monitoring kangaroo populations in southeastern …...monitoring kangaroo populations in...

Monitoring Kangaroo Populations in Southeastern New SouthWales

A. R. Pople1, S. C. Cairns2 and N. Menke1

1The Ecology Centre, University of Queensland Q 4072

2Zoology, School of Environmental Sciences and Natural Resources Management,University of New England, Armidale NSW 2351

Prepared for the New South Wales National Parks and Wildlife Service

July 2003

Pople et al.: Monitoring kangaroos in southeast NSW 2

Summary

1. This study was undertaken to assess the feasibility of conducting kangaroosurveys in an area encompassed by the Braidwood, Cooma, Goulburn, Gundagaiand Yass Rural Lands Protection Boards (RLPBs). National Parks and StateForests comprise almost 30% of the region and were excluded from the surveyarea because they preclude commercial harvesting.

2. The area was visited and discussions were held with NSW National Parks andWildlife Service (NPWS) and RLPB staff to develop some idea of the relativedensity distribution of kangaroos within the area. The distribution of licences forthe non-commercial cull of kangaroos provided further indication of kangaroodistribution.

3. Because of the relatively high relief of the landscape and the relatively highproportion of tree cover in the area, a helicopter rather than a fixed-wing aircraftwas suggested as the most suitable survey platform. Line transect samplingconducted as double counts should minimise bias and maximise repeatability ofthe population estimates.

4. Using information on tree cover and the relief of the landscape, the five RLPBswere stratified into probable areas of high, medium and low kangaroo density.This enabled survey effort to be allocated to each stratum to minimise thevariance of the population estimate.

5. Based upon the density stratification of the area and information drawn fromhelicopter surveys conducted in the northern tablelands of NSW, a surveycomprising 900 km of transect line is proposed to estimates kangaroo densities inthe agricultural and grazing lands of the five RLPBs.

6. A risk assessment conducted to examine to combined effects of survey precision,survey frequency and harvest rate suggested an appropriate survey strategywould be to carry out triennial helicopter line transect surveys with a precision of20%.


1. Introduction

Currently, commercial harvesting of kangaroos in New South Wales is restricted tothe inland sheep-wheat belt, the western pastoral areas and the northern tablelands(Fig. 1). No commercial harvesting of kangaroo occurs east of the Great DividingRange, nor in the central and southern tablelands. There has, however, been anexpression of interest, largely from sheep and cattle graziers, in allowing commercialharvesting in the southeast of the state. These landholders presently cull kangarooson pest destruction (shoot-and-let-lie, SLL) licences, but recognise that this is bothan inefficient means of pest control and wasteful, and could be redressed throughthe introduction of commercial harvesting. A further advantage of introducingcommercial harvesting is that animals are likely to be killed more humanely byqualified field processors than if killed by landholders under SLL licences, and thiscan be policed to a far greater extent than a non-commercial cull (RSPCA 2002).

If commercial harvesting is to occur in the region, then state and federal governmentconservation agencies require kangaroo populations to be monitored. Elsewhere inthe state, harvesting of kangaroo populations is regulated by quotas that are set asproportions of population estimates determined usually by aerial survey. This is aproportional harvesting strategy employed throughout Australia for kangaroomanagement (Pople in press).

Aerial surveys are conducted annually by fixed-wing aircraft in the western plainswhere most harvesting occurs. In the northern tablelands and the Barrier Rangessurveys are less frequent. This is because harvesting is less intensive in theseareas and either the rugged terrain or heavy vegetation cover requires surveys to beundertaken by helicopter (Southwell and Sheppard 2000) or by ground counts(Southwell et al. 1995), which are costlier than surveys by fixed-wing aircraft.

This report examines the feasibility of monitoring kangaroo populations in thesoutheast of New South Wales, specifically in the five Rural Lands Protection Boards(RLPBs) shown in Figure 2. This region comprises rugged terrain and heavilytimbered areas, making aerial survey difficult. However, where surveys are feasible,a monitoring system is suggested. This includes survey design (i.e. transect numberand placement) and survey frequency.

In wildlife management, the appropriate survey frequency and precision hasgenerally been considered with the aim of detecting trends (e.g. Caughley 1979;Gerrodette 1987; Harris 1986). In kangaroo management, harvest regulation isprimarily through quotas that are set as proportions of absolute estimates ofpopulation size. Trends are of secondary importance. Imprecision in populationestimates and infrequent estimates will risk applying a quota that is either too high ortoo low (i.e. over- or underharvest, see Pople 2003). By not harvesting at thedesired rate (e.g. 15%), costs are incurred to the kangaroo industry through reducedand more variable yield, to graziers through increased competition with sheep anddamage to crops and there is a social cost if kangaroos are reduced below somearbitrarily low density (i.e. quasiextinction). These costs must be balanced againstthe cost of more intensive and frequent surveys (i.e. there is a trade-off). Alternativeharvest strategies can also be considered in order to reduce the risk of over- or


underharvest. These include harvesting at a different, and even variable, rate,regulating effort or incorporating spatial reserves.

Figure 1. Zones within New South Wales where commercial harvesting is allowed.Fixed-wing aerial surveys are conducted in zones 1-8, 10 and 11. Helicoptersurveys are conducted in zones 9, 13 and 14 (After New South Wales NationalParks and Wildlife Service 2001).

