This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/authorsrights

Author's personal copy

Modelling pastoral farm systems — Scaling from farm to region

Iris Vogeler a,⁎, Ronaldo Vibart a, Alec Mackay a, Samuel Dennis b, Vicki Burggraaf c, Josef Beautrais a

a AgResearch, Grasslands Research Centre, Palmerston North, New Zealandb AgResearch, Lincoln Research Centre, Lincoln, Lincoln, New Zealandc AgResearch Limited, Ruakura Research Centre, Hamilton, New Zealand

H I G H L I G H T S

• An up-scaling approach was developed that linked farm modelling with land databases.• The up-scaling approach was compared to using representative farm systems.• Forty-five percent of New Zealand's Southland area has the potential for dairying.• Conversion into dairying could increase regional profit from farming by up to 75%.• Conversion into dairying would increase regional environmental impacts.

a b s t r a c ta r t i c l e i n f o

Article history:Received 3 December 2013Received in revised form 10 February 2014Accepted 21 February 2014Available online 20 March 2014

Keywords:DairyingSheep and beefLand conversionFarmax® Pro and Farmax® Dairy ProOverseer®

Farm system and nutrient budget models are increasingly being used to inform and evaluate policy options onthe impacts of land use change on regional environmental and economic performance. In this study, the commonapproach of up-scaling representative farm systems to a regional scale, with a limited input of resource informa-tion, was compared with a new approach that links a geospatial land resource information data base (NZLRI,Agribase™) that includes independent estimates of the productive capacity of land parcels, with individualfarm-scale simulation (Farmax® Pro and Farmax® Dairy Pro) and nutrient budgeting models (Overseer®). TheSouthland region of New Zealand, which is currently undergoing enormous land use change, was used as acase study. Model outputs from the new approach showed increased profit of about 75% for the region if the cur-rent land area under dairying increases from 16% to 45%, with the shift to dairy constrained to high pasture pro-duction classes only. Environmental impacts associated with the change were substantial, with nitrate leachingestimated to increase by 35% and greenhouse gas emissions by 25%. Up-scaling of representative farm systemsto the regional scale with limited input of resource information predicted lower potential regional profit andhigher N leaching from dairy conversion. The new approach provides a farm scale framework that could easilybe extended to include different systems, different levels of farming performance and the use of mitigationtechnologies.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

New Zealand (NZ) contains 14.6 million ha of farm land, approxi-mately 75% of which is hill country, characterised by predominantland slopes above 15°. The major land use on hill country in NZ issheep and beef (S&B) farming, whereas on flat and rolling landscapes,land use is more intensive, comprising dairying and a diversity of horti-cultural and arable farming. In the last years NZ has seen a large shift inland use, with a large number of dairy conversions drivenmainly by thegreater profitability from dairying relative to S&B (Beukes et al., 2011a).

The national sheep flock has decreased in the last 5 years from 38 mil-lion to 31 million sheep,whereas thenational dairy herd increased from5.3 million to 6.4 million cows (Statistics New Zealand, 2012).

Southland, NZ's southernmost region, has a long tradition of sheepand beef farming (Monaghan et al., 2005). In the last 20 years(1990–2010) dairying has expanded from 100 to 850 herds, makingup 16% of the pastoral farmed area (New Zealand Dairy Statistics,2010–11). In the Southland region there is still a potential for furtherdairy conversion. Conversion from S&B to dairying sees a shift fromgrazed legume based pastures, relying largely on N input by legume fix-ation, to a greater use of nitrogen (N) fertiliser, irrigation, supplements(including concentrates) and off-farm grazing, enabling large increasesin per hectare production levels. As inmany parts of theworld concernsabout environmental effects from these intensive livestock operations,especially on nutrient enrichment of water bodies are increasing

Science of the Total Environment 482–483 (2014) 305–317

⁎ Corresponding author at: AgResearch, Grasslands Research Centre, Tennent Drive,Private Bag 11008, Palmerston North, New Zealand. Tel.: +64 6 3518328; fax: +64 63518003.

E-mail address: iris.vogeler@agresearch.co.nz (I. Vogeler).

http://dx.doi.org/10.1016/j.scitotenv.2014.02.1340048-9697/© 2014 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Author's personal copy

(Hamill andMcBride, 2003; Monaghan et al., 2007; Smith andWestern,2013).

The National Policy Statement for Freshwater Management (NPS,2011) directs Regional Councils to set water quality limits for freshwa-ter bodies. In cases where these limits are not met policies and plansare implemented to ensure that these are met in the future. Policy de-velopment to achieve these targets will require an extension of currentcontrols around point source discharges, which represent only a verysmall percentage of total discharges, to diffuse losses from agriculturalland (Elliott et al., 2008). Farm system and nutrient budget models areincreasingly being used to inform and evaluate policy options. Suchmodels can be used to explore the impacts of policies on the financialperformance of land owners, sector and the regional economy andemissions to the environment.

Farm system and nutrient budget models widely used to evaluatethe influence of farm systems and practices on on-farm production,profitability and emissions include the DairyNZ Whole Farm Model(Berntsen et al., 2003; Vogeler et al., 2012), DairyMod (Johnson et al.,2008) and Fasset (Berntsen et al., 2003; Beukes et al., 2008; Johnsonet al., 2008), Farmax® Pro and Farmax® Dairy Pro (www.farmax.co.nz), and Overseer (Wheeler et al., 2006). The Farmax models arewhole-farmdecision supportmodels that usemonthly estimates of pas-ture growth, and farm and herd information to determine the produc-tion and economic outcomes of managerial decisions. Further detailsand evaluation of the models for selected farm scenarios in NZ can befound in Bryant et al. (2010) and White et al. (2010). Overseer®, anutrient budget model, has been widely used in NZ as an agriculturalmanagement tool to support decision-making for on-farm nutrientmanagement. The model can also be used to explore the relationshipbetween production and emissions, including leaching, run-off andgreenhouse gas (GHG) emissions from the farm (Wheeler et al.,2008), and has been shown similar N leaching as the process basedSPASMO (Soil Plant Atmosphere System Model) model (Rutherford,2012). The model also contains a suite of N, phosphorus (P) and GHGmitigation options.

Aggregating on-farm performance, outputs and emissions producedby the modelling tools across landscapes to explore the implications ofland use and practice change at scales beyond individual farms requiresthe linking of the farm-based models to land resource information. Thisensures that interactions between diverse landscape units and farm sys-tems and practices are captured in the modelling exercise and reflectedin the outputs. Common practice in the investigation of land use changeis the up-scaling of outputs from modelling a “typical” farm system,which is assembled to represent the types of farms that would beexpected to be found in the region. This approach is relatively easy toimplement but does not account for variation in land resources andthe associated impacts the required changes in farm systems designhave on performance, outputs and emission levels. Alternatively, re-gions can be subdivided into smaller “uniform” areas with specific soilcharacteristics, land use and farm management. Models can then berun for these different areas and the outputs aggregated. Several catch-ment models, including, CLUES (Catchment Land Use EnvironmentalSustainability (Elliott et al., 2011)), SWAT (Soil and Water and Assess-ment Tool (Neitsch et al., 2011)), and the NIWA land use model(Rutherford, 2012) use this approach. CLUES has been applied in severallocations in NZ to assess the impact of land use change onwater qualityand socioeconomic factors at a regional or national scale and has beenextensively calibrated and tested by NIWA. The NIWA land use modelhas been used to assess the impact of land management on the qualityof receiving waters and aquatic plant growth in the Hawkes Bay regionof New Zealand. SWAT has been used extensively in the USA and inter-nationally, as well as in NZ. However, themodel is physically-based, anddifficulties with obtaining suitable input data and parameter valueshave been reported. Ekanayake and Davie (2004) found that themodel could predict mean annual N concentrations reasonably well,but not inter-annual variations. From this the authors concluded that

simpler models might be equally suited for predicting the impact ofland use change.

Another catchment model that has been used to inform policies andto assess relative costs of alternative environmental policies for ‘landuse optimisation’ is the New Zealand Forest and Agriculture RegionalModel (NZFARM). The model is designed to optimise economic out-come, based on various policy scenarios, and has been used to modeleconomic impacts of nutrient reduction policies in the Canterburyarea of NZ (Samarasinghe et al., 2011).

For catchment modelling estimation of deep drainage, stream andgroundwater attenuation is needed, which adds additional complexity.Catchment models also require external input data of N loss from land.Both Overseer and the process based SPASMO (Green et al., 2003) havebeen used to provide such data (Rutherford, 2012). Alternatively out-puts from other process based, such as the widely used open accessAPSIM (Agricultural Production Systems sIMulator) could be used toprovide N loss data from different land management (Vogeler et al.,2013). The drawback of such process basedmodels for estimating lossesof nutrients from land for different farm systems and different manage-ment is that it would be very time consuming and costly, and they alsorequire accurate model parameterisation and calibration.

