’E Gentle slope, adjacent to hill Antrim High Structured red earth

2 Mt Sanford 170 12’ 17’’S 1300 36’ 41’’E Gentle slope, adjacent to hill Antrim High Structured red earth

3 Mt Sanford 170 11’ 15’’S 1300 36’ 08’’E Lower plain, 100m from watercourse Wave Hill Low Red cracking clay

4 Mt Sanford 170 13’ 12’’S 1300 38’ 21’’E Mid-plain Wave Hill High Red cracking clay

5 Mt Sanford 170 13’ 23’’S 1300 34’ 56’’E Gentle crest Wave Hill Low Black cracking clay

6 Mt Sanford 170 13’ 10’’S 1300 33’ 49’’E Mid-plain Wave Hill Medium Red cracking clay

7 Kidman Springs 160 05’ 08’’S 1310 00’ 54’’E Lower plain Argyle Medium Grey cracking clay

8 Kidman Springs 160 05’ 08’’S 1310 01’ 00’’E Lower plain Argyle Low Grey cracking clay

9 Kidman Springs 160 04’ 11’’S 1300 59’ 40’’E Lower plain Argyle Medium Grey cracking clay

10 Kidman Springs 160 04’ 32’’S 1300 59’ 40’’E Lower plain Argyle Low Grey cracking clay

11 Kidman Springs 160 05’ 04’’S 1300 56’ 59’’E Mid-plain Humbert Low Calcareous red earth

12 Kidman Springs 160 05’ 24’’S 1300 55’ 15’’E Mid-plain Dinnabung Medium Structured red earth

13 Kidman Springs 160 06’ 27’’S 1310 00’ 23’’E Mid-plain Humbert High Structured red earth

14 Kidman Springs 160 08’ 03’’S 1300 55’ 20’’E Mid-plain Dinnabung Low Massive red earth

15 Victoria River Downs 160 23’ 24’’S 1300 58’ 48’’E Lower plain, 1km from watering point Ivanhoe High Red cracking clay

16 Victoria River Downs 160 24’ 02’’S 1300 57’ 40’’E Lower plain, 2km from watering point Ivanhoe Low Grey cracking clay

17 Rosewood 160 24’ 24’’S 1290 02’ 57’’E Gentle crest Wave Hill Medium Black cracking clay

18 Rosewood 160 26’ 56’’S 1290 02’ 29’’E Lower plain Wave Hill Low Red cracking clay

19 Rosewood 160 28’ 40’’S 1290 08’ 15’’E Moderate slope, adjacent to hill Antrim Medium Structured red earth

20 Auvergne 150 42’ 45’’S 1290 57’ 29’’E Lower plain Ivanhoe Medium Grey cracking clay

21 Auvergne 150 47’ 19’’S 1290 59’ 50’’E Lower plain Ivanhoe Low Grey cracking clay

1 Land system according to Stewart et al. (1970); 2 Visual assessment of the impact of grazing at each site based mainly on pasture species composition; 3 Soil description from Isbell (1996).


45

3.3.4 Grazing history prior to establishment of study sites

Mt Sanford

Study site locations were all part of a large paddock (212km2) that had been grazed for

many years. During the year prior to site establishment the paddock was partially

subdivided to establish a stocking rate demonstration, and the sites for this study were

located within this area. The short period of subdivision prior to site establishment is

assumed not to have impacted on the initial state of the sites. There was evidence (e.g. few

perennial grasses and low ground cover) of past localised heavy grazing where Sites 1, 2

and 4 were located. Sites 3, 5 and 6 were established in areas that appeared to have been

grazed only moderately.

Kidman Springs

This property has experienced periods of heavy grazing pressure in the past, particularly

the red earth pastures, when the land had a large feral animal population (mainly donkeys)

several decades ago. Since becoming a government research station in the late 1960’s

Kidman Springs has experienced moderate stocking rate and pastures generally appeared

to be in good condition.


The paddock in which the study sites were located is of medium size (60km2) and has had

past grazing pressure varying from none or light in some years, to heavy in others. Grazing

pressure was moderate at the time of site establishment.

Rosewood

The sites were all located within paddocks used for grazing for many years. The previous

decade had seen moderate stocking rates, although the property had been heavily grazed

prior to then, due in part to a large feral animal population (cattle, horses and donkeys).

Auvergne

The sites were within medium-sized paddocks (56 km2 and 39 km2) grazed for many years.

The large number of juvenile native woody species around Site 20 was probably due to the

impact of grazing and/or changes to the natural fire regime of the landscape. The paddock


46

containing Site 21 had experienced heavy grazing in the past but it is not known to what

extent this has influenced species composition at the time of site establishment.

3.3.5 Collection of data

Rainfall

All sites had a raingauge installed at establishment and, where possible, rainfall was

recorded after each rain event. Sites varied in the reliability of recorded gaugings. At Mt

Sanford sometimes several rain events occurred between readings. In these cases, the total

amount was apportioned into relative amounts for each event according to the nearest

reliable daily recording station (Mt Sanford Station homestead, 35km to the north).

Otherwise, they were a reliable record. At Kidman Springs due to the closeness (about

500m) of Sites 7 and 8, only one raingauge was used for these two sites. All other sites

have individual gauges. All records at Kidman Springs were considered reliable. At

Victoria River Downs, Rosewood and Auvergne, gaugings were recorded so infrequently

that they were considered unreliable and the daily records from the respective homesteads

(5km, 5-15km, and 10km away from the sites respectively) are the closest reliable record

of rainfall at these sites. Victoria River Downs homestead is a Bureau of Meteorology

(BoM) recording station and other climatic variables such as temperature and pan

evaporation were collected in addition to rainfall.

Soil measurements

Soil measurements were confined to the top metre of the soil profile as pasture root depth

in the nutrient-poor and highly weathered ancient soils dominating the region tends to

concentrate plant root systems in the surface layers (Schmidt and Lamble 2002). Three soil

cores were taken at random locations within the exclosure (but not inside the pasture

sampling cells) at each harvest to determine soil moisture content (Plate 8 in Appendix 1).

Soil cores were hand-augered to a depth of 1.0m, unless rock or hard dry soil prevented

reaching this depth. The soil core was collected in 10cm graduations. In the laboratory the

samples were weighed, dried at 1050C for 48 hrs, and weighed again to determine

gravimetric soil moisture content.

GSM (g/g) = (wet soil wt (g) – dry soil wt (g)) / dry soil wt (g) (Equation 2) GSM= gravimetric soil moisture content


47

Soil bulk density (BD) was measured using ring samples taken from the face of a pit dug

adjacent to each site (Plate 8). Samples were collected at several depths (0-10cm, 10-20cm,

20-30cm, 50-60cm, and 80-90cm) using a standard procedure (Rowell 1994, and in more

detail with illustrations in Dalgliesh and Foale 1998). Bulk density is calculated as the

weight of oven-dry soil per volume of soil and is expressed as grams per cubic centimetre

(g/cm3).

Once bulk density was determined, volumetric soil moisture contents for each 10cm soil

increment were calculated. For an increment where a bulk density sample was not taken

(e.g. 30-40cm), the closest BD value was used (e.g. 20-30cm). Where a soil bulk density

sample was unable to be obtained below a certain depth, the value for the deepest BD

sample taken was used to convert all gravimetric values to volumetric values for

increments collected deeper in the profile.

VSM (mm) = GSM (g/g) * SBD (g/cm3) (Equation 3) VSM = volumetric soil moisture content

GSM= gravimetric soil moisture content

SBD = soil bulk density

The soil profile at each site was described using the same pit as for bulk density sampling.

Resource Assessment Officers from NT Dept. Lands described the soil profile using the

methodology of McDonald et al. (1990) and the classification system of Isbell (1996). Soil

chemistry properties of each site were also measured from samples collected at these soil

pits. A wide range of chemical properties were determined by laboratory analysis for the 0-

10cm and 10-20cm layers using standard analysis techniques of Rayment and Higginson

(1982). The results of laboratory analysis make only a minor contribution to the focus of

this thesis and, as such, are presented in the appendices (Table 9.5 to Table 9.8 in

Appendix 3). A brief discussion of the analysis results as they relate to this study appears

in Section 3.5.1.

Pasture measurements

Carryover pasture material was removed at establishment (Section 3.3.6) and the initial

conditions at the site described. This approach allows for subsequent pasture measurements


48

to reflect the current season without the potentially complicating factor of plant material

present that has originated in previous seasons.

During the study period, pasture measurements (called ‘harvests’) were taken at four

strategic times during each year.

• The first harvest generally took place about 6 weeks after the break of the season when

the initial growth flush is peaking. In the GRASP model, growth during this period is

dependent primarily on the ability of the pasture to regrow from stored reserves (i.e.

after dormancy or defoliation).

• The second harvest was timed mid-wet season, around the general onset of pasture

flowering. In GRASP, growth during this second period is governed primarily by

transpiration-use-efficiency and soil water limitations.

• The third harvest was at maximum standing biomass, after the wet season rains had

ceased. In GRASP, growth in this third period is often limited by nitrogen availability,

water stress and physiological maturity.

• The fourth harvest (end of the dry season) indicates detachment rates, or natural pasture

breakdown. Sites were once again cleared of remnant vegetation by burning, and the

data collection repeated for a second year.

Plant data collected at each harvest was comprehensive (Plate 9 in Appendix 1). A 1m2

quadrat was placed in a unique position within a cell. In order to minimize bias in pasture

data due to the potential effect a cut quadrat can have on neighbouring uncut quadrats, a

buffer zone of 1 metre between quadrats was enforced (Figure 3.3).

• Each quadrat was photographed from as close to vertical as possible.

• Visual estimations of cover (green plant material, dead plant material, detached litter,

rocks, and bare ground percentages) and pasture composition (dominant grass, sub-

dominant grass, other grasses, and dicots, as percentage of total yield) were recorded.

• Plant height was measured by gently lowering a light plate (560mm by 420mm, 210g)

down a metre rule onto the pasture and recording height (cm) above the ground. For

each quadrat, two measurements of plant height were taken so that most of the quadrat

was covered.

• The quadrat was cut at close to ground level (about 2cm) and the plants separated into

four categories (dominant grass, sub-dominant grass, other grasses, and dicots), placed

in sealed plastic bags, and weighed within 6 hours.


49

• Litter was collected from each quadrat, bulked into the one bag for the site, weighed,

and averaged across quadrats.

• After weighing, samples collected from all eight harvested quadrats were bulked

according to their species category and thoroughly mixed. Sub-samples were taken

from each bulked category and oven dried at 800C for 48hrs to determine the average

dry matter content. These sub-samples were then ground and chemically analysed to

determine nitrogen and phosphorus contents.

Once per year, by mid-way through the growing season, the basal area of perennial grasses

within each site was measured (Plate 9). Only perennial grasses were recorded except

where none were present. In these cases, annual grasses were recorded. Basal area was

measured by placing a 5-pin point frame 20 times within each cell (100 observations) and

recording the number of direct ‘strikes’. A strike was defined as when a pin rests exactly at

a point where the green stem of a grass plant emerges from the soil. This measure is at risk

of operator bias as judgements are involved (what constitutes a strike?) and as such is a

partially subjective assessment. To reduce the impact of subjectivity, the same operator

assessed all sites in this study, and it is assumed any risk of bias is equal for all recordings.

Perennial grass basal area (PGBA) was then calculated as a percentage value:

PGBA (%) = (no. of strikes / no. of point samples)*100 (Equation 4)

The entire suite of plant species present was recorded once during the study period at each

site. The species identification nomenclature of Dunlop (1990) and Barritt (1996) is used

and all species presented in this thesis are found in Appendix 4. An example of the plant

species record for a site appears in Appendix 5.

3.3.6 Removal of carryover material

All sites were cleared of any carryover material at the beginning of the study period and at

the end of the first study year. Burning was prescribed in the methodology, but as sites

were established on grazed land, many did not have sufficient pasture present to carry a

fire. Thus, methods of initially removing plant material depended on the amount of

material present. Some sites were bare and did not require any removal. Other sites were

patchy and raking was required to draw together enough material for burning. Others could


50

support a fire. Individual site details are given in the following paragraphs. All sites were

burnt at the end of the first study year as sufficient pasture was present to support a fire in

all cases.

Sites 1 and 2 at Mt Sanford were raked clear of carryover plant material at establishment

while for Sites 3 to 6, carryover material was removed by burning.

Sites 7-10 at Kidman Springs were burnt at establishment while Sites 11 and 12 were raked

clean of carryover material. During establishment of Site 12 several trees and woody forbs

were removed but these partially regrew over the study period. This regrowth is assumed

not to have impacted on site measurements. At establishment of Site 13, the area had been

grazed bare of all plant material. Site 14 was raked and burnt at establishment.

Sites 15 and 16 at Victoria River Downs were burnt at establishment and at the end of the

first study year. Sites 17-19 at Rosewood were raked and burnt at establishment. Little

carryover material was present at the establishment of Site 19 and no removal was

necessary.

3.4 Field results

Results from the field study of 21 sites over two years are presented in this section.

Climate, soil and pasture results are presented in considerable detail as they are crucial to

calibration of GRASP model parameters in Chapter 4. Some additional data, related but

not crucial to the focus of this thesis, are presented in Appendix 3 to Appendix 6.

The data collection period was not identical for all sites, thus results presented as time-

series sometimes span different periods e.g. 1993-95 or 1994-96. This must be kept in

mind when comparing data across sites or locations. Figure 3.4 shows the timing of

management and data collection at each site.


51

Figure 3.4 Time-series of management and data collection from each site over the study period.

3.4.1 Climate data

Climatic conditions during the measurement period are crucial to understanding the

processes that drive plant growth (Section 2.3.1). Apart from rainfall, climate variables

exhibit little spatial variability across the landscape unless a dominant topographic feature

exists. Hence, one set of meteorological data for each of the five major site locations

describes the climate conditions other than rainfall. This data is sourced from the Qld Dept

of Natural Resources Data Drill (DataDrill 2005) that contains interpolated data derived

from Bureau of Meteorology (BoM) weather recording stations in the district. While daily

rainfall records exist for all sites at Mt Sanford and Kidman Springs, only averages are

presented here for these locations. Attempts were made to collect daily rainfall for each

individual site at Victoria River Downs, Rosewood and Auvergne, but these were

unreliable and station homestead records are the best data available.

3.4.1.1 Rainfall during the measurement period

Seasonal rainfall totals (July to June) recorded in this study are ranked against historical

rainfall data at the same location using the percentiles approach (Table 3.3). For example,

seasonal rainfall ranked as the 90th percentile indicates that 90% of historical rainfall totals

123456789

101112131415161718192021Jul-93 Jan-94 Jul-94 Jan-95 Jul-95 Jan-96 Jul-96

Site

No.

Soil and pasture measurementsCarryover material removedGrass basal areaFloristicsBulk density and soil chemistry


52

at that location are lower than the total recorded in this study. Median rainfall is expressed

as the 50th percentile.

The 1993/94 wet season was around median at the Mt Sanford sites (seasonal totals = 38th

to 55th percentiles - Table 3.3). Rainfall equalled or exceeded the median in every month

except January 1994, and stopped abruptly in March (Figure 3.6). Kidman Springs sites

experienced median to above median rainfall (55th to 78th percentiles) including a well

above median December 1993 (Figure 3.8) and a well below median in February 1996.

The 1994/95 wet season was well above the median at Mt Sanford (72nd to 89th

percentiles). Monthly totals from December 1994 to March 1995 were all well above

median values (Figure 3.6). At Kidman Springs, rainfall totals were median to above

median, (50th to 79th percentiles). Victoria River Downs (87th percentile) received above

median rainfall from November 1994 to March 1995. An extremely wet January and

February 1995 (Figure 3.9 and Figure 3.10) resulted in much of the area being waterlogged

for weeks. Rosewood (83rd percentile) had close to median monthly values (Figure 3.12)

except for January 1995, which was the fourth wettest January in 109 years. Auvergne

(34th percentile) was just below median (only 8% less than the 47-year median value), also

with a wet January (Figure 3.14).

The 1995/96 wet season varied across the district with Kidman Springs (45th to 60th

percentiles) receiving about median rainfall, Rosewood (54th percentile) about median,

Auvergne (70th percentile) above median, and Victoria River Downs (97th percentile) had

well above its median total (Table 3.3). A dry spell occurred during January and February

1996 at each location except Auvergne (Figure 3.7, Figure 3.9, Figure 3.11 and Figure

3.13).

A summary of rainfall over the observation period is shown in Table 3.3. For convenience,

Figure 3.5 to Figure 3.14 show only one value for daily or monthly rainfall at each of the

five locations, although individual site data exists for Mt Sanford and Kidman Springs. For

these two locations, sites were averaged to produce the single values presented in Figure

3.5 to Figure 3.8. The individual site data is used when calculating the soil water balances

in Chapter 4.

Chapter 3 A field study of pasture growth in the VRD ------------------------------------------------------------------------------------------------------------------------------------------

Tabl2003 ____


e 3.3 Measured seasonal rainfall (July - June) over the study period, the 47-year (1957/58 – /04) median value, and rainfall percentiles. Median data calculated from DataDrill (2005).

_________________________________________________________________ 53

Chapter 3 A field study of pasture growth in the VRD ------------------------------------------------------------------------------------------------------------------------------------------ Mt Sanford


Figure 3.5 Maximum and minimum daily temperatures (0C) at Mt Sanford (17012’S, 130036’E) over the study period (upper graph) compared to the 7-day moving averages of 47-year values (1957/58 – 2003/04); and measured daily rainfall averaged across Sites 1-6 (lower graph). Temperature data derived from DataDrill (2005).

Figumedi ____


re 3.6 Total monthly measured rainfall at Mt Sanford averaged across Sites 1-6, and 47-year an values (1957/58 – 2003/04) derived from DataDrill (2005).

_____________________________________________________________ 54

Chapter 3 A field study of pasture growth in the VRD ------------------------------------------------------------------------------------------------------------------------------------------ Kidman Springs


Figure 3.7 Maximum and minimum daily temperatures (0C) at Kidman Springs (16006’S, 131000’E) over the study period (upper graph) compared to the 7-day moving averages of 47-year values (1957/58 – 2003/04); and measured daily rainfall averaged across Sites 7-12 for 1993/94, Sites 7-14 for 1994/95 and Sites 13-14 for 1995/96 (lower graph). Temperature data derived from DataDrill (2005)

FiguSitesderiv ____


re 3.8 Total monthly measured rainfall at Kidman Springs averaged across Sites 7-12 for 1993/94, 7-14 for 1994/95 and Sites 13-14 for 1995/96, and 47-year median values (1957/58 – 2003/04) ed from DataDrill (2005).

_____________________________________________________________ 55

Chapter 3 A field study of pasture growth in the VRD ------------------------------------------------------------------------------------------------------------------------------------------ Victoria River Downs


Figure 3.9 Maximum and minimum daily temperatures (0C) at Victoria River Downs (16024’S, 131006’E) over the study period (upper graph) compared to the 7-day moving averages of 47-year values (1957/58 – 2003/04); and measured daily rainfall at the station homestead (lower graph). Temperature data derived from DataDrill (2005).

Figuvalue ____


re 3.10 Total monthly measured rainfall at Victoria River Downs homestead, and 47-year median s (1957/58 – 2003/04) derived from DataDrill (2005).

_____________________________________________________________ 56

Chapter 3 A field study of pasture growth in the VRD ------------------------------------------------------------------------------------------------------------------------------------------ Rosewood


Figure 3.11 Maximum and minimum daily temperatures (0C) at Rosewood (16030’S, 129000’E) over the study period (upper graph) compared to the 7-day moving averages of 47-year values (1957/58 – 2003/04); and measured daily rainfall at the station homestead (lower graph). Temperature data derived from DataDrill (2005).

Figu(1957 ____


re 3.12 Total monthly measured rainfall at Rosewood homestead, and 47-year median values /58 – 2003/04) derived from DataDrill (2005).

_________________________________________________________________ 57

Chapter 3 A field study of pasture growth in the VRD ------------------------------------------------------------------------------------------------------------------------------------------ Auvergne


Figure 3.13 Maximum and minimum daily temperatures (0C) at Auvergne (15024’S, 130000’E) over the study period (upper graph) compared to the 7-day moving averages of 47-year values (1957/58 – 2003/04); and measured daily rainfall at the station homestead (lower graph). Temperature data derived from DataDrill (2005).

Figu(1957 ____


re 3.14 Total monthly measured rainfall at Auvergne homestead, and 47-year median values /58 – 2003/04) derived from DataDrill (2005).

_________________________________________________________________ 58


59

3.4.1.2 Other climatic variables

Examination of temperature data in Figure 3.5, Figure 3.7, Figure 3.9, Figure 3.11 and

Figure 3.13 reveal several notable features. Daily maximum temperatures during the

pasture growth period are consistently high but only occasionally exceed 400C. Most

recordings are close to average, however several exceptions occur. During sustained

monsoonal activity (e.g. March 1994 at Mt Sanford, Figure 3.5), a series of below average

maximums have been recorded, likely due to continuous cloud cover. Conversely, a hot

dry spell occurred during January and February 1996 at each location, most notably

Rosewood, (Figure 3.11) with a cluster of maximum temperatures close to 400C.

During the mid-dry season of each study year (June to August) a number of maximum and

minimum temperature observations are well below the 47-year average values at all

locations. These cool days and cold nights resulted from strong weather systems from the

temperate south pushing their way into this tropical region and displacing the normally

mild conditions at this time of the year.

Other climatic variables are shown in Table 3.4. This data is interpolated from the nearest

recording stations using QDNR Data Drill and is available for each study location,

although only data from Victoria River Downs are presented. The much gentler gradient in

these variables across the VRD suggest the other localities experienced similar general

trends during the study period.

Radiation levels are quite consistent throughout the year and close to long-term means.

Evaporation demand is very high throughout the year, particularly during the growing

season. At Victoria River Downs, annual pan evaporation (2772mm) is about four times

the median rainfall (696mm) and evaporation during each study year was similar. Vapour

pressure is highly seasonal with daily averages ranging from 10hPa in July to 28hPa in

February. Observations throughout the growing seasons roughly follow long-term trends

with some departures from average due to dominant weather conditions at the time.

Chapter 3 A field study of pasture growth in the VRD ------------------------------------------------------------------------------------------------------------------------------------------ Table 3.4 Monthly values for radiation, evaporation and vapour pressure deficit at Victoria River Downs during the study period (source: DataDrill 2005).


3.4.2 Soil data Results of soil measurements are presented within a framework of site groups

according to the soil type and its geologic origin. These groups have close agreement

with the land system framework initially used to select sites. This is not surprising

given that soil type and geology are two main factors in the land system classification

of Stewart et al. (1970). Antrim land system soils are described here as red earths

overlying basalt, Humbert and Dinnabung land system soils are described as red

earths overlying limestone, Wave Hill land system soils are described as cracking

clays overlying basalt, and Argyle and Ivanhoe land system soils are described as

alluvial cracking clays.

When interpreting the results presented in this section, particularly soil water

observations, several points must be remembered. Due to the isolated location of

some study sites (Mt _____________________________________________________________________

60


61

Sanford Sites 1-6 and Rosewood Sites 17-19), access following major rain events is only

possible after a certain amount of surface drying has occurred. Thus, the actual field

capacity may differ from the highest measured value. Soil water wilting points were

measured at each of these sites some considerable time (months) after the last rain and are

reliable. However for Kidman Springs, Victoria River Downs and Auvergne, more reliable

access to study sites allows soil water measurements to be collected at times that give

reasonable estimates of both field capacity and wilting point.

All cracking clay soil sites crack deeply (to at least 1.0m) when drying and take

considerable rain (an estimated 100mm or more) to close the cracks. The cracks usually

close during November or December storms. The red earth sites are non-cracking.

In this study, the difference in soil water content between field capacity and wilting point is

termed the ‘water holding capacity’ (WHC) of the soil, after Mott et al. (1985). This soil

characteristic is also known as ‘plant available water capacity’ (e.g. in the APSIM model,

McCown et al. 1996), but this term is not used here in order to prevent confusion with the

term ‘plant available water content’ (PAWC, MLA 2004a) which is the actual amount of

soil water above wilting point at any point in time, used in later chapters of this thesis.


62

3.4.2.1 Red earths overlying basalt

Three sites (Sites 1, 2 and 19) are located on red earth soils of basalt origin (Plate 10 in

Appendix 1). The soil depths given in Table 3.5 are those recorded during the soil profile

descriptions by Department of Lands officers (Section 3.3.5). However, during the field

measurements of soil moisture, soil cores were able to be extracted down to 1.0m most

times. The red earths at these sites overlay weathering basalt that is somewhat soft and

porous and able to be penetrated using a hand auger. Plant roots are likely to enter and

extract water from this weathering basalt layer and therefore the effective pasture rooting

depths reflect soil moisture measurement results rather than those of the profile

descriptions.

Table 3.5 Soil description for sites located on red earths overlying basalt.

Site Description Soil order 1

Soil texture Profile Soil depth

Assumed pasture

rooting depth

1 Structured red

earth

Dermosol Heavy clay throughout Slightly gravelly 0.65m 1.0m

2 Structured red

earth

Dermosol Sandy loam over heavy

clay sub-soil

Non-gravelly 0.35m 1.0m

19 Structured red

earth

Dermosol Light-medium clay

throughout

Gravelly 0.3m 1.0m

1 From Isbell (1996).

Table 3.6 Physical properties of soils at sites located on red earths overlying basalt.

Site Air Dry Moisture

Content (%)

Bulk Density (g/cm3)

Water Holding Capacity (mm)

0-10 cm

10-20 cm

0-10 cm

10-20 cm

20-30 cm

50-60 cm

80-90 cm

0-10 cm

10-50 cm

50-100 cm

Total

1 4.5 5.1 1.44 1.51 1.65 1.65 1 1.65 1 15.6 68.4 66.3 150.2

2 3.9 4.9 1.48 1.44 1.51 1.51 1 1.51 1 14.3 76.4 57.9 148.6

19 2.1 5.2 1.58 1.43 1.51 1.51 1 1.51 1 6.7 70.1 56.2 133.0

1 Bulk density unable to be measured at this depth and value from previous depth has been used.


63

Figure 3.15 Field-measured soil water contents of sites located on red earths overlying basalt.

Solid vertical lines with data points represent measured soil water contents throughout the study period;

shaded areas indicate derived water holding capacity based on wettest and driest recorded profiles; and

dashed horizontal lines represent a texture change from structured red earth to weathering basalt.

The format of Figure 3.15 to Figure 3.18 is adapted from McCown (1971) and Greacen

and Williams (1983). In these figures the estimated soil water capacity for the surface layer

(0-10cm) is estimated from the soil layer immediately below it (10-20cm) rather than the

measured values. This approach is taken because it is assumed the lowest values measured

at the surface represent air-dry rather than wilting point. Similarly, the maximum measured

values of the 0-10cm layer were taken after a certain amount of surface drying had

occurred (necessary for access to sites). Therefore field capacity for the soil surface is also

estimated from the layer immediately below it.

Site 1

0

20

40

60

80

100

0.0 0.2 0.4 0.6Volumetric water content (mm/mm)

Dep

th (c

m)

16/11/1993 2/02/199411/03/1994 5/05/199414/09/1994 4/01/199514/03/1995 26/04/1995

Site 2

0

20

40

60

80

100


Dep

th (c

m)

16/11/1993 2/02/199411/03/1994 4/05/199414/09/1994 4/01/199515/03/1995 27/04/199529/11/1995

Site 19

0

20

40

60

80

100


Dep

th (c

m)

8/11/1994 14/12/19947/02/1995 19/04/19952/11/1995 4/01/199614/02/1996 6/06/1996


64

3.4.2.2 Red earths overlying limestone

Four sites (Sites 11, 12, 13 and 14) are located on red earth soils of limestone origin (Plate

11 in Appendix 1). These soils are quite shallow (Table 3.7) and the soil water results

reflect this (Table 3.8). Unlike the red earths in the previous section, the underlying rock

layer at these sites could not be penetrated during sample collection and therefore little soil

water data was obtained below 50cm at Site 13, or below 60cm at Sites 11, 12 and 14.

Pasture rooting depth and the approximation of plant available water capacity has been

limited to these depths as represented by the lower limit of the shaded areas in Figure 3.16.

Table 3.7 Soil description for sites located on red earths overlying limestone.


Soil texture Profile Soil depth

Assumed pasture

rooting depth

11 Calcareous red

earth

Calcarosol Not given 50% CaCo3 nodules

between 0.2-0.4m

0.4m 0.6m

12 Structured red

earth

Dermosol Light clay over medium-

heavy clay sub-soil


13 Structured red

earth

Dermosol Sandy clay loam over

medium-heavy clay sub-

soil


14 Massive red

earth

Kandosol Sandy clay loam over

light-medium clay sub-

soil

Very gravelly 0.65m 0.6m

1 From Isbell (1996).

Table 3.8 Physical properties of soils at sites located on red earths overlying limestone.


Content (%)



0-10 cm

10-20 cm

0-10 cm

10-20 cm

20-30 cm

50-60 cm

0-10 cm

10-50 cm

50-60 cm

Total

11 2.6 2.7 1.59 1.48 1.48 1 1.48 1 21.7 44.7 7.0 73.4

12 2.6 3.1 1.46 1.58 1.47 1.56 15.9 49.0 12.1 77.0

13 3.4 9.1 1.70 1.65 1.69 - 26.0 83.6 - 109.6

14 3.5 6.9 1.57 1.57 1.68 1.92 21.7 44.7 7.0 73.4

1 Bulk density unable to be measured at this depth and value from previous depth has been used.


65

Figure 3.16 Field-measured soil water contents of sites located on red earth overlying limestone.



dashed horizontal lines represent a texture change from red earth to limestone.

Site 14

0

20

40

60

80

100


Dep

th (c

m)

8/09/1994 12/01/19956/03/1995 3/05/199517/10/1995 10/01/199629/02/1996 28/05/199626/11/1996

Site 11

0

20

40

60

80

100


Dep

th (c

m)

25/11/1993 20/01/199416/03/1994 10/05/199422/09/1994 10/01/19956/03/1995 3/05/199517/10/1995

Site 12

0

20

40

60

80

100


Dep

th (c

m)

25/11/1993 20/01/199417/03/1994 10/05/199422/09/1994 12/01/19957/03/1995 3/05/199517/10/1995

Site 13

0

20

40

60

80

100


Dep

th (c

m)

7/09/1994 12/01/19957/03/1995 4/05/199517/10/1995 9/01/199629/02/1996 28/05/199626/11/1996


66

3.4.2.3 Cracking clays overlying basalt

Six sites (Sites 3, 4, 5, 6, 17 and 18) are located on cracking clay soils of basalt origin

(Plate 12 in Appendix 1). These soils are deeper than the red earths of similar origin

(Section 3.4.2.1). In contrast to the red earth sites in Section 3.4.2.1, penetrating the

underlying basalt layer with a soil auger was very difficult and few samples were collected

from this basalt layer. It is therefore assumed that plant roots do not extend into this layer

at these sites, and effective pasture rooting depth is limited to the deepest reliable sampling

depth as shown in Table 3.9 and indicated by the dashed lines in Figure 3.17.

Table 3.9 Soil description for sites located on cracking clays overlying basalt.


Soil texture Profile 2 Depth Assumed pasture

rooting depth

3 Red cracking

clay

Vertosol Heavy clay throughout Gravelly 1.4m 1.0m

4 Red cracking

clay

Vertosol Medium clay over heavy

clay sub-soil


5 Black cracking

clay


clay sub-soil

Gravelly 1.25m 1.0m

6 Red cracking

clay

Vertosol Heavy clay throughout Gravelly 0.75m 0.7m

17 Black cracking

clay

Vertosol Heavy clay throughout Moderately gravelly 0.9m 0.9m

18 Red cracking

clay

Vertosol Heavy clay throughout Non-gravelly 0.8m 0.8m

1 From Isbell (1996). 2 All these cracking clays have a self-mulching surface.


67

Table 3.10 Physical properties of soils at sites located on cracking clays overlying basalt.


Content (%)



0-10 cm

10-20 cm

0-10 cm

10-20 cm

20-30 cm

50-60 cm

80-90 cm

0-10 cm

10-50 cm

50+ cm

Total

3 6.3 7.0 1.18 1.22 1.51 1.44 1.53 14.8 148.7 129.5 2 292.9

4 7.1 7.3 1.35 1.43 1.52 1.47 1.62 21.3 137.8 117.6 3 276.7

5 6.8 7.7 1.39 1.49 1.52 1.54 1.66 30.5 147.5 132.1 2 310.1

6 7.6 7.4 1.26 1.42 1.55 1.57 - 21.2 173.4 60.0 5 254.5

17 6.4 6.5 1.35 1.47 1.57 1.76 1.76 1 26.4 155.3 133.4 3 315.2

18 7.5 7.9 1.29 1.34 1.39 1.50 - 20.3 131.1 63.0 4 214.4

1 Bulk density unable to be measured at this depth and value from previous depth has been used. 2 WHC 50-100cm; 3 WHC 50-90cm; 4 WHC 50-80cm; 5 WHC 50-70cm.

Figure 3.17 Field-measured soil water contents of sites located on cracking clays overlying basalt.



dashed horizontal lines represent a texture change from cracking clay to basalt. (Figure continued overleaf)

Site 3

0

20

40

60

80

100


Dep

th (c

m)

16/11/1993 2/02/199412/03/1994 4/05/199415/09/1994 4/01/199514/03/1995 26/04/199517/11/1995

Site 4

0

20

40

60

80

100


Dep

th (c

m)

17/11/1993 1/02/199412/03/1994 4/05/199413/09/1994 5/01/199513/03/1995 27/04/199515/11/1995


68

Figure 3.17 (cont.)

Site 5

0

20

40

60

80

100


Dep

th (c

m)

17/11/1993 1/02/199413/03/1994 4/05/199413/09/1994 5/01/199514/03/1995 26/04/199515/11/1995

Site 6

0

20

40

60

80

100


Dep

th (c

m)

17/11/1993 1/02/199413/03/1994 4/05/199413/09/1994 5/01/199514/03/1995 26/04/199528/11/1995

Site 17

0

20

40

60

80

100


Dep

th (c

m)

8/11/1994 14/12/19948/02/1995 20/04/19953/11/1995 5/01/199615/02/1996 6/06/199628/11/1996

Site 18

0

20

40

60

80

100


Dep

th (c

m)

8/11/1994 15/12/19948/02/1995 19/04/19952/11/1995 4/01/199614/02/1996 5/06/199628/11/1996


69

3.4.2.4 Alluvial cracking clays

Eight sites (Sites 7, 8, 9, 10, 15, 16, 20 and 21) are located on cracking clay soils of

alluvial origin (Plate 13 in Appendix 1). These soils are very deep as shown in Table 3.11.

Sampling depth is restricted to 1.0m in accordance with the Swiftsynd methodology and

limitation of sampling equipment. Many pasture species growing on these alluvial soils are

long-lived (e.g. Astrebla spp. - Williams and Roe 1975; Orr and Evenson 1991b) and their

root systems may extend deep into the soil profile. However, Christie (1981) reports 85%

of root biomass in an Astrebla grassland in Queensland occurring in the top 40cm of the

soil and considers this zone the most important for pasture growth. Accordingly, effective

pasture rooting depth on alluvial soils is limited to 1.0m in this study as most water use

will be from this zone.

Table 3.11 Soil description for sites located on cracking clays of alluvial origin.


Soil texture Profile 2 Soil depth

Assumed pasture

rooting depth

7 Grey cracking

clay

Vertosol Heavy clay throughout Non-gravelly >2.0m 1.0m

8 Grey cracking

clay


9 Grey cracking

clay


10 Grey cracking

clay


15 Red cracking

clay


clay sub-soil

Non-gravelly >2.0m 1.0m

16 Grey cracking

clay

Vertosol Heavy clay throughout,

overlying weathering

calcium carbonate layer


20 Grey cracking

clay

Vertosol Heavy clay throughout,

overlying red clay sub-

soil


21 Grey cracking

clay

Vertosol Medium clay surface

over heavy clay.

Overlying massive to

weakly structured buried

soil from 0.9m


1 From Isbell (1996). 2 All these cracking clays have a self-mulching surface.


70

The plant available water capacity shown for Site 15 in Figure 3.18 and Table 3.12 is likely

to be an over-estimation. The figure shows one measurement (9 March 1995) recorded

very high values. This data was collected soon after several days of heavy rain and the site

and surrounding area were inundated with water. Many surface depressions still contained

water that had neither run off nor infiltrated (Plate 22 in Appendix 1). While collecting soil

samples, water flowed into the core holes laterally from the surrounding soil. Thus, data

represents waterlogged conditions rather than field capacity. While this data is included in

Figure 3.18 and Table 3.12 as measured, it is considered an anomaly and excluded during

the calibration of the GRASP model for this parameter in Chapter 4. Data from the same

date at Site 16, a kilometre away, is a closer value for true field capacity at both these sites.

Sites 20 and 21 were very difficult to access after rain and considerable drying of the

profile had occurred by the time sampling was possible. This drying has lead to lower

measures WHC’s than for other sites in this group.

Table 3.12 Physical properties of soils at sites located on cracking clays of alluvial origin.


Content (%)



0-10 cm

10-20 cm

0-10 cm

10-20 cm

20-30 cm

50-60 cm

80-90 cm

0-10 cm

10-50 cm

50-100 cm

Total

7 5.5 5.6 1.50 1.40 1.42 1.47 1.68 32.4 117.3 129.9 279.5

8 5.0 5.2 1.54 1.68 1.68 1.70 1.75 27.7 111.6 104.2 243.6

9 4.4 4.4 1.52 1.79 1.75 1.83 1.87 24.1 97.3 96.3 217.7

10 4.4 4.6 1.58 1.53 1.73 1.79 1.74 31.6 113.5 130.6 275.7

15 8.1 8.6 1.50 1.57 1.47 1.47 1.71 48.0 1 195.5 1 60.6 304.2 1

16 7.3 7.4 1.32 1.54 1.54 1.54 1.54 33.3 119.3 116.2 268.8

20 3.8 3.9 1.68 1.80 1.80 1.71 1.71 12.5 75.3 63.1 150.9

21 2.8 3.1 1.52 1.52 1.52 1.76 1.84 21.0 89.2 75.8 186.0

1 data likely to be an over-estimation due to waterlogged conditions.


71

Figure 3.18 Field-measured soil water contents of sites located on cracking clays of alluvial origin.

Solid vertical lines with data points represent measured soil water contents throughout the study period; and

shaded areas indicate derived water holding capacity based on wettest and driest recorded profiles. (Figure

continued overleaf)

Site 7

0

20

40

60

80

100


Dep

th (c

m)

24/11/1993 21/01/199417/03/1994 9/05/199420/09/1994 11/01/19957/03/1995 17/10/1995

Site 8

0

20

40

60

80

100


Dep

th (c

m)

21/11/1993 21/01/199417/03/1994 9/05/199420/09/1994 10/01/19958/03/1995 3/05/199517/10/1995

Site 9

0

20

40

60

80

100


Dep

th (c

m)

24/11/1993 19/01/199418/03/1994 10/05/199421/09/1994 11/01/19958/03/1995 4/05/199518/10/1995

Site 10

0

20

40

60

80

100


Dep

th (c

m)

25/11/1993 18/01/199418/03/1994 10/05/199421/09/1994 11/01/19958/03/1995 4/05/199517/10/1995


72

Figure 3.18 (cont.)

The dotted vertical line in the Site 15 graph represents a likely upper limit to water holding capacity although

the shaded area covers the entirety of the range of soil water contents measured during this study.

Site 15

0

20

40

60

80

100


Dep

th (c

m)

9/09/1994 13/01/19959/03/1995 2/05/199518/10/1995 10/01/199628/02/1996 30/05/199626/11/1996

Site 16

0

20

40

60

80

100


Dep

th (c

m)

9/09/1994 13/01/19959/03/1995 2/05/199518/10/1995 10/01/199628/02/1996 30/05/1996

Site 20

0

20

40

60

80

100


Dep

th (c

m)

1/09/1994 13/12/19949/02/1995 21/04/19951/11/1995 3/01/199613/02/1996 4/06/1996

Site 21

0

20

40

60

80

100


Dep

th (c

m)

1/09/1994 13/12/19949/02/1995 20/04/19951/11/1995 3/01/199613/02/1996 4/06/1996


73

3.4.2.5 Accounting for missing soil water values

Soil water values were determined by extracting three soil cores at each harvest (Section

3.3.5). Data presented in Section 3.4.2.1 to Section 3.4.2.4 reveals that the depth to which

soil was sampled varied between measurement times at some sites. This was due to both

minor variations in the depth of the soil profile across some sites, and physical difficulty in

extracting soil at certain times. Johnston (1996) and Taddese et al. (2002) both experienced

similar difficulty in extracting soil under very dry conditions. In order to present total soil

water for a consistent depth at each site throughout the entire field study, a method of

extrapolating soil moisture in shallow holes beyond the deepest measured values down to

the assumed depth of the profile has been developed.

The method for extrapolation of ‘missing’ soil data is presented below.

1. For a depth increment where two soil cores have measured values and one does not,

the average trend of change in soil moisture from the preceding depth increment for

the two measured holes is applied to the missing increment value (Equation 5).

2. For a depth increment where one soil core has a measured value and two do not, the

trend of change in soil moisture from the preceding depth increment for the

measured hole is applied to both the missing increment values (Equation 6)

3. For a depth increment where no measured values exist, the values from the

preceding increment are applied without change to the missing increment values

(Equation 7). This approach has been adopted as no indication of any trend in

moisture change is available.

Table 3.13 Example to demonstrate extrapolating soil moisture values to missing data points.

Depth X

(e.g. 80 - 90cm)

Depth Y

(e.g. 90 – 100cm)

Soil core 1 A D

Soil core 2 B E

Soil core 3 C F

A, B, C, D, E and F represent soil moisture values (gravimetric or volumetric).

When value F has not been measured, but all other values are known:

F = (C / (A + B)) * (D + E) (Equation 5)


74

When values E and F have not been measured but all other values are known:

E = (B / A) * D

F = (C / A) * D (Equation 6)

When values D, E and F have not been measured but all other values are known:

D = A

E = B

F = C (Equation 7)

3.4.3 Pasture data

Results of pasture measurements are presented within a framework of site groups based on

pasture species composition during the study period. These groups do not necessarily

reflect the original site classifications based on land systems as differences in species

composition can occur within a single land system and similarities can occur on different

land systems. Additionally, some changes in composition from initial conditions occurred

at some sites. The dominant species were defined as those that contributed, on average, at

least 40% of the total pasture biomass throughout the measurement period. For two groups

of sites the dominant species was a perennial grass. These species were barley Mitchell

grass (Astrebla pectinata) and ribbon grass (Chrysopogon fallax). A third group contains

sites where differing perennial grasses were the dominant species, and this group is

presented as ‘other perennial grass dominated pastures’. In the last two groups several

species combine as a single dominant functional group (annual grasses, and forbs).

3.4.3.1 Species dynamics during the measurement period

At individual sites, some pasture species (mainly annuals) varied in their abundance during

the measurement period. In one season of measurement some species were large

contributors to total standing dry matter and yet were barely present (or even absent) in the

other season. Conversely, species composition was quite static in most perennial grass

dominated sites. This section provides a qualitative description of changes in species

composition during the measurement period where such changes occurred. Full details of

the main species present at each site are given under the relevant results sections (Section

3.4.3.4 to Section 3.4.3.8).


75

At Mt Sanford in 1994/95, a season of above median rainfall, a greater abundance of

annual species (mainly Iseilema vaginiflorum, Flemingia pauciflora and Rhyncosia

minima) were measured on the barley Mitchell grass dominated Sites 3, 5 and 6 than

during 1993/94 when only median rainfall was received. The increased abundance of

annual species seemingly came at the expense of the perennial grasses, which grew less in

this season. Orr (1981) and Foran and Bastin (1984) noted similar trends in Astrebla spp.

pasture during favourable rainfall periods. The abundance of annual species in 1994/95 at

Sites 3, 5 and 6 suggests that under favourable conditions annual species can successfully

compete with perennials for limiting resources. Such competition occurs in the inter-

tussock spaces. Dry periods at Rosewood in early 1996 reduced the contribution of the

annual grasses Iseilema vaginiflorum and I. fragile, and forbs to total pasture biomass at

Site 18 at the end of the 1995/96 season.

On ribbon grass dominated sites at Kidman Springs, one species (Eragrostis tenellula – an

annual short grass) was quite abundant in Site 7 at the time of establishment but then

absent for the entire period of data collection. This is most likely a response to within-

season rainfall distribution favouring species that out-competed E. tenellula, and not a

response to exclosure from grazing as this species was also absent in the surrounding

grazed area. Foran and Bastin (1984) and Foran et al. (1985) also report fluctuating

contribution of annual grass species to total biomass on cracking clay soils as a result of

seasonal variation. The annual grasses Iseilema spp. and Dichanthium sericeum and forbs

Abelmoschus ficulneus and Trichodesma zeylanicum increased in abundance in some sites

(Sites 7 and 9) during the second measurement year. Species composition in Sites 8 and 10

varied little throughout the study. At Victoria River Downs the annual grass species

present at the establishment of Site 16 all increased in abundance during the study and

another annual grass (D. sericeum) became common. (Note: D. sericeum is classified as a

perennial species by many references, however it was observed to grow only as an annual

during this study. Jacobsen (1981) reports anecdotal evidence of other perennial

Dichanthium species behaving as annuals in Queensland.)

The annual or short-lived perennial grass, Enneapogon polyphyllus, along with a few

scattered forbs, initially dominated pasture in Site 11. Pasture cover was sparse early in the

second study year, probably due to the burning at the end of the first year reducing the E.


76

polyphyllus population, and reducing germination by destroying the pasture seed heads and

any seed lying on the soil surface. The forb Bonamia pannosa increased considerably in

the second year, and the perennial grass Aristida inaequiglumis appeared. Sites 12 and 14

experienced little species composition change over the study period.

At Victoria River Downs, forb species were particularly abundant during the well above

median 1994/95 growing season. Species composition at the Rosewood sites changed little

during the study period except for some increase in forbs at Site 19. Similarly, Auvergne

sites were relatively stable in terms of species composition.

Three sites dominated by annual short grass species (Sites 1, 2 and 13) displayed

considerable changes in species composition and patterns of growth in the second study

year. At Sites 1 and 2, the annual short grass Sporobolus australasicus that was co-

dominant with Brachyachne convergens during 1993/94, was almost entirely absent in

1994/95. Also in 1994/95, a few individual perennial grass plants (Aristida latifolia)

appeared and a greater abundance of forbs (Sida cleistocalyx and Salsola kali) contributed

to the total pasture biomass. At Site 13, one species (Tragus australianus – an annual

grass) was relatively abundant at the time of establishment, but was then absent for the

period of data collection. The forbs Bonamia pannosa, Heliotropium tenuifolium and

Neptunia monosperma all increased during the second growing season.

3.4.3.2 Phenological observations

At all locations in each study year, green growth was initiated during November or early

December (Table 3.14). Annual short grasses (e.g. Brachyachne convergens and

Sporobolus australasicus) and perennial grasses generally emerged within two weeks of

‘opening’ rains. These observations are consistent with the many studies of tropical grasses

showing high rates of germination associated with the onset of wet season rainfall (e.g.

Lazarides et al. 1969; Mott 1972; Mott 1978; McIvor and Gardener 1991; Graham et al.

2004) However, some mid-height annual grasses (e.g. Iseilema spp. and Dichanthium

sericeum) did not emerge until later in the growing season when considerable rain had

fallen. Watt (1978) found that D. sericeum has an intermediate stationary stage during the

germination process termed ‘hydropedesis’. This process allows partially germinated seed

to remain viable during wetting and drying cycles of the surface of cracking clay soils


77

early in the growing season. It is possible hydropedesis caused the delay in emergence of

D. sericeum observed in this study, and that other species also possess this ability.

Emergence of forbs is species dependent (Mott 1972; Graham et al. 2004) and no

generalizations are given here.

The beginning of the pasture growing season has been defined by a number of authors

(Norman 1963a; Rose et al. 1972; McCown 1973; McCown et al. 1974; McCown 1981a)

using a range of criteria all based on rainfall or a soil moisture index exceeding a certain

value for a specified time period. Based on such previous work, criteria for the beginning

of pasture growth for each season in this study are defined as the beginning of the period in

which both 50mm of rain within 14 days and 75mm of rain within 28 days occurs. Table

3.14 shows the calculated date of commencement of growth for each location in this study.

Using this approach, the initiation of growth is associated with the first period of

substantial rainfall rather than the light and isolated falls that often precede this time.

McCown (1981b) gives median dates for the beginning of the growing season across

northern Australia. While different criteria to McCown (1981b) has been used here to

determine the initiation of growth, results show that the seasons during the study period

show close agreement with the median starting dates described by McCown (1981b).

Table 3.14 Estimated dates of initiation of pasture growth for each study location, based on criteria of

receiving both 50mm of rain within 14 days and 75mm within 28 days of the initiation date.

Growing season Location

1993/94 1994/95 1995/96

Mt Sanford 1/11/1993 7/12/1994

Kidman Springs 10/11/1993 15/11/1994 3/12/1995

Victoria River Downs 29/11/1994 7/11/1995

Rosewood 18/11/1994 3/12/1995

Auvergne 22/11/1994 26/11/1995

Rainfall was generally adequate in each season at all locations to sustain vegetative growth

following germination or re-shooting after plant dormancy. However, times and duration

of flowering, and cessation of growth are influenced by individual species and climatic

conditions. These are presented on a location-by-location basis in the following

paragraphs. Pastures, especially annual species, mature soon after the cessation of rainfall.


78

Perennial grasses are able to remain green some time after rain stops due to their deep roots

accessing stored soil water. Growth can be sustained, but at a much-reduced rate as

available nutrients are generally scarce and temperature begins to decrease.

Mt Sanford

The annual grasses that dominated Sites 1 and 2 began flowering mid-season and

continued until dry conditions matured the plants. This is the typical pattern observed for

these species at most locations in this study. At Sites 3 to 6, Astrebla pectinata only

flowered modestly during the 1993/94 season and barely at all in 1994/95, probably a

response to exclosure after many years of grazing. Orr and Evenson (1991a) and Orr and

Evenson (1991b) also found reduced seed production in Astrebla pastures protected from

the effects of grazing. B. convergens followed its usual pattern (Figure 3.19) of mid-season

flowering, and the other annuals all set their seed from mid-March onwards.

Kidman Springs

Chrysopogon fallax flowered and then set and dropped seed in a matter of weeks during

the mid-season in both study years. It is assumed this mid-season flowering is a

characteristic trait of C. fallax (Figure 3.19). Most other perennial grasses flowered from

March onwards. Bulky annual grasses such as Iseilema spp. and D. sericeum tended to

grow late in the growing season and did not flower until late March or April. At Site 11

Enneapogon polyphyllus began flowering mid-way through each season and flowering

continued until dry conditions matured the pasture in April or May. The mixture of species

at Site 12 meant plants were flowering at a range of times and for differing lengths of time.

The flowering pattern of individual species at Site 12 was similar to other sites on red earth

soils. At Site 13, B. convergens began flowering mid-way through each season and

continued until dry conditions matured the plants in April. This continual flowering from

mid-season through to late-season seems to be characteristic of B. convergens. The dry

spell in January and February 1996 caused many plants at Site 13 to mature and only

young or subsequently germinating plants were able to take advantage of the rainfall that

followed. Most perennial grasses at Site 14 flowered during March and April each year

except Eriachne aristidae, which began earlier (February).



79

Most perennial grasses flowered during March and April in both study years except C.

fallax, which flowered in February. The annual grass B. convergens began flowering at a

similar time to C. fallax and continued until dry weather induced physiological maturity in

May each year. Other annuals did not flower until closer to the end of the growing periods.

Rosewood

Most perennial grasses flowered during March and April each study year except E.

polyphyllus, which began flowering earlier and continued to flower for much of the

growing season. B. convergens began flowering in February, but the other annuals were at

least a month later. During the 1995/96 growing season, the annual grasses Iseilema spp.

largely died off during a hot and dry period in March 1996, prior to late rain. The Iseilema

spp. exhibited minimal flowering and the late rain was of little benefit to these species.

However, the late rain prolonged the growth period of perennial grasses and extended the

period of seed setting.

Auvergne

Most perennial grasses flowered during March and April of 1995, except C. fallax

(February). The annual grass B. convergens flowered at a similar time to C. fallax, but

Iseilema spp. and Sorghum spp. flowered closer to the end of the growing period.

Flowering times in 1996 were nearly a month later for most species.

Figure 3.19 shows the typical observed responses of four common native pasture species to

the strongly seasonal rainfall pattern of the VRD. Climate is the overriding factor

determining the timing of plant growth. However, under a given set of conditions the

timing of germination or re-shooting, length of rapid vegetative growth, initiation and

duration of flowering, and cessation of growth at plant maturity is dependent upon the

individual species. Perennial grasses generally have a longer period of active growth and

less rapid maturation than annual grass and forb species. A descriptive account of seasonal

plant growth in the semi-arid tropics is offered by Eyles et al. (1984). While Eyles et al.

(1984) treatment of the subject is general, it none-the-less considers the stages of plant

growth for some of the major pasture species covered in this study. A description of the

growth pattern and associated nutrient and biomass dynamics of the perennial tropical tall

grass Heteropogon contortus is given by Norman (1963b).


80

Figure 3.19 The observed phases of plant growth during the study period for: a) the annual short grass

Brachyachne convergens; b) the annual mid-height grass Iseilema vaginiflorum; c) the perennial tussock

grass Astrebla pectinata; and d) the perennial tuft grass Chrysopogon fallax.

Figure layout adapted from Harrington et al. (1984).

3.4.3.3 Notes on the presentation of pasture results

The amount of standing pasture present in the field at any one time includes carryover

material from previous growing seasons, and excludes losses associated with grazing or

detachment. This amount is called ‘total standing dry matter’ (TSDM) and refers to the

oven-dry weight of above ground plant material. ‘Growth’ differs from TSDM in that it

excludes carryover material, instead referring to the total production of plant material

during the current season including that lost by grazing and detachment. The Swiftsynd

methodology prescribes exclosure from grazing and removal of carryover material prior to

field measurements, and detachment in ungrazed pastures during the growing season is

very small. Consequently, TSDM is a close approximation for growth in the results

reported in this section. As no measure of root material was made during this study, all

terms relating to plant material refer to the above ground portion only.

Wet season Rainfall ceases anytime

Rainfall begins anytime

Emergence

Vegetative growth

Flowering

Maturity

Emergence

Vegetative growth

Flowering

Maturity

Re-shooting

Vegetative growth

Flowering

Maturity

Oct Nov Dec Jan Feb Mar Apr May Jun July Aug Sep

Re-shooting

Vegetative growth Maturity

a)

b)

c)

d) Flowering


81

Figures in the following sections show TSDM and nitrogen uptake across the measurement

period. In these figures TSDM is reset to zero at the end of the first measurement year as

sites are burnt to remove carryover material (Section 3.3.6). TSDM is again reset to zero at

the end of the second measurement year to signify the end of the study period. Several sites

(Sites 1, 16, 19, 20 and 21) show zero TSDM earlier in the second study year due to no

final measurements being recorded. These are explained under the relevant sections.

3.4.3.4 Barley Mitchell grass dominated pasture

Four sites (Sites 3, 5, 6 and 18) are located on pastures dominated by barley Mitchell grass

(Astrebla pectinata - Plate 14 and Plate 15 in Appendix 1). Figure 3.20 and Figure 3.21

display some results from Sites 5, 6 and 18. While a full dataset exists for Site 3, results are

not included in these figures as the other sites adequately represented them. Data from Site

3 not shown here appears in Appendix 6. Figure 3.22 shows results from all sites in this

group. Table 3.15 lists a few of the common species found at each site. While the pasture is

termed ‘barley Mitchell grass dominant’, other pasture species are also present in some

abundance. Many of these are annual species and occupy the inter-tussock spaces. The

high perennial grass content of these sites can be seen in Figure 3.20.

Table 3.15 Main pasture and tree species present during the study period for sites dominated by barley

Mitchell grass (Astrebla pectinata).

Site Perennial grasses Annual grasses Forbs Trees Distance to closest tree

3 Aristida latifolia

Astrebla pectinata 1

Chrysopogon fallax 2

Brachyachne convergens

Iseilema vaginiflorum

Neptunia monosperma 2

Rhyncosia minima

Acacia spp.

Terminalia volucris

20m



Chrysopogon fallax

Iseilema fragile 2


Flemingia pauciflora 2

Rhyncosia minima

None present >500m

6 Aristida latifolia 2



Iseilema fragile 2


Chrysogonum

trichodesmoides


Rhyncosia minima 2

Lysiphyllum cunninghamii

Terminalia volucris

15m

18 Astrebla pectinata 1

Panicum decompositum

Chionachne hubbardiana 2

Iseilema fragile

Cyperus spp. 2

Polymeria ambigua 2

Rhyncosia minima

Trichodesma zeylanicum


Terminalia volucris

20m

1 Most dominant species throughout (>40% of total biomass). 2 Other species abundant at different times (estimated at least 20% of total biomass).


82

Figure 3.20 Pasture composition of three sites dominated by barley Mitchell grass. Error bars indicate the

standard error of the site mean for total standing dry matter at each sampling time (harvest).

Site 5

0

1000

2000

3000

4000

Jul-93 Oct-93 Jan-94 Apr-94 Jul-94 Oct-94 Jan-95 Apr-95 Jul-95 Oct-95

TSD

M (k

g/ha

)

ForbsAnnual GrassPerennial Grass

Site 6

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 18

0

1000

2000

3000

4000


TSD

M (k

g/ha

)



83


dominated by barley Mitchell grass.

The dotted lines represent expected nitrogen contents during periods where no data was collected (explained

in the text at the end of this section).

Site 5

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 6

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 18

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)


84

Table 3.16 Results for N and P status of pasture at sites dominated by barley Mitchell grass.

Site Nitrogen Phosphorus

Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

3 28.7 2.49 0.49 2.44 0.146 0.047

5 33.2 2.12 0.46 2.63 0.113 0.032

6 26.5 3.11 0.50 3.39 0.207 0.037

18 20.8 2.04 0.36 2.74 0.192 0.051

Table 3.17 Some important pasture variables at sites dominated by barley Mitchell grass.

Site Year PGBA 1

(%)

Peak TSDM 2

(kg/ha)

Growth rate 3

(kg/ha/day)

3 1993/94 4.50 2058 ± 278 9.0

1994/95 3.75 2544 ± 309 18.2

5 1993/94 3.63 3291 ± 392 12.7

1994/95 3.75 3471 ± 305 24.8

6 1993/94 1.50 2502 ± 355 12.9

1994/95 1.13 3330 ± 520 23.8

18 1994/95 2.38 3359 ± 336 22.1

1995/96 2.13 2254 ± 432 12.2

1 PGBA - perennial grass basal area. 2 TSDM- total standing dry matter (± value is the standard error of the site mean). 3 Growth Rate – average daily growth rate calculated using TSDM accumulation from the date of initiation of

growth (Table 3.14) to the end-of-growing-season measurement (Harvest 3).


sites dominated by barley Mitchell grass.

y = 43.35x - 88.43R2 = 0.71

0

1000

2000

3000

4000

0 20 40 60 80 100

Total Plant Cover (%)

TSD

M (k

g/ha

)

Site 3Site 5Site 6Site 18

y = 110.54x - 364.70R2 = 0.71

0

1000

2000

3000

4000

0 10 20 30 40

Plant Height (cm)

TSD

M (k

g/ha

)

Site 3Site 5Site 6Site 18

a. b.


85

N uptake and N content data for barley Mitchell grass dominated sites are shown in Figure

3.21 and Table 3.16. Heavy early wet season rainfall during 1993/94 at Mt Sanford

delayed data collection past the ideal time for the first harvest (6 weeks after the initiation

of growth, Table 3.14). Consequently, considerable pasture growth had occurred and the

peak nitrogen concentration period had likely already passed. The heavier dashed lines in

Figure 3.21 represent a smoothed interpolation between data points and probably do not

represent the true pattern of pasture nitrogen content during the interval between site

establishment and the first harvest. For this reason, a second ‘hand-drawn’ line of nitrogen

content (fainter dotted line) is included which is thought to better indicate the actual pattern

of N content change (after Norman 1963a). This approach has been applied, where

appropriate, to all presentations of nitrogen content results in the following sections.

Data in Table 3.16 suggest that Site 18 has a lower nitrogen supply that the other sites.

Sites 3, 5 and 6 are located about 5km from each other within the same land system at Mt

Sanford, and Site 18 is several hundred kilometres away at Rosewood. Hence, Site 18 is

expected to be less similar to Sites 3, 5 and 6 than these sites are to each other.

As inter-tussock spaces occupy a large part of the total ground surface even in ‘dense’

Mitchell grass pastures, the growth habit of the species that occupy these spaces will

heavily influence any overall yield-cover relationship. Prostrate species (like many forbs)

result in much lower TSDM values at a given plant cover than erect species (like some

annual grasses). The relationship between plant height and TSDM (Figure 3.22) appears

strong and this relationship becomes an important component of the modelling process in

Chapter 4.


86

3.4.3.5 Ribbon grass dominated pasture

Five sites (Sites 7, 8, 9, 10 and 16) are located on pastures dominated by ribbon grass

(Chrysopogon fallax - Plate 15 and Plate 16 in Appendix 1). Site 16 contains a lower

proportion of C. fallax than the other sites in this group. While a full dataset exists for Sites

7 and 16, these data have not been included in Figure 3.23 or Figure 3.24 as results from

Sites 8, 9 and 10 adequately represent all sites in this group. Data from Sites 7 and 16 not

shown here appear in Appendix 6. A paddock wildfire went through Site 16 during the late

dry season of 1996 and burnt all standing pasture. No final end-of-dry-season

measurements were possible at Site 16.

Table 3.18 Main pasture and tree species present during the study period for sites dominated by ribbon grass

(Chrysopogon fallax).


7 Chrysopogon fallax 1


Dichanthium sericeum 2

Iseilema fragile

Abelmoschus ficulneus

Euphorbia schultzii


Polymeria ambigua 2


None present >500m



Dichanthium fecundum

Dichanthium sericeum

Iseilema fragile



None present >500m


Aristida holathera


Eulalia aurea


Iseilema fragile 2

Euphorbia schultzii


Polymeria ambigua

Acacia spp.


Terminalia volucris

30m



Dichanthium fecundum



Iseilema fragile




Acacia spp.


