Commercial harvesting of Kangaroos in Australia

by Tony Pople and Gordon Grigg
Department of Zoology, The University of Queensland
for Environment Australia, August 1999
Chapters 10,11,12 and 13 and Appendix 1 provided by staff at Environment Australia



If there is to be a harvest of free-ranging kangaroos, whether as a renewable resource or for pest control, or both, effective population monitoring is a necessary prerequisite. Without that, a wildlife manager has no yardstick against which to judge the effectiveness of any management actions and whether or not any of the identified management goals are being achieved.

Employing a tracking strategy (i.e. variable quota) to harvest kangaroos (see Chapter 2), requires, by definition, regular estimates of population size. Furthermore, an understanding of the forces driving kangaroo population dynamics, upon which kangaroo management also depends, relies on the monitoring of population fluctuations.

Monitoring of macropod populations has been reviewed extensively by Southwell (1989). Since this review, and partly as a consequence, development of monitoring techniques has proceeded on a number of fronts. Monitoring techniques can be classified as either direct or indirect. Direct monitoring entails counting actual animals whereas indirect monitoring involves inferring population size and trend from counts of animal signs (such as scats or tracks) or from information on animals taken by shooters. The latter has been developed extensively for fisheries management (Hilborn and Walters 1992) and for harvests of some ungulates (e.g. white-tailed deer, Odocoileus virginianus; Roseberry and Woolf 1991). An important distinction needs to be made between absolute density, which is an estimate of the actual numbers in a population, and relative density, which is an index of population size. Relative density is generally adequate for monitoring population trends, but absolute density will be required to set a quota as a proportion of population size (Caughley and Sinclair 1994).

Quota derivation

Prescription of harvest quotas for an MSY is not possible. What is relevant here is not whether to harvest at the MSY, but whether the MSY can actually be identified. Caughley's (1987a) model could not pin down a single MSY even after several 100-year simulations (see Figure 2). A stable MSY was denied by environmental stochasticity, primarily variable rainfall. Added to this are the effects of a selective harvest and, more generally, the variable behaviour of shooters. By harvesting males preferentially, shooters can increase rp and therefore the MSY. However, the extent of this male bias will vary with population density and composition, and a number of environmental factors including habitat and weather (see below). Furthermore, shooter behaviour will vary depending upon whether they are the sole harvester of a population (and essentially have property rights) or whether the population is treated as a common resource (see above).

Nevertheless, wildlife managers can set a range of harvest quotas which, under a range of harvest conditions, will produce a sustained yield somewhere below the maximum. This is certainly the safe option. Any attempt to implement the MSY is fraught with the danger of overharvesting (see above). The corollary is that any optimisation should be approached cautiously.

Caughley (1987a) suggested that an instantaneous harvest rate of 10-15% per year represented the MSY for red kangaroo populations subjected to an unselective harvest (see Figure 2). The MSY for grey kangaroos and wallaroos would be slightly lower. Current quotas for each of the Australian States allowing a commercial harvest of kangaroos are generally higher, at about 15-20% (see Figures 16, 17, 18 and 19). Although not stated explicitly, this implies either that a selective harvest is allowing such an increase, that the quota represents a sustained yield below the MSY, or that the population estimates are conservative.

Whatever percentage of the population is considered appropriate to harvest, some estimate of population size is usually required. In practice, quotas have been derived from a range of sources, depending upon the Australian state. Since the late 1970s, they have been based upon population estimates obtained primarily from aerial surveys, with some consideration being given to factors such as overall population trends, climatic conditions (mainly rainfall) and trends in various harvest statistics. These harvest statistics include carcass weight, sex ratio, skin size and the size of the overall offtake. Since the quotas are usually set for a calendar year from an aerial survey undertaken in the previous year, managers are projecting the population a short way into the future. Therefore, given certain seasonal conditions, plus the results of additional surveys and monitoring of the harvest throughout the year, quotas may be modified. As a result, managers may place restrictions on the harvest such as closing certain areas down or placing weight and size limits on the animals entering the industry.