As the focus of this study was to estimate the impact of land useand management on economic impacts, N losses and GHG emissionsfrom land, rather than looking at the effect on receiving waters, a newapproach was developed rather than using catchment models. Theapproach uses a geospatial land resource information database (NZLRI,Agribase™) that includes independent estimates of the productive ca-pacity of each land parcel, to inform individual farm-scale simulations(Farmax® Pro and Farmax® Dairy Pro) and nutrient budgeting models(Overseer®). Impacts were assessed from individual farms, using actualfarm boundaries. The approach was compared to an approach based onup-scaling representative farm systems to a regional scale, with limitedinput of land resource and pasture production information, to quantifythe impact of land use change on the economy and environment ofthe Southland region in NZ.

2. Methods

To estimate the regional economic and environmental impacts ofagricultural farming (dairying and S&B farming) in the SouthlandRegion of NZ two different scenarios were investigated, with:

• Scenario 1: the current mix of dairying and S&B• Scenario 2: conversion of S&B farm land into dairying using ageospatial land resource information database (NZLRI, Agribase™)that includes independent estimates of the productive capacity ofland to inform the farm-scale simulations.

Two different approaches for estimating the regional impacts of landuse change were then compared. The first approach (Approach 1) wasbased on linking farm system models with land resource and pastureproduction information, whereas the second (Approach 2) was basedon up-scaling representative farm systems to a regional scale.

The scenarios and approaches were explored in the following steps,with steps 4 to 6 being only used for Approach 1:

1. Defining the currentmix of land use based on land resource informa-tion, which provides the information for Scenario 1;

2. Estimating the productive potential of the land across the regionbased on land resource information, and classification into PastureProduction Classes (PPCs);

3. Estimation of the potential conversion of land into dairying, whichprovides information for Scenario 2;

4. Farm system modelling for dairying and S&B for various PPCs;5. Aggregation of the model results for individual farms;6. Aggregation of model outputs for the regional scale for Scenarios 1

and 2;

306 I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

7. Comparison of the newly developed up-scaling method (Approach 1)with the simpler method based on regional averages (Approach 2).

These steps are described in detail below.

2.1. Case study and associated land resource information

Southland, NZ's southernmost region (Fig. 1), was used to exam-ine the new approach of upscaling. The region covers an area of1.7 million ha, 65% of which was under pastoral farming in 2007(Statistics NZ, 2012). Based on information held in Agribase™(AsureQuality, 2010), a spatial and demographic census of all knownNZ farms 84% of this farmed area was used for mixed S&B, about 15%for dairy and the remainder for “other” stock (alpacas, beef, deer,goats, horses). By overlaying the individual farm parcels from Agribasewith information from the Land Resource Information (LRI) system forNZ (Landcare Research), we obtained the land area, Land Use Capability(LUC) classes and topography for each farm (Fig. 1), focusing here onthe two major farm types (S&B and dairy) and soil types (brown andpallic). This provided the information required for the current mix offarm enterprises in the region (Scenario 1).

The LUC classes were used, as described below, to estimate the pro-ductive potential of the land, as required for (i) settingup the farm-scalesimulation modelling, (ii) defining the land that can potentially be con-verted into dairying (Scenario 2), and (iii) up-scaling model outputs toobtain regional estimates of profit and environmental outputs.

The LUC, a land classification system, has been developed inNZ to as-sist sustainablemanagement of farm enterprises (Lynn et al., 2009). TheLUC classes are based on five factors including rock type, soil, slopeangle, erosion type and severity, and vegetation cover. Using the LUC,land is categorised into eight classes reflecting its potential sustainableuse, with classes 1 to 7 being potentially suitable for pastoral grazing,and LUC 1 having no limitations to use and the highest productivepotential.

The potential pasture production was estimated from the extendedlegend of the LUC worksheets, which provide the potential productivecapacity of a ryegrass/white clover mixed pasture (with the only N ad-dition beingby biologicalfixation)under a “typical” S&B farming systemfor each LUC land unit. These spatially discrete estimates of the sheep-carrying capacity (ewes/ha) for each land unit from the 1980swere con-verted into pasture production (kg drymatter (DM)/ha), assuming thaton LUC classes 1 to 4, the pasture consumed per ewewas 600 kgDMperannum compared with 550 kg DM/ewe on LUC classes 5 to 7, whereewe performance was lower (100% lambing). It was presumed thatonly 70% of the pasture grown was actually consumed by these ewes.A further assumption in the calculation of pasture production was a7% upwards adjustment of pasture production on existing dairyfarms to account for improved pasture genetics and differing grazingmanagement (Smith, 2012), so the pasture yield for S&B for each LUCwas multiplied by 1.07 to determine the base dairy pasture yield.In addition to this, dairy farms apply more N fertiliser than S&B farms,and an N response rate of 10 kg DM/kg N was assumed for this N;up to 150 kg N/ha/year was applied, producing an additional1500 kg DM/ha/year. The calculated pasture production figures(Fig. 2) agree with observations, with well managed dairy pastures onflat land achieving 16–18 t DM/ha/year and hill pastures typically inthe range of 5–12 t DM/ha/year (Valentine and Kemp, 2007).

The potential pasture production was then classified into PastureProduction Classes (PPCs), with PPC 1 having the highest productionof about 14 t/ha under S&B and 16 t/ha under dairy (Fig. 2). Of thefarmed area in Southland the percentage of land in each of these PPCclasses are: PPC1: 0.1%; PPC2: 13.4%; PPC3: 29.8%; PPC4: 20.3%; PPC5:2.2%; PPC6: 25.1%; and PPC7 9.1%. Together, PPCs 1–4, which are poten-tially suitable for intensive dairy use, account for 64% of the farmed areain Southland.

To further characterise the land on each farm for the purpose ofmodelling, the predominant soil order, topography and drainage pro-files were identified. The main soil order on each farm was definedusing the New Zealand Soil Classification (NZSC) soil orders (Landcare

Fig. 1. Land use classification (LUC) classes of land currently used for sheep and beef (left) and dairy farming (right) for the Southland region of NZ.

307I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

Research). For Southland, the predominant soil orders are brown(Dystrudepts in the US Soil Taxonomy) and pallic soils (AericFragiaqualf in the US Soil Taxonomy), covering about 50% and 25% ofthe area, respectively (Table 1). For simplicity, only the dominant soilorder on a farm was considered across the entire property. Modellingmultiple soils across a farm would be an additional step that would in-crease the complexity of the exercise.

The predominant topographies within a farm were obtained byoverlaying Agribase™ with slope data based on the digital elevationmodel (100 m DEM (Landcare Research, 2013)), and grouping theminto four categories corresponding to those in Overseer®: flat, 0–7°;rolling, 8–15°; easy hill, 16–25°; and hard hill, N25°. For each farm, theproportion of land in each slope category was determined and themodel scenarios were set up accordingly.

Two drainage classes were also defined based on the NZSC drainageclasses (Landcare Research), as follows: poorly drained: NZSC drainageclasses 1–3 (very poor, poor and imperfectly drained); and well-drained: NZSC drainage classes 4–5 (moderatelywell andwell drained).About 70% of the farmed area was identified as having well-drainedsoils, with the remainder having poorly drained soils. In Southland,poorly drained soils are generally mole and tile drained when used fordairying (J. Risk, Environment Southland, pers. comm.). For the purposeof the current modelling exercise, poorly drained soils under dairying(but not under S&B) had an artificial drainage system in place. The cap-ital and maintenance costs of implementing these drainage systemswere not considered.

For the purpose of this study, it was also assumed that there was nodifference in pasture production betweenwell and poorly drained soils.This could be added as another variable to the analysis, but it should

have aminor effect andwould increase the complexity of themodellingexercise.

The estimation of the farms that can potentially be converted intodairying (Scenario 2) was based on the actual individual farms fromthe Agribase database and used the characteristics and coverage(in hectares) of each PPC × drainage class on each farm. For this, we as-sumed that only farms that have a minimum of 230 ha in PPCs 1–4 canconvert into dairying. Farms that were already in dairying and did notmeet these assumptions remained in dairying. Furthermore, it wasassumed that all potentially convertible dairy farms that had an areaof ≥450 ha in PPCs 5–6 kept this land under S&B; if the area was inPPCs 5–6 was b450 ha, it went into unproductive land with no profitand environmental impact on the region.