Terminalia volucris

15m


Astrebla pectinata



Iseilema fragile

Iseilema vaginiflorum 2

Sorghum stipoideum 2

Evolvulus alsinoides 2

Ipomoea polymorpha

None present >500m

1 Most dominant species throughout (>40% of total biomass). 2 Other species abundant at different times (estimated at least 20% of total biomass).


87

Figure 3.23 Pasture composition of three sites dominated by ribbon grass. Error bars indicate the standard

error of the site mean for total standing dry matter at each sampling time (harvest).

Site 8

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 9

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 10

0

1000

2000

3000

4000


TSD

M (k

g/ha

)



88


dominated by ribbon grass.


in the text at end of Section 3.4.3.4).

Site 8

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 9

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)Site 10

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)


89

Table 3.19 Results for N and P status of pasture at sites dominated by ribbon grass.


Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

7 30.5 2.11 0.54 1.73 0.129 0.039

8 36.6 1.87 0.43 2.64 0.134 0.023

9 22.5 2.01 0.47 1.17 0.096 0.015

10 20.8 1.67 0.37 1.40 0.107 0.010

16 28.6 2.14 0.45 1.45 0.105 0.014

Table 3.20 Some important pasture variables at sites dominated by ribbon grass.

Site Year PGBA 1

(%)

Peak TSDM 2

(kg/ha)

Growth rate 3

(kg/ha/day)

7 1993/94 2.25 2595 ± 202 13.6

1994/95 3.25 3661 ± 497 21.5

8 1993/94 6.00 3426 ± 226 18.5

1994/95 4.25 4282 ± 212 25.3

9 1993/94 0.88 2008 ± 214 10.0

1994/95 2.13 2527 ± 208 14.9

10 1993/94 4.25 2666 ± 547 14.5

1994/95 3.50 2763 ± 337 16.2

16 1994/95 1.63 3093 ± 91 20.1

1995/96 1.50 2682 ± 288 13.1




90


sites dominated by ribbon grass.

Typical maximum TSDM on ribbon grass dominated sites is around 3000kg/ha (Figure

3.23). Exceptions are Site 9 (both years) and Site 8 (1994/95). Data in Table 3.19 offers

some explanation. Site 9 had a lower maximum nitrogen uptake and a lower PGBA than

other sites (Table 3.20), contributing to the generally lower maximum TSDM. The

increased contribution of forbs at Site 8 in 1994/95 resulted in the highest TSDM recorded

in this study (4282kg/ha). The legume Flemingia pauciflora was the main forb species

present and is likely to have contributed to the high total nitrogen uptake at Site 8 in the

1994/95 growing season.

The relationships between standing dry matter and plant cover (Figure 3.25) appear similar

to barley Mitchell grass sites (Section 3.4.3.4). However, the scatter of data points in the

TSDM vs. height plot (Figure 3.25) indicates some variability in plant growth habit.

Chrysopogon fallax is a tuft grass achieving high yields at a low height. The variability

exhibited in Figure 3.25 reflects the difference between sites in species composition of the

remainder of the sward (i.e. other perennial grasses, annual grasses and forbs - Table 3.18)

and their individual growth habits.

3.4.3.6 Other perennial grass dominated pasture

Six sites (Sites 11, 12, 14, 17, 19 and 20) are located on pastures dominated by perennial

grasses other than barley Mitchell grass or ribbon grass (Plate 17 and Plate 18 in Appendix

1). A mixture of perennial species makes up the majority of TSDM at Sites 12, 14, 17, 19

y = 40.92x - 331.54R2 = 0.66

0

1000

2000

3000

4000

0 20 40 60 80 100


TSD

M (k

g/ha

)

Site 7Site 8Site 9Site 10Site 16

y = 108.53x - 232.20R2 = 0.64

0

1000

2000

3000

4000

0 10 20 30 40

Plant Height (cm)

TSD

M (k

g/ha

)

Site 7Site 8Site 9Site 10Site 16

a. b.


91

and 20, while Enneapogon polyphyllus is the single dominant perennial at Site 11. While a

full dataset exists for Sites 11, 12 and 19, these have not been included in Figure 3.26 and

Figure 3.27 as results from Sites 14, 17 and 20 span the range of results from all sites in

this group. Data from Sites 11, 12 and 19 not shown here appear in Appendix 6. Cattle

broke into Site 19 at Rosewood during the late dry season of 1996 and consumed all the

pasture. Localised early storms at Auvergne dropped about 200mm of rain on Site 20

before the end of dry season measurements were due, destroying much of the standing dry

pasture. No final measurements were collected from Sites 19 or 20.

Table 3.21 Main pasture and tree species present during the study period for sites dominated by other

perennial grass species.


11 Enneapogon polyphyllus 1

Aristida inaequiglumis

None present Bonamia pannosa 2

Indigofera hirsuta 2

Carissa lanceolata

Corymbia terminalis

Eucalyptus pruinosa

15m


Dichanthium fecundum 1

Heteropogon contortus 1

Sehima nervosum 1

Brachyachne convergens 2

Dichanthium sericeum

Bonamia pannosa

Gossypium australe 2

Polymeria ambigua 2

Carissa lanceolata

Eucalyptus pruinosa

10m

14 Sehima nervosum 1

Eriachne aristidae 1

Aristida holathera

Heteropogon contortus

Brachyachne convergens Gomphrena canescens 2

Ptilotus exaltatus

Carissa lanceolata

Eucalyptus spp.

Hakea arborescens

Terminalia canescens

10m

17 Dichanthium fecundum 1

Aristida latifolia 1



Abutilon otocarpum

Heliotropium tenuifolium 2


Corymbia terminalis

Terminalia arostrata

20m


Enneapogon polyphyllus 1

Aristida inaequiglumis


Sporobolus australasicus

Bonamia pannosa

Sida cleistocalyx 2

Corymbia terminalis

Eucalyptus brevifolia

Eucalyptus pruinosa

2 trees within

the site


Panicum decompositum 1

Dichanthium fecundum 1


Sorghum spp. 2

Neptunia monosperma

Sesbania cannabina 2

Atalaya hemiglauca

Eucalyptus spp.


Terminalia volucris

10m

1 Most dominant species throughout. 2 Other species abundant at different times.


92

Figure 3.26 Pasture composition of three sites dominated by other perennial grass species. Error bars

indicate the standard error of the site mean for total standing dry matter at each sampling time (harvest).

Site 14

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 17

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 20

0

1000

2000

3000

4000


TSD

M (k

g/ha

)



93


dominated by other perennial grass species.



Site 14

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 17

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 20

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)


94

Table 3.22 Results for N and P status of pasture at sites dominated by other perennial grass species.


Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

11 17.7 1.73 0.36 1.53 0.084 0.023

12 21.2 1.43 0.54 1.53 0.078 0.028

14 13.0 1.99 0.45 0.68 0.053 0.012

17 15.3 2.00 0.40 1.53 0.160 0.029

19 9.6 2.00 0.46 1.13 0.105 0.046

20 18.3 2.15 0.32 1.17 0.135 0.020

Table 3.23 Some important pasture variables at sites dominated by other perennial grass species.

Site Year PGBA 1

(%)

Peak TSDM 2

(kg/ha)

Growth rate 3

(kg/ha/day)

11 1993/94 2.88 2707 ± 420 14.7

1994/95 2.75 2365 ± 297 13.9

12 1993/94 1.25 2418 ± 366 12.8

1994/95 2.13 3032 ± 477 17.9

14 1994/95 2.88 2396 ± 378 14.2

1995/96 2.13 1581 ± 346 8.6

17 1994/95 3.00 2315 ± 497 15.1

1995/96 1.13 1762 ± 311 9.5

19 1994/95 1.88 1873 ± 677 12.3

1995/96 0.63 736 ± 125 4.0

20 1994/95 0.88 2621 ± 260 17.5

1995/96 2.00 3973 ± 448 20.8




95


sites dominated by other perennial grass species.

Pasture yield in 1994/95 exceeded that in 1995/96 at sites shown in Figure 3.26; however,

at Site 20 this trend is reversed. Examining the associated rainfall data at Kidman Springs

(Site 14), Rosewood (Site 17) and Auvergne (Site 20) - Figure 3.7, Figure 3.11 and Figure

3.13 respectively – reveals that a prolonged dry spell occurred in early 1996 at Kidman

Springs and Rosewood, but was less pronounced at Auvergne. This dry spell brought on

physiological maturity of pasture early and reduced the maximum TSDM (Section 3.4.3.2)

at Sites 14 and 17. Growth was less interrupted at Auvergne and pasture here was able to

take advantage of the high rainfall that occurred late in the 1995/96 growing season.

Species composition varies considerably between sites in this group (Table 3.21), however

the relationship between TSDM and plant height is quite consistent (Figure 3.28). This

may be attributed to most sites containing some mid-height grass species (perennial and/or

annual) and this has resulted in similar TSDM to height relationships across all sites in this

group.

y = 39.88x - 39.05R2 = 0.71

0

1000

2000

3000

4000

0 20 40 60 80 100


TSD

M (k

g/ha

)Site 11Site 12Site 14Site 17Site 19Site 20

y = 110.19x - 596.25R2 = 0.87

0

1000

2000

3000

4000

0 10 20 30 40

Plant Height (cm)

TSD

M (k

g/ha

)

Site 11Site 12Site 14Site 17Site 19Site 20

a. b.


96

3.4.3.7 Annual short grass dominated pasture

Three sites (Sites 1, 2 and 13) are located on pastures dominated by annual short grass

species (mainly Brachyachne convergens and Sporobolus australasicus - Plate 18 and

Plate 19 in Appendix 1). Each of these sites had a history of preferential grazing for many

years prior to site establishment. The sites in this group represent a species change from

high perennial grass content that occurs on lightly grazed areas, rather than land types

naturally dominated by annual grasses.

Cattle broke into Site 1 during the late dry season of 1995 and consumed all pasture. No

final end-of-dry-season measurement was possible at Site 1.

Table 3.24 Main pasture and tree species present during the study period for sites dominated by annual short

grass species.


1&2 Aristida latifolia



Sporobolus australasicus 2

Abutilon otocarpum

Indigofera trita

Salsola kali

Sida cleistocalyx 2

Atalaya hemiglauca

Corymbia terminalis

Eucalyptus brevifolia

Grevillea striata

30m

13 Aristida holathera

Aristida latifolia



Dactyloctenium radulans

Bonamia pannosa

Heliotropium tenuifolium 2

Neptunia monosperma

Corymbia terminalis

Eucalyptus pruinosa

20m



97

Figure 3.29 Pasture composition of sites dominated by annual short grass species. Error bars indicate the

standard error of the site mean for total standing dry matter at each sampling time (harvest).

Site 1

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 2

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 13

0

1000

2000

3000

4000


TSD

M (k

g/ha

)



98


annual short grass species.



Site 1

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 2

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)Site 13

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0N

Con

tent

(%)


99

Table 3.25 Results for N and P status of pasture at sites dominated by annual short grass species.


Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

1 25.5 2.50 0.87 1.78 0.129 0.042

2 23.3 3.13 0.79 1.54 0.125 0.042

13 17.9 2.54 0.69 1.39 0.086 0.024

Table 3.26 Some important pasture variables at sites dominated by annual short grass species.

Site Year AGBA 1

(%)

Peak TSDM 2

(kg/ha)

Growth rate 3

(kg/ha/day)

1 1993/94 8.38 2165 ± 205 10.5

1994/95 1.63 2952 ± 399 21.1

2 1993/94 6.00 1977 ± 239 9.9

1994/95 0.38 2424 ± 263 17.2

13 1994/95 5.75 2593 ± 195 15.2

1995/96 0.38 854 ± 100 4.8

1 AGBA – annual grass basal area. No measurable perennial grasses were present at these sites. 2 TSDM- total standing dry matter (± value is the standard error of the site mean). 3 Growth Rate – average daily growth rate calculated using TSDM accumulation from the date of initiation of



sites dominated by annual short grass species.

Peak pasture TSDM in the annual grass dominated sites (Figure 3.29) appears to be

generally lower than for the perennial grass dominated sites (Section 3.4.3.4 to Section

y = 24.13x + 128.58R2 = 0.62

0

1000

2000

3000

4000

0 20 40 60 80 100


TSD

M (k

g/ha

)

Site 1Site 2Site 13

y = 122.36x - 252.21R2 = 0.82

0

1000

2000

3000

4000

0 10 20 30 40

Plant Height (cm)

TSD

M (k

g/ha

)

Site 1Site 2Site 13

a. b.


100

3.4.3.6). The major feature of Figure 3.29 is the very low TSDM at Site 13 throughout

1995/96 (maximum TSDM = 854kg/ha). This is likely a result of the combination of two

events. Standard methodological practice (Day and Philp 1997) dictates removal of

standing biomass at the end of the end of the first measurement year. Fire was used to

remove this carryover material. In undisturbed annual short grass pastures, such as at Sites

1, 2 and 13, the seed bank providing next season’s growth is mostly held in the seed heads

of standing pasture. The fire most likely destroyed much of this seed bank, thus reducing

germination and plant density in the second year (Plate 23, Plate 24 and Plate 25 in

Appendix 1). Mott and Andrew (1985) and McIvor et al. (1994) also recognise the

vulnerability of annual grasses to seed bank destruction by fire. In the case of Site 13, the

already-reduced plant density was further hindered by the dry conditions of early 1996

(Figure 3.7). Thus, the combination of these two events resulted in very low pasture

growth at Site 13 in the 1995/96 growing season.


101

3.4.3.8 Forb dominated pasture

Three sites (Sites 4, 15 and 21) are located on pastures dominated by forb species (Plate 20

and Plate 21 in Appendix 1). Like the annual grass dominated sites, each of these sites

appear to have been preferentially grazed by cattle prior to site establishment. While forbs

are a normal component of native pastures in the region, they are rarely the major

component in unmodified pasture communities. Thus, the sites in this group represent a

species change from the high perennial grass content that occurs on lightly grazed areas,

rather than land types naturally dominated by forb species.

Localised early storms at Auvergne dropped about 200mm of rain on Site 21 before the

end of dry season measurements were due. Much of the dry pasture was destroyed, and

new green growth had occurred by the time the site was accessible. No final measurements

were collected as biomass reflected both new seasons growth and an accelerated

detachment and breakdown rate of the previous season’s growth.

Table 3.27 Main pasture and tree species present during the study period for sites dominated by forb species.



Astrebla elymoides



Iseilema fragile


Abutilon otocarpum 1

Basilicum polystachyon

Rhyncosia australis 1


Terminalia arostrata 50m

15 Astrebla pectinata 2

Chrysopogon fallax

Aristida latifolia

Astrebla squarrosa 2



Iseilema fragile


Evolvulus nummalaris



Rhyncosia minima 1

Sesbania cannabina

None present >500m


Aristida latifolia

Brachyachne convergens 2 Euphorbia schultzii 1


Polymeria ambigua 1

Terminalia volucris


50m



102

Figure 3.32 Pasture composition of sites dominated by forb species. Error bars indicate the standard error of

the site mean for total standing dry matter at each sampling time (harvest).

Site 4

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 15

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 21

0

1000

2000

3000

4000


TSD

M (k

g/ha

)



103


forb species.



The high values in May 1995 at Site 15 are discussed in the text at the end of this section.

Site 4

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 15

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 21

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)


104

Table 3.28 Results for N and P status of pasture at sites dominated by forb species.


Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

Max. uptake (kg/ha)

Max. content (%)

Min. content (%)

4 36.4 2.35 0.85 3.32 0.157 0.049

15 56.4 2.67 0.72 2.64 0.132 0.035

21 19.3 2.55 0.51 0.87 0.174 0.015

Table 3.29 Some important pasture variables at sites dominated by forb species.

Site Year PGBA 1

(%)

Peak TSDM 2

(kg/ha)

Growth rate 3

(kg/ha/day)

4 1993/94 1.00 2343 ± 161 12.5

1994/95 1.00 2950 ± 295 18.9

15 1994/95 0.75 2758 ± 284 17.5

1995/96 1.50 2680 4 13.1

21 1994/95 3.25 2031 ± 172 13.6

1995/96 1.63 1588 ± 111 8.3


growth (Table 3.14) to the end-of-growing-season measurement (Harvest 3). 4 No standard error can be calculated for this measurement.


sites dominated by forb species.

y = 25.81x + 261.44R2 = 0.67

0

1000

2000

3000

4000

0 20 40 60 80 100


TSD

M (k

g/ha

)

Site 4Site 15Site 21

y = 126.07x - 251.44R2 = 0.67

0

1000

2000

3000

4000

0 10 20 30 40

Plant Height (cm)

TSD

M (k

g/ha

)

Site 4Site 15Site 21

a. b.


105

In Figure 3.33, the May 1995 data for Site 15 is shown as recorded. The nitrogen uptake

value (column) is 56.4kg/ha, far higher (some 20kg/ha higher) than at any other time at this

site or any other site. Such an unusual value requires some investigation to determine

whether it is realistic or contains some error in measurement. Over 2000kg/ha of forbs

were present at this harvest (Plate 21 in Appendix 1), and the sub-sample used to determine

N content was only a small portion of this standing dry matter. While every care was taken

to make this sub-sample as representative as possible, the diversity of forb species present

made this a difficult task. It is conceivable that the sub-sample analysed may have

contained a higher proportion of high N content plant material than was truly

representative. When multiplied by the large dry matter yield, this potential error may have

contributed to the unusually high value shown. Data is reported in this section as collected,

but due to the atypical nature of this value, it is considered an anomaly and omitted when

deriving GRASP model parameters for this site in Chapter 4.

3.5 Discussion of field study results

3.5.1 Climate

Rainfall

Rainfall followed patterns considered typical of the climate in the VRD (e.g. Slatyer 1970).

The first falls were recorded in October or November each year, usually the result of

thunderstorms. These storms are often localised and rainfall received is highly variable and

infrequent. As the growing season progresses, more substantial and sustained falls occur

during December to March resulting from increased storm frequency and waves of

monsoon activity. Daily falls sometimes exceeded 75mm and occasionally more than

100mm was recorded. Rainfall totals were generally close to median values at all study

sites, with some locations receiving well above median rainfall in some seasons (1993/94

at Kidman Springs; 1994/95 at Mt Sanford, Victoria River Downs and Rosewood; and

1995/96 at Victoria River Downs). Short dry periods occurred during the growing season,

with one such period during late January and early February 1996 being the most

significant. Rainfall tended to cease abruptly in March or April, although several isolated

falls after this time were recorded.

One of the most noticeable features of the climate graphs in Section 3.4.1 is the general

gradient in median rainfall, with highest values nearest the coast (Auvergne, Figure 3.14)

Chapter 3 A field study of pasture growth in the VRD ------------------------------------------------------------------------------------------------------------------------------------------ and declining as distance from the coast increases to the furthest inland (Mt Sanford,

Figure 3.6). Variability also increases as median rainfall declines (Table 3.30). This

gradient impacts upon the timing of initiation, duration and cessation of rainfall at

each location (McCown 1981b). This gradient was less pronounced during the

seasons measured in this study as most locations experienced similar rainfall patterns. Table 3.30 Variability in rainfall at the five study locations in the VRD. Rainfall data from DataDrill (2005).

Temperature

The temperature graphs reveal little difference between study locations in either

maximum or minimum temperatures (both the 47-year averages and the values

reported during the study period) throughout the wet season months (November -

May). The dry season months see greater diurnal fluctuations and lower absolute

values for daily maximums and minimums at the more inland locations. Temperatures

during the period of measurement are considered to be close to optimal growth and

the cooler conditions during June to August each year are of little consequence as

pastures are dormant at this time of the year.

Other climate variables

Examination of radiation, evaporation and vapour pressure data (Table 3.4) shows

values during the study period being close to long-term averages. Radiation levels in

the Victoria River District are generally high and are thought not to restrict plant

growth except for short periods during the wet season when heavy monsoonal cloud

cover may persist for several days. These times are associated with the periods of

heavy rainfall as evident in Figure 3.5 to Figure 3.13. Annual potential (Class A pan)

evaporation rates are very high in the VRD; about four times the annual rainfall at

Victoria River Downs. Rainfall exceeds pan evaporation only during a few months of

the year. Thus, the climate in this region is termed semi-arid. Vapour pressure values

were close to long term averages except for

_____________________________________________________________________

106

NOTE: This table is included on page 106 in the print copy of the thesis held in the University of Adelaide Library.


107

some variation between years in the May to August period (Table 3.4). Plants are dormant

at this time of year and variation in vapour pressure has no impact on plant growth.

3.5.2 Soils

Soil bulk density

Soil bulk density (BD) reported by Day (1977) on red earths are close to those measured

on comparable sites in this study (Table 3.6 and Table 3.8), ranging from 1.4 - 1.5g/cm3 at

the surface to 1.5 - 1.8g/cm3 at depth. Lodge et al. (2003b) found BD values of 1.4g/cm3 at

the surface and 1.6 to 1.7g/cm3 below 40cm depth in a red Chromosol soil in northern

NSW. Bulk density values for cracking clay sites in this study are similar to or higher than

published data. Dalgliesh and Foale (1998) provide BD values for black vertosols of 1.0 -

1.2g/cm3 at the surface and 1.2 - 1.4g/cm3 at 1.0m.; and grey vertosols have BD’s in the

order of 1.4 – 1.6g/cm3 and changing little throughout the profile. The cracking clay soils

(vertosols) in this study have measured bulk densities of about 1.3g/cm3 and 1.5g/cm3 near

the surface, and about 1.65g/cm3 and 1.7g/cm3 at depth for basalt and alluvial clays

respectively (Table 3.31). These denser soil profiles contribute to the higher water holding

capacities of soils in this study than in the published data. Berndt and Coughlan (1976)

note some issues with measuring bulk density of cracking clay soils using small-diameter

core-sampler such as used in this study (Plate 8 in Appendix 1), as bulk density varies with

water content due to shrinkage and swelling of the soil being three-dimensional (Talsma

and van der Lelij 1976; Smiles 1997).

Soil water content

Slatyer (1954) estimated plant available water holding capacity (WHC) of 85mm/m on the

red limestone soils (Tippera) surrounding Katherine. This value is much less than the

values found for similar soils as part of this study. Day (1977) reported red earth soils with

WHC’s of 77 to 135mm/m in the Daly Basin, north-east of the VRD. The WHC’s of red

earths measured in this study (mostly 120 to 150mm/m, Table 3.6 and Table 3.8, averaging

139mm/m, Table 3.31) are at the upper end of the values reported by Day 1977.

The highest values of WHC for this study were found on the cracking clays overlying

basalt (Table 3.10 and Table 3.31). Mott et al. (1985) give some average values for WHC

in self-mulching cracking clays of 220mm/m, values similar to those reported by McKeon


108

et al. (1990). Dalgliesh and Foale (1998) list water holding capacities of 150 to 250mm/m

and Hochman et al. (2001) report WHC’s of 124mm/m for black vertosols (similar to

basalt clays), compared to 215 to 315mm/m measured here. Dalgliesh and Foale (1998)

also quote WHC’s of 130 to 200mm/m and Hochman et al. (2001) report WHC’s of

101mm/m on grey vertosols (similar to alluvial clays), compared with the measured values

in this study of 150 to 280mm/m (Table 3.12). Taddese et al. (2002) measured water

holding capacity of 140mm/m in a vertosol soil in Ethiopia.

Determining WHC from field measurements is difficult (Dalgliesh and Foale 1998),

particularly for the cracking clays (Greacen and Williams 1983; Isbell 1983; Dalgliesh et

al. 1998). Isbell (1983) comments that “application of arbitrary upper and lower limits

based on water contents derived from the moisture characteristic are of limited value in

describing the water store of a cracking clay”. The nature of cracking clay soils is such

that once the soil profile is full, drainage is slow and true drained upper limit is difficult to

assess. The highest soil water contents encountered during this study (cracking clays

overlying basalt, Figure 3.17) were all measured some 7 days after the last rain, and yet

conditions were sometimes still boggy and sites difficult to access. It is conceivable that

some of the high soil water values assumed to be ‘field capacity’ may in fact still be

saturated. A more accurate estimate of field capacity would require detailed measurement

of soil water content and soil water potential as the soil dries from ‘saturated’ to ‘field

capacity’. Given the broad scale and remote nature of the current study, it was not possible

to make these more detailed measurements. The soil data presented is the most detailed

available for VRD soils.

Wilting point in cracking clays can also be problematic to ascertain. Their cracking nature

results in soil water loss by evaporation directly from the deep cracks, lowering the soil

water content in the soil surrounding the cracks to below true wilting point. Thus, sampling

results can vary from hole to hole at the same site, depending on whether the sample is

collected from the middle of a large ped, or closer to the ped edge.

The red earth soils are easier to measure for WHC due to their higher position in the

landscape (better drainage) and lower clay content making them easier to traverse.

However, the isolated location of many of the sites meant access to them required crossing


109

watercourses and cracking clay soils and, therefore, sampling was sometimes not possible

at the wettest times of the year.

Another potential contributing factor to the higher WHC’s reported here is plant root

density. Baker and Ahern (1989) emphasise that “estimates of rooting depth are necessary

parameters in predicting available water capacity of soils”. Isbell (1983) and Hochman et

al. (2001) discuss the impact of plant root patterns on wilting point values when

determining WHC, suggesting that fibrous-rooted and tap-rooted plants will have different

wilting points in the same soil. Similarly, annual crops will have different wilting points to

perennial plants. One feature of the soil water charts presented in this chapter is the relative

uniformity of soil water availability down the soil profile. Established perennial grass

plants on undisturbed soils are likely to have greater root density deep in the soil profile

than the annually cropped soils that are the source of some published WHC data. Greater

root density results in a higher level of water extraction and therefore lowers the effective

wilting point. Low wilting point values contribute to the higher total WHC’s presented

here in comparison to those reported in the literature.

Table 3.31 Summary of results for important soil variables.

Soil bulk density (g/cm3) Water Holding Capacity

Soil group Assumed pasture rooting depth

0-10cm 50-60cm 80-90cm Total (mm)

Average (mm/m)

Red earths – basalt 1.0m 1.50 1.56 1.56 145.7 145.7

Red earths – limestone 0.6m 1.58 1.65 - 83.4 138.9

Cracking clays – basalt 0.9m 1.30 1.55 1.64 277.3 308.1

Cracking clays - alluvial 1.0m 1.52 1.66 1.73 231.7 231.7

Soil chemistry and nutrients

Soil chemistry and nutrient analyses were restricted to the top 20cm of soil, as this zone

contains the most organic matter (Spain et al. 1983) and greatest concentrations of the

major elements (Christie 1979). Organic carbon content of the soils in this study ranged

from 12752kg/ha to 26771kg/ha in the top 20cm of soil (Table 3.32) (0.5% – 1.1%, Table

9.5 to Table 9.8), values within the range given by Spain et al. (1983) for these soil types.

Total soil nitrogen contents were 650kg/ha to 1222kg/ha in the top 20cm (0.02% – 0.04%),

values much lower than those of Spain et al. (1983) and Mott et al. (1985), but similar to


110

the results of Christie (1979). Nitrate nitrogen levels were extremely low, ranging from

3.1kg/ha to 4.3kg/ha (1.1 – 1.4 ppm) in the top 20cm. Total phosphorus levels were also

very low at 303kg/ha to 679kg/ha (0.010% – 0.023%), similar to the ‘infertile sandy red

earth’ studied by Christie (1979).

Chemical properties for red earths, and to a lesser extent cracking clays, have been

reported for nearby regions. In general, reported values for base exchange capacity and

total and exchangeable N, P, S and K that are similar or less than those recorded in Table

9.5 to Table 9.8. The red earths in the Daly Basin (Day 1977) were slightly acid (pH 5.7 –

6.7) as were those at Katherine (Williams et al. 1985), while similar soils in this study are

neutral to slightly alkaline (pH 7.2 – 8.4). Williams et al. (1985) report a cracking clay soil

as slightly alkaline (pH 8.0), consistent with results here (Table 9.7 and Table 9.8).

Table 3.32 Summary of soil chemistry results for each of the major soil types in this study.

Soil type pHwater OC 1 NO3-N 2 N(t) 3 P(bicarb) 4 P(t) 5 C:N ratio 6

Red earths – basalt 7.5 18603 3.4 1081 18.5 679 17.2

Red earths – limestone 7.7 26771 4.3 1222 32.0 605 21.9

Cracking clays – basalt 8.1 12752 3.1 650 22.0 305 19.6

Cracking clays - alluvial 8.1 15643 4.3 756 19.2 303 20.7

1 OC - organic carbon; 2 NO3-N – nitrate nitrogen; 3 N(t) - total nitrogen; 4 P(bicarb) – bicarbonate

extractable (Colwell); 5 P(t) - total phosphorus; 6 C:N ratio – carbon to nitrogen ratio. All units are kg/ha in

the top 20cm of soil, except the pH and C:N ratio, which are dimensionless.

Summary

Results from this study indicate that soils in the VRD have generally higher plant available

water capacities than those reported for comparable soils in other regions. Cracking clay

soils of basalt origin have higher water holding capacities than soils of alluvial origin. The

shallow red earths have the lowest WHC’s. The studied soils contain similar levels of

many nutrients as those in the literature, except for the cracking clay soils, which contained

lower amounts of total nitrogen than published data. As with most soils in Australia’s

semi-arid tropics, natural fertility levels are very low, particularly deficient for N and P.

These low levels of N and P have been identified as a major constraint to pasture growth in

the VRD (Mott et al. 1985).


111

Overall, the soils measured in this study possess physical and chemical properties that are

(with the exception of the high field capacities) essentially similar to comparable soils in

other regions of northern Australia. Much of the data presented in this chapter and in

Appendix 3 are the first of their kind to be published for the VRD.

3.5.3 Pastures

Observations of species dynamics

At some sites, species composition and phenology varied throughout the study period.

Such variations were probably due to either exclosure from grazing, prevailing seasonal

conditions, response to fire after burning to remove carryover material, seasonal variations

in nutrient supply, inter-plant competition, or some combination of these factors. While a

general understanding exists of how these factors affect pasture communities, the

individual species-specific responses are not well understood.

Many studies have examined the biology and ecology of individual native grass and forb

species in northern Australia (e.g. Norman 1963b; Mott 1978; Jacobsen 1981; Hacker

1984; Andrew 1986; Cook and Andrew 1991; Orr 1991; Stockwell et al. 1994). These

studies have provided a good understanding of many species and an overall general

appreciation for the factors influencing pasture species dynamics. However, observations

during this study noted a variety of responses, some of which are inconsistent with these

published works. Some gaps in knowledge were also revealed. Suggested further research

to improve our understanding of the pasture system includes: nitrogen fixation by common

legumes such as Flemingia pauciflora, Rhyncosia minima and Neptunia monosperma; the

germination and emergence of Dichanthium sericeum, Iseilema vaginiflorum and similar

mid height grasses under field conditions in the VRD; and the seed bank dynamics of

annual short grasses such as Brachyachne convergens and Sporobolus australasicus and

how they are affected by fire.

Total standing dry matter

Maximum total standing dry matter (TSDM) across all sites and years varied from 736 to

4282kg/ha. Throughout this study the highest mean maximum TSDM values were

measured on sites dominated by perennial grasses, followed by forb sites, with annual

grass dominated pastures having the lowest (Table 3.33). TSDM’s of pastures in this study


112

are generally consistent with those reported by Foran et al. (1985) (800 to 2100kg/ha),

McIvor et al. (1994) (1490 to 2240kg/ha), MacDonald et al. (1997) (1530 to 2560kg/ha)

and Bastin et al. (2003) (1700 to 3000kg/ha), albeit at the upper end of the ranges they

provide (Table 2.2). This is mainly attributable to the favourable seasonal conditions that

prevailed during the study period, the high water holding capacity of the soils, and the

absence of trees and grazing on the sites measured.

In the annual short grass dominated sites, burning the undisturbed pasture to remove

carryover material at the end of the first study year (Sites 1 and 2 in 1994, and Site 13 in

1995) is suspected of destroying much of the seed bank needed for the following season.

Germination was sparse and few individual plants grew during the second study year at

these sites (Table 3.26). While end of growing season TSDM in 1995 was not reduced at

Sites 1 and 2, this was a season of above median rainfall. Site 13, however, shows

drastically reduced growth in the second study year (Figure 3.29), a season of

approximately median rainfall during which periods of hot and dry weather affected plant

growth. These observations highlight the vulnerability of pastures containing high

proportions of annual species to fire and less than favourable seasonal conditions.

Table 3.33 Summary of field results for important pasture variables.

Species PGBA 1 (%)

Max. TSDM 2 (kg/ha)

Max. N uptake (kg/ha)

Max. N content

(%)

Min. N content

(%)

Barley Mitchell grass 2.85 2851±141 24 1.96 0.48

Ribbon grass 2.96 2970±117 24 1.74 0.50

Other perennial grasses 1.96 2315±135 15 1.66 0.50

Annual short grasses 0.02 2161±136 19 2.27 0.88

Forbs 1.52 2392±111 30 2.37 0.77

1 PGBA - perennial grass basal area.

2 TSDM- total standing dry matter (± value is the standard error of the group mean).

Measuring TSDM introduces the issue of measurement variance. Variance occurs because

of spatial heterogeneity in native pasture communities; that is, pasture density differs

considerably over short distances. Section 3.3.5 details the methodology used in this study

for collecting TSDM data, prescribing eight individual measurements that are averaged to

provide a mean value of TSDM for each site at that harvest. Results from the field study


113

show considerable spatial variation in TSDM at many sites throughout the growth cycle.

This variation is illustrated by the relatively large standard error bars in Figure 3.20, Figure

3.23, Figure 3.26, Figure 3.29, and Figure 3.32; and in the individual TSDM results in

Table 9.12 (Appendix 7).

Table 3.34 summarises measurement variance for TSDM in this study, represented by 95%

confidence limits about the harvest mean (Goldman and Weinberg 1985). Early wet season

measurements display greater variance that at any other time. Overall, measurement

variance was ±35.4% of the TSDM harvest mean, indicating that many pasture yield

results contain sizeable statistical error. This level of measurement variance in the field

data becomes important when evaluating modelling results in Chapter 4 and Chapter 5.

Table 3.34 Variance in field measurements of total standing dry matter (TSDM) for 21 sites over two years.

Values represent 95% confidence limits expressed as a proportion of the harvest mean.

n Minimum 95% confidence limit value for 21

sites over two years

Maximum 95% confidence limit value for 21


Mean 95% confidence limit value for 21


Early wet (H1 and H5) 42 ±16.7% ±162.2% ±44.8%

Mid wet (H2 and H6) 42 ±4.5% ±89.4% ±30.6%

End of wet (H3 and H7) 42 ±7.0% ±85.5% ±30.4%

End of dry (H4 and H8) 37 ±13.5% ±102.4% ±35.8%

All data 163 ±35.4%

Data summarised from Table 9.12. H1 is harvest 1, and so on.

Perennial grass basal area

Perennial grass basal area (PGBA) measured during this study varied from 0.0 - 6.0% of

the soil surface, with most values between 2 - 3%. Mott et al. (1985) report a typical value

of around 4% for Australian savannas with the inter-tussock spaces supporting a wide

variety of ephemeral grasses and forbs. Observations of basal area made during this study

are consistent with Mott et al. (1985), and a number of other studies in northern Australia.

A study on native pastures on northern Western Australia by Holm and Eliot (1980)

recorded PGBA on cracking clays of 2.96%; and another study on the Barkly Tableland by

Foran and Bastin (1984) also found PGBA values similar to those in this study (1.0 –


114

4.0%). Foran et al. (1985) report total basal cover (annuals plus perennials) of 0.06 –

3.39% in pastures on red earths at Kidman Springs.

McIvor et al. (1995a) report a positive relationship between PGBA and TSDM on tropical

tall-grass pastures near Katherine. Results from this study show a similar trend, although

the relationship is poor (y = 207x + 2143, r2 = 0.171, Figure 3.35c). Not surprisingly, as

PGBA increases so does perennial grass production (y = 457x + 512, r2 = 0.534, Figure

3.35a), and sites dominated by perennial grasses have highest TSDM (Table 3.33). Foran

and Bastin (1984) report similar findings on Mitchell grass pastures in the Barkly

Tableland. Conversely, as PGBA increases the production of annual grasses and forbs

declines, but the relationship is less pronounced (y = -250x + 1630, r2 = 0.248, Figure

3.35b). TSDM of annual grasses and forbs is more likely related to factors other than

PGBA, such as seed bank density, nutrient availability and evapotranspiration.

Reduction in PGBA by grazing results in species composition change, with perennial

grasses declining and annual grasses and forbs increasing in abundance (Gardener and

McIvor 1985). Observations in this study suggest that soil type is influential on how

species change is exhibited. Low PGBA on red earth soils results in dominance of the

sward by annual short grasses. However, on cracking clays, forbs are the main contributor

to total biomass when perennial grasses are sparse.


115

Figure 3.35 Relationships between perennial grass basal area (PGBA) and end of growing season standing

dry matter (SDM) of a) perennial grasses; b) annual grasses and forbs; and c) total pasture.

Nutrients

End of growing season total plant nitrogen uptake varied from 9.6kg/ha to 56.4kg/ha, and

nitrogen content varied from 0.32% to 3.13%. The highest total N uptake and N content

values recorded in this study occur on forb dominated sites and annual grass dominated

sites respectively (Table 3.28 and Table 3.25). Friedel (1981) reports nitrogen levels in

combined herbage and litter of 30 to 72kg/ha, while Ash and McIvor (1995) report 5 to

21kg/ha of nitrogen uptake in tropical tallgrass pastures depending on pasture condition.

They also report an increase in plant N concentration as the proportion of annual grasses

and forbs in a pasture sward increase. Schmidt et al. (1998) also found that annual species

have a high capacity for nitrogen assimilation. Many forbs are legumes and supplement

soil-available N by fixing nitrogen from the atmosphere, thus increasing the total N supply

available for growth (Peoples and Baldock 2001). No data is available on the contribution

y = 207x + 2143R2 = 0.171

0

1000

2000

3000

4000

5000

0 1 2 3 4 5 6

PGBA (%)

Tota

l pas

ture

SD

M (k

g/ha

) Total SDMn = 42

c.

y = -250x + 1630R2 = 0.248

0

1000

2000

3000

4000

5000

0 1 2 3 4 5 6

PGBA (%)

Ann

ual g

rass

es a

nd fo

rbs

SDM

(kg/

ha)

Annual Grasses andForbs SDM

n = 42

b.y = 457x + 512

R2 = 0.534

0

1000

2000

3000

4000

5000

0 1 2 3 4 5 6

PGBA (%)

Pere

nnia

l gra

ss S

DM

(kg/

ha)

PerennialGrass SDM

n = 42

a.


116

of native legumes to plant N supply in the VRD, but results of this study suggest they

contribute to the total supply of N available for plant growth.

Results of nitrogen uptake during the growing season show a rapid increase in total plant

nitrogen yield during the early to mid wet season, then little further increase for the

remainder of the growing season (e.g. Figure 3.21). This pattern is consistent with Christie

(1981) who reports nitrogen uptake rates declining rapidly after 8 continuous weeks of

growth in a Mitchell grassland in Queensland.

The lowest values for minimum N content occur in perennial grass pastures (Table 3.33).

This implies that perennial grasses are able to continue growth after soil N supply is

exhausted by internally diluting nitrogen to a lower concentration than can annual grasses

and forbs. In an environment where soil nutrients are often a major limit to pasture growth

(Section 2.3.2), internal N dilution is influential on end of growing season TSDM. Christie

(1981) reports higher conversion efficiency of nitrogen into pasture biomass by a perennial

C4 grass than for other species under the same conditions, supporting the results of this

study. Hendricksen et al. (1992) and Holm and Eliot (1980) both report declining plant N

concentration as the growing season progresses, findings also consistent with results of this

study.

Summary of field study

• seasonal rainfall totals (July - June) recorded in the study varied from 538mm to

1062mm, being influenced by location and by year. All sites received rainfall totals

near to, or well above, long-term medians (34th to 97th percentiles);

• plant available water capacity of soil profiles varies from 123mm/m to 350mm/m and

is a function of soil texture, soil depth, drainage and plant rooting pattern. These values

are similar to, or greater than, other published data from comparable soil types;

• soil fertility, as reflected in the soil nutrient and chemical properties, indicates these

soils to be typical for the northern Australian rangelands. Available nitrogen and

phosphorus are generally extremely low for much of the year;

• Pasture species composition at any individual location can vary from year to year

depending upon grazing pressure, prevailing seasonal conditions, fire, available


117

nutrient supply, and competition with other species. Some of these responses are not

well understood;

• maximum pasture TSDM varies from 736kg/ha to 4282kg/ha and is influenced by

location, year, soil type, plant density, species composition, and nutrient supply and

dilution;

• TSDM measured using the Swiftsynd methodology results in variance of ±35% of the

mean, and this becomes important during the modelling process in the following

chapters;

• perennial grass based pastures generally have highest TSDM (average of 2712kg/ha),

and annual species (grasses and forbs) are more prone to variations in abundance due to

seasonal conditions and/or management (e.g. burning); and

• Nitrogen supply increases with content of annual grasses and forbs, but this is offset by

a lower efficiency at converting N into plant biomass.

3.6 Conclusions

The field study results presented in this chapter provide comprehensive detail for a range

of land types on the climatic conditions, soil properties and resulting pasture responses

during the study period. Rainfall and other climate variables during the study period were

generally favourable for pasture growth and, apart from a dry period during one growing

season, no effect of drought or other extreme weather was experienced.

Soil water holding capacity is dependent on soil type, and WHC’s of soils in this study

were generally high compared to other published data from comparable soils. This

difference is attributed to high soil bulk density and the increased soil water extraction of

established perennial grasses compared with the annually-cropped soils that are the source

of much of the published water holding capacity data. Levels of soil nitrogen and

phosphorus are naturally very low.

Pasture biomass production is similar to, or higher than other studies in and around the

VRD. Two main factors contribute to this result:

• pastures in this study were measured without the effect of grazing or the presence of

trees, both which reduce pasture biomass accumulation; and


118

• the favourable growing conditions resulted in a considerable contribution of annual

grasses and forbs to total pasture production, even on perennial-dominated pasture

types.

A number of limitations in the accuracy of the field data have been revealed:

• determining soil water holding capacity of cracking clay soils from field measurements

is difficult, and the isolated nature of the study sites has exacerbated this problem;

• the methodology employed to measure pasture biomass has resulted in sizeable

measurement error due to the spatial heterogeneity of pasture types in the VRD; and

• plant nitrogen uptake is also subject to some error due to difficulty in obtaining

representative subsamples in cases where pastures contain high biomass with many

species.

It is concluded that, with some noted exceptions, the dataset produced as a results of this

field study is suitable for developing local values for input parameters of the GRASP

pasture growth model. In addition, the dataset represents the most detailed description of

soil water capacity, pasture growth and N uptake yet assembled for the VRD.

Chapter 4 Analysis of field measurements using a systems modelling approach

119

4.0 Analysis of field measurements using a systems modelling approach

4.1 Introduction

Analysis of real systems using simulation modelling involves a number of phases. Grant et

al. (1997) identify four phases of systems analysis:

• conceptual model formulation (abstracting from the real system to develop a

conceptual, qualitative model of the system of interest);

• quantitative model specification (development of the mathematical equations and their

parameter values that collectively form the quantitative model);

• model evaluation (the process of determining the relative usefulness of a model for

meeting specific objectives); and

• model use (answering the questions that were identified at the beginning of the

modelling project).

The four phases of systems analysis are highly interconnected and model development may

cycle through several phases more than once. The GRASP pasture growth model has

already passed through the model evaluation and use phases numerous times and is

therefore considerably robust in structure for the environments in which it has been

developed and applied. However, GRASP has not yet been evaluated under the unique

conditions of the VRD.

This chapter aims to calibrate the GRASP model on a site-by-site and year-by-year basis

using the field data reported in Chapter 3, and assess the performance of the model when

predicting key soil and pasture variables. To achieve this aim, Section 4.2 first provides an

overview of the GRASP model to provide the reader with an understanding of the model

structure and the need to derive values for a number of important parameters. While the

actual calibration process partially depends on characteristics of the site where field data

was collected, it is an essentially systematic procedure, and the general approach is

described in Section 4.3. The individual site and season parameter values resulting from

calibration are presented in Section 4.4. GRASP’s ability to mimic the pasture system at

the field study sites is assessed by comparison of model output with field data for several

key variables (i.e. standing dry matter, soil water content, nutrient uptake and plant cover)

the results of which are presented in Section 4.5. The capability of the model and its


120

limitations are discussed in Section 4.6 and conclusions presented in Section 4.7. Figure

4.1 illustrates the structure of this chapter and Figure 4.2 shows the phases of systems

analysis and model development, and how the components of this study relate to each

phase.


Figure 4.2 The four phases of systems analysis (bolded text in box) and an indication of where the

components of this study fit within this framework (after Grant et al. 1997).

Conceptual model formulation

Quantitative model specification

Model evaluation

Model use

Collection of local field data – Chapter 3

Calibration of model from field results – Chapter 4

Sensitivity testing of model to determine factors most influential on pasture growth – Chapter 6 Application of model to management of grazing land in the VRD – Chapter 7

Validation of model with independent data to test its capability – Chapter 5

Overview of the calibration procedure (Section 4.3)

Derive GRASP parameters that describe each individual site (Section 4.4)

Overview of the GRASP model (Section 4.2)

Assessing the fit between model results and field data

(Section 4.5)



121

4.2 An overview of the GRASP pasture growth model

The GRASP model is documented by McKeon et al. (1990) and in more detail by

Littleboy and McKeon (1997). A simplified version including the rationale of many

components is provided by Rickert et al. (2000). The core of the model, ‘Subroutine

GRASP’, has remained largely unchanged since the early 1990’s and is the focus of this

overview. Many developmental versions based upon Subroutine GRASP now exist.

‘Spaghetti GRASP’ is Dr Greg McKeon’s most advanced form of GRASP, and a

personalised copy, ‘GVT89MQ3’ is used in this study (G.M. McKeon pers.comm.).

The objective of the GRASP model is to simulate the soil water balance, pasture growth

and other important biophysical processes by using data from short-term field observations

and extrapolating over many years. Outputs from the model can then be used in decision-

making processes by researchers, policy makers and grazing land managers. Day et al.

(1993) state that a fundamental step in providing recommendations for pasture

management is to assess likely spatial and temporal variability in pasture production. The

limited availability of pasture productivity data in most cases means assessments of such

variability can only be made by extrapolating from long-term climatic data in association

with estimates of site productivity.

In GRASP, pasture growth is related to transpired water by a model parameter termed

‘transpiration-use-efficiency’ (TUE). Transpiration is a component of the formal water

balance calculated by GRASP (Figure 2.6). Pasture growth is assumed to be directly

proportional to the rate of transpiration (Arkley 1963; Rose et al. 1972; Tanner and

Sinclair 1983; Walker and Richards 1985). Calculating transpiration-use-efficiency

requires knowledge of plant growth and soil water. This is the basis for data collection in

the field study reported in Chapter 3.

4.2.1 Soil water balance

The soil water balance is simulated using three soil layers (0 - 10cm, 10 - 50cm, 50cm to

pasture rooting depth (assumed maximum depth is 100cm). Studies of pasture in northern

Australia report low root densities in the soil profile below one metre (McCown 1971;

Christie 1978; Christie 1981). A fourth layer (grass rooting depth to tree rooting depth) can

be simulated but is not incorporated in this study as the focus is on pasture rather than


122

trees. The processes of runoff, infiltration, drainage, soil evaporation and transpiration are

calculated separately for each day from inputs of rainfall and pan evaporation. Surface

cover, rainfall intensity and soil moisture are part of the runoff calculation (McIvor et al.

1995b; Scanlan et al. 1996b; Owens et al. 2003). Runoff relationships apply to mid-slope

locations. Little quantitative data exists for deep drainage in northern Australian

rangelands, but Harrington et al. (1984) and Christie (1981) suggest that evaporation and

transpiration are the main pathways of rainfall, and drainage is low in semi-arid rangelands

(except under prolonged saturation, Talsma and van der Lelij 1976). Lodge et al. (2003a)

also found that evapotranspiration was the largest output term in the soil water balance of a

native pasture in New South Wales, and that runoff and drainage were episodic events.

The water balance equation is given by Rickert et al. (2000):

SWt = SWt-1 + Raint + Irrigt - Draint - Runofft - Esoilt - Transt – Ecanopyt - Elittert

(Equation 8)

where SWt and SWt-1 are the soil water contents for t and t-1 days respectively, and Raint,

Irrigt, Draint, Runofft, Esoilt, Transt, Ecanopyt and Elittert are daily flows of rainfall,

irrigation, deep drainage, runoff, soil evaporation, transpiration, canopy interception and

litter interception respectively. All components must be in the same units (i.e. millimetres).

In this study, irrigation and canopy interception are of little relevance (no sites were

irrigated and only one site contained trees). Rickert et al. (2000) state that locations with a

high proportion of rainfall events of less than 25mm may experience substantial

evaporative losses by water adhering to surface litter. In tropical northern Australia, a high

proportion of rainfall received comes from heavy storms (e.g. 44% of total seasonal

rainfall at Victoria River Downs Station comes in falls exceeding 25mm - DataDrill 2005),

thus minimising the influence of evaporation from litter on the soil water balance. With

these factors having minimal influence in this study, the main factors contributing to

variation in the soil water balance are rainfall, runoff, evaporation and transpiration.

Drainage will be important in very wet periods, but no data is available for the region.

Calculating the soil water balance:

• recorded daily rainfall is partitioned into runoff and infiltration;


123

runoff is estimated using relationship derived from Scanlan et al. (1996a), and has

recently been improved using a ‘curve number runoff method’ (Owens et al. 2003).

Runoff is affected by cover (high cover = low runoff), soil water deficit (dry soil =

low runoff), and rainfall intensity (high intensity = high runoff). Slope is a factor in

the runoff calculation but most simulations assume a gentle gradient;

water not run off is infiltrated into the surface layer;

• when trees are present, tree basal area is used to determine transpiration by trees from a

set relationship (Calder 1992). At a tree basal area of 20m2/ha transpiration from trees

will equal pan evaporation;

• pasture transpiration is calculated from green plant cover, pan evaporation and vapour

pressure deficit. It has priority of water use over soil evaporation;

• soil evaporation occurs in the 0 - 50cm (mostly 0 - 10cm) layers, and is inversely

related to surface cover. Evaporation from the soil surface decreases as surface water

contents decrease (Eastham et al. 1992). Consideration is given to the cracking soils

where some moisture loss directly from the 50 - 100cm layer occurs through

evaporation;

• infiltration from one soil layer to the next occurs when soil water content exceeds the

field capacity of that layer; and

• deep drainage is estimated from the amount of soil water above field capacity in the

deepest soil layer for pasture growth (i.e. 50 - 100cm).

4.2.2 Pasture growth, death, detachment and breakdown

Once the soil water balance (Equation 8) has been quantified and transpiration by pasture

has been calculated from green cover and available soil water, the processes of pasture

growth, death and detachment can be quantified (McKeon et al. 1982b). In ideal conditions

pastures grow at a potential rate determined by their physiology. Under variable

conditions, growth is constrained by supply of soil water, ambient temperature, solar

radiation, and nutrient supply (Rickert et al. 2000).

Pasture growth is determined for water limiting and radiation limiting conditions, with the

most limiting factor determining growth (Littleboy and McKeon 1997). Under water

limiting conditions, growth is calculated from transpiration and transpiration-use-

efficiency (dry matter accumulation per unit of transpiration). Under radiation limiting


124

conditions, growth is calculated from intercepted radiation and radiation-use-efficiency

(dry matter accumulation per unit of radiation). Growth occurs within constraints of

nitrogen availability and temperature. The nitrogen availability function (Figure 4.3)

assumes some N is stored internally (perennial roots or seeds), and uptake of N per unit of

transpiration is constant until transpiration ceases due to soil water limitations or a user-

defined maximum amount of available N has been reached. Calculation of regrowth from

stored reserves is required when little or no green material is present to transpire water or

intercept radiation (i.e. after plant dormancy or defoliation by grazing or fire).

Figure 4.3 Relationship between accumulated pasture transpiration and nitrogen uptake (from Littleboy and

McKeon 1997). The user-defined coefficients for stored plant N reserves, N uptake per 100mm of

transpiration, and maximum N available for uptake shown in this figure are realistic for the semi-arid tropics,

but are examples only.

Dry matter is simulated in three biomass pools: green, dead, and litter (detached dead); and

the processes of growth, death, and detachment transfer biomass from one pool to another

(Figure 2.6 and Figure 4.4). Biomass in the green pool is partitioned into leaf and stem

pools via a user-defined ratio. Breakdown of litter returns dry matter to the soil. Soil

organic matter dynamics are not simulated by GRASP.

0

10

20

30

40

50

0 100 200 300 400 500 600 700 800 900 1000

Accumulated transpiration (mm)

Nitr

ogen

upt

ake

(kg/

ha)

Stored plant N reserves = 5kg/haN uptake per 100mm transpiration = 5kg/haMaximum N available for uptake = 30 kg/ha


125

Figure 4.4 Flow of dry matter through the biomass pools (bolded) in the pasture system (modified from

Littleboy and McKeon 1997). Processes that transfer dry matter from one biomass pool to another are

presented in italics.

Calculating pasture growth:

• initial plant regrowth is calculated as a function of the plant growth index, plant

density, and potential regrowth rate;

• pasture transpiration (from the soil water balance) and growth (derived from field

observations) are proportional. The relationship between them is quantified, providing

transpiration-use-efficiency (TUE);

• green cover directly affects transpiration, and transpiration is proportional to growth,

therefore the relationship between green cover and growth (green yield) becomes a key

part of the model. The green yield at 50% green cover determines the shape of this

function;

• as green cover increases, growth under water limiting conditions is calculated from

transpiration and transpiration-use-efficiency. Radiation and temperature may constrain

growth under certain conditions, but they are not a major limiting factors in the VRD

(Section 2.3.1);

• atmospheric vapour pressure deficit (VPD) influences transpiration and changes with

plant height. The relationship between plant height and green yield adjusts transpiration

to daily VPD. Plant height at 1000kg/ha defines the height-yield relationship.

• growth is partitioned into leaf and stem material;

• minimum plant nitrogen content and maximum nitrogen uptake, both derived from

plant analysis, provide an upper limit to accumulated plant yield under conditions

where soil water does not limit plant growth; and

• grazing reduces green yield and green cover, and therefore subsequent transpiration

and growth.

Plant reserves

(seed or perennial base)

+ soil N

Green pool

death Dead pool

detachment Litter pool

breakdown

Animal intake

growth Soil organic matter


126

Plant death can occur in three ways: 1) a background death rate is applied, being directly

related to green yield; 2) an inverse relationship between soil water supply and plant death

is applied (plant death increases once soil water supply falls below a critical threshold);

and 3) frost may cause plant death, but this is not relevant to the VRD (Section 2.3.1).

Detachment of standing dead pasture due to wind and rain is calculated from a background

rate that is derived from field measurements. Mechanical detachment from disturbance by

grazing animals is also calculated, but pastures in this study were exclosed from grazing,

eliminating this form of detachment.

Breakdown of detached litter is determined from temperature and rainfall data, being

accelerated in warm, wet conditions (Robbins et al. 1989; Cooksley et al. 1993).

4.2.3 Assumptions and limitations of the model

A number of assumptions about the soil-plant-climate system are included in the GRASP

model. These assumptions are:

• for a given pasture type, there is no differentiation between species. All pasture is

treated as a green pool, dead pool, and litter pool;

• rainfall is the only source of soil water. There is no net run-on. Irrigation can be

incorporated but it is not applicable in this study;

• no use of water by grasses beyond a soil depth of 100cm, although the model can be

parameterised to access soil water below this depth;

• below-ground biomass accumulation (root growth) is not simulated, nor is seed

production;

• tree growth and senescence are not included;

• nutrients other than nitrogen are non-limiting;

• no reduction in nitrogen uptake occurs as a result of green material consumption by

grazing animals; and

• no impact on pasture growth by pests and disease.


127

4.3 Method for deriving model parameters and calibrating GRASP

The first major step in applying an empirically derived model such as GRASP to a new

situation is to calibrate the relationships it is built upon with local data (Figure 4.2). Rykiel

(1996) defines model calibration: “calibration is, in essence, the step of making a model as

consistent as possible with the data set from which the parameters are estimated”.

Calibration is often referred to as ‘model fitting’ as the process involves a certain amount

of parameter manipulation to produce the best fit between model output and the field

measured data. The calibration procedure in this study aims to provide sets of input

parameter values that best represent measured data for each site in each study year. This is

a variation from Day et al. (1997a) where a single set of parameters was established to

represent all years at a site. This approach was taken in order to maximise the number of

specific calibrated parameters available when deriving more generalised parameter sets

later in this thesis.

An optimisation technique such as least squares regression is often recommended for

calibration of model parameter values (e.g. Jones and Carberry 1994), but is not adopted

here. Optimisation has limitations as an appropriate technique for deriving parameters for

GRASP, due in part to the large number of parameters requiring simultaneous calibration.

Formal optimisation techniques may result in unrealistic parameter values even though

they produce acceptable results. This is expressed well by Carter et al. (1996b) who

suggest that it is possible at times to simulate the correct answer with incorrect parameters.

No formal algorithm yet exists to objectively calibrate GRASP from a measured dataset.

Consequently, the calibration procedure is manual and somewhat subjective. While the

procedure is essentially systematic, it does involve some judgement on:

• interpretation of the field results when making estimates of parameter values;

• specific parameters to target for individual site calibration;

• when ‘acceptable’ agreement between measured and modelled data is achieved; and

• deciding if individual field data points are atypical and excluding them from the

calibration procedure.

A large number of parameters in GRASP (100+) could potentially be calibrated. All

parameters have default values derived from field studies or literature. In practice, far


128

fewer than 100 parameters are actually calibrated for a given site. In this study 35

parameters were calibrated for each site, with the balance remaining at default values.

Individual calibrations of 21 sites over two years were undertaken, and in each case the

actual process was partially influenced by site characteristics. Hence, the exact calibration

procedure varied from site to site, but followed a general approach. An overview of this

approach to calibrating specific groups of parameters is provided to demonstrate how the

data presented in Chapter 3 was used to adapt the GRASP model to the grazing lands of

the VRD.

4.3.1 Input files

GRASP requires climate data from which to calculate the soil water balance and plant

growth, death and detachment processes. Spanning the two-year period of study at each

site, a climate file containing daily records of rainfall (measured at each of the study sites

or the nearest station homestead), pan evaporation, maximum and minimum temperature,

radiation, and vapour pressure (all from DataDrill 2005) defines the climatic environment

within which simulations will be made.

Model parameter values are specified in two separate input files. First, a file containing

initial (default) values for all parameters required for running GRASP has been developed

from previous studies. These parameters were established from the GUNSYNpD project

(McKeon et al. 1988; McKeon and Johnston 1990) and have been continually revised and

updated based upon subsequent studies and accumulating knowledge. Most parameters are

generally held at their default values. The default parameter file used in this study is shown

in Appendix 10.

A second file contains a number of more site-specific parameters that are either taken

directly, or derived, from field observations collected using the methodology of Day and

Philp (1997). As many parameters as possible are derived from the field data. These site-

specific parameter values overwrite the equivalent parameters in the default parameter file.

An example of a site-specific parameter file is shown in Appendix 10.


129

The following section details those parameters taken directly from field data, and those

parameters that used field data as an initial estimate and were then refined to provide a

better fit between the field dataset and model outputs. At first mention, individual model

parameters frequently referred to in the text are described in full with parameter names in

italics, followed by an abbreviation and a parameter reference number (e.g. p999).

Abbreviations are used thereafter.

4.3.2 Deriving individual parameter values

Soil water parameters

GRASP calculates the soil water balance for three layers: 1 (0 - 10cm), 2 (10 - 50cm) and 3

(50+cm). The depth of layers 1 and 2 (L1, p020 and L2, p021) are fixed, while the depth of

layer 3 (L3, p022) was determined from field data (pasture rooting depth e.g. Table 3.9).

Field capacity (FC) and Wilting point (WP) for each layer (p028 to p031) were also

sourced from field data presented in Section 3.4.2. Data collected on WP was considered

reliable but field measurements occasionally did not provide true FC (discussed in Section

3.5.1). Where model predictions of soil water did not coincide with field measurements,

adjustment to FC values was sometimes necessary to improve model performance. Runoff

of surface water (Runoff, p270) was allowed on red earths but not on cracking clays.

Calibrated values of soil parameters for all sites are presented in Table 4.1.

Parameters for soil layers 1 and 2 influence soil water in lower layers but parameters for

lower layers do not influence soil water in the layers above them; i.e. water only flows

downwards in GRASP. Cracking clays are often problematic to calibrate as moisture can

enter and leave the lower soil layers through the cracks without passing through the layers

above. Evaporation from cracks (Cracking, p035) is accounted for in GRASP, however

simultaneous wetting of the whole profile is not. Other soil types in the VRD do not crack.

Sward structure parameters

For much of the growing season, plant growth is related to transpiration. The relationships

between cover and yield, and height and yield, both influence transpiration and vary

considerably from pasture type to pasture type (Day et al. 1997a). Quantification of these

relationships for each study site requires specific values for Green plant yield at 50% green

cover (GY50GC, p045), and Plant height at 1000kg/ha (Ht1000kg, p096). Where field


130

measurements did not provide these values directly, they were estimated from regressions

of the data (e.g. Figure 3.22), and refined to optimise model performance. Calibrated

values of sward structure parameters for all sites are presented in Table 4.2.

Plant growth parameters

The Growth Index (Section 2.4) is the underlying principle governing plant growth in

GRASP and in the VRD soil water supply is the dominant component. However, other

factors are influential at different times during the growth cycle. Plant growth early in the

growing season depends upon plant density and regrowth from stored reserves. Perennial

grass basal area (PGBA, p005) was a direct field measurement of plant density (e.g. Table

3.17) and Potential regrowth rate per unit of PGBA (PotRegrow, p006) was estimated

from plant yield at the first harvest and the number of days elapsed since the initiation of

plant growth (Table 3.14).

After initial regrowth from stored reserves, the main phase of pasture growth is primarily

transpiration-dependent. Therefore, a key growth parameter in GRASP is Transpiration-

use-efficiency (TUE, p007). Preliminary estimates of TUE were derived from field data.