Quotas are administered differently amongst the States. In Queensland and Western Australia, quotas are not set regionally. Relative to the other States, this allows some areas to be harvested at much higher rates than others. In New South Wales and South Australia, regional quotas are set, but not necessarily at the same harvest rate. This has resulted in quite different systems of issuing tags to shooters (see Chapter 5).. There are several reasons for these differences in quota allocation amongst the States. Regional quotas are harder to administer with increased administrative costs and also require regional population estimates.

Quotas are administered differently amongst the States. In Queensland and Western Australia, quotas are not set regionally. This allows some areas to be harvested at much higher rates than others. In New South Wales and South Australia, regional quotas are set, but not necessarily at the same harvest rate. This has resulted in quite different systems of issuing tags to shootersThere are several reasons for these differences in quota allocation amongst the States. Regional quotas are harder to administer with increased administrative costs and also require regional population estimates. Nevertheless, all States are being encouraged to pursue a regional approach to management, with the goals of kangaroo management potentially differing across regions.

Recently, there has been a push largely from the grazing community and associated government departments to reduce kangaroo numbers in order to reduce 'total grazing pressure'(see Chapter 7), resulting in a rise in quotas as a percentage of population size (Figure 15f). Clearly, if quotas are maintained above the MSY and are met, such a move will compromise the goal of harvesting solely for a sustained yield. Harvesting solely for damage mitigation requires no quota, rather an assessment of damage. Treating kangaroos solely as pests without regard to their conservation and value as a resource would probably be publicly unacceptable (see Chapters 7, 8 and 9). At the very least there would need to be a quota providing a safety net, and population monitoring to ensure that it worked.

Direct monitoring

Unfortunately, the population density of most macropods is difficult to assess from direct counts. Many are small, cryptic species which become active only at night. Many species inhabit the thickly vegetated and hilly fringes of the continent, areas which include parts of the ranges of the commercially harvested species (see Chapter 3).

Luckily, the large commercial species occur in more open, relatively flat rangeland habitat and are still active during daylight hours near sunrise and sunset (crepuscular). This makes them amenable not only to ground survey, but also aerial counts which can be performed relatively cheaply over a large area.

Aerial surveys

Strip transect sampling with fixed-wing aircraft has become an established method for the broad scale monitoring of kangaroo populations (Caughley et al. 1976; Caughley and Grigg 1981). As shown in Figure 11, a fixed-wing aircraft (usually a Cessna 182) is flown at a ground speed of 185 km h-1 (100 kts), 76 m (250 ft) above the ground. Two observers occupying the rear seats count kangaroos seen on either side of the aircraft in strips that are 200 m wide on the ground. These strips are delineated by streamers or fibreglass rods trailing parallel to the fuselage and attached to the wing struts on either side of the aircraft. Observers count in 97 s units with a 7 s break between them, giving a sampling unit of 1 km2 (Caughley and Grigg 1981). Sightings can be recorded onto data sheets during the break or recorded directly into micro- cassette recorders. Global positioning receivers and radar altimeters help maintain constant height and speed, and permit precise navigation over often featureless landscapes. Surveys are conducted in winter in sunny conditions within three hours after sunrise or before sunset.

diagram of strip transect sampling from fixed-wing aircraft

Figure 11. Strip transect sampling from fixed-wing aircraft (after Poole 1984).

Trained observers are needed for survey work. It usually takes a new observer about 50 h to get used to the discomfort, maintain a steady high level of concentration and sort and store separately the images an observer identifies as red kangaroos, grey kangaroos or wallaroos. An observer in training sits on the same side of the aircraft as an experienced observer, with both scanning the same strip. At the beginning, the novice may see <50% of the trained observer's score. This gradually improves over many sessions and the scores eventually stabilise. Luckily, most people seem to show similar skills, but a separate correction factor can be calculated for any observer whose performance is consistently different from the rest. Not surprisingly some people prove unsuitable.

Parallel transects are generally flown east-west (reducing problems with the sun) across survey areas (Figure 12). In a typical survey session, about 200 km2 can be scanned. For broad scale surveys (> 100,000 km2), sampling intensities of <2% generally provide adequate precision for kangaroo populations in the sheep rangelands (Caughley 1979). For smaller blocks, such as a 150 km2 grazing property, sampling intensities >10% may be required.