2.2. Model setup

To examine the financial and environmental performance of repre-sentative S&B and dairy systems in Southland, the farm-scale simulationmodels Farmax® Pro (Version 6.4.6.07, AgResearch Science Edition forS&B) and Farmax® Dairy Pro (Version 6.4.0.12, AgResearch ScienceEdition for Dairy), both referred to as Farmax (www.farmax.co.nz),were linked with Overseer® (Version 6.0, 2012; www.overseer.co.nz),a nutrient budget model (Wheeler et al., 2008), referred to as Overseer.

Farmax was used to investigate feed supply, key physical indicatorsand financial outputs from the S&B and dairy farms. A key feature ofFarmax is the biological feasibility (i.e. matching feed supply with feeddemand) of farm systems with varying stocking policies, which havebeen determined according to monthly pasture growth rates and theuse of supplemental feed, as described by White et al. (2010) for S&Bsystems and Bryant et al. (2010) for dairy systems. Farmax can eitherbe run in the short-term mode for designing farm management forthe forthcoming season, or in a long-term mode for describing thefarm's average year. The latter was used for the study described here.

Overseer was used to examine some of the environmental impactsand nutrient losses of the farms. Overseer was developed to aid soilnutrient management and budgeting for the main soil nutrients (N,phosphorus, potassium, sulphur, calcium, magnesium and sodium) ap-plicable tomost NZ farming enterprises (Wheeler et al., 2006). The GHGmodel built within Overseer is based on the NZ IntergovernmentalPanel on Climate Change inventory for agricultural methodology(Vogeler et al., 2012), modified and updated to allow for on-farm man-agement strategies (Beukes et al., 2011b).

Information on a farm's physical characteristics, livestock policy andmanagement practices was assembled in Farmax and then exported toparameterise Overseer. On-farm nitrate N leaching and GHG emissionsas the sum of methane, nitrous oxide and carbon dioxide emissions

Fig. 2. Pasture Production Classes (PPC) calculated from the attainable sheep-carrying ca-pacities listed in the extended legends of the LUC worksheets, assuming a utilisation of70%, an increase of 7% under dairying and a N fertilisation response of 10 kg DM/kg N.

Table 1Characteristics, actual and those used for the model setup, of sheep and beef (S&B) and dairy farms in Southland based on Pasture Production Class (PPC), predominant soil order, andtopography. The areas currently under S&B or dairying are also given.Source: Landcare Research; AsureQuality, 2010.

PPC Soil order Area (ha) Area (%) within different slope classes

Actual Model Setup

S&B Dairy S&B Dairy

S&B Dairy 0–7° 8–15° 16–25° 0–7° 0–7° 8–15° 16–25° 0–7°

1 Brown 209 443 100.0 0.0 0.0 100.0 100 0 0 1002 Brown 53,403 42,874 99.6 0.4 0.0 100.0 100 0 0 1002 Pallic 26,240 22,075 100.0 0.0 0.0 100.0 100 0 0 1003 Brown 136,259 30,806 97.7 2.3 0.0 99.7 100 0 0 1003 Pallic 113,301 40,833 98.0 2.0 0.0 99.7 100 0 0 1004 Brown 125,635 12,158 90.1 9.8 0.1 98.6 100 0 0 1004 Pallic 65,897 14,986 92.0 7.8 0.2 99.3 100 0 0 1005 Brown 19,309 761 53.5 44.1 2.4 87.7 55 45 0 –

5 Brown 2847 333 83.1 16.6 0.3 97.8 85 15 0 –

6 Brown 226,587 6325 27.1 47.3 23.8 71.3 25 50 25 –

6 Brown 35,081 2329 50.1 46.8 3.1 76.3 50 50 0 –

308 I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

expressed as CO2 equivalents (CO2-e) were assessed (Wheeler et al.,2008).

Representative farms for each enterprise were used in themodellingexercise. Representative dairy systems were sourced from DairyNZ(2010a), and expert opinion from DairyNZ (D. Dalley, pers. comm.)andAgResearch scientistswhohaveworked in the region (R.Monaghanand D. Stevens, pers. comm.). Data for representative S&B systemsweresourced from Beef and Lamb New Zealand (B + LNZ) (2010), as de-scribed by Vibart et al. (2011).

2.2.1. Farmax and Overseer — dairyNZ dairy farm systems are based on grazing pasture throughout the

year, with varying amounts of conserved and bought in supplements.Dairy enterprises in NZ have been classified into five production sys-tems based on the timing and amount of feed brought onto the milkingplatform, an area for lactating cows exclusively (Hedley and Bird, 2006;Hedley and Kolver, 2006; DairyNZ, 2010a). The most dominant dairyenterprise, which represented 44% of Southland dairy farms in the2007/08 season, is described as a System 3 (Dairybase, 2008). In a Sys-tem 3 dairy farm, 10–20% of the feed is imported to extend lactation,typically as autumn feed.

Dairy cows are typically wintered off the milking platform duringthe dry season, increasing the amount of feed available for lactating an-imals in early spring by minimising pasture utilisation and damage(Beukes et al., 2011a). To account for a support area (area for young live-stock and dry cows), a 230-ha System 3 dairy farm was modelled to in-clude a milking platform area of 190 ha and an adjacent support area of40 ha (Beukes et al., 2011a) (Table 2). The size of the milking platformwas chosen based on the average milking platform size for Southlandduring the 2009/10 season (DairyNZ, 2010b).

Also, to accommodate all dairy livestock categories within the re-gion, replacement heifers (assuming a 21% replacement rate) weresent off-farm on December 16 (114 kg live weight (LW)), managed byneighbouring S&B farmers for 17.5 months and returned pregnant tothe dairy farms on May 31 (445 kg LW). All other heifer and bull calveswere sold weekly during calving. This system (lactating plus dry cowskept on farm)was considered for regional up-scaling, sincemaintainingwintering cows on the farm ensured that cow numbers were kept pro-portional to dairy farm areas and allowed for a more comprehensivecapture of nutrient losses and financial outcomes at a regional level.

The System 3 enterprise was set up on dairy farms with PPCs 1–4(94% of the total area currently under dairy is on LUC classes 1–4;Table 1), resulting in four separate Farmax simulations. Slight adjust-ments were made to account for differences in pasture productionusing industry statistics and scientific expertise for the region (Table 2).The annual pasture production yields for the various PPCs (Fig. 2) wereused with the seasonal pattern of pasture growth obtained from two

Southland dairy farms (Vibart et al., 2012) and scaled to the correspond-ing PPC.

A medium to high level of productivity was assumed for the dairyfarms (kg milk solids (MS)/cow, where MS is defined as milk fat plusprotein yield) to represent slightly above average production and prof-itability (DairyNZ, 2010b). Operating expenses used were the Farmaxdefault 2009/10 values for dairying in the South Island as provided byDairyNZ (2010b). Prices were not adjusted specifically for Southlandor for the systems involved. The milk price was set at NZ$6/kg MS.

Whole-farm feed supply and demand were balanced in the simula-tions. To allow for consistent pasture management as the PPC declinedfrom 1 to 4, dairy cow numbers were scaled accordingly to achieve sim-ilar pasture use across farms. The followingmanagement variableswereset for each farm (further information is provided in Table 2): (a) amat-ing season that started on October 28 and calving that started August03; (b) a body condition score (from 1 to 10, where b3 is consideredemaciated and N7 is considered obese) of 5.0 was set at calving; (c) acow of average genetic merit, namely a Holstein Friesian–Jersey crosscowwith a breedingworth of 88; and (d) a lactation length of 265 days.

After the identification of the predominant soil drainage profiles(i.e. well or poorly drained soils) to inform the Overseer model, sevenOverseer simulations were conducted to represent the PPC × drainagecombination [(PPC 1 × 1 drainage profile) + (PPC 2, 3 and 4 × 2 drain-age profiles)].

Liquid effluent was stored and sprayed on 28% of the total farm area(64 ha). The dairy effluent system included a holding pond,with no sep-aration of solids before effluent entered the holding pond. Liquid efflu-ent was disposed by frequent stirring and spraying, and applicationdepth was set at 12–24 mm, representing a travelling irrigator set atmedium speed. Effluent applications used current best managementpractices, with additional care to ensure a minimal risk of effluent lossthrough runoff or the drainage system. The amount of N fertiliserapplied annually to the effluent block was adjusted to achieve150 kg total N/ha, including theN applied as effluent. Nitrogen fertiliserwas applied as four equal dressings throughout the growing season inthe non-effluent block, and adjusted accordingly for the effluent block.