�

�

� ��

��

�

�

�

�

�

�

�

��

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

��

�

�

�

�

�

�

�

�

�

� �

�

��

�

��

��

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

��

� �

Zone 2

Zone 11

Non CommercialZone

Zone 8

Zone 4

Zone 1Zone 7

Zone 10Zone 6

Zone 13

Zone 9

Zone 14

HAY

MOREE

LOUTH

TILPA

GUYRA

COBAR

DUBBO

BOOMI

LEETON

BOURKE

COROWA

WALCHA

WAKOOL

BARHAM

NYNGAN WARREN

COOLAH

YETMAN

FORBES

PARKES

SYDNEY

BINGARA

BURONGA

WALGETT

CARINDA BARRABAPILLIGA

WEE WAA

IVANHOE

NYMAGEE DUNEDOO

COOMBAH

MARSDEN

GOODOOGA

INVERELL

WARIALDA

GRIFFITH

QUAMBONE

BUNDARRA

BOGGABRI

GUNNEDAH

BAAN BAA

NARRABRI

ARMIDALE

TOCUMW AL

LOCKHART

COOLABAH

QUIRINDI

TAMW ORTH

MUNGINDI

BOOLIGAL

MENINDEE

ENNGONIA

POONCARIE

WENTWORTH

ARDLETHAN

BALRANALD

COONAMBLE

NECKARBOO

MOULAMEIN

HERMIDALE

TOMINGLEYTULLAMORE

PEAK HILL

NARROMINE

TOTTENHAM

WILCANNIA

NEVERTIRE

GILGANDRA

WILLANDRA

WANAARING

BARRINGUN

MILPARINKA

TIBOOBURRA

GLEN INNES

NARRANDERA

PACKSADDLE

DENILIQUIN

MOUNT HOPE

BOGGABILLA

BREWARRINA

BOGAN GATECONDOBOLIN

BINNA BUNNA

TENTERFIELD

BROKEN HILL

COLLARENEBRI

WHITE CLIFFS

WEILMORINGLE

WEST W YALONG

COONABARABRAN

TAMBARSPRINGS

BURREN JUNCTION

HUNGERFORD GATE

LAKE CARGELLIGO

RANKINS SPRINGS

ACT

0 50 100 150 200 Kilometers

N

EW

S

Kangaroo Management Zones

Kangaroo ManagementZones

Zone 1 TibooburraZone 2 Broken HillZone 4 Lower DarlingZone 6 CobarZone 7 Cobar NorthZone 8 NarrabriZone 9 ArmidaleZone 10 CoonabarabranZone 11 GriffithZone 13 Glen InnesZone 14 Upper Hunter


Figure 2. The study area in southeastern NSW. The eastern third of the BraidwoodRural Lands Protection Board (RLPB) is dominated by National Parks, State Forests,areas of high relief and heavy forest cover, and so has been excluded from the studyarea. CR, Crookwell; GO, Goulburn; YA, Yass; CT, Cootamundra; TU, Tumut; CM,Cooma. Areas of high relief and open forest or woodland were determined from therelief coverage (Geoscience Australia 2001), which is © Commonwealth of Australia(Geoscience Australia) 2001, and landuse coverage (Bureau of Rural Sciences2001), © Commonwealth of Australia (National Land and Water Resources Audit)2001.


2. Survey feasibility

Ground assessment of the study area was made in February 2003. This involvedfield inspection of the major environments in the region and discussions with NewSouth Wales National Parks and Wildlife Service (NSW NPWS) staff at Bungonia,Queanbeyan, Jindabyne and Tumut, discussions with RLPB rangers at Goulburn,Cooma, Gundagai and Yass, and discussions with graziers at Yass. Thesediscussions identified likely areas of relatively high kangaroo density within theregion, and therefore those areas that should be targeted for assessment of surveyfeasibility. Obviously, these areas could also be used as a basis for stratification(see Section 3 below). Field inspection and discussions were also designed toidentify and hopefully confirm the likely factors determining kangaroo distribution andabundance (Pople 1989; Southwell et al. 1999) within the region, thus providing afurther basis for stratification.

Forest cover, land use and areas of high relief within the study area are shown inFigure 2 and Appendices 1 and 2. A large proportion (28%, Table 1) of the region isin National Park, particularly Kosciuszko and Budawang National Parks, or StateForest from which commercial harvesting of kangaroos is precluded. The accuracyand repeatability of aerial surveys diminishes in areas of high relief or heavy forestcover because of low and variable visibility (Clancy et al. 1997; Pople 1999;Southwell 1989; Southwell and Sheppard 2000). Safety may also be compromisedin particularly rugged terrain. Such areas occur in the region, but are largely withinNational Parks or State Forests (Appendices 1 and 2). In particular, there is heavytree cover in the northwestern area of Kosciuszko National Park, the adjoining andnearby State Forests between Tumut and Wee Jasper, and the eastern slopes of theGreat Dividing Range that are within the Braidwood RLPB (Appendix 1). Areas ofhigh relief occur in the eastern slopes of the Great Dividing Range, south from theSnowy River within Kosciuszko National Park in the southwest of the Cooma RLPB(Appendix 2).

The coastal area of the Braidwood RLPB stretches from Sussex Inlet to Moruya.Aside from the difficulties of surveying this area of often-heavy forest cover, the areais unlikely to be suitable for commercial harvesting because the area is heavilypopulated with numerous small landholdings making administration of any harvestdifficult and potentially contentious. Further, the area outside National Park andState Forest is small so the absolute kangaroo population size is unlikely to supporta commercial industry. The eastern third (38%) of the Braidwood RLPB, effectivelyeast of and including Budawang National Park, was therefore excluded from thestudy area.