It has long been recognised that sample counts obtained using this method require adjustment by correction factors to ensure repeatability, for monitoring population trends, and accuracy (or bias), and for determining absolute population size (Pollock and Kendall 1987). Controlling for such factors as survey strip width and height above the ground (Caughley et al. 1976), temperature (Bayliss and Giles 1985; Caughley 1989), time of day (Hill et al. 1985; Short and Hone 1988), cloud cover (Short and Bayliss 1985; Caughley 1989) and season (Caughley 1989) ensures constancy of this bias and therefore the repeatability of the method. Here, repeatability is distinguished from precision or sampling error (Southwell 1989). Attempting to control for the influence of vegetation density and canopy cover on the sight-ability of kangaroos through the use of habitat correction factors (Caughley et al. 1976; Short and Bayliss 1985; Bayliss 1987; Southwell et al. 1986; Short and Hone 1988; Grice et al. 1990) further ensures repeatability, by accounting for altered habitat associations of animals between surveys, and allows the conversion of aerial counts to absolute densities.

Habitat correction factors have been developed using a number of methods including an indirect regression technique, the index- manipulation-index method, double counting and from the comparison of aerial counts of kangaroos with corresponding ground counts (Southwell 1989; Pople et al. 1993). The latter has received the most attention and appears the most reliable and generally applicable. Habitat correction factors derived from a number of studies are shown in Table 3. Until recently, most States have used the correction factors determined by Caughley et al. (1976), partly to maintain comparable estimates from year-to-year and to ensure conservative management. For both species of grey kangaroos and common wallaroos, these are clearly underestimates. Recent re-assessments of correction factors have seen the use of revised correction factors in most States (see Figures 16, 17, 18 and 19).

Aerial surveys of kangaroos using helicopters have been conducted in a number of areas using strip transect sampling (e.g. Choquenot 1991, 1995). Helicopters enable surveys to be conducted in more rugged terrain than with fixed-wing aircraft. Wallaroos in the Barrier, Mt Darling and Coonabaralba Ranges of western New South Wales have been surveyed by NSW NPWS using helicopters since 1993. Helicopters also provide greater visibility, partly through slower speeds and partly because doors can be removed. A combination of helicopters and line transect methodology has been used for surveying kangaroos in Queensland (Figure 12).

map of main areas covered by aerial survey

Figure 12. Main areas covered by aerial survey using fixed-wing aircraft (shaded) and helicopters (solid), and by ground surveys (stippled). Additional helicopter and ground surveys are conducted over smaller areas in each state.

Line transect sampling involves determining the distances of animals from a transect line. The further animals are away from the transect line, the more likely they are to be missed. This decline in sight-ability or the detection function can be modelled, allowing density to be calculated. If there is no decline in sightability, the count is equivalent to a strip transect. Detection functions will differ between species, habitats and time of day. The advantage of the technique is that correction for visibility bias is survey specific. However, to estimate density accurately by this method, a number of assumptions need to be satisfied. In order of importance, they are:

  1. Animals on the transect line are always seen
  2. Animals do not move before being counted, and none are counted twice
  3. Distances and angles are measured without error (or, for aerial counts, sightings are placed into correct distance classes)
  4. Sightings are independent events.

Sample sizes of >60 animals, which are rarely a problem for aerial counts, are required to adequately model sight-ability. Buckland et al. (1993) provide a full discussion of the technique and provide a computer program DISTANCE (Laake et al. 1993) for modelling sightability and estimating density. Examples of line transect counts of kangaroos collected on aerial and ground surveys are shown in Figure 13.

sighting histogram

Figure 13a. Sighting histograms for line transect surveys of red kangaroos, eastern grey kangaroos and common wallaroos conducted by helicopters. All surveys were conducted in the Blackall region of central-western Queensland (QDoE, unpublished data).

sighting histogram

Figure 13b. Sighting histograms for line transect survey of red kangaroos, conducted on foot (arrow indicates truncation point used in analysis) in the Blackall region of central-western Queensland (after Clancy et al. in press; QDoE, unpublished data).