Supplements were purchased as baleage (wrapped silage) andwerefed to the animals to fill feed deficits during the milking season. Excesspasture growth was cut and conserved as pasture silage and fed to dry

Table 2Size, livestock policies, feed consumption, animal performance and profitability of aSystem 3 dairy farm on Pasture Production Classes (PPCs) 1–4 in Southland.

PPC

1 2 3 4

Effective area (ha) 230 230 230 230Cows, July 1 626 621 504 409Pasture consumed (kg DMa/ha) 11,014 10,930 8725 7037Total feed consumed (kg DM/ha) 13,098 12,997 10,458 8473MSb (kg/ha) 1128 1120 901 733kg DMIc/kg MS 12.0 11.9 11.9 11.9kg LWd/t available DM 81.9 81.9 82.4 82.7Operating profite (NZ$/ha) 2898 2858 1732 825

a Dry matter.b Milk solids.c Dry matter intake.d Live weight.e Farm profit before tax.

Table 3Size, livestock policies, feed consumption, animal performance andprofitability of a class 6sheep and beef (S&B) farm on Pasture Production Classes (PPCs) 2–6 in Southland.

PPC

2 3 4 5 6

Effective area (ha) 450 450 450 450 450Stocking rate (SUa/ha) 17.0 14.2 12.6 9.4 5.7Species ratio (%)b

Sheep 86 86 86 77 77Beef 10 10 10 20 20Deer 4 4 4 3 3

Animal production (kg/ha)Meat 252.2 211.3 186.3 137.6 83.4Wool 72.9 60.4 53.2 35.4 21.3Total 325.1 271.7 239.4 173.0 104.7

Animal reproductionEwe efficiency indexc (%) 60.8 60.8 60.8 60.8 60.8Cow efficiency indexd (%) – – – 37.8 37.8

Financial indicators (NZ$/ha)Gross margine 715 599 517 398 225Operating profitd 434 315 236 147 −26

a SU, stock units.b The species ratio is based on DM consumption.c Ewe efficiency index = total standardised lamb weaning weight (at 90 days, in kg)

per kg ewe mated, expressed as a percentage.d Cow efficiency index = total standardised calf weaning weight (at 200 days, in kg)

per kg cow mated, expressed as a percentage.e Farm profit before tax.

309I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

cows. Additionally, a winter fodder crop (kale, Brassica oleracea L.) wassown on 10% of the non-effluent area (17 ha) and was assumed to sup-ply 13 t DM/ha. The fodder cropwas grazed fromMay to late Septemberby lactating and, predominantly, by non-lactating cows.

Dairy farm physical characteristics and livestock policies wereexported from Farmax to parameterise Overseer, with the followingoverall assumptions: (a) the mean annual rainfall was set at 1000 mm;(b) all soil fertility test values were typical values suggested by Overseer,including Olsen P (30; Overseer default); (c) for well-drained brownsoils, the occurrence of pugging or treading damage was considered tobe rare,whereas for poorly drained pallic soils, the occurrence of puggingor treading damage was considered to occur during winter and aftermost rainfall events; and (d) maintenance phosphorus and sulphurwere applied in November as soluble fertiliser; the amount of fertiliserapplied was tailored to individual nutrient maintenance requirementsas suggested by Overseer.

2.2.2. Farmax and Overseer — sheep and beefA class 6 (South Island finishing–breeding (Beef and Lamb New

Zealand, 2010)) farm was chosen as a representative S&B farming sys-tem in Southland. Briefly, class 6 farms comprise an extensive type offinishing operation, with a carrying capacity that typically ranges from6 to 11 stock units (SU)/ha on non-irrigated farms. The B + LNZ class6 farm is the predominant S&B farm class in the South Island (Beefand Lamb New Zealand, 2010). The S&B farms were modelled on PPCs2–6, resulting infive separate Farmax simulations. All farmswere equal-ly sized (effective area: 450 ha), but with varying percentages of land ineach slope class (Table 1). All feed was produced on-farm, and animal

productivity was assumed to represent average production and profitsimilar to that from Beef and Lamb New Zealand (2010) (Table 3). Aswith dairy systems, the farms were assumed to be in a steady stateboth in terms of opening and closing numbers of livestock and corre-sponding body weights. Also, it was assumed that there was no differ-ence in annual pasture production from the well-drained and poorlydrained soils under S&B farming.

The annual pasture production yields for the various PPCs (Fig. 2)were used with the seasonal pattern of pasture growth obtained fromVibart et al. (2011) for a flat to rolling topography (slopes b 15°) inSouthland, and scaled to the corresponding PPCs. Also, as for the dairyfarms, a non-irrigated South Island ryegrass/white clover-type pasturewas used in Farmax for PPCs 2–4, whereas a browntop (Agrostiscapillaris L.) type pasture was set for PPCs 5 and 6.

Livestock policies of the S&B farming systems on PPCs 2–4comprised an animal species ratio of 86:10:4 (sheep:cattle:deer;based on total DM intake) including breeding Romney ewes, tradingAngus × Hereford bull calves and contracted dairy heifers, and reddeer (D. Stevens, AgResearch, pers. comm.) (Table 3). Bulls were pur-chased in April and sold the followingApril/May. Similarly, the livestockpolicies of the S&B farming systems on PPCs 5 and 6 comprised an ani-mal species ratio of 77:20:3, with a larger beef component from replac-ing purchased bulls with breeding Angus × Hereford cows. As with thedairy farms, livestock numbers were scaled according to PPC to achievesimilar pasture use across farms.

Reproductive efficiencies were considered equivalent for all PPCs.For all S&B PPCs, ewe pregnancy, lambing and weaning percentageswere set at 166%, 130% and 126%. The corresponding cow pregnancy,calving and weaning percentages for PPCs 5 and 6 were set at 92%,85% and 84%. Overall ewe efficiency (%), calculated from the totalweight of weaned lambs (at 90 days old) and the total weight ofmated ewes, was 61%. Similarly, overall cow efficiency, calculatedfrom the total weight of weaned calves (at 200 days old) and the totalweight of mated cows, was 38% (Table 3). Ewe, cow and hind replace-mentswere set at 28%, 18% and 18%,which iswithin the ranges reportedby the Farm Technical Manual (Fleming, 2003).

In addition to supplementary hay from excess spring pasturegrowth, a winter fodder crop (swedes, Brassica napus L.) with an as-sumed DM yield of 12.0 t DM/ha was grown on 28 ha (PPCs 2–4) oron 11 ha (PPCs 5–6), and grazed from May to early October.

As with dairy farms, once the predominant soil drainage profileswere identified, these were used to inform the Overseer model, tenOverseer simulationswere conducted to represent the S&BPPC× drain-age combinations (PPC 2 to 6 × 2 drainage profiles). Physical character-istics and livestock policies for each of the S&B farms were exportedfrom Farmax to parameterise Overseer, with the same assumptions asused for dairy farms, except for the soil fertility tests, where typicalvalues for S&B farming systems were used.

Fig. 4. Associated farm-scale nitrate leaching (left) and GHG emissions (right) simulated by Overseer for hypothetical dairy farms in Southland as a function of Pasture Production Class(PPC) and for well (open symbols) and poorly (closed symbols) drained soils.

Fig. 3. Farm-scale profit (closed symbols) andmilk solid (MS) production (open symbols)simulated by Farmax for hypothetical dairy farms in Southland as a function of PastureProduction Class (PPC).

310 I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

For each PPC, farm pre-tax profit was calculated as the difference be-tween total farm gross revenue (from Farmax) and total farm expenses(from Farmax and the B+ LNZ Economic Service's Sheep and Beef FarmSurvey Forecast 2010–11), as described in Vibart et al. (2011) (Table 3).

2.3. Aggregating model outputs — farm-scale and regional up-scaling

2.3.1. Farm-scale aggregationThe farm profit before tax for the individual farms (FPi) for the two

different scenarios was calculated as:

FPi ¼X

FPD � AD þ FPS&B � AS&Bð Þ ð1Þ

where FPD and FPS&B are the profits before tax on a per-hectare basis, fordairying and S&B, respectively; and AD and AS&B represent the farmareas under dairying and S&B, respectively (in ha). The profit per hect-are for either farm system was calculated using regression analyses ofproduction in each PPC.

N leaching and GHG emissions for the individual farms were calcu-lated in the same way as the profit, but separate regression analyseswere undertaken for well-drained and poorly drained soils.

2.3.2. Regional up-scalingFor estimating the effect of land conversion on the profit and envi-

ronmental impacts at the regional scale, two different approacheswere used. For both approaches, the area in dairying was increasedstepwise and the outcomes for the different proportions of dairying ver-sus S&B were compared. Approach 1 linked outputs from farm systemmodelling with land resource information as described above. Dairyfarming was first limited to land within PPCs 1 and 2, then increasedstepwise to PPCs 1–3; to PPCs 1–4, with PPC 4 limited to well-drainedsoils; and then extended to poorly drained soils. As before, a minimumof 230 ha in PPCs 1–4 was required for dairying. This resulted in an in-crease in land area under dairying from 4% to 41%.