The composition of the remaining area is shown in Figure 2 and quantified in Table1. The three density strata are described in Section 3 below. A small area ofrelatively high relief (>500 m) occurs within the area outside National Parks andState forests. This is likely to be unsuitable for aerial survey and so was excludedfrom the overall survey area, effectively assuming a density of zero in areas of highrelief. High relief areas are generally associated with high tree cover and thecombination means kangaroo density is likely to be low in these areas andharvesting is unlikely to be feasible. The impact of this small deletion wouldtherefore be slight.


The survey area shown in Figure 2 ranges from low, moderately timbered rangesbordering largely open plains in the southeast (Cooma and Braidwood RLPB) andaround Tumut and Wee Jasper, to rolling hills of scattered woodland in the north andwest. The northern and western areas could be surveyed by fixed-wing aircraft, butcorrection factors would need to be developed for this and the remaining part of theregion is unsuitable for this survey method. A helicopter offers a number ofadvantages over fixed-wing aircraft as a platform for surveying this region. Itprovides greater visibility and can fly lower and slower, which is particularlyadvantageous in rugged terrain and in areas with heavy tree cover. The greatermanoeuvrability of helicopters allows surveys to be designed with shorter and moreclosely spaced transect lines, which is an advantage in heterogeneous landscapes.Finally, helicopters provide an appropriate platform for line transect sampling (Clancyet al. 1997), more so than fixed-wing aircraft (Pople et al. 1998), potentially allowingsurvey-specific correction for variations in visibility bias. Ideally, helicopter linetransect surveys should be conducted as double counts (Borchers et al. 1998; Poplein press), which would allow some assessment of bias due to incomplete sighting ofkangaroos on the transect line.

Table 1. Areas (km2) of five Rural Lands Protection Boards (RLPBs) proposed forcommercial harvesting. Harvesting is precluded from National Parks (NPs) andState Forests (SFs) and there are some areas of high relief outside NPs and SFsthat are unsuitable for aerial survey. The remaining area is divided into three stratarepresenting likely high, medium and low kangaroo density.

Braidwood Cooma Goulburn Gundagai Yass TOTAL

RLPB area 8,824 11,375 6,426 9,507 6,305 42,438

High relief outsideNPs and SFs 261 431 35 174 146 1,047

High density 135 0 56 3,691 3,067 6,949Medium density 3,848 5,981 4,520 1,723 1,451 17,524

Low density 301 1,291 1,396 959 1,097 5,044

Survey area 4,284 7,271 5,973 6,373 5,615 29,516

3. Survey design

3.1 Objective

The overall objective of a survey of kangaroos in the study area is to estimatepopulation size from a representative sample. This initially provides an indication ofwhether or not harvesting is commercially viable and worth the administrative costsin the long term. If harvesting is viable, an estimate allows quotas to be set as aproportion of population size. Population size needs to be estimated efficiently, interms of cost and time, and with a precision (i.e. confidence interval) sufficient tokeep the risk of over- or underharvest to an acceptable level (see Section 4 below).


A survey will also indicate the distribution of kangaroos, allowing spatial allocation ofquotas or effort.

Precision can be greatly improved by stratifying sampling. As a rough rule of thumb,sampling effort should be allocated in proportion to the size of the population in eachstratum (Thompson 1992). For this, some indication is required of the relativedistribution of kangaroos amongst strata in the study area as well as strata area.Ideally, strata should be some mappable component of the environment such asvegetation cover or soil type. However, arbitrary strata can be defined so long asthey can be mapped, their area is known and they are defined before a survey.Post-sampling stratification is possible, but it must obviously be based on informationindependent of the survey and it is likely to be well short of optimal.

3.2 Kangaroo distribution

Discussions with NSW NPWS staff, RLPB rangers and graziers suggested thefollowing distribution of eastern grey kangaroos in the study area:

Goulburn. Tags issued under SLL licences (~3,000 in 2002) have beendistributed throughout the district, with possible concentrations around Taralgaand Lake Bathurst. There is a patchy distribution of kangaroos in the districtassociated with cover.

Yass. Most tags issued under SLL licences (~6,000 issued in 2002 to shiresadministered by the Queanbeyan NPWS office) have been issued in a bandrunning from Gunning through Crookwell to Boorowa. Graziers mapped what theyconsidered an area of high kangaroo density in the RLPB. They considereddensities were higher in Yass and Gundagai RLPBs than in Goulburn, Braidwoodand Cooma RLPBs.

Gundagai. The Tumut NPWS office issued ~5,000 tags under SLL licencesthroughout the Gundagai and Hume RLPBs in 2001. Many of these were issuedin the Holbrook area (Hume RLPB) reflecting relatively high numbers. Within theGundagai RLPB, kangaroos occur in moderate densities in valleys south of theHume Highway where grazing properties adjoin State Forest and National Park.There are relatively high densities in hilly areas north and south of Gundagai, butmuch lower densities in the west towards Junee. Essentially, kangaroos wereclosely associated with moderate tree cover. The RLPB ranger mapped thesestrata.

Braidwood. Relatively high kangaroo density in valleys to the south andsouthwest of Braidwood. Anecdotal evidence suggested an increase in greykangaroos in the Queanbeyan district since the early 1960s.