In Queensland, the survey method involves a helicopter (usually a Bell 47 KH4) with the doors removed being flown at a ground speed of 93 km h-1 (50 kts), 61 m (200 ft) above the ground. Two observers occupying the rear seats count kangaroos seen on either side of the aircraft. The sightings of kangaroos are placed into 25 m distance classes up to 125 m perpendicular to the transect line measured from directly below the observer. Sightings are recorded into micro-cassette recorders. The distance classes are delineated on aluminium poles extending perpendicularly from either side of the helicopter. Observers count in five minute units with a 30 s break between them. A full description of this method is given in Clancy et al. (1997).

Results obtained using this technique with helicopters have been found to compare favourably with those of walked line transect surveys (Clancy et al. 1997), which are considered accurate (see below). The exception is for common wallaroos, where density estimates are still negatively biased by a factor of 2-3 (Clancy et al. 1997). However, this bias appears to be relatively consistent. Helicopter line transect surveys would therefore appear to be an accurate and repeatable method for kangaroo surveys. However, because of the high running costs of helicopters, it is unlikely that they will ever replace fixed-wing aircraft in broad scale surveys. There is a further reluctance to change because there are now long runs of data (almost 20 years in some cases) collected along fixed transect lines using a standard method. Helicopter line transect surveys can, nevertheless, provide estimates of kangaroo population size that can be used to evaluate those obtained from strip transect surveys conducted on a relatively broad scale with fixed-wing aircraft.

The use of line transect methodology in surveys of kangaroos by fixed-wing aircraft has been examined (Southwell 1993; Pople et al. 1998) and, while reducing bias, has generally proven less successful than with helicopters. This is likely to be the result of not all animals on the transect line being seen with certainty. Ultralight aircraft have proven more successful, providing density estimates of kangaroos that are similar to those obtained from helicopter surveys and walked line transects (Grigg et al. 1997). Ultralights have the advantage over helicopters through being cheaper to operate, and may find application for surveys on individual properties.

At worst, fixed-wing aerial surveys have led to underestimates of population size and an inability to track short-term fluctuations in numbers. In terms of setting harvest quotas, which are based mainly upon population estimates, this would result in conservative management practice. Higher correction factors could be applied to reduce bias. Further comparisons of helicopter and fixed-wing estimates will hopefully show the full extent of any disparities in the survey methods and the ability of temperature correction factors and broadly-based habitat correction factors to account for them.

Pople et al. (1998) have suggested that an alternative to standard 200 m strip transects in fixed-wing aircraft surveys would be to use 100 m strip transects. These have the advantage over 200 m strips of improved visibility, leading to smaller correction factors and a reduction in random errors. They would be particularly attractive if they returned more repeatable estimates than do 200 m strips. However, this would need to be assessed in relation to continuing long runs of data using the standard 200 m strip transect. Future surveys will hopefully clarify this as well.

Ground surveys

Direct counts of kangaroos can be made from vehicles or on foot. The use of roads is likely to lead to biased estimates as roads are rarely random samples of an area (Caughley and Sinclair 1994). This has been demonstrated in surveys of whiptail wallabies by Southwell and Fletcher (1990). In other situations, the bias is less pronounced and, if this bias is consistent, vehicle surveys can still be used to track population size.

Walked line transect surveys of populations of known size have been found to yield accurate population estimates of kangaroos (Southwell 1989, 1994). The method usually involves one or two observers walking compass bearings and recording angles (with a sighting compass) and distances (with a rangefinder) to all animals seen.

Direct ground counts of kangaroos are commonly used for small populations, forming the basis of many studies on the habitat use, population dynamics and social organisation of large macropods. However, ground counts are generally not logistically feasible for broad scale monitoring. An exception has been surveys by Southwell et al. (1995a, 1995b, 1997), who provided population estimates for large macropods over an area of 350,000 km2 in the eastern highlands of Australia (Figure 12). These surveys have provided the basis for setting quotas in New South Wales and have given confidence to the sustainability of whiptail wallaby harvest in Queensland. Nevertheless, these surveys took five years to complete and have not been repeated.