Estimates of evapotranspiration were calculated from rainfall received and changes in soil

water between one harvest and the next. Initial estimates of transpiration were set at 0.6

times evapotranspiration (K.A. Day pers. comm. based on extensive experience of

calibrating the model in Queensland, and supported by data in Murphy et al. 2004). Pasture

growth was estimated from changes in plant yield from one harvest to the next, with the

period between first and second harvests (H1 to H2 and H5 to H6, e.g. Figure 3.20 and

Table 9.12) being most important as growth is unlikely to be influenced by critical soil

water or nutrient limitations at this stage. Starting estimates of TUE were calculated from

pasture growth and transpiration values, and then refined during the calibration process to

provide better simulation results.

As the growing season concludes, soil water limitations reduce growth and senesce the

pasture. Values for the parameter Soil water index at which growth stops (SWIX0Grow,

p149) were held at default values unless model output showed poor correspondence with

field measurements. Soil water field data collected at the end of the growing season

provided a guide to refining these parameters for better model performance. Calibrated

values of plant growth parameters for all sites are presented in Table 4.3.


131

Nitrogen parameters

Nutrient supply (mainly nitrogen) is an important factor in pasture growth in the VRD

(Section 2.3.2). A number of parameters in GRASP specify the role of nitrogen, and values

for these come directly from field measurements. Examples of field data that provide these

parameter values are seen in Figure 3.21 and Table 3.16.

An internally stored supply of N initiates plant growth before root density is sufficient to

capture available N in the soil. The parameter N uptake at zero transpiration (Nup0Trans,

p097) represents this stored N supply and, along with Maximum N content in plant

material (Max%N, p100), was calculated from plant N analysis results at the first harvest

(e.g. Table 3.16).

Plant nitrogen status changes rapidly during the early growing season, and field

measurements did not always capture data that truly represents these parameters. This is

apparent in Figure 3.21 where the first harvest at Site 5 occurred in early February 1994,

some three months after growth began. Data recorded at this harvest did not provide

suitable values for Max%N (likely to have already peaked and declined), nor Nup0Trans

(uptake of N from soil already well under way). Consequently, values for these parameters

required refinement during the calibration procedure to provide better model predictions of

early season growth.

Other critical nitrogen parameters are also derived from plant analysis. At the end of the

growing season, values for Maximum N uptake (MaxN, p099), and N content at which

growth stops (%N0Grow, p101) and Minimum N content in dead plant material (Min%N,

p111) come directly from field data (e.g. Figure 3.24). Values for the parameter N uptake

per 100mm of transpiration (Nup100T, p098) is derived from field measurements of N

uptake throughout the growing season along with preliminary estimates of transpiration.

For field data to provide accurate values for these critical N parameters, plant growth

should not have ceased before they are fully expressed: i.e. drought conditions can senesce

plants before all available N is taken up by plants, or before they have internally diluted N

to their physiological limit. Seasonal conditions during the study period were generally

favourable and it is therefore assumed that field measurements provide suitable values for


132

the nitrogen parameters. Calibrated values of nitrogen parameters for all sites are presented

in Table 4.4.

Detachment parameters

Dead plant material detaches due to the action of wind and rain. The main period of

detachment is over the dry season. Differences in total plant yield between the end of the

growing season harvest, and end of the dry season harvest provide Detachment rate

parameter values (Detachment, p128 to p131). Calibrated values of detachment parameters

for all sites are presented in Table 4.6.

Tree parameters

Trees compete with pasture for soil water and nutrients. In this study, only one site

contained trees (Site 19). Tree basal area (TBA, p291) was estimated in the field, and Tree

wilting points (TWP) for the three soil layers (p292, p293 and p294) were set at the same

values as for the pasture. Table 4.5 contains tree basal area and tree wilting point values for

Site 19.

4.3.3 Summary

The first stage in calibrating GRASP for native pastures in the VRD is to derive a set of

initial parameter values based on local field data. Many parameter values are direct field

measurements (e.g. WP, PGBA, Max%N, Min%N and Detachment); others can be derived

from a series of field measurements (e.g. Ht1000kg, and GY50GC); and some parameters

(e.g. TUE) can only be calibrated by running the model with the best initial estimates of all

parameters, then refining their value to achieve more satisfactory simulation results.

The previous sections show how parameters that describe growth in GRASP have

influence at differing times throughout the growth cycle. Figure 4.5 illustrates typical daily

biomass accumulation simulated by GRASP for a native pasture in the VRD, and includes

parameters that are important at critical times.


133

Figure 4.5 Typical daily biomass accumulation curve for native pasture in the Victoria River District,

including approximate times of field measurements and the parameters in GRASP that are calibrated from

data collected at these times.

In a complex environment it is expected that discrepancies will occur between field

measured data and simulation model outputs. These discrepancies arise from deficiencies

in the model structure that do not account for the influence of some factors on site

processes, and measurement errors or omissions in data collection that set limits on the

accurate calibration of the model. In this study, the abbreviated ‘Swiftsynd’ methodology

(Day and Philp 1997) prescribes three measurement times during the growing season and

while every effort was made to coincide these measurements with critical times in the

pasture growth cycle, this was not always possible. Interpretation and judgement on those

situations where initial parameter values may need refinement requires an in-depth

understanding of the field data and the circumstances under which it was collected.

Reliable calibration is dependent on appropriate use and interpretation of field data, and

this is a major advantage of integrating the field experimental research with simulation

modelling, as undertaken in this study.

4.4 Results of calibration: Final model parameters

The parameter values established during the calibration process are specific to the

individual sites during the season of measurement. Many values will be the same or very

similar between years at the same site, but for some parameters (e.g. PotRegrow, TUE, and

0

1000

2000

3000

4000

Oct Jan Apr Jul Oct

Tota

l Sta

ndin

g D

ry M

atte

r (kg

/ha) Cumulative pasture biomass

Field measurements

Initial conditions described - soil water - total soil depth

First harvest - early wet season - max %N in plant - N uptake zero trans

Second harvest - mid wet season - soil water field capacity - grass basal area - cover vs yield relationship - height vs yield relationship

Third harvest - end of wet season - SWIX when growth stops - %N when growth stops - potential N uptake Fourth harvest - end of dry

season - soil water wilting point - leaf and stem detachment - min %N in plant


134

GY50GC) values can vary at the same site in different growing seasons. The final

calibrated parameter values for each site and each year are shown in Table 4.1 to Table 4.6

and these form the basis of all simulation studies in this thesis.

In a few instances during the calibration procedure, model predictions did not correspond

at all with field observations and special measures were required to improve the

simulations. These measures took two forms: 1) early growing season predictions of

TSDM were grossly under-predicted at Sites 9, 15 and 21 and the relationship between

green cover and green yield was reduced during this time to improve results (Table 4.2);

and 2) soil water values during December 1994 at Site 18 were much higher than predicted

and it was deduced that 100mm of surface run-on and/or unrecorded rainfall had occurred,

and the model was parameterised to account for this. Possible reasons for these difficulties

are discussed later in this chapter.

Finally, it is important to note that the final parameter calibration process is not completely

separate for each year at a site. Soil water parameter values (Table 4.1) are determined

from field measurements in both years and kept constant for all simulations at each site. In

a similar manner, some pasture growth parameters (i.e. TUE, PotRegrow, SWIX0Grow,

Table 4.3) and all nitrogen parameters (Table 4.4) are kept constant for both years at a site

unless conditions were so different between seasons that a change in value was necessary

to improve simulation results.


135

Table 4.1 Site-by-year calibrated GRASP soil parameters. (Table continued overleaf)

Parameter Pmtr No. S01Y11 S01Y2 S02Y1 S02Y2 S13Y1 S13Y2 S11Y1 S11Y2 S19Y1 S19Y2 S14Y1 S14Y2 S20Y1 S20Y2 Soil type REB2 REB REB REB REL REL REL REL REB REB REL REL ACC ACC Pasture type ASG3 ASG ASG ASG ASG ASG OPG OPG OPG OPG OPG OPG OPG OPG

Depth of layer 1 (mm) 020 100 100 100 100 100 100 100 100 100 100 100 100 100 100

Depth of layer 2 (mm) 021 400 400 400 400 400 400 400 400 400 400 400 400 400 400

Depth of layer 3 (mm) 022 500 500 500 500 0 0 100 100 500 500 100 100 500 500

Total depth of soil profile (mm) 1000 1000 1000 1000 500 500 600 600 1000 1000 600 600 1000 1000

Air dry layer 1 (mm) 019 3.5 3.5 3.5 3.5 3 3 3 3 4 4 2.5 2.5 5 5

Wilting point layer 1 (mm) 029 10 10 7 7 8 8 8 8 8 8 7 7 10 10

Field capacity layer 1 (mm) 026 35 35 35 35 35 35 33 33 35 35 25 25 35 35



Wilting point layer 3 (mm) 031 45 45 35 35 1 1 4.5 4.5 40 40 17 17 65 65


Cracking (Yes/No) 035 No No No No No No No No No No No No Yes Yes

Runoff (Yes/No) 270 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No

WHC (fc-wp) layer 1 (mm) 25 25 28 28 27 27 25 25 27 27 18 18 25 25

WHC (fc-wp) layer 2 (mm) 86 86 80 80 98 98 46 46 88 88 90 90 85 85

WHC (fc-wp) layer 3 (mm) 85 85 80 80 1 1 11.5 11.5 80 80 18 18 85 85

WHC (fc-wp) layers 1-3 (mm) 196 196 188 188 126 126 82.5 82.5 195 195 126 126 195 195

WHC (fc-wp) average (mm/10cm) 19.6 19.6 18.8 18.8 25.2 25.2 13.8 13.8 19.5 19.5 21 21 19.5 19.5

1 S01Y1 is Site 1, Year 1; S01Y2 is Site 1, Year 2; and remaining headings follow this logic. Sites do not appear in numerical order across the table; instead, the order of

presentation is closely associated with soil type and species composition presented later in this thesis (Table 5.1 in Chapter 5). 2 REB is red earth overlying basalt; REL is red earth overlying limestone; CCB is cracking clay overlying basalt; and ACC is alluvial cracking clay. 3 ASG is annual short grasses; OPG is other perennial grasses; RG is ribbon grass; BMG is barley Mitchell grass; and F is forbs.


136

Table 4.1 (cont.) (Table continued overleaf)

Parameter Pmtr No. S12Y1 S12Y2 S08Y1 S08Y2 S10Y1 S10Y2 S16Y1 S16Y2 S07Y1 S07Y2 S09Y1 S09Y2 S21Y1 S21Y2 Soil type REL REL ACC ACC ACC ACC ACC ACC ACC ACC ACC ACC ACC ACC Pasture type OPG OPG RG RG RG RG RG RG RG RG RG RG F F

Depth of layer 1 (mm) 020 100 100 100 100 100 100 100 100 100 100 100 100 100 100

Depth of layer 2 (mm) 021 400 400 400 400 400 400 400 400 400 400 400 400 400 400

Depth of layer 3 (mm) 022 100 100 500 500 500 500 500 500 500 500 500 500 500 500


Air dry layer 1 (mm) 019 4 4 5 5 4 4 5 5 5 5 5 5 2.5 2.5

Wilting point layer 1 (mm) 029 7 7 12 12 10 10 10 10 12 12 12 12 7.5 7.5






Cracking (Yes/No) 035 No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Runoff (Yes/No) 270 Yes Yes No No No No No No No No No No No No

WHC (fc-wp) layer 1 (mm) 23 23 33 33 35 35 40 40 33 33 33 33 32.5 32.5

WHC (fc-wp) layer 2 (mm) 60 60 120 120 115 115 110 110 115 115 110 110 95 95

WHC (fc-wp) layer 3 (mm) 17 17 120 120 140 140 130 130 110 110 120 120 100 100

WHC (fc-wp) layers 1-3 (mm) 100 100 273 273 290 290 280 280 258 258 263 263 227.5 227.5

WHC (fc-wp) average (mm/10cm) 16.7 16.7 27.3 27.3 29 29 28 28 25.8 25.8 26.3 26.3 22.8 22.8


137

Table 4.1 (cont.)

Parameter Pmtr No. S04Y1 S04Y2 S15Y1 S15Y2 S05Y1 S05Y2 S03Y1 S03Y2 S18Y1 S18Y2 S06Y1 S06Y2 S17Y1 S17Y2 Soil type CCB CCB ACC ACC CCB CCB CCB CCB CCB CCB CCB CCB CCB CCB Pasture type F F F F BMG BMG BMG BMG BMG BMG BMG BMG OPG OPG

Depth of layer 1 (mm) 020 100 100 100 100 100 100 100 100 100 100 100 100 100 100

Depth of layer 2 (mm) 021 400 400 400 400 400 400 400 400 400 400 400 400 400 400

Depth of layer 3 (mm) 022 400 400 500 500 500 500 500 500 300 300 200 200 400 400


Air dry layer 1 (mm) 019 5 5 4 4 5 5 5 5 5 5 4 4 4 4







Cracking (Yes/No) 035 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Runoff (Yes/No) 270 No No No No No No No No No No No No No No

WHC (fc-wp) layer 1 (mm) 35 35 35 35 40 40 35 35 40 40 35 35 40 40

WHC (fc-wp) layer 2 (mm) 120 120 120 120 160 160 140 140 140 140 140 140 120 120

WHC (fc-wp) layer 3 (mm) 120 120 120 120 150 150 150 150 90 90 70 70 120 120

WHC (fc-wp) layers 1-3 (mm) 275 275 275 275 350 350 325 325 270 270 245 245 280 280

WHC (fc-wp) average (mm/10cm) 30.6 30.6 27.5 27.5 35 35 32.5 32.5 33.8 33.8 35 35 31.1 31.1


138

Table 4.2 Site-by-year calibrated GRASP sward structure parameters. (Table continued below)

Parameter Pmtr No. S01Y1 S01Y2 S02Y1 S02Y2 S13Y1 S13Y2 S11Y1 S11Y2 S19Y1 S19Y2 S14Y1 S14Y2 S20Y1 S20Y2 Soil type REB REB REB REB REL REL REL REL REB REB REL REL ACC ACC Pasture type ASG ASG ASG ASG ASG ASG OPG OPG OPG OPG OPG OPG OPG OPG Green standing dry matter at 50% green

cover for calculation of transpiration,

radiation interception and runoff (kg/ha)

045, 046 and

271 700 1500 500 1600 800 1500 1800 1650 1450 2050 1900 2550 1050 1600

Height of 1000kg/ha TSDM (cm) 096 10 10 12 10 11 15 10 13 16 24 13 22 16 14

Table 4.2 (cont.) (Table continued below)

Parameter Pmtr No. S12Y1 S12Y2 S08Y1 S08Y2 S10Y1 S10Y2 S16Y1 S16Y2 S07Y1 S07Y2 S09Y1 S09Y2 S21Y1 S21Y2 Soil type REL REL ACC ACC ACC ACC ACC ACC ACC ACC ACC ACC ACC ACC Pasture type OPG OPG RG RG RG RG RG RG RG RG RG RG F F Green standing dry matter at 50% green



045, 046 and

271 1800 1750 1500 1300 1650 1650 1450 1700 1200 1450 1550 1750 1350 1200

Until 1st harvest change yield at 50% cover

to this value (kg/ha) 500 500


Table 4.2 (cont.)

Parameter Pmtr No. S04Y1 S04Y2 S15Y1 S15Y2 S05Y1 S05Y2 S03Y1 S03Y2 S18Y1 S18Y2 S06Y1 S06Y2 S17Y1 S17Y2 Soil type CCB CCB ACC ACC CCB CCB CCB CCB CCB CCB CCB CCB CCB CCB Pasture type F F F F BMG BMG BMG BMG BMG BMG BMG BMG OPG OPG Green standing dry matter at 50% green



045, 046 and

271 850 850 1650 1450 2300 1650 2050 1600 1100 1650 1600 1400 1900 2200

Until 1st harvest change yield at 50% cover

to this value (kg/ha) 500



139

Table 4.3 Site-by-year calibrated GRASP plant growth parameters. (Table continued below)

Parameter Pmtr No. S01Y1 S01Y2 S02Y1 S02Y2 S13Y1 S13Y2 S11Y1 S11Y2 S19Y1 S19Y2 S14Y1 S14Y2 S20Y1 S20Y2 Soil type REB REB REB REB REL REL REL REL REB REB REL REL ACC ACC Pasture type ASG ASG ASG ASG ASG ASG OPG OPG OPG OPG OPG OPG OPG OPG

Perennial grass basal area (%) 005 8.4 1.63 6 0.38 5.75 0.38 2.88 2.75 1.88 0.63 2.88 2.13 0.88 2

Potential regrowth rate / unit PGBA

(kg/ha/day/basal%) 006 3 8 2 34.2 2.5 13 8 8 9 9 4 4 10 10

Transpiration-use-efficiency (kg/ha/mmT) 007 8 13.5 9 13.5 9 9 15 8 10 9 12 10 12 12

Soil water index at which growth stops 149 0.01 0.01 0.01 0.01 0.01 0.01 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3



Perennial grass basal area (%) 005 1.25 2.13 6 4.25 4.25 3.5 1.63 1.5 2.25 3.25 0.88 2.13 3.25 1.63


(kg/ha/day/basal%) 006 15 10 6 6 6 6 10 6 9 9 8 8 6 6

Transpiration-use-efficiency (kg/ha/mmT) 007 13 10 9 9 9 9 16 9 8 8 14 14 8 10


Table 4.3 (cont.)


Perennial grass basal area (%) 005 1 1 0.75 1.5 3.63 3.75 4.5 3.75 2.38 2.13 1.5 1.3 3 1.33


(kg/ha/day/basal%) 006 14 14 10 10 8 8 5 5 10 10 10 10 10 12

Transpiration-use-efficiency (kg/ha/mmT) 007 16 16 10 10 10 10 8 8 10 10 16 16 10 10



140

Table 4.4 Site-by-year calibrated GRASP nitrogen parameters. (Table continued below)


N uptake at 0 mm of transpiration (kg/ha) 097 4.0 4.0 2.0 2.0 1.0 1.0 7.0 7.0 2.0 3.0 1.0 1.0 2.0 2.0

N uptake per 100mm of transpiration

(kg/ha/100mmT) 098 10 10 10 10 8 8 10 10 8 8 7 7 8 8

Maximum nitrogen uptake (kg/ha) 099 21 25 24 20 18 10 18 18 10 10 13 13 19 19

Maximum nitrogen content in plants (%) 100 2.5 3.5 2.5 3.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5

N content at which growth stops (%) 101 0.8 0.8 0.8 0.8 0.65 0.65 0.65 0.7 0.5 0.5 0.45 0.45 0.7 0.45

N content at which growth is restricted (%) 102 0.9 0.9 0.9 0.9 0.75 0.75 0.75 0.8 0.6 0.6 0.55 0.55 0.8 0.55

Minimum N content in dead (%) 111 0.8 0.9 0.9 0.8 0.7 0.7 0.4 0.5 0.5 0.9 0.4 0.6 0.4 0.3





(kg/ha/100mmT) 098 14 14 7 7 10 10 14 6 14 14 10 10 8 8




N content at which growth is restricted (%) 102 0.8 0.8 0.7 0.7 0.8 0.8 1 0.9 0.8 0.8 0.95 0.95 1 1.2



141

Table 4.4 (cont.)




(kg/ha/100mmT) 098 14 14 12 12 10 10 9 9 10 10 16 16 14 18




N content at which growth is restricted (%) 102 1.3 1.3 1.45 1.3 0.75 0.75 0.6 0.6 0.7 0.7 0.9 0.9 0.7 0.9


Table 4.5 Tree parameters for Site 19.

Parameter Pmtr No. S19Y1 S19Y2 Soil type REB REB Pasture type OPG OPG

Tree Basal Area (m2) 291 0.9 0.9

Wilting point layer 1 for trees 292 4 4



Note: No other sites contained trees.


142

Table 4.6 Site-by-year calibrated GRASP detachment parameters. (Table continued below)


Leaf detachment - wet season (kg/kg/day) 128 0.0005 0.002 0.001 0.003 0.0015 0.005 0.0005 0.0005 0.003 0.003 0.0025 0.0005 0.0015 0.0015

Stem detachment - wet season (kg/kg/day) 129 0.0005 0.002 0.001 0.003 0.0015 0.005 0.0005 0.0005 0.003 0.003 0.0025 0.0005 0.0015 0.0015

Leaf detachment - dry season (kg/kg/day) 130 0.0005 0.002 0.001 0.003 0.0015 0.005 0.0005 0.0005 0.003 0.003 0.0025 0.0005 0.0015 0.0015

Stem detachment - dry season (kg/kg/day) 131 0.0005 0.002 0.001 0.003 0.0015 0.005 0.0005 0.0005 0.003 0.003 0.0025 0.0005 0.0015 0.0015



Leaf detachment - wet season (kg/kg/day) 128 0.0005 0.001 0.0015 0.003 0.004 0.004 0.003 0.003 0.001 0.0035 0.001 0.004 0.003 0.003

Stem detachment - wet season (kg/kg/day) 129 0.0005 0.001 0.0015 0.003 0.004 0.004 0.003 0.003 0.001 0.0035 0.001 0.004 0.003 0.003

Leaf detachment - dry season (kg/kg/day) 130 0.0005 0.001 0.0015 0.003 0.004 0.004 0.003 0.003 0.001 0.0035 0.001 0.004 0.003 0.003

Stem detachment - dry season (kg/kg/day) 131 0.0005 0.001 0.0015 0.003 0.004 0.004 0.003 0.003 0.001 0.0035 0.001 0.004 0.003 0.003

Table 4.6 (cont.)


Leaf detachment - wet season (kg/kg/day) 128 0.005 0.003 0.002 0.005 0 0.0025 0.0002 0.004 0.001 0.002 0.001 0.001 0.0035 0.003

Stem detachment - wet season (kg/kg/day) 129 0.005 0.003 0.002 0.005 0 0.0025 0.0002 0.004 0.001 0.002 0.001 0.001 0.0035 0.003

Leaf detachment - dry season (kg/kg/day) 130 0.005 0.003 0.002 0.005 0 0.0025 0.0002 0.004 0.001 0.002 0.001 0.001 0.0035 0.003

Stem detachment - dry season (kg/kg/day) 131 0.005 0.003 0.002 0.005 0 0.0025 0.0002 0.004 0.001 0.002 0.001 0.001 0.0035 0.003


143

4.5 Comparing model outputs with field data

The calibration process includes comparison of model outputs of important variables with

the field measurements reported in Chapter 3. Four main variables were compared:

• plant available soil water content (PAWC, from MLA 2004a), the amount of soil water

above wilting point and therefore available for transpiration and plant growth;

• green cover (Cover), the proportion of the ground surface covered by transpiring green

plant material;

• nitrogen uptake (N uptake), the total amount of nitrogen the pasture plant contains; and

• total standing dry matter (TSDM), the net result of pasture growth, death and

detachment (largely dependent on the three previous variables).

Three measures of comparing the level of agreement between model outputs and field

measurements are used to evaluate model performance in this section.

Literature relevant to this study uses specific terms relating to the comparison of modelling

outputs and field measured data. Modelling outputs, or simulation results, are expressed as

‘predictions’ as the intention of models like GRASP is to extrapolate variables such as

TSDM beyond field measurements. Field-measured data are commonly expressed as

‘observations’. These terms are applied accordingly throughout this thesis.

4.5.1 Methods

Time-series plots

Mayer and Butler (1993) and Rykiel (1996) recognise visual techniques as one means of

assessing model performance. Time-series plots of predictions and observations of

important variables form the basis of comparisons between model and real system. This

method of assessing the adequacy of calibration parameters is a useful tool for gaining an

initial view of how well the real system is being replicated.

Time series data, while visually informative for individual sites, do not provide objective

evaluation of the overall performance of the model. Two statistical approaches are used for

more meaningful analysis of simulation results: 1) regression analysis; and 2) measures of

prediction deviance. These approaches are explained and applied below.


144

Regression analysis

The statistical correlation between predicted and observed data is a common objective

measure of model performance during calibration (Jones and Carberry 1994; Rykiel 1996;

Grant et al. 1997). Mayer et al. (1994) list key assumptions for using regression statistics

to test model performance, including that all points within a dataset are independent of

each other. Time-series data, like those in this study, are auto-correlated and therefore not

independent; i.e. TSDM at time 2 (e.g. end of wet season) is dependent on TSDM at time 1

(e.g. mid wet season), as well as growth and detachment between time 1 and time 2.

Observations made at different sites or in different seasons are, however, statistically

independent of each other. Thus, by using subsets consisting of data points from different

sites at the same time of the growing season, and from the same site and time but in

different growing seasons, datasets of ‘independent’ points can be constructed and

subjected to statistical analysis. The four subsets of independent data in this study are

‘Early wet’ (data from Harvest 1 and Harvest 5 at all sites), ‘Mid wet’ (Harvest 2 and 6),

‘End of wet’ (Harvest 3 and 7), and ‘End of dry’ (Harvest 4 and 8).

Measuring model performance using regression statistics often involves assessing

predictions against the ‘perfect fit’ function of unity slope and zero intercept (y = x, or the

one-is-to-one line, Mayer et al. 1994; Asseng et al. 1998). As data points are compared to a

set function and not the line of best fit, values for the one-is-to-one statistic are usually

lower than for the more widely used coefficient of determination (r2) if calculated from the

same dataset. The GRASP model has been tested previously using the one-is-to-one

statistic (McKeon et al. 1990; Johnston 1996; Day et al. 1997a; Dyer et al. 2001a). While

many studies report results using this statistic, few provide details of its calculation. In this

study, the one-is-to-one statistic is calculated according to Mayer and Butler (1993), who

define it as ‘modelling efficiency’ (EF):

EF = 1 - (SS about y = x) / (Corrected SS of y)

= 1 - Σ(yi – xi)2 / Σ(yi – mean of y)2 (Equation 9)

where SS is the sum of squares, yi are observed values and xi are corresponding predicted

values.


145

Mayer and Butler (1993) explain the interpretation of EF: “r2 is interpreted as the

proportion of variation explained by the fitted regression line, and EF is a similar measure

against the set line y = x. For regression, the line of best fit cannot be worse than y = mean

of x, so r2 has a lower bound of zero. In Equation 9 the data are compared with a fixed

line, so this restriction is removed. Thus, EF has a (theoretical) lower bound of negative

infinity. A value of zero indicates the fit to y = x is equal to the fit to y = mean of x, with

values of EF less than zero resulting in a worse fit. The calculated EF is thus an overall

indication of goodness of fit. Any model giving a negative value cannot be recommended,

with preferable values close to one indicating a ‘near-perfect’ model.” In this thesis, EF is

referred to by the more common label, r2(1:1).

Prediction variance: Root Mean Squared Error

Another common measure of model performance is based on the actual differences

between observed values and the associated model predictions. Wallach and Goffinet

(1989) and Jones and Carberry (1995) use prediction variance, or the mean squared error

of prediction (MSEP) to measure model performance. Mayer and Butler (1993) present the

related value, root mean squared error (RMSE), sometimes called root mean squared

deviation or difference (RMSD), and GRASP has previously been assessed using this

statistic (Day et al. 1993). RMSE is a useful measure of model performance as it “indicates

the relative degree of model deviation from field observations, and can be used to

determine further statistical properties” (Mayer and Butler 1993).

Mayer and Butler (1993) define RMSE as:

RMSE = {(Σ(yi – xi)2) / n}0.5 (Equation 10)

where yi are observed values, xi are corresponding predicted values, and n is the number of

data pairs.

The higher the value for RMSE, the greater the overall difference between model

predictions and field observations. Relative prediction error can be determined from


146

comparing RMSE to the mean of observed data. These values have been calculated for the

four main variables and are presented in Table 4.7.

Prediction variance: Deviation of predictions from observed data

Mitchell and Sheehy (1997) present an alternative measure of model performance based

upon deviation of predictions from their associated observation values. These deviations

are analogous to the ‘error’ in RMSE. This method plots deviations (as prediction minus

observation) against the observations. The two criteria used to assess model performance

are the ‘envelope’ of acceptable precision about the line of zero deviation, and the

proportion of points that lie within it. Instances where the prediction values are less than

the envelope of acceptable precision are considered under-predictions. Over-prediction

occurs when the predictions exceed the envelope. Model performance can be assessed for

individual growth phases, and across the whole growth cycle using this method.

Rykiel (1996) suggests measurement variance (e.g. 95% confidence limits of the observed

data) as criteria for acceptable model precision. The method of field data collection in this

study has restricted calculation of reliable measurement variance to TSDM data. (Table

9.12 in Appendix 7). Therefore, assessing model performance using the deviations method

is only possible for predictions of TSDM. Two approaches to using measurement variance

as the envelope of acceptable precision have been adopted in this study:

• Each prediction was compared to the individual 95% confidence limits of the

corresponding observation to determine whether it was inside or outside the limits; and

• Each prediction was also compared to the average 95% confidence limits for all

observations (that is, ±35% of the observation value - Table 3.34 in Section 3.5.3) to

determine whether it was inside or outside the limits.

Where predictions fall within the confidence limits, they are considered as reliable an

estimate of TSDM as actually measuring this variable in the field using the methodology

applied in this study.


147

4.5.2 Results

Time-series plots

Figure 4.6 shows time-series plots of TSDM, N uptake, PAWC and Cover for several sites

with differing soil types and species compositions as examples of this method of

comparison. Model predictions are presented as a continuous line and field observations

are displayed as isolated points. The TSDM plots also include 95% confidence limits for

the observation values, and where the prediction line passes within these limits the

prediction is not significantly different to the observation.

Both TSDM time-series plots in Figure 4.6 display close visual agreement between

predictions and observations, except for points early in each growing season at Site 4.

Overall, the time-series plots of TSDM resulting from all individual site and year

simulations were similar in appearance to the two presented here, with observations and

predictions generally corresponding except for a few individual points.

Model predictions of PAWC also show close agreement with observations, with departures

of model predictions from the measured data apparent in some instances. The two

examples of PAWC time-series illustrated in Figure 4.6 represent a very high water

holding capacity soil (Site 5) and a very low water holding capacity soil (Site 12), and the

good visual results suggest GRASP can adequately simulate water balances across a range

of soil types in the region when using local parameter sets.

N uptake and Cover show the most discrepancies between predictions and observations in

the examples presented here. The N uptake predictions for Site 1 clearly correspond with

measured data and demonstrate how well GRASP can simulate this variable. Alternatively,

predictions of N uptake at Site 14 deviate considerably from the measured data and are an

example of poor model performance. Green plant cover also proved difficult to simulate

and some predictions are inconsistent with the field measurements. This has implications

for the calculations of the water balance and plant growth as green cover is an important

variable in calculating evaporation and transpiration, and hence pasture growth. These

implications are discussed later in this chapter.


148

Figure 4.6 Examples of time-series plots of prediction curves (lines) with observed values (points, including

95% confidence limits in the TSDM plots). Data presented are from sites of different soil type and pasture

species composition: Site 4 (forbs on basalt clay); Site 8 (ribbon grass on alluvial clay); Site 1 (annual short

grasses on basalt red earth); Site 14 (other perennial grasses on limestone red earth); Site 5 (barley Mitchell

grass on basalt clay); Site 12 (other perennial grasses on limestone red earth); Site 3 (barley Mitchell grass on

basalt clay); and Site 15 (forbs on alluvial clay).

Site 4

0

1000

2000

3000

4000

5000

Jul-93 Jan-94 Jul-94 Jan-95 Jul-95

Tota

l sta

ndin

g dr

y m

atte

r (kg

/ha)

Observed

PredictedSite 8

0

1000

2000

3000

4000

5000


Tota

l sta

ndin

g dr

y m

atte

r (kg

/ha)

Observed

Predicted

Site 1

0

5

10

15

20

25

30


Nitr

ogen

upt

ake

(kg/

ha)

Observed

PredictedSite 14

0

5

10

15

20

25

30


Nitr

ogen

upt

ake

(kg/

ha)

Observed

Predicted

Site 5

0

100

200

300

400


Plan

t ava

ilabl

e w

ater

con

tent

(mm

) Observed

PredictedSite 12

0

100

200

300

400


Plan

t ava

ilabl

e w

ater

con

tent

(mm

) Observed

Predicted

Site 3

0

20

40

60

80

100


Gre

en p

lant

cov

er (%

)

Observed

PredictedSite 15

0

20

40

60

80

100


Gre

en p

lant

cov

er (%

)

Observed

Predicted


149

Regression statistics

Figure 4.7 displays predictions against observations as a measure of assessing the overall

performance of the final calibration parameters for all individual sites. Four variables

presented are presented: TSDM, N Uptake, PAWC, and Cover. PAWC is presented as

plant available soil water content per 50cm of soil depth. Sites differ in their total soil

profile depth, but the 0 - 50cm soil depth is common to all sites and is the most important

for pasture growth (Christie 1981). Consequently, data for this profile depth are presented

to maintain consistency across sites.

In Figure 4.7, the predicted values are plotted on the x-axis and the observed values on the

y-axis. This format has been adopted according to Mayer and Butler (1993), who reason

that “to conform with statistical assumptions, it is usual to take the observations as the Y-

variate (as these data contain natural variability) and the model predictions as the non-

variable X’s”. The statistical assumptions they refer to are that errors on dependent

variables (field observations) are normally and independently distributed, and that

independent variables are without error (i.e. model predictions are single values and have

no statistical error).

The TSDM plot in Figure 4.7 displays a good overall fit between model predictions and

field observations, particularly at the end of the wet season and end of the dry season

where predictions lie very close to the 1:1 line. Statistics in Table 4.7 confirm that GRASP

simulated this variable with great accuracy at these times (r2(1:1) = 0.966 and 0.954

respectively). Some individual prediction errors are evident, particularly early in the wet

season when a number of over-predictions of TSDM occur, reducing model performance at

this time (r2(1:1) = 0.588).

Figure 4.7 also appears to show good model performance for prediction of PAWC in many

instances. Closer inspection, however, reveals a wide scatter of points at values of less than

100mm/50cm. These soil water contents mostly occurred at the beginning and end of the

growing season and regression statistics confirm that model performance was poorest at

these times (r2(1:1) = 0.510 for the early wet and r2(1:1) = 0.280 for the end of the wet -

Table 4.7). Importantly, during the mid wet season when plant water use is high and

pasture growth rapid, predictions of PAWC were accurate (r2(1:1) = 0.874). The low water

contents at the end of the dry season were also closely predicted (r2(1:1) = 0.926).


150

Predictions of N uptake in Figure 4.7 show greater spread about the 1:1 line than do

predictions of the previous two variables, but many points still lie close to the line,

indicating fair overall model performance. End of wet season N uptake is well simulated

by GRASP (r2(1:1) = 0.774). Early wet season and mid wet season predictions showed

greater variability (r2(1:1) = 0.417 and 0.518 respectively), while end of dry season

predictions of N uptake contained many errors and had no relationship with the observed

data (r2(1:1) = -4.460).

Predictions of green plant cover did not correspond well with observations. Many large

predictions errors occurred in the early wet season (r2(1:1) = 0.284), and almost total over-

prediction of mid wet season cover is an obvious feature of Figure 4.7 (r2(1:1) = -0.576).

End of wet season and end of dry season cover predictions show little agreement with

observed data (r2(1:1) = -0.026 and r2(1:1) = -1.208 respectively).


151

Figure 4.7 Observed vs. predicted data for all sites. Model predictions are the results of using individual site

parameter sets. Variables presented are total standing dry matter (TSDM), nitrogen uptake (N uptake), plant

available water content in the 0-50cm layer of the soil (PAWC), and green plant cover. Associated statistics

are presented in Table 4.7.

Prediction variance: Root Mean Squared Error

RMSE values presented here have been calculated from the same dataset as presented in

Figure 4.7 and are used in conjunction with the regression statistics to build a picture of

model performance.

Calculation of RMSE for predictions of TSDM produced relatively high error values for

the early and mid wet season (RMSE = 273 and 363 kg/ha respectively), and much lower

values at the end of the wet and end of the dry season (RMSE = 132 and 129kg/ha)

0

1000

2000

3000

4000

5000

0 1000 2000 3000 4000 5000

Predicted TSDM (kg/ha)

Obs

erve

d TS

DM

(kg/

ha)

Early wetMid wetEnd of wetEnd of dry

1:1n=163

0

10

20

30

40

50

0 10 20 30 40 50

Predicted N uptake (kg/ha)

Obs

erve

d N

upt

ake

(kg/

ha)


1:1n=163

0

50

100

150

200

250

300

0 50 100 150 200 250 300

Predicted PAWC (mm/50cm)

Obs

erve

d PA

WC

(mm

/50c

m)


n=162 1:1

0

20

40

60

80

100

0 20 40 60 80 100

Predicted green cover (%)

Obs

erve

d gr

een

cove

r (%

)


n=163 1:1


152

indicating better model performance at these times. Results of RMSE showed the same

trends in model performance for TSDM as did the regression statistics.

Results for mid wet season and end of wet season predictions of PAWC have an interesting

feature. Using RMSE alone as an indicator of model performance suggests GRASP

simulates soil water at these times with similar prediction error (RMSE = 24.8 and

25.1mm/50cm, Table 4.7). However, comparing these values to their corresponding

observation means show that in fact mid wet season predictions had a much lower relative

error (24.8 in 111.0mm/50cm) than did end of wet season predictions (25.1 in

48.8mm/50cm). This highlights the point that single measures of model performance can

be misleading. Generally, results of both regression statistics and RMSE indicate similar

model performance when simulating PAWC; that is, predictions during the middle of the

wet season and during the late dry season are most accurate, while those for the beginning

and end of the wet season contain many errors.

Using RMSE as a measure of model performance for predictions of N uptake produced

very similar results to using regression statistics. The lowest value (and best model

performance) was at the end of the wet season (RMSE = 4.2 in 19.9kg/ha), and highest

value (and worst model performance) was at the end of the dry season (RMSE = 9.6 in

10.5kg/ha), with other times being between these extremes (RMSE = 5.0 and 4.7 for early

wet and mid wet season predictions).

Cover is again shown to be a poorly simulated variable throughout the entire growth cycle,

with consistently large error values (RMSE = 18, 24 and 20% for early, mid and end of wet

season predictions).


153

Table 4.7 Results of statistical comparison of predictions and observed values for all sites, using individual site parameter sets. Data presented is total standing dry matter

(TSDM), plant available water content of the 0-50cm layer of the soil (PAWC), nitrogen uptake (N uptake), and green plant cover (Cover).

n TSDM PAWC N uptake Cover

Mean of observed

(kg/ha)

Mean of predicted

(kg/ha)

r2(1:1) RMSE (kg/ha)

Mean of observed

(mm/50cm)

Mean of predicted

(mm/50cm)

r2(1:1) RMSE (mm/50cm)

Mean of observed

(kg/ha)

Mean of predicted

(kg/ha)


Mean of observed

(%)

Mean of predicted

(%)

r2(1:1) RMSE (%)

Early wet 42 561 571 0.588 273 64.7 74.8 0.510 30.7 9.5 7.2 0.417 5.0 29 30 0.284 18

Mid wet 42 1896 1774 0.678 363 111.0 99.9 0.874 24.8 19.1 18.3 0.518 4.7 50 70 -0.576 2 24

End of wet 42 1 2513 2553 0.966 132 48.8 31.5 0.280 25.1 19.9 22.3 0.774 4.2 29 39 -0.026 2 20

End of dry 37 1902 1887 0.954 129 20.2 14.9 0.926 13.3 10.5 9.8 -4.460 2 9.6 2 3 -1.208 2 4

1 n = 41 for PAWC 2 Negative values for r2(1:1) indicate poor model performance (Section 4.5.1).


154

Prediction variance: Deviation of predictions from observed data

Figure 4.8 shows the deviation of each prediction (point) from the corresponding

observation (line of zero deviation) for all TSDM measurements. This figure is an

alternative presentation of the TSDM data in Figure 4.7, with the magnitude of deviations

being more obvious than on a diagonal regression. For clarity of display, the envelope of

acceptable precision in Figure 4.8 represents average measurement variance (±35% of the

observation value), as plotting individual 95% confidence limits for every observation

results in a very crowded figure that is difficult to interpret. Using average measurement

variance as an example of an envelope of acceptable model precision has resulted in most

predictions being within the envelope, indicating very good overall model performance

when predicting TSDM.

Figure 4.9 shows a subset of data from Figure 4.8, and is an alternative view of the TSDM

site data presented in Figure 4.6. In this figure, 95% confidence limits for each observation

have been displayed and form individual envelopes of acceptable precision for each

prediction. It can be clearly seen that two predictions at Site 4 and one prediction at Site 8

fall outside the 95% confidence limits of their corresponding observations, indicating

GRASP has not adequately simulated the processes affecting TSDM at these times.

A summary of the performance of GRASP when predicting TSDM throughout the pasture

growth cycle is presented in Table 4.8. Results show that 83% (135 of 163) of all

predictions of TSDM fell within the 95% confidence limits of their associated

observations, while 85% of all predictions of TSDM (139 of 163) fell within average

measurement variance. Greatest discrepancies occur at low values early in the wet season

(36%, or 15 of 42 predictions are outside individual 95% confidence limits, with 11 of the

15 being over-predictions). Mid wet season growth was better simulated as only 26%, or

11 of 42 predictions were outside individual 95% confidence limits, with 7 of the 11 being

under-predictions. All predictions except one were inside the individual 95% confidence

limits for the end of the wet season and end of the dry season, indicating a high level of

model performance when predicting TSDM at these times.

Results using ±35% of observation values as the envelope of acceptable precision are very

similar to using individual 95% confidence limits (Table 4.8), suggesting that an envelope


155

of acceptable precision set at ±35% of observation values is a useful measure of model

performance in this study.

Source data for the figures and tables presented in this section are found in Appendix 8

(Table 9.13).


dry matter for all sites and years. Predictions are generated by GRASP using the individual parameter sets

presented in Table 4.1 to Table 4.6. Dashed lines indicate the envelope of acceptable precision, equal to the

average magnitude of measurement variance (±35% of observation values).

Figure 4.9 Deviation of prediction values (points) from their corresponding observations of TSDM (x on line

of zero deviation and including 95% confidence limits) for Site 4 and Site 8.

Site 4

-1500

-1000

-500

0

500

1000

1500

0 1000 2000 3000 4000 5000

Observed TSDM (kg/ha)

Dev

iatio

n of

pre

dict

ion

from

ob

serv

ed T

SDM

(kg/

ha)

Observed

Prediction deviationSite 8

-1500

-1000

-500

0

500

1000

1500

0 1000 2000 3000 4000 5000


Dev

iatio

n of

pre

dict

ion

from

ob

serv

ed T

SDM

(kg/

ha)

Observed

Prediction deviation

-2500

-2000

-1500

-1000

-500

0

500

1000

1500

2000

2500

0 1000 2000 3000 4000 5000

Observed value (kg/ha)

Dev

iatio

n of

pre

dict

ion

from

obs

erve

d va

lue

(kg/

ha)

Early wet

Mid wet

End of wet

End of dry

n = 163


156

Table 4.8 Summary of the number of predictions of TSDM that fell outside the envelopes of acceptable

precision.

Under-prediction Over-prediction Total outside envelope

n 95% cl 1 ±35% obs 2 95% cl ±35% obs 95% cl ±35% obs

Early wet 42 4 4 11 13 15 (36%) 17 (40%)

Mid wet 42 7 3 4 2 11 (26%) 6 (14%)

End of wet 42 1 1 0 0 1 (2%) 1 (2%)

End of dry 37 0 0 1 0 1 (2%) 0 (0%)

Total 163 12 (7%) 8 (5%) 16 (10%) 15 (9%) 28 (17%) 24 (15%)

1 95% cl - envelope of acceptable precision equals individual 95% confidence limits for the corresponding

observation. 2 ±35% obs - envelope of acceptable precision equals ±35% of the observation value.

4.6 Discussion of the calibration procedure and modelling results

4.6.1 Calibration

Calibration of GRASP to simulate pasture growth at the study sites required the assembly

of sets of input parameter values. The resultant sets of parameter values (Table 4.1 to Table

4.6) are the first of their kind to be constructed for pastures in the VRD. A set of constant

parameter values for a ‘standard’ tropical grassland (C4 parameters, G.M. McKeon,

pers.comm.) has been developed from studies in Queensland (e.g. Day et al. 1997a) and

are compared with results from this study (Table 4.9). For most parameters in Table 4.9 the

standard grassland values fall within the range of values resulting from site calibration in

this study. However, the gap between the mean parameter values from this study and the

C4 parameter values indicate some underlying differences between soils and pastures in the

VRD compared with those contributing to the standard grassland parameters. These are

discussed here.

The most apparent difference between land types in the VRD and those contributing to the

C4 parameters is soil water holding capacity. As discussed in Chapter 3, the water holding

capacities (WHC) of soils in this study were relatively high compared with published data

from other regions of northern Australia. Consequently, the parameter values defining

WHC for each soil layer are also higher than those of the C4 parameter set. This is further


157

exaggerated by the fact that a higher proportion of cracking clay soils contribute to the

mean values for this study than is the case for the C4 parameters.

Some pasture parameters also differ between the two locations. Pastures in the VRD are

more dense than those in Queensland (PGBA = 2.57 vs. 1.00, Ht1000kg = 12.5 vs. 20cm).

Differences in species composition are expressed as a more erect plant habit (GY50GC =

1529 vs. 1000kg/ha).

Values for TUE demonstrate differences between the VRD (10.9kg/ha/mm at vapour

pressure deficit of 20hpa) and the C4 parameters (13.5kg/ha/mm). Day et al. (1997a)

reports TUE showing considerable variation between pasture communities in Queensland

(7 to 25kg/ha/mm) and attributes this variation to differences in species composition.

Downes (1969) found that C4 species have a higher water use efficiency (which includes

evaporation) than C3 species, but design of this study (measuring whole pasture swards

rather than individual species) makes such a comparison difficult. Ludlow (1985) states

that the main reason for higher growth rates in C4 plants is that the absence of

photorespiration in the photosynthetic process leads to greater conversion efficiencies.

Christie 1978 gives water use efficiency of 3.9kg/ha/mm of stored water for a C3 native

grassland and 6.9kg/ha/mm of stored water for a buffel grass (C4) pasture. Bolger and

Turner (1999) report TUE values for temperate pasture of around 30kg/ha/mm, but

includes root biomass accumulation, a variable not measured in this study. Thumma et al.

(1998) point out that measuring TUE at the field level is difficult and measurements taken

at one time in the growing period may not accurately represent water use efficiency over

the whole season. Their study reports a TUE value of 3.6g/kg in pot trials using

Stylosanthes scabra but such values are difficult to compare with data in this study. Walker

and Richards (1985) suggest that nutrient status is a contributing factor to TUE, with a

general deficiency sometimes lowering transpiration efficiency. The composition of

pastures, methods and timing of measurements, and the effect of nutrient status account for

some of the variation between values reported in this and other studies.

The rate of N uptake per unit of transpiration is higher for pastures in the VRD than in

Queensland (Nup100T = 10.5 vs. 6.0kg/ha/100mm) although little difference in the total

nitrogen supply (MaxN = 24.1 vs. 20.0kg/ha) or internal dilution (%N0Grow = 0.72 vs.


158

0.68) are evident. The influence that these individual parameter values have on predictions

of pasture growth is explored later in this thesis (Chapter 6).

Table 4.9 Summary of parameter values derived during calibration of GRASP to the study sites.

Parameter number

Minimum Maximum Mean Standard C4 grassland1

WHC Layer 1 (mm) p026 & p029 18 40 31.6 15

WHC Layer 2 (mm) p027 & p030 46 160 106.6 60

WHC Layer 3 (mm) p028 & p031 12 150 95.9 50

GY50GC (kg/ha) p045 500 2550 1529 1000

Ht1000kg (cm) p096 8 24 12.5 20

PGBA (%) p005 0.63 6.00 2.57 1.00

PotRegrow

(kg/ha/day/%PGBA)

p006 4 15 8.85 3.5

TUE (kg/ha/mmT) p007 8 16 10.9 13.5

SWIX0Grow p149 0.01 0.3 0.16 0.3

NUp0T (kg/ha) p097 1 10 3.4 5.0

NUp100T (kg/ha/100mmT) p098 6 18 10.5 6.0

MaxN (kg/ha) p099 10 38 24.1 20.0

%N0Grow (%) p101 0.5 1.35 0.72 0.68

Detachment (kg/ha/day) p128 to p131 0 0.005 0.0023 0.003

1 General parameters for a ‘standard’ C4 grassland in Queensland (G.M. McKeon pers.comm.) 2 Depth of Layer 3 varies from 1cm to 50cm.

During the calibration process, achieving a good fit between predictions and observations

of early wet season growth across many sites was difficult to achieve. Results presented in

Section 4.5 show 40% of predictions of TSDM at this time did not correspond with field

observations. Additionally, several instances occurred during the site calibrations where

predictions of early wet season TSDM so poorly corresponded to the observed values that

special measures were required to enable GRASP to better simulate pasture growth at this

time. Two measures were used: one focused on the yield-cover relationship; the other on

soil water.

Three sites (Sites 9, 15 and 21) required very low initial values for GY50GC (Table 4.2) in

order to best simulate early wet season growth, then required an increase in this parameter

value to produce acceptable predictions for the remainder of the growth cycle. A number


159

of common features across the three sites provide some indication as to why these

measures were necessary. Each instance occurred only at the first harvest after site

establishment and only involved sites on cracking clay soils with pastures containing

roughly equal proportions of perennial grasses and forbs at the time. It is feasible that on a

soil surface free of carryover material and in the absence of grazing, forbs with a prostrate

habit dominated the yield-cover relationship early in the growing season to such an extent

that a very low value for GY50GC was required for GRASP to replicate pasture growth at

this time. As the season progressed, other more-erect species increased in size and altered

the underlying yield-cover relationship. An increase in value was required mid way

through the growing season for GRASP to simulate pasture growth for the remainder of the

season.

Given than GRASP is not capable of simulating a mixed pasture sward, and that other sites

share similar characteristics to Sites 9, 15 and 21, it is probable that some of the poor early

wet season simulation results presented in Section 4.5 are due to the array of processes

driving growth in mixed pasture swards at this time being too complex to replicate with a

mono-specific model such as GRASP.

The second special calibration measure involved soil water. During the calibration

procedure for Site 18 at Rosewood Station, predictions of soil water content were

considerably lower than the measured data on 8 February 1995 (Harvest 1). Recorded

rainfall at the station homestead (5km away) did not fully account for the increase in soil

water since the previous measurement, and 100mm of ‘irrigation’ were added to the soil

profile at this site during the period prior to Harvest 1 to improve the model’s calculation

of the soil water balance. This discrepancy likely occurred through two events, either alone

or in combination: 1) rainfall amounts can vary considerably over short distances as a

result of thunderstorms during the early wet season (Slatyer 1960), and more rainfall could

have fallen at the study site than at the homestead; and 2) high intensity thunderstorms can

produce substantial surface runoff, even on cracking clay soils, and Site 18 may have

experienced net run-on during such a storm, adding unrecorded water to the soil profile.

The importance of a detailed understanding of the dataset used when calibrating GRASP is

illustrated in Figure 4.7. A substantial under-prediction has occurred for one mid wet

season data point (field measurement = 240mm/50cm, model prediction = 144mm/50cm).


160

This data point relates to the inundation event at Site 15 highlighted in the field results

(Figure 3.18 and Plate 22 in Appendix 1). During calibration, this data point was ignored

as it was a known anomaly and model prediction was not expected to concur with the field

measurement in this instance. Similarly, on one occasion plant nitrogen yield at Site 15

was measured as 56.4kg/ha, an extremely high value compared with other results in this

study. This data point was also considered an anomaly and disregarded during calibration

of nitrogen parameters at this site.

Summary of the calibration procedure

In this study, calibration of GRASP using field data was a complex, partially subjective

procedure and while it followed a structured approach, it was not exactly the same each

time. Some expertise was required. Previous experience at calibrating GRASP (Ken A.

Day) was combined with an in-depth knowledge of the field dataset (myself) to derive the

required input parameter values. The subjective component of the calibration procedure

allowed the unquantifiable elements of the skilled researchers experience and judgement to

be expressed, and these have been recognised as having benefits for accurate simulation

and associated understanding of the operation of the real system (Rykiel 1996). Provision

of a fully objective framework for more objective calibration (such as developing an

algorithm) would, however, enable a greater number of researchers to undertake

calibration of GRASP and potentially lead to its more widespread use.

4.6.2 Comparison of model predictions with field data

The accuracy with which GRASP simulates values for important soil and pasture variables

is the primary basis for assessing the models capability when simulating the processes of

pasture growth at the study sites.

Total standing dry matter

True pasture growth is difficult to measure directly and is more usually represented by

change in total standing dry matter from one point in time to another during the growing

season, allowing for detachment and defoliation. In pastures protected from grazing, losses

of dry matter during the growing season are usually very small (Vallentine 1990), making

TSDM a good surrogate for cumulative pasture growth. Losses during the dry season are

mostly due to the natural detachment processes of wind and rain. Close prediction of


161

TSDM is, therefore, a strong indication that GRASP is capable of accurately simulating the

processes governing native pasture growth, death and detachment during the study period.

Table 4.10 summarises the performance of GRASP when simulating TSDM for each phase

of the pasture growth cycle. It is clearly apparent that end of wet season and end of dry

season TSDM’s have been closely predicted. Each measure of model performance

indicates a very high degree of accuracy with r2(1:1) values high (0.966 and 0.954

respectively), RMSE values low (132 and 129kg/ha) and a very high proportion of

predictions within measurement variance (98% and 98%). These results exceed those of a

comparable study by Day et al. (1997a), who reported that GRASP calibrated to native

pastures in Queensland accounted for 86% of the variation in yield at the end of the

summer growing period, with an average error of 445kg/ha, while Richards et al. (2001)

reported that GRASP accounted for 53 to 86% of the variation in pasture yields in central

Australia. The better performance of GRASP in this study is attributed to a different

approach to calibration than Day et al. (1997a), who calibrated GRASP based on

developing a single parameter set capable of describing growth across all years at a

location. In this study, separate parameter sets were constructed for each individual season

at a location, allowing parameterisation to account for year-to-year variability in factors

affecting pasture growth. Richards et al. (2001) used data not specifically collected for the

purpose of calibrating GRASP that did not include some of the required variable (e.g.

nitrogen content and yield).

Criteria for determining acceptable model performance are dependent on the intended

application of the simulation results, but some general guides are offered in the literature.

Mayer and Butler (1993) suggest that regressing predictions against the one-is-to-one

function should produce positive values (i.e. r2(1:1) > 0), and Rykiel (1996) suggests that

75% of predictions should fall within measurement variance for acceptable model

performance. Applying these criteria indicates that GRASP has acceptably simulated the

processes determining standing pasture at the end of the wet season and end of the dry

season.

Goodness of fit comparisons for the early wet season indicate that prediction errors were

largest during this period (r2(1:1) = 0.588, RMSE = 273kg/ha, and only 64% of predictions

are within measurement variance). These errors are visible as the over-predictions in


162

Figure 4.8. Improved predictive skill is apparent for mid wet season TSDM (r2(1:1) =

0.678, RMSE = 363kg/ha, and 74% of predictions are within measurement variance).

Using the criteria of Mayer and Butler (1993) and Rykiel (1996) suggests that some early

and mid wet season predictions cannot be considered accurate.

Table 4.10 Summary of model calibration results for total standing dry matter.

n Mean of observations

(kg/ha)

Mean of predictions

(kg/ha)


Proportion of predictions within 95%

conf limits of observations

(%)

Early wet 42 561 571 0.588 273 64

Mid wet 42 1895 1774 0.678 363 74

End of wet 42 2513 2553 0.966 132 98

End of dry 37 1781 1887 0.954 129 98

TSDM was over-predicted for eight of the eleven observations below 250kg/ha (Figure

4.10). Six of these eight over-predictions involved sites on cracking clay soils and, on each

occasion, soil water in Layers 1 and 2 were also over-predicted while soil water in Layer 3

was under-predicted. These over-predictions of soil water likely resulted in over-prediction

of transpiration, and hence pasture growth. Errors in soil water predictions on cracking

clay soils are attributed to surface water from rainfall running into deep soil cracks and

directly entering Layer 3 without infiltrating through Layers 1 and 2 above (discussed

further in the next section). The water balance module in GRASP does not account for

such direct entry of water into the lower soil profile, but rather infiltrates it in the

conventional ‘layered bucket model’ manner. Consequently, GRASP over-predicted soil

water availability in the upper soil profile, and under-predicted soil water in the 50-100cm

layer of the profile.

The remaining two instances of over-prediction of TSDM below 250kg/ha occurred on red

earth sites that contained a high portion of annual grasses. It was found during the field

study that burning these sites to remove carryover material at the end of the first study year

caused a large reduction in plant density the following season (discussed in Section 3.5.3).

While plant density was reduced, individual plant growth was observed to be very


163

vigorous, presumably because of little inter-plant competition for resources.

Parameterising a perennial grassland model such as GRASP to simulate an annual grass

pasture regrowing from a fire-affected seed bank approaches the limits of the model’s

intended application, and sizeable prediction errors have occurred in these instances.

To summarise, GRASP tended to over-predict pasture biomass when observation values

were less than 250kg/ha (Figure 4.10). Carter et al. (1996b) also reported over-prediction

of low pasture biomasses, but Day et al. (1997a) found that GRASP performed well when

predicting low yields. Johnston (1996) also found that GRASP model tended to

overestimate yield early in the growing season.

An alternative perspective on interpreting prediction errors early in the wet season is that

the initiation of pasture growth in the field is affected by many factors, some of which are

not included in the GRASP model. Many tropical pasture plants are known to have

dormancy requirements that must be fulfilled before germination begins (Mott 1978;

Jacobsen 1981; Andrew and Mott 1983; McIvor and Gardener 1994; Graham et al. 2004).

These mechanisms are important to prevent excessive germination after isolated early wet

season storms so that failure to receive follow-up rain does not lead to high seedling death

rates, severely jeopardising the future of the species. It is conceivable that dormancy

mechanisms also exist in established perennial plants to prevent premature re-shooting for

the same reasons. The existence and operation of these mechanisms in the VRD is

unknown, but they are another potential source of error between predictions and

observations of TSDM early in the growing season.


164

Figure 4.10 Deviation of predictions (points) from observed values of total standing dry matter (x on line of

zero deviation and including 95% confidence limits) for all observations less than 500kg/ha.

Plant available soil water content

Results in Table 4.7 show that soil water content was best predicted during the mid wet

season (r2(1:1) = 0.874, RMSE = 24.8mm/50cm) when the soil profile was quite wet

(111.0mm/50cm), and at the end of the dry season (r2(1:1) = 0.926, RMSE =

13.3mm/50cm) when the soil profile was almost dry (20.2mm/50cm). These results show

that the calibration values for field capacity and wilting point of each layer are accurate,

and that GRASP adequately simulated the important water balance components (runoff,

infiltration, evaporation, transpiration and drainage) at these times.

However, simulations of soil water were much poorer in the early wet season (r2(1:1) =

0.510, RMSE = 30.7mm/50cm), and at the end of the wet season (r2(1:1) = 0.280, RMSE =

25.1mm/50cm). Simulating the wetting up of the soil profile at the beginning of the

growing season, and drying out at the end of the growing season proved difficult,

particularly on the cracking clay soils. A likely contributor to early wet season prediction

errors of soil water in these soils is the direct entry of water into the deepest soil layer

(50+cm) without having to pass through those layers above it. Surface runoff during

intense thunderstorms at the beginning of the wet season runs into the deep soil cracks that

have opened during the dry season, and infiltrate laterally into the soil profile at depth.

This has also been observed in other studies (van Dam 2000; Taddese et al. 2002; Bethune

and Wang 2004). Direct wetting of the lowest soil layer before upper layers have reached

field capacity is not accounted for in GRASP and this leads to discrepancy between

-500

-250

0

250

500

0 100 200 300 400 500


Dev

iatio

n of

pre

dict

ion

from

obs

erve

d TS

DM

(k

g/ha

)

Observed

Prediction deviation


165

observations and predictions of soil water. Predictions of soil water are much more

accurate during the mid wet season when cracks have closed and direct wetting of Layer 3

from surface runoff no longer occurs.

Close examination of end of wet season simulation results for soil water showed that most

errors at his time involved under-prediction in cracking clay soils. Several locations

received high amounts of rainfall in just a few days near the end of the growing season

(e.g. Victoria River Downs in April 1996 - Figure 3.9). GRASP responds by infiltrating

this rainfall until the entire soil profile is at field capacity, and drains through any

remaining water. This has the effect of removing excess soil water very quickly. If no

further rainfall occurs, evapotranspiration then proceeds to dry the soil profile. In the field,

however, drainage of excess water is often much slower, particularly during the second

half of the wet season when monsoon rains have often already filled the soil profile.

Consequently, soil water values measured in the field some weeks after such events were

often higher than predicted values.

Another likely contributor to discrepancies between predicted and observed PAWC is

spatial variation in rainfall. While many individual sites had raingauges, resources

available to this study did not permit manual recording of individual rainfall events, or

installation of sufficiently reliable automatic weather recording stations at Victoria River

Downs, Rosewood, and Auvergne (Site 15 to Site 21). Simulations for these sites used

station homestead rainfall records, some kilometres from the study sites (Section 3.3.5).

Differences in rainfall received at the homesteads and at the study sites will lead to errors

in calculating the water balance.

Overall, PAWC was predicted less accurately than Day et al. (1997a) (r2 = 0.95 and mean

error of 15mm/m). However, their study contained few of the cracking clay soils that

experienced most of the prediction errors recorded here. Johnston (1996) also found that

the GRASP model had greatest difficulty predicting soil moisture at sites with cracking

clay soils due to soil cracks allowing water to wet the soil profile at depth at the same time

as the surface layers.


166

Plant nitrogen uptake

In contrast to TSDM and PAWC, many discrepancies between observed and predicted N

uptake data are evident throughout the entire growth cycle (Figure 4.7). Best results for

prediction of N uptake occur at the end of the wet season (r2(1:1) = 0.774, RMSE =

4.2kg/ha), demonstrating the capability of the model when simulating total seasonal N

uptake. However, it must be remembered that values for the parameter maximum N uptake

(MaxN) are taken directly from field measurements, and varied between sites and

sometimes between years at the same site (Table 4.4). Thus, predictions of N uptake at the

end of the growing season are heavily influenced by the data they are compared with.

Under these circumstances, strong correlations between predictions and observed data are

inevitable.