Cooma. The number of tags issued under SLL licences annually was notavailable, but 3,000 is a very rough guess. Kangaroos occur throughout theRLPB. One argument was that kangaroos were concentrated in the foothills andmore heavily timbered areas on the margins of the RLPB, but density has becomemore homogeneous. Other people argued that densities of kangaroos were stillhigher in the higher rainfall area in the east (e.g. Kybeyan) and partly timbered


areas in the south and west. Overall this suggests a higher density stratum on themargins of the RLPB.

Shoot-and-let-lie licences were considered to be only a rough guide to kangaroodistribution and abundance. The number of tags issued in 2002 was unusually highbecause of the drought. Nevertheless, it appears <20,000 tags are issued annually(2001-2002) for an area larger than the five RLPBs here. This compares with acombined commercial and non-commercial take of 35-50,000 kangaroos (easterngrey and wallaroos) each year (1998-2001) in the Armidale and Glenn Innes zones(NSW NPWS unpublished data), encompassing an area of ~37,000 km2 in similarterrain and vegetation. The tag issue would have to grossly underestimate thenumber shot each year in the southern tablelands for the take to be comparable tothat in the northern tablelands. The comparison is rough and clearly a more rigorouspopulation estimate is required. The offtake, or ideally population density, belowwhich a harvest will no longer be commercially viable will be somewhere below thatin the northern tablelands. Hacker et al. (2003) suggest 5 kangaroos km-2 as aminimum density.

Discussions also revealed that there are similar densities of kangaroos and similarconcerns from graziers and crop farmers in districts northwest and southwest of thestudy area. This suggests that there may be interest in further expanding thecommercial zone if commercial harvesting occurs in the five RLPBs examined here.

3.3 Stratification and allocation of survey effort

Ground assessment of the region did confirm the association between eastern greykangaroos and partly cleared forest (cf. predominantly open or predominantly closedareas) and areas of greater pasture productivity (Hill 1981a; Hill 1981b; Scott-Kemmis 1979). Three strata were recognised and these are shown in Figure 3 andtheir areas given in Table 1. Relatively high kangaroo densities are predicted in theGundagai and Yass RLPBs and these were mapped by graziers and RLPB staff.Relatively moderate density is expected in partly timbered areas outside this stratum.A buffer of 3 km was therefore placed around wooded (open forest and woodland)areas (Fig. 2 and Appendix 1). The remaining open areas comprised a low-densitystratum. For the total survey area, 24% was high-density stratum, 59% was mediumdensity and 17% was low density (Table 1).

Survey effort using line transects is measured by total line length. If 100 km oftransect is adequate to survey a species’ population in a small area, then 100 km willbe adequate over a larger area, assuming similar dispersion and random sampling oftransects. Increasing line length will improve precision. The line length required fora target precision can be determined from a pilot study or drawn from surveys in asimilar environment. The target precision (i.e. standard error/mean) for theseproposed surveys is 20% (see section 4).


Figure 3. Design for a survey of kangaroos in the study area in southeastern NSW.The survey area is restricted to areas outside National Parks and State Forests andareas without high relief, which are shown in Figure 2. The survey area is dividedinto three strata according to expected kangaroo density and transects placedrandomly. CR, Crookwell; GO, Goulburn; YA, Yass; CT, Cootamundra; TU, Tumut;CM, Cooma


The total line length desired for these surveys was determined in two ways. Firstly, itwill be constrained by cost. In the northern tablelands, 1,480 km were surveyed in2001 and 2002 in three zones covering an area of 52,000 km2 (Cairns 2003). Thesurvey area in southern NSW is smaller and so a total line length of ~900 km wouldbe in proportion to area. However, randomly placed transects (proposed here, seebelow) are likely to be costlier to survey than parallel transects in blocks because ofincreased ferry distance between lines. The second method for determining theappropriate line length involves using the precision of the surveys in the northerntablelands as a guide. The relevant equation is (Buckland et al. 1993):

2

2

)()(

ST

NTNTST Dcv

DcvLL =

where cv(DNT) and LNT are the respective precision and line length of the northerntablelands survey, and cv(DST) and LST are the corresponding values for theproposed southern tablelands surveys here. In the northern tablelands,cv(DNT)=12% for LNT=1480 (Cairns 2003). In the southern tablelands, ifcv(DST)=20% (see Section 4), then LST = 510 km. The precision of the northerntablelands surveys is based on transects in nine blocks of 3-6 × parallel transects of30-50 km each separated by 10 km. Precision is likely to have been poorer iftransects were placed throughout the survey area. However, stratification is likely toimprove precision, so a line length of ~500 km may be sufficient to achieve aprecision of 20%. A line length of 500 km should also be more than adequate toachieve 60-80 sightings, which are considered a minimum for distance sampling(Buckland et al. 1993).

If the study area is administered as more than one zone, then each will require 500km of survey line to achieve a precision of 20%. If 900 km is surveyed, then there islikely to be the option of breaking the region into two zones. A total line length of 900km was therefore used to design the survey here.

Transect lines were located within strata by randomly selecting intersection points ina 10 km × 10 km grid placed over the study area. The selected point formed themidpoint of a 30 km east-west transect. If the transect extended beyond a stratum inone direction it was truncated and extended in the other direction. If it could not fitinto the stratum on a gridline, the remainder of the line was placed on theneighbouring gridline to the north or south. A transect length of 30 km ensured >20replicate lines over the study area for adequate precision and a representativesample.