Most state management programs use ground counts in small, key areas, as both a check on aerial survey estimates and a way of estimating kangaroo density in habitat unsuitable for aerial surveys. Ground counts can also provide an estimate of the age structure and sex ratio of populations (Clancy and Croft 1992; Moss 1995; Pople 1996; Shelly 1997). This can help explain recent population trends, make projections about future population trends, and interpret harvest data.

Indirect monitoring

Characteristics of the actual harvest (harvest parameters), such as sex ratio and average weight, could provide a cost-effective means of monitoring kangaroo populations. Furthermore, such indirect monitoring may provide a more extensive coverage of harvest areas than direct monitoring methods such as aerial survey, offering potentially continuous monitoring of a population while it is being harvested. Harvest data are routinely collected by management authorities in all Australian States where kangaroos are commercially harvested. In the larger States of Western Australia and Queensland, because aerial surveys are expensive and logistically difficult, harvest data have been used in population monitoring (Anon. 1984; Poole 1984; Southwell 1989). In Tasmania, direct monitoring of commercially harvested wallaby populations is difficult because of the habitat and the behaviour of the animals, forcing managers to rely heavily on the analysis of harvest statistics (Anon. 1984). However, as pointed out by Southwell (1989), there has been no empirical investigation to validate the use of harvest parameters as a monitoring tool.

Prior to 1991, there had been a great reluctance by what is currently QDoE to use aerial survey methods to monitor populations. Kirkpatrick and Nance (1985) considered estimates derived from fixed-wing aircraft to be negatively biased, thereby overestimating harvest rate, and imprecise, leading to misleading population trends. QDoE therefore relied on indirect monitoring methods to determine the size of annual quotas (Anon. 1984, 1989)

Various harvest parameters are collected routinely by management authorities, including sex ratio, average size and weight of harvested animals, as well as catch-effort data such as animals shot per unit time. In turn, these data may provide information on population abundance, composition (i.e. sex ratio and age structure) or status (i.e. under- or overharvest). Kangaroos are markedly sexually dimorphic in size and exhibit continued growth (particularly for males) well into adulthood (Jarman 1989). Because shooters are paid by carcass weight or skin size, they strongly select large animals and therefore males. Harvest composition in terms of average size, average weight and sex ratio therefore offers a potentially useful index of kangaroo population parameters.

The advantages and disadvantages of monitoring kangaroo populations indirectly have been discussed by Southwell (1989). For harvest parameters to be an effective index of any population parameters, a number of assumptions must be satisfied. Harvesting equipment, efficiency and conditions (e.g. weather, access to animals and market prices) must all be standardised. In addition, it is assumed that the harvesting of each animal is independent and that animals do not learn to avoid kangaroo shooters.

Southwell (1989) identified several potential violations of these assumptions. The relationships between harvest parameters and population abundance are unlikely to be linear, because of time saturation (Caughley 1977), whereby harvest parameters eventually reach a plateau at some level of population size. Harvesting of each animal is therefore not independent. Changes in abundance in this zone of saturation will not be reflected in the harvest parameters. Time saturation would be unimportant if population status or composition was being monitored (Southwell 1989). Furthermore, shooter success will vary with weather conditions. It is generally accepted that shooting is difficult in wet conditions when the ground is boggy (Prince 1984a) or under windy conditions when animals seek cover and are restless, making them difficult targets. Shooting is considered easier during dry times when access is easier, ground cover is low and kangaroos tend to concentrate at watering points and on remaining food patches. Other potentially confounding factors are the effect of the environment, particularly drought, on population composition, market activity and economics.

QDoE have used a range of harvest parameters to monitor the status of harvested kangaroo populations. This was supplemented by unbiased samples of animals from harvested and unharvested populations and a simulation model relating various harvest regimes to population trends and harvest offtake (Kirkpatrick and Nance 1985).

An age-structured model of an eastern grey kangaroo population was used by Nance (1985) to identify potential indirect monitoring methods. Nance argued that harvest rate and any overharvesting would be reflected in the sex ratio of the harvest. The model assumed juvenile survival rate was density dependent and that harvest mortality was fully compensated by a reduction in this natural mortality at sustainable harvest levels. The harvest was assumed to be strongly selective, targeting males and older animals. The model was criticised by de la Mare (1988) who proposed an alternative model based on size-selective harvesting. Harvest selectivity may be strong if only carcasses are taken, or weak if skin-only shooting is involved. In the latter case, harvest sex ratio will not reflect an exploited population's plight. Neither model operated in a stochastic environment and the vital rates were constant over time for an unharvested population. As pointed out by Cairns (1989), both models need to be validated and presumably incorporate the effects of a variable environment to be accepted as management tools.