Predictions of N uptake during the wet season deviate considerably from observed values

(r2(1:1) = 0.417, RMSE = 5.0kg/ha for early wet season and r2(1:1) = 0.518, RMSE =

4.7kg/ha for mid wet season predictions). One source of prediction errors at these times is

the manner in which GRASP calculates N uptake. GRASP assumes nitrogen uptake is

related to cumulative transpiration, with the parameters NUp100T and MaxN governing

the relationship (Figure 4.3). Assumptions in the model include a constant rate of N uptake

per unit of transpiration until available N is exhausted, and that the available N supply

remains constant for each growing season unless otherwise specified by the user. These

assumptions are an over-simplification of the processes determining plant N uptake.

Timing of mineralisation of soil organic matter, decomposition of plant litter, N fixation by

legumes, addition of N from rainfall, and losses of N through leaching, volatilisation and

removal of biomass all affect the amount of nitrogen available for plant uptake (Abbadie et

al. 1992). Additionally, the ability of plants to acquire nutrients differs between species

(Schmidt and Lamble 2002). How these factors combine to affect the availability and

uptake of nitrogen throughout the growing season is not well understood in the region, and

they are not part of the current model structure.

Over the longer term, similar lack of knowledge about the location-to-location and year-to-

year variation in total available nitrogen also poses challenges to the modelling process.

Further research into these aspects of nitrogen dynamics in the VRD is recommended, and

some modification of the model structure may be necessary in order to better simulate N

uptake in the region.


167

Results showed no correlation between observations and predictions of N uptake at the end

of the dry season. Changes in N uptake (total plant N yield) from the end of the growing

season to the end of the dry season occur in two main ways: 1) internal translocation of

nutrients to basal reserves or into seeds which are shed following plant maturity; and 2)

direct loss of plant N through detachment of leaf and stem material (Vallentine 1990). The

amount of N lost depends upon the content of the detached material which is unlikely to be

uniform over time or space. Simulating such complex N pathways is not possible with

GRASP, and this contributes to the prediction errors.

An additional cause of discrepancy between predictions and observations of N uptake is

the method of calculating plant N from field data. Plant nitrogen concentration is

determined at each harvest from a pasture sub-sample, and then multiplied by plant

biomass to give total N uptake (Section 3.3.5). Errors associated with measuring plant

biomass, obtaining a representative pasture sub-sample, and laboratory analysis of N

concentration may sometimes compound and lead to sizeable error in the field estimate of

N uptake. One observation of N uptake in this study was so anomalous (3.4.3.8) that it was

assumed an error of this kind occurred and the data point was ignored during the

calibration procedure.

GRASP assumes nutrients other than nitrogen are non-limiting (Section 4.2.3). However,

phosphorus is known to limit native pasture growth in northern Australia (Section 2.3.2).

Norman (1962) found that N uptake interacted with P uptake at Katherine and it is

conceivable that similar interactions occur in the VRD. Consequently, simulating N uptake

as the sole limiting nutrient may result in prediction errors under certain conditions. Better

understanding the role of P (and other nutrients) in pasture growth in the region, and the

incorporation of a phosphorus index into GRASP could improve pasture growth

simulations.

To summarise, potential errors exist in both measuring N uptake in the field and simulation

by GRASP of the biological processes involved. Considering the importance of nitrogen in

the pasture system, greater emphasis on understanding the factors involved in nitrogen

dynamics is recommended for future investigation.


168

Cover

Cover was a poorly simulated variable. One likely reason is that the different growth

patterns of individual species in a mixed sward is beyond the capacity of GRASP to

simulate as it treats pastures as one species pool. Prostrate forbs (e.g. Rhyncosia australis,

Abutilon otocarpum and Neptunia monosperma) were observed to increase their green leaf

area very quickly early in the wet season compared to grass species growing in the same

location. These same forbs readily senesced and detached leaves soon after the onset of

water stress, causing rapid decreases in green plant cover relative to grasses. Attempting to

fully account for such processes would require considerable additional data on the growth

and death characteristics of individual species, and inter-plant competition. Modification of

GRASP to allow multiple species pools to be simulated simultaneously would also be

required. These are substantial tasks and unlikely to be solved in the foreseeable future,

thus some error is always likely when simulating green cover using GRASP in mixed

pasture swards.

Errors in prediction of green plant cover have significance because of the central role cover

plays in determining soil water balance and plant growth in GRASP. Green plant cover and

the soil water supply index determine transpiration in the model, which in turn determines

growth. Growth adds to the green pool of dry matter, and a set proportion of this green

pool is transpiring green material. This additional transpiring green material is partitioned

into leaf and stem, therefore increasing the green cover available for transpiration and

growth next day. Thus, errors in green cover will lead to errors in plant growth, either

through over-predictions compounding to produce even more green cover and plant

growth, or under-predictions failing to provide enough green cover to produce the changes

in TSDM measured in the field.

Summary

The prediction errors for green cover likely resulted in subsequent errors in calculating

transpiration, N uptake and TSDM. Greatest errors for N uptake and TSDM occurred

during the growing season, reflecting the importance of correctly simulating transpiration.

However, by the end of the wet season TSDM is very well predicted despite poor

correlations for cover and soil water. This suggests another factor is determining TSDM at

this time. Analysis of predictions of N uptake at the end of the wet season showed that 40

of 42 predictions equalled the value for the parameter MaxN (Table 4.4), indicating that all


169

available nitrogen had been taken up, and therefore became a limitation to plant growth.

This issue will be investigated further in Chapter 6.

4.7 Conclusions

Conclusions from the calibration of GRASP and performance of the model in predicting

field data are:

• GRASP is capable of simulating pasture TSDM throughout the growth cycle at

selected times and locations in the VRD with a high level of reliability when calibrated

using site-specific field data;

• TSDM is better simulated at the end of growing season than during the growing

season, with TSDM values <250kg/ha most poorly predicted;

• soil water content is well simulated. Better accounting for direct entry of water in the

deepest soil layer of cracking clay soils would likely improve GRASP’s ability to

replicate the soil water balance for these soils;

• cover was a problematic variable to predict, largely due to the different growth patterns

of perennial grasses and annual species in mixed pasture swards; and

• plant nitrogen uptake at the end of the growing season is well predicted but the rate of

N uptake during the growing season was more difficult to simulate.

• during the growing season, predictions of TSDM depend upon accurate simulation of

soil water and green plant cover. End of growing season TSDM is more closely

determined by total nitrogen supply.

Results of comparing model predictions with field observations demonstrate that GRASP

is a suitable modelling framework for simulating pasture growth in the VRD when

calibrated using local field data. However, it must be remembered that the field

observations used for comparison are the same data from which the model calibration

parameters are derived. Thus, exaggerated model performance is highly likely. In order to

assess the actual reliability of GRASP when extrapolated to locations where no field data

contributes to the development of model parameters, some assessment of the model against

independent data is required. Such an assessment is the focus of the next chapter.

Chapter 5 Testing the performance of GRASP for application to the wider landscape

170

5.0 Testing the performance of GRASP for application to the wider landscape

5.1 Introduction

The previous chapter described the process and outcomes of calibrating the GRASP model

to individual locations in the VRD from field measured data. In order to assess the

adequacy of the model for simulating pasture growth at times and locations not covered by

field measurements (i.e. most of the region), some evaluation of model predictions against

independent field data is necessary. This process is known as empirical validation, or

model testing. Rykiel (1996) describes validation as “a testing process on which to base an

opinion of how well a model performs so that a user can decide whether the model is

acceptable for its intended purpose. A model only needs to be good enough to accomplish

the goals of the task to which it is applied”.

This chapter aims to validate GRASP using independent field data in order to assess the

capacity of the model when predicting pasture growth across the wider landscape of the

VRD.

Before model testing can occur, the many site-specific values for parameters developed in

the previous chapter need to be reduced to single, generic values. A framework for

developing generic parameter sets based on grouping sites according to similar soil type

and pasture composition is described and the resultant parameters presented in Section 5.2.

Data from other studies suitable for validation of GRASP is not available, so a technique is

explained in Section 5.3.2 that allows the same field data used for calibration to again be

used as independent validation data. Results of simulations using generic parameter sets

are compared with field data in Section 5.3.3. Implications of these results on the

application of GRASP to the wider VRD landscape are discussed in Section 5.4 and

conclusions presented in Section 5.5. The structure of this chapter is illustrated in Figure

5.1.


171


5.2 Generic parameters suitable for extrapolation across the landscape

The parameter sets developed in Chapter 4 (Table 4.1 to Table 4.6) are specific to the

individual site in the season of measurement and, on their own, are of limited value for

application of the model to other times and locations. Developing generic parameters sets

that represent common land types in the region allows extrapolation of GRASP beyond the

boundaries of the current study.

5.2.1 Classifying study sites by land type

Grouping the study sites according to common features, then summarising their calibration

results is the approach taken to developing generic land type parameter sets. A framework

is required for this task. Sites in this study have been classified using many criteria

(Appendix 2), but the most suitable option for describing land types from a pasture growth

modelling perspective is not known for the VRD. The most common options for describing

the land types in the VRD are based on: 1) soil type (e.g. Northcote 1979; Isbell 1996); 2)

vegetation composition (e.g. Wilson et al. 1990); or 3) some combination of the two (e.g.

Using Chapter 3 results as the source of independent model testing data

(Section 5.3)

VRD as a region Pasture types Soil types

Evaluation of the capability of GRASP to predict independent field data (Section 5.3.3)

Final generic parameter sets for application across the VRD (Section 5.2.2)

Discussion and conclusions on independent validation

(Section 5.4 and Section 5.5)

Assembling parameter sets and data for model testing (Section 5.3.2)

Framework for aggregating sites to develop generic parameter sets (Section 5.2.1)

Soil x pasture combinations


172

Stewart et al. 1970; Tothill and Gillies 1992). To determine the most appropriate option for

describing land types in future modelling studies, sites are grouped in each of four ways:

• by soil type (Soil);

• by pasture species composition (Species);

• by a combination of both (Soil x Species); and

• all sites together in one group (Regional VRD).

Based on the field results in Chapter 3, a matrix of sites classified according to soil types

and species compositions is presented in Table 5.1. This matrix is the foundation for

grouping sites when deriving the generic land type parameters.

Table 5.1 Matrix of study sites (numbers in table) as they relate to soil and species groups for development

of generic parameter sets.

Species groups Soil groups

Structured red earth

Alluvial cracking clay

Basalt cracking clay

Annual short grasses 1, 2, 13

White grass 14, 19 20

Ribbon grass 7, 8, 9, 10, 16

Forbs 15, 21 4

Barley Mitchell grass 3, 5, 6, 18

Other 1 11, 12 17

1 This species group will be excluded from analyses of predictions against observed data as it contains sites

of unique and diverse species composition, and no independent field data is available for comparison.

All sites are included in one of three broad Soil groups as there is a limited range of soil

types in this study. The three Soil groups are:

• structured red earth (Sites 1, 2, 11, 12, 13, 14 and 19);

• alluvial cracking clay (Sites 7, 8, 9, 10, 15, 16, 20 and 21); and

• basalt cracking clay (Sites 3, 4, 5, 6, 17 and 18).

The two separate red earth classifications in the field results section of Chapter 3 (red earth

overlying basalt, and red earth overlying limestone) have been joined to form a single soil

type here (structured red earth) as, for the purposes of simulation modelling, they are

similar enough in their water holding characteristics (Table 4.1) to be considered as one

soil type.


173

Species composition is more diverse and, as a result, five Species groups are presented:

• barley Mitchell grass (Sites 3, 5, 6 and 18);

• ribbon grass (Sites 7, 8, 9, 10, and 16);

• white grass (Sites 14, 19 and 20);

• annual short grasses (Sites 1, 2 and 13); and

• forbs (Sites 4, 15 and 21).

The sites in each Species group are the same as the pasture groups presented in Chapter 3

(Section 3.4.3.4 to Section 3.4.3.8), with the exception of the ‘white grass’ group, which is

a subset of the ‘other perennial grass dominated sites’ classification of Section 3.4.3.6.

Three sites (Sites 11, 12 and 17) are not included in the white grass group or any other

group due to their unique species composition. These three sites are excluded from the

construction of Species group parameter sets as no suitable independent field data is

available for model testing in these cases.

Soil type and species composition can be combined to produce more specific Soil x Species

groups. Seven combinations are possible in this study:

• barley Mitchell grass on basalt cracking clay (Sites 3, 5, 6 and 18);

• ribbon grass on alluvial cracking clay (Sites 7, 8, 9, 10 and 16);

• white grass on structured red earth (Sites 14 and 19);

• white grass on alluvial cracking clay (Site 20);

• annual short grasses on structured red earth (Sites 1, 2 and 13);

• forbs on alluvial cracking clay (Sites 15 and 21); and

• forbs on basalt cracking clay (Site 4).

Again, Sites 11, 12 and 17 are not included due to their unique species composition.

In addition to the Soil, Species, and Soil x Species groups, aggregation of all 21 sites into a

single group (Regional VRD) provides representation for the whole region, irrespective of

soil type or pasture species composition.

5.2.2 Developing generic parameter sets for common land types

For each group of sites identified in Section 5.2.1, datasets of parameter values are

assembled. For example, parameters from both study years at Site 1 are grouped with those


174

of Sites 2, 11, 12, 13, 14 and 19 to form the ‘structured red earth’ Soil group. Site 1

parameters are also grouped with those of Sites 2 and 13 to form the ‘annual short grasses’

Species group. All twenty one sites are combined to form the Regional VRD group. Soil x

species parameter sets are constructed based on the premise that generic soil parameters

(Table 4.1) are best determined from sites of similar soil type, and generic pasture

parameters (Table 4.2 to Table 4.6) are best determined from sites of similar species

composition. Using Site 1 as the example, soil parameters are grouped with those of Sites

2, 11, 12, 13, 14 and 19, and pasture parameters are grouped with those of Sites 2 and 13.

Once the datasets have been assembled, all values for each individual parameter require

summarising into a single value for the group. The approach taken in this thesis to

calculate the average across each year and each site in the dataset to produce the single

group value. The result is a single set of parameter values that can be used by GRASP to

simulate pasture growth at any location where that land type occurs. Table 5.2 presents the

generic parameter values for three approaches to describing land types: i.e. Soil, Species,

and Regional VRD. The Soil x Species parameter sets were constructed by combining the

generic soil parameters from the relevant Soil group with the plant growth, nitrogen, sward

structure and detachment parameters from the relevant Species group.

For some parameters, it is not appropriate to calculate a single generic value by averaging

results from all years at all sites in the group. For some nitrogen parameters, full

expression of the characteristic may not occur each year (e.g. growth ceases before all

available N is taken up, Day et al. 1997a). Such parameters are more suitably determined

from a single year at each site. The highest value over the two measurement years at each

is selected for maximum nitrogen uptake (MaxN) and maximum N content (Max%N), and

then site values are averaged to determine the generic value. The lowest value for N

content at which growth stops (%N0Grow), N content at which growth is restricted

(%Nrestrict), and minimum N content in dead plant material (Min%Ndead) is selected,

then site values are averaged.


175

Table 5.2 Generic parameter values for Soil, Species, and Regional VRD groups. (Table continued overleaf)

Parameter Parameter No. in

GRASP

Regional VRD


Alluvial cracking

clay

Basalt cracking

clay

Barley Mitchell grass

Ribbon grass

White Grass

Annual short

grasses

Forbs

Soil parameters

Depth of layer 3 (mm) 022 405 284 500 383 375 500 367 333 467

Air dry layer 1 (mm) 019 4.1 3.4 4.4 4.7 4.8 4.8 3.8 3.3 3.8

Wilting point layer 1* (mm) 029 9.5 7.8 10.4 10.0 10.0 11.2 8.3 8.3 9.2

Field capacity layer 1 (mm) 026 41.1 32.5 43.8 47.5 47.5 46.0 31.7 35.0 43.3

Wilting point layer 2 (mm) 030 40.8 35.6 43.8 43.3 45.0 48.0 32.3 38.7 35.0


Wilting point layer 3 (mm) 031 56.7 24.4 71.9 67.5 71.3 76.0 40.7 26.7 63.3


Runoff (Yes/No) 270 No Yes No No No No Yes Yes No

Cracking (Yes/No) 035 Yes No Yes Yes Yes Yes No No Yes

WHC Layer 1 (mm) 31.6 24.7 33.3 37.5 37.5 34.8 23.3 26.7 34.2

WHC Layer 2 (mm) 106.6 76.6 108.8 136.7 145.0 114.0 87.7 88.0 111.7

WHC Layer 3 (mm) 95.9 45.8 115.6 116.7 115.0 124.0 61.0 55.0 113.3

Total WHC (mm) 234.0 147.1 257.7 290.8 297.5 272.8 172.0 169.7 159.2

Average WHC (mm/10cm) 25.9 18.8 25.8 33.0 34.1 27.3 20.0 21.1 26.9

Plant growth parameters

Perennial grass basal area (%) 005 2.57 2.79 2.48 2.44 2.87 2.96 1.73 3.761 1.52


(kg/ha/day/basal%) 006 8.85 9.26 7.88 9.67 8.25 7.40 7.67 10.451 10.00

Potential regrowth rate (kg/ha/day) 22.7 25.8 19.5 23.6 23.7 21.9 13.3 39.31 15.2

Transpiration-use-efficiency (kg/ha/mmT) 007 10.9 10.6 10.4 11.7 11.0 10.5 10.8 10.3 11.7

Soil water index at which growth stops 149 0.16 0.17 0.26 0.01 0.01 0.30 0.30 0.01 0.11

Soil water index at which cover is restricted 009 0.96 0.88 1.00 1.00 1.00 1.00 1.00 0.72 1.00


176

Table 5.2 (cont.)

Parameter Parameter No. in

GRASP

Regional VRD


Alluvial cracking

clay

Basalt cracking

clay

Barley Mitchell grass

Ribbon grass

White Grass

Annual short

grasses

Forbs

Nitrogen parameters

N uptake at 0 mm of transpiration (kg/ha) 097 3.4 2.6 4.9 2.3 2.8 4.9 1.8 2.3 5.0


(kg/ha/100mmT) 098 10.5 9.6 9.9 12.5 11.3 10.2 7.7 9.3 11.3

Maximum nitrogen uptake (kg/ha) 099 24.1 18.4 26.8 27.1 27.9 27.6 14.0 22.3 31.0

Maximum nitrogen content in plants (%) 100 2.64 2.79 2.50 2.70 2.75 2.50 2.50 3.17 2.50

N content at which growth stops (%) 101 0.72 0.65 0.78 0.73 0.64 0.73 0.47 0.75 1.10

N content at which growth is restricted (%) 102 0.82 0.75 0.88 0.83 0.74 0.83 0.57 0.85 1.20

Minimum N content in dead (%) 111 0.53 0.59 0.48 0.53 0.48 0.46 0.40 0.77 0.70

Sward structure parameters

Green standing dry matter at 50% green cover

(kg/ha) 045, 046, 271 1529 1539 1469 1596 1669 1520 1767 1100 1225

Height of 1000kg/ha standing DM (cm) 096 12.5 13.9 11.7 12.1 12.5 11.6 17.5 11.3 9.5

Detachment parameters

Leaf detachment (kg/kg/day) 128, 130 0.0023 0.0018 0.0028 0.0022 0.0015 0.0028 0.0020 0.0022 0.0035

Stem detachment (kg/kg/day) 129, 131 0.0023 0.0018 0.0028 0.0022 0.0015 0.0028 0.0020 0.0022 0.0035

1 Represents annual grass basal area and related values, not perennial grass basal area.


177

5.3 Testing model performance using independent data

5.3.1 Use of data for both model fitting and model testing

Having now developed generic parameter sets that enable extrapolation of GRASP to new

times and locations in the VRD, some indication of model performance when using these

parameter sets to predict pasture growth is necessary: i.e. the model requires testing

(validation) with independent data. Considerable literature is dedicated to the issue of

systems model validation, with many opposing or alternative views expressed (e.g.

Harrison 1990; Mayer et al. 1994; Mitchell 1997). One common point throughout the

literature is that no universally accepted method of model validation is available, as the

process is dependent upon the eventual use of the model. Like the calibration procedure,

validation is usually based on comparison of predictions with field observations.

A common approach to validation is using separate datasets for the model fitting and

model testing stages, as demonstrated by Carter et al. (1996b) in the validation of a spatial

version of GRASP. However, Jones and Carberry (1994) point out that a major constraint

in independent model validation is the scarcity of appropriate data. This is the case for the

VRD where published figures of native pasture production in the VRD are sparse. Foran et

al. (1985), McIvor et al. (1995a), MacDonald et al. (1997), Dyer et al. (2001a) and Bastin

et al. (2003) report local studies that include some pasture production component, but these

data have limitations for model validation in this study. Their studies mainly report end of

growing season pasture yields, often under grazed conditions. Here, GRASP is being

evaluated for its ability to replicate the whole pasture growth cycle, not just peak pasture

yield. For this reason, appropriate validation data would include observations throughout

the annual growth cycle in the absence of grazing. Consequently, no separate dataset is

available for validation of GRASP for common land types in the VRD. This means we are

unable to directly test model performance when using the generic parameter sets developed

in Section 5.2.2 (Table 5.2). So how can the reliability of these predictions be assessed?

Jones and Carberry (1994) suggest an approach, based upon the work of Efron and Gong

(1983), where the same data can be used in both the estimation of model parameters

(calibration) and the testing of model predictions (validation). The basis of their approach

is resampling from a complete dataset where parameter values are derived from one subset

of the available data, and the resulting model predictions compared with the field


178

measurements sourced from another subset; a procedure known as ‘jackknifing’.

Jackknifing, and the related variance measure, ‘cross-validation’, have been used as model

testing procedures in a number of scientific disciplines including viticulture, (Lu et al.

2003), water quality research (Tibby and Reid 2004), soil science (Viscarra Rossel and

McBratney 2001), and wildlife and marine ecology (Kalish et al. 1996; Kirsch et al. 1997).

The resampling approach allows model performance to be assessed in a way that is akin to

directly testing the generic parameter sets in Table 5.2 if suitable external validation data

was available. This thesis undertakes model testing using the jackknifing technique as a

guide and assumes the results produced are the best indication available of the capacity of

GRASP to predict pasture growth across the VRD.

5.3.2 Assembling data for model testing

The first stage of jackknifing involves omitting the parameter sets of one site in a group

dataset, and producing a single parameter set by averaging values from the remaining sites.

The result is known as a ‘partial estimate’ parameter set. For example, Site 1 parameters

are removed from the annual short grasses group dataset and the partial estimate parameter

set is constructed by averaging the individual parameter values of Sites 2 and 13. As no

parameters from Site 1 contribute to the partial estimate parameter set, the field data from

Site 1 is independent from the partial estimate parameters and can be used as validation

data.

GRASP is then run using the partial estimate parameter set with the climate data of the

omitted site (in this example, Site 1). Outputs then represent the models prediction of

pasture growth at Site 1 during the study period. Predictions of TSDM are paired with the

corresponding independent observations from Site 1 and these pairs form the basis of

assessing the models performance.

The next stage of jackknifing is to omit the parameter sets of a different site from the group

dataset, calculate a new partial estimate parameter set, run GRASP with this dataset and

the climate of the omitted site, and assemble pairs of observations and predictions for

evaluation. For example, Site 2 parameters are removed from the annual short grasses

group dataset, partial estimates calculated from Sites 1 and 13, and GRASP is run with this


179

new partial estimate parameter set and the climate for Site 2. Resulting TSDM predictions

are paired with the observed data from Site 2 for analysis.

The process of removing individual site parameters, calculating partial estimates from the

remaining sites, and running GRASP with the partial estimates to produce data pairs of

model predictions and independent observations is repeated for all groups. Figure 5.2

illustrates this procedure using the annual short grasses group as an example.

Table 5.3 shows for each site the related sites from which partial estimate parameters are

calculated for Soil, Species, and Regional VRD groups. Partial estimates for the Soil x

Species approach are calculated from the relevant sites in the Soil and Species groups.

Completion of the partial estimates process results in pairs of observations and predictions

for analysis. The aim of this study is to evaluate the overall performance of GRASP in

simulating pasture growth, therefore results from all groups within a classification are

analysed together (e.g. structured red earth, alluvial clay and basalt clay pairs are analysed

collectively as Soils).

An issue arising when validating GRASP is initial values for the main soil and pasture

variables at the beginning of the simulation period. All field data was collected from a

position of zero initial biomass, so no account of carryover pasture is necessary. However,

soil moisture requires some consideration. The climatic regime of the VRD (Section 2.1

and Section 3.4.1) typically sees a long period of warm temperatures and virtually no

rainfall from May to October. During this time, the soil surface usually dries below wilting

point (air dry), and the remainder of the profile dries to wilting point. For validation in this

study it is assumed that starting soil moisture equals air dry at the surface (Layer 1, 0-

10cm) and wilting point for the remainder of the soil profile in all soil types. Examination

of daily rainfall records prior to the beginning of the field study reveals no falls of

significance at any of the study sites, supporting this assumption.


180

Figure 5.2 Illustration of the procedure for assembling data for the independent validation of GRASP, using

the annual short grasses group as an example. S01Y1 refers to Site 1, Year 1; S01Y2 refers to Site 1, Year 2;

and so on.

S01Y1 S01Y2 S02Y1 S02Y2 S13Y1 S13Y2

Mean of individual parameter values = partial estimates

Run GRASP with partial estimates parameter set and Site 1 climate

Output of TSDM

Site 1 parameters not used

Field observations of TSDM for Site 1

Parameter values in Table 4.1 to Table 4.6

Data pairs for analysis Data pairs for analysis Data pairs for analysis

Comparisons of observed and predicted values for TSDM (Section 5.3.3)

S01Y1 S01Y2 S02Y1 S02Y2 S13Y1 S13Y2




Output of TSDM


S01Y1 S01Y2 S02Y1 S02Y2 S13Y1 S13Y2



Output of TSDM




181

Table 5.3 Source of parameters used to calculate partial estimates for independent model validation. This

table is based upon the matrix presented in Table 5.1.

Site Soil groups Species groups Regional VRD group

soil type related soil sites pasture type related pasture sites related sites

1 red earth 2, 11, 12, 13, 14, 19 annual grasses 2, 13 2 - 21

2 red earth 1, 11, 12, 13, 14,19 annual grasses 1, 13 1, 3 - 21

3 basalt clay 4, 5, 6, 17, 18 Mitchell grass 5, 6, 18 1 – 2, 4 - 21

4 basalt clay 3, 5, 6, 17, 18 forbs 15, 21 1 – 3, 5 - 21



7 alluvial clay 8, 9, 10, 15, 16, 20, 21 ribbon grass 8, 9, 10, 16 1 – 6, 8 - 21

8 alluvial clay 7, 9, 10, 15, 16, 20, 21 ribbon grass 7, 9, 10, 16 1 – 7, 9 - 21

9 alluvial clay 7, 8, 10, 15, 16, 20, 21 ribbon grass 7, 8, 10, 16 1 – 8, 10 – 21


11 red earth 1, 2, 12, 13, 14, 19 other 1 No related sites 1 – 10, 12 – 21

12 red earth 1, 2, 11, 13, 14, 19 other 1 No related sites 1 – 11, 13 – 21

13 red earth 1, 2, 11, 12, 14, 19 annual grasses 1, 2 1 – 12, 14 – 21

14 red earth 1, 2, 11, 12, 13, 19 white grass 19, 20 1 – 13, 15 – 21

15 alluvial clay 7, 8, 9, 10, 16, 20, 21 forbs 4, 21 1 – 14, 16 – 21


17 basalt clay 3, 4, 5, 6, 18 other 1 No related sites 1 – 16, 18 – 21

18 basalt clay 3, 4, 5, 6, 17 Mitchell grass 3, 5, 6 1 – 17, 19 – 21

19 red earth 1, 2, 11, 12, 13, 14 white grass 14, 20 1 – 18, 20 – 21

20 alluvial clay 7, 8, 9, 10, 15, 16, 21 white grass 14, 19 1 – 19, 21

21 alluvial clay 7, 8, 9, 10, 15, 16, 20 forbs 4, 15 1 - 20

1 These sites were excluded from Species group validation due to their unique species composition.

5.3.3 Results of model testing (validation)

The performance measures employed during model calibration (i.e. regression analysis and

prediction variance) are also used for assessing the results of the independent validation

procedure. Both the model and the field data used here are the same as during calibration,

so comment is not provided on limitations of the model structure or on specific instances

where predictions do not correspond with the field data. These have been covered in the

previous chapter. The focus of this chapter is the effect of using generic parameter values

rather than site-specific values when using GRASP to predict pasture growth in the VRD.


182

All measures of model performance show much larger prediction errors for TSDM

throughout the pasture growth cycle when using generic parameter sets compared with

results for site-specific calibration in Chapter 4. The loss of predictive skill during model

validation is well illustrated by the scatter of data points about the one-is-to-one lines in

Figure 5.3.

Table 5.4 contains results of statistical analyses of data presented in Figure 5.3. For end of

wet season TSDM, r2(1:1) values range from –0.265 (Regional VRD) to –0.535 (Soils) and

RMSE varies from 807kg/ha (Regional VRD) to 912kg/ha (Soil x Species) compared to

132kg/ha during calibration (Table 4.7), indicating much poorer performance of GRASP

when using generic parameters to predict pasture yield at this time. For early wet season

TSDM, r2(1:1) is greater than zero, suggesting better model performance. However, the

mean observation value for early wet season TSDM was 561kg/ha and RMSE = 395kg/ha

(Soil groups) to 417kg/ha (Soil x Species groups) compared to 273kg/ha at calibration,

again indicating relatively large prediction errors. Predictions of mid wet season and end of

dry season TSDM show similar trends.

Other performance measures also show reduced model performance when predicting

independent field data. Results of deviation of predictions from observed values are shown

in Figure 5.4 (envelope of acceptable precision = ±35% of the observation value), and a

summary of these figures appears in Table 5.5. Only 47% to 64% of predictions of end of

wet season growth were within the 95% confidence limits of the independent observations

compared with 98% of predictions during calibration (Table 4.8). Predictions of TSDM at

other times of the pasture growth cycle show similarly reduced levels of model

performance when compared with using site-specific parameter values. Overall, however,

no trend to under- or over-prediction occurs and the mean of all predictions and the mean

of all observations at each of the growth phases are very similar (Table 5.4).


183

Figure 5.3 Observed vs. predicted TSDM for independent model validation using four approaches to

developing generic parameter sets.

0

1000

2000

3000

4000

5000

0 1000 2000 3000 4000 5000


Obs

erve

d TS

DM

(kg/

ha)


1:1n=163 Soil groups

0

1000

2000

3000

4000

5000

0 1000 2000 3000 4000 5000


Obs

erve

d TS

DM

(kg/

ha)


1:1n=139 Species groups

0

1000

2000

3000

4000

5000

0 1000 2000 3000 4000 5000


Obs

erve

d TS

DM

(kg/

ha)


1:1n=139Soil x Species

0

1000

2000

3000

4000

5000

0 1000 2000 3000 4000 5000


Obs

erve

d TS

DM

(kg/

ha)


1:1n=163 Regional VRD


184


dry matter (TSDM) during independent validation. Predictions are generated by GRASP using four

approaches to developing generic group parameter sets (Table 5.2). Dashed lines indicate the envelope of

acceptable precision, equal to the average magnitude of measurement variance (±35% of observation values).

-2500

-1500

-500

500

1500

2500

0 1000 2000 3000 4000 5000


Dev

iatio

n of

pre

dict

ion

from

ob

serv

ed T

SDM

(kg/

ha)


n =

Soil x Species

-2500

-1500

-500

500

1500

2500

0 1000 2000 3000 4000 5000


Dev

iatio

n of

pre

dict

ion

from

ob

serv

ed T

SDM

(kg/

ha)


n =

Regional VRD

-2500

-1500

-500

500

1500

2500

0 1000 2000 3000 4000 5000


Dev

iatio

n of

pre

dict

ion

from

ob

serv

ed T

SDM

(kg/

ha)


n =

Soil groups

-2500

-1500

-500

500

1500

2500

0 1000 2000 3000 4000 5000


Dev

iatio

n of

pre

dict

ion

from

ob

serv

ed T

SDM

(kg/

ha)


n =

Species groups


185

Table 5.4 Results of comparing model predictions with observed values of TSDM for independent validation using four approaches to developing generic parameter sets.

Soils Species Soil x Species Regional VRD

n Mean of obs (kg/ha)

Mean of pred (kg/ha)











Early wet 42 561 634 0.139 395 36 562 609 0.135 414 36 562 618 0.123 417 42 561 658 0.102 403

Mid wet 42 1896 1799 -0.735 842 36 1899 1752 -1.092 954 36 1899 1769 -1.124 961 42 1896 1876 -0.615 812

End of wet 42 2513 2646 -0.535 890 36 2527 2505 -0.434 908 36 2527 2498 -0.447 912 42 2513 2706 -0.265 807

End of dry 37 1902 1972 -0.212 667 31 1870 1938 -0.395 727 31 1870 1917 -0.287 698 37 1902 1997 -0.108 638

Note: obs = observed values, pred = predicted values. Negative values for r2(1:1) indicate poor model performance (Section 4.5.1).

Table 5.5 Summary of predictions of TSDM that fell outside the envelopes of acceptable precision (95% confidence limits of the corresponding observation) using four

approaches to developing generic parameter sets.

Soils Species Soil x Species Regional VRD

n Under-prediction

Over-prediction

Total outside

envelope

n Under-prediction

Over-prediction

Total outside

envelope

n Under-prediction

Over-prediction

Total outside

envelope

n Under-prediction

Over-prediction

Total outside

envelope

Early wet 42 6 17 23 (55%) 36 6 13 19 (53%) 36 6 13 19 (53%) 42 6 16 22 (52%)

Mid wet 42 12 11 23 (55%) 36 13 12 25 (69%) 36 13 13 26 (72%) 42 11 11 22 (52%)

End of wet 42 7 10 17 (40%) 36 9 10 19 (53%) 36 9 9 18 (50%) 42 7 8 15 (36%)

End of dry 37 3 12 15 (41%) 31 5 9 14 (45%) 31 6 9 15 (48%) 37 4 13 17 (46%)

Total 163 28 (17%) 50 (31%) 78 (48%) 139 33 (23%) 44 (32%) 77 (55%) 139 34 (24%) 44 (32%) 78 (56%) 163 28 (17%) 48 (29%) 76 (47%)


186

5.4 Discussion of independent validation of GRASP

Ability to independently predict pasture growth

Considerable loss of predictive skill occurs when extrapolating GRASP to times and

locations where no field data contributes to the input parameter values. A summary of

prediction results for end of wet season TSDM is presented in Table 5.6, including data

from the calibration phase (Chapter 4) for comparative purposes. Results show that the

three measures employed in this study all indicate poor model performance when using

generic parameters. The negative values for r2(1:1) imply that the mean of observations

from the field data is a better predictor of pasture growth at other locations than using the

model (Section 4.5.1 in Chapter 4). Similarly, RMSE values are 30 to 37% of the

observation mean, and only about half of predictions fall within measurement variance;

each measure signifying large departures of model predictions from the observations.

These results all indicate that using generic parameters to predict pasture growth in the

VRD under the conditions tested is not reliable.

The notion that the mean of observations is a better predictor of pasture growth at other

locations (and by extension, in other seasons) is obviously inappropriate. Variation in

pasture growth as a response to prevailing seasonal conditions is a feature of the semi-arid

tropics, and GRASP has previously performed well at accounting for this variation (Day et

al. 1997a). Thus, the results of model testing in this thesis reflect the limitation of the

jackknifing technique for providing suitable validation data when the original dataset does

not contain data representative of the range of seasons over which the model is likely to be

applied. Because all the data used in both calibration of parameters and testing of model

performance was collected during median to above median rainfall seasons, most end of

wet season TSDM values fall in the 2000 to 4000kg/ha range. When calculating

performance measures such as regression statistics for a dataset that is clustered in a

relatively small range, poor fits can often result. If additional data from below-median

seasons was available for model testing so that the TSDM range was, for example, 500 to

4000kg/ha, then results of measures such as regression analysis would likely show better

model performance. Lack of data from below median seasons is one limitation of the

model testing results presented in this chapter.


187

One measure that does provide confidence in the model’s performance is comparison of

the overall prediction mean with the observation mean. For each approach to developing

generic parameters, the mean of all predictions and the mean of all observations for end of

wet season TSDM are very similar. This suggests that while individual predictions often

deviate considerably from observations, when summarised, under- and over-predictions

effectively negate each other to produce acceptable results. Therefore, where an average or

summary of many predictions is required, rather than individual predictions, the model can

be used with some confidence.

In other studies, performance of GRASP has generally been better than here. Johnston

(1996) reports acceptable prediction of TSDM when using independent grazing trial data to

validate GRASP in the mulga zone of south west Queensland (slope of line-of-best-fit not

significantly different (nsd) to 1.0 and intercept nsd to 0.0, r2 = 0.15 to 0.91). Day et al.

(1997a) found that using the model with a constant set of parameters accounted for a high

proportion of the variation of both within year and between year variation in pasture yield

in south east Queensland (r = 0.849 to 0.973, mean error = 170 to 457kg/ha). Richards et

al. (2001) reports good correlation between predicted pasture yield and independent

observed data (r2 = 0.717 to 0.852) using GRASP with a constant parameter set in central

Australia, although it is important to note that in this case the parameter set and field

observations were not fully independent. Day et al. (2003) successfully calculated pasture

growth over 16 years in open country in Zimbabwe using independent data (RMS(E) =

724kg/ha which was considered reasonable).

Less than complete agreement between predicted and observed variables indicates

limitation in the model’s capacity to represent the real system. Such limitation is due to: 1)

deficiency in the structure of the model preventing it from sufficiently representing all

influential processes or elements in the system; 2) errors associated with the data used to

represent the real system; or 3) errors associated with parameter values used in the

simulation (Grant et al. 1997). In this study, the errors associated with model structure and

with field data are the same for validation as for calibration. Therefore, loss of predictive

skill compared to calibration is a result of errors in parameter values alone. These errors

arise during the averaging of individual site parameter values into generic parameter sets. It

is apparent that using generic values can sometimes lead to large prediction errors.


188

It is known that some parameters have a greater effect on predictions of pasture growth

than others (Littleboy and McKeon 1997), with parameters affecting evapotranspiration

being most influential (Day et al. 1997a) and those affecting nutrient supply important at

times (Day et al. 2003). Quantifying the effect of these important parameters on model

predictions would improve our understanding of the factors affecting pasture growth in the

VRD and provide guidance to model users of the parameters needing most consideration

when simulating pasture growth in the VRD.

Approaches to grouping sites into common land types

Four approaches to grouping study sites were explored to determine the most appropriate

way of differentiating land types for future modelling studies. It was found that each

approach to developing the generic parameters resulted in similar predictive skill. This was

not expected as it is reasonable to assume that deriving generic parameters from sites with

similar features would lead to improved predictions over using generic parameters derived

from all sites regardless of land type. These results reinforce the earlier finding that

attempting to use generic values for influential parameters instead of site-specific values

causes direct loss of predictive skill.

Table 5.6 Summary of results of independent validation of GRASP (including comparative results from

calibration in Chapter 4) for end of wet season total standing dry matter.

n Mean of observations

Mean of predictions

(kg/ha)


Proportion of predictions within 95%

conf limits of observations

(%)

Soils 42 2513 2646 -0.535 890 60

Species 36 2527 2505 -0.434 908 47

Soil x Species 36 2527 2498 -0.447 912 50

Regional VRD 42 2513 2706 -0.265 807 64

Calibration 42 2513 2553 0.966 132 98

Implications for applying the model

The purpose of developing generic parameter sets and testing the performance of the

model against independent data was to determine the capability of GRASP for predicting

pasture growth on the commonly grazed land types in the Victoria River District. There are


189

many applications of a well-tested model such as GRASP to grazing management (Section

2.6), often differing in terms of time scale and spatial resolution. Local producer

experience and scientific research both recognise that individual land types have different

capacities for pasture and livestock production (Smith 1998; Hasker 2000). Therefore,

determining year to year variability in pasture growth, and deriving safe utilisation levels

requires separate simulations for each land type if differences in productivity between land

types at the same location are to be expressed. These applications also require simulations

to be run over many years. Applications such as incorporating seasonal climate forecasting

into tactical stocking rate decisions often use predictions of pasture growth from much

shorter periods (up to 12 months) and may use either land type or regional average

parameters.

5.5 Conclusions

Validation of the GRASP model using generic parameter sets and independent data

resulted in a considerable loss of predictive skill compared to using site-specific

parameters derived from field measurements. The loss of predictive skill occurs at all

stages of the pasture growth cycle and is similar for all approaches used to develop generic

parameter sets. However, any tendency of the model to under- or over-predict is negated

and the average of all predictions is close to the average of all independent observations

throughout the growing season.

These results have significant implications for the use of GRASP to independently predict

pasture growth in the VRD. Where the desired outcome of simulation studies is a relative

difference in pasture growth (e.g. the effect of a change in climate), generic parameter sets

can be very useful. Using generic parameters is questionable, however, if a high degree of

prediction accuracy is required (e.g. predicting the amount of pasture present on a specific

land type in the current season). In these circumstances, determining location-specific

values for the parameters most influential on pasture growth would improve the

performance of GRASP. Identifying the parameters most influential on predictions of

pasture growth, and quantifying their effect on model behaviour is therefore an important

step in developing GRASP for application to grazing management in the VRD. This step is

the focus of the next chapter.

Chapter 6 Determining parameters most influential on predictions of long-term pasture growth

190

6.0 Determining parameters most influential on predictions of long-term pasture growth

6.1 Introduction

The previous chapter concluded that using generic parameters to simulate pasture growth

in the VRD sometimes leads to considerable prediction errors. These errors arise because

some parameters have greater influence on model behaviour than others (McKeon et al.

1990), and the most influential parameters require site-specific (rather than generic) values

for best model performance. Quantifying the impact that specific parameters have on

predictions of pasture growth in the VRD provides guidance to future model users of the

limitations of using generic parameter values when undertaking simulation studies, and

indicates where effort should be placed when measuring site-specific parameters.

Determining the influential parameters by comparing changes in model output caused by

specific changes in parameter values is known as sensitivity analysis (Rickert et al. 2000).

This chapter aims to determine, through sensitivity analysis, the most influential model

parameters on predictions of pasture growth in the VRD, and quantify the relative effect of

changes in their value.

To achieve this aim, sensitivity analysis is undertaken in two steps: 1) factors most

influential on year-to-year variability in pasture growth are first identified (Section 6.2);

and 2) the effects of varying parameter values describing these factors are then quantified

(Section 6.3). Trees are also an important factor in pasture growth. So far in this thesis, no

account of the competition for water and nutrient between trees and pasture has been

included in simulations. Therefore, Section 6.4 examines the impact of trees on predictions

of pasture growth using GRASP. Results of predicting year-to-year variability in pasture

growth, sensitivity analysis of important parameters, and the impact of trees on pasture

growth are discussed in Section 6.5, and conclusions drawn in Section 6.6. Figure 6.1

illustrates the structure of this chapter.


191


6.2 Modelling year-to-year variability in pasture growth across a climate

gradient

So far in this study, pasture growth has only been simulated during the two growing

seasons corresponding to the field measurements reported in Chapter 3. For GRASP to be

applied to the management of grazing land, simulation over much longer time periods is

often required. GRASP has the capacity to calculate pasture growth for any length of time

where daily climate records are available. Daily climate data are available for any location

in the country since 1889. Up to 1956, data for variables other than rainfall are averages

and therefore do not vary from year to year, although some real data back to 1900 has

recently been incorporated. Since 1957 however, all climate variables are derived from

actual recordings and reflect prevailing conditions at the time. This data is available from

the combined QDNR/BoM Internet resource, the Data Drill, and consists of interpolated

meteorological data (DataDrill 2005). In the VRD, climate recording stations are sparse

and it is expected that, in some instances, the interpolated data for a remote location may

differ to some extent from actual conditions. Nevertheless, the Data Drill provides the best

climate data available.

6.2.1 Method

The Regional VRD parameter set developed in Section 5.2.2 is used to simulate pasture

growth at three locations (Auvergne, Victoria River Downs and Inverway - Figure 1.1).

Sensitivity analysis to identify influential model parameters

(Section 6.3)

Running GRASP over the historical climate record

(Section 6.2)

Generic parameter sets from Chapter 5

(Table 5.2)

Incorporating the effect of trees into predictions of pasture growth

(Section 6.4)



192

These locations represent a climate gradient across the VRD with rainfall and humidity

decreasing, maximum temperatures increasing, and minimum temperatures decreasing as

distance from the northern coastline increases (DataDrill 2005). Parameter values used to

calculate growth remain constant across the three locations, and only climate data changes

between simulations. Therefore, any differences between simulation results are entirely

due to climatic factors.

Additionally, the five Species group parameter sets (Barley Mitchell grass, Ribbon grass,

White grass, Annual short grasses and Forbs) and three Soil group parameter sets

(Structured red earth, Alluvial cracking clays and Basalt cracking clays) are used to

simulate pasture growth at a single location (Victoria River Downs) to examine the effect

of predicting pasture growth using different parameters sets under the same climate

conditions.

Long-term simulations are run using daily climate data from the period 1 July 1957 to 30

June 2004. However, the results only show predictions from 1959/60 to 2003/04, a period

of 45 years. Starting conditions for soil and pasture variables are initially set at defaults

(soil water at wilting point, and no carryover of dry matter from the previous season),

which may differ from actual conditions at the time. In circumstances where soil water

content exceeds wilting point or considerable dry matter is present at the start of the

simulation period, prediction errors are likely to occur. These errors arise because the

presence of available soil water permits greater transpiration to occur in the field than is

predicted by the model and the presence of carryover dry matter can decrease soil

evaporation rates which also affects the soil water balance. The presence of dry matter can

also reduce surface runoff, but in the Regional VRD parameter set, runoff is deactivated as

the majority of soils contributing to these parameters are cracking clays on very gentle

slopes where runoff is assumed to be negligible. Once the simulation has run for two years,

the effect of any variation between initial conditions in the field and those used in the

simulation is negligible. Values for soil water and dry matter at the beginning of the

simulation period are not known, so simulation results from the first two growing seasons

(1957/58 and 1958/59) are discarded as they contain no ‘memory’ of past seasons.

Several other issues require consideration before undertaking long-term simulations.

GRASP requires cumulative transpiration (and therefore nitrogen uptake) to be reset


193

annually, thereby making N available for growth the following year. Littleboy and

McKeon (1997) suggest reset occur in September or October, but examination of long-

term rainfall records for the VRD showed several occasions where rainfall sufficient to

initiate and sustain new seasons pasture growth occurred earlier in the year. For this

reason, reset takes place on 1 July each year. Soil water is not reset during long-term

simulations. Another consideration is how best to simulate perennial grass basal area

(PGBA). In the field, PGBA is likely to change over time as a result of seasonal

conditions, and the impact of management practices such as fire and grazing pressure.

GRASP has the capacity to simulate dynamic basal area based on pasture growth in the

current and preceding seasons. Preliminary simulations showed that activating the dynamic

grass basal area routine resulted in predictions of PGBA that were unrealistic when

compared with measured data (predicted values were 7.0 to 9.0%, compared with field

measured values of 0.0 to 6.0% averaging 2.6%, Section 3.5.3). As a consequence,

predictions of seasonal pasture growth were generally greater when the dynamic basal area

routine was activated than when using a constant basal area. These results suggest the

current basal area routine requires modification to better replicate changes in PGBA from

one season to another in the environment of the VRD. For this reason, long-term

simulations were run with a constant PGBA (2.6%), a more realistic value providing

greater confidence in model predictions of seasonal pasture growth.

Running GRASP over many years with historical climate data provides predictions of

seasonal pasture growth. In this chapter, the term ‘simulated seasonal pasture growth’

(SSPG) is used to refer to the model output which is a close approximation to the ‘total

standing dry matter’ (TSDM) at the end of the growing season that has been presented in

previous chapters. As no further field data is presented in this thesis, the distinction in

terminologies helps clarify that results presented from here on are model output and not

field observations.

Results of long-term pasture growth simulations can be presented in many ways. Three

approaches are used in this chapter:

• time-series;

• probability distributions; and

• regressions.


194

Time-series are useful for visually assessing the variability of environmental factors

between seasons, while probability distributions provide a particularly useful means of

summarising this variability (Rickert et al. 2000). In this study, the main format for

presenting probability distributions is as ‘probabilities of exceedence’ (after Stone et al.

2000); that is, the chance of a particular amount of pasture growth being exceeded.

Probabilities are expressed as fractional values (e.g. a value that has a 90% chance of being

exceeded is expressed as 0.9). This form of presentation is the inverse of the commonly

presented ‘cumulative probability’ (e.g. Hill et al. 2006). ‘Percentiles’ are another common

expression in the literature. Median pasture growth is usually referred to as the 50th

percentile, and other values follow the usual convention where, for example, the amount of

growth in the lowest 10% of seasons is referred to as the 10th percentile. Results presented

in the following section follow these conventions.

While results of long-term seasonal pasture growth quantify year-to-year variability, they

do not provide insight to why this variability occurs. To understand which factors are

responsible for seasonal variability in pasture growth, comparing time-series and

regressions of the main environmental factors with predictions of pasture growth is

necessary. For this reason, predictions of seasonal pasture growth using the Regional VRD

parameter set are compared to rainfall (the main climatic determinant of plant growth),

pasture transpiration (a model output), and nitrogen uptake (also a model output but

dependent on field measured data).

6.2.2 Results of long-term simulations: limits to pasture growth

Simulated seasonal pasture growth (SSPG) from 1959/60 to 2003/04 at three locations is

presented in Figure 6.2. Two main features are apparent in Figure 6.2: 1) an upper limit to

SSPG is expressed in many seasons; and 2) considerable year-to-year variability occurs at

some locations. The upper limit to SSPG is reached in most seasons at Auvergne (41 of 45

years, or 91%), less frequently at Victoria River Downs (25 of 45, 56%) and on fewest

occasions at Inverway (12 of 45, 25%). Year-to-year variability is greatest at Inverway

with predictions of seasonal pasture growth ranging from 378 to 3345kg/ha/year. At

Victoria River Downs the range is 1784 to 3345kg/ha/year, and least variability occurs at

Auvergne (2566 to 3345kg/ha/year).


195

When the simulation results are presented as probability distributions, the impact of the

climate gradient can be clearly seen (Figure 6.3). The upper limit to SSPG is again

apparent and the climate regime dictates how often this limit is reached. Using

1500kg/ha/year as an example (the threshold for standing dry matter required to carry a

fire, Dyer et al. 2001b), Figure 6.3 shows that this amount of pasture growth (or more) will

occur every year at Auvergne and Victoria River Downs, and in 82% of years at Inverway.

Another way of interpreting the graph is to look at median growth (horizontal dashed line).

Median growth at Auvergne and Victoria River Downs is 3345kg/ha/year, while at

Inverway it is 2567kg/ha/year.

The landscape of the VRD is not homogenous, but instead consists of many land types that

differ in topography, soil texture, soil profile depth, drainage and pasture species

composition. These factors interact, resulting in differences in pasture production under the

same climate regime. This is demonstrated in Figure 6.4 where long-term production of

five pasture types and threes soil types at Victoria River Downs are presented. Using the

generic Species and Soil parameter sets developed in Chapter 5 (Table 5.2) results in very

different predictions of historical pasture growth at the same location. The magnitude of

the upper limit to pasture growth, and the frequency with which it is reached differs

between pasture and soil types. For example: SSPG for white grass pastures varies from

871 to 3000kg/ha/year, only reaching its upper limit in 18% of seasons (PE = 0.18); while

barley Mitchell grass SSPG ranges from 2740 to 4373kg/ha/year, reaching the upper limit

in 60% of seasons (PE = 0.60). SSPG for pastures on structured red earths ranges from

2234 to 2797kg/ha/year, upper limit reached in 74% of seasons (PE = 0.74); while basalt

clay pastures SSPG ranges from 2024 to 3712kg/ha/year, upper limit reached in 50% of

seasons (PE = 0.50). As the climate data is identical for each simulation, variation of

parameter values is responsible for the differences in predictions of long-term seasonal

pasture growth between these soil and pasture types.

Plotting rainfall as a time-series (Figure 6.5a) and as a regression with seasonal pasture

growth (Figure 6.5b) shows that some of the year-to-year variability in pasture growth

evident in Figure 6.2 is explained by variation in rainfall. During low rainfall seasons

(<500mm) rainfall and pasture growth are closely related. Considerable variability in

SSPG occurs once seasonal rainfall exceeds 500mm. The upper limit to pasture growth

(3345kg/ha/year) first becomes active once rainfall exceeds 600mm. The relationship


196

between seasonal rainfall and SSPG for all years when SSPG < 3345kg/ha/year is not

strong (y = 3.4x + 600, r2 = 0.550, n = 57). The absence of a distinct upper limit to rainfall

in Figure 6.5a and the scatter of points in Figure 6.5b suggests that variation in rainfall

amount only accounts for some of the year-to-year variation in pasture growth. Two

instances where rainfall and SSPG appear poorly related are examined below to help

clarify why seasonal rainfall alone is not a good predictor of pasture growth.

Two seasons where the relationship between rainfall and SSPG was very different from the

overall trend are 1986/87 at Auvergne (2733kg/ha pasture growth from 921mm of rainfall)

and 1992/93 at Inverway (2138kg/ha from 785mm). Simulation results were examined in

detail to determine the cause. In both cases, a large proportion of the total rainfall was

received in a short period (two months), and little further rain was received for the

remainder of the season. At Auvergne 771mm was received from 1 January to 28 February

1987 (84% of the seasonal total), and only a further 9mm was received to the end of June

that year. Model output showed that this resulted in the lowest ratio of pasture transpiration

to rainfall in 45 years (24%, compared to the median of 48%), the second highest ratio of

drainage to rainfall (41%, median = 16%) and the lowest ratio of transpiration to

evapotranspiration (41%, median = 59%). At Inverway 654mm was received from 1

January to 28 February 1993 (83% of the seasonal total) and only a further 19mm was

received to the end of June. This resulted in a low transpiration to rainfall ratio (24%,

median = 41%), the highest drainage to rainfall ratio (32%, median = 0%), and a low

transpiration to evapotranspiration ratio (35%, median = 44%). These results indicate that

in some seasons the partitioning of rainfall to the various water balance pathways leads to

variable relationships between rainfall and predicted pasture growth, and highlights the

importance of within-season rainfall distribution.

As pasture growth is more closely related to actual plant water use than to rainfall, and

given the close association between these two factors in the structure of GRASP, it is

expected that variation in pasture growth would be closely related to variation in

transpiration. Results in Figure 6.6 confirm this close relationship. The time-series of

transpiration more closely resembles that of pasture growth although, as with rainfall,

absence of a distinct upper limit to transpiration indicates that some other factor is limiting

seasonal pasture growth. Figure 6.6b, shows that the relationship between seasonal

transpiration and SSPG for years when SSPG < 3345kg/ha/year is very close (y = 9.0x +


197

535, r2 = 0.929, n = 57). The remaining unaccounted variation in seasonal pasture growth

when SSPG < 3345kg/ha/year is mostly due differences in vapour pressure deficit (which

affects transpiration use efficiency) from year to year (Littleboy and McKeon 1997). For

years when SSPG = 3345kg/ha/year, transpiration is a poor predictor of pasture growth.

It is immediately obvious from the strong similarity between time-series plots of nitrogen

uptake (Figure 6.7a) and pasture growth (Figure 6.2) that restricted N uptake is responsible

for the upper limit to seasonal pasture growth. As nitrogen uptake in GRASP is linked to

transpiration, it is expected that the close relationship in Figure 6.6b would also occur in a

regression between N uptake and SSPG. Figure 6.7b confirms this close relationship. The

difference, however, is the point at which the relationship breaks down. While

transpiration and SSPG form a linear regression until SSPG reaches 3345kg/ha/year, the

regression between N uptake and SSPG is only valid to 2206kg/ha/year, when N uptake

reaches 24kg/ha. Up to this point, the relationship is explained by the function y = 80x +

332, r2 = 0.922, n = 20. Average plant nitrogen content in standing pasture at the end of

the growing season during these years is 0.97%. After 2206kg/ha/year of SSPG, additional

pasture growth occurs without further N uptake. Therefore, additional pasture growth is

only possible by plants internally diluting nitrogen to continue growth. This dilution takes

place until SSPG reaches 3354kg/ha/year, the upper limit to pasture growth already

observed in the other regressions. At this point, average plant nitrogen content is 0.72%.

These two critical values, 24kg/ha of total N uptake and 0.72% plant N content are, in fact,

the values of two input parameters in the Regional VRD parameter set: maximum N uptake

(MaxN); and nitrogen content at which growth stops (%N0Grow). Of 135 individual

predictions of SSPG in Figure 6.7b, 82 (61%) reach 3345kg/ha/year SSPG and 24kg/ha of

N uptake. These results demonstrate very clearly that the amount of nitrogen available for

plant uptake, and the physiological limit to which plants can internally dilute N before

growth is no longer possible, are major influencing factors on GRASP’s prediction of

pasture growth in the VRD.


198

Figure 6.2 Time-series of simulated seasonal pasture growth (SSPG, July to June) over a 45-year period

(1959/60 to 2003/04) using the Regional VRD parameter set at three locations in the VRD.

Figure 6.3 Probability distribution of simulated seasonal pasture growth (SSPG, July to June) over a 45-year

period (1959/60 to 2003/04) using the Regional VRD parameter set at three locations in the VRD. The

horizontal dashed line represents the median (50th percentile) value.

Figure 6.4 Probability distributions of simulated seasonal pasture growth (SSPG, July to June) over a 45-

year period (1959/60 to 2003/04) using a) five Species group parameter sets; and b) three Soil group

parameter sets at Victoria River Downs. Horizontal dashed lines represents median (50th percentile) values.

0

1000

2000

3000

4000

5000

1959/60 1969/70 1979/80 1989/90 1999/2000

SSPG

(kg/

ha/y

r)

AuvergneVictoria River DownsInverway

0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000

SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce


0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000

SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

Red earths

Alluvial clays

Basalt clays0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000

SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

Annual short grassesForbsBarley Mitchell grassWhite grassRibbon grass

a. b.


199

Figure 6.5 a) Time-series of seasonal rainfall (July to June); and b) relationship between seasonal rainfall

and simulated seasonal pasture growth (SSPG) over a 45-year period (1959/60 to 2003/04) using Regional

VRD parameters at three locations in the VRD. The plotted regression is for seasons when SSPG was less

than the upper limit of 3345kg/ha.

Figure 6.6 a) Time-series of cumulative seasonal transpiration (July to June); and b) relationship between

seasonal transpiration and simulated seasonal pasture growth (SSPG) over a 45-year period (1959/60 to

2003/04) using Regional VRD parameters at three locations in the VRD. The plotted regression is for seasons

when SSPG was less than the upper limit of 3345kg/ha.

Figure 6.7 a) Time-series of nitrogen uptake (July to June); and b) relationship between N uptake and

simulated seasonal pasture growth (SSPG) over a 45-year period (1959/60 to 2003/04) using Regional VRD

parameters at three locations in the VRD. The plotted regression is for seasons when N uptake was less than

the upper limit of 24kg/ha/year.

0

250

500

750

1000

1250

1500

1750

2000

1959/60 1969/70 1979/80 1989/90 1999/2000

Seas

onal

rain

fall

(mm

) AuvergneVictoria River DownsInverway

y = 3.4x + 603r2 = 0.550

0

1000

2000

3000

4000

5000

0 250 500 750 1000 1250 1500

Seasonal rainfall (mm)

SSPG

(kg/

ha/y

r)


1986/87

1992/93

a. b.

0

200

400

600

800

1000

1959/60 1969/70 1979/80 1989/90 1999/2000

Seas

onal

tran

spira

tion

(mm

)


y = 9.0x + 535r2 = 0.929

0

1000

2000

3000

4000

5000

0 100 200 300 400 500 600 700 800

Seasonal transpiration (mm)

SSPG

(kg/

ha/y

r)


a. b.

0

5

10

15

20

25

30

35

1959/60 1969/70 1979/80 1989/90 1999/2000

N u

ptak

e (k

g/ha

/yr)


y = 80x + 332r2 = 0.922

0

1000

2000

3000

4000

5000

0 5 10 15 20 25 30

N uptake (kg/ha/yr)

SSPG

(kg/

ha/y

r)


82 datapoints at 24kg/ha of N and 3345kg/ha SSPG

a. b.


200

6.3 Sensitivity of model predictions to changes in value of influential

parameters

The review of literature and the results in the previous section both highlighted that plant

water use and nitrogen supply are important factors driving pasture growth in northern

Australia, and this is reflected in the structure of GRASP. It follows that parameters for

transpiration rate and plant nitrogen status will affect predictions of pasture growth.

However, the effect of variation in the values for these parameters on predictions of

pasture growth in the VRD is not known. This section quantifies those effects.

6.3.1 Method

Three transpiration parameters are examined for model response to changes in their value:

• Green yield at 50% green cover, identified during calibration as being influential on

pasture growth, particularly early in the wet season (Section 4.6.1);

• Transpiration use efficiency, the main parameter affecting growth rate (Littleboy and

McKeon 1997); and

• Soil water index at which growth stops, which defines the lower limit to soil water at

which transpiration ceases.

Three nitrogen parameters are also examined:

• N uptake per 100mm of transpiration (Nup100T), which is the link between nitrogen

supply and plant uptake via transpired soil water;

• Maximum N uptake (MaxN), shown to limit pasture growth even when transpiration is

still occurring (Figure 6.7b); and

• N content at which growth stops (%N0Grow), also identified as a factor in defining the

upper limit to pasture growth.

The Regional VRD parameter set and the climate records from Victoria River Downs were

used in the sensitivity analysis. Values of each parameter were sequentially changed while

all others remained constant, and the impact on probabilities of simulated seasonal pasture

growth (SSPG) examined. The values investigated for each parameter fall within the range

of values found in the individual site parameter sets in Chapter 4.


201

6.3.2 Results of sensitivity testing

Transpiration parameters

The main characteristic of Figure 6.8 is the effect of transpiration parameters on SSPG

during low SSPG seasons. By increasing values of TUE, GRASP simulates an increased

efficiency of pasture converting transpired soil water into dry matter, and greater seasonal

growth results from the same amount of transpiration. For example, a change in TUE from

9kg/ha/mmT to 15kg/ha/mmT results in SSPG increasing from 2433kg/ha/year to

3345kg/ha/year when probability of exceedence (PE) = 0.8.