Allocation of line length to strata was based on ‘rough’ estimates of the relativedensities in each stratum. In the northern tablelands, surveys recorded areas of lowdensity of ~1 kangaroo km-2, areas of medium density of 5 kangaroos km-2 and high-density areas of ~10 kangaroos km-2 (Cairns 2003). These were used in the stratahere. The required line length in each stratum in each RLPB was determined byallocating the 900 km in proportion to the estimated population size in each stratumin each RLPB (Tables 2 and 3). Transects were selected in each RLPB in turn,rather than the entire survey area at once, to ensure good coverage, at least at theRLPB scale. Only short line lengths were required in the low-density stratum, so


these were usually just an extension of a line in an adjoining stratum to reduce ferrytime between transects. This may introduce some bias and underestimate variancebecause edge habitat may be oversampled. However, the problem is likely to beonly slight because density should be low in this stratum and the medium densitystratum includes a liberal buffer around wooded areas. The resulting survey designis shown in Figure 3.

Table 2. ‘Rough’ estimates of abundance of eastern grey kangaroos in five RuralLands Protection Boards in the southern tablelands of NSW. These are calculatedfrom densities in three strata shown in Figure 3 and their areas (see Table 1).Densities are extrapolations from surveys in the northern tablelands of NSW (Cairns2003).

Abundance ‘guess’Stratadensity 'guess' km-2

Braidwood Cooma Goulburn Gundagai Yass TOTAL

(High density) 10 1,347 0 565 36,911 30,669 69,493(Medium density) 5 19,241 29,903 22,602 8,616 7,256 87,618

(Low density) 1 301 1,291 1,396 959 1,097 5,044

Survey area 20,889 31,194 24,563 46,486 39,023 162,154

Table 3. Allocation of survey line length (km) according to estimated abundance(see Table 2) in each stratum in each Rural Lands Protection Board.

Strata Braidwood Cooma Goulburn Gundagai Yass TOTAL

High density 7 0 3 205 170 386Medium density 107 166 125 48 40 486

Low density 2 7 8 5 6 28

Survey area 116 173 136 258 217 900

These very rough estimates of kangaroo numbers in each RLPB suggest that thearea could be divided into two zones of similar kangaroo abundance. Yass andGundagai could form one zone while Goulburn, Braidwood and Cooma could formthe other. Survey-based estimates are obviously needed before a decision is madeon this basis.

Surveys could be conducted over five days, excluding weather delays. Thisassumes 200-250 km could be flown in a survey session.


4. Survey precision and frequency

4.1 Methods

The most appropriate harvest strategy, frequency and precision can be selectedfrom various alternatives by comparing the risks of undesirable managementoutcomes from each alternative. An undesirable outcome may be couched in termsof the goals of conservation, sustained-yield harvesting or pest control, such asquasiextinction risk, reduced yield or excessive kangaroo density. Selection of thebest harvest strategy and monitoring system will be constrained by feasibility,primarily cost.

Risk of over- and underharvest was assessed through simulating a kangaroopopulation under various management scenarios of survey frequency, precision andharvest strategy. Caughley’s (1987) interactive model was used to provideconservative estimates of risk (Pople 2003) (i.e. tend to overestimate risk). Althoughthe model was developed for an arid-zone population of red kangaroos, it can yieldresults relevant to eastern grey kangaroos in a more mesic environment. Firstly,predictions of such entities as harvest rates for maximised yield, average populationsize and harvest yield were not sought; rather the objective of the modelling was tocompare management scenarios. It is the qualitative rather than the quantitativeresults of the modelling that are of relevance. Nevertheless, the modelled populationneeded to show similar dynamics and a similar response to harvesting to an easterngrey kangaroo population in southeastern NSW. The population fluctuations inCaughley’s (1987) interactive model therefore required dampening by reducing thestandard deviation in seasonal rainfall by 37% so that it was equivalent to thataround Yass.

Modelling follows a modification of the methods described by Pople (2003). Briefly,in the interactive model, rainfall drives pasture biomass, which in turn determines therate of increase of the kangaroo population. There are two negative feedback loops.The first is pasture biomass reducing pasture growth. The second is kangaroodensity reducing kangaroo rate of increase by reducing the available biomass byeating it. Seasonal rainfall was drawn from a lognormal distribution with a meanfrom Menindee Post Office in western NSW (annual mean = 244 mm, s.d. = 106mm), where Caughley’s (1987) interactive model was parameterised, but a standarddeviation of 63% of the Menindee value to reflect the more stable environment (seeabove). The kangaroo population was harvested and the pasture grew and wasgrazed down in weekly time steps. Sheep competed with the kangaroo population,consuming 1.5 times the pasture eaten by the kangaroos. The initial population sizefor each simulation was 10 kangaroos km-2.

Instantaneous harvest rates were converted to isolated rates of harvesting(Caughley 1977), appropriate to the time step, to simulate harvesting spread evenlythroughout each year. Harvest offtake for each three months was determined fromthe most recent population estimate. Harvest offtake was therefore the same foreach three months of the year, despite population size changing at each threemonthly time step. Population estimates were drawn from a lognormal distributionwith a standard deviation dependent upon the survey precision (standarddeviation/mean). Survey precision varied from 0 to 2 (i.e. 0-200%), but was set at


0.2 unless otherwise stated. For each scenario, the average of 1,000 simulations isreported. Population models were run in Excel with the add-in POPTOOLS (GregHood CSIRO 2002, http://www.dwe.csiro.au/vbc/poptools/index.htm).