Approach 2 was based on simple regional “representative” meanvalues that did not consider the characteristics of the underlying landresources. For profits, we used annual values for Southland of $1224/ha for dairying, $330/ha for S&B on PPCs 2–4, and $174/ha for S&B onPPCs 5 and 6, as reported for intensive S&B and hill country S&B(Ministry of Agriculture and Forestry, 2010). The area that could be po-tentially converted into dairyingwas limited to 60%, as dairyingwas re-stricted to land in PPCs 1–4. Note that this does not account for thespatial distribution of the PPCs or current farm boundaries. For Nleaching, average annual losses for Southland of 26 kgN/ha under dairy-ing and 13 kg N/ha for S&B were used (Monaghan et al., 2010).

3. Results and discussion

3.1. Farm system modelling

3.1.1. DairyingTotalMSproductionwas greater on farmswith higher productive po-

tential, enabling a higher carrying capacity (Table 2). As MS sales com-prised 95% of the farm revenue, farm profit also increased with lowerPPC, and ranged from $825 to nearly $2900/ha from PPC 4 to 1 (Fig. 3).These results are consistent with findings from van Bysterveldt et al.(2006), who found a positive correlation between dairy farm profitabil-ity and pasture consumed on a per hectare basis. These results, alongwith the ones reported for S&B, are also in line with a comparison of po-tential annual profitability for land use in southern NZ of about $450/hafor S&B, stocked at 11.4 SU/ha (equivalent to PPC4 in ourmodelling) and$1000/ha for dairy (Copland and Stevens, 2012). For the regional up-scaling (Section 2.2.2) a linear regressionwas fitted to the farm profit re-sults from PPCs 2–4, with the profit from PPC 1 being equal to that of PPC2. The resulting regression equations are given in Section 2.2.1.

Since the environmental impacts of farming systems are dependenton the stocking rate as well as soil drainage characteristics, Overseersimulation results are presented for the poorly drained pallic soils andwell-drained brown soils in Southland (Fig. 4). Modelled nitrateleaching ranged from 21 to 44 kg N/ha, and was higher under thefarms in lower PPCs, mainly because these allow higher stocking rates.These values are within the range reported for dairy systems in NZ; forinstance, Monaghan et al. (2008) reported a value of 30 kg N/ha/yearfrom a dairy farm in Southland under a poorly drained silt loam, andDi and Cameron (2002) reported N leaching ranges from dairying of15–115 kg/ha/year. Regardless of PPC, leaching was lower from thepoorly drained soils than from the well-drained soils. Annual GHG

Fig. 6. Farm-scale profit (closed symbols) and production of meat and wool (open sym-bols) simulated by Farmax for hypothetical S&B farms inSouthland as a function of PastureProduction Class (PPC).

0

4000

8000

12000

16000G

HG

[kg

CO

2-e/

ha

/yr]

CO2N2OCH4

0

4000

8000

12000

16000

1 2 3 4 2 3 4

GH

G [

kg C

O2-

e/h

a /y

r]

CO2N2OCH4

PPC PPC

Fig. 5. Farm-scale GHG emissions simulated by Overseer for hypothetical dairy farms in Southland as a function of Pasture Production Class (PPC), and for well (left) and poorly (right)drained soils.

311I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

emissions were also higher for lower PPC classes, due to the higherstocking rate and associated livestock dry matter (DM) intake as theforemost driver of methane emissions.

In contrast to leaching, GHG emissions were higher in the poorlydrained soils compared with the well-drained soils. Denitrification pro-cesses, leading to higher nitrous oxide (N2O) emissions (Fig. 5), areknown to be prominent in wet, poorly drained soils (De Klein et al.,2003). Nitrous oxide emissions comprised, on average, 22% of totalGHG emissions with a well-drained soil and 35% with poorly drainedsoil. Again, for the regional up-scaling, a linear regression was fitted tothe N leaching and GHG emission results from PPCs 2–4.

3.1.2. Sheep and beefSimulations from the modelled S&B farms show higher farm

profit before tax under lower PPCs (Fig. 6), again mainly due to thehigher pasture production (Fig. 2), allowing for a higher stockingrate (ranging from 5.7 SU/ha in PPC 6 to 17.1 in PPC2), and therebyincreasing meat and wool production per hectare (Fig. 6). Revenuefor all PPCs was primarily from sheep sales (62–70% of revenue),whereas revenue from wool sales contributed 14%, beef sales con-tributed 4–14%, deer sales contributed 6–7% and contract grazing con-tributed 5%. Expenditure was slightly higher for lower PPCs due toincreased costs for expenses like animal health and feed crops. On theless productive PPCs, the model results showed lower profits or even anet loss.

The range of pre-tax profit simulated here agreeswell with the profitreported by Beef and Lamb New Zealand (2010) for class 6 farms of$131/ha. It should, however, be noted that these financial results forS&B were assembled from total gross revenue outputs from FarmaxPro with long-term pricing assumptions and expense items mostly

from the BLNZ Economic Service Forecast (Beef and Lamb NewZealand, 2010), and corrections for inflation or capital expenditure forconversion to dairy.

Farm-scale nitrate leaching as simulated by Overseer was lower onpoorly drained soils compared with well-drained soils and, due to alower stocking rate, lower on higher PPC soils (Fig. 7). Leaching for theS&B farms ranged from 8 kg/ha/year to 17 kg/ha/year, which is in ac-cordance with reported values from sheep-grazed pastures in NZ of7–20 kg N/ha/year (Heng et al., 1991; Ruz Jerez et al., 1995; Magesanet al., 1996).

Farming on lower PPC soils was also associated with increasedmethane (CH4) and nitrous oxide (N2O) emissions at a farm level.Total annual GHG emissions ranged from 2179 kg CO2-e/ha to5825 kg CO2-e/ha for well-drained soils and from 2281 kg CO2-e/ha to7061 kg CO2-e/ha for poorly drained soils. The N2O emissions frompoorly drained soils contributed 40% of the total farm-scale emissions(Fig. 8).

3.2. Farm-scale modelling

3.2.1. Individual farmsModel results from Farmax andOverseerwere related to actual farms

obtained from the Agribase database in Southland (see Section 2.1), forScenario 1 with the current mix of dairy and S&B, and for Scenario 2with the farms that could potentially convert into dairying based onPPCs. To account for the various land units with different PPC's anddrainage characteristics on individual farms linear regression analysiswas performed on the model outputs for farm profit, N leaching andGHG emissions.

0

2000

4000

6000

8000

GH

G [

kg C

O2-

e/h

a /y

r]

0

2000

4000

6000

8000

2 3 4 5 6 2 3 4 5 6

GH

G [

kg C

O2-

e/h

a /y

r]

PPC PPC

CO2

N2O

CH4

CO2

N2O

CH4

Fig. 8. Farm-scale GHG emissions simulated by Overseer for hypothetical S&B farms in Southland on different Pasture Production Classes (PPC), for well (left) and poorly (right) drainedsoils.

Fig. 7.N leaching (left) and farm-scale GHG emissions (right) simulated by Overseer for hypothetical S&B farms in Southland as a function of Pasture Production Class (PPC) for well andpoorly drained soils.

312 I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

The profit per hectare for an individual farm under either farm sys-tem is given, according to Figs. 3 and 6, by the following relationships:

FPD ¼ −1016PPC2−4 þ 4855 ð2Þ

FPS&B ¼ −109PPC2−6 þ 656 ð3Þ

where PPC2−4 is the area-weighted average PPC for any individual farmunder dairying with PPC1 set to PPC2, and PPC2−6 is the area-weightedaverage PPC for S&B. As for Scenario 2, it was assumed that all potential-ly convertible dairy farms that had an area of≥450 ha in PPCs 5–6 keptthis land under S&B, inwhich case, the additional profit from S&Bwouldbe:

FPS&B ¼ −109PPC5−6 þ 656: ð4Þ

The profit for each individual farm was then calculated according toEq. (1).