However, increasing values for GY50GC has the opposite effect. Changing the value of

GY50GC from 500kg/ha to 2000kg/ha (e.g. from a prostrate broadleaf forb to an erect

narrow-leaved grass) reduces SSPG from 3247kg/ha/year to 2607kg/ha/year when PE =

0.8. This effect is due to erect species intercepting a lesser proportion of radiation than

prostrate species, thus slowing the growth rate early in the season. Some effect of increased

soil evaporation due to lower ground cover is also likely. Littleboy and McKeon (1997)

acknowledge that determining radiation interception from leaf area rather than a set

proportion of the green biomass pool is preferable, but little information is available on

leaf:stem ratios for native tropical pastures.

Increasing values of SWIX0Grow also reduces pasture growth in low rainfall years,

although the effect is less pronounced than for GY50GC. Changing SWIX0Grow from

0.01 to 0.1 made little difference to predictions of SSPG, while a further increase to 0.3

reduced growth from 2756kg/ha/year to 2523kg/ha/year when PE = 0.8. Increasing values

for this parameter has the effect of raising the soil water threshold at which pasture growth

ceases. This threshold does not prevent further plant water use, but it does prevent further

biomass from accumulating when the soil water index is less than the defined parameter

value. Therefore, the higher the value of this parameter, the lower the total pasture growth

resulting from transpiration. At the index values tested in this study, the effect was minor

as very little additional pasture growth occurred at SWIX0Grow = 0.01 compared with 0.1.

For all transpiration parameters, changes in value did not result in any change to the upper

limit of pasture growth, only the frequency with which it was reached.


202

Nitrogen parameters

The main feature of Figure 6.9 is the effect that two of the nitrogen parameters have on the

upper limit to pasture growth. When MaxN = 15kgN/ha the upper limit of SSPG is

2089kg/ha/year; and doubling MaxN to 30kgN/ha doubles the upper limit of SSPG to

4176kg/ha/year and reduces the probability of reaching this limit from 0.87 to 0.33. In

GRASP, growth occurs within the limits of nitrogen availability (Figure 4.3), and results in

Section 6.2.2 show that this limit is reached often in the VRD. Lifting the upper limit of

nitrogen availability leads to a direct increase in pasture growth during seasons not

constrained by soil water supply.

In contrast to above, increasing values for %N0Grow reduces the upper limit to pasture

growth. When %N0Grow = 0.5% the upper limit to SSPG is 4804kg/ha/yr; and increasing

%N0Grow to 1.0% halves the upper limit of SSPG to 2402kg/ha/yr, while increasing the

probability of reaching this limit from 0.2 to 0.71. Figure 6.7b showed that during low

rainfall years, predicted end of growing season plant N content is about 0.97%. However,

when all N available for uptake is exhausted, plants internally dilute N to continue growth.

Increasing values for %N0Grow effectively limits the amount of internal dilution that can

occur, and therefore reduces the additional pasture growth possible.

Probabilities of SSPG change little with variations in the rate of Nup100T indicating it is

not a major factor influencing seasonal pasture growth in the VRD. Some impact in lower

rainfall years is apparent when Nup100T falls below 11kg/ha/100mmT, with the number of

years when maximum SSPG is attained being reduced (Figure 6.9).While Nup100T does

not impact on the upper limit to SSPG, it is influential on how often that upper limit is

reached. Probabilities of seasonal pasture growth are not affected by changes in value of

Nup100T over 11kg/ha/100mmT when using the Regional VRD parameter set.

To summarise, the parameters MaxN and %N0Grow are highly influential on both the

magnitude of the upper limit to pasture growth and the frequency with which it is reached,

while Nup100T affects the frequency with which the upper limit is reached when values

are less than 11kg/ha/100mmT, but has no effect on the magnitude of the upper limit.

All sensitivity analysis results are tabulated in Table 9.14 in Appendix 9.


203

Figure 6.8 Sensitivity of simulated seasonal pasture growth results (SSPG) using a standard parameter set at

Victoria River Downs to changes in values of transpiration-use-efficiency, green yield at 50% green cover,

and soil water index at which growth stops. The horizontal dashed lines represent the median (50th percentile)

value.

Note: In the SWIX0Grow chart, simulation results are virtually identical at values of 0.01 and 0.1.

Consequently, the probabilities of exceedence lines almost coincide.

0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

GY50GC=500kg/ha

GY50GC=1000kg/ha

GY50GC=2000kg/ha

Green yield at 50% green cover(GY50GC)

0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

9kg/ha/mmT

12kg/ha/mmT

15kg/ha/mmT

Transpiration useefficiency (TUE)

0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

SWIX=0.01

SWIX=0.1

SWIX=0.3

Soil water index at which growth stops

(SWIX0Grow)


204

Figure 6.9 Sensitivity of simulated seasonal pasture growth (SSPG) using a standard parameter set at

Victoria River Downs to changes in values of maximum N uptake, N content at which growth stops, and N

uptake per 100mm of transpiration. The horizontal dashed lines represent the median (50th percentile) value.

Note: In the Nup100T chart, simulation results are virtually identical with uptake rates of 11kgN/ha/100mmT

and 15kgN/ha/100mmT. Consequently, the probability of exceedence lines coincide.

0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

7kgN/ha/100mmT

11kgN/ha/100mmT

15kgN/ha/100mmT

N uptake per 100mmtranspiration

(Nup100T)

0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

15kgN/ha20kgN/ha25kgN/ha30kgN/ha

Maximum N uptake (MaxN)

0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

0.5%N0.75%N1.0%N

N content at whichgrowth stops (%N0Grow)


205

6.4 The effect of trees on predictions of pasture growth

Trees are known to compete with pasture for water and nutrients in the semi-arid tropics of

the Northern Territory (Winter et al. 1989b; Cafe et al. 1999). Trees also intercept

radiation, but at the tree density levels found in the VRD the effect on pastures is likely to

be very small (Dupont et al. 1996). So far in this thesis, trees have been excluded from

simulations of pasture growth to simplify water balance calculations. However, trees are

present on most land types across northern Australia (Section 2.3.2 in Chapter 2) and

application of GRASP to grazing management often requires accounting for their effect on

pasture growth. This section quantifies the impact of the presence of trees on predictions of

pasture growth in two different climates in the VRD.

6.4.1 Method

The Regional VRD parameter set and the climate files from Auvergne (mean annual

rainfall = 900mm/year) and Inverway (577mm/year) were used to determine the impact of

trees on seasonal pasture growth. In GRASP, tree density is represented by the parameter

Tree basal area (TBA). Tree density varies across the VRD and values for Auvergne

(TBA = 6.0m2/ha) and Inverway (TBA = 2.0m2/ha) were estimated from Stewart et al.

(1970), Wilson et al. (1990) and Carter et al. (1996a). It was assumed that soil water

wilting points for trees are the same as for pastures, and that trees take up nitrogen via soil

water at the same rate as pastures (i.e. Nup100T = 10.5kg/ha/100mmT for both vegetation

groups). Tree water use was drawn from the same soil profile as pastures.

Some data is available for end of growing season pasture yield under trees from 1982 to

1989 at Manbulloo, 40km west of Katherine (McIvor et al. 1994). Even though cattle were

present during the study, the data was collected from areas that experienced such little

grazing as to be considered negligible. GRASP was used to predict seasonal pasture

growth at the same location over the same time period. The simulation was based on the

Regional VRD parameter set, Katherine climate data (mean annual rainfall = 970mm), and

a tree basal area (TBA) of 7.0m2/ha (Manbulloo lies about halfway between Auvergne,

TBA = 6.0m2/ha; and Howard Springs near Darwin, TBA = 8 to 10m2/ha, Hutley et al.

2000). Comparison of these two sets of pasture yields provides some indication of the

reliability of GRASP when incorporating trees into predictions of pasture growth.


206

6.4.2 Results

The presence of trees results in GRASP predicting reduced SSPG across all years at both

locations (Figure 6.10), with the effect more pronounced at Auvergne where tree density is

higher than at Inverway. At Auvergne, SSPG ranges from 1338 to 2044kg/ha/year in the

presence of 6.0m2/ha of trees, much lower than predictions when trees are not included

(2566 to 3345kg/ha/year). Median pasture growth is reduced from 3345 to 1790kg/ha/year,

a decrease of 1555kg/ha/year, or 46%.

At Inverway, SSPG ranges from 303 to 2872kg/ha/year under 2.0m2/ha of trees compared

to 378 to 3345kg/ha/year when trees are not included. The reduction in median pasture

growth from 2567 to 1964kg/ha/year represents a 603kg/ha/year, or 23% decrease.

In GRASP, trees have priority of water use over pastures; i.e. if soil water is available,

demand from trees is satisfied first and pastures have access to what remains. Therefore,

trees also access available nitrogen before pastures. Within-season variation in soil water

availability due to rainfall patterns results in year-to-year variation in total tree water use

and, consequently, in the amount of soil water and N available to pastures. Figure 6.11

offers some indication on how trees affect pasture growth in these two environments. At

Auvergne where nitrogen is almost always limiting, trees lower SSPG by reducing the

amount of N available to pasture as a result of their water use. At Inverway where water

supply and nutrients combine to determine SSPG, the presence of trees affects pasture

growth by competing directly for both these resources.

Comparing field data with model predictions at Manbulloo shows GRASP closely predicts

end of growing season standing dry matter over an 8-year period when accounting for the

presence of trees. Greater year-to-year variation was observed in the field than was

predicted, and this is potentially due to changes over time in nitrogen supply and use, as

Manbulloo is a severely N-limited environment similar to Auvergne.

To summarise, trees have a large effect on predictions of pasture growth in high rainfall

locations in the VRD (median pasture growth reduced by 46%) due to their effect of

reducing nitrogen supply in an already nitrogen limited environment; and a lesser, but still

significant, effect in drier locations (median growth reduced by 23%).


207

Figure 6.10 Time-series of simulated seasonal pasture growth (SSPG, July to June) in the absence and the

presence of trees over 45 years (1959/60 to 2003/04) at a) Auvergne (mean annual rainfall = 900mm); and b)

Inverway (mean annual rainfall = 577mm). Tree basal area (TBA) was set at 6.0m2/ha at Auvergne and

2.0m2/ha at Inverway.

Figure 6.11 Probability distribution of simulated seasonal pasture growth (SSPG) in the absence and the

presence of trees over 45 years (1959/60 to 2003/04) at two locations in the VRD.

Figure 6.12 Presentation yield of tallgrass pasture at the end of growing season in lightly grazed native

pasture woodlands at Manbulloo, near Katherine NT (McIvor et al. 1994) and GRASP predictions of

seasonal pasture growth (July to June) using Regional VRD parameters and tree basal area of 7.0m2/ha.

0.0

0.2

0.4

0.6

0.8

1.0

0 1000 2000 3000 4000 5000

SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

Auvergne (no trees)

Auvergne (TBA 6.0sqm/ha)

Inverway (no trees)

Inverway (TBA 2.0sqm/ha)

0

1000

2000

3000

4000

5000

1959/60 1969/70 1979/80 1989/90 1999/2000

SSPG

(kg/

ha/y

r)

No treesTBA = 6.0sqm/ha

Auvergne

0

1000

2000

3000

4000

5000

1959/60 1969/70 1979/80 1989/90 1999/2000

SSPG

(kg/

ha/y

r)

No trees

TBA = 2.0sqm/haInverwaya. b.

0

1000

2000

3000

4000

1982 1983 1984 1985 1986 1987 1988 1989

TSD

M (k

g/ha

)

McIvor et al. (1994)

GRASP predictions

Manbulloo


208

6.5 Discussion

Year-to-year variability in seasonal pasture growth

Year-to-year variability in pasture growth has major implications for livestock production

and resource condition (Day et al. 1999), and knowledge of historical variability in pasture

production is essential for the development of sustainable grazing management practices.

In this chapter, predictions of historical pasture growth show that year-to-year variability in

pasture growth is mainly a feature of the central and southern VRD (e.g. Victoria River

Downs and Inverway) while variability is low in the north (e.g. Auvergne, Figure 6.2 and

Figure 6.3). This difference in variability is primarily due to the impact of a rainfall

gradient where mean annual totals decrease and the coefficient of variation increases as

distance from the northern coastline increases (Section 3.5.1 in Chapter 3).

However, seasonal rainfall is not the best predictor of pasture growth as evidenced by the

relatively weak regression in Figure 6.5b. The relationship breaks down in circumstances

where within-season rainfall distribution results in a high proportion of rainfall lost through

the deep drainage, surface runoff, or soil evaporation components of the water balance. In

the two cases explored in Section 6.2.2, a large amount of rainfall was recorded over a

relatively short period, followed by little further rain. In these cases, a high proportion of

rainfall was predicted to pass though the deep drainage pathway and transpiration

accounted for less than one quarter of total rainfall. The scatter of points in Figure 6.5b

shows that such circumstances are not uncommon in the VRD.

Transpiration is a better predictor of pasture growth. The relationship between these two

factors is a well-established scientific principal and this is reflected in the structure of

GRASP (Section 4.2) where growth under water limiting conditions is a product of

transpiration (from the water balance) and transpiration efficiency (from the parameter

TUE and corrected for atmospheric vapour pressure deficit). Figure 6.6b shows a strong

linear relationship up to 300mm of plant water use before an upper limit to pasture growth

becomes active. Bolger and Turner (1999) also report a strong linear relationship between

plant water use and pasture production. In their study, the relationship held for up to

440mm of growing season water use where nutrients were not limiting (high fertiliser

input) in temperate annual pastures in Western Australia.


209

Table 6.1 Summary of year-to-year variation in predictions of pasture growth over 45 years at three locations

in the VRD.

Auvergne Victoria River Downs

Inverway

Range (kg/ha) 2566 - 3343 1784 – 3343 378 - 3343

Pasture growth limited by soil water supply (% of seasons) 9 44 75

Pasture growth limited by nitrogen uptake (% of seasons) 91 56 25

The upper limit to pasture growth

An upper limit to seasonal pasture growth occurs in many seasons. This upper limit is

reached in over 90% of seasons at Auvergne and in as few as 25% of seasons at Inverway

(Table 6.1). Rickert et al. (2000) explains that in GRASP, when there are no climatic

limitations, the upper limit to seasonal growth of unfertilised pasture is determined by the

ability of the soil to supply mineralised nitrogen. Results in Figure 6.7b and Figure 6.9

reflect this, showing that the magnitude of the upper limit is determined by the amount of

nitrogen available for uptake (MaxN) and the level to which it can be diluted by pasture

before further growth is no longer possible (%N0Grow). Figure 6.4 shows that these

nitrogen characteristics are affected by both soil type and pasture species composition. In

Figure 6.4a, white grass (WG) and barley Mitchell grass (BMG) have very different

probability distributions. Examining the parameters for these two pasture types shows large

differences in both MaxN (WG = 14.0, BMG = 27.9kg/ha) and %N0Grow (WG = 0.57,

BMG = 0.74%). In Figure 6.4b, pasture growth on structured red earths (SRE) and basalt

cracking clays (BCC) also exhibit different distributions which are again largely explained

by availability of nitrogen (MaxN for SRE = 18.4, for BCC = 27.1kg/ha). Much of the

remaining differences in predictions of SSPG between soil and pasture types at the same

location are attributed to soil water holding capacity.

The nitrogen-induced limitation to pasture growth is also supported by other studies, many

of which have been discussed in the literature review (Section 2.3.2 of Chapter 2). For

example, Mott et al. (1985) and McIvor et al. (1994) both report that seasonal pasture

growth in tropical tallgrass pastures is quite consistent due to nutrient limits rather than

water limits; and Ash and McIvor (1995) report differences in nitrogen uptake and content

due to changes in species composition.


210

The unchanging upper limit to seasonal pasture growth so obvious in the results of Section

6.2.2 is unlikely, however, to be as uniform in the field. Year-to-year variation in perennial

grass basal area, pasture species composition, the effect of trees, and mineralisation of

organic matter to supply nutrients all combine to affect the amount of pasture grown in any

season. Figure 6.12 shows that GRASP predicts very minor year-to-year variation in end of

growing season standing dry matter compared with greater variation in the data of McIvor

et al. (1994). The greater variation observed by McIvor et al. (1994) is likely due in large

part to annual differences in nitrogen supply and dilution. That GRASP does not fully

account for the observed variation indicates that the model is limited in its ability to

simulate the processes affecting N supply and dilution over time.

Sensitivity of SSPG to changes in parameter values

Assessing the sensitivity of predictions of SSPG to transpiration involved three parameters:

1) transpiration use efficiency (TUE); 2) green yield at 50% green cover (GY50GC); and

3) soil water index at which growth stops (SWIX0Grow) (Figure 6.8). Results showed that

TUE was a sensitive parameter and increased SSPG, but only in 50% of low yielding

seasons, before the nitrogen supply constrained further growth. Christie (1978) found that

pasture species possessing the C4 photosynthetic pathway (most tropical grasses) had about

a 60% higher water use efficiency that C3 species (forbs). Therefore, while increasing TUE

has benefits to pasture growth in low SSPG years, it is of little practical application as TUE

is a physiological trait of species rather than one able to be manipulated by management.

Any changes in sward TUE from one year to another are due to changes in species

composition rather than an effect of season or other environmental factors.

Changes to GY50GC also affect pasture growth in low SSPG years. This parameter

represents the predominant growth habit of the pasture, with low values representing

prostrate species (mainly broadleaf forbs and annual short grasses) and high values

representing erect species (perennial grasses and annual mid-height and tall grasses).

Prostrate species have greater solar radiation interception and therefore faster growth rates.

Encouraging prostrate species may seem an attractive managerial goal in water-limited

environments, but in reality increasing the proportion of annual short grasses and forbs in

pasture leads to reduced pasture growth and increased risk of land degradation (Ash and

McIvor 1995) as forbs (C3 species) have a lower TUE.


211

Little effect was found when varying SWIX0Grow at the levels tested. Decreasing from

0.3 to 0.1 resulted in small increases in SSPG at low yields, but a further decrease to 0.01

had no effect. This occurs because the actual difference in soil water content between

SWIX values of 0.1 and 0.01 is very small and little transpiration is possible at these

levels.

Sensitivity to variation in nitrogen parameters was also investigated for three nitrogen

parameters: 1) MaxN; 2) %N0Grow; and 3) nitrogen uptake per 100mm of transpiration

(Nup100T) (Figure 6.9). SSPG is very sensitive to changes in Max N. As previously

discussed, nitrogen supply constrains growth at many locations and in many seasons in the

VRD. It is not surprising that changes to this parameter have a large impact on the upper

limit to SSPG. Increasing nitrogen supply allows pasture to continue using transpired soil

water for dry matter production beyond previous limits. So long as sufficient soil water is

available, increasing N supply will have benefits for pasture growth.

Decreasing the values for %N0Grow allows plants to make more effective use of already-

captured nitrogen once uptake from the soil is exhausted. The lower the level to which it

can be diluted before plants cease to grow or soil water also becomes exhausted, the

greater the amount of SSPG possible. Like TUE, %N0Grow is a physiological trait of

individual species and little can be done managerially to influence this plant characteristic.

The rate at which nitrogen is taken up in soil water has little effect on SSPG. A low rate of

uptake decreases SSPG in water-limited years, but changes to Nup100T do not affect the

upper limit to pasture growth. The model is not very sensitive to values for this parameter.

Of the parameters tested for model sensitivity, MaxN and %N0Grow have the greatest

influence on predictions of pasture growth, particularly in when soil water supply is

plentiful. TUE and GY50GC have some influence in lower SSPG years. Particular care is

required when determining these parameters, and they probably require site-specific values

for best model performance.

The only sensitive parameter open to manipulation by management is MaxN. Examples of

options to improve nitrogen supply are fertiliser application (Norman 1962) and

introduction of legumes (Winter et al. 1989a). Introduction of legumes has been the focus


212

of past research in northern Australia, but due to poor competition with native species and

susceptibility to disease, few species have persisted. However, increasing the total pasture

production is not necessarily the main aim of grazing managers in the region. Dry matter

production is often not the most limiting resource for beef cattle production. Deficiencies

of phosphorus during the wet season, energy during the dry season and protein (nitrogen)

for most of the year are of greater significance (Norman 1965; Niethe et al. 1988; Hasker

2000). Therefore, the ability of pastures to supply adequate nutrition is a greater limitation

to animal production than pasture quantity alone. One approach to improving nitrogen

content of pasture is to manipulate species composition in favour of annual grasses and

forbs, as shown in the results of Chapter 3, and supported by Ash and McIvor (1995).

However, such a species change is associated with increased risk of resource damage, and

is not a recommended practice for long-term sustainable pasture management.

The effect of trees on pasture growth

In a severely nitrogen-limited environment the presence of trees has a large impact on

pasture growth. Figure 6.11 shows that median SSPG at Auvergne is reduced by 46%

when trees are incorporated into simulations. This is consistent with Cafe et al. (1999) who

found an increase in pasture dry matter yield of 56% in a cleared plot compared with

uncleared pasture in a monsoonal tallgrass pasture at Katherine. This increase was

attributed to removal of competition for nutrients, particularly nitrogen, between trees and

pasture. Comparison of model predictions including the effect of trees, and independent

data collected on an open woodland at Manbulloo showed close agreement (Figure 6.12),

indicating that GRASP was adequately accounting for the effect of trees on pasture growth

in this environment.

A range of views exist in the literature on the impact of trees on pasture growth in northern

Australia. Hutley et al. (2000) reports that in a Eucalypt open-forest savanna in the Top

End of the NT, annual transpiration by trees was significantly less than for the grassy

understorey, and that water use by trees did not differ between the wet and dry seasons.

This is most likely due to trees accessing water from deeper in the soil profile that grasses.

This differs from the approach used by GRASP in simulations in this study where trees and

pasture compete for the same water store and annual transpiration was similar for both

vegetation groups at Auvergne (median pasture transpiration = 233mm/year, median tree

transpiration = 238mm/year). Other studies indicate trees can have both competitive and


213

complementary effects on pasture growth (Jackson and Ash 2001; Scanlan 2002; Schmidt

and Lamble 2002). Despite these alternate views, GRASP appears to adequately account

for the presence of trees in simulations of pasture growth in this study.

Summary

To summarise, factors affecting plant transpiration influence how often the upper limit to

pasture growth is reached, and factors affecting nitrogen supply and use affect the

magnitude of this upper limit.

Findings of this study point towards the need for a much better picture of nitrogen supply

across the landscape within seasons, and from year to year, if the absolute values of

individual simulation results are to be used confidently in grazing management. Such an

understanding will permit both better relationships within the structure of GRASP that

describe the processes involved in nitrogen supply and uptake, and more accurate

quantification of these relationships.

6.6 Conclusions

Climate, soil type, and pasture species composition are all sources of year-to-year

variability in seasonal pasture growth in the VRD. Variability is greater in the south than

the north of the region, and this is primarily due to factors affecting pasture transpiration

(i.e. soil water availability due to rainfall amount and distribution). Limited supply of

nitrogen for pasture growth results in little year-to-year variability in the north.

Nitrogen uptake and internal dilution are critical factors affecting predictions of pasture

growth throughout the region. Reliable values for these parameters are necessary if

quantitative accuracy of pasture growth predictions is required. Seasonal pasture

production in the VRD would benefit from increasing the supply of nitrogen available for

plant growth, particularly in the wetter north. Trees also have a significant impact on

pasture growth predictions and need to be accounted for when applying to model to

analysis of grazing management systems.

Chapter 7 Implications for analysing grazing practices in the VRD: examples of model application

214

7.0 Implications for analysing grazing practices in the VRD: examples of model application

7.1 Introduction

It was concluded in Chapter 2 that little data is available on the productivity of native

pastures in the VRD and this is a severe limitation to the development and analysis of

sustainable grazing practices. This thesis has now developed the capacity to simulate

pasture growth for commonly grazed pasture communities in the region. But how is this

new capacity beneficial to the pastoral industry?

This chapter aims to demonstrate ways the newly calibrated and tested model can be

applied to the analysis of grazing management practices in the VRD that were not possible

before this study. Two applications of GRASP are presented. Section 7.2 uses the model to

provide long term historical pasture growth predictions across the VRD for calculating

current and expected future levels of pasture utilisation in the region. Results will indicate

whether current grazing practices are considered safe in the long term, and quantify the

impact of future stocking rate changes. Section 0 investigates the potential benefits to the

pastoral industry of alleviating the current constraints on pasture growth imposed by

limited nitrogen supply. Implications of the results from these two applications are

discussed in Section 7.4 and conclusions presented in Section 7.5. The structure of this

chapter is illustrated in Figure 7.1.


Current and future levels of pasture utilisation

(Section 7.2)

Stocking rate data from producer surveys

Alleviating the nitrogen constraint to pasture growth

(Section 7.3)

Generic parameter sets developed in this study

(Chapter 5)

Discussion and conclusions (Sections 7.4 and 7.5)

Identification of nitrogen as a major factor affecting pasture growth

(Chapter 6)


215

7.2 Pasture utilisation in the VRD

Efficient and sustainable use of native pastures for grazing in northern Australia is based

upon principles aimed at controlling the timing, amount and evenness of pasture use (MLA

2004a). One of the keys to good grazing management is balancing the expected forage (dry

matter) production with forage use. This balance is sometimes expressed as the long term

‘safe’ level of pasture utilisation. MLA (2004b) define safe utilisation rate as “the

maximum rate of average annual pasture use consistent with maintaining or encouraging

good land condition”. Good land condition includes elements such as stable soil,

groundstorey dominance by perennial grasses, and a balance between pasture and woody

species.

Many studies across northern Australia have explored pasture utilisation (e.g. Scanlan et

al. 1994; Johnston et al. 1996a; MacDonald et al. 1997; Hall et al. 1998), and safe

utilisation levels have been estimated for many regions across northern Australia, including

the VRD (Table 7.1). These studies indicate that maximum safe utilisation levels for native

pastures are 25 to 30% across a wide range of climates. Within each region, variation

between land types is apparent. Pasture utilisation continues to be a focus for research in

the Victoria River District. For example, a large-scale project currently under way at

Pigeon Hole Station (16049’S, 131013’E) in the central VRD is investigating pasture

utilisation at the commercial cattle station scale. The Pigeon Hole project is aiming to

evaluate strategies for achieving uniform grazing distribution, and determine whether this

has benefits for livestock production (HeytesburyBeef and MLA 2005). Results will not be

available until after 2007.

Evaluating the sustainability of current grazing practices requires knowledge of pasture

utilisation levels. Until now, determining current levels of pasture utilisation on cattle

properties in the VRD has been limited by the availability of accurate pasture production

data. As a result of this study, pasture growth for individual land types and locations can be

now predicted using GRASP. Reliable calculations of current levels of pasture utilisation

are now possible.


216

Buxton and Stafford Smith (1996) highlight the value of linking producer knowledge with

a systems modelling framework, including incorporating the effects of climate variability

when assessing grazing management options. This section aims to determine current levels

of pasture utilisation across the VRD by combining stocking rate data from recent producer

surveys with predictions of long-term pasture growth. Additionally, data is available that

indicates producers in the region intend to increase stocking rates over the next decade

(Oxley 2006). The impact of this expected increase in cattle numbers on pasture utilisation

levels is also explored.

Table 7.1 Some estimates of long-term safe levels of pasture utilisation across northern Australia

Region Safe utilisation rate Source

Northern Territory

Katherine/Victoria River District 10-30% MLA (2004a)

Kidman Springs (central VRD) 10-25% R.A. Cowley (pers. comm.)

Sturt Plateau 12-20% SPLBPG (1995)

Queensland

Various regions 13-23% Hall et al. (1998)

Central Burnett 27% Day et al. (1997b)

SW Qld 15-20% Johnston et al. (1996a)

Mitchell grasslands 30% Wilson et al. (1984)

Western Australia

Kimberley 0-25% Novelly and Baird (2001)

7.2.1 Methods

Pasture utilisation is a function of the feed intake requirements of grazing animals and the

amount of pasture available for consumption. Animal intake depends upon the class of

animal (young growing steers, lactating cows, etc) and the stocking rate (number of

animals per unit area), while available pasture can now be predicted using GRASP.

Adapting the safe stocking rate function of MLA (2004a) gives us:

Utilisation (%) = intake / pasture growth *100 (Equation 11)

where intake (kgDM/ha/yr) = average dry matter (DM) consumption per year; and pasture

growth (kgDM/ha/yr) = average annual pasture growth.


217

7.2.1.1 Animal intake

The formula for calculating annual animal intake is given by MLA (2004a):

Annual intake = forage demand per animal unit * stocking rate (Equation 12)

where: intake is kgDM/ha/year; forage demand per animal is kgDM/AE/year; stocking

rate is AE/ha; and AE is Animal Equivalents.

Animal Equivalents

Dry matter intake by grazing cattle is influenced by breed, size, and reproductive status.

Animals of different classes can be standardised so that they are easily comparable; a

concept known as Animal Equivalents (AE’s). One AE is defined as “the intake of a 450kg

dry cow, and intakes of other animals are relative to this” (MLA 2004b). In this study, a

450kg cow raising a calf to 6 months is equated to 1.35AE’s, a yearling steer (8 to 18

months) equals 0.68AE’s, and an older steer (19-30 months) equals 0.8AE’s. All

calculations of dry matter intake are based on these values.

Forage demand per AE

Many authors provide estimates of annual forage consumption by a ‘standard’ animal (e.g.

Holechek 1988; SCA 1990; Vallentine 1990; Hall et al. 1998; MLA 2001), varying from

2700 to 4100kgDM/AE/year. Average forage consumption in the Katherine region is

estimated at 3650kgDM/AE/year (MLA 2004a), and this value is used in all calculations in

this study.

Stocking rate

Two sources of recent stocking rate data in the VRD are available. In 1997, fourteen cattle

producers in the Victoria River District were surveyed about their land management

practices (Smith 1998). One part of the survey related to their estimations of long-term

average stocking rates for the different land systems on their property. Ten respondents

provided estimations of stocking rates on some or all of the main land systems on their

property (Table 7.2).


218

In 2004, twenty four cattle producers in the VRD were surveyed as part of a Territory-wide

effort to document the current management practices and attitudes of the cattle industry

(Oxley 2006). Included in the survey responses were data from twenty two properties on

current grazing area and carrying capacity. From the survey data, average property

stocking rates were calculated (Table 7.4). Many respondents indicated they intended to

increase their carrying capacity over the next decade and provided estimates of expected

property carrying capacities in five years (2009) and ten years (2014) time. Average future

stocking rates were calculated using this data.

In both surveys, producers provided data with the understanding that individual properties

were not identified when analysing and publishing survey results. Their wishes are

respected in this thesis.

The different scales of stocking rate/carrying capacity data provided in the surveys require

separate approaches when predicting pasture growth for the calculation of utilisation rates.

The approaches used are detailed in the following sections.

7.2.1.2 Predicting pasture growth on specific land systems - the 1997 survey

Records of long term pasture production do not exist for the VRD, and is only now

possible to predict pasture growth for any land system using the GRASP model and the

generic parameter sets developed in this study (Table 5.2). While GRASP has been

calibrated for five species groups and three soil groups at five locations in the VRD,

application across the wider pastoral landscape of the region encounters other pasture

communities, soil types, and climatic regimes. The closest association between an untested

land type and the newly-developed generic parameter sets must be established before

simulation modelling is possible.

The land system mapping of CSIRO (1970) and pasture lands mapping of Tothill and

Gillies (1992) provide descriptions of soils and vegetation in the VRD from a pastoral

industry perspective, and Wilson et al. (1990) also provides herbaceous species

information. Using these resource descriptions, the generic soil and pasture parameter sets

developed in this study were linked to the land systems provided by producers (Table 7.2).


219

Given the diversity of land types for which producers provided estimates of ‘safe’ carrying

capacity, the Soil x Species approach described in Section 5.3.2 was used. This method

allows differences in productivity between land systems at the same location to be

expressed. To account for the effects of trees and management practices not included in the

generic parameters, a number of modifications were necessary. For each land system:

• tree basal area (1.0 to 6.0m2/ha) was estimated from Stewart et al. (1970), Wilson et al.

(1990) and Carter et al. (1996a) and kept constant across all years;

• timing and frequency of burning (May to November; 0 to 40% per annum) was

estimated from Smith (1998); and

• the threshold for standing dry matter required to carry a fire (1500kg/ha) was taken

from Dyer et al. (2001b).

Burning is a common management practice in the region and affects calculations of pasture

growth and nitrogen uptake. Producers provided data on the timing and frequency of

burning each land system in the survey, and this has been incorporated into the

simulations.

Another issue requiring attention was how to estimate soil water holding capacities (WHC)

for land systems not represented in the field study or the generic parameters in Table 5.2.

Using descriptions in Stewart et al. (1970), soil depths were estimated for all land systems.

Where necessary, WHC’s for red earths were proportionally adjusted from the values in

Table 5.2 to account for differences in estimated profile depth. Sandy soils were assumed

to have half the WHC of red earths of the same depth. Values for WHC for each affected

land system are:

• Cockburn and Pinkerton land systems, WHC = 63mm;

• Tanmurra, Franklin, Napier and Antrim land systems, WHC = 101mm;

• Dinnabung and Frayne land systems, WHC = 147mm; and

• Cockatoo land system, WHC = 182mm.

Generic parameter sets were prepared and relevant daily climate data extracted from the

Data Drill website. Simulations were run from 1957 to 2004 for each land system at each

location, and median pasture growth calculated.


220

7.2.1.3 Predicting average pasture growth for the VRD – the 2004 survey

Stocking rate data for separate land systems was not collected in the 2004 survey, so the

approach used in the previous section is not applicable here. Individual properties were not

identified, so using location-specific climate data to predict pasture growth at each

property was again not possible. Consequently, pasture growth was predicted for a single

point and applied to all properties when calculating utilisation rates. Pasture growth was

predicted using the Regional VRD parameter set (Table 5.2) at Victoria River Downs (a

central location) and the climate file from 1957 to 2004. Tree basal area was set at

4.0m2/ha, and 25% of pasture burnt in October each year. Long-term median pasture

growth was predicted as 2060kg/ha/year.

7.2.1.4 Calculating utilisation

Utilisation rates were calculated for each land system (1997 survey) and each property

(2004 survey) using Equation 11 and Equation 12. During calculations, a number of

assumptions were made:

• pastures are used evenly, with no accounting for differences in utilisation rates due to

topography or distance from cattle watering points;

• for calculations of future utilisation rates:

the area grazed remains the same as in 2004; and

long term climate (and subsequent pasture growth) is also unchanged.


221

Table 7.2 Stocking rate and burning frequency for each land system in the 1997 survey; and estimations of tree basal area, closest associated generic pasture and soil

parameters, and estimated soil depth for each land system.

Property Land system 1 Safe stocking rate 2 (breeders/km2)

Burning regime 2 Tree basal area 3 (m2/ha)

Pasture parameter group 4

Soil parameter group 4

Soil depth 5 (cm)

1 Angallari 3.5 30% burnt in October 6 Ribbon grass Yellow earth 100 1 Cockatoo 2 30% in May 6 White grass Red earth 100 1 Cockburn 1 30% in May 3 White grass Red earth 30 1 Dinnabung 3.5 30% in October 5.5 Ribbon grass Red earth 72 1 Ivanhoe 10 20% in October 2.5 Ribbon grass Alluvial cracking clay 100 1 Ivanhoe 2 (forbs) 13 20% in October 2.5 Forbs Alluvial cracking clay 100 2 Inverway 4.5 15% in November 1 Barely Mitchell grass Alluvial cracking clay 100 3 Dinnabung 12.5 steers 10% in May 5.5 Ribbon grass Red earth 72 3 Ivanhoe 25 steers 10% in May 2.5 Ribbon grass Alluvial cracking clay 100 4 Dinnabung 10 25% in October 5.5 Ribbon grass Red earth 72 4 Pinkerton 1.5 25% in October 5.5 White grass Red earth 30 4 Tanmurra (Best) 6.5 25% in October 5.5 Annual short grass Red earth 50 5 Dinnabung 5 40% in May 5.5 Ribbon grass Red earth 72 5 Frayne 4 20% in May 5.5 Annual short grass Red earth 72 5 Willeroo 6 Unburnt 6 Ribbon grass Basalt cracking clay 88 6 Franklin 4.5 10% in November 3 Annual short grass Red earth 50 6 Inverway 8 5% in November 1 Barely Mitchell grass Alluvial cracking clay 100 6 Wave Hill 9 5% in November 2 Barely Mitchell grass Basalt cracking clay 88 7 Argyle 10 Unburnt 2.5 Barely Mitchell grass Basalt cracking clay 88 7 Dinnabung 5 35% in November 5.5 Ribbon grass Red earth 72 7 Ivanhoe 8 Unburnt 2.5 Ribbon grass Alluvial cracking clay 100 7 Napier 3.5 25% in November 5.5 White grass Red earth 50 8 Franklin 2.5 10% in November 3 Annual short grass Red earth 50 8 Inverway 7 10% in November 1 Barely Mitchell grass Alluvial cracking clay 100 8 Wave Hill 8 10% in November 2 Barely Mitchell grass Basalt cracking clay 88 9 Dinnabung 22.5 steers 40% in May 5.5 Ribbon grass Red earth 72 9 Willeroo 15 Unburnt 6 Ribbon grass Basalt cracking clay 88

10 Antrim 2.5 10% in November 3 Annual short grass Red earth 50 10 Inverway 5 10% in November 1 Barely Mitchell grass Alluvial cracking clay 100 10 Wave Hill 10 10% in November 2 Barely Mitchell grass Basalt cracking clay 88

1 Stewart et al. (1970); 2 from Smith (1998): 3 estimated from Stewart et al. (1970), Wilson et al. (1990), Carter et al. (1996a) and R.A. Cowley (pers. comm.); 4 generic

parameter sets in Table 5.2, estimated using Tothill and Gillies (1992) and Stewart et al. (1970); 5 estimated from Stewart et al. (1970).


222

7.2.2 Results – the 1997 survey

Long-term pasture growth and utilisation results calculated from the land system and

stocking rate data provided in the 1997 survey are shown in Table 7.3. Long-term pasture

utilisation varied from 5.1% (Cockburn land system on Property 1) up to 45.2%

(Dinnabung on Property 9). Average pasture utilisation across all properties and land

systems was 16.4%.

Low utilisation rates (<10%, eight of thirty cases) occurred on five properties and over

many land systems (Angallari, Antrim, Cockatoo, Cockburn, Franklin, Inverway, and

Pinkerton), pasture production levels (954 to 2915kg/ha/year) and soil types (yellow earth,

shallow to deep red earth, and cracking clays).

A high proportion of utilisation rates were between 10 and 20% (sixteen of thirty cases).

These also occurred across many properties (eight of ten), land systems (Argyle,

Cockburn, Dinnabung, Franklin, Frayne, Inverway, Ivanhoe, Napier, Tanmurra, Wave Hill

and Willeroo), pasture production levels (1176 to 3173kg/ha/year), and soil types (cracking

clays and moderately deep to deep red earths).

High utilisation rates (>25%, six of thirty cases) occurred on just four properties and three

land systems (Dinnabung, Ivanhoe and Willeroo), but over a wide range of pasture

production levels (1127 to 2328kg/ha/year). Soil types were restricted to deep red earths

and cracking clays.

These results show that land systems with white grass (Sehima), barley Mitchell grass

(Astrebla), annual short grasses (Brachyachne) and similar species support low to medium

utilisation rates, and land systems containing ribbon grass (Chrysopogon) and forbs

support medium to high utilisation rates. Utilisation results also revealed that low to

medium utilisation rates occur throughout the region, while high utilisation rates were

restricted to the high rainfall zone (data not shown to protect property identities).

Utilisation is the ratio between animal intake and pasture growth, so regressing these two

variables is an alternative view of the data presented in Table 7.3. Results of this regression

clearly show that utilisation rates can be separated into two distinct groups: 1) utilisation


223

rates 20% or below; and 2) utilisation rates above 25% (Figure 7.2). Also presented in

Figure 7.2 is the function y = 0.21x + 4.5, from Hall et al. (1998) who summarised

utilisation data from three regions in Queensland. It is apparent that the majority of

properties in the 1997 survey utilise pastures at a lower rate than their eastern state

counterparts, while those cases where utilisation exceeds 25% are at the high end of

pasture utilisation rates by northern Australian standards.

Figure 7.2 Relationship between intake at long-term stocking rate and median simulated seasonal pasture

growth (SSPG) across 10 properties in the VRD using stocking rate data provided in the 1997 survey (Smith

1998) and pasture growth predicted using the Soil x Species approach to developing generic parameters. The

dashed line represents a similar relationship from three regions in Queensland (Hall et al. 1998).

y = 0.21x + 194r2 = 0.677

y = 0.16x - 58r2 = 0.673

0

250

500

750

1000

0 1000 2000 3000 4000

Median SSPG (kg/ha/yr)

Inta

ke a

t lon

g-te

rm s

tock

ing

rate

(k

g/ha

/yr)

Utilisation <20%

Utilisation >25%

y = 0.21x + 4.5


224

Table 7.3 Median simulated seasonal pasture growth (SSPG) calculated using the Site x Species approach to

developing generic parameters; and long term utilisation rates for specific land systems on 10 properties in

the VRD, using stocking rate data provided in the 1997 survey (Smith 1998).

Property Land system 1 Median SSPG

(kg/ha/yr)

Long term stocking rate

(AE/km2)

Animal intake at stocking

rate (kg/ha/yr)

Utilisation rate

(% of median

SSPG)

1 Angallari 2003 4.7 172 8.6

1 Cockatoo 1633 1.4 99 6.1

1 Cockburn 954 2.7 49 5.1

1 Dinnabung 1361 4.7 172 12.7

1 Ivanhoe 2828 13.5 493 17.4

1 Ivanhoe (frontage) 2 2328 17.6 641 27.5

2 Inverway 2290 6.1 222 9.7

3 Dinnabung 1127 10.0 310 27.5

3 Ivanhoe 2351 20.0 621 26.4

4 Dinnabung 1396 13.5 493 35.3

4 Pinkerton 1271 2.0 74 5.8

4 Tanmurra 1806 8.8 320 17.7

5 Dinnabung 1248 6.8 246 19.7

5 Frayne 1660 5.4 197 11.9

5 Willeroo 2234 8.1 296 13.2

6 Franklin 1608 6.1 222 13.8

6 Inverway 2925 10.8 394 13.5

6 Wave Hill 3024 12.2 443 14.7

7 Argyle 3173 13.5 493 15.5

7 Dinnabung 1267 6.8 246 19.4

7 Ivanhoe 2704 10.8 394 14.6

7 Napier 1176 4.7 172 14.7

8 Franklin 1715 3.4 123 7.2

8 Inverway 2812 9.5 345 12.3

8 Wave Hill 2942 10.8 394 13.4

9 Dinnabung 1235 18.0 558 45.2

9 Willeroo 2297 20.3 739 32.2

10 Antrim 1860 3.4 123 6.6

10 Inverway 2915 6.8 246 8.5

10 Wave Hill 3012 13.5 493 16.4

Mean 2039 8.9 326 16.4

1 From Stewart et al. (1970) 2 Frontage refers to floodout levees associated with a major river


225

7.2.3 Results – the 2004 survey

Current property utilisation rates calculated from the results of the 2004 survey, and those

forecast for 2009 and 2014, are presented in Table 7.4. Current whole-property pasture

utilisation levels range from 7.4 to 32.9%, averaging 16.3% across twenty two properties.

Most utilisation rates are between 10 and 20% (sixteen of twenty two properties). These

results are remarkably similar to those of the 1997 survey, despite being calculated from a

different producer-provided dataset and using median pasture growth from a single

location rather than separate values for individual land systems.

Utilisation rates are expected to rise over the next decade as a result of properties

increasing their carrying capacities. In 2009, property utilisation rates will range from 8.5

to 38%, averaging 19.1%; and in 2014 the range will be the same (the two properties

representing either end of the range do not anticipate further increases after 2009), while

the average will increase to 19.9%. At this time, fourteen properties will still have

utilisation levels less than 20%.

Utilisation levels have so far been calculated based on median pasture growth at a single

location (Victoria River Downs). However, results in Chapter 6 showed that pasture

growth can vary from year to year in the VRD, particularly in the south. This variation will

impact upon annual utilisation rates. To illustrate this effect, Figure 7.3 shows how

utilisation varies over time at three locations in the region if a constant stocking rate is

maintained (in this case 9.2AE/km2, the district average in Table 7.4). Results show that

for most years utilisation varies from 10 to 25%, within the safe levels for the VRD (Table

7.1). Exceptions occur, mainly at Inverway during the mid 1960’s (a well-recorded drought

and pasture degradation period in central Australia - McKeon et al. (2004)), and to a lesser

extent in the early 1970’s and early 1990’s when annual utilisation levels exceeded 30%.

An alternative view is to examine the year-to-year variability in stocking rate required to

maintain a constant level of pasture utilisation (in this case 16.3%, Table 7.4). Figure 7.4

shows that across the district, stocking rates would need to vary from 7 to 12AE/km2 to

achieve constant utilisation in the vast majority of seasons. Figure 7.4 also shows that

stocking rate needs to be reduced to just 3 to 5AE/km2 to avoid pasture degradation during

periods of drought.


226

Table 7.4 Current and expected future stocking rates (SR) and levels of pasture utilisation (Util) on 22

properties in the VRD using property carrying capacity and grazing area data from the 2004 survey (Oxley

2006). Pasture growth used to determine utilisation calculated from the Regional VRD parameter set at

Victoria River Downs.

Property Current SR (AE/km2)

Current Util (%)

SR in 2009 (AE/km2)

Util in 2009 (%)

SR in 2014 (AE/km2)

Util in 2014 (%)

1 4.2 7.4 20.0 35.4 20.0 35.4

2 4.8 8.5 4.8 8.5 4.8 8.5

3 10.7 19.0 10.7 19.0 10.7 19.0

4 18.6 32.9 21.4 38.0 21.4 38.0

5 10.4 18.5 12.5 22.1 15.6 27.7

6 7.1 12.5 7.1 12.5 7.1 12.5

7 11.1 19.7 12.8 22.6 12.8 22.6

8 8.0 14.2 10.0 17.7 10.0 17.7

9 13.4 23.7 13.4 23.7 13.4 23.7

10 9.0 15.9 9.7 17.3 10.5 18.6

11 9.1 16.1 9.5 16.9 9.5 16.9

12 8.9 15.8 8.9 15.8 8.9 15.8

13 10.6 18.8 10.6 18.8 10.6 18.8

14 5.3 9.4 6.0 10.6 6.7 11.8

15 9.0 15.9 9.0 15.9 9.0 15.9

16 10.0 17.7 11.7 20.7 15.0 26.6

17 8.2 14.6 10.0 17.7 10.7 19.0

18 10.0 17.7 11.4 20.2 11.4 20.2

19 10.5 18.7 10.5 18.7 10.5 18.7

20 5.3 9.4 6.5 11.5 8.1 14.4

21 10.0 17.7 10.0 17.7 10.0 17.7

22 7.7 13.6 10.8 19.1 10.8 19.1

Mean 9.2 16.3 10.8 19.1 11.3 19.9


227

Figure 7.3 Effect of climate variability on annual pasture utilisation over a 45 year period (1959/60 to

2003/2004) when a constant stocking rate is maintained (district average = 9.2AE/km2, Table 7.4).

Figure 7.4 Effect of climate variability on annual stocking rate over a 45 year period (1959/60 to 2003/2004)

when a constant utilisation rate is maintained (district average = 16.3%, Table 7.4).

0

5

10

15

20

1959/60 1969/70 1979/80 1989/90 1999/2000

Ann

ual s

tock

ing

rate

(AE/

km2)

Northern VRDCentral VRDSouthern VRDDistrict average

0%

10%

20%

30%

40%

50%

60%

1959/60 1969/70 1979/80 1989/90 1999/2000

Ann

ual p

astu

re u

tilis

atio

n Northern VRDCentral VRDSouthern VRDDistrict average


228

7.3 Potential benefits of alleviating the nitrogen limitation to pasture

growth

Nitrogen supply is known to be a limiting factor for pasture growth in the semi-arid tropics

of northern Australia (Section 2.3.2 in Chapter 2). During sensitivity analysis in Chapter 6

it was shown that nitrogen supply constrains pasture growth in up to 91% of seasons at

Auvergne in the northern VRD. Alleviating this constraint has likely benefits for increased

pasture growth, and subsequent animal production through increased property carrying

capacities. This section aims to determine the potential benefits of alleviating the nitrogen

constraint to pasture growth in the northern VRD.

To achieve this aim, literature relating to improving nitrogen status in tropical pasture

systems is first reviewed in Section 7.3.1. Then, by modifying parameters in GRASP to

remove the nitrogen constraint (Section 7.3.2), pasture growth is simulated for the high

rainfall zone of the northern VRD. In effect, this quantifies the climatic potential for

pasture production in this environment (Section 7.3.3). The resulting predictions of pasture

growth and nitrogen uptake provide a basis for undertaking a benefit-cost analysis of

improving nitrogen supply, a process described in Section 7.3.4.

7.3.1 Approaches to improving nitrogen status of pastures in the semi-arid tropics

Considerable research has focussed on methods to improve the nitrogen status of pastures

in the semi-arid tropics, including:

• addition of nitrogen fertilisers (Norman 1962);

• extended periods of fallow allowing mineralised N to accumulate (Wetselaar and

Norman 1960);

• augmentation of native pastures with introduced legume species (Gardener et al. 1993);

• replacement of native pastures with sown legume-grass pastures (Winter et al. 1996);

• ley farming systems incorporating a legume phase (Dimes et al. 1996); and

• reducing competition from other vegetation groups (i.e. killing or removal of trees,

Cafe et al. 1999).


229

From a pastoral industry perspective, many of these options are not feasible. The high cost

of transporting and applying fertiliser (Dimes et al. 1996) and the relatively low recovery

rates of the additional nutrients by pasture (<30%, Blunt and Haydock 1978; Beech 1990)

preclude using fertilisers. The practice of extended fallowing involves land being out of

production for some years and this is not practical in a grazing enterprise, nor

environmentally sound. Ley farming requires large capital investment, ongoing inputs,

intensive management practices, and has not been a successful venture in the region (Bauer

1985). Winter et al. (1989b) found that, under grazing, the grass yield advantage of killing

trees was short-term and declined over time. Additionally, large-scale killing or removing

trees has considerable environmental implications and is not a recommended practice. Full

pasture replacement is only suitable to small areas of better quality land used for special

purposes (Parker and Cobiac 2001). This leaves augmentation of native pastures with

introduced legumes as the most practical option available in the northern VRD to improve

the nitrogen status of pastures.

Considerable work was done on tropical legumes at Katherine Research Station by CSIRO,

much of it involving Townsville stylo (Stylosanthes humilis), a self-seeding annual

legume. After 7 years of S. humilis pasture, annual cereal crops took up 66kg/ha of

nitrogen, about half the soil nitrogen available (Wetselaar and Norman 1960), compared to

just 7kg/ha of nitrogen uptake by an undisturbed native pasture (Norman 1963a).

Wetselaar (1967) reported a mean yield of 5250kg/ha for S. humilis over a 10-year period

under dryland conditions, with N fixation of 73kg/ha/year. Stewart (1970a) give a similar

mean dry matter yield (5300kg/ha) for a mixed S. humilis - Cenchrus ciliaris (buffel grass)

pasture ‘under reasonable agronomic treatments’. Despite the benefits to soil fertility,

pasture production and subsequent animal production, the use of legumes in pastoral

enterprises declined in the 1970’s. Changes in pastoral industry economics, the poor

competitive ability of S. humilis with annual grasses, and its susceptibility to the fungal

disease anthracnose (Colletotrichum gloeosporioides) all contributed to the decline.

New species were investigated including Caribbean stylo (S. hamata) and shrubby or

perennial stylo (S. scabra). These species were used to augment existing native pastures,

resulting in advantages for animal liveweight gain in the NT (Winter et al. 1989a) and

Queensland (Gardener et al. 1993), but responses for pasture herbage yield were mixed

and depended partially upon intensity of grazing and rates of phosphorus fertiliser


230

application. Pasture yields in an augmented pasture over a 13-year period averaged in

excess of 5000kg/ha in a study near Townsville in Queensland (Gardener et al. 1993).

Mean annual rainfall during their study period (1973 to 1986) was 854mm, suggesting

similar results could be expected in the northern VRD. Nitrogen uptake in a maize (Zea

mays) crop increased by 30kg/ha following a 1-year S. hamata pasture, and by 55kg/ha

following a 3-year legume phase compared to a crop grown following a grass pasture

(Jones et al. 1996), highlighting the effectiveness of legumes at improving nitrogen

nutrition in subsequent crops. Annual legumes are known to progressively improve the soil

organic N status of many soils by between 30 and 80kgN/ha per annum in temperate and

tropical pastures (Peoples and Baldock 2001).

Most studies emphasize the necessity of adequate phosphorus nutrition (superphosphate

application) for successful establishment and persistence of legumes, and this is

particularly important in the VRD as soils are very low in P (Table 3.32 in Chapter 3).

Additional management considerations include appropriate grazing management, control

of weeds (Eyles et al. 1984) and appropriate use of fire (Mott et al. 1989).

The contribution of native legumes to improving soil nitrogen status in the VRD is not

known, but two cases in the field study of this thesis imply they are capable of increasing

the amount of N available for plant growth. The first case involves the unusually high N

uptake value at Site 15 (Section 3.4.3.8) where 56.4kg/ha was measured. While this value

was considered an anomaly during calibration of GRASP in Chapter 4, it does suggest that

high levels of N uptake are possible when forbs (presumably including some legumes) are

plentiful. The second case involves another high N uptake value, this time at Site 8

(Section 3.4.3.5) where 36.6kg/ha was measured. This high N uptake was attributed to the

abundance of the native legume Flemingia pauciflora.

Presently, about one third of stations in the Katherine Region have some small areas of

augmented pasture. Oxley (2006) reports that those properties in the VRD with some form

of pasture improvement average 38ha of fully improved pastures and 80ha of augmented

pastures. However, survey respondents expressed difficulty in providing reliable figures as

much of the pasture improvement was scattered thinly over large areas. The main

impediment to wide scale adoption seem to be based on producers perceiving greater

returns from their investment dollars by increasing station carrying capacity through means


231

other than pasture augmentation (i.e. increasing the evenness of grazing on existing native

pastures using additional watering points and fencing, T.J. Oxley pers. comm.). Also,

establishment and persistence of legumes introduced to established native pastures has

proved problematic in the past (R.T. Andison pers. comm.), further detracting from the use

of augmented pastures in the region.

It is concluded from this review that, despite past problems, augmentation of native

pastures with legumes has the potential to improve pasture productivity in the northern

VRD. In the next section, simulation of pasture growth when the nitrogen constraint is

removed is undertaken to quantify the level of nitrogen supply required for pastures to

grow unrestricted by this nutrient.

7.3.2 Method

In GRASP, nitrogen uptake is a passive process linked to transpiration. The relationship

between N uptake and transpiration is outlined in Section 4.2.2. Essentially, pasture N

yield during the growing season depends on both the amount of internally-stored N at the

beginning of the season (Nup0Trans) and the N uptake per unit of transpiration (Nup100T)

since the beginning of the growing season. N uptake continues until transpiration ceases or

the available N supply (MaxN) is exhausted (Figure 4.3). MaxN is the maximum potential

pasture nitrogen yield, and is a surrogate for nitrogen supply from the soil.

In an environment where nitrogen is non-limiting, maximum potential growth is related to

maximum transpiration. Long-term pasture growth simulations at Auvergne in the northern

VRD (Section 6.2.2) resulted in maximum seasonal transpiration of 675mm/year

(1975/76). The average value for Nup100T was 10.5kg/ha/100mm of transpiration. This

gives a preliminary estimate of 71kg/ha as the amount of N required for unrestricted

growth. However, high amounts of available soil nitrogen in the field would likely increase

the concentration of N in soil water, and consequently the rate of N uptake per unit of

pasture transpiration. Additionally, increased growth in the absence of a nitrogen

restriction may lead to greater cumulative transpiration by increasing the amount of

transpiring green plant material and reducing soil evaporation. Thus, the N supply required

to achieve unrestricted growth is likely to be higher than 71kg/ha.


232

To simulate pasture growth when nitrogen is non-limiting, GRASP was run using the

Auvergne climate file (1957 to 2004) and the Regional VRD parameter set (Table 5.2) with

a MaxN value of 100kg/ha. The rate of N uptake was set at 11kg/ha/100mm of

transpiration, as sensitivity analysis showed predictions of pasture growth do not change at

levels above this value (Figure 6.9).

Trees compete with pasture for available soil water and, as a consequence, available

nitrogen. Given the widespread occurrence of trees in the northern VRD, any scenario

development based on simulation studies of pasture growth in this area should account for

their effect. However, the inclusion of trees in a study to determine the amount of nitrogen

required for unrestricted pasture growth adds an unnecessary layer of complexity to results.

For this reason, two simulations were run: 1) in the absence of trees, for determining the

level of N supply at which pasture growth is no longer restricted; and 2) in the presence of

trees, for determining the actual amount of pasture growth likely under field conditions if

N supply was non-limiting (this becomes important in subsequent sections of this chapter).

7.3.3 Results

Figure 7.5 shows results of simulating pasture growth at Auvergne in the absence of a

nitrogen limitation to pasture growth. When compared with predictions under current

levels of N supply (24kg/ha), the effect of removing the N limitation can be clearly seen.

Pasture growth exhibits considerable year-to-year variability, and the upper limit to pasture

growth imposed by inadequate N (as expressed by the flat vertical line in the probability

distribution for N-limited conditions) is no longer apparent. Figure 7.6a shows that most of

the variability in pasture growth can now be explained by changes in seasonal transpiration

(r2 = 0.862). This is further demonstrated by the virtually identical pattern in the

relationship between N uptake and SSPG (r2 = 0.861, Figure 7.6b).

Table 7.5 shows that in the absence of trees the maximum level of nitrogen uptake over the

45-year period from 1959/60 to 2003/04 is 96kg/ha, resulting in 9597kg/ha of dry matter

production (1975/76). The minimum predicted N uptake over that period is 38kg/ha,

resulting in 3436kg/ha of SSPG (1986/87). Another effect of increasing the amount of N

available for pasture growth is an increase in the N content of pasture at the end of the

growing season. When N is still available for uptake at the end of the growing season,


233

internal dilution to continue growth is unnecessary (Figure 6.7b). This explains the higher

values of end of season N content for unlimited-N compared with N-limited simulations.

Results in Table 7.5 also show that total yearly transpiration increases as a result of

increasing available N. This occurs because under N-limited conditions once N limits

growth, green cover cannot increase further and an upper limit to transpiration rate occurs.

Under unlimited-N conditions, green cover continues to increase so long as sufficient soil

water is available to support additional transpiration. Effectively, this increases the

proportion of evapotranspiration that is transpired by pasture and reduces the proportion

lost as direct evaporation from the soil. Consequently, total transpiration (and therefore

pasture growth) increases as a direct result of alleviating the N limitation.

When trees are incorporated into the simulation, the potential benefits to pasture growth of

removing the N limitation are much reduced (Table 7.5). As N supply is no longer

restricted, competition for available soil water becomes the primary effect of trees on

predicted pasture growth. Under these conditions, SSPG ranges from 1475 to

7456kg/ha/year, with a median of 3824kg/ha/year. Maximum N uptake in pastures is

reduced to 65kg/ha, with a median of 33kg/ha. As shown in Chapter 6, competition

between trees and pastures has a significant impact on the outcome of predictions of

pasture growth in the northern VRD.

Figure 7.5 a) Time-series; and b) probability distribution of simulated seasonal pasture growth (SSPG) in the

absence of trees over 45-year period (1959/60 to 2003/04) using Regional VRD parameters at Auvergne

when maximum nitrogen supply (MaxN) is limited (24kg/ha, the observed district average) and theoretically

unlimited (96kg/ha).

0.0

0.2

0.4

0.6

0.8

1.0

0 2000 4000 6000 8000 10000

SSPG (kg/ha/yr)

Prob

abili

ty o

f exc

eede

nce

MaxN = 24kg/ha

MaxN = 96kg/ha0

2500

5000

7500

10000

12500

1959/60 1969/70 1979/80 1989/90 1999/2000

SSPG

(kg/

ha/y

r)

MaxN = 96kg/ha

MaxN = 24kg/ha a. b.


234

Figure 7.6 Relationship between simulated seasonal pasture growth (SSPG, July to June) and a) seasonal

transpiration (July to June); and b) total nitrogen uptake in the absence of trees over 45 years (1959/60 to

2003/04) at Auvergne when maximum nitrogen supply (MaxN) is limited (24kg/ha, the observed district

average) and theoretically unlimited (96kg/ha).

Table 7.5 Comparison of simulated key nitrogen and pasture variables under nitrogen-limited and unlimited

conditions, and in the presence and absence of trees at Auvergne over a 45 year period (1959/60 to 2003/04).

Values across each row are not necessarily from the same season.

N uptake

(kg/ha/yr)

N content at end of growing season 1

(%)

Pasture transpiration

(mm/yr)

SSPG 2

(kg/ha/yr)

With Trees

N-limited conditions

Minimum 10 0.65 82 1338

Median 12 0.69 233 1790

Maximum 15 0.84 369 2044

Unlimited N

Minimum 13 0.81 94 1475

Median 33 0.91 283 3824

Maximum 65 1.03 585 7456

Without Trees

N-limited conditions

Minimum 24 0.72 226 2359

Median 24 0.72 417 3345

Maximum 24 0.93 675 3345

Unlimited N

Minimum 38 0.91 297 3436

Median 61 1.10 488 5747

Maximum 96 1.34 791 9597

1 calculated by dividing N uptake by pasture growth for each season. 2 simulated seasonal pasture growth

y = 11.6x - 108r2 = 0.862

0

2000

4000

6000

8000

10000

12000

0 200 400 600 800 1000

Seasonal transpiration (mm)

SSPG

(kg/

ha/y

r)

MaxN = 96kg/ha

MaxN = 24kg/ha

y = 98.6x - 405r2 = 0.861

0

2000

4000

6000

8000

10000

12000

0 20 40 60 80 100 120

Total N uptake (kg/ha)

SSPG

(kg/

ha/y

r)

MaxN = 96kg/ha

MaxN = 24kg/haa. b.


235

7.3.4 Augmenting native pastures with legumes: a benefit-cost analysis

Benefit-cost analysis is defined as “a method to assess the relative desirability of

competing alternatives, where desirability is measured as economic worth” (Sinden and

Thampapillai 1995). Benefit-cost analysis involves a number of steps:

• identify the problem and define alternatives to resolve it;

• value the benefits and costs of each alternative;

• calculate the net benefits;

• compare alternatives based on their net benefits;

• test for the effect of changes in assumptions and data; and

• make a final recommendation.