4.2 Results

4.2.1 Survey frequency

As expected, the risk of quasiextinction is lower than that reported for the populationmodelled with rainfall variability for Menindee (Pople 2003). Reducing surveyfrequency not surprisingly increases the probability of quasiextinction (Fig. 4).Although the increase does not appear dramatic, a five-year survey frequencyreturns a risk of ~10% of the population falling below 2 kangaroos km-2 and ~20% fora threshold of 5 kangaroos km-2. The equivalent risks for an annual surveyfrequency are <1% and 10%, respectively.

Reducing survey frequency results in only a slight increase in population and harvestvariability and has little effect on average harvest offtake and population density.Quasiextinction is likely to be of concern to the kangaroo industry and landholders aswell as conservationists, because harvesting is unlikely to be commercially viablebelow some threshold density. Again, Hacker et al. (2003) suggest this is around 5kangaroos km-2. However, in practice there are regions where harvesting occurs atdensities as low as 2-3 kangaroos km-2, such as in the northwest of the SouthAustralian Pastoral zone (Cairns and Grigg 1993).

These simulations assume that quotas remain constant between surveys. In thepast, quotas have been adjusted in the northern tablelands for changes in kangaroonumbers in adjoining management zones and for recent rainfall (Gilroy 1999).Quotas have been similarly adjusted in Western Australia where the harvest areawas surveyed triennially, but now there are three survey areas with a different areasurveyed each year (Pople and Grigg 1998). The risks of overharvest described bythese simulations are therefore worst-case scenarios. A simple rule wasincorporated into the simulations whereby a survey was undertaken at set frequency(three or five years) unless rainfall in the previous year had been ≥ 50% belowaverage, in which case the population was surveyed. This made little difference toquasiextinction risk compared to annual surveys.

If the harvest rate is reduced to 10%, then the risk of quasiextinction is greatlyreduced (Fig.5). It is only at the lower threshold of 2 kangaroos km-2 that reducingsurvey frequency with a 10% harvest becomes riskier than a 15% harvest withannual surveys. Even then, it is only when surveys are as infrequent as every fiveyears. On average, reducing the harvest rate from 15% to 10% results in anincrease in average population density of 27% and a reduction in average harvestofftake of 16%.


Fig. 4. Probability of quasiextinction over 20 years from a harvest rate of 15% with asurvey frequency of 1-5 years.

Fig. 5. Probability of quasiextinction over 20 years from a harvest rate of 10% with asurvey frequency of 1-5 years. The risk for a 15% harvest with an annual surveyfrequency is shown as a dashed line.

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0 1 2 3 4 5 6 7 8 9 10

Threshold density (kangaroos km-2)

Ris

kof

falli

ngbe

low

thre

shol

din

20y

1

3

5

15% annual

Surveyfrequency

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0 1 2 3 4 5 6 7 8 9 10


Ris

kof

falli

ngbe

low

thre

shol

din

20y

1

2

3

4

5

Surveyfrequency


4.2.2 Survey precision

Pople (2003) found that precision of ≤ 50% had little effect on the probability ofquasiextinction under harvesting. Here, a broader range of possible precisions anddrawing population size from a lognormal rather than a normal distribution identifieswhen there is a substantial risk (Fig. 6). Above 50%, precision increases theprobability of quasiextinction dramatically. Interestingly, there is an interactionbetween survey precision and frequency (Fig.7). Reducing survey frequency fromevery one to every five years when precision is 50% has a much greater impact onquasiextinction risk than if precision is only 20%.

4.2.3 Trade-off between survey frequency and cost

The appropriate survey frequency will depend on the costs of the surveys and thecosts of compromising the harvest strategy through not always knowing populationsize. Imprecision can also incur costs, but this only seems a problem in this casewhen precision is >50% (Fig. 6). To make costs comparable, the potential gains andlosses incurred by the kangaroo industry, graziers, conservationists and any otherstakeholders from adopting various monitoring programs need to be put onto thesame scale. This would require detailed exploration and so is not determined here.Instead, a simple estimate of survey cost is contrasted with quasiextinction risk forincreasing survey frequency. This should provide a first approximation of theappropriate survey frequency and the nature of the trade-off.

There are fixed costs associated with conducting aerial surveys that do not change ifsurveys are conducted less frequently. These include labour costs and can beassumed to be $50,000 per annum. The variable costs primarily comprise aircraftrunning costs and are assumed to be $50,000 per survey. The total five-year surveycost (fixed + variable x no. of surveys) is shown in Figure 8. This shows the ever-diminishing savings from reducing survey frequency. Quasiextinction risk fromFigure 4 is also shown in Figure 8, but only for a threshold density of 2 kangarooskm-2. Risk of quasiextinction increases dramatically beyond a survey frequency ofevery four years. The actual probability should be interpreted cautiously as it isdependent on the model structure and parameters and these have not beenassessed for this population. Furthermore, the acceptable level of risk is a valuejudgement.


Fig. 6. Probability of quasiextinction over 20 years from a harvest rate of 15% with aprecision (SD/mean) of 0-2.

Fig. 7. Probability of quasiextinction over 20 years from a harvest rate of 15% with aprecision (CV=SD/mean) of 0.2 and 0.5 and survey frequency of 1 and 5 years.

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0 1 2 3 4 5 6 7 8 9 10


Ris

kof

falli

ngbe

low

thre

shol

din

20y

2.0

1.0

0.5

0.2

0.0

CV(N)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0 1 2 3 4 5 6 7 8 9 10


Ris

kof

falli

ngbe

low

thre

shol

din

20y

5 and 0.5

5 and 0.2

1 and 0.5

1 and 0.2

Surveyfrequencyand CV


Fig. 8. Probability of the population dropping below 2 kangaroos km-2 (i.e.quasiextinction) over 20 years (solid line) from a harvest rate of 15% with a precision(SD/mean) of 0.2 and survey frequency of 1 and 5 years. Also shown is the cost ofaerial surveys over five years (dotted line) given survey frequencies of 1-5 years.