For N leaching and GHG emissions, separate regression analyseswere undertaken for well and poorly drained soils. On a per hectarebasis, N leaching for dairying under System 3 under well-drained(NDW) and poorly drained (NDP) soils was calculated according toFig. 4 as:

NDW ¼ −6:5PPCw2−4 þ 51 and NDP ¼ −5PPCp2−4 þ 41: ð5Þ

N leaching for S&B on well-drained (NS&BW) and poorly drained(NS&BP) soils was calculated according to Fig. 7:

NS&BW ¼ −2:2LUCw2−6 þ 21 and NS&BP ¼ −1:45LUCp2−6 þ 16: ð6Þ

GHG emissions on a per hectare basis for dairying under System 3under well (GHGDW) and poorly (GHGDP) drained soils were calculatedaccording to Fig. 4 as:

GHGDW ¼ −1568PPCw2−4 þ 13298 and GHGDP ¼ −1795PPCp2−4 þ 15;682:

ð7Þ

GHG emissions for S&B on well-drained (GHGS&BW) and poorlydrained (GHGS&BP) soils were calculated according to Fig. 7 as:

GHGS&BW ¼ −901LUCw2−6 þ 7709 and GHGS&BP ¼ −1308LUCp2−6 þ 9904:

ð8Þ

These regression equations were used to estimate individual farm-scale profits and environmental impacts based on the characteristicsand coverage (in hectares) of each PPC × drainage class on each farmfor the two scenarios investigated. Table 4 shows the number of farmsand the total area under dairy for the two scenarios, alongwith averageherd size, stocking rate, required replacement numbers of heifers andthe percentage of those that can be locally carried by S&B farms.

According to Agribase, Southland has about 750 dairy farms, slightlylower than the 872 reported for 2010/11 by DairyNZ (2011). The aver-age farm size and herd size (212 ha and 555 cows) are comparable tothe DairyNZ figures. Under Scenario 2, the average herd and farm sizewould increase to 350 ha and 750 cows, but with a slightly lower aver-age stocking rate, due to land with higher PPCs being converted todairying.

To maximise profit and minimise pasture use and damage, South-land dairy enterprises often rear young stock and send dry cows to win-ter off the milking platform. A 3-year study in Southland (Dalley et al.,2008) showed that many farmers opt for a dairy support block toachieve control of stock management and costs. But to account for alldairy livestock categories, in our modelling exercise, we assumed thatdairy farms have an adjacent support block for winter-feeding drycows, based partly on forage crops (most commonly brassicas such askale or swedes). This ensures that the environmental impacts of thewintering-off are accounted for at the farm scale. Heifer replacementrearing in our modelling is, however, carried by the S&B farms. Thismeans that under Scenario 1, almost all of the required heifers can beraised by the S&B farms in Southland. Under Scenario 2, only a smallportion of the heifers are accounted for in the current S&B farmbudgets;so most of the heifers would need to be raised outside the region(Table 5).

The distribution of current (Scenario 1) and potential farming(Scenario 2) enterprises shows that the new dairy farms that wouldbe created through conversion from S&B farming are mainly in thelower to middle range of profit (Fig. 9), with the mean annual dairyfarm profit before tax across the region decreasing from $2039 to$1858/ha, due to the average PPC of land available for conversionbeing higher than the current average PPC of dairy farms. S&B farmsthat could potentially convert are more evenly spread with regard toprofit, with the mean profit remaining almost the same with $302and $312/ha for Scenarios 1 and 2. Note that this calculation is notarea-averaged and the results are partly due to the restriction thatfarms needed to have at least 230 ha in PPCs 1–4 to be able to convert.This meant that a number of small farms that have a high profitcould not convert. It should also be noted that the distribution shownhere is only due to the difference in the distribution of PPCs within theindividual farms. Individual farm performance and management,whichwere not accounted for in the current simulation study,would re-sult in different profits for any individual property. Possible changes infarm boundaries to optimise the distribution of land were also notconsidered.

Mean annual N leaching estimates from a dairy farm equated to30.6 kg/ha for Scenario 1 and 29.3 kg/ha for Scenario 2. Mean annualGHG emissions would be 9808 kg CO2-e/ha/year for Scenario 1 and9393 kg CO2-e/ha/year after implementing dairy over all potentiallyconvertible land (Scenario 2). For S&B, the mean annual N leachingwould equate to 12.9 kg/ha for Scenario 1, and 13 kg/ha for Scenario2. Mean annual S&B GHG emissions are 5156 kg CO2-e/ha/year for Sce-nario 1 and 5256 kg CO2-e/ha/year for Scenario 2. These increased

Table 4Dairy farming under land use scenarios for Southland, with Scenario 1 being the currentmix of dairying and sheep and beef (S&B), and Scenario 2 being the potential dairycoverage according to land resources.

Scenario 1 Scenario 2

Area under dairying (%) 15 46No. of farms 754 1421Average farm size (ha) 220 346Average herd size 506 743Average stocking rate 2.32 2.15Heifers required 80,172 221,862Heifers carried by S&B in current budgets (%) 95 11

Table 5Regional average profit, N leaching and GHG emissions calculated from the model outputs of Overseer and Farmax under two different scenarios with (1) the current mix of dairy andsheep and beef (S&B) farms and (2) the potential mix according to land resources.

Scenario Area in dairy (thousand ha) Area in S&B (thousand ha) Profit ($/ha) N leaching (kg/ha) GHG emissions (kg CO2-e/ha)

1 176 901 462 13.1 46482 492 585 815 17.5 5750

313I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

values for both N leaching and GHG emissions are, as discussed above,due to the restriction for conversion of an area of at least 220 ha inPPCs 1–4.

The new dairy farms resulting from conversion would mainly be inthe N leaching range of 25–31 kg/ha/year (Fig. 10) and have GHG emis-sions of 7000–10,500 kg CO2-e/ha/year (Fig. 11). This analysis did notconsider mitigation options as part of the study, but there are regionsin the country where limits have or are being set. For example, in theTaupo catchment of NZ, an N cap is already in operation (EnvironmentWaikato, 2008). In the Manawatu–Wanganui region N leaching limitsare proposed dependent on LUC classes (e.g. a maximum N leachingof 30 kg/ha on LUC 1, and of 15 kg/ha for LUC 6) (Horizons RegionalCouncil, 2012). One strength of the new approach is that mitigation op-tions – ranging from restricted grazing to the use of nitrification inhibi-tors, which have been shown to reduce N leaching (Monaghan et al.,2009; Christensen et al., 2010; Vogeler et al., 2013) – could be easily in-corporated into the framework.

S&B farms that would potentially convert into dairying would comemainly from the middle range of GHG emissions but as dairy farmswould be at the lower to middle range (Fig. 11). While there are cur-rently no limits on GHG emissions for agriculture in NZ, understandinghow land use changes affect these and how to quantify these changes isimportant for the future.

3.2.2. Regional up-scalingThe profit at a regional scale was estimated by summing up the indi-

vidual farm profits of all 4150 individual farms, calculated by Eq. (1),and dividing this by the total area in dairying plus S&B farming. Nleaching and GHG emissions were up-scaled in a similar way to profit.Conversion of more farms into dairying (Scenario 2) would result in

an increased profit, but also higher N leaching and GHG emissions inthe region compared with the current situation (Scenario 1; Table 5).Mitigation technologies could reduce these, but were not included inthe current modelling.

It must be noted that this up-scaling is based on potential dairyconversion based on land resources as defined by the productivepotential of the landscapes found in Southland and with the currentfarm boundaries. Further refinement of the up-scaling approachcould be achieved by taking the direct influence of soil propertiesand pasture production into account, as well as environmental impactsand variations in climate within the region. The actual level of conver-sion could also differ substantially if additional or different farming sce-narios were tested. Features of the farms, such as the physical shape ofthe farmand the degree towhich the landscape is dissectedwith streamor roads, and social drivers (e.g. farmers' preferences) were also notaccounted for. For example, the dairy lifestyle does not suit every farm-er, high capital cost associated with conversion, and the more volatilelong-term returns from dairy than from S&B (Ministry of Agricultureand Forestry, 2010) could discourage some from dairy conversions.Also, no timeframe on the conversion can be indicated. For such pur-poses, the use of models that include social drivers, such as MultiAgent Simulation models, would be needed (Gaube et al., 2009).

As such, the approach only provides an indication of potential ef-fects, and the actual level of conversion could differ substantially fromthe single scenario that has been presumed here. This study, however,demonstrates the usefulness of the spatially referenced approach foranalysing changes in land use on a regional scale.

Finally, the new regional up-scaling approach (Approach 1) wascompared with an approach based on regional averages (Approach 2).According to Approach 1, the percentage of the region in dairying

Fig. 9.Distribution of farm profit as predicted by Farmax for hypothetical dairy (left) and S&B farms (right) based on the current (Scenario 1) and potential (Scenario 2) number of farms.

Fig. 10.Distribution of N leaching as predicted byOverseer for hypothetical dairy (left) and S&B farms (right) based on the current (Scenario 1) and potential (Scenario 2) number of farms.