A benefit-cost analysis of alleviating the nitrogen constraint to pasture growth compared

with maintaining existing native pasture is undertaken in this section using these steps as a

framework.

7.3.4.1 The problem and the alternatives

It was identified in Chapter 6 that the availability of nitrogen in the high rainfall zone of

the northern VRD is a major constraint to pasture growth. This constraint limits the level of

animal production possible from these pastures. Alternatives for improving the nitrogen

status of pastures in this zone have been reviewed in the preceding sections, with

augmentation of pastures using introduced legumes being the most practical option

available. Maintaining existing native pastures is the ‘do nothing’ option. Two other

aspects require quantification: 1) the scale of the problem; and 2) the likely level of

adoption of the proposed alternative.

Scale

30 000km2 (3 million hectares) in the Ord-Victoria Area was identified by CSIRO (1970)

as being suitable for introduced stylo pastures. The total area of such land across northern

Australia is obviously much greater, but this analysis is restricted to the area covered by

the CSIRO study. Land systems suitable for pasture augmentation in the Ord-Victoria Area

are the Upland tall grass plains (Wingate, Moyle, Cockatoo, Litchfield and Macphee land

systems) and the Tippera tall grass plains (Matheson, Jindara, Wriggley, Dinnabung,

Frayne and Angallari land systems) (Table 3.1 in Chapter 3). These land systems occur in

areas receiving greater than 750mm of annual rainfall, the threshold for establishment of


236

stylo pastures (Stewart 1970a). Land systems receiving less than 750mm per annum are not

considered in this analysis although it is acknowledged that pasture augmentation may be

possible in this zone.

Adoption

An adoption rate of 25% is used; i.e. it is assumed that, over time, 25% (750 000ha) of the

land suitable for pasture augmentation in the Ord-Victoria Area would be developed in

such a manner. This adoption rate is considered reasonable (if not conservative) and is

based on diffusion of innovation theory (Moore 1991). This adoption rate is consistent with

rates proposed for new research initiatives by MLA (2005). Adoption rates normally

increase over time, but no attempt to estimate such rates has been made here.

7.3.4.2 Valuing benefits and costs

Benefits of increasing the nitrogen status of pastures are determined using a case study

approach. Section 7.3.3 shows the results of a simulation study examining the effect of

removing nitrogen as a limitation to pasture growth at Auvergne. Auvergne is an

appropriate case study location as it is both geographically and climatically central in the

area identified by Stewart (1970a) as suitable for pasture augmentation using stylos. It is

assumed augmented pastures will last 10 years before renovation and re-establishment is

required. Benefits and costs are calculated using a partial budgeting approach; that is,

normal station running costs common to both alternatives have not been included.

Differences in animal management costs (e.g. mustering, animal husbandry, etc.) between

the alternatives are not included although these costs would require consideration in a more

comprehensive analysis. Similarly, benefits associated with increased property values due

to higher carrying capacity of augmented pastures are also not included.

Benefits

Benefits are calculated based on changes in stocking rate and liveweight gain of steers

destined for the live export trade to South East Asia, the primary turnoff market for this

region (Bortolussi et al. 2005a). In reality, much of the land in the northern VRD is

currently grazed by breeding herds but the impact of improved nutrition on breeders is

complex and difficult to quantify. Steers provide a simple means of comparison. The

analysis was based on the following assumptions.


237

Native pasture is grazed at 4AE/km2 (8% annual utilisation of 1790kg/ha of annual pasture

growth). Annual weight gain is 110kg/AE/year. Pastures are grazed at a constant stocking

rate from Year 1 onwards. The sources of these values are described below.

• 8% utilisation is derived by averaging the calculated utilisation rates for the land

systems in Table 7.4 that are suitable for legume augmentation (i.e. Angallari,

Cockatoo, Cockburn, and Frayne). The Dinnabung land system was not included as

utilisation rates are already very high and augmentation would be of little benefit.

• 1790kg/ha is the median pasture growth of native pasture at Auvergne in the presence

of trees and the current nitrogen limitation (Table 7.5).

• Annual liveweight gain of 90kg/steer/year (Bortolussi et al. 2005b) has been converted

to gain per AE using 1 steer = 0.8AE.

Augmented pasture is grazed at 17AE/km2 (16% annual utilisation of 3824kg/ha). Annual

weight gain is 170kg/AE/ha. Pastures are grazed at 4AE/km2 in Year 1, 8AE/km2 in Year

2, and 17AE/km2 from Year 3 onwards.

• 16% is the current average level of pasture utilisation across all land types in the VRD

(Section 7.2.2) and is representative of the more productive land types in the region.

Long-term sustainable utilisation rates for augmented pastures are not known for the

northern VRD, and are a major source of uncertainty in this analysis.

• 3824kg/ha is the median pasture growth of native pasture at Auvergne in the presence

of trees, but without a nitrogen limitation (Table 7.5), and is considered a good

estimate for augmented pasture on the land types considered in this analysis.

• Annual liveweight gain of 135kg/steer/year (Winter et al. 1989a) has been converted to

gain per AE using 1 steer = 0.8AE.

Sale price is $1.65/kg liveweight for both options, a relatively conservative price net of

costs of transportation from the property to the export terminal in Darwin or Wyndham

(Elders, Katherine pers.comm.).

Costs

Native pasture is the current situation and therefore has no additional costs.

Pasture augmentation practices and costs are:


238

• pasture is burnt at the beginning of the establishment season (Year 1). Cost is $10/ha;

• Stylo seed (4kg/ha @ $19/kg) and superphosphate (100kg/ha @ $760/tonne) broadcast

onto the burnt ground in Year 1. Broadcasting cost is $6/ha;

• weed control ($20/ha) undertaken in Year 2; and

• superphosphate (25kg/ha) broadcast in Year 2 and every second year thereafter.

These costs have been adapted from an economic analysis of pasture improvement in the

Katherine Region (Oxley and Walker 2003). Costs have been adjusted using changes in the

Consumer Price Index (CPI) from September 2002 (when their costs were accurate) to

September 2006. During this period, the CPI has increased by 12.5% (ATO 2006).

7.3.4.3 Calculating the net benefit and comparing alternatives

Year-to-year costs of pasture augmentation are presented in Table 7.6. Averaged over the

10 year life of the pasture, total annual costs equal $31/ha. Net benefit values are shown in

Table 7.7. Net benefit is calculated by deducting average costs from the additional benefit

gained by augmenting pasture. Net regional benefit is the net benefit multiplied by the

expected area over which augmentation is projected to occur (in this case, 750 000ha, or a

25% adoption rate over 3 million hectares).

Table 7.6 Year-to-year costs of augmenting native pasture with legumes in the northern Ord-Victoria Area

(adapted from Oxley and Walker (2003). Pastures require renovation and re-establishment after 10 years. All

values are $/ha.

Treatment Year 1

Year 2

Year 3

Year 4

Year 5

Year 6

Year 7

Year 8

Year 9

Year 10

Burning 10

Stylo seed (4kg/ha)

74

Superphosphate (100kg/ha Yr1, 25kg/ha Yr2, Yr4, etc.)

76 19 19 19 19 19

Broadcasting seed and fertiliser

6 6 6 6 6 6

Amicide (2l/ha) spot spray year after burn

20

Total 166 45 0 25 0 25 0 25 0 25

Mean total cost per year when averaged across the 10 year expected life of the pasture = $31/ha


239

Table 7.7 Projected benefit of augmenting native pastures with legumes in the northern Ord-Victoria Area.

Native pasture

Augmented pasture

Gain from augmentation

Stocking rate (AE/ha) 0.04 0.17 0.13

Liveweight gain (kg/AE/year) 110 170 60

Sale price ($/kg) 1.65 1.65 0 3

Gross value ($/ha/year) 7 43 2 36

Average costs ($/ha/year) 0 1 31 31

Net benefit ($/ha/year) 5

Area (ha) 750 000

Net regional benefit ($/year) 4.54 million 1 No additional costs to normal station management are incurred when maintaining the current native pasture

situation. 2 Takes into account the reduced stocking rate in Year 1 and Year 2. 3 Sale price is the same for both options.

7.3.4.4 Testing the effect of changes in input data

When undertaking benefit-cost analyses, Sinden and Thampapillai (1995) state: “very

rarely can all the necessary data be estimated and even more rarely can they be estimated

accurately”. Assumptions of input values are therefore necessary, and numerous

assumptions of both the benefits expected from pasture augmentation and the costs

associated with realising those benefits are included in this analysis (Section 7.3.4.2). This

section explores the effect of changes in value of the major assumptions to determine the

relative importance of each input to the eventual net regional benefit (NRB) value. The

sensitivity of each component was tested by both decreasing and increasing its value by

25% while keeping all other values constant. Assumptions tested are the stocking rates on

native and augmented pastures, liveweight gains, sale price of turnoff livestock, fertiliser

costs, and pasture augmentation adoption rate. In all cases, a 25% change is considered

feasible for all components in this analysis, justifying the use of this amount. Results are

presented in Table 7.8.


240

Table 7.8 Sensitivity of benefit-cost analysis to changes in input values. Base values are those used in

calculation of costs (Table 7.6) and projected benefits (Table 7.7). Percentage change for net regional benefit

is relative to the base value of $4.54million.

Input value Net regional benefit

($million)

Base = $4.54m

Relative change in net regional benefit

(%)

Stocking rate

Native pasture (AE/ha) Base = 0.04

-25% 0.03 5.65 24%

+25% 0.05 3.43 -24%

Augmented pasture (AE/ha) Base = 0.17

-25% 0.13 -3.63 -180%

+25% 0.21 12.71 180%

Weight gain

Native pasture (kg/AE/yr) Base = 110

-25% 83 5.65 24%

+25% 137 3.43 -24%

Augmented pasture (kg/AE/yr) Base = 170

-25% 128 -3.63 -180%

+25% 212 12.71 180%

Sale price ($/kg) Base = 1.65

-25% 1.24 -2.43 -154%

+25% 2.06 11.51 154%

Fertiliser costs ($/ha/year) Base = 17

-25% 13 7.75 71%

+25% 21 1.33 -71%

Adoption rate (%) Base = 25

-25% 19 3.41 -25%

+25% 31 5.68 25%

The greatest influence on the benefit-cost analysis is stocking rate and liveweight gain on

augmented pastures. A 25% increase in either of these components leads to a very large

increase in NRB (an increase of 180% from $4.54m to $12.71m). However, a decline of

25% leads to a net loss of $3.63m. Similarly, a change in sale price by 25% can result in

either a large gain (NRB = $11.51m if sale price increases to $2.06/kg) or a net loss (NRB

= -$2.43m if sale price falls to $1.24/kg). Increasing stocking rate or liveweight gain on

native pastures has a much lesser effect on NRB (increase of just 24% to $5.65m), but the

risk of a negative return is eliminated even if a 25% decrease occurs.


241

Changes in the major cost, fertiliser, also has a significant effect on NRB. For example, if

fertiliser costs increased by 25%, the net regional benefit declines to $1.33m. A change in

the adoption rate has a relatively low impact on NRB for the values tested here.

7.3.4.5 Final recommendation

This benefit-cost analysis shows a positive outcome from augmenting native pastures in

the northern Ord-Victoria Area, with a potential benefit of $4.54 million per year to the

regions pastoral industry. It is recommended that renewed effort by government and

funding agencies be made to facilitate the wider use of augmented pastures across the

northern VRD and other similar environments. Investing $500 000 per annum by these

agencies to bring about the potential gains identified in this analysis would result in a

benefit-cost ratio of 9:1.

Options for increasing the uptake of pasture augmentation include:

• quantification of stocking rates and liveweight gains to prove the conceptual benefits;

• improved establishment and management techniques to reduce risks of failure; and

• continued search for suitable species and cultivars (this has benefits across the whole of

northern Australia):

7.4 Discussion

7.4.1 Pasture utilisation

Calculating utilisation rates in the VRD using two separate surveys and simulation

approaches produced remarkably similar results. The first approach used GRASP to predict

pasture growth on individual land systems at specific locations, and related these to

producer’s estimates of long-term stocking rates for each land system. The second

approach used pasture growth predicted at a single location and related this to current

property-level stocking rates across the district. The outcomes of these two approaches are

discussed below.


242

The 1997 survey

Utilisation levels on individual land systems varied from 5 to 45%, averaging 16%. Some

association between land system and utilisation rate was revealed, but the relationship was

not strong. Land systems with very low utilisation rates (<10%) included several that are

commonly regarded as poor grazing country (Angallari, Cockburn, Cockatoo and

Pinkerton). These land systems are collectively known as tallgrass pasture lands (Perry

1970) and have been previously recognised as areas of low cattle productivity because the

pastures they support have low nutrient content for much of the year (Norman 1965;

Winter et al. 1989c).

Land systems with high utilisation rates (>25%) have deep soils and are only found in the

high rainfall zone of the northern VRD (Dinnabung, Ivanhoe and Willeroo). Two cases

involved utilisation rates above 35%, and Ash and Stafford Smith (1996) suggest that these

rates are not sustainable in the long term. Potential explanations for these very high

utilisation calculations include:

• They occur on small pockets of fertile grazing land in an otherwise unproductive

landscape, leading to heavy utilisation to assist with economic viability;

• producers may have provided inappropriate figures in the survey (e.g. they were unsure

of actual stocking rates, they referred to animal classes with different intake

requirements to those used in the calculations, or they quoted stocking rates that were

not sustained for the entire year);

• land types exist in a modified state due to sustained heavy utilisation while still

producing acceptable animal production. Changes in pasture composition to a more

annual species based sward increases the nutritional supply available for grazing (Ash

and McIvor 1995), and higher animal production is possible;

• relative experience of managers and their attitude to pasture condition and risk of soil

erosion; and

• errors associated with simulating pasture growth (suitability of selected parameters);

Generally, however, little difference in pasture utilisation was found across most land

systems, with over half of all calculated utilisation rates being between 10 and 20% of

average annual pasture grown. This does not indicate, however, that these land systems all


243

have a similar capacity for livestock production. Rather, utilisation rate is affected by a

number of factors, including:

• animal production being constrained by adequate nutrient levels in the pasture during

the dry season, causing cattle producers to adopt low stocking rates in order to achieve

economically viable liveweight gains, and reproductive rates high enough to maintain

self-replacing breeding herds (Norman 1965; McCown 1981a). This principally applies

to the region’s north;

• continuous grazing strategy used by most producers requires a conservative approach

to stocking rates to prevent overgrazing of preferred areas, particularly in locations

experiencing considerable year-to-year variability in pasture growth (Oxley et al.

2006). This principally applies to the region’s south; and

• differences in evenness of pastures use in large paddocks affects livestock production

efficiency (MLA 2004b).

The 2004 survey

Property-level utilisation rates showed similar variation to the land systems approach.

Calculated utilisation rates using total property grazing area and total carrying capacity

varied from 7 to 33%, averaging 16%. Over two-thirds of properties have average

utilisation rates between 10 and 20%. The strong similarity between these results and those

calculated using the land systems approach imparts considerable confidence in their

accuracy. When the expected property infrastructure improvements and associated increase

in property carrying capacities takes place over the next decade, the average utilisation rate

across the district will rise to 20%. This is approaching, but still within, the maximum level

currently considered safe for pastures in the region (Table 7.1).

Interpretation of these results does, though, require careful consideration. Scanlan et al.

(1994) points out that carrying capacity values that are fixed do not account for changes in

resource condition over time, and some producers may provide values that refer to

‘normal’ years while others refer to dry or drought years. In this study it is assumed that

producers refer to median years and therefore calculated utilisation rates relate to pasture

growth in these seasons. However, pasture growth varies from year to year in much of the

district (Figure 6.2 in Chapter 6) and as a consequence utilisation levels will also vary if a

constant stocking rate is maintained (Figure 7.3). Similarly, stocking rates will need to vary


244

from year to year if a constant utilisation level is maintained (Figure 7.4). Evaluating the

risk of exceeding desired levels of pasture utilisation at a specific stocking rate is an

important part of grazing management but until now no such means of evaluation has been

available. Table 7.9 summarises the effect of year-to-year variation in pasture growth on

utilisation rates at four different stocking rates at Victoria River Downs. From this table,

the chance of achieving or exceeding any chosen utilisation level at any of the stocking

rates included can be determined. For example: a stocking rate of 10AE/km2 will result in

utilisation rates being less than 20% in at least 70% of seasons (30th percentile); or to

achieve a utilisation rate of no more than 15% in a median season (50th percentile), a

stocking rate of 8.5AE/km2 or less is required. This is a valuable tool in analysing

proposed grazing practices.

Table 7.9 Predicted percentiles of: utilisation at four constant stocking rates; and stocking rates at four

constant levels of utilisation. Calculations based on pasture growth predictions over a 45 year period

(1959/60 to 2003/2004) at Victoria River Downs using the Regional VRD parameter set.

Season Utilisation at four stocking rates (%) Stocking rate at four utilisations (AE/km2)

percentile 5AE/km2 10AE/km2 15AE/km2 20AE/km2 10% 15% 20% 25%

1st 22 43 65 86 2.3 3.5 4.6 5.8

10th 12 23 35 47 4.3 6.4 8.6 10.7

20th 10 21 31 41 4.8 7.2 9.7 12.1

30th 10 19 29 39 5.1 7.7 10.3 12.9

40th 9 18 28 37 5.4 8.1 10.9 13.6

50th 9 18 27 35 5.6 8.5 11.3 14.1

60th 9 17 26 34 5.8 8.7 11.7 14.6

70th 8 17 25 33 6.0 9.0 12.0 15.0

80th 8 16 24 33 6.1 9.2 12.3 15.3

90th 8 16 24 31 6.4 9.5 12.7 15.9

100th 7 15 22 29 6.9 10.3 13.7 17.1

General comments on utilisation calculations

The utilisation rates calculated in this chapter include a safety factor. Annual intake of dry

matter used was 3650kgDM/AE/year (MLA 2004a) compared with 2700kgDM/AE/year

used in Queensland (Johnston et al. 1996a; Hall et al. 1998). Thus, higher utilisation rates

will result from the same stocking rate and pasture growth data when using higher animal

intake. When this is taken into consideration, the generally low utilisation rates compared


245

with those in Queensland (Figure 7.2) would be lower still if calculated using the same

intake values quoted in Hall et al. (1998).

Trees and shrubs occur on most land systems in the VRD (Stewart et al. 1970, Wilson et

al. 1990) and cattle are known to eat their leaves and soft branches, collectively known as

‘browse’ (Squires and Low 1987). The contribution of browse to cattle diets throughout the

year and from different land systems is not well understood in the VRD and, as a

consequence, is not accounted for in the utilisation calculations. However, as utilisation

rates increase in the VRD and approach maximum safe limits, accounting for consumption

of trees and shrubs may become important. Greater understanding of how cattle use this

class of vegetation in the VRD is likely to be necessary in the future, and is a

recommended focus for future research work.

Achieving a certain safe level of utilisation has both managerial and ecological difficulties.

The variation in pasture growth from year to year in the southern VRD would require large

fluctuations in animal numbers to maintain constant utilisation (Figure 7.4), and this is

practically very difficult to achieve. Bridge et al. (1983) found that sustained defoliation of

perennial tallgrass pastures on a red earth eucalypt woodland in northern Australia resulted

in the death of many individuals in a single season, and almost all plants in two years.

Results in Figure 7.3 show that several periods of potentially high utilisation have occurred

in the past 45 years and stock numbers would need to have been reduced substantially to

avoid pasture degradation. Thus, long-term sustainability of grazing in locations that

experience seasonal variability poses significant challenges to landholders, and the

capacity to account for seasonal variability using modelling has considerable application in

these circumstances.

The idea of a stable, unchanging pasture resulting from a permanent equilibrium between

pasture growth and animal consumption is disputed (Westoby et al. 1989; Behnke et al.

1993). These authors consider rangelands as dynamic systems fluctuating between a series

of states, depending on prevailing climate and disturbance factors, and any attempts to

define set rules for grazing are inappropriate. Behnke and Scoones (1993) propose a more

‘opportunistic’ approach to grazing management in African rangelands where fluctuations

in forage production are exploited, similar to the concept of maintaining constant

utilisation by varying stocking rate. The difficulty in reducing cattle numbers in times of


246

low pasture supply is acknowledged. The development of more flexible grazing practices

is likely to have benefits for cattle producers in climatically variable areas such as the

southern VRD.

Vallentine (1990) concludes that the assignment of safe utilisation rates (called ‘proper use

factors’ in his publication) for individual species, even within percentage ranges, is quite

arbitrary and open to many exceptions and variations. Utilisation rates are, nevertheless,

considered useful management tools when these limitations are recognised.

7.4.2 Alleviating the nitrogen constraint to pasture growth

The potential pasture growth outcomes of removing the nitrogen limitation that exists in

the northern VRD were explored using GRASP. Results showed that to fully realise the

climatic potential for pasture growth at Auvergne, up to 96kg/ha of plant-available

nitrogen would be required, and 61kg/ha of available N would allow unrestricted growth in

a median pasture-growing season (Table 7.5). Much of this additional N (about half) would

be captured by trees but considerable pasture growth benefits still result. For example,

median pasture growth would double from the current 1790 to 3824kg/ha. Past research on

legumes across northern Australia has shown that obtaining this level of production and the

proposed scenario to achieve it (legume augmentation) are realistic (Section 7.3.1).

Using the simulation results, a benefit-cost analysis of augmenting native pastures on low

productivity land types in the northern Ord-Victoria Area showed positive economic

benefits to the pastoral industry. Calculated benefits to the region are large ($4.54million

per year), based on a modest advantage per hectare ($5/ha) but encompassing the large

area identified by CSIRO (1970) as suitable for augmentation (30 000km2 of tallgrass

pasture lands) and an adoption rate of 25% (7 500 km2). Considerably greater area across

northern Australia is suitable for legumes and, therefore, potential benefits to the wider

pastoral economy are even larger than those proposed in this analysis.

The benefit-cost analysis includes many assumptions, and testing the sensitivity of the

analysis to each assumption indicated that stocking rate and liveweight gain from

augmented pastures are critical variables impacting on the calculation of benefits. While

some existing data is available (e.g. Winter et al. 1989c), more testing of these assumptions


247

is required. Oxley and Walker (2003) also recognise the difficulty in defining the benefits

of pasture augmentation in the region and acknowledge that it has significant impact on the

outcomes of any economic analysis.

The use of legumes to increase the productivity of pastures in the monsoon tropics is not

new. However, success has been limited due to problems with establishment and

persistence of legumes, and in recent years the focus for increasing productivity has been

through more effective use of existing native pastures. As utilisation of native pastures

approaches safe limits in the coming years, further gains in livestock production will

require exploiting the landscape in other ways. Thus, the use of legume-augmented pasture

is anticipated to re-emerge as a desired option.

A research program aimed at providing cattle producers with greater confidence to invest

in pasture augmentation should encompass several components. First, better quantification

the animal production advantages of augmented pastures would provide improved data on

which to assess benefits of widespread adoption of the practice. Such research would

demonstrate whether the concept of pasture augmentation as a means of increasing the

productivity of pastoral enterprises is valid. While most existing work has been based on

steer performance, breeders currently graze most of the area considered in this study and

defining the gains for this class of animal has rarely been explored. Defining the

productivity gains by breeders on augmented pastures would greatly improve the ability to

evaluate the suitability of pasture augmentation on breeding properties.

The second component involves improving the reliability of establishment and persistence

of legumes introduced into existing native pastures to reduce the risk of failure. Variable

establishment success, poor competitive ability, and disease susceptibility of the once-

widespread legume, Townsville stylo, was a contributing factor in its decline (Winter et al.

1989c). Overcoming these problems is critical if producers are to have confidence in

adopting the practice of pasture augmentation.

The last component should focus on continued development of robust species and cultivars

for use in an augmentation program to support would support the first two components.

The search for better legumes has benefits for the whole of northern Australia.


248

Based on an investment of $500 000 per year in the Ord-Victoria region, a benefit-cost

ratio of 9:1 was calculated if funding bodies were to invest in the proposed research. This

level of return is sufficient to encourage research agencies and funders to consider the

development of the suggested research program.

7.5 Conclusions

Current pasture utilisation rates in the VRD average 16% across a range of land systems

and climate zones. This level is conservative by northern Australian standards and within

the experimentally-derived safe limits of 25 to 30%. Utilisation rates are projected to

increase to about 20% over the next decade through more efficient use of pasture

resources, and this planned increase appears to be sustainable.

Wide-scale augmentation of native pastures with legume species has the potential to

increase the productivity and value of the pastoral industry in the region. However, pasture

augmentation is not without risk as costs associated with establishment and management,

and potential benefits to be realised contain some uncertainty. Research to reduce this

uncertainty will have positive benefits and is recommended.

The two examples presented in this chapter demonstrate that the new capacity to predict

pasture growth in the VRD using GRASP has application for analysing grazing

management systems at a range of scales. At the regional scale, exploring the outcomes of

scenarios that affect the primary factors influencing pasture growth can be used in the

development of policies to bring about any desired changes. At the property and land

system level, GRASP can be used to analyse current and proposed grazing practices.

Chapter 8 Integrating discussion and final conclusions of study

249

8.0 Integrating discussion and final conclusions of study

8.1 Introduction

The development and improvement of sustainable grazing practices in the Victoria River

District (VRD) is essential for the continued prosperity of the region’s pastoral industry.

With a district cattle population of over 500 000 and the likelihood of an increase over the

coming years, analysis of current and proposed grazing practices to evaluate their long-

term sustainability is essential. Such analyses require reliable data on pasture production

for the main land systems, and over the range of seasons likely to be experienced. Prior to

this study the availability of pasture production data in the region did not adequately fulfil

these requirements and was a severe limitation when analysing grazing practices. This

study has overcome that limitation by developing the capacity to predict native pasture

growth in the VRD. With this new capacity, pasture growth can now be predicted for

individual land types and properties, and at the broader regional scale.

The aim of this chapter is to discuss the significance of this predictive capacity to analysis

of grazing practices in the VRD. It is not intended to repeat the specific findings of the

individual chapters, but rather to discuss them in a broader context. Readers should refer to

previous chapters for greater detail. A brief overview of the outcomes of this thesis is

presented in Section 8.2, before the implications to the pastoral industry in the region are

discussed. The systems modelling approach adopted in this study is limited by both the

accuracy of the data used to quantify the model relationships, and the adequacy of the

model structure itself. Understanding these limitations is important when applying and

interpreting simulations results and Section 8.3 addresses this issue. A number of

recommendations for future research to build upon the outcomes of this study are outlined

in Section 8.4. Finally, the major conclusions from this study are presented in Section 8.5.

8.2 The capacity to predict pasture growth in the VRD

Overview

By using the GRASP model with input parameters derived from local data, pasture growth

can now be predicted in the VRD for any time period; from the pattern of growth within a

single growing season, to seasonal pasture production over many years. When GRASP is

calibrated using site-specific field data, the soil water balance, pasture growth and nutrient


250

uptake are predicted with a high level of skill. These results indicate that the modelling

framework is capable of mimicking the main processes influencing pasture growth on a

variety of soil and pasture types in the region.

Using generic parameters to predict pasture growth in specific instances is often much less

reliable than using site-specific parameters and significant prediction errors can occur.

Pasture growth in the VRD is often limited by the supply of nitrogen, particularly in the

northern part of the district where available soil water is rarely a limiting factor to plant

growth. Thus, accurate quantification of soil nitrogen supply and the physiological limit to

which pastures can dilute nitrogen before growth stops are essential for best model

performance. Trees also have a significant impact on predictions of pasture growth, and for

GRASP to reliably predict pasture growth when analysing grazing practices, their effect

must be included.

Applications of the current capacity to predict pasture growth

Many potential applications exist for the capacity to predict pasture growth under a wide

variety of conditions. However, the suitability of GRASP for these applications depends

upon the scale of the issue (location-specific or a more regional view) and the reliability

with which the model can simulate the desired conditions. Chapter 5 found that using

generic parameters often resulted in individual model predictions of pasture growth

deviating considerably from field measurements, but using the average of many predictions

better matched the observations. This result indicates that the most appropriate applications

of GRASP in the VRD are those where summary values of many predictions are used

rather than any single prediction. Such an approach assumes that individual prediction

errors (under- and over-predictions) negate each other when the average or median is

calculated.

Chapter 6 showed that predictions of pasture growth in the VRD are sensitive to the

relationship between green cover and green yield, transpiration use efficiency, nitrogen

uptake, and internal nitrogen dilution. Using site-specific values for these parameters in an

otherwise generic parameter set would improve the predictive skill of GRASP (although

this approach has not been specifically tested in this thesis), and expand the potential

applications of the model. Where site-specific values for these parameters are available,

individual predictions can be used with much greater confidence.


251

A summary of potential applications of a predictive model of pasture growth, and the

current suitability of GRASP for these applications is presented in Table 8.1. For each

application, comment is provided on the considerations necessary when attempting to use

GRASP with parameters developed in this study.

Table 8.1 Summary of the current suitability of GRASP for application to analysis of grazing practices in the

VRD when using parameters developed in this study.

Scale Application Current suitability?

Considerations

Land system or paddock

Defining year-to-year variability in seasonal

pasture growth

Yes Only for land systems included in

this study (and those similar), and

where reasonable estimates of N

supply are available. Some testing

of model performance in low rainfall

years is required

Long-term utilisation levels Yes Use average or median seasonal

pasture growth values (e.g. Section

7.2 in Chapter 7)

Analysis of alternative grazing options Yes Where relative differences between

options is desired rather than

quantitative accuracy

Tactical grazing decisions No Quantitative accuracy for individual

predictions is required, and

therefore site-specific parameters

are needed. This is difficult to

achieve in real-time

Property Using seasonal climate forecasting for

management decisions

No As above

Defining risks associated with variability in

the commencement, duration and cessation

of the growing season

Yes Pattern of growth due to climate

influences is adequately simulated

using generic parameters

Region Investigating effects of climate change Yes Where relative differences between

scenarios is desired rather than

quantitative accuracy

Monitoring seasonal conditions Yes General trends okay, but need

further testing for low rainfall

seasons

Development of grazing policies Yes Use average or median seasonal

pasture growth values

Extrapolating grazing trial results Yes Dependent on generality of grazing

trial


252

8.3 Limitations of the systems modelling approach used in this study

Modelling is a simplification of real life so there will always be limitations in the ability of

any modelling framework to account for all the components that interact in a system. The

intention of the GRASP model (as applied in this study) was to combine established

biophysical relationships with climate data to provide estimates of pasture production,

accounting for differences in soil type, species composition and prevailing climate. The

biophysical relationships were quantified using purpose-collected field data, and climate

data was either collected on-site or derived from an interpolated meteorological dataset.

Using this framework introduces two forms of limitation: 1) the accuracy of the data used

in calibrating the relationships that make up the model and the data defining the

environment within which the model operates; and 2) the structure of the model and its

ability to mimic the real system. These limitations are discussed in the following

paragraphs.

Limitations associated with field data

Chapter 3 described the outcomes of the field study designed to collect the data necessary

for quantifying the relationships in GRASP. A number of features of the field study

impacted upon the calibration of model parameters. Seasonal rainfall was median to above

median, and this has advantages for deriving model parameters such as soil water holding

capacity, transpiration use efficiency, and potential nitrogen uptake and dilution. However,

when validating the model using the jackknifing technique (using the same dataset for both

calibration and validation), the model was not able to be tested for performance in low

rainfall seasons. Dimes et al. (1996) emphasises that field studies designed to provide data

for model quantification and use should, ideally, span the extremes of conditions over

which the model is intended to be applied. This was not accomplished during this study

and, therefore, the reliability of GRASP when applied during dry or drought periods in the

VRD has not been determined. It was shown, however, that simulations using the historical

climate record indicate model output is sensitive to low rainfall years using the parameters

derived in this study (Figure 6.2 in Chapter 6).

Soil water holding capacities of the cracking clays in this study were high compared to

published data from comparable soils across northern Australia, and this was largely

attributed to the difficulty in accurately determining field capacity in these slow-draining


253

soils. This is a common problem for cracking clays (Berndt and Coughlan 1976; Isbell

1983; Smiles 1997) and is a limitation to accurate prediction of the water balance on these

soils, thereby potentially affecting predictions of pasture growth. More research is required

to accurately define the upper and lower limits of available soil water in cracking clays in

the VRD.

A limited number of land systems were studied due to time, resource, and site access

restrictions. Many other land systems exist in the VRD and these contain a wide variety of

soil and pasture types, and within these land systems species composition and soil surface

characteristics can vary due to effects of grazing and seasonal conditions. Calibrating

GRASP using data from just a few locations allows general relationships to be established,

but does not provide for all soil conditions, pasture types or species mixes. For example,

pastures dominated by spinifex (Triodia Spp. and Plechtrachne spp.), and annual sorghum

(Sorghum spp.) have not been measured and these are likely to exhibit different growth

traits to the species included in this study. Similarly, the field study did not include yellow

earths, sandy soils or desert soils; all likely to exhibit some physical and chemical

properties different to those reported in Chapter 3. Reliable application of the parameter

sets developed in this study is therefore limited to soil and pasture types with

characteristics similar to those from which the parameters have been derived.

Chapter 2 discussed how rainfall is the most important climate variable influencing pasture

growth, and that other variables (mainly temperature) can be influential at times. The field

study collected rainfall from each site or the closest station homestead, providing a mostly

accurate record. Other variables (maximum and minimum temperature, pan evaporation,

solar radiation and atmospheric vapour pressure) were obtained from the Data Drill

(DataDrill 2005). This data is interpolated from weather recording stations often many

(sometimes hundreds) of kilometres away from the study sites and it is reasonable to

assume that, on occasions, the interpolated data differs to some extent from the actual

conditions at the sites. Accuracy of data used to define the climatic environment within

which the model is being applied is potentially another limit to accurate model

performance. This is exacerbated when rainfall data is also derived from interpolated data.

Therefore, for best model performance, the climate data (particularly rainfall) should be

sourced from as close to the location of interest as possible.


254

Limitations associated with model structure

GRASP is described as a ‘simple model of pasture production’ (Rickert et al. 2000) despite

the many equations it contains and the 100+ parameters requiring input values for its

operation. This description infers that processes involved in pasture production have been

omitted, are general in nature, or are yet to be incorporated into the model structure. The

authors of GRASP acknowledge a range of limitations in the model structure (Section

4.2.3 in Chapter 4), including no provision for changes over time of species composition or

tree density, and no accounting for root growth or seed production (Littleboy and McKeon

1997). The section restricts discussion to the limitations most relevant to the VRD, both

previously acknowledged and those only now arising from the results of this study.

Chapter 4 to Chapter 6 document the calibration and testing of GRASP when applied to the

VRD, and in so doing, a number of deficiencies in the model structure were identified.

Plant density (measured as perennial grass basal area, PGBA) is used by the model to

estimate growth early in the growing season when low green cover is present to intercept

radiation or transpire soil water. Predictions of pasture growth for the remainder of the

season are dependent on the ability of plants to grow early in the season, and therefore

PGBA has some influence on predictions of pasture growth throughout the whole growth

cycle. Results of the field study showed little correlation between grass basal area and total

pasture growth in the VRD, primarily due to the variable contribution of annual grasses

and forbs to standing biomass. Additionally, these species groups have different

temperature response, radiation use, transpiration use and nutrient acquisition

characteristics (Fitzpatrick and Nix 1970; Ludlow 1985; Schmidt et al. 1998), and

attempting to use single values for mixed swards inevitably leads to prediction errors.

Predictions of plant cover in mixed swards are also difficult to simulate due to differences

in species architecture and growth patterns. It has been previously noted that simulation of

perennial grasses and more ephemeral species (annual grasses and forbs) as separate

biomass pools would likely improve model predictions (Johnston 1996), and to date this

issue remains unresolved. Results in this study reinforce Johnston’s position by again

highlighting the difficulty in accurately predicting pasture production using one biomass

pool in swards containing a high proportion of species whose abundance varies

considerably from year to year.


255

Simultaneous wetting of the entire soil profile in cracking clays through direct entry of

surface water to the lower soil layers was highlighted in Chapter 4 as the likely source of

prediction errors for soil water and subsequent pasture growth on these soils early in the

wet season. Presently, GRASP does not include the capacity to predict simultaneous

profile wetting. The SWAP model (van Dam et al. 1997) does include this process and

uses rainfall intensity to determine the amount of water exceeding the surface infiltration

rate, which is then run into soil cracks. However, their model uses one minute time steps

and is reliant on accurate rainfall intensity data, neither of which is practical at northern

Australian scales. GRASP already contains a rainfall intensity function as part of

calculating surface runoff, although it is a relatively basic function operating in 15 minute

intervals and based on the time of the year, with rainfall assumed more intense in summer

than winter. Modifying this function to predict water running into soil cracks may have

some application and is worthy of further investigation.

Nitrogen uptake and internal dilution were identified in Chapter 6 as major limiting factors

to plant growth, particularly in the northern VRD, and this is supported by previous

experimental work (e.g. Norman 1962). Comparison of model predictions and field

measurements of annual pasture production over eight years in a highly nitrogen limited

environment showed that greater year-to-year variability was measured in the field than

was predicted by GRASP (Figure 6.12). This suggests that maximum N supply and the

point to which plants can dilute nitrogen before growth ceases is highly unlikely to be

constant from year to year. Therefore, a more dynamic and mechanistic model of nitrogen

dynamics is likely to lead to better long-term predictions of pasture growth in nitrogen

limited environments. Development of such a model is strongly recommended based on

the results of this study.

Currently GRASP only recognises nitrogen as a limiting nutrient to pasture growth so, in

effect, nitrogen acts a proxy for all nutrients. For grass dominated pastures, this approach

has been generally been adequate. Past research shown that phosphorus is also important to

pasture growth in semi-arid Australia (Norman 1962; McIvor 1984), but these trials show

the response of pastures to N and P supply often differ, with forbs (mainly legumes)

responding to additions of P, but less so to N (e.g. Friedel et al. 1980). As pastures in the

VRD often contain significant proportions of forbs, the current absence of a separate


256

phosphorus supply index when calculating pasture growth is another factor likely to lead to

prediction errors.

8.4 Recommendations for future work

The previous section discussed the limitations of the field data available for accurate

calibration of GRASP in the VRD, and limitations of the model structure itself. During this

discussion, a number of recommendations for improving the structure of the model were

made to address these limitations. Recommendations fall into two categories: field studies

to improve calibration of the current model; and improvements to model structure. These

are summarised here.

Field studies to improve model calibration

• Testing model performance with field data collected during low rainfall years;

• expanding the number of land systems for which calibration data is available;

• defining the water holding capacity of cracking clay soils more precisely; and

• increasing the spatial intensity of data collection on nitrogen supply and internal plant

dilution.

Improvements to model structure

• Simulating separate biomass pools for perennial grasses and ephemeral species;

• developing a dynamic model for year-to-year variation in nitrogen supply and dilution;

• modifying the rainfall intensity and runoff components to simulate simultaneous

wetting of the whole soil profile in cracking clays; and

• incorporating a phosphorus index to better describe the effect of nutrient supply on

pasture growth.

These lists are not exhaustive and other less critical suggestions for future work are made

throughout the thesis.

This section has recommended a series of future research projects aimed at improving the

capability of GRASP when predicting pasture growth in the VRD, and some of these

recommendations include modifications to the model structure. One consequence of

increasing the complexity of the model by including these recommended changes is the


257

effect on its calibration. A more complex model requires a larger number of parameters to

be calibrated, and therefore more field data to derive values for these parameters. Thus,

increasing the complexity of GRASP must be balanced against the additional effort

required to calibrate it so that the model become more useful, rather than more accurate but

less able to be calibrated and applied.

8.5 Final conclusion

The capacity to predict pasture growth in the VRD has now been developed and tested,

overcoming the lack of measured pasture production data available for the VRD.

Predictions can now account for variability in pasture growth due to:

• soil type;

• pasture species composition;

• climate regime; and

• time.

Limitations in the ability to predict pasture growth have been identified, based primarily on

accounting for the spatial and temporal variation in nitrogen supply and dilution by

pastures. As much of the VRD experiences a nitrogen-induced constraint to pasture

growth, model performance will always be determined by the adequacy of the field data

used to initially describe nitrogen supply, and the ability of the model to predict the

dynamics of this supply over time.

Despite this limitation, the outcome of this study represents an important advance in the

provision of pasture production data for use in the development and analysis of grazing

practices in the Victoria River District.

Appendices

Appendices

Appendix 1 Photographs of study sites and specific events

Plate 1 Nalluvial fl

Plate 2 Cspp. pastu

258

orthern Victoria River District landscape showing escarpment country, woodland plains and

ats.

entral Victoria River District landscape showing cattle grazing Chrysopogon fallax and Iseilema

re on cracking clay soil.

Appendices

259

Plate 3 Southern Victoria River District landscape showing open grassy plains with scattered trees and

occasional rocky basalt outcrops.

Plate 4 Land showing evidence of past heavy grazing: bare and scalded ground, dead trees, and annual

pasture species.

Appendices

260

Plate 5 Fencing a study site to exclude cattle from grazing.

Plate 6 Burning to remove carryover material at commencement of the study period (Site 16).

Plate 7 Marking out sampling cells at site establishment (Site 2).

Appendices

261

Plate 8 a) Bulk density sampling in soil profile pit; and b) hand-augering soil moisture cores during field

measurements.

Plate 9 a) Measuring perennial grass basal area using a 5-point frame; and b) pasture sampling during field

measurements.

a. b.

a. b.

Appendices

262

Plate 10 Red earth overlying basalt (Site 19).

Plate 11 Red earth overlying limestone (Site 11).

Appendices

263

Plate 12 Cracking clay overlying basalt (Site 4).

Plate 13 Alluvial cracking clay (Site 7).

Appendices

264

Plate 14 Barley Mitchell grass pasture (Site 18).

Plate 15 Overhead view of 1m2 quadrat in a) barley Mitchell grass pasture (Site 18); and b) ribbon grass

pasture (Site 8).

Plate 16 Ribbon grass pasture (Site 8).

a. b.

Appendices

265

Plate 17 White grass pasture (Site 14).

Plate 18 Overhead view of 1m2 quadrat in a) white grass pasture (Site 14); and b) annual short grass pasture

(Site 1).

Plate 19 Annual short grass pasture (Site 1).

a. b.

Appendices

266

Plate 20 Forb pasture (Site 4).

Plate 21 a) Overhead view of 1m2 quadrat in a forb pasture (Site 4); and b) an example of the abundant,

diverse forbs present at Site 15 on 2 May 1995.

Plate 22 Inundated conditions at Site 15 on 9 March 1995.

a. b.

Appendices

267

Plate 23 Annual short grass pasture early in the wet season with no burning at site establishment (Site 13, 12

Jan 1995).

Plate 24 Overhead view of 1m2 quadrats in an annual short grass pasture (Site 13) during early wet season a)

with no burning at site establishment; and b) after burning to remove carryover material.

Plate 25 Annual short grass pasture early in the wet season after site burnt to remove carryover material (Site

13, 9 Jan 1996). Fire most likely destroyed much of the seed bank, resulting in poor germination.

a. b.

Appendices

268

Appendix 2 Study sites classified according to previous resource

inventories of the VRD

A number of systems exist for classification of land in the VRD using soil, vegetation and

topography. These have been described in Section 2.1 and 2.3.3 of Chapter 2. Each site in

this study has been classified using these systems so results can be easily compared with

future work (Table 9.1 to Table 9.4).

Table 9.1 Study sites classified using: 1 Perry (1970); 2 Stewart et al. (1970); and 3 DIPE (unpublished).

Site Code Pasture Land 1 Land System 2 Land Unit 3 Geology 3

1 An Hilly Country with Useful Lowlands Antrim G6a Basalt

2 An Hilly Country with Useful Lowlands Antrim G6a Basalt

3 Wv Mitchell and Other Grass Plains Wave Hill G6a Basalt




7 A Mitchell and Other Grass Plains Argyle L7l Alluvium




11 Hu Hilly Country with Useful Lowlands Humbert G6d2 Limestone

12 D Tippera Tall Grass Plains Dinnabung G6d2 Limestone

13 Hu Hilly Country with Useful Lowlands Humbert G5d Limestone

14 D Tippera Tall Grass Plains Dinnabung G5d Limestone

15 Iv Blue Grass Plains Ivanhoe L7l Alluvium

16 Iv Blue Grass Plains Ivanhoe L7l Alluvium


18 Wv Mitchell and Other Grass Plains Wave Hill L6a3 Basalt

19 An Hilly Country with Useful Lowlands Antrim G5a9 Basalt

20 Iv Blue Grass Plains Ivanhoe L7l* Alluvium

21 Iv Blue Grass Plains Ivanhoe L7l* Alluvium

Appendices

269

Table 9.2 Study sites classified according to: 1 Northcote (1979); 2 Northcote et al. (1975); 3 Stace et al.

(1968); 4 Isbell (1996); 5 McDonald et al. (1990).

Site No.

Factual Key 1

Soil Texture 1

A Description of Aust. Soils 2

Handbook of Aust. Soils 3

Classification 4 Landform 5

1 Gn 3.12 Gradational Red Smooth-ped Earths Euchrozems Dermosols Hills


3 Ug 5.37 Uniform Red Self-mulching

Cracking Clays

Red clays Vertosols Plains


Cracking Clays


5 Ug 5.32 Uniform Brown Self-mulching

Cracking Clays

Brown clays Vertosols Plains


Cracking Clays


7 Ug 5.24 Uniform Grey Self-mulching

Cracking Clays

Grey clays Vertosols Plains


Cracking Clays



Cracking Clays



Cracking Clays


11 Gc 1.22 Gradational Calcareous earths Solonized brown

soils

Calcarosols Plains

12 Gn 3.12 Gradational Red Smooth-ped Earths Euchrozems Dermosols Plains

13 Gn 3.12 Gradational Red Smooth-ped Earths Euchrozems Dermosols Rises

14 Gn 2.13 Gradational Red Massive Earths Calcareous red

earths

Kandosols Rises


Cracking Clays



Cracking Clays


17 Ug 5.32 Uniform Brown Self-mulching

Cracking Clays

Brown clays Vertosols Plains


Cracking Clays




Cracking Clays



Cracking Clays


Appendices ------------------------------------------------------------------------------------------------------------------------------------------ Table 9.3 Study sites classified according to Wilson et al. (1990).

____


_________________________________________________________________ 270

Appendices ------------------------------------------------------------------------------------------------------------------------------------------ Table 9.4 Study sites classified according to Tothill and Gillies (1992).

____


_________________________________________________________________ 271

Appendices

272

Appendix 3 Results of soil chemistry analysis

Soil chemical properties were measured for the 0-10cm and 10-20cm soil layers (Section 3.3.5) and full results are presented here. Summary

values appear in the main body of the thesis.

Table 9.5 Laboratory analysis results of soil samples from sites located on red earths overlying basalt.

Site Depth pH1 EC2 Cl3 C Sand4 F Sand5 Silt6 Clay7 Ca8 Mg9 Na10 K11 ECEC12 OC13 NO3-N14 P15 SO4-S16 N(t)17 P(t)18 K(t)19 S(t)20 Cu21 Zn22 Mn23

1 0-10cm 7.5 0.05 1.0 17 44 10 31 10.0 19.0 0.07 1.30 30 0.73 2.0 6.0 3.0 0.05 0.022 2.280 0.014 2.40 0.89 32.0

10-20cm 7.7 0.04 1.0 15 39 13 35 12.0 21.0 0.13 0.52 34 0.64 1.0 2.0 6.0 0.03 0.016 2.050 0.014 2.20 2.20 23.0

2 0-10cm 7.8 0.02 1.0 23 45 8 26 11.0 16.0 0.10 0.74 28 0.51 1.0 6.0 2.0 0.04 0.025 2.100 0.013 2.00 0.44 24.0

10-20cm 7.5 0.03 1.0 14 40 12 35 13.0 19.0 0.15 0.58 33 0.53 1.0 4.0 4.0 0.03 0.022 2.000 0.016 5.00 18.00 5.1

19 0-10cm 7.3 0.06 1.0 14 47 14 27 13.0 14.0 0.07 0.76 28 0.67 1.0 13.0 11.0 0.03 0.031 1.690 0.019 2.00 3.90 81.0

10-20cm 7.2 0.08 1.0 13 37 12 41 15.0 17.0 0.13 0.49 33 0.69 1.0 6.0 15.0 0.04 0.021 1.380 0.020 2.70 3.10 83.0

Units of measurement in Table 9.5 to Table 9.8 1 pH - soil pH 9 Mg - magnesium (mequiv/100g) 17 N(t) - total nitrogen (%) 2 EC - electrical conductivity (dS m–1) 10 Na - sodium (mequiv/100g) 18 P(t) - total phosphorus (%) 3 Cl - soluble chloride (mg kg–1) 11 K - potassium (mequiv/100g) 19 K(t) - total potassium (%) 4 C Sand - coarse sand fraction (%) 12 ECEC - effective cation exchange capacity (mequiv/100g, Ca + Mg + Na + K) 5 F Sand - fine sand fraction (%) 13 OC - organic carbon (%) 20 S(t) - total sulphur (%) 6 Silt - silt fraction (%) 14 NO3-N - nitrate-nitrogen (mg kg–1) 21 Cu - copper (mg kg–1) 7 Clay - clay fraction (%) 15 P - phosphorus, bicarbonate extractable (Colwell) (mg kg–1) 22 Zn - zinc (mg kg–1) 8 Ca - calcium (mequiv/100g) 16 SO4-S – sulphate-sulphur, extractable (mg kg–1) 23 Mn - Manganese (mg kg–1)

Standard analysis methods of Rayment and Higginson (1982).

Appendices

273

Table 9.6 Laboratory analysis results of soil samples from sites located on red earths overlying limestone.


11 0-10cm 8.3 0.09 1.0 2 55 12 33 16.0 5.1 0.03 0.79 22 0.97 1.0 11.0 2.0 0.06 0.020 2.660 0.019 0.47 0.64 15.0

10-20cm 8.4 0.12 1.0 3 51 10 38 28.0 5.0 0.05 0.64 34 0.81 1.0 8.0 2.0 0.02 0.022 2.730 0.016 0.91 0.80 13.0

12 0-10cm 7.6 0.03 1.0 3 33 15 50 9.5 4.9 0.04 1.00 15 1.10 1.0 13.0 4.0 0.05 0.021 3.060 0.015 1.50 0.50 35.0

10-20cm 7.6 0.03 1.0 2 27 13 60 11.0 4.8 0.02 0.92 17 0.89 1.0 8.0 6.0 0.04 0.020 3.100 0.014 1.50 0.31 31.0

13 0-10cm 7.3 0.04 1.0 3 53 14 31 8.9 3.6 0.03 0.76 13 0.73 1.0 13.0 5.0 0.04 0.018 1.916 0.014 0.51 5.48 18.0

10-20cm 7.9 0.03 3.0 2 44 9 43 14.0 4.4 0.04 0.66 19 0.70 1.0 12.0 5.0 0.03 0.017 2.300 0.014 0.45 3.50 5.8

14 0-10cm 7.3 0.03 3.0 5 57 17 26 5.2 3.3 0.03 0.44 9 0.87 1.0 11.0 4.0 0.03 0.018 1.860 0.013 0.48 3.70 13.0

10-20cm 7.5 0.03 3.0 2 44 10 46 7.2 4.1 0.03 0.66 12 0.76 4.0 5.0 3.0 0.04 0.018 2.390 0.014 0.77 1.20 6.5

Units of measurement are shown in the footnote of Table 9.5.

Table 9.7 Laboratory analysis results of soil samples from sites located on cracking clays overlying basalt.


3 0-10cm 8.3 0.10 1.0 4 29 13 52 23.0 20.0 0.68 1.10 45 0.53 1.0 8.0 2.0 0.03 0.013 1.210 0.013 1.50 0.83 14.0

10-20cm 8.7 0.12 1.0 6 26 12 56 25.0 21.0 1.30 0.90 48 0.59 2.0 6.0 3.0 0.02 0.010 1.150 0.011 1.90 0.85 8.8

4 0-10cm 8.5 0.07 1.0 6 34 10 52 35.0 16.0 0.13 1.00 52 0.45 1.0 7.0 2.0 0.03 0.013 1.040 0.011 0.89 0.49 9.2

10-20cm 8.4 0.08 2.0 6 31 10 52 36.0 16.0 0.18 0.93 53 0.36 1.0 7.0 2.0 0.02 0.011 1.000 0.011 1.40 0.89 9.0

5 0-10cm 8.3 0.07 1.0 4 30 12 53 31.0 17.0 0.22 0.97 49 0.37 1.0 7.0 3.0 0.02 0.012 0.661 0.013 1.20 0.64 12.0

10-20cm 8.5 0.08 1.0 3 30 12 55 31.0 18.0 0.33 0.77 50 0.41 1.0 7.0 4.0 0.02 0.009 0.627 0.012 1.20 0.58 9.5

6 0-10cm 8.0 0.04 1.0 4 35 13 50 24.0 16.0 0.06 1.30 41 0.48 1.0 8.0 2.0 0.02 0.012 1.020 0.013 5.20 19.00 13.0

10-20cm 7.9 0.03 1.0 4 34 11 52 28.0 18.0 0.13 1.10 47 0.41 1.0 6.0 2.0 0.03 0.009 1.030 0.010 1.40 0.86 19.0

17 0-10cm 7.3 0.07 1.0 10 29 15 48 23.0 16.0 0.22 0.81 40 0.64 1.0 13.0 11.0 0.03 0.014 0.868 0.015 5.20 15.00 131.0

10-20cm 7.2 0.09 1.0 9 27 17 50 25.0 17.0 0.30 0.56 43 0.50 1.0 8.0 15.0 0.02 0.011 0.839 0.020 4.10 21.00 77.0

18 0-10cm 7.8 0.09 1.0 6 26 13 57 34.0 19.0 0.19 1.30 54 0.55 2.0 13.0 18.0 0.03 0.012 0.905 0.013 1.00 5.00 43.0

10-20cm 7.7 0.08 1.0 6 25 13 56 34.0 19.0 0.23 1.10 54 0.42 1.0 8.0 14.0 0.02 0.010 0.875 0.012 0.87 5.90 36.0


Appendices

274

Table 9.8 Laboratory analysis results of soil samples from sites located on cracking clays of alluvial origin.


7 0-10cm 8.3 0.06 1.0 1 35 13 51 28.0 12.0 0.38 0.86 41 0.55 1.0 8.0 3.0 0.02 0.010 1.320 0.011 1.00 1.80 14.0

10-20cm 8.4 0.06 1.0 1 33 13 53 29.0 12.0 0.59 0.75 42 0.47 1.0 8.0 3.0 0.02 0.008 1.310 0.010 1.10 1.10 11.0

8 0-10cm 8.3 0.10 1.0 3 36 14 50 28.0 9.7 0.19 0.96 39 0.59 2.0 8.0 4.0 0.04 0.014 1.620 0.013 1.40 0.88 12.0

10-20cm 8.5 0.09 1.0 2 33 12 52 30.0 11.0 0.23 0.83 42 0.47 1.0 7.0 5.0 0.02 0.013 1.630 0.010 1.20 0.58 9.0

9 0-10cm 8.1 0.05 1.0 2 40 3 57 23.0 8.9 0.17 0.45 33 0.45 1.0 4.0 31.0 0.02 0.007 0.360 0.010 1.00 1.10 30.0

10-20cm 8.2 0.04 1.0 2 39 13 48 25.0 9.2 0.27 0.33 35 0.33 1.0 4.0 4.0 0.01 0.006 0.324 0.009 1.00 0.87 16.0

10 0-10cm 8.2 0.04 1.0 2 42 13 44 22.0 9.1 0.22 0.41 32 0.45 1.0 4.0 4.0 0.02 0.005 0.334 0.009 1.00 0.96 26.0

10-20cm 8.1 0.04 1.0 2 40 13 46 23.0 9.0 0.26 0.35 33 0.33 1.0 3.0 5.0 0.01 0.006 0.317 0.008 0.93 0.70 18.0

15 0-10cm 8.7 0.06 1.0 3 20 16 61 28.0 14.0 0.50 0.62 43 0.67 1.0 7.0 6.0 0.04 0.017 1.790 0.011 0.62 5.20 3.6

10-20cm 8.9 0.07 1.0 3 20 14 63 27.0 14.0 0.94 0.50 42 0.66 5.0 7.0 2.0 0.03 0.016 1.740 0.011 0.54 2.60 2.5

16 0-10cm 7.9 0.03 1.0 6 40 11 46 20.0 16.0 0.08 0.63 37 0.70 1.0 6.0 1.0 0.04 0.008 0.989 0.011 0.53 5.80 4.8

10-20cm 8.1 0.03 1.0 6 39 11 46 20.0 16.0 0.08 0.65 37 0.45 1.0 4.0 1.0 0.03 0.008 1.020 0.011 0.53 7.80 5.0

20 0-10cm 8.0 0.07 1.0 3 44 18 39 20.0 7.5 0.08 0.49 28 0.58 2.0 10.0 7.0 0.03 0.011 0.519 0.012 0.59 7.20 20.0

10-20cm 8.4 0.07 1.0 3 41 15 42 21.0 8.6 0.13 0.34 30 0.51 1.0 8.0 4.0 0.03 0.009 0.524 0.011 0.47 5.10 10.0

21 0-10cm 6.9 0.07 6.0 4 48 12 36 10.0 7.3 0.10 0.62 18 0.48 1.0 6.0 9.0 0.02 0.010 0.507 0.011 0.89 5.70 119.0

10-20cm 7.1 0.04 1.0 4 46 13 39 11.0 7.2 0.11 0.48 19 0.37 1.0 4.0 4.0 0.01 0.007 0.491 0.009 0.65 4.60 47.0


Appendices

275

Many of the soil nutrient analysis results in Table 9.5 to Table 9.8 are reported as weight of

nutrient per unit weight of soil (i.e. mg/kg, equivalent to ‘parts per million’). Converting

these values to weight of nutrient per unit area (i.e. kg/ha) provides quantitative values for

the amount of nutrients present in the landscape. The major nutrients of carbon, nitrogen

and phosphorus have been converted in this way and are presented in Table 9.9. The

conversion procedure is detailed in Dalgliesh and Foale (1998).

Table 9.9 Amount of soil nutrients present in surface layers of study sites. (Table continued overleaf)

Site Depth OC 1 NO3-N 2 N(t) 3 P(t) 4 C:N ratio 5

Red earths overlying basalt

1 0-10cm 10512 2.88 720 317 14.6

10-20cm 9664 1.51 453 242 21.3

20-30cm 10560 1.65 495 264 21.3

2 0-10cm 7548 1.48 592 370 12.8

10-20cm 7632 1.44 432 317 17.7

20-30cm 8003 1.51 453 332 17.7

19 0-10cm 10586 1.58 474 490 22.3

10-20cm 9867 1.43 572 300 17.3

20-30cm 10419 1.51 604 317 17.3

Red earths overlying limestone

11 0-10cm 15423 1.59 954 318 16.2

10-20cm 11988 1.48 296 326 40.5

20-30cm 11988 1.48 296 326 40.5

12 0-10cm 16060 1.46 730 307 22.0

10-20cm 14062 1.58 632 316 22.3

20-30cm 13083 1.47 588 294 22.3

13 0-10cm 12410 1.70 680 306 18.3

10-20cm 11550 1.65 495 280 23.3

20-30cm 11830 1.69 507 287 23.3

14 0-10cm 13659 1.57 471 283 29.0

10-20cm 11932 6.28 628 283 19.0

20-30cm 12768 6.72 672 302 19.0

Units of measurement: 1 OC - organic carbon (kg/ha) 2 NO3-N - nitrate-nitrogen (kg/ha) 3 N(t) - total nitrogen (kg/ha) 4 P(t) - total phosphorus (kg/ha) 5 C:N ratio – carbon to nitrogen ratio

Appendices

276



Cracking clays overlying basalt

3 0-10cm 6254 1.18 354 153 17.7

10-20cm 7198 2.44 244 122 29.5

20-30cm 8909 3.02 302 151 29.5

4 0-10cm 6075 1.35 405 176 15.0

10-20cm 5148 1.43 286 157 18.0

20-30cm 5472 1.52 304 167 18.0

5 0-10cm 5143 1.39 278 167 18.5

10-20cm 6109 1.49 298 134 20.5

20-30cm 6232 1.52 304 137 20.5

6 0-10cm 6048 1.26 252 151 24.0

10-20cm 5822 1.42 426 128 13.7

20-30cm 6355 1.55 465 139 13.7

17 0-10cm 8640 1.35 405 189 21.3

10-20cm 7350 1.47 294 162 25.0

20-30cm 7850 1.57 314 173 25.0

18 0-10cm 7095 2.58 387 155 18.3

10-20cm 5628 1.34 268 134 21.0

20-30cm 5838 1.39 278 139 21.0

Appendices

277

Table 9.9 (cont.)


Cracking clays of alluvial origin

7 0-10cm 8250 1.50 300 150 27.5

10-20cm 6580 1.40 280 112 23.5

20-30cm 6674 1.42 284 114 23.5

8 0-10cm 9086 3.08 616 216 14.8

10-20cm 7896 1.68 336 218 23.5

20-30cm 7896 1.68 336 218 23.5

9 0-10cm 6840 1.52 304 106 22.5

10-20cm 5907 1.79 179 107 33.0

20-30cm 5775 1.75 175 105 33.0

10 0-10cm 7110 1.58 316 79 22.5

10-20cm 5049 1.53 153 92 33.0

20-30cm 5709 1.73 173 104 33.0

15 0-10cm 10050 1.50 600 255 16.7

10-20cm 10362 7.85 471 251 22.0

20-30cm 9702 7.35 441 235 22.0

16 0-10cm 9240 1.32 528 106 17.5

10-20cm 6930 1.54 462 123 15.0

20-30cm 6930 1.54 462 123 15.0

20 0-10cm 9744 3.36 504 185 19.3

10-20cm 9180 1.80 540 162 17.0

20-30cm 9180 1.80 540 162 17.0

21 0-10cm 7296 1.52 304 152 24.0

10-20cm 5624 1.52 152 106 37.0

20-30cm 5624 1.52 152 106 37.0

Appendices

278

Appendix 4 Plant species nomenclature

Species identification nomenclature used in this study is taken from Dunlop (1990) and

Barritt (1996).

Table 9.10 Plant species nomenclature for all individual species presented in this study.