0.00

0.02

0.04

0.06

0.08

0.10

0 1 2 3 4 5 6Survey frequency

Pro

bab

ility

ofD

<2k

m-2

250

300

350

400

450

500

x$1

,000

P$


5. Management recommendations

Whether the study area should be divided into two or more zones will depend partlyon whether management objectives vary within the area, partly on ease ofadministration and partly on the precision of population monitoring. The surveydesign shown in Figure 3 should provide population estimates with an acceptableprecision of 20% in two zones in the study area, but this needs to be realised.Shoot-and-let-lie licences are currently issued from four NPWS offices. It wouldseem logical to issue commercial harvesting tags and licences and SLL licencesfrom the same office. The problem is that NPWS administrative boundaries for theissue of SLL licences are quite different to RLPB boundaries. There are a number ofoptions. Quotas could be issued for each RLPB and commercial tags and licencesissued by all four NPWS offices. This has the advantage of tags being issued byNPWS staff most familiar with a property. There is the complication of allocating aregional quota to smaller areas, but this is done presently in the commercial zonebecause tags are issued to properties. The difference here is that there is anadditional step in dividing the zone quotas among the four NPWS offices. This couldbe done fairly easily by overlaying the NPWS administrative boundaries over themap showing density strata in Figure 3. Alternatively, commercial tags and licencescould be issued centrally, say from the Queanbeyan office.

A survey frequency of three years with a precision of 20% appears to increase therisk of overharvest only slightly, while substantially reducing survey costs. Whetherthe increase in risk is acceptable is a decision for management in consultation withstakeholders.

Other harvest strategies could be employed including a lower harvest rate (e.g. 10%)or perhaps limiting harvest effort by restricting the number of shooters. Given thatalmost 30% of the area is in National Park or State Forest, and terrain andvegetation will limit shooter access to other areas, there is a considerable safety netto guard against overharvesting. It is now well known that such spatial refuges canminimise the risk of overexploitation (Hall 1998; Lauck et al. 1998). To be effective,protected areas need to contain source populations that can repopulate areasdepleted through harvesting. In the short term (i.e. <5 years), harvested areasadjoining National Parks and State Forests are likely to have their populationsbolstered in this way. However, migration to areas remote from refuges will onlyoccur over the longer term, perhaps >10 years. A further qualifying point is thatoverall kangaroo densities within these refuges are likely to be lower than thesurrounding regions. The higher densities will be on the margins of refuges, whereforested areas abut open grazing lands. In short, spatial refuges in the region shouldensure regional persistence and will reduce the risk of overharvesting in the vicinityof a refuge, but be less effective with distance from a refuge.

A further safety net would be a threshold population density below which there is noharvesting. In theory, when there is uncertainty about a population’s size anddynamics, harvest offtake can be maximised and low probability of quasiextinctionachieved by imposing a threshold density and allowing higher rates of harvest abovethe threshold (Engen et al. 1997; Milner-Gulland et al. 2001). The problem is thatthe temporal variation in harvest offtake increases with the density at which thethreshold is set. This has been discussed for kangaroos by Pople (2003) who


considered that, given that the kangaroo industry is unlikely to operate at densitiesbelow 2-5 kangaroos km-2, imposing a threshold around this density would ensurethat this in fact occurs. This would guard against an increase in the value ofkangaroo products allowing shooters to operate at lower densities. It would alsoincrease public confidence that the harvest is sustainable, because they would notbe relying on the industry’s decision to cease harvesting being determined by marketforces.

Acknowledgements

Numerous people provided helpful discussion on the distribution and abundance ofkangaroos in the district, including Lisa Collins, Andrew Moore, Simon Allender andPam O’Brien from NPWS offices, Mark McGaw, Tim Sears, Col Elphick, IanKlingham and Kevin Baker from RLPB offices, and landholders Lindsey Butt, GeoffMcFarlane, Tony O’Shea, Andrew Bain and John Betts, and Bruce Hazell fromBookham Agricultural Bureau. We also thank Stephen Naven for providingcoverages of RLPBs, National Parks and State Forests.

References

Borchers DL, Zucchini W, Fewster RM (1998) Mark-recapture models for linetransect surveys. Biometrics 54, 1207-1220.

Buckland ST, Anderson DR, Burnham KP, Laake JL (1993) 'Distance Sampling.'(Chapman and Hall: London)

Bureau of Rural Sciences (2001) '1996/97 Land Use of Australia, Version 2.'National Land and Water Resources Audit, Canberra.

Cairns SC (2003) 'A report to the New South Wales National Parks & Wildlife Serviceon the consultancy: Kangaroo Monitoring New England TablelandsHelicopter Survey.' Unpublished report to New South Wales National Parksand Wildlife Service, Dubbo, NSW.

Cairns SC, Grigg GC (1993) Population dynamics of red kangaroos (Macropusrufus) in relation to rainfall in the South Australian pastoral zone. Journal ofApplied Ecology 30, 444-458.

Caughley G (1977) 'Analysis of Vertebrate Populations.' (Wiley and Sons: London)

Caughley GJ (1979) Design for aerial censuses. In 'Aerial Survey of FaunaPopulations' pp. 15-23. (Australian Government Publishing Service: Canberra)

Clancy TF, Pople AR, Gibson LA (1997) Comparison of helicopter line transects withwalked line transects for estimating densities of kangaroos. Wildlife Research24, 397-409.