314 I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

increases from4% to 40%, by stepwise conversion of less productive landinto dairying. The remainder stays under S&B. Regional profit increaseswith an increased proportion under dairying; however, due to lowerpotential productive land coming into dairying, the increase is not line-ar. Nitrate leaching increases linearly with an increasing proportion ofland in dairying (Fig. 12). Based on Approach 2, regional N leaching in-creases linearly from 15 kg N/ha/year to 20 kg N/ha/year, with the areain dairying increasing from 10% to 60%. Similarly, the regional profit in-creases linearly from about $400/ha/year to $800/ha/year. Approach 1consistently estimates a higher profit than Approach 2, e.g. with 20%of the area under dairying, the regional profit under Approach 1 is 23%higher. The differences between the two approaches are moredominant with lower, but more productive area in dairying, due to thehigh profit for dairying under high productive PPC's with nearly NZ$2900/ha (Fig. 3), compared to the average farm profit of $1224/hasued for Approach 2. The differences between regional profits becomesmaller as less productive land is converted into dairying. Nitrateleaching, on the other hand, is about 10% lower when using Approach1 compared to Approach 2. This is because Approach 2 does not accountfor the effects of land resources on environmental or economic impacts;rather, it treats the entire area as uniform. Note also, that under Ap-proach 1 only about 40% of the total area can be converted into dairying,as the remaining higher productive land parcels are too small to be con-verted. Under Approach 2, 60% of the area can be converted, as it doesnot account for the spatial distribution of the land resources or currentfarm boundaries.

Thus the more detailed up-scaling approach, based on outputs fromfarm system models linked with spatial resource information, givesmore positive – and probably realistic – outcomes for dairy conversion

in the Southland region, both economically and environmentally, com-pared to up-scaling based simply on using regional averages.

4. Conclusions

Farm-scale modelling was linked with spatially based land and pas-ture resource information to estimate the regional effects of dairyconversion on economic and environmental impacts. The Southlandregion of NZ was used as a case study. This up-scaling approach,which accounts for land resources, indicates higher potential regionalprofit and lower N leaching losses from dairy conversion, comparedwith up-scaling representative regional farm systems, that does not ac-count for the influence of the underlying resources. Based on PPCs de-rived from estimates of sheep-carrying capacity from the extendedlegend of the LUC worksheets, there is the potential for 45% of theSouthland region to be in dairying in the future.While such a conversionwould increase the regional profit, the environmental impacts wouldalso be substantial, with increases in N leaching of nearly 35% and in-creases of nearly 25% for GHG emissions.

While for the current modelling study, based on outputs from farmsystem models linked with spatial resource information, only onefarmmanagement systemwas considered, this could easily be extendedto e.g. include different systems, different levels of farming performanceand the use of mitigation technologies.

Conflict of interest

I'm not aware of any conflict of interest with the current manuscripton “Modelling Farm Systems — Scaling From Farm to Region” by Iris

Fig. 11. Distribution of GHG as predicted by Overseer for hypothetical dairy (left) and S&B farms (right) based on the current (Scenario 1) and potential (Scenario 2) number of farms.

Fig. 12. Average N leaching and profit for the Southland case study, dependent on the area in dairying and S&B, as obtained by linking farmmodel outputs with spatially based land re-source information (Approach 1), and by using reported regional mean values for dairying and S&B, and increasing the area in dairying (Approach 2).

315I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

Vogeler, Ronaldo Vibart, Alec Mackay, Samuel Dennis, Vicki Burggraaf,and Josef Beautrais.

Acknowledgements

This work was conducted under the Rural Futures Programme,which was initiated with FRST (PROJ-13807-PPS-AGR) funding andhas since been transferred to AgResearch's own core funding. The au-thors would like to thank Rogerio Cichota and Jane Chrystal for helpfuldiscussions.

References

AsureQuality. Agribase. Database of NZ farms. www.asurequality.com/geospatial-services/agribase.cfm, 2010. [accessed 31 July 2013].

Beef and Lamb New Zealand. Economic service survey. http://www.beeflambnz.com,2010. [accessed 31 July 2013].

Berntsen J, Petersen BM, Jacobsen BH, Olesen JE, Hutchings NJ. Evaluating nitrogen taxa-tion scenarios using the dynamic whole farm simulation model FASSET. Agr Syst2003;76:817–39.

Beukes PC, Palliser CC, Macdonald KA, Lancaster JAS, Levy G, Thorrold BS, et al. Evaluationof a whole-farmmodel for pasture-based dairy systems. J Dairy Sci 2008;91:2353–60.

Beukes PC, Gregorini P, Romera AJ, Dalley DE. The profitability and risk of dairy cow win-tering strategies in the Southland region of New Zealand. Agr Syst 2011a;104:541–50.

Beukes PC, Gregorini P, Romera AJ. Estimating greenhouse gas emissions from NewZealand dairy systems using amechanistic whole farmmodel and inventorymethod-ology. Anim Feed Sci Technol 2011b;166–167:708–20.

Bryant JR, Ogle G, Marshall PR, Glassey CB, Lancaster JAS, Garca SC, et al. Description andevaluation of the Farmax Dairy Pro decision support model. N Z J Agric Res 2010;53(1):13–28.

Christensen CL, Hanly JA, Hedley MJ, Horne DJ. Using duration-controlled grazing to re-duce nitrate-N leaching from dairy farms. In: Currie LD, Christensen CL, editors.Farming's future: minimising footprints and maximising margins. Occasional reportno. 23Palmerston North, New Zealand: Fertilizer and Lime Research Centre, MasseyUniversity; 2010. p. 46–52.

Copland RJ, Stevens DR. The changing face of southern New Zealand farming: opportuni-ties of land use change. Proceedings of the New Zealand Grassland Association, vol.74. Gore: New Zealand Grassland Association Inc.; 2012. p. 1–6.

Dairybase. About Dairybase. http://www.dairybase.co.nz/page/pageid/2145841076/About_DairyBase, 2008. [accessed 31 July 2013].

DairyNZ. Pasture and nutrients. Facts and figures for New Zealand dairy farmers. Version1 http://www.dairynz.co.nz/Publications/FactsFigures/2010/, 2010a. [accessed 31July 2013].

DairyNZ. New Zealand dairy statistics 2009–10. http://www.dairynz.co.nz/file/fileid/34215, 2010b. [accessed 31 July 2013].

DairyNZ. New Zealand dairy statistics, 2010–11. http://www.dairynz.co.nz/file/fileid/39959, 2011. [accessed 31 July 2013].

Dalley DE, Wilson DR, Edwards G, Judson G. Getting the most from your dairy supportland — tips for allocating winter forages. http://www.side.org.nz/IM_Custom/ContentStore/Assets/9/7/d9b068fcb54f0d4ec903b4e9a125b896/Getting%20the%20best%20from%20your%20support%20land.pdf, 2008. [accessed 31 July 2013].

De Klein C, Barton L, Sherlock RR, Li Z, Littlejohn RP. Estimating a nitrous oxide emissionfactor for animal urine from someNewZealand pastoral soils. Aust J Soil Res 2003;41:381–99.

Di HJ, Cameron KC. Nitrate leaching and pasture production from different nitrogensources on a shallow stony soil under flood-irrigated dairy pasture. Aust J Soil Res2002;40:317–34.

Ekanayake J, Davie T. The SWAT model applied to simulating nitrogen fluxes in theMotueka River catchment. Prepared for the Stakeholders of the Motueka IntegratedCatchmentManagement Programme. Landcare ICM report no. 2004–2005/04Lincoln,New Zealand: Landcare Research; 2004.

Elliott S, McBride G, Shankar U, Semandi-Davies A, Quinn J,Wheeler D, et al. CLUES SpatialDSS: from farm-scale leaching models to regional decision support. In: Sanches-MarreM, Bejar J, Comas J, Rizzoli A, Guariso G, editors. Proceedings of the Internation-al Modelling Congress on Environmental Modelling and Software Integrating Sci-ences and Information Technology for Environmental Assessment and DecisionMaking 4th Biennial Meeting. International Environmental Modelling and SoftwareSociety (iEMS); 2008.

Elliott, S., Semadeni-Davies, A., Shankar, U., 2011. CLUES Catchment Modelling— Lessonsfrom Recent Applications. In: Currie, L.D.C., C.L. (eds). Adding to the Knowledge Basefor the Nutrient Manager. (Ed.), FLRC. Occasional Report No. 24, pp. Fertilizer andLime Research Centre, Massey University, Palmerston North, New Zealand., MasseyUniversity, Palmerston North.