Abelmoschus ficulneus (L.) Wight & Arn. ex Wight

Abutilon andrewsianum W. Fitzg.

Abutilon otocarpum F. Muell.

Aristida holathera Domin

Aristida inaequiglumis Domin

Aristida latifolia Domin

Astrebla elymoides F. Muell.

Astrebla pectinata (Lindley) F. Muell. ex Benth.

Astrebla squarrosa C. Hubb.

Atalaya hemiglauca (F. Muell.) F. Muell. ex Benth.

Basilicum polystachyon (L.) Moench

Boerhavia dominii Meikle & Hewson

Bonamia pannosa (R. Br.) Hallier f.

Brachyachne convergens (F. Muell.) Stapf

Carissa lanceolata R. Br.

Chionachne hubbardiana Henrard

Chrysogonum trichodesmoides (F. Muell.) F. Muell.

Chrysopogon fallax S.T. Blake

* Citrullus colocynthis (L.) Schrader

Corymbia terminalis (F. Muell.) K.D. Hill & L.A.S. Johnson

Crotalaria medicaginea Lam.

Dactyloctenium radulans (R. Br.) Beauv.

Dichanthium fecundum S.T. Blake

Dichanthium sericeum (R. Br.) A. Camus

Echinochloa elliptica Michael & Vickery

Enneapogon polyphyllus (Domin) N. Burb.

Eriachne aristidae F. Muell.

Eucalyptus brevifolia F. Muell.

Eucalyptus pruinosa Schauer

Eulalia aurea (Bory) Kunth

Euphorbia schultzii Benth.

Euphorbia vachellii Hook. & Arn.

Evolvulus alsinoides L.

Flemingia pauciflora Benth.

Gomphrena canescens R. Br.

Gossypium australe F. Muell.

Grevillea striata R. Br.

Hakea arborescens R. Br.

Heliotropium tenuifolium R. Br.

Heteropogon contortus (L.) Beauv. ex Roemer & Schultes

Indigofera colutea (Burman f.) Merrill

Indigofera hirsuta L.

Indigofera trita L. f.

Ipomoea polymorpha Roemer & Schultes

Iseilema fragile S.T. Blake

Iseilema vaginiflorum Domin

Jacquemontia browniana Ooststr.

Lysiphyllum cunninghamii (Benth.) De Wit

Neptunia monosperma F. Muell.

Panicum decompositum R. Br.

Phyllanthus urinaria L.

Polymeria ambigua R. Br.

Psoralea pustulata F. Muell.

Ptilotus exaltatus Nees

Rhyncosia australis Benth.

Rhyncosia minima (L.) DC.

Salsola kali L.

Sehima nervosum (Rottler.) Stapf

Sesbania cannabina (Retz.) Poiret

Sida cleistocalyx F. Muell.

Sorghum stipoideum (Ewart & J.W. White) C. Gardner &

C. Hubb

Sporobolus australasicus Domin

* Stylosanthes humilis Kunth

Tephrosia macrocarpa Benth.

Terminalia arostrata Ewart & O. Davies

Terminalia canescens (DC.) Radlk.

Terminalia volucris R. Br. ex Benth.

Trichodesma zeylanicum (Burman f.) R. Br.

(* denotes introduced species)

Appendices

279

Appendix 5 Description of complete species composition at the study sites

The entire floristic composition of each site was described once (July 1995). A botanist

(Jack Cusack, CSIRO Division of Wildlife and Ecology) identified all tree, shrub, and

pasture species present in the immediate area, and provided an indication of their

abundance. Table 9.11 shows an example of the floristic description for a site (Site 1).

Table 9.11 Complete list of plant species present at Site 1, July 1995.

Abundance Scientific Name Common Name Common Patchy Sparse

Brachyachne convergens Native couch X

Aristida latifolia Feathertop wiregrass X

Enneapogon polyphyllus Limestone grass X

Indigofera trita X

Salsola kali Roly poly X

Sida cleistocalyx X

Abutilon otocarpum Desert lantern X

Abutilon andrewsianum X

Astrebla elymoides Weeping Mitchell grass X

Atalaya hemiglauca Whitewood X

Boerhavia dominii Boerhavia X

Carissa lanceolata Conkerberry X

Chrysopogon fallax Golden beard grass X

Citrullus colocynthis Bush cucumber X

Corymbia terminalis Northern bloodwood X

Crotalaria medicaginea Rattlepod X

Dichanthium sericeum Queensland bluegrass X

Echinochloa elliptica Nasty grass X

Euphorbia vachellii X

Evolvulus alsinoides X

Grevillea striata X

Heliotropium tenuifolium Heliotrope X

Indigofera colutea X

Iseilema fragile Flinders grass X

Iseilema vaginiflorum Red flinders grass X

Jacquemontia browniana Snake stem X

Panicum decompositum Native millet X

Phyllanthus urinaria Phyllanthus X

Psoralea pustulata X

Rhyncosia minima Native rhyncosia X

Sporobolus australasicus Fairy grass X

Tephrosia macrocarpa X

Terminalia arostrata Nutwood X

Appendices

280

Appendix 6 Pasture field data in addition to that presented in Chapter 3

Most pasture data from the study sites are presented in Section 3.4.3. Results of species

composition and nitrogen dynamics from three sites representative of each group appear in

the relevant sections (Section 3.4.3.4 to Section 3.4.3.8). Data from sites not included in

those sections are presented here. In addition, data on the flow of dry matter through the

three biomass pools (green material, dead material and detached litter) are also presented.

Figure 9.1 Pasture composition results not presented in Chapter 3. Error bars indicate the standard error of

the site mean for total standing dry matter at each sampling time (harvest). (Figure continued overleaf)

Site 3

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 7

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 16

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Appendices

281

Figure 9.1 (cont.)

Site 11

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 12

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Site 19

0

1000

2000

3000

4000


TSD

M (k

g/ha

)


Appendices

282

Figure 9.2 Pasture nitrogen contents (dashed lines) and nitrogen uptake (vertical bars) of sites not presented

in Chapter 3. (Figure continued overleaf)



Site 3

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 7

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 16

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0N

Con

tent

(%)

Appendices

283

Figure 9.2 (cont.)

Site 11

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)

Site 12

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

N C

onte

nt (%

)Site 19

0

10

20

30

40

50

60


N U

ptak

e (k

g/ha

)

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0N

Con

tent

(%)

Appendices

284

Appendix 7 Descriptive statistics for total standing dry matter

measurements

Table 9.12 Descriptive statistics for total standing dry matter data collected at each harvest for all sites. Eight

1m2 quadrats were cut at each harvest. All units are kg/ha. (Table continued overleaf)

Site Harvest Date Mean quadrat

yield

Minimum quadrat

yield

Maximum quadrat

yield

Standard error of mean

95% confidence limit of mean

1 1 02/02/1994 735 354 1577 ±140.6 ±332.5

2 11/03/1994 1887 1185 3023 ±215.3 ±509.1

3 05/05/1994 1951 1119 2651 ±165.5 ±391.2

4 14/09/1994 2165 1416 2938 ±204.6 ±483.7

5 04/01/1995 294 87 821 ±92.4 ±218.4

6 14/03/1995 2233 453 4098 ±409.8 ±969.0

7 26/04/1995 2952 1955 4797 ±398.9 ±943.4

8 15/11/1995 Not measured

2 1 02/02/1994 533 185 1263 ±127.3 ±301.0

2 11/03/1994 1013 783 1273 ±66.3 ±156.8

3 04/05/1994 1977 1147 3351 ±238.7 ±564.5

4 14/09/1994 1931 892 2648 ±218.0 ±515.5

5 04/01/1995 71 0 270 ±33.3 ±78.8

6 15/03/1995 2022 462 3508 ±339.3 ±802.3

7 27/04/1995 2424 1427 3434 ±262.9 ±621.6

8 29/11/1995 1413 541 2252 ±193.7 ±458.0

3 1 02/02/1994 1091 754 1615 ±115.8 ±273.7

2 12/03/1994 1430 562 2235 ±178.4 ±421.8

3 04/05/1994 1861 1124 2322 ±132.8 ±314.1

4 15/09/1994 2058 1288 3396 ±277.9 ±657.1

5 04/01/1995 76 31 115 ±9.8 ±23.3

6 14/03/1995 1654 1079 2125 ±122.0 ±288.6

7 26/04/1995 2544 1600 3796 ±308.8 ±730.2

8 17/11/1995 1573 1097 2002 ±97.4 ±230.2

4 1 01/02/1994 1121 745 1696 ±101.7 ±240.4

2 12/03/1994 1807 1395 2618 ±146.4 ±346.3

3 04/05/1994 2343 1805 3122 ±160.6 ±379.7

4 13/09/1994 1232 756 1913 ±150.6 ±356.1

5 05/01/1995 78 21 157 ±20.0 ±47.2

6 13/03/1995 2950 2212 4282 ±294.7 ±696.8

7 27/04/1995 2672 2074 4941 ±327.9 ±775.3

8 15/11/1995 1747 986 3565 ±335.4 ±793.0

Appendices

285



yield

Minimum quadrat

yield

Maximum quadrat

yield



5 1 01/02/1994 1288 749 1838 ±150.0 ±354.6

2 13/03/1994 2515 1056 4374 ±399.1 ±943.8

3 04/05/1994 2738 1253 3797 ±311.2 ±736.0

4 13/09/1994 3291 2005 4958 ±392.3 ±927.7

5 05/01/1995 174 115 303 ±21.3 ±50.4

6 14/03/1995 2228 1127 3856 ±338.9 ±801.3

7 26/04/1995 3471 2143 5234 ±304.5 ±720.1

8 15/11/1995 2719 1772 5364 ±397.8 ±940.7

6 1 01/02/1994 1551 395 3710 ±378.9 ±895.9

2 13/03/1994 2502 1208 4230 ±355.3 ±840.0

3 04/05/1994 2013 658 4804 ±452.0 ±1068.9

4 13/09/1994 2270 872 5628 ±538.5 ±1273.3

5 05/01/1995 38 12 69 ±7.5 ±17.7

6 14/03/1995 2040 657 3350 ±320.9 ±758.8

7 26/04/1995 3330 1179 5602 ±520.2 ±1230.2

8 28/11/1995 2658 823 4568 ±444.7 ±1051.6

7 1 21/01/1994 1534 712 3592 ±309.8 ±732.6

2 17/03/1994 2595 2039 3487 ±202.1 ±477.9

3 09/05/1994 2553 1688 3873 ±241.8 ±571.8

4 20/09/1994 2248 1524 3815 ±272.9 ±645.2

5 11/01/1995 641 446 890 ±61.6 ±145.8

6 07/03/1995 2557 2201 3129 ±117.7 ±278.4

7 04/05/1995 3661 2134 6014 ±496.6 ±1174.3

8 17/10/1995 2033 953 2762 ±210.5 ±497.8

8 1 21/01/1994 1561 878 2024 ±132.7 ±313.9

2 17/03/1994 2744 1031 3427 ±281.0 ±664.4

3 09/05/1994 3426 2934 4792 ±226.0 ±534.5

4 20/09/1994 2847 1985 3436 ±161.5 ±382.0

5 10/01/1995 727 406 1283 ±110.4 ±261.2

6 08/03/1995 2501 1719 2959 ±130.5 ±308.6

7 03/05/1995 4282 3545 5056 ±212.2 ±501.8

8 17/10/1995 2682 1710 3758 ±284.9 ±673.6

Appendices

286



yield

Minimum quadrat

yield

Maximum quadrat

yield



9 1 19/01/1994 826 320 1299 ±112.3 ±265.6

2 18/03/1994 2008 1259 3238 ±213.5 ±504.9

3 10/05/1994 1862 1388 2970 ±171.2 ±404.8

4 21/09/1994 1626 325 3310 ±318.0 ±751.9

5 11/01/1995 546 183 858 ±75.6 ±178.7

6 08/03/1995 2298 1304 4221 ±340.3 ±804.7

7 04/05/1995 2527 1874 3789 ±208.2 ±492.4

8 18/10/1995 1314 651 1568 ±105.0 ±248.4

10 1 18/01/1994 1060 681 1726 ±134.1 ±317.0

2 18/03/1994 2279 1842 3173 ±175.0 ±413.8

3 10/05/1994 2666 963 5343 ±547.2 ±1293.9

4 21/09/1994 1590 1037 2347 ±135.6 ±320.7

5 11/01/1995 476 232 857 ±81.1 ±191.9

6 08/03/1995 2233 1330 3047 ±211.5 ±500.2

7 04/05/1995 2763 1790 4417 ±336.7 ±796.3

8 17/10/1995 1542 1232 2097 ±102.0 ±241.1

11 1 20/01/1994 897 439 1426 ±115.6 ±273.3

2 16/03/1994 2683 1673 3856 ±259.8 ±614.4

3 10/05/1994 2707 1595 5311 ±420.1 ±993.4

4 22/09/1994 2567 1135 4247 ±369.1 ±872.8

5 10/01/1995 700 56 1348 ±145.8 ±344.7

6 06/03/1995 2039 686 3080 ±266.3 ±629.8

7 03/05/1995 2345 1343 3830 ±370.6 ±876.4

8 17/10/1995 2365 786 3060 ±297.2 ±702.8

12 1 20/01/1994 768 66 1621 ±179.2 ±423.7

2 17/03/1994 1923 1162 2563 ±164.8 ±389.6

3 10/05/1994 2418 896 4121 ±365.8 ±865.1

4 22/09/1994 2248 446 4622 ±486.7 ±1150.8

5 12/01/1995 630 133 1647 ±183.7 ±434.4

6 07/03/1995 2003 984 4700 ±413.4 ±977.6

7 03/05/1995 3032 1100 4804 ±477.1 ±1128.1

8 17/10/1995 2537 586 5568 ±717.9 ±1697.5

Appendices

287



yield

Minimum quadrat

yield

Maximum quadrat

yield



13 1 12/01/1995 518 238 894 ±80.5 ±190.4

2 07/03/1995 2529 1577 3431 ±293.9 ±694.9

3 04/05/1995 2593 1656 3315 ±194.5 ±459.8

4 17/10/1995 2085 1187 4072 ±318.6 ±753.3

5 09/01/1996 132 13 268 ±36.5 ±86.2

6 29/02/1996 530 55 877 ±113.5 ±268.4

7 28/05/1996 854 375 1250 ±100.0 ±236.4

8 26/11/1996 366 116 829 ±80.9 ±191.4

14 1 12/01/1995 316 0 610 ±73.9 ±174.7

2 06/03/1995 1858 1735 1988 ±35.6 ±84.1

3 03/05/1995 2396 187 3503 ±377.9 ±893.7

4 17/10/1995 1627 804 2176 ±181.6 ±429.3

5 10/01/1996 385 181 721 ±72.3 ±171.0

6 29/02/1996 920 151 1509 ±176.7 ±417.7

7 28/05/1996 1531 202 2737 ±310.6 ±734.4

8 26/11/1996 1581 0 2747 ±346.4 ±819.2

15 1 13/01/1995 419 195 655 ±52.3 ±123.6

2 09/03/1995 2758 2025 4118 ±284.1 ±671.9

3 02/05/1995 2696 1339 4985 ±398.0 ±941.2

4 18/10/1995 1965 440 7689 ±851.3 ±2012.9

5 10/01/1996 1071 781 1435 ±75.7 ±179.0

6 28/02/1996 1499 648 2174 ±184.9 ±437.1

7 30/05/1996 2680 1689 4102 ±274.2 ±648.5

8 26/11/1996 1138 0 2402 ±368.8 ±872.2

16 1 13/01/1995 357 129 593 ±57.5 ±135.9

2 09/03/1995 2840 1927 3366 ±160.0 ±378.2

3 02/05/1995 3093 2619 3474 ±91.0 ±215.2

4 18/10/1995 1882 1556 2569 ±112.9 ±266.9

5 10/01/1996 566 370 849 ±68.9 ±162.9

6 28/02/1996 1044 557 1801 ±132.8 ±314.1

7 30/05/1996 2682 1877 4469 ±287.9 ±680.9


Appendices

288



yield

Minimum quadrat

yield

Maximum quadrat

yield



17 1 14/12/1994 160 101 378 ±32.2 ±76.2

2 08/02/1995 1543 1032 2579 ±181.5 ±429.1

3 20/04/1995 2315 877 4645 ±497.4 ±1176.1

4 03/11/1995 1379 573 3183 ±302.6 ±715.6

5 05/01/1996 193 76 311 ±27.6 ±65.3

6 15/02/1996 1053 492 2929 ±276.5 ±653.8

7 06/06/1996 1762 970 3734 ±311.2 ±736.0

8 28/11/1996 1313 794 1976 ±135.4 ±320.2

18 1 15/12/1994 405 115 564 ±46.2 ±109.3

2 08/02/1995 2290 342 3750 ±358.1 ±846.7

3 19/04/1995 3359 2469 5436 ±336.2 ±795.1

4 02/11/1995 2696 1394 3726 ±278.8 ±659.3

5 04/01/1996 360 234 472 ±36.2 ±85.5

6 14/02/1996 1384 940 1848 ±126.0 ±298.0

7 05/06/1996 2254 705 3648 ±432.4 ±1022.5

8 28/11/1996 1521 1044 1981 ±106.5 ±251.9

19 1 14/12/1994 138 17 303 ±34.8 ±82.4

2 07/02/1995 1281 200 2168 ±211.6 ±500.4

3 19/04/1995 1873 278 6211 ±677.0 ±1600.8

4 02/11/1995 1054 186 2225 ±235.8 ±557.6

5 04/01/1996 269 10 1546 ±184.5 ±436.3

6 14/02/1996 355 43 954 ±134.2 ±317.4

7 06/06/1996 736 207 1173 ±124.5 ±294.5


20 1 13/12/1994 130 8 239 ±28.3 ±67.0

2 09/02/1995 1599 618 2727 ±227.0 ±536.8

3 21/04/1995 2621 1371 3496 ±260.4 ±615.8

4 01/11/1995 1993 1038 2888 ±230.9 ±546.1

5 03/01/1996 443 163 683 ±65.1 ±153.9

6 13/02/1996 1518 487 3229 ±328.4 ±776.5

7 04/06/1996 3973 2697 6556 ±447.9 ±1059.2


Appendices

289

Table 9.12 (cont.)


yield

Minimum quadrat

yield

Maximum quadrat

yield



21 1 13/12/1994 42 14 64 ±6.2 ±14.6

2 09/02/1995 1302 743 1906 ±142.0 ±335.9

3 20/04/1995 2031 1189 2747 ±172.4 ±407.6

4 01/11/1995 1125 648 2000 ±158.7 ±375.3

5 03/01/1996 260 129 427 ±43.7 ±103.2

6 13/02/1996 1164 685 1650 ±113.1 ±267.4

7 04/06/1996 1588 1148 2054 ±110.8 ±262.1


Appendices

290

Appendix 8 Results of model calibration

Table 9.13 Observed and predicted total standing dry matter (TSDM) values from GRASP model

calibration. (Table continued overleaf)

Site Harvest Observed value

(kg/ha)

95% Conf limit

(kg/ha)

95% Conf limit as

proportion of

observed value (%)

Predicted value

(kg/ha)

Predicted value as a proportion

of the observed

value

Deviation of

prediction from

observed value

(kg/ha)

Predicted value

within 95% conf limit

of observed

value?

1 1 735 ±332.5 ±45.2% 754 1.03 19 Yes

2 1887 ±509.1 ±27.0% 1654 0.88 -233 Yes

3 1951 ±391.2 ±20.1% 2256 1.16 305 Yes

4 2165 ±483.7 ±22.3% 2223 1.03 58 Yes

5 294 ±218.4 ±74.2% 336 1.14 42 Yes

6 2233 ±969.0 ±43.4% 2021 0.90 -212 Yes

7 2952 ±943.4 ±32.0% 2978 1.01 26 Yes

8 Not measured

2 1 533 ±301.0 ±56.4% 507 0.95 -26 Yes

2 1013 ±156.8 ±15.5% 1372 1.35 359 No

3 1977 ±564.5 ±28.5% 2080 1.05 103 Yes

4 1931 ±515.5 ±26.7% 1928 1.00 -3 Yes

5 71 ±78.8 ±111.8% 342 4.85 271 No

6 2022 ±802.3 ±39.7% 1744 0.86 -278 Yes

7 2424 ±621.6 ±25.6% 2410 0.99 -14 Yes

8 1413 ±458.0 ±32.4% 1584 1.12 171 Yes

3 1 1091 ±273.7 ±25.1% 697 0.64 -394 No

2 1430 ±421.8 ±29.5% 1172 0.82 -258 Yes

3 1861 ±314.1 ±16.9% 1902 1.02 41 Yes

4 2058 ±657.1 ±31.9% 2035 0.99 -23 Yes

5 76 ±23.3 ±30.7% 461 6.07 385 No

6 1654 ±288.6 ±17.5% 1609 0.97 -45 Yes

7 2544 ±730.2 ±28.7% 2493 0.98 -51 Yes

8 1573 ±230.2 ±14.6% 1681 1.07 108 Yes

Appendices

291



(kg/ha)

95% Conf limit

(kg/ha)

95% Conf limit as

proportion of

observed value (%)

Predicted value

(kg/ha)


of the observed

value

Deviation of

prediction from

observed value

(kg/ha)

Predicted value


of observed

value?

4 1 1121 ±240.4 ±21.4% 563 0.50 -558 No

2 1807 ±346.3 ±19.2% 1680 0.93 -127 Yes

3 2343 ±379.7 ±16.2% 2277 0.97 -66 Yes

4 1232 ±356.1 ±28.9% 1231 1.00 -1 Yes

5 78 ±47.2 ±60.9% 398 5.13 320 No

6 2950 ±696.8 ±23.6% 2979 1.01 29 Yes

7 2672 ±775.3 ±29.0% 2836 1.06 164 Yes

8 1747 ±793.0 ±45.4% 1633 0.93 -114 Yes

5 1 1288 ±354.6 ±27.5% 1348 1.05 60 Yes

2 2515 ±943.8 ±37.5% 2225 0.88 -290 Yes

3 2738 ±736.0 ±26.9% 3152 1.15 414 Yes

4 3291 ±927.7 ±28.2% 3388 1.03 97 Yes

5 174 ±50.4 ±28.9% 455 2.61 281 No

6 2228 ±801.3 ±36.0% 2110 0.95 -118 Yes

7 3471 ±720.1 ±20.7% 3352 0.97 -119 Yes

8 2719 ±940.7 ±34.6% 2779 1.02 60 Yes

6 1 1551 ±895.9 ±57.8% 634 0.41 -917 No

2 2502 ±840.0 ±33.6% 1427 0.57 -1075 No

3 2013 ±1068.9 ±53.1% 2524 1.25 511 Yes

4 2270 ±1273.3 ±56.1% 2309 1.02 39 Yes

5 38 ±17.7 ±46.6% 261 6.87 223 No

6 2040 ±758.8 ±37.2% 2041 1.00 1 Yes

7 3330 ±1230.2 ±36.9% 3281 0.99 -49 Yes

8 2658 ±1051.6 ±39.6% 2802 1.05 144 Yes

Appendices

292



(kg/ha)

95% Conf limit

(kg/ha)

95% Conf limit as

proportion of

observed value (%)

Predicted value

(kg/ha)


of the observed

value

Deviation of

prediction from

observed value

(kg/ha)

Predicted value


of observed

value?

7 1 1534 ±732.6 ±47.8% 1654 1.08 120 Yes

2 2595 ±477.9 ±18.4% 2014 0.78 -581 No

3 2553 ±571.8 ±22.4% 2535 0.99 -18 Yes

4 2248 ±645.2 ±28.7% 2231 0.99 -17 Yes

5 641 ±145.8 ±22.8% 1067 1.67 426 No

6 2557 ±278.4 ±10.9% 2900 1.13 343 No

7 3661 ±1174.3 ±32.1% 3693 1.01 32 Yes

8 2033 ±497.8 ±24.5% 2319 1.14 286 Yes

8 1 1561 ±313.9 ±20.1% 1275 0.82 -286 Yes

2 2744 ±664.4 ±24.2% 3324 1.21 580 Yes

3 3426 ±534.5 ±15.6% 3433 1.00 7 Yes

4 2847 ±382.0 ±13.4% 2836 1.00 -11 Yes

5 727 ±261.2 ±35.9% 890 1.23 163 Yes

6 2501 ±308.6 ±12.3% 2991 1.20 490 No

7 4282 ±501.8 ±11.7% 4330 1.01 48 Yes

8 2682 ±673.6 ±25.1% 2684 1.00 2 Yes

9 1 826 ±265.6 ±32.2% 578 0.70 -248 Yes

2 2008 ±504.9 ±25.1% 2042 1.02 34 Yes

3 1862 ±404.8 ±21.7% 2004 1.08 142 Yes

4 1626 ±751.9 ±46.3% 1774 1.09 148 Yes

5 546 ±178.7 ±32.8% 540 0.99 -6 Yes

6 2298 ±804.7 ±35.0% 2389 1.04 91 Yes

7 2527 ±492.4 ±19.5% 2539 1.00 12 Yes

8 1314 ±248.4 ±18.9% 1490 1.13 176 Yes

Appendices

293



(kg/ha)

95% Conf limit

(kg/ha)

95% Conf limit as

proportion of

observed value (%)

Predicted value

(kg/ha)


of the observed

value

Deviation of

prediction from

observed value

(kg/ha)

Predicted value


of observed

value?

10 1 1060 ±317.0 ±29.9% 803 0.76 -257 Yes

2 2279 ±413.8 ±18.2% 2409 1.06 130 Yes

3 2666 ±1293.9 ±48.5% 2760 1.04 94 Yes

4 1590 ±320.7 ±20.2% 1698 1.07 108 Yes

5 476 ±191.9 ±40.3% 664 1.39 188 Yes

6 2233 ±500.2 ±22.4% 1745 0.78 -488 Yes

7 2763 ±796.3 ±28.8% 2778 1.01 15 Yes

8 1542 ±241.1 ±15.6% 1644 1.07 102 Yes

11 1 897 ±273.3 ±30.5% 791 0.88 -106 Yes

2 2683 ±614.4 ±22.9% 2530 0.94 -153 Yes

3 2707 ±993.4 ±36.7% 2723 1.01 16 Yes

4 2567 ±872.8 ±34.0% 2552 0.99 -15 Yes

5 700 ±344.7 ±49.3% 685 0.98 -15 Yes

6 2039 ±629.8 ±30.9% 1952 0.96 -87 Yes

7 2345 ±876.4 ±37.4% 2519 1.07 174 Yes

8 2365 ±702.8 ±29.7% 2484 1.05 119 Yes

12 1 768 ±423.7 ±55.1% 644 0.84 -124 Yes

2 1923 ±389.6 ±20.3% 2138 1.11 215 Yes

3 2418 ±865.1 ±35.8% 2390 0.99 -28 Yes

4 2248 ±1150.8 ±51.2% 2241 1.00 -7 Yes

5 630 ±434.4 ±69.0% 717 1.14 87 Yes

6 2003 ±977.6 ±48.8% 2220 1.11 217 Yes

7 3032 ±1128.1 ±37.2% 2914 0.96 -118 Yes

8 2537 ±1697.5 ±66.9% 2656 1.05 119 Yes

Appendices

294



(kg/ha)

95% Conf limit

(kg/ha)

95% Conf limit as

proportion of

observed value (%)

Predicted value

(kg/ha)


of the observed

value

Deviation of

prediction from

observed value

(kg/ha)

Predicted value


of observed

value?

13 1 518 ±190.4 ±36.8% 585 1.13 67 Yes

2 2529 ±694.9 ±27.5% 2726 1.08 197 Yes

3 2593 ±459.8 ±17.7% 2596 1.00 3 Yes

4 2085 ±753.3 ±36.1% 2149 1.03 64 Yes

5 132 ±86.2 ±65.4% 186 1.41 54 Yes

6 530 ±268.4 ±50.6% 327 0.62 -203 Yes

7 854 ±236.4 ±27.7% 815 0.95 -39 Yes

8 366 ±191.4 ±52.4% 480 1.31 114 Yes

14 1 316 ±174.7 ±55.2% 406 1.28 90 Yes

2 1858 ±84.1 ±4.5% 1382 0.74 -476 No

3 2396 ±893.7 ±37.3% 2458 1.03 62 Yes

4 1627 ±429.3 ±26.4% 1785 1.10 158 Yes

5 385 ±171.0 ±44.4% 448 1.16 63 Yes

6 920 ±417.7 ±45.4% 786 0.85 -134 Yes

7 1531 ±734.4 ±48.0% 745 0.49 -786 No

8 1581 ±819.2 ±51.8% 1581 1.00 0 Yes

15 1 419 ±123.6 ±29.5% 212 0.51 -207 No

2 2758 ±671.9 ±24.4% 2028 0.74 -730 No

3 2696 ±941.2 ±34.9% 2763 1.02 67 Yes

4 1965 ±2012.9 ±102.4% 2107 1.07 142 Yes

5 1071 ±179.0 ±16.7% 971 0.91 -100 Yes

6 1499 ±437.1 ±29.2% 2054 1.37 555 No

7 2680 ±648.5 ±24.2% 2705 1.01 25 Yes

8 1138 ±872.2 ±76.6% 1257 1.10 119 Yes

Appendices

295



(kg/ha)

95% Conf limit

(kg/ha)

95% Conf limit as

proportion of

observed value (%)

Predicted value

(kg/ha)


of the observed

value

Deviation of

prediction from

observed value

(kg/ha)

Predicted value


of observed

value?

16 1 357 ±135.9 ±38.0% 371 1.04 14 Yes

2 2840 ±378.2 ±13.3% 2384 0.84 -456 No

3 3093 ±215.2 ±7.0% 3105 1.00 12 Yes

4 1882 ±266.9 ±14.2% 2077 1.10 195 Yes

5 566 ±162.9 ±28.8% 441 0.78 -125 Yes

6 1044 ±314.1 ±30.1% 962 0.92 -82 Yes

7 2682 ±680.9 ±25.4% 2940 1.10 258 Yes

8 Not measured

17 1 160 ±76.2 ±47.7% 306 1.92 146 No

2 1543 ±429.1 ±27.8% 1698 1.10 155 Yes

3 2315 ±1176.1 ±50.8% 2344 1.01 29 Yes

4 1379 ±715.6 ±51.9% 1335 0.97 -44 Yes

5 193 ±65.3 ±33.9% 298 1.55 105 No

6 1053 ±653.8 ±62.1% 745 0.71 -308 Yes

7 1762 ±736.0 ±41.8% 1785 1.01 23 Yes

8 1313 ±320.2 ±24.4% 1254 0.95 -59 Yes

18 1 405 ±109.3 ±27.0% 420 1.04 15 Yes

2 2290 ±846.7 ±37.0% 1953 0.85 -337 Yes

3 3359 ±795.1 ±23.7% 3427 1.02 68 Yes

4 2696 ±659.3 ±24.5% 2841 1.05 145 Yes

5 360 ±85.5 ±23.8% 519 1.44 159 No

6 1384 ±298.0 ±21.5% 1090 0.79 -294 Yes

7 2254 ±1022.5 ±45.4% 2331 1.03 77 Yes

8 1521 ±251.9 ±16.6% 1868 1.23 347 No

Appendices

296

Table 9.13 (cont.)


(kg/ha)

95% Conf limit

(kg/ha)

95% Conf limit as

proportion of

observed value (%)

Predicted value

(kg/ha)


of the observed

value

Deviation of

prediction from

observed value

(kg/ha)

Predicted value


of observed

value?

19 1 138 ±82.4 ±59.5% 177 1.28 39 Yes

2 1281 ±500.4 ±39.1% 988 0.77 -293 Yes

3 1873 ±1600.8 ±85.5% 1892 1.01 19 Yes

4 1054 ±557.6 ±52.9% 1128 1.07 74 Yes

5 269 ±436.3 ±162.2% 183 0.68 -86 Yes

6 355 ±317.4 ±89.4% 333 0.94 -22 Yes

7 736 ±294.5 ±40.0% 884 1.20 148 Yes

8 Not measured

20 1 130 ±67.0 ±51.7% 127 0.98 -3 Yes

2 1599 ±536.8 ±33.6% 869 0.54 -730 No

3 2621 ±615.8 ±23.5% 2664 1.02 43 Yes

4 1993 ±546.1 ±27.4% 2016 1.01 23 Yes

5 443 ±153.9 ±34.8% 664 1.50 221 No

6 1518 ±776.5 ±51.2% 1480 0.98 -38 Yes

7 3973 ±1059.2 ±26.7% 3963 1.00 -10 Yes

8 Not measured

21 1 42 ±14.6 ±35.0% 264 6.35 222 No

2 1302 ±335.9 ±25.8% 1190 0.91 -112 Yes

3 2031 ±407.6 ±20.1% 2017 0.99 -14 Yes

4 1125 ±375.3 ±33.4% 1206 1.07 81 Yes

5 260 ±103.2 ±39.7% 329 1.27 69 Yes

6 1164 ±267.4 ±23.0% 814 0.70 -350 No

7 1588 ±262.1 ±16.5% 1624 1.02 36 Yes

8 Not measured

Appendices

297

Appendix 9 Results of sensitivity testing

A tabulated summary of the results from the sensitivity testing of important parameters

undertaken in Chapter 6 is presented here.

Table 9.14 Summary of results from parameter sensitivity analysis. Data presented is end of growing season

pasture TSDM. All units are kg/ha.

Minimum Maximum Median

Transpiration related parameters

Green yield at 50% green cover

500 kg/ha 1695 3344 3343

1000 kg/ha 1712 3344 3342

2000 kg/ha 1677 3344 3259

Variation

Transpiration use efficiency

9 kg/ha/mmT 1703 3344 3121

12 kg/ha/mmT 1820 3344 3343

15 kg/ha/mmT 1895 3344 3343

Variation 192 0 222

Soil water index at which growth stops

0.01 1877 3344 3343

0.1 1841 3344 3342

0.3 1562 3344 3342

Variation 315 0 1

Nitrogen related parameters

N uptake per 100mm transpiration

7 kgN/ha/100mmT 1257 3344 3061

11 kgN/ha/100mmT 1806 3344 3342

15 kgN/ha/100mmT 1808 3344 3342

Variation 551 0 281

Maximum N uptake

15 kg/ha 1540 2089 2088

20 kg/ha 1689 2785 2783

25 kg/ha 1786 3480 3478

30 kg/ha 1796 4176 3760

Variation 107 1391 977

N content at which growth stops

0.5 % 1818 4805 3806

0.75 % 1732 3204 3202

1.0 % 1031 2403 2402

Variation 787 2402 1404

Appendices

298

Appendix 10 Parameter files used in modelling process

Default parameter file used to initialise GRASP gvt89def.prv default PARAMETER values for gunsynd from ron gun 21/7/1998 ron gun is at brian pastures research station, gayndah First number = parameter code number Second number = parameter value 001 000.0 Site number allocated by Kenneth A Day 07 38969576 002 000.0 Tothill & Gillies local production unit SOIL PARAMETERS 020 100.0 Thickness (mm) of soil layer 1 (surface 100mm approx) which can be air dried. Nemonic = SW(8,1). 021 400.0 Thickness (mm) of soil layer 2. This layer cannot dry below permanent wilting point, and is the main zone of root activity. Nemonic = SW(8,2). 022 500.0 Thickness (mm) of soil layer 3. The lower limit of this layer is the limit of root penetration ( =SW(8,3)). 026 36.0 SW(2,1) Layer 1 maximum soil moisture (mm). 027 174.0 SW(2,2) Layer 2 maximum soil moisture (mm). 028 105.0 SW(2,3) Layer 3 maximum soil moisture (mm). 019 10.0 AIRDRY Layer 1 air dry soil moisture content (mm). 029 15.0 SW(3,1) Layer 1 wilting point soil moisture (mm). 030 70.0 SW(3,2) Layer 2 minimum soil moisture (mm). 031 65.0 SW(3,3) Layer 3 minimum soil moisture (mm). STARTING SOIL MOISTURE also used when p290 is a date 023 00.0 SW(9,1) Starting value for soil moisture layer 1 (mm). 024 000.0 SW(9,2) Starting value for soil moisture layer 2 (mm). 025 00.0 SW(9,3) Starting value for soil moisture layer 3 (mm). TREE WATER USE 291 0.0 MATURE TREE BASAL AREA square metres/ha 292 0.0 Layer 1 minimum soil moisture (mm) with trees 293 0.0 Layer 2 minimum soil moisture (mm) with trees. 294 0.0 Layer 3 minimum soil moisture (mm) with trees. 295 000.0 Layer 4 available water (trees only) 296 100.0 Maximum rooting depth of trees in cm 299 00.0 Starting value for soil moisture layer 4 (mm), trees only asw4 297 1.44 Tree Root length at surface, rl= p297*exp(-p298*z) 298 0.61 Tree Root length exponent, rl= p297*exp(-p298*z) 168 20.00 Tree basal area at which potential tree transpiration=pan SOIL EVAPORATION 033 4.0 EPLIM Upper limit to daily BARE soil evaporation (mm/day) 1.0 mm/day for surface sealed soils, or large cover of rocks 8.0 mm/day for sand 035 0.0 If=1 cracking occurs with increased soil evaporation 036 0.0 Soil evaporation when soil cracked (mm/day) RUNOFF 270 0.0 0 for free draining soils, 1for runoff as a f(yield) 271 1150.0 Tsdm yield at 50% cover for run-off calculation 272 0.95 k value in cover=y**k / (y**k + p271**k) 273 1.00 Maximum runoff of rainfall at zero cover, wet soil 278 1.00 % slope of land 0-20% 279,280,281 DO NOT RE-USE

Appendices

299

I15 is intensity & is a function of latitude & season I15 Brian Pastures= 1.016+0.465*cos, I15 Charters = 0.9+0.7*cos I15 capella = 0.867+0.582*cos 104 1.016 Constant in I15 equation I15=p104+p105*cos(dayno+15) 105 0.465 Slope in I15 equation I15=p104+p105*cos(dayno+15) 243 -0.1880 Cover sensitivity coef (MS calls b2) sed_cov = -0.05 ! perfect code sed_cov = -0.095 ! `springvale' data, -0.188 Burdekin data 244 0.0550 Bare soil effeciency of entrainment kusle = 0.1 ! springvale guess kusle = 0.055 ! Burdekin Scanlan&McIvor 245 1.000 Soil loss If=1 use Rose model, 2=Bob Miles, 3=Joe Scanlan 173 0.000 Initial depth of soil loss in mm 174 0.999 Relative effect of soil loss at infinite soil loss 175 1000.0 Depth of soil loss giving half of effect ie (1+p174)/2 176 0.000 Bulk density (g/cc) to convert soil loss to depth if=0.0 no effect of soil loss calculated OBSERVED SOIL MOISTURE 282 0.0 If=1.0 reset to observed soil moisture in management file ROOT DISTRIBUTION 106 0.5 Relative supply of layer 3 cf layers 1,2. Usually 0.5 PLANT COVER 210 2.0 Selector for cover fuction; 1=f(time) P(38...43) ,2=f(yields). p210=3 use observed green & bare cover 038 0.625 SCOV mean ) SCOV = P38 +P39*cos(0.01720*(idayno+P40)) 039 0.325 SCOV amplitude ) = total surface cover calculated as a 040 -30.0 SCOV lag ) function of time.(-30 : max cover=Jan30 041 0.425 GCOV mean ) GCOV = P41 +P42*cos(0.01720*(idayno+P43)) 042 0.325 GCOV amplitude ) = green surface cover calculated as a 043 -30.0 GCOV lag ) function of time.(-30 : max cover=Jan30 045 1600.0 Green yield (kg/ha) when green cover for transpiration is 50% 107 1.0 A value to transform green cover to POT TRANS/PAN PLANT TEMPERATURE INDEX selection parameters. C3 : p61= 9.0, p62= 18.0, p63= 30.0, p64= 45.0; Fitz & Nix 1970 C4 : p61=14.0, p62= 24.0, p63= 45.0, p64= 50.0; McCown 1980 209 4.0 TIX 1=FSS, 2=GP , 3= NP ,4= use p61 and p62 ,5= tix=1.0 6=maize, 7=combined NP, 8=NP f(max,min) 061 14.0 If temp is less than P61, temperature index (TIX) is zero. 062 24.0 As temp increases from P61 to P62, TIX increases from 0 to 1. 063 35.0 As temp increases from P62 to P63, TIX remains at 1. 064 45.0 As temp increases from P63 to P64, TIX decreases from 1 to 0.0 CLIMATE CHANGE or PAN CALCULATION 003 0.0 If >0.0 <1.0 use .p51 met files & calc mean daytime VPD if(p(3).gt.0.0.and.p(3).lt.1.0.and.m(50).eq.1)then dewpt=dewsin(vp) vpmin=vpsat(tmin) vpd=vpmin+p(3)*(vpsat(tmax)-vpmin)-vp 068 0.0 If=1 call climate change If=2 calculate pan from vpd If=3 use p51, calculate vpd from tmax etc, pan from vpd, pan=f(vpd) 233 0.0 if p68=2 epan=p(233)+p(234)*vpd+p(237)*rs 234 0.0 if p68=2 epan=p(233)+p(234)*vpd+p(237)*rs 237 0.0 if p68=2 epan=p(233)+p(234)*vpd+p(237)*rs epan= -2.610+0.1879*vpd+0.2545*rs

Appendices

300

065 1.0 Multiplier on summer rainfall 1Oct-30Apr 1.0= no change 240 1.0 Proportion of original summer 1Oct-30Apr rain days retained 241 1.0 Multiplier on winter rainfall 1May-30Sept 1.0= no change 242 1.0 Proportion of original winter 1May-30Sept rain days retained 066 0.0 Change in minimum temperature 238 0.0 Change in dewpoint temperature 239 1.0 Multiplier on Solar Radiation 1.0= no change 067 0.0 Change in maximum temperature if p67=-999 calculate tmax from regression f(tmin,dewpt,solarrad) PLANT SOLAR RADIATION INDEX & INTERCEPTION 046 1600.0 Green yield (kg/ha) when radiation interception is 50% 008 12.0 Radiation use efficiency kg/ha per MJ/sqm of solar radiation PLANT GROWTH 005 6.8 Initial plant density e.g. % basal area 006 2.0 Potential daily regrowth rate (kg/ha/day/unit of density) This is with water, temperature and light non-limiting, (growth index = 1), and represents the potential rate at which a pasture will regrow in the first few weeks after burning or cutting. Density unit is same as P5 007 15.0 Transpiration efficiency (kg/ha/mm of transpired at vpd 20hPa Daily growth =p(7)*vpdix*daily transpiration vpd is vapour pressure deficit input from met data with .v51 vpdix=10/(vpd*f(height)) te=p(7)*vpdix 094 1.5 Multiplier of VPD for zero height 095 20.0 Height at which VPD multiplier = 1.0 096 10.0 Height (cm) of 1000 kg/ha height=p(96)*(tsdm/1000.0) vpdhgt=amax1(1.0,amin1(p(94), $ 1.0+(height-p(95))*(p(94)-1.0)/(0.0-p(95)) )) if(vpd.gt.10.0)vpdix=amax1(0.0,amin1(1.0,10.0/(vpd*vpdhgt) )) SOIL MOISTURE SUPPLY EFFECT ON PLANT GROWTH 274 0.0 If=1 use denmead and shaw for limiting soil moisture index 275 13.0 Layer 1 p275*awr1**2 mm/day 276 13.0 Layer 2 p276*awr2**2 mm/day 277 3.6 Layer 3 p277*awr3**2 mm/day 149 0.4 Soil water index at which above-ground growth stops. NEW SWARD MODEL 123 0.50 Proportion of leaf of total growth (L/(L+S)) 124 0.00 Green yield at which stem growth begins FROST p125 must be greater than p11 011 0.00 Minimum screen temperature (c) at which green cover = 0% 125 2.00 Minimum screen temperature (c) at which green cover =100%=no deat 053 2.00 Daily minimum screen temperature for frost effect on quality 009 0.40 Soil water index. Maximum green cover = amin1(0.99,swix/p(9)) 047 0.00 Scale (0-1) for effectiveness of tree litter in runoff covermin=p(47)*(1.0-baremax) 132 75.0 Percentage of leaf/(leaf+stem) in diet at 50% leaf in sward PLANT SENESCENCE AND LITTER BREAKDOWN 010 0.002 Death constant ) DEATH = (P51*(1-swix) + P10) * green pool 051 0.013 Death slope ) where swix = soil water index 133 1.0 Multiplier on total death for DM death of leaf 134 1.0 Multiplier on total death for DM death of stem e.g. leaf lasts 60 days, stem lasts 360 days DETACHMENT

Appendices

301

128 0.002 Prop of Dead leaf detached per day from 1Dec to 30 April 129 0.002 Prop of Dead stem detached per day from 1Dec to 30 April 130 0.002 Prop of Dead leaf detached per day from 1May to 30 November 131 0.002 Prop of Dead stem detached per day from 1May to 30 November Rainfall effect on detachment 154 0.000 Prop of Dead leaf detached per day per 100 mm of (rain-p156) 155 0.000 Prop of Dead stem detached per day per 100 mm of (rain-p156) 156 50.0 Two-day rainfall required to initiate detachment caused by rain 015 1.00 Proportion of pasture which can be eaten by stock. The rest is lost by trampling. DEIN2 = TINT/P15 - TINT P16 and P18 are constants for litter breakdown BREAK = (SW(6,1) * temp/25 * P16 + P18 * stockrate)*litter pool Thus, breakdown is rapid when soil water in layer 1 is high, when temperature is high, and the pasture is grazed. 016 0.04 Rate of litter breakdown when hot and wet 018 0.0 Coefficient of stocking rate on litter breakdown NITROGEN UPTAKE 090 0.0 N kg/ha per 1000 mm of rain 097 5.0 N uptake (kg/ha) at zero transpiration, N=p(97)+p(98)*(trans/100 098 5.8 N uptake per 100 mm of GRASS transpiration 167 0.0 Prop of p98 for N uptake in TREE transpiration from layers 1&2&3 099 23.0 Maximum N uptake (kg/ha) 100 2.5 Maximum % N in growth 101 0.4 % N at zero growth Nitrogen index = (%N-p101)/(p102-p101) 102 0.5 % N at maximum growth Nitrogen index = (%N-p101)/(p102-p101) 103 1.0 N uptake per 100 mm of soil water 108 0.000 Proportional decline per day in % N for green material 109 0.015 Proportional decline per day in % N for dead material 110 1.0 Minimum % N in green & maximum in dead 111 0.4 Minimum % N in dead 112 0915.0 Date for resetting Nitrogen uptake GRAZING P142, P143 & P144 define an intake restriction index from the proportion of pasture eaten (PCON) and total standing DM. RESTR = max(0.0, min( P142 + P143 * PCON, TSDM/P144, 1.0) When RESTR = 0, intake is fully restricted, and when RESTR = 1, intake is not restricted LWGYEAR1 sub for annual lwg calculation 214 300.0 Pasture yield limiting lwg in annual lwg calculation 215 15.0 LWG advantage due to burning used in lwgyear1 sub alwgmax=alwgdays*(wtint*0.304-0.800) alwgmax=alwgdays*(wtint*p(216)+p(217)) 216 0.304 Slope in LWG maximum possible for given dry matter intake 217 -0.800 Intercept in LWG maximum possible for given dry matter intake alwgyear1=alwgdays*(0.0239-0.002117*wutil+0.005586*wgixdays) alwgyear1=alwgdays*(p(228)+p(229)*wutil+p(230)*wgixdays & +p(231)*wthiacc) general regression from bp,galloway,khills lwgyea5b.col alwgyear1=alwgdays*(0.0590-0.002107*wutil+0.004889*wgixdays) general regression from bp,galloway,khills lwgyea7b.col alwgyear1=alwgdays*(0.0603-0.002061*wutil+0.004833*wgixdays) new regression specifically for kangaroo hills trees lwgyear6.col alwgyear1=alwgdays*(0.0239-0.002117*wutil+0.005586*wgixdays) 228 0.0239 Intercept in annual lwg regression 229-0.002117 Coeff for %utilisation in annual lwg regression 230 0.005586 Coeff for %green days in annual lwg regression 231 0.000000 Coeff for THI(temperature-humidity) in annual lwg regression McCaskill's model 117 0.0 Output for animal model to lw21.ogp 0=no output

Appendices

302

Options same for p247 118 18.0 Default animal age (months) when LW's are reset 119 1.0 No. of experimental treatment for liveweight gain 120 1.0 Animal model; 0 =0.0,or 1 for utilization model, 2 for GRASP green days, 3 for WATBAL green days, 4 for new greenday_frost method, 5 for new diet N method diet quality, 6 old grasp method 7=f(ET last month), 8=%green leaf andrew ash, 9=khills reln 8=%green leaf andrew ash, 9=khills reln Model 2&3 056 0.050 Growth index for greenday/frost & wool climatic index 121 0.493 Animal growth rate for green days (native 0.493, imp 0.613) 122 -0.163 Animal growth rate for non-green days (native -0.163, imp -0.043 Daily lwg from f(greenleaf) New Model 4 (4 Feb 1996) greenleaf, dead feed 113 0.900 Animal growth rate when greenleaf > glx2 eg 100 to 200 kg/ha 114 0.000 Not used 115 0.000 Animal growth rate for when zero greenleaf tsdm=300 kg/ha 116 -1.000 Animal growth rate for when zero greenleaf & tsdm=0, eg -1kg/day Model 1 utilisation These values are used only to adjust intake. 142 1.05 Intercept in equation of reln between intake and utilisation 143 -.3000 Slope in equation of reln between intake and utilisation 144 300.0 Yield (kg/ha) at which intake restriction no longer operates 145 70.0 Expected live weight gain (kg/hd) in summer at low stocking rate 146 35.0 Expected live weight gain (kg/hd) in autumn at low stocking rate 147 10.0 Expected live weight gain (kg/hd) in winter at low stocking rate 148 35.0 Expected live weight gain (kg/hd) in spring at low stocking rate Model 5 diet %N 092 1.000 Slope in dietn=p92*%dietquality+p93 093 0.000 Intercept in dietn=p92*%dietquality+p93 054 0.657 Slope in lwg=p54*dietn+p55 055 -0.405 Intercept in lwg=p54*dietn+p55 Forage quality, for example green leaf 057 12.15 Slope in %Dry Matter Digestibility DMD=p57*%Ndietn+p58 058 39.19 Intercept in %Dry Matter Digestibility DMD=p57*%Ndietn+p58 059 0.150 Slope in %Phosphorus =p59*%Ndietn+p60 060 0.020 Intercept in %Phosphorus =p59*%Ndietn+p60 150 0.0 Initial stocking rate (weaners/ha, live weight = 200 kg) 170 20.0 Dry matter yield for each age class at which intake not limited in dietsel 171 -0.0286 K value for decline in diet preference with age 172 -0.1024 K value for decline in diet nitrogen with age SIMULATION CONTROL 261 1.0 Batch operation=0 ; interactive=1 ; replace original pars=2 batch means no parameters, mrx file starts with management title set p261 =1 in default file if(n.eq.261.and.line.eq.2)call replac : replace with original par for gvt89 reset m(61)=0 as most common operation interactive option is NOT available if(m(61).eq.1)m(61)=0 203 1900.0 Starting year of simulation; 1800 to begin at start of metfile. 204 07.0 Starting month of simulation 206 00.0 Number of days in simuln run,last date : 1st Mar 1986=198603 if=0 150 years

Appendices

303

CLIMATE STATIONS 263 00000.0 Station no. of AUSTCLIM station from menu option(39039=GAYNDAH) 263 39039.0 Station no. of AUSTCLIM station from menu option(39039=GAYNDAH) 264 42.0 No. of daily (rainfall) station in pmbstat2.pat, 1=BrianPastures 250 6.0 If=1 full daily met data, if=3 weekly austclm 4=daily rain in dr2 format, with either AUSTCLIM or station p269 6=daily rain + daily climate, no in p269 type 1 269 8.0 Monthly climate station type 5 in pmbstat2, if=0 AUSTCLIM gvt89ma3 allows the pathway to be put in the parameter file but 264 or 269 must be set to -99 (or <0.0) Example for typical Gunsynpd analysis 264 -99.0 ron8697.dr2 250 6.0 if=1 full daily met data, if=3 weekly austclm 157 4=daily rain in dr2 format, with either AUSTCLIM or station p269158 6=daily rain + daily climate, no in p269 type 1 159 269 -99.0 c:\p51\gayndah.p51 235 0.0 No of file with record no.s for re_ordering station 264, p250=1 In pmbstat2.pat type=7 OUTPUT CONTROL mndy=monthday eg 0315 is 15th March P247 - Output of totals : 365=yearly,91=seasonally,30=monthly, 7=weekly,1=daily,999=each observation. If P247=mndy and 182 Nov-Apr, 183 Dec-May, BUT no winter output 988 for each management date 15 detachment analysis on yield dates code 15, 286=0 90 for calender quarters ie 3-31,6-30,9-30,12-31 P249 = 0 then probabilities will be printed. p247=9901 for totals between srate change for totals between obs,code 15 must be first rec on date P248 - Output of model : 365=yearly,91=seasonally,30=monthly, 7=weekly,1=daily,999=each observation, 988=soil moisture P249 - if = 1,totals are summed; if = 0 and P247 = mndy then probabilities will be printed P262 - Output to screen : 365=yearly,91=seasonally,30=monthly, 7=weekly,1=daily,999=each observation,988=soil moisture 11=growth model, 12= total days for each growth limit 988=soilmoisture, 977=runoff, 976=water balance on p247 246 132.0 Output type: 80=80 column output, 132= 132 output 0=132 col 247 0.0 Output of totals:365 - 999=yr - obs.If=mndy & P249=0,print prob note! multiple management dates can prevent required output 999 for pasture observations, 988 for any management date 133 output on Images dates ie 4 monthly 30/4 31/8 31/12 For probabilities p247=date e.g 930 is 30th September 9901 for output on change of stocking rate 15 detachment analysis on yield dates code 15, 286=0 9927 to output totals on date of code 27 248 0.0 Output of model:365=yr,91=seas,30=mthly,7=wkly,1=daily,999=obs 999 for pasture observations, 988 for soil moisture 9977 for climatic averages on s8.ogp 249 1.0 if=1,totals are summed;if=0 and P247=mndy then probs are printed 262 979.0 Output to screen:365=yr,91=seas,30=mthly,7=wkly,1=daily,999=obs 999 for soil moisture & pasture observations, 998 for detailed pasture observations, green, dead, litter 988 for soil moisture 979 for all gunsynd analysis ie p19, screen=mrx 978 for nitrogen change in TSDM 977 for runoff output to screen 976 for water balance in long term simulations & tree water use 975 for tree water use 974 for nitrogen uptake 973 for output of both water balance & nitrogen uptake on p247 972 for surface conditions and litter breakdown

Appendices

304

971 for simulation output for GRASSMAN 969 for liveweight output on observation date 968 for diet selection output 284 for TE & RUE growth analysis 286 for rainfall use efficiency on screen 11 for growth model debugging 12 for growth model debugging 259 0.0 Output to screen: 1= stop screen scrolling 283 1.0 If=1 ET output to file s18.ogp, p246 must be 132 if=2 Drought days output to s18.ogp 284 2.0 If=1 TE output to file p9.ogp, p246 must be 132 if=2 RFUE,ETUE,TUE if=3 Comparison of N models, %N code=26 if=4 Output of totals for LWG to p19.ogp on p247 if=5 Output of growth, utilisation, grass basal area on gba obs if=6 Output of growth, utilisation, grass basal area on yield obs if=7 Output of growth, utilisation, as for=2 on yield obs if=8 Output of many growths, utilisation, on gba obs, use p284=8 & p268=1,2,4,5 to compare with obs tgba if=9 Output of climatic indices for wool growth code 27 if=10 Output of wool growth,lwc on p(81) if=11 Output of LWG analysis on SR change on unit 17 if(m(84).eq.11.and.m(89).eq.4) if=12 Output of parameter change with resource change if=970 for daily output used in estab4 If=13 output of XO variables to p9.ogp 211 7.0 If=1-365 gives output of observed & predicted , and simulated : 365=yearly,91=seasonally,30=monthly, 7=weekly,1=daily,999=each observation,mndy=monthday 133 output on Images dates ie 4 monthly 30/4 31/8 31/12 output in m15.ogp or m1 285 0.0 If=1 monthly growth output to file m15.ogp, p211 must be 0 if=2 monthly growth in rainman output 286 0.0 If=1 rainfall use efficiency to r17.ogp, p246 must be 132 if=2 p247 =15 for detachment analysis if=3 output for developing forecast equations if=4 p289=0, p81 ne 0 summary output on r17.ogp of lw21.ogp if=5 summary output of lwg analysis for multiple regression 287 00.0 If=1 runoff output to p19.ogp, only days with rain GE p287 p270 must be 1 for output, p246 must be 132 soil water deficit, cover, I15, rain, run-off are output if=999 output observed runoff to p19.ogp 289 0.0 Output options for unit 21 If=0 Only summary of probability table If=1 Liveweight gain calculation If=2 Diet %N, green in diet code 36 If=3 Analysis of wool data for days>260 If=4 Liveweight gain calculation, on SR change If=5 Surfair parameter set If=6 Surfair data output comparison If=7 Forecasting data output for regression analysis If=8 Daily output for pat pepper If=9 Output stored simulation results If=10 output runoff & soil loss If=11 output of XO variables If=12 annual liveweight models, components If=13 annual output on p247 for rank, SW,growth,animal If=14 Drought periods when layers2+3<p(162) If=15 Output for comparison with sheep Herdecon If=16 Output of all grass basal calculations If=17 Output of transpiration from each layer on p247 If=18 Output for at APSIM on 211 If=19 Output of leaf and stem If=20 Output of recored no of year for re_ordering met data if=21 comparison with mss If=22 indices for growth limiting comparison If=23 output of soil moisture , sdm

Appendices

305

If=24 Jyo output If=25 accumulated soil loss output If=26 output of wool and sheep models 236 15.0 For storing simulation output from probability array XO Default is 15 = lwg/ha Also parameter no for identifying variable from xo averages 169 9999.0 Value of goal for pest optimisation 9999 for maximising, -9999 for minimising, or an actual value eg obs value 227 0.0 Parameter no for output when p289=0 208 0.0 Parameter no for output when p289=0 212 000.0 Too few days for analysis of lwg when p289=4 213 000.0 Too many days for analysis of lwg when p289=4 ANNUAL CROP MANAGEMENT To be used only in simulation studies of annual forage P151, P152 & P153 control 3 decision rules for planting a crop. Crop can be planted if the avail water ratio in the whole profile > P151, in layer 1 > P152 251 0.0 If=1 call crop emergence subroutine and use options P252to 260. 151 0.500 Min available water ratio in total profile required for planting 152 0.900 Min available water ratio in layer one required for planting 153 0.300 Max awr in layer one so it`s dry enough to plant 252 0901.0 First date for planting; month day 0901 = 1st Sept. 253 1231.0 Last date for planting; month day 1231 = 31st Dec. 254 25.0 Yield at emergence (kg/ha) 255 10.0 Number of days from planting to emergence 256 90.0 Number of days from emergence to end of crop growth 260 70.0 Number of days from emergence to end of green growth 257 401.0 End of crop on month, day due to temperature PASTURE BURNING MANAGEMENT 265 0.0 If=1 call pasture burning subroutine and use options 266-7 052 100.0 Percentage of pasture burnt 266 1001.0 First date of burning; month day 1001 = 1st Oct 267 0.0 Threshold yield required for burn; total standing DM kg/ha DYNAMIC PASTURE BASAL AREA 268 0.0 If=1 call dynamic basal area subroutine & reset when mrx=11 if=2 call dynamic but do NOT reset when mrx=11 if=0 CONSTANT basal area(p5) & reset when mrx=11 if=3 CONSTANT basal area(p5) & do NOT reset when mrx=11 if=4 call tussock grass basal area=f(growth,util) & reset when mrx=11 if=5 call tussock grass basal area=f(growth,util) & do NOT reset when mrx=11 if=6 call tussock grass basal area=f(yield430) & reset when mrx=11 if=7 call tussock grass basal area=f(yield430) & do NOT reset when mrx=11 if=8 call GENERAL grass basal area=f(yield430,trans,growth) & do NOT reset when mrx=11 if=10 grass basal area=ET & utilisation) & do NOT reset when mrx=11 MSS uses options 100=0,3 ; 101=1,2 ; 102= 4,5 ; 103= 6,7 288 0.0 Water (ET) use efficiency for basal area change 268=1 166 0.00 Slope on ET for 100% green utilisation, basal area change,268=10 if(p(288).le.0.0)p(288)=2.0+p(99)/5.0 barea=((1.0-p(164)*growly+p(164)*etsu*p(288))/500.0 164 0.5 Proportion of this year's growth affecting sward basal area use p284=8 & p268=1,2,4,5 to compare with obs tgba if p268=4,5 change tussock grass basal area in subroutine TGBAGZ 157 430.0 Date (monthday) for annual change in tussock grass basal area 158 0.30 Minimum tussock grass %BA, TGBA=p158+p159*(growth or yield430) 159 2.0 Slope , TGBA=p158+p159*(average growth year1+2 /1000 or yield430) +p(165)*utilisation, ifp268=4,5 165 0.00 Coefficient(-ve) on green utilisation in basal area change,268=4,5