Engen S, Lande R, Saether B-E (1997) Harvesting strategies for fluctuatingpopulations based on uncertain population estimates. Journal of TheoreticalBiology 186, 201-212.

Geoscience Australia (2001) ’Relief. Digital dataset.’ National Land and WaterResources Audit, Canberra.

Gerrodette T (1987) A power analysis for detecting trends. Ecology 68, 1364-1372.

Gilroy J (1999) Kangaroo monitoring in relation to the kangaroo management plan inNew South Wales. Australian Zoologist 31, 306-308.

Hall SJ (1998) Closed areas for fisheries management - the case consolidates.Trends in Ecology and Evolution 13, 297-298.

Harris RB (1986) Reliability of trend lines obtained from variable counts. Journal ofWildlife Management 50, 165-171.

Hill GJE (1981a) Distribution of grey kangaroos in south, inland Queensland.Australian Rangeland Journal 3, 58-66.

Hill GJE (1981b) A study of the habitat preferences in the grey kangaroo. AustralianWildlife Research 8, 245-254.

Lauck T, Clark CW, Mangel M, Munro GR (1998) Implementing the precautionaryprinciple in fisheries management through marine reserves. EcologicalApplications 8, s72-s78.

Milner-Gulland EJ, Shea K, Possingham H, Coulson T, Wilcox C (2001) Competingharvesting strategies in a simulated population under uncertainty. AnimalConservation 4, 157-167.

New South Wales National Parks and Wildlife Service (2001) ’The New South WalesKangaroo Management Program. A management program for the utilisationof four kangaroo species in NSW.’ (New South Wales National Parks andWildlife Service: Sydney)http://www.nationalparks.nsw.gov.au/npws.nsf/Content/Kangaroo+management+program

Pople A (1989) Habitat associations of Australian Macropodoidea. In ’Kangaroos,Wallabies and Rat-kangaroos’. (Eds G Grigg, P Jarman and I Hume) pp. 755-766. (Surrey Beatty and Sons: Sydney)

Pople A (2003) ’Harvest management of kangaroos during drought.’ Unpublishedreport to New South Wales National Parks and Wildlife Service, Dubbo, NSW.

Pople A (in press) Population monitoring for kangaroo management. AustralianMammalogy.

Pople AR (1999) Repeatability of aerial surveys. Australian Zoologist 31, 280-286.


Pople AR, Cairns SC, Clancy TF, Grigg GC, Beard LA, Southwell CJ (1998)Comparison of surveys of kangaroos in Queensland using helicopters andfixed-wing aircraft. The Rangeland Journal 20, 92-103.

Pople AR, Grigg GC (1998) ’Commercial harvesting of kangaroos in Australia.’Environment Australia, Canberra. http://www.ea.gov.au/biodiversity/trade-use/wild-harvest/kangaroo/harvesting/index.html

RSPCA (2002) ’Kangaroo Shooting Code Compliance: A survey of the extent ofcompliance with the requirements of the Code of Practice for the HumaneShooting of Kangaroos.’ Environment Australia, Canberra.

Scott-Kemmis D (1979) The distribution and abundance of grey kangaroos in relationto environment in arid and semi-arid areas of New South Wales. M.Sc. thesis,University of Sydney, Sydney.

Southwell C (1989) Techniques for monitoring the abundance of kangaroo andwallaby populations. In ’Kangaroos, Wallabies and Rat-kangaroos’. (Eds GGrigg, P Jarman and I Hume) pp. 659-693. (Surrey Beatty and Sons: Sydney)

Southwell C, Sheppard N (2000) Assessing harvested populations of the euro(Macropus robustus erubescens) in the Barrier ranges of western NSW.Australian Mammalogy 21, 165-171.

Southwell CJ, Cairns SC, Pople AR, Delaney R (1999) Gradient analysis ofmacropod distribution in open forest and woodland of eastern Australia.Australian Journal of Ecology 24, 132-143.

Southwell CJ, Weaver KE, Cairns SC, Pople AR, Gordon AN, Sheppard NW, BroersR (1995) Abundance of macropods in north-eastern New South Wales, andthe logistics of broad-scale ground surveys. Wildlife Research 22, 757-766.

Thompson SK (1992) ’Sampling.’ (John Wiley and Sons: New York)


Appendices

Appendix 1. Forest type and land use in the study area. CR, Crookwell; GO,Goulburn; YA, Yass; CT, Cootamundra; TU, Tumut; CM, Cooma. Forest type iscategorised by crown cover: closed forest, 80-100%; open forest 50-80%; woodland20-50%; open, 0-20%. The landuse coverage (Bureau of Rural Sciences 2001) is ©Commonwealth of Australia (National Land and Water Resources Audit) 2001.


Appendix 2. Landscape relief, National Parks and State Forests in the study area.Relief is calculated as the difference in elevation between the nearest ridge uphilland the nearest stream downhill in the catchment of the each stream. Calculationsused the interim 9 sec DEM. CR, Crookwell; GO, Goulburn; YA, Yass; CT,Cootamundra; TU, Tumut; CM, Cooma. The relief coverage (Geoscience Australia2001) is © Commonwealth of Australia (Geoscience Australia) 2001.

monitoring kangaroo populations in southeastern …...monitoring kangaroo populations in...

Documents