Environment Waikato. The condition of rural water and soil in the Waikato region. Risksand opportunities. Environment Waikato; 2008.

Fleming P. Farm technical manual. Lincoln: FarmManagement Group, Lincoln University;2003.

Gaube V, Kaiser C, Wildenberg M, Adensam H, Fleissner P, Kobler J, et al. Combiningagent-based and stock-flow modelling approaches in a participative analysis of theintegrated land system in Reichraming, Austria. Landsc Ecol 2009;24:1149–65.

Green SR, Vogeler I, Clothier BE, Mills TM, Van Den Dijssel C. Modelling water uptake by amature apple tree. Aust J Soil Res 2003;41:365–80. [http://www.waikatoregion.govt.nz/PageFiles/10480/Soil%20and%20water%20issues.pdf].

Hamill KD, McBride GB. River water quality trends and increased dairying in Southland,New Zealand. N Z J Mar Freshw Res 2003;37:323–32.

Hedley P, Bird P. Achieving high profit from your farm system. In: Brookes IM, editor. Pro-ceedings of the 4th Dairy3 Conference. Hamilton, 3–5 April 2006. Hamilton: Centrefor Professional Development and Conferences, Massey University; 2006. p. 167–74.

Hedley P, Kolver E. Achieving high performance from a range of farm systems in South-land. SIDE Proceedings : The Positive SIDE. South Island Dairy Event, Invercargill;2006. p. 141–67.

Heng LK, White RE, Bolan NS, Scotter DR. Leaching losses of major nutrients from a mole-drained soil under pasture. N Z J Agric Res 1991;34:325–34.

Horizons Regional Council. One plan. Chapter 13: discharges to land andwater, vol. 3. ; 2012.[http://www.horizons.govt.nz/assets/publications/about-us-publications/one-plan-publications-and-reports/proposed-one-plan/Chapter13_dischargestolandandwater.pdf (accessed 31 July 2013)].

Johnson IR, Chapman DF, Snow VO, Eckard RJ, Parsons AJ, Lambert MG, et al. DairyModand EcoMod: biophysical pasture-simulation models for Australia and NewZealand. Aust J Exp Agric 2008;48:621–31.

Landcare Research. Land and Environments of New Zealand. www.landcareresearch.co.nz/resources/maps-satellites/lenz, 2013. [accessed 31 July 2013].

Landcare Research. New Zealand Land Resource Inventory database. http://www.landcareresearch.co.nz/resources/data/lris. [accessed 31 July 2013].

Lynn IH, Manderson AK, Page MJ, Harmsworth GR, Eyles GO, Douglas GB, et al. Land usecapability survey handbook— a New Zealand handbook for the classification of land.Lincoln: Agresearch Ltd Hamilton, Landcare Research NZ Ltd; 2009 [Institute ofGeological and Nuclear Science Ltd Lower Hutt, New Zealand].

Magesan GN, White RE, Scotter DR. Nitrate leaching from a drained, sheep-grazed pas-ture. I. Experimental results and environmental implications. Aust J Soil Res 1996;34:55–67.

Ministry of Agriculture and Forestry. Farm monitoring reports. http://maxa.maf.govt.nz/mafnet/rural-nz/statistics-and-forecasts/farm-monitoring/, 2010. [accessed 31 July2013].

Monaghan RM, Paton RJ, Smith LC, Drewry JJ, Littlejohn RP. The impacts of nitrogenfertiliser and increased stocking rate on pasture yield, soil physical condition and nu-trient losses in drainage from cattle grazed pasture. N Z J Agric Res 2005;48:227–40.

Monaghan RM, Wilcock RJ, Smith LC, Tikkisetty B, Thorrold BS, Costall D. Linkages be-tween land management activities and water quality in an intensively farmed catch-ment in southern New Zealand. Agr Ecosyst Environ 2007;118:211–22.

Monaghan RM, de Klein CAM, Muirhead RW. Prioritisation of farm scale remediation ef-forts for reducing losses of nutrients and faecal indicator organisms to waterways: acase study of New Zealand dairy farming. J Environ Manage 2008;87:609–22.

Monaghan RM, Smith LC, Ledgard SF. The effectiveness of a granular formulation ofdicyandiamide (DCD) in limiting nitrate leaching from a grazed dairy pasture. N Z JAgric Res 2009;52:145–59.

Monaghan RM, Semadeni-Davies A, Muirhead RW, Elliott S, Shankar U. Land use andland management risks to water quality in Southland. AgResearch report preparedfor Environment Southland; 2010. [www.es.govt.nz/media/21906/agresearch-monaghan-land-use-risks.pdf (accessed 31 July 2013)].

Neitsch SL, Arnold JG, Kiniry JR, Srinivasan R, Williams JR. Soil and water assessment toolinput/output file documentation: version 2009. Texas Water Resources InstituteTechnical Report 365. College Station (Texas): Texas A&M University System; 2011.

NPS. National policy statement for freshwater management. http://www.mfe.govt.nz/publications/rma/nps-freshwater-management-2011/docs/nps-freshwater-mgnt-2011.pdf, 2011. [accessed 12 May 2011].

Rutherford K. Modelling the effects of land use on nutrients entering the Tukituki River,Hawkes Bay; 2012.

Ruz Jerez BE, White RE, Ball PR. A comparison of nitrate leaching under clover-based pas-tures and nitrogen fertilized grass grazed by sheep. J Agric Sci 1995;125:361–9.

Samarasinghe O, Daigneault A, Greenhalgh S, Sinclair R. Modelling economic impacts ofnutrient reduction policies in the Hurunui Catchment, Canterbury. New ZealandAssociation of Economists Annual Meeting, Wellington, New Zealand 29 June–1July; 2011.

Smith AP. Long term pasture growth patterns for Southland New Zealand: 1978 to 2012.Proceeding of the New Zealand Grassland Association, vol. 74. Gore: New ZealandGrassland Association Inc.; 2012. p. 147–52.

Smith AP, Western AW. Predicting nitrogen dynamics in a dairy farming catchment usingsystems synthesis modelling. Agr Syst 2013;115:144–54.

Statistics NZ. Dairy industry ‘mooooving’ forward. http://www.stats.govt.nz/browse_for_stats/snapshots-of-nz/yearbook/environment/agriculture/dairy.aspx, 2012. [accessed31 July 2013].

Valentine I, Kemp P. Pasture and supplement resources. In: Rattray PV, Brookes IM, NicolAM, editors. Pasture and supplements for grazing animals. Hamilton, NZ: NewZealand Society of Animal Production; 2007. p. 3–11.

van Bysterveldt A, Moir J, Metherell A. Nutrient management on the Lincoln Universitydairy farm. Prim Ind Manag 2006;9(2):17–22.

Vibart R, Vogeler I, Devantier B, Dynes R, Rhodes T, Allan W. Impact of carbon farming onperformance, environmental and profitability aspects of sheep and beef farmingsystems in Southland. FLRC Workshop, 8–10. February. Palmerston North: MasseyUniversity; 2011.

Vibart RE,White T, Smeaton D, Dennis S, Dynes R, BrownM. Efficiencies, productivity, nu-trient losses and greenhouse gas emissions from New Zealand dairy farms identifiedas high production, low emission systems. In: Currie LD, Christensen CL, editors. Ad-vanced nutrient management. Occasional report no. 25Palmerston North: Fertilizerand Lime Research Centre, Massey University; 2012. p. 1–11.

316 I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317

Author's personal copy

Vogeler I, Beukes P, Romera A, Cichota R. Estimating nitrous oxide emissions from a dairyfarm based on a mechanistic whole farm model and segregated emission factors forNew Zealand. Soil Res 2012;50:188–94.

Vogeler I, Beukes P, Burggraaf V. Evaluation ofmitigation strategies for nitrate leaching onpasture-based dairy systems. Agr Syst 2013;115:21–8.

Wheeler DM, Ledgard SF, Monaghan RM, McDowell R, de Klein CAM. Overseer® nutrientbudget model — what it is, what it does. In: Currie LD, Hanly JA, editors.

Implementing sustainable nutrient management strategies in agriculture. Occasionalreport no. 19Palmerston North: Fertilizer and Lime Research Centre, Massey Univer-sity; 2006. p. 231–6.

Wheeler DM, Ledgard SF, DeKlein CAM. Using the Overseer nutrient budget model to es-timate on-farm greenhouse gas emissions. Aust J Exp Agric 2008;48:99–103.

White TA, Snow VO, King WM. Intensification of New Zealand beef farming systems. AgrSyst 2010;103:21–35.

317I. Vogeler et al. / Science of the Total Environment 482–483 (2014) 305–317