Appendices

306

160 2.5 Maximimum possible increase per year in TGBA e.g. 2.5 %units 161 9.0 Maximimum possible tussock grass basal area 162 0.30 Swix 2+3 <threshold for counting drought days 163 0.50 Swix 2+3 >threshold for turning off drought RESET STOCKING RATE, LIVEWEIGHT, BREED for constant SR if average growth has not been calculated set stock at the no in p83 if(averagegrowth.le.0.0)then srwean=p(83)/100.0 079 0.0 TSDM yield for destocking at change date 080 0.0 Date for shearing,p80 before p247 and= p81 930 is 30th Sept 081 0.0 Date for resetting stocking rate & shearing, 930 is 30th Sept 082 0.0 1=Constant %of AveGR, 2=%Use of TSDM for next year, 3=%Use of Growth 083 0.0 %Utilisation of average growth, pasture yield, forecast growth 084 0.0 Liveweight kg 085 0.0 Breed 1=XBRED 2=BRITISH 3=G2 XBRED, =11 for dry sheep equivalents 086 0.0 Age in months Forecasting stocking rate multiplier for each year type 219 1.0 Scaling factor used in option 3 to correct for bias in the multiple regression forecasting growth 220 1.0 For year type 0 forecasting stocking rate multiplier 221 0.0 For year type 1 forecasting stocking rate multiplier 222 0.0 For year type 2 forecasting stocking rate multiplier 223 0.0 For year type 3 forecasting stocking rate multiplier 224 0.0 For year type 4 forecasting stocking rate multiplier 225 0.0 For year type 5 forecasting stocking rate multiplier 226 0.0 For year type 6 forecasting stocking rate multiplier 232 0.0 Option 4 Foremultipl=p(220+jj)*( tsdm/(p(232)+0.01) ) 225 For p82=9 low value of random SR multiplier 226 For p82=9 high value of random SR multiplier RESET POOLS TO SAME YIELD on a date each year if p81 =0 , reset animal vars p83,p84,p85,p86 For grass basal area(p5) to be a constant, p268=0 140 0.0 Date for resetting DRY MATTER, 930 is 30th Sept 135 0.0 Green leaf 136 0.0 Green stem 137 0.0 Dead leaf 138 0.0 Dead stem 139 0.0 Litter 088 0.0 Accumulated transpiration since p112 089 0.0 Accumulated growth since p112 087 0.0 Accumulated growth since 1201 ie 1st Dec 141 0.0 Accumulated consumption since 1201 ie 1st Dec 069 0.0 tgrowtgba3 total growth for basal calc 070 0.0 teatgrgba green eaten used in pargrz 071 0.0 teatdmgba 072 0.0 proputilold used pargraz 073 0.0 transtgba used for tgba calculation 074 0.0 yld430 yield at April 30 075 0.0 tgrowtgba2 total growth for basal calc 076 0.0 tgrowtgba1 total growth for basal calc 077 0.0 growly growth last year calc from accumulated ET 078 0.0 etsu accumulated ET 290 0.0 Date for resetting soil moisture to p23,24,25,299 930 is 30th Sept HEAVY UTILISATION PARAMETERS 180 0.0 If=1 change parameters as a function of utilisation 30th April If=2 change parameters f(green util & % perennials 30th April) If=3 change parameters f(total util & % perennials 30th April) 191 0.50 proportion green_eaten/growth at which pasture has heavy utilisation parameters if p180=1 Change in perennial index if p180=0 or 3 use %DM utilisation, 1 or 2 use green utilisation 194 0.00 Initial pasture condition 0=90% perennials,11=heavily grazed

Appendices

307

195 1.00 Resilience rate %util<22.5 1= 1 year equivalent on AA scale 196 1.00 Degradation rate %util>34 1= 1 year equivalent on AA scale 197 0.00 If=0 resource cannot return from heavily grazed state 198 30.0 % DM Utilisation for increase in %perennials to occur. %UGrn if p180=2 199 45.0 % DM Utilisation for decrease in %perennials to occur. %UGrn if p180=2 181 23.0 p99 Maximum N uptake (kg/ha) 182 1600.0 p45 Green yield (kg/ha) when green cover for transpiration is 50% 183 10.0 p96 Height (cm) of 1000 kg/ha 184 0.4 p101 % N at zero growth Nitrogen index = (%N-p101)/(p102-p101) 185 0.5 p102 % N at maximum growth Nitrogen index = (%N-p101)/(p102-p101) 186 0.002 p128 Prop of Dead leaf detached per day from 1Dec to 30 April 187 0.002 p129 Prop of Dead stem detached per day from 1Dec to 30 April 188 0.002 p130 Prop of Dead leaf detached per day from 1May to 30 November 189 0.002 p131 Prop of Dead stem detached per day from 1May to 30 November 190 0.99 p149 Soil water index at which above-ground growth stops. 192 300.0 p144 Yield (kg/ha) at which intake restriction no longer operates 193 0.40 p009 Soil water index. Maximum green cover = amin1(0.99,swix/p(9)) 200 0.20 p056 Growth index for greenday/frost & wool climatic index 300 0.0 End of parameters

Constant parameters for the VRD xxxxxxxxxxxxxOUTPUT CONTROLxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 246 132.0 Output type: 80=80 column output, 132= 132 output 0=132 col 247 000.0 Output of totals:365 - 999=yr - obs.If=mndy & P249=0,print prob 248 0.0 Output of model:365=yr,91=seas,30=mthly,7=wkly,1=daily,999=obs 249 1.0 if=1,totals are summed;if=0 and P247=mndy then probs are printed 262 999.0 Output to screen:365=yr,91=seas,30=mthly,7=wkly,1=daily,999=obs 259 0.0 Output to screen: 1= stop screen scrolling 283 1.0 if=1 ET output to file s18.ogp, p246 must be 132 284 2.0 if=5 basal area utilisation 211 1.0 if=1-365 gives output of observed & predicted , and 285 0.0 if=2 monthly growth in rainman output 286 0.0 if=1 rainfall use efficiency to r17.ogp, p246 must be 132 287 00.0 if=1 runoff output to p19.ogp, only days with rain GE p287 289 19.0 Output options for unit 21 CONSTANT FOR VRD 112 1001.0 Date for resetting Nitrogen uptake 265 0.0 If=1 call pasture burning subroutine and use options 266-7 268 0.0 If=1 call dynamic basal area subroutine & reset when mrx=11 104 0.9 Constant in I15 equation I15=p104+p105*cos(dayno+15) 105 0.7 Slope in I15 equation I15=p104+p105*cos(dayno+15) 003 0.75 If >0.0 <1.0 use .p51 met files & calc mean daytime VPD 033 4.0 EPLIM Upper limit to daily BARE soil evaporation (mm/day) 1.0 mm/day for surface sealed soils, or large cover of rocks 8.0 mm/day for sand 036 0.5 Soil evaporation when soil cracked (mm/day) 227 264.0 climate station 250 6.000 If=1 full daily met data, if=3 weekly austclm 26314834.000 Station no. of AUSTCLIM station from menu option(39039=GAYNDAH)

Appendices

310

246 132.0 Output type: 80=80 column output, 132= 132 output 0=132 col 247 701.0 Output of totals:365 - 999=yr - obs.If=mndy & P249=0,print prob 248 0.0 Output of model:365=yr,91=seas,30=mthly,7=wkly,1=daily,999=obs 249 0.0 if=1,totals are summed;if=0 and P247=mndy then probs are printed 262 979.0 Output to screen:365=yr,91=seas,30=mthly,7=wkly,1=daily,999=obs 259 0.0 Output to screen: 1= stop screen scrolling 283 0.0 if=1 ET output to file s18.ogp, p246 must be 132 284 13.0 if=13 output of XO to p9.ogp 211 0.0 if=1-365 gives output of observed & predicted 285 0.0 if=2 monthly growth in rainman output 286 0.0 if=1 rainfall use efficiency to r17.ogp, p246 must be 132 287 00.0 if=1 runoff output to p19.ogp, only days with rain GE p287 289 0.0 Output options for unit 21 If=4 Liveweight gain calculation, on SR change If=9 Output stored simulation results 203 1957.0 Starting year of simulation; 1800 to begin at start of metfile. 204 01.0 Starting month of simulation 206 200407 Number of days in simuln run,last date : 1st Mar 1986=198603 250 1 269 0 264 206 C:\CLIMATE\INVERWAY.P51 INTRODUCED PARAMETERS 291 1.0 MATURE TREE BASAL AREA square metres/ha 150 0.0 Initial stocking rate (weaners/ha, live weight = 200 kg) 265 1.0 If=1 call pasture burning subroutine and use options 266-7 052 5.0 Percentage of pasture burnt 266 1101.0 First date of burning; month day 1001 = 1st Oct 267 1500.0 Threshold yield required for burn; total standing DM kg/ha 268 1.0 If=1 call dynamic basal area subroutine & reset when mrx=11 167 1.0 Prop of p98 for N uptake in TREE transpiration from layers 1&2&3 112 1101.0 Date for resetting Nitrogen uptake RESET STOCKING RATE, LIVEWEIGHT, BREED 140 0000.0 Date for resetting DRY MATTER, 930 081 701.0 Date for resetting stocking rate 082 2.0 1=Constant, 2=%Use of TSDM for next year, 3=%Use of Growth 083 10.0000 %Utilisation of AveGR, or pasture yield 084 400.0 Liveweight kg 085 3.0 Breed 1=XBRED 2=BRITISH 3=G2 XBRED, =11 for dry sheep equivalents 086 30.0 Age in months 120 1.0 Animal model; 0 =0.0,or 1 for utilization model, 056 0.20 was 0.05 Growth index for greenday/frost & wool climatic index 300 0.000 End of parameters 14 99990000 file_end

References

311

References Abbadie L, Mariotti A, Menaut J-C (1992) Independence of savanna grasses from soil organic matter for their nitrogen supply. Ecology 72, 608-613. Alcock DJ (2006) Using grazing systems modelling to assess economic, production and environmental risks to aid in selecting appropriate stocking rates. Australian Journal of Experimental Agriculture 46, 841-844. Aldrick JM, Dunlop CR, Howe DF (1978) 'Report on the lands of the Ord River catchment, Northern Territory.' Animal Industry & Agriculture Branch, Dept. of the Northern Territory, Technical Bulletin No. 24, Darwin, NT. Andrew MH (1986) Survivorship of the grass Enneapogon purpurascens in the Victoria River District, north-west Australia. Australian Rangelands Journal 8, 63-66. Andrew MH, Gowland PN, Holt JA, Strickland GR (1985) Vegetation and Fauna. In 'Agro-research in the Semi-Arid Tropics: North West Australia'. (Ed. RC Muchow) pp. 93-111. (University of Queensland Press, St Lucia: Brisbane, Qld) Andrew MH, Mott JJ (1983) Annuals with transient seed banks: the population biology of indigenous Sorghum species of tropical north-west Australia. Australian Journal of Ecology 8, 265-276. Arkley RJ (1963) Relationships between plant growth and transpiration. Hilgardia 34, 559-584. Ash AJ, McIvor JG (1995) Land condition in the tropical tall-grass pasture lands. 2. Effects on herbage quality and nutrient uptake. Rangeland Journal 17, 86-98. Ash AJ, McIvor JG, Corfield JP, Winter WH (1995) How land condition alters plant-animal relationships in Australia's tropical rangelands. Agricultural Ecosystems and Environment 56, 77-92. Ash AJ, Stafford Smith DM (1996) Evaluating stocking rate impacts in rangelands: Animals dont practice what we preach. Rangeland Journal 18, 216-243. Asseng S, Keating BA, Fillery IRP, Gregory PJ, Bowden JW, Turner NC, Palta JA, Abrecht DG (1998) Performance of the APSIM-wheat model in Western Australia. Field Crops Research 57, 163-179. ATO (2006) 'http:\\www.ato.gov.au.' Baker DE, Ahern CR (1989) Estimates of effective rooting depth for predicting available water capacity of Burdekin soils, Queensland. Australian Journal of Soil Research 27, 439-454. Barritt MJ (1996) 'Checklist of the vascular plants of the Victoria River District, Northern Territory, Australia (Draft only).' Parks and Wildlife Commission of the NT, Darwin, NT. Bastin GN, Andison RT (1990) Kidman Springs country - 10 years on. Range Management Newsletter 90, 15-19. Bastin GN, Ludwig JA, Eager RW, Liedloff AC, Andison RT, Cobiac MD (2003) Vegetation changes in a semi-arid tropical savanna, northern Australia:1973-2002. Rangeland Journal 25, 3-19.

References

312

Bauer FH (1985) A brief history of agriculture in north-west Australia. In 'Agro-research in the Semi-Arid Tropics: North West Australia'. (Ed. RC Muchow) pp. 13-28. (University of Queensland Press, St Lucia: Brisbane, Qld) Beale IF (1973) Tree density effects on yield of herbage and tree components in south west Queensland mulga (Acacia aneura F. Muell.) scrub. Tropical Grasslands 7, 135-142. Beech DF (1990) The effect of carrier and rate of nitrogen application on the growth and oil production of lemongrass (Cymbopogon citratus) in the Ord Irrigation Area, Western Australia. Australian Journal of Experimental Agriculture 30, 243-250. Behnke RH, Scoones I (1993) Rethinking range ecology: Implications for range management in Africa. In 'Range ecology at disequilibrium'. (Eds RH Behnke, I Scoones and C Kerven). (Overseas Development Institute: London, UK) Behnke RH, Scoones I, Kerven C (Eds) (1993) 'Range ecology at disequilibrium.' (Overseas Development Institute: London, UK) Berndt RD, Coughlan KJ (1976) The nature of changes in bulk density with water content in a cracking clay. Australian Journal of Soil Research 15, 27-37. Bethune MG, Wang QJ (2004) Simulating the water balance of border-check irrigated pasture on a cracking soil. Australian Journal of Experimental Agriculture 44, 163-171. Biondini ME, Lauenroth WK, Sala OE (1991) Correcting estimates of net primary production: are we overestimating plant production in rangelands? Journal of Range Management 44, 194-198. Blunt CG, Haydock KP (1978) Effect of irrigation, nitrogen and defoliation on pangola grass in the dry season at the Ord Valley, north-western Australia. Australian Journal of Experimental Agriculture and Animal Husbandry 18, 825-833. Bolger TP, Turner NC (1999) Water use efficiency and water use of Mediterranean annual pastures in southern Australia. Australian Journal of Agricultural Research 50, 1035-1046. Bortolussi G, McIvor JG, Hodgkinson JJ, Coffey SG, Holmes CR (2005a) The northern Australian beef industry, a snapshot. 1. Regional enterprise activity and structure. Australian Journal of Experimental Agriculture 45, 1057-1073. Bortolussi G, McIvor JG, Hodgkinson JJ, Coffey SG, Holmes CR (2005b) The northern Australian beef industry, a snapshot. 3. Annual liveweight gains from pasture based systems. Australian Journal of Experimental Agriculture 45, 1093-1108. Bridge BJ, Mott JJ, Hartigan RJ (1983) The formation of degraded areas in the dry savanna woodlands of northern Australia. Australian Journal of Soil Research 21, 91-104. Burrows WH, Carter JO, Scanlan JC, Anderson ER (1990) Management of savannas for livestock production in northeast Australia: Contrasts across the tree grass continuum. Journal of Biogeography 17, 503-512. Buxton R, Stafford Smith DM (1996) Managing drought in Australia's rangelands: Four weddings and a funeral. Rangeland Journal 18, 2292-2308. Cafe LM, McKeon GM, Dyer RM, Heywood JM (1999) Trees compete with pastures in the semi-arid tropics of the Northern Territory. In 'People and rangelands: building the future - VI

References

313

International Rangeland Congress'. Townsville, Queensland, Australia. (Eds D Eldridge and D Freudenberger) pp. 291-292 Calder IR (1992) A model of transpiration and growth of Eucalyptus plantation in water-limited conditions. Journal of Hydrology 130, 1-15. Carter JO, Flood N, et al. (1996a) Development of data rasters for model inputs. In 'Development of a National Drought Alert Strategic Information System. Volume 3. Final report on QPI 20'. (Land and Water Resources Research and Development Corporation) Carter JO, Flood N, McKeon GM, Peacock A, Beswick A (1996b) 'Development of a National Drought Alert Strategic Information System Volume 4: Model framework, parameter derivation, model calibration, model validation, model outputs, web technology.' Dept of Natural Resources, Indooroopilly, Final Project Report, QPI20 to the Land and Water Resources Research and Development Corporation (LWRRDC), Brisbane, Qld. Christian CS, Stewart GA (1953) 'General Report on survey of Katherine-Darwin region, 1946.' Division of Land Research and Regional Survey, CSIRO, Land Research Series No.1, Melbourne, Vic. Christie EK (1975a) Physiological responses of semiarid grasses. II. The pattern of root growth in relation to external phosphorus concentration. Australian Journal of Agricultural Research 26, 437-446. Christie EK (1975b) Physiological responses of semiarid grasses. III. Growth in relation to temperature and soil water deficit. Australian Journal of Agricultural Research 26, 447-457. Christie EK (1978) Ecosystem processes in semiarid grasslands. I. Primary production and water use of two communities possessing different photosynthetic pathways. Australian Journal of Agricultural Research 29, 773-787. Christie EK (1979) Ecosystem processes in semi-arid grasslands. II. Litter production, decomposition and nutrient dynamics. Australian Journal of Agricultural Research 30, 29-42. Christie EK (1981) Biomass and nutrient dynamics in C4 semi-arid Australian grassland community. Journal of Applied Ecology 18, 907-918. Christie EK, Moorby J (1975) Physiological responses of semiarid grasses. I. The influence of phosphorus supply on growth and phosphate absorption. Australian Journal of Agricultural Research 26, 423-436. Colls K, Whittaker R (2001) 'The Australian weather book.' (Reed New Holland: Sydney, NSW) Condon RW (1986) 'A reconnaissance erosion survey of part of the Victoria River District, NT.' Hassall and Associates for the Conservation Commission of the Northern Territory, Canberra, ACT. Conroy J, Seneweera S, Ghannoum O, Rudman S, Camilleri M (1998) The effect of rising atmospheric CO2 concentrations on tropical grasses. Tropical Grasslands 11, 26-28. Cook GD, Andrew MH (1991) The nutrient capital of indigenous Sorghum species and other understory components of savannas in north-western Australia. Australian Journal of Ecology 16, 375-384. Cooksley DG, Prinsen JH, Paton CJ, Lisle AT (1993) Rate of decomposition and release of nutrients from litter in a Heteropogon contortus - Bothriochloa bladhii dominant native pasture. In 'Grasslands

References

314

for our world - proceedings of the XVII International Grasslands Congress'. Palmerston North, New Zealand pp. 1557-1559. (SIR Publishing, Wellington) CSIRO (1970) 'Lands of the Ord-Victoria Area, Western Australia and Northern Territory.' Division of Land Research and Regional Survey, CSIRO, Land Research Series No. 28, Melbourne, Vic. Dalgliesh NP, Foale MA (1998) 'Soil matters - monitoring soil water and nutrients in dryland farming.' (Agricultural Production Systems Research Unit: Toowoomba, Qld) Dalgliesh NP, McCown RL, Bridge B, Probert ME, Cawthray S (1998) Characterising soils for plant available water capacity - challenges on shrink/swell soils. In '9th Australian Agronomy Conference'. Wagga Wagga, New South Wales pp. 793-794 DataDrill (2005) 'http://www.nrm.qld.gov.au/silo/datadrill.' Day KA, Dube S, Unganai L, MacLaurin AR, Dye PJ, Hall WB, Crimp SJ, McKeon GM (1999) Accounting for year-to-year variability in pasture growth: Applying a Queensland approach to Zimbabwe. In 'People and rangelands: building the future - VI International Rangeland Congress'. Townsville, Queensland, Australia. (Eds D Eldridge and D Freudenberger) pp. 859-860 Day KA, MacLaurin AR, Dube S, Hlatshwayo A, Trevor C (2003) 'Capturing the benefits of seasonal climate forecasts in agricultural management: Sub-project 3 - grazing systems in Zimbabwe.' Qld Department of Natural Resources and Mines, Indooroopilly, Final Project Report, LWR2/96/215 to the Australian Centre for International Agriculture Research (ACIAR), Brisbane, Qld. Day KA, McKeon GM, Carter JO (1997a) 'Evaluating the risks of pasture and land degradation in native pastures in Queensland.' Dept of Primary Industries and Dept Natural Resources, Indooroopilly, Final Project Report, DAQ124A to the Rural Industries Research and Development Corporation (RIRDC), Brisbane, Qld. Day KA, McKeon GM, Orr DM (1993) Comparison of methods for assessing productivity of native pastures in Queensland. In 'Grasslands for our world - proceedings of the XVII International Grasslands Congress'. Palmerston North, New Zealand. (Ed. MJ Baker) pp. 784-785. (SIR Publishing, Wellington) Day KA, Philp MW (1997) 'Appendix 3: SWIFTSYND methodology - a methodology for measuring a minimum data set for calibrating pasture and soil parameters of the pasture growth model GRASP.' Dept of Primary Industries and Dept Natural Resources, Indooroopilly, Final Project Report, DAQ124A to the Rural Industries Research and Development Corporation (RIRDC), Brisbane, Qld. Day KA, Scattini WJ, Osborne JC (1997b) Extending carrying capacity calculations to the central Burnett region of Queensland. In 'DroughtPlan Working Paper No 10'. (Ed. R Clark). (CSIRO: Alice Springs, NT) Day KJ (1977) 'Fertility studies on three red earth soils of the Day Basin, Northern Territory.' Animal Industry and Agriculture Branch, Department of the Northern Territory, Technical Bulletin No. 22, Darwin, NT. Dimes JP, McCown RL, Saffigna PG (1996) Nitrogen supply to no-tillage crops, as influenced by mulch type, soil type and season, following pasture leys in the semi-arid tropics. Australian Journal of Experimental Agriculture 36, 937-946. DIPE (unpublished) Land units of the Victoria River District, NT.

References

315

Doley D, Trivett NBA (1974) Effects of low water potentials on transpiration and photosynthesis in Mitchell grass (Astrebla lappacea). Australian Journal of Plant Physiology 1, 539-550. Downes RW (1969) Differences in transpiration rates between tropical and temperate grasses under controlled conditions. Planta 88, 261-273. DPIF, ORD (2001) 'Northern Territory Primary Industry and Fisheries Statistics 1999-2000.' Department of Primary Industry and Fisheries and The Office of Resource Development, Technical Bulletin No. 284, Darwin, NT. Dunlop CR (1990) 'Checklist of vascular plants of the Northern Territory, Australia.' (Parks and Wildlife Commission of the Northern Territory: Darwin, NT) Dupont G, Carter JO, Mott JJ, Yates D (1996) Tree micro-climate effects on understorey vegetation: their impact on spatial and temporal modelling of pasture production in Queensland. In '0th Biennial Conference'. Port Augusta, South Australia pp. 223-224. (Australian Rangeland Society) Dyer RM, Cafe LM, Craig A (2001a) 'Australian Grassland and Rangeland Assessment by Spatial Simulation (AussieGRASS): NT and Kimberley Rangeland Sub-project.' Department of Natural Resources and Mines, Indooroopilly, Final Project Report, QNR-9 for the Climate Variability in Agriculture Program (CVAP), Brisbane, Qld. Dyer RM, Jacklyn P, Partridge IJ, Russell-Smith J, Williams RJ (2001b) 'Savanna burning: Understanding and using fire in northern Australia.' (Tropical Savannas CRC: Darwin, NT) Eastham J, Gregory PJ, Williamson D, Watson GD (1992) Evaporation and transpiration components of evapotranspiration from early and late sown wheat and lupins on a duplex soil. Australian Journal of Experimental Agriculture 32, 993. Efron B, Gong G (1983) A leisurely look at the bootstrap, the jackknife, and cross-validation. American Statistician 37, 36-48. Evans DG, Gruba P (2002) 'How to write a better thesis.' (Melbourne University Press: Carlton, Vic.) Eyles AG, Cameron DG, Hacker JB (1984) 'Pasture research in northern Australia - its history, achievements and future emphasis.' Division of Tropical Crops & Pastures, CSIRO, Research Report No. 4, Melbourne, Vic. Farre I, Robertson MJ, Asseng S, French RJ, Dracup M (2004) Simulating lupin development, growth and yield in a Mediterranean environment. Australian Journal of Agricultural Research 55, 863-877. Fitzpatrick EA, Nix HA (1970) The climatic factor in Australian grassland ecology. In 'Australian grasslands'. (Ed. RM Moore) pp. 1-26. (Australian National University Press: Canberra, ACT) Foran BD, Bastin GN (1984) The dynamics of a Mitchell grass (Astrebla spp.) rangeland on the Barkly Tableland, Northern Territory. Australian Rangelands Journal 6, 92-97. Foran BD, Bastin GN, Hill B (1985) The pasture dynamics and management of two rangeland communities in the Victoria River District of the Northern Territory. Australian Rangelands Journal 7, 107-113. Forster BA, Laity JR (1972) 'Report on the land units of Victoria River Research Station - Kidman Springs.' Land Conservation Section, Primary Industry Branch, Northern Territory Administration, Darwin, NT.

References

316

Friedel MH (1981) Studies of central Australian semidesert rangelands. II. Range condition and the nutrient dynamics of the herbage layer, litter and soil. Australian Journal of Botany 29, 233-245. Friedel MH, Cellier KM, Nicolson KP (1980) Nutrient deficiencies in central Australian semi-desert rangelands, with reference to decline in range condition. Australian Rangelands Journal 2, 151-161. Gardener CJ, McCaskill MR, McIvor JG (1993) Herbage and animal production from native pastures and pastures oversown with Stylosanthes hamata 1. Fertiliser and stocking rate effects. Australian Journal of Experimental Agriculture 33, 651-570. Gardener CJ, McIvor JG (1985) Pasture stability in the semi-arid tropics of north eastern Australia. In 'Ecology and Management of the Worlds Savannas - International Savanna Symposium in Brisbane, Qld, 1984'. (Eds JC Tothill and JJ Mott) pp. 314-317. (Australian Academy of Science in conjunction with CAB: Canberra, ACT) Goldman RN, Weinberg JS (1985) 'Statistics: An introduction.' (Prentice Hall: Englewood Cliffs, New Jersey, USA) Graham RA, Florentine SK, Fox JED, Luong TM (2004) The germinable soil seedbank of Eucalyptus victrix grassy woodlands at Roy Hill station, Pilbara district, Western Australia. Rangeland Journal 26, 17-33. Grant WE, Pedersen EK, Marin SL (1997) 'Ecology and natural resource management: systems analysis and simulation.' (John Wiley and Sons: New York, USA) Greacen EL, Williams J (1983) Physical properties and water relations. In 'Soils: an Australian viewpoint'. (Ed. CSIRO) pp. 509-530. (Division of Soils, CSIRO: Melbourne, Vic) Hacker RB (1984) Vegetation dynamics in a grazed mulga shrubland community. II. The ground story. Australian Journal of Botany 32, 251-161. Hall WB, Bruget D, Carter JO, McKeon GM, Yee Yet JS, Peacock A, Hassett RC, Brook KD (2001) 'Australian grassland and rangeland assessment by spatial simulation (Aussie GRASS).' Dept of Natural Resources and Mines, Final Project Report, QNR-9 to the Climate Variability in Agriculture Program (CVAP), Brisbane, Qld. Hall WB, McKeon GM, Carter JO, Day KA, Howden SM, Scanlan JC, Johnston PW, Burrows WH (1998) Climate change and Queensland's grazing lands: II. An assessment of the impact on animal production from native pastures. Rangeland Journal 20, 174-202. Hammer GL, Nicholls N, Mitchell C (Eds) (2000) 'Applications of seasonal climate forecasting in agricultural and natural ecosystems - the Australian experience.' (Kluwer Academic Publishers: Dordrecht, The Netherlands) Harrington GN, Wilson AD, Young MD (Eds) (1984) 'Management of Australia's rangelands.' (CSIRO Division of Wildlife and Rangelands Research: Melbourne, Vic) Harrison SR (1990) Regression of a model on real-system output: an invalid test of model validity. Agricultural Systems 34, 183-190. Hasker PJS (2000) 'Beef cattle performance in northern Australia: summary of recent recent research.' (Queensland Department of Primary Industry and Meat and Livestock Australia)

References

317

Hendricksen R, Rickert KG, Ash AJ, McKeon GM (1982) Beef production model. In 'Animal Production in Australia - 14th Biennial Conference'. Brisbane, Queensland pp. 204-208. (Australian Society of Animal Production) Hendricksen RE, Gilbert MA, Punter LD (1992) Effect of superphosphate application on macro-nutrient and micro-nutrient concentrations in grazed stylo-native grass pasture in tropical Australia. Australian Journal of Agricultural Research 43, 1725-1738. HeytesburyBeef, MLA (2005) 'Pigeon Hole Project 2002 - 2007, Pastoral Field Day Program - August 2005.' Hill BA (1976) 'Victoria River District 1976 survey (reprint).' Division of Primary Production, Technical Bulletin No. 82, Katherine, NT. Hill JO, Robertson MJ, Pengelly BC, Whitbread AM, Hall CA (2006) Simulation modelling of lablab (Lablab purpureus) pastures in northern Australia. Australian Journal of Agricultural Research 57, 389-401. Hochman Z, Dalgliesh NP, Bell KL (2001) Contributions of soil and crop factors to plant available soil water capacity of annual crops on black and grey vertosols. Australian Journal of Agricultural Research 52, 955-961. Holechek JL (1988) An approach for setting the stocking rate. Rangelands 10, 10-14. Holm AM, Eliot GJ (1980) Seasonal changes in the nutritive value of some native pasture species in north-western Australia. Australian Rangelands Journal 2, 175-182. Hutley LB, O'Grady AP, Eamus D (2000) Evapotranspiration from eucalypt open-forest savanna of northern Australia. Functional Ecology 14, 183-194. Isbell RF (1983) Kimberley-Arnhem-Cape York (III). In 'Soils: an Australian viewpoint'. (Ed. CSIRO) pp. 189-199. (Division of Soils, CSIRO: Melbourne, Vic) Isbell RF (1996) 'The Australian soil classification.' (CSIRO: Melbourne, Vic) Ivory DA, Whiteman PC (1978) Effect of temperature on growth of five subtropical grasses. I. Effect of day and night temperature on growth and morphological development. Australian Journal of Plant Physiology 5, 131-148. Jackson J, Ash AJ (1998) Tree-grass relationships in open eucalypt woodlands of north-eastern Australia: Influence of trees on pasture productivity, forage quality and species distribution. Agroforestry Systems 40, 159-176. Jackson J, Ash AJ (2001) The role of trees in enhancing soil nutrient availability for native perennial grasses in open eucalypt woodlands of north-east Queensland. Australian Journal of Agricultural Research 52, 377-386. Jacobsen CN (1981) A review of the species of Dichanthium native to Australia with special reference to their occurrence in Queensland. Tropical Grasslands 15, 84-95. Johnson D (2004) 'Geology of Australia.' (Cambridge University Press) Johnson RW, Tothill JC (1985) Definition and Broad Geographic Outline of Savanna Lands. In 'Ecology and Management of the Worlds Savannas - International Savanna Symposium in Brisbane,

References

318

Qld, 1984'. (Eds JC Tothill and JJ Mott) pp. 1-13. (Australian Academy of Science in conjunction with CAB: Canberra, ACT) Johnston PW (1996) Grazing capacity of native pastures in the mulga lands of south-western Queensland: a modelling approach. PhD thesis, University of Queensland. Johnston PW, McKeon GM, Day KA (1996a) Objective 'safe' grazing capacities for south-west Queensland Australia: Development of a model for individual properties. Rangeland Journal 18, 244-258. Johnston PW, Tannock PR, Beale IF (1996b) Objective 'safe' grazing capacities for south-west Queensland Australia: Model application and evaluation. Rangeland Journal 18, 259-269. Jones PN, Carberry PS (1994) A technique to develop and validate simulation models. Agricultural Systems 46, 427-442. Jones PN, Carberry PS (1995) Optimal use of data in crop model development and validation. Environmetrics 6, 491-496. Jones RK, Probert ME, Dalgliesh NP, McCown RL (1996) Nitrogen inputs from a pasture legume in rotations with cereals in the semi-arid tropics of northern Australia: experimentation and modelling on a clay loam soil. Australian Journal of Experimental Agriculture 36, 985-994. Kalish JM, Livingston ME, Schofield KA (1996) Trace elements in the otoliths of New Zealand blue grenadier (Macruronus novaezelandiae) as an aid to stock discrimination. Marine and Freshwater Research 47, 537-542. Kirsch JAW, Springer MS, Lapointe F-J (1997) DNA-hybridisation studies of marsupials and their implications for metatherian classification. Australian Journal of Zoology 45, 211-280. Kraatz M (Ed.) (2000) 'Managing for healthy country in the VRD.' (Tropical Savannas CRC: Darwin, NT) Lawrie AC (1981) Nitrogen fixation by native Australian legumes. Australian Journal of Botany 29, 143-157. Lazarides M, Norman MJT, Perry RA (1969) 'Wet season development pattern of some native grasses in Katherine, NT.' CSIRO Division of Land Research, Technical Paper No. 26. Littleboy M, McKeon GM (1997) 'Appendix 2: Subroutine GRASP - grass production model, Marcoola version of subroutine GRASP.' Dept of Primary Industries and Dept Natural Resources, Indooroopilly, Final Project Report, DAQ124A to the Rural Industries Research and Development Corporation (RIRDC), Brisbane, Qld. Lodge GM, Murphy SR, Harden S (2003a) Effects of grazing and management on herbage mass, persistance, animal production and soil water content of native pastures. 1. A redgrass-wallaby grass pasture, Barraba, North West Slopes, New South Wales. Australian Journal of Experimental Agriculture 43, 875-890. Lodge GM, Murphy SR, Harden S (2003b) Effects of grazing and management on herbage mass, persistance, animal production and soil water content of native pastures. 2. A mixed native pasture, Manilla, North West Slopes, New South Wales. Australian Journal of Experimental Agriculture 43, 891-905.

References

319

Lu P, Yunusa IA, Walker RR, Muller WJ (2003) Regulation of canopy conductance and transpiration and their modelling in irrigated grapevines. Functional Plant Biology 30, 689-698. Ludlow MM (1985) Photosynthesis and dry matter production in C3 and C4 pasture plants, with special emphasis on tropical C3 legumes and C4 grasses. Australian Journal of Plant Physiology 12, 557-572. MacDonald RN, Orton J, Hearnden M (1997) 'Mt Sanford project pasture trends and calving patterns 1993-96.' Dept of Primary Industry and Fisheries, Katherine, NT. MacLeod ND, Ash AJ, McIvor JG (2004) An economic assessment of the impact of grazing land condition on livestock performance in tropical woodlands. Rangeland Journal 26, 49-71. Mayer DG, Butler DG (1993) Statistical validation. Ecological modelling 68, 21-32. Mayer DG, Stuart MA, Swain AJ (1994) Regression of real-world data on model output: an appropriate overall test of validity. Agricultural Systems 45, 93-104. McAlpine JR (1970) Estimating pasture growth periods and droughts from simple water balance models. In '11th International Grassland Congress'. Surfers Paradise, Queensland, Australia pp. 484-487 McCown RL (1971) Available water storage in a range of soils in north-eastern Queensland. Australian Journal of Experimental Agriculture and Animal Husbandry 11, 343-348. McCown RL (1973) An evaluation of the influence of available soil water storage capacity on growing season length and yield of tropical pastures using simple water balance models. Agricultural Meteorology 11, 53-63. McCown RL (1981a) The climatic potential for beef cattle production in tropical Australia: Part I - Simulating the annual cycle of liveweight change. Agricultural Systems 6, 303-317. McCown RL (1981b) The climatic potential for beef cattle production in tropical Australia: Part III - Variation in the commencement, cessation and duration of the green season. Agricultural Systems 7, 163-178. McCown RL, Gillard P, Edye LA (1974) The annual variation in yield of pastures in the seasonally dry tropics of Queensland. Australian Journal of Experimental Agriculture and Animal Husbandry 14, 328-333. McCown RL, Hammer GL, Hargreaves JNG, Holzworth DP, Freebairn DM (1996) APSIM: A novel software system for model development, model testing and simulation in agricultural systems research. Agricultural Systems 50, 255-271. McDonald RC, Isbell RF, Speight JG, Walker J, Hopkins MS (1990) 'Field handbook (2nd Ed.) - Australian soil and land survey handbook.' (Inkata Press) McIvor JG (1984) Phosphorus requirements and responses of tropical pasture species: Native and introduced grasses, and introduced legumes. Australian Journal of Experimental Agriculture and Animal Husbandry 24, 370-378. McIvor JG, Ash AJ, Cook GD (1995a) Land condition in the tropical tall-grass pasture lands. 1. Effects on herbage production. Rangeland Journal 17, 69-85.

References

320

McIvor JG, Ash AJ, Mott JJ, Andrew MH (1994) Ecology of grazed tropical tallgrass rangelands in northern Australia. In '8th Biennial Conference'. Katherine, Northern Territory pp. 115-124. (Australian Rangeland Society) McIvor JG, Gardener CJ (1991) Soil seed densities and emergence in pastures in the seasonally dry tropics of northeastern Australia. Australian Journal of Ecology 16, 159-169. McIvor JG, Gardener CJ (1994) Germinable soil seed banks in native pastures in north-eastern Australia. Australian Journal of Experimental Agriculture 34, 1113-1119. McIvor JG, Williams J, Gardener CJ (1995b) Pasture management influences runoff and soil movement in the semi-arid tropics. Australian Journal of Experimental Agriculture 35, 55-65. McKeon GM, Brook KD, Carter JO, Day KA, Howden SM, Johnston PW, Scanlan JC, Scattini WJ (1994) Modelling utilisation rates in the black speargrass zone of Queensland. In '8th Biennial Conference'. Katherine, Northern Territory pp. 128-132. (Australian Rangeland Society) McKeon GM, Carter JO, Day KA, Hall WB, Howden SM (1998) 'Evaluation of the impact of climate change on northern Australian grazing industries.' Dept of Natural Resources, Indooroopilly, Final Project Report, DAQ139A to the Rural Industries Research and Development Corporation (RIRDC), Brisbane, Qld. McKeon GM, Carter JO, Johnston PW, Orr DM (1988) Native pasture production in Queensland - a methodology for research (GUNSYND). In '5th Biennial Conference'. Longreach, Queensland pp. 115-118. (Australian Rangeland Society) McKeon GM, Day KA, Howden SM, Mott JJ, Scattini WJ, Weston EJ (1990) Northern Australian savannas: Management for pastoral production. Journal of Biogeography 17, 355-372. McKeon GM, Hall WB, Henry B, Stone G, Watson I (Eds) (2004) 'Pasture degradation and recovery in Australia's rangelands: Learning from history.' (Dept of Natural Resources, Mines and Energy, Indooroopilly: Brisbane, Qld) McKeon GM, Johnston PW (1990) Native pasture production in Queensland - GUNSYND (Grass Under Nutritional Stability: Yield Nitrogen and Development). In 'Native Grass Workshop'. Dubbo, New South Wales p. 46. (Australian Wool Corporation) McKeon GM, Paull CJ, Peacock A (1996) 'Development of a National Drought Alert Strategic Information System Volume 5: Evaluation of model performance relative to rainfall, inter-state calibration, extension and state comments.' Dept of Natural Resources, Indooroopilly, Final Project Report, QPI20 to the Land and Water Resources Research and Development Corporation (LWRRDC), Brisbane, Qld. McKeon GM, Rickert KG (1984) A computer model of the integration of forage options for beef production. In 'Animal Production in Australia - 15th Biennial Conference'. Brisbane, Queensland pp. 15-19. (Australian Society of Animal Production) McKeon GM, Rickert KG, Ash AJ (1982a) Simulation studies. In '14th Biennial Conference'. Brisbane, Queensland pp. 208-210. (Australian Society of Animal Production) McKeon GM, Rickert KG, Ash AJ, Cooksley D, Scattini WJ (1982b) Pasture production model. In '14th Biennial Conference'. Brisbane, Queensland pp. 201-204. (Australian Society of Animal Production)

References

321

Mitchell PL (1997) Misuse of regression for empirical validation of models. Agricultural Systems 54, 313-326. Mitchell PL, Sheehy JE (1997) Comparison of predictions and observations to assess model performance: A method of empirical validation. In 'Second International Symposium on Systems Approaches for Agricultural Development'. Los Banos, Philippines. (Eds MJ Kropff, PS Teng, PK Aggarwal, J Bouma, BAM Bouman, JW Jones and HH van Laar) pp. 437-451. (Kluwer Academic Publishers, Dordrecht, Netherlands) MLA (2001) 'Northern Nutrition: An educational package for beef producers of northern Australia.' (Meat and Livestock Australia) MLA (2004a) 'Grazing land management, Katherine version - workshop notes.' (Meat and Livestock Australia) MLA (2004b) 'Grazing land management, technical manual.' (Meat and Livestock Australia) MLA (2005) 'Livestock production research and development strategic plan 2006-2011.' Meat and Livestock Australia, Sydney, NSW. Moore GA (1991) 'Crossing the chasm: marketing and selling high-tech products to mainstream customers.' (Harper Business: New York, USA) Moore P (2005) 'A guide to plants of inland Australia.' (Reed New Holland: Sydney, NSW) Mott JJ (1972) Germination studies on some annual species from an arid region of Western Australia. Journal of Ecology 60, 293-304. Mott JJ (1978) Dormancy and germination in five native grass species from savanna woodland communities of the Northern Territory. Australian Journal of Botany 26, 621-631. Mott JJ, Andrew MH (1985) The effect of fire on the population dynamics of native grasses in tropical savannas of north-west Australia. In 'Proceedings of the Ecological Society of Australia' pp. 231-239 Mott JJ, Williams J, Andrew MH, Gillison AN (1985) Australian Savanna Ecosystems. In 'Ecology and Management of the Worlds Savannas - International Savanna Symposium in Brisbane, Qld, 1984'. (Eds JC Tothill and JJ Mott) pp. 56-82. (Australian Academy of Science in conjunction with CAB: Canberra, ACT) Mott JJ, Winter WH, McLean RW (1989) Management options for increasing the productivity of tropical savanna pastures. IV Population biology of introduced Stylosanthes spp. Australian Journal of Agricultural Research 40, 1227-1240. Muchow RC (Ed.) (1985) 'Agro-research in the Semi-Arid Tropics: North West Australia.' (University of Queensland Press, St Lucia: Brisbane, Qld) Murphy SR, Lodge GM, Harden S (2004) Soil surface water dynamics in pastures in northern New South Wales. 3. Evapotranspiration. Australian Journal of Experimental Agriculture 44, 571-583. Niethe G, Bertram J, Carpenter M (1988) Animal health and production in the Northern Territory. In 'Animal Production in Australia - 17th Biennial Conference'. Sydney, New South Wales pp. 74-76. (Australian Society of Animal Production)

References

322

Norman MJT (1960) 'Grazing and feeding trials with beef cattle at Katherine, NT.' CSIRO, Melbourne, Vic. Norman MJT (1962) Response of native pasture to nitrogen and phosphate fertiliser at Katherine, NT. Australian Journal of Experimental Agriculture and Animal Husbandry 2, 221-227. Norman MJT (1963a) The pattern of dry matter and nutrient content changes in native pastures at Katherine, NT. Australian Journal of Experimental Agriculture and Animal Husbandry 3, 119-124. Norman MJT (1963b) The short-term effects of time and frequency of burning on native pastures at Katherine, NT. Australian Journal of Experimental Agriculture and Animal Husbandry 3, 26-29. Norman MJT (1965) Seasonal performance of beef cattle on native pasture at Katherine, NT. Australian Journal of Experimental Agriculture and Animal Husbandry 5, 227-231. Norman MJT (1967) The comparative performance of tropical crossbred and local shorthorn beef cattle on native pastures at Katherine, NT. Australian Journal of Experimental Agriculture and Animal Husbandry 7, 217-224. Norman MJT (1969) The effect of burning and seasonal rainfall on native pasture at Katherine, NT. Australian Journal of Experimental Agriculture and Animal Husbandry 9, 295-298. Norman MJT, Wetselaar R (1960) Losses of nitrogen on burning native pasture at Katherine, NT. Journal of Australian Institute of Agricultural Science 26, 272-273. Northcote KH (1979) 'A factual key for the recognition of Australian soils - 4th edition.' (Rellim Technical Publications, Glenside: Adelaide, SA) Northcote KH, Hubble GD, Isbell RF, Thomson CH, Bettenay E (1975) 'A description of Australian soils.' (CSIRO: Melbourne, Vic) Novelly P, Baird P (2001) A process for re-assessing the potential carrying capacity of Kimberley pastoral leases. In 'Northern Australia Beef Industry Conference'. Kununurra, Western Australia pp. 41-47 O'Rourke PK, Winks L, Kelly AM (1992) 'North Australia Beef Industry Survey 1990.' Queensland Department of Primary Industry and the Meat Research Corporation, Brisbane, Qld. Orr DM (1981) Changes in the quantitative floristics in some Astrebla spp. (Mitchell grass) communities in south-western Queensland in relation to trends in seasonal rainfall. Australian Journal of Botany 29, 533-545. Orr DM (1991) Trends in the recruitment of Astrebla spp. in relation to seasonal rainfall. Rangeland Journal 13, 107-117. Orr DM, Evenson CJ (1991a) Effect of a single clipping on the seed production response of Astrebla spp. Rangeland Journal 13, 57-60. Orr DM, Evenson CJ (1991b) Effects of sheep grazing Astrebla grasslands in central western Queensland III. Dynamics of Astrebla spp. under grazing and exclosure between 1975 and 1986. Rangeland Journal 13, 36-46. Owens JS, Silburn DM, McKeon GM, Carroll C, Willcocks J, deVoil R (2003) Cover-runoff equations to improve simulation of runoff in pasture growth models. Australian Journal of Soil Research 41, 1467-1488.

References

323

Oxley T (2006) 'Pastoral industry survey 2004, Katherine.' Department of Primary Industry, Fisheries and Mines, Katherine, NT. Oxley T, Leigo S, Hausler P, Bubb A, MacDonald RN (2006) 'Pastoral industry survey 2004, NT.' Department of Primary Industry, Fisheries and Mines, Darwin, NT. Oxley T, Walker S (2003) 'Economic appraisal of pasture improvement in the Katherine Region.' Department of Primary Industry, Fisheries and Mines, Internal report, Katherine, NT. Parker D, Cobiac MD (2001) 'An introduction to pastures of the Katherine region Part 2. Common uses for introduced pasture species.' Department of Primary Industry and Fisheries, Agnote No. E69, Katherine, NT. Partridge IJ (1994) 'Will it rain? - The effects of the Southern Oscillation and El Nino on Australia.' (QDPI: Brisbane, Qld) Paterson SJ (1970) Geomorphology of the Ord-Victoria Area. In 'Lands of the Ord-Victoria Area, Western Australia and Northern Territory'. (Ed. CSIRO) pp. 83-91. (Division of Land Research and Regional Survey, CSIRO: Melbourne, Vic) Peoples MB, Baldock JA (2001) Nitrogen dynamics of pastures: nitrogen fixation inputs, the impact of legumes on soil nitrogen fertility, and the contributions of fixed nitrogen to Australian farming systems. Australian Journal of Experimental Agriculture 41, 327-346. Perry RA (1960) 'Pasture Lands of the Northern Territory.' Division of Land Research and Regional Survey, CSIRO, Land Research Series No. 5, Melbourne, Vic. Perry RA (1970) Pasture Lands of the Ord-Victoria Area. In 'Lands of the Ord-Victoria Area, Western Australia and Northern Territory'. (Ed. CSIRO) pp. 120-125. (Division of Land Research and Regional Survey, CSIRO: Melbourne, Vic) Petheram RJ, Kok B (1983) 'Plants of the Kimberley region of Western Australia.' (University of Western Australia Press: Perth, WA) Power M (1993) The predictive validation of ecological and environmental models. Ecological modelling 68, 33-50. Rayment GE, Higginson FR (1982) 'Australian laboratory handbook of soil and water chemical methods - Australian soil and land survey handbook vol.3.' (Inkata Press) Richards R, Watson I, Bean J, Maconochie J, Clipperton S, Beeston G, Green D, Hacker RB (2001) 'Australian grassland and rangeland asessment by spatial simulation (AussieGRASS), Southern Pastures Sub-project.' Department of Natural Resources and Mines, Indooroopilly, Final Project Report, QNR-9 (Southern Pastures Sub-project) for the Climate Variability in Agriculture Program (CVAP), Brisbane, Qld. Rickert KG, McKeon GM (1982) Soil water balance model: WATSUP. In 'Animal Production in Australia - 14th Biennial Conference'. Brisbane, Qld pp. 198-200. (Australian Society of Animal Production) Rickert KG, McKeon GM (1985) Models for native pasture management and development in south-east Queensland. In 'Ecology and Management of the Worlds Savannas - International Savanna Symposium in Brisbane, Qld, 1984'. (Eds JC Tothill and JJ Mott) pp. 299-302. (Australian Academy of Science in conjunction with CAB: Canberra, ACT)

References

324

Rickert KG, Stuth JW, McKeon GM (2000) Modelling pasture and animal production. In 'Field and laboratory methods for grassland and animal production research'. (Eds L 't Mannetje and RM Jones) pp. 29-64. (CABI Publishing: Wallingford, UK) Robbins GB, Bushell JJ, McKeon GM (1989) Nitrogen immobilisation in decomposing litter contributes to productivity decline in ageing pastures of green panic (Panicum maximum var. trichoglume). Journal of Agricultural Science 113, 401-406. Robertson DJ (1980) 'The Victoria River District cattle industry survey 1979.' Division of Primary Production, Technical Bulletin No. 35, Katherine, NT. Robertson JA (1982) 'The Victoria River District cattle industry survey 1982.' Department of Primary Production, Technical Bulletin No. 66, Katherine, NT. Robertson MJ, Carberry PS, et al. (2002) Simulation of growth and development of diverse legume species in APSIM. Australian Journal of Agricultural Research 53, 429-446. Robinson CS (1971) 'Ecology of the Hardman Basin, NT.' Animal Industry and Agriculture Branch, Northern Territory Administration, Technical Bulletin No. 6, Darwin, NT. Rose CW, Begg JE, Byrne GF, Torssell BWR, Goncz JH (1972) A simulation model of growth-field relationships for Townsville style (Stylosanthes humilis H.B.K.) pasture. Agricultural Meteorology 10, 161-183. Rowell DL (1994) 'Soil science - methods and applications.' (Longman Scientific & Technical, UK) Rykiel EJ (1996) Testing ecological models: The meaning of validation. Ecological modelling 90, 229-244. SCA (1990) 'Feeding standards for Australian livestock: Ruminants.' (CSIRO: Melbourne, Vic.) Scanlan JC (2002) Some aspects of tree-grass dynamics in Queensland's grazing lands. Rangeland Journal 24, 56-82. Scanlan JC, McKeon GM, Day KA, Mott JJ, Hinton AW (1994) Estimating safe carrying capacities of extensive cattle-grazing properties within tropical, semi-arid woodlands of north-eastern Australia. Rangeland Journal 16, 64-76. Scanlan JC, Pressland AJ, Myles DJ (1996a) Grazing modifies woody and herbaceous components of north Queensland woodlands. Rangeland Journal 18, 47-57. Scanlan JC, Pressland AJ, Myles DJ (1996b) Runoff and soil movement on mid-slopes in north-east Queensland grazed woodlands. Rangeland Journal 18, 33-46. Scattini WJ, Powell EE (1982) Simulation of production variability from native pastures in south east Queensland. In 'Animal Production in Australia - 14th Biennial Conference'. Brisbane, Queensland p. 197. (Australian Society of Animal Production) Schmidt S, Lamble RE (2002) Nutrient dynamics in Queensland savannas: implications for the sustainability of land clearing for pasture production. Rangeland Journal 24, 96-111. Schmidt S, Stewart GR, Turnbull MH, Erskine PD, Ashwath N (1998) Nitrogen relations in natural and disturbed plant communities in tropical Australia. Oecologia 117, 95-105.

References

325

Scholes RJ (1993) Nutrient cycling in semi-arid grasslands and savannas: its influence on pattern, productivity and stability. In 'Grasslands for our world - proceedings of the XVII International Grasslands Congress'. Palmerston North, New Zealand. (Ed. MJ Baker) pp. 484-487. (SIR Publishing, Wellington) Shaw NH, Norman MJT (1970) Tropical and sub-tropical woodlands and grasslands. In 'Australian grasslands'. (Ed. RM Moore) pp. 112-122. (Australian National University Press: Canberra, ACT) Sinden JA, Thampapillai DJ (1995) 'Introduction to benefit-cost analysis.' (Longman Australia Pty Ltd: Melbourne, Vic) Slatyer RO (1954) A note on the available moisture range of some northern Australian soils. Journal of Australian Institute of Agricultural Science 20, 46-47. Slatyer RO (1960) 'Agricultural climatology of the Katherine area, NT.' Division of Land Research and Regional Survey, CSIRO, Technical Paper No.13, Melbourne, Vic. Slatyer RO (1970) Climate of the Ord-Victoria Area. In 'Lands of the Ord-Victoria Area, Western Australia and Northern Territory'. (Ed. CSIRO) pp. 62-74. (Division of Land Research and Regional Survey, CSIRO: Melbourne, Vic) Smiles DE (1997) Water balance in swelling materials: some comments. Australian Journal of Soil Research 35, 1143-1152. Smith EL (1960) Effects of burning and clipping at various times during the wet season on tropical tall grass range in northern Australia. Journal of Range Management 13, 197-203. Smith H (1998) 'The Victoria River District Land and Livestock Survey 1997.' NT Department of Primary Industry and Fisheries, Internal report, Katherine, NT. Spain AV, Isbell RF, Probert ME (1983) Soil organic matter. In 'Soils: an Australian viewpoint'. (Ed. CSIRO) pp. 551-. (Division of Soils, CSIRO: Melbourne, Vic.) SPLBPG (1995) 'Sustainable land management for beef production on the Sturt Plateau - based on producer experience.' Sturt Plateau Local Best Practice Group, Katherine, NT. Squires VR, Low WA (1987) Botanical and chemical composition of the diet selected by cattle in three range types in central Australia. Australian Rangelands Journal 9, 86-95. Stace HCT, Hubble GD, Brewer R, Northcote KH, Sleeman JR, Mulcahy MJ, Hallsworth EG (1968) 'A handbook of Australian soils.' (Rellim Technical Publications: Adelaide, SA) Stewart GA (1970a) Agricultural potential of the Ord-Victoria Area. In 'Lands of the Ord-Victoria Area, Western Australia and Northern Territory'. (Ed. CSIRO) pp. 126-134. (Division of Land Research and Regional Survey, CSIRO: Melbourne, Vic) Stewart GA (1970b) Soils of the Ord-Victoria Area. In 'Lands of the Ord-Victoria Area, Western Australia and Northern Territory'. (Ed. CSIRO) pp. 92-103. (Division of Land Research and Regional Survey, CSIRO: Melbourne, Vic) Stewart GA, Perry RA, Paterson SJ, Sleeman JR, Traves DM (1970) Land systems of the Ord-Victoria Area. In 'Lands of the Ord-Victoria Area, Western Australia and Northern Territory'. (Ed. CSIRO) pp. 11-61. (Division of Land Research and Regional Survey, CSIRO: Melbourne, Vic)

References

326

Stockwell TGH, Andison RT, Ash AJ, Bellamy JA, Dyer RM (1994) State and transition models for rangelands. 9. Development of state and transition models for pastoral management of the golden beard grass and limestone grass pasture lands of NW Australia. Tropical Grasslands 28, 260-265. Stone R, Smith I, McIntosh P (2000) Statistical methods for deriving seasonal climate forecasts from GCM's. In 'Applications of seasonal climate forecasting in agricultural and natural ecosystems - the Australian experience'. (Eds GL Hammer, N Nicholls and C Mitchell) pp. 135-147. (Kluwer Academic Publishers: Dordrecht, The Netherlands) Sweeney FC, Hopkinson JM (1975) Vegetative growth of nineteen tropical and sub-tropical pasture grasses and legumes in relation to temperature. Tropical Grasslands 9, 209-217. Taddese G, MohamedSaleem MA, Ayalneh W (2002) Effect of livestock grazing on physical properties of a cracking and self-mulching Vertisol. Australian Journal of Experimental Agriculture 42, 129-133. Talsma T, van der Lelij A (1976) Infiltration and water movement in an in situ swelling soil during prolonged ponding. Australian Journal of Soil Research 14, 337-349. Tanner CB, Sinclair TR (1983) Efficient water use in crop production: research or re-search? In 'Limitation to efficient water use in crop production'. (Eds HM Taylor, WR Jordan and TR Sinclair). (American Society of Agronomy: Madison, Wisconsin, USA) Taylor JA, Tulloch D (1985) Rainfall in the wet-dry tropics: Extreme events at Darwin and similarities between years during the period 1870-1983 inclusive. Australian Journal of Ecology 10, 281-295. Thumma BR, Naidu BP, Cameron DF, Bahnisch LM (1998) Transpiration efficiency and its relationship with carbon isotope discrimination under well-watered and water stressed conditions in Stylosanthes scabra. Australian Journal of Agricultural Research 49, 1039-1045. Tibby J, Reid MA (2004) A model for inferring past conductivity in low salinity waters derived from Murray River (Australia) diatom plankton. Marine and Freshwater Research 55, 597-607. Timmers PK, Clewett JF, Day KA, McKeon GM, Pinington GK, Scanlan JC (1999) WinGRASP - a modelling package for pasture growth and grazing systems in northern Australia. In 'People and rangelands: building the future - VI International Rangeland Congress'. Townsville, Queensland. (Eds D Eldridge and D Freudenberger) p. 881 Tothill JC, Gillies C (1992) 'The pasture lands of northern Australia - their condition, productivity and sustainability.' Tropical Grasslands Society, Occasional Publication No. 5, Brisbane, Qld. Tothill JC, Nix HA, Stanton JP, Russell MJ (1985) Land use and productive potentials of Australian savanna lands. In 'Ecology and management of the worlds savannas'. (Eds JC Tothill and JJ Mott) pp. 125-141. (Australian Academy of Science, in conjunction with CAB: Canberra, ACT) Traves DM, Dunn PR, Jones PJ (1970) Outline of the Geology of the Ord-Victoria Area. In 'Lands of the Ord-Victoria Area, Western Australia and Northern Territory'. (Ed. CSIRO) pp. 75-82. (Division of Land Research and Regional Survey, CSIRO: Melbourne, Vic) TSCRC (2006) 'http://savanna.ntu.edu.au/.' Vallance HJ, Cobiac MD, Andison RT, Stockwell TGH (1993) 'Important pasture species of the Victoria River District.' (NT Department of Primary Industry and Fisheries: Darwin, NT)

References

327

Vallentine JF (1990) 'Grazing management.' (Academic Press: New York, USA) van Dam JC (2000) Simulation of field-scale water-flow and bromide transport in a cracked clay soil. Hydrological Processes 14, 1101-1117. van Dam JC, Huygen J, Wesseling JC, Feddes RA, Kabat P, van Walsum PEV, Groenendijk P, van Diepen CA (1997) 'SWAP Version 2.0, Theory. Simulation of water flow, solute transport and plant growth in the Soil-Water-Atmosphere-Plant environment.' Department of Water Resources, Agriculture University, Technical Document 45, Wageningen, The Netherlands. Viscarra Rossel RA, McBratney AB (2001) A response-surface calibration model for rapid and versatile site-specific lime-requirement predictions in south eastern Australia. Australian Journal of Soil Research 39, 185-201. Walker BH, Ludwig D, Holling CS, Peterman RM (1981) Stability of semi arid savanna grazing systems. Journal of Ecology 69, 473-498. Walker GK, Richards JE (1985) Transpiration efficiency in relation to nutrient status. Agronomy Journal 77, 263-269. Wallach D, Goffinet B (1989) Mean squared error of prediction as a criterion for evaluating and comparing system models. Ecological modelling 44, 299-306. Watt LA (1978) Some characteristics of the germination of Queensland blue grass on cracking black earths. Australian Journal of Agricultural Research 29, 1147-1155. Westoby M, Walker B, Noy-Meir I (1989) Opportunistic management for rangelands not at equilibrium. Journal of Range Management 42, 266-274. Wetselaar R (1967) Estimation of nitrogen fixation by four legumes in a dry monsoonal area of north-western Australia. Australian Journal of Experimental Agriculture and Animal Husbandry 7, 518-522. Wetselaar R, Hutton JT (1963) The ionic composition of rainwater at Katherine, NT, and its part in the cycling of plant nutrients. Australian Journal of Agricultural Research 14, 319-329. Wetselaar R, Norman MJT (1960) Recovery of available soil nitrogen by annual fodder crops at Katherine, Northern Territory. Australian Journal of Agricultural Research 11, 693-704. Wheaton T (Ed.) (1994) 'Plants of the northern Australian rangelands.' (Dept of Lands, Housing and Local Government: Darwin, NT) Whiteman PC (1968) The effects of temperature on the vegetative growth of six tropical legume pastures. Australian Journal of Experimental Agriculture and Animal Husbandry 8, 528-532. Williams CH, Andrew CS (1970) Mineral nutrition of pastures. In 'Australian grasslands'. (Ed. RM Moore) pp. 321-338. (Australian National University Press: Canberra, ACT) Williams CH, Raupach M (1983) Plant nutrients in Australian soils. In 'Soils: an Australian viewpoint'. (Ed. CSIRO) pp. 777-793. (Division of Soils, CSIRO: Melbourne, Vic) Williams J, Day KJ, Isbell RF, Reddy SJ (1985) Soils and climate. In 'Agro-research in the Semi-Arid Tropics: North West Australia'. (Ed. RC Muchow) pp. 31-68. (University of Queensland Press, St Lucia: Brisbane, Qld)

References

328

Williams OB, Roe R (1975) Management of arid grasslands for sheep: plant demography of six grasses in relation to climate and grazing. Proceedings of the Ecological Society of Australia 9, 142-156. Wilson AD, Harrington GN, Beale IF (1984) Grazing management. In 'Management of Australia's rangelands'. (Eds GN Harrington, AD Wilson and MD Young) pp. 129-140. (CSIRO: Melbourne, Vic) Wilson BA, Brocklehurst PS, Clark MJ, Dickinson KJM (1990) 'Vegetation survey of the Northern Territory, Australia.' (Conservation Commission of the NT: Darwin, NT) Winter WH, McCosker TH, Pratchett D, Austin JDA (1985) Intensification of beef production. In 'Agro-research in the Semi-Arid Tropics: North West Australia'. (Ed. RC Muchow) pp. 395-418. (University of Queensland Press, St Lucia: Brisbane, Qld) Winter WH, McCown RL, Zuill D (1996) Legume-based pasture options for the live cattle trade from the Australian semi-arid tropics. Australian Journal of Experimental Agriculture 36, 947-955. Winter WH, Mott JJ, McLean RW (1989a) Evaluation of management options for increasing the productivity of tropical savanna pastures 2. Legume species. Australian Journal of Experimental Agriculture 29, 623-630. Winter WH, Mott JJ, McLean RW (1989b) Evaluation of management options for increasing the productivity of tropical savanna pastures 3. Trees. Australian Journal of Experimental Agriculture 29, 631-634. Winter WH, Mott JJ, McLean RW, Ratcliff D (1989c) Evaluation of management options for increasing the productivity of tropical savanna pastures 1. Fertiliser. Australian Journal of Experimental Agriculture 29, 613-622.

predicting native pasture growth in the victoria river ... · caz smith who are the current...

Documents