Carcass weight, skin size and sex ratio (% male) were identified by Pople (1996) as potential indicators of the harvest rate of eastern grey kangaroo populations. However, the harvest rate of red kangaroo populations was only reflected in skin sizes. By monitoring population composition, which changes in spite of harvesting, Pople (1996) was able to identify the basis for some of these relationships.

Trends in population size and harvest offtake

Commercial harvesting of kangaroos is restricted to commercial zones in each state (Figure 14). Within these zones are conservation reserves, where commercial harvesting is not permitted (Chapter 12). There is direct monitoring of the kangaroo populations over much of this commercial zone (Figure 12), being conducted largely by annual aerial surveys. In each state, these surveys are supplemented with ground surveys and indirect monitoring. Full details of survey and harvest areas, harvest restrictions and various monitoring methods can be found in the relevant management programs for each state.

map showing where kangaroos are commercially harvested

Figure 14. Main area or zone in Australia where kangaroos are commercially harvested. Relatively small numbers are harvested outside this zone including the Northern Territory, northern South Australia, far western Queensland and occasionally Victoria.

Figures 15, 16, 17, 18 and19 describe the trends in harvest offtake, quotas and population size for commercially harvested kangaroo species in each state since 1975. Explanatory notes for each figure briefly describe the survey methods used and how data were compiled. To help interpretation of these data, annual rainfall for 1970-96 within the commercial zone of each state (averaged across four recording stations within each commercial zone) are shown in Figure 20. This is presented as standardised rainfall, allowing rainfall to be averaged across regions and compared between States. The data indicate years of above and below average rainfall. This is an oversimplification of the influence of rainfall on kangaroo dynamics. The effect of rainfall will depend on its timing and there will be differences between regions and kangaroo species (Cairns and Grigg 1993). Averaging what is invariably patchy rainfall over such large areas is a further simplification.

Figure 15a suggests that Australia's kangaroo population that is exposed to harvesting has fluctuated between 15 and 35 million animals. Eastern grey kangaroos have been the most abundant species, although by 1996 the numbers of red kangaroos were roughly similar (Figure 15b). Additional populations occur outside this area and estimates of these were given by Caughley et al. (1983). The most significant in terms of numbers are eastern grey kangaroos in the eastern highlands. Southwell et al. (1997) covered approximately 40% of this area, and their estimate has been included in Figures 15a and 15b. Southwell et al. (1997) suggested a further 4 million eastern grey kangaroos occurred in the remaining area at the time of their surveys (1987-92).

An interpretation of the trends in Australia's kangaroo population up until 1987 was given by Fletcher et al. (1990). Populations in eastern Australia increased through the late 1970s following years of above average rainfall. Populations declined following the 1982- 3 drought, but then recovered to predrought levels by 1990. There was a decline through the early 1990s which was most pronounced in the eastern grey kangaroo population. In Western Australia, kangaroo populations recovered from a drought in the late 1970s, increasing to a peak around 1990 following above average rainfall. Populations have since declined with years of both good and, probably more importantly, poor rainfall. While state kangaroo populations have fluctuated markedly, their net result has been a national population that has been relatively stable.

Over the past 22 years the national harvest of each species has been increasing (Figure 15c). Ignoring state boundaries, red kangaroos and eastern grey kangaroos have been harvested in similar numbers over the past 22 years and dominated the harvest (Figure 15c). On average, 58% (range 43-68%) of the annual harvest in Australia has been taken in Queensland during this time (Figure 15d). The harvest of red kangaroos in Queensland is unique among the state macropod harvests in that each year since the mid-1980s the quota has been met and the season closed (Figure 16). Harvest rates have generally been highest for red kangaroos and lowest for western greys (Figure 15e). Harvest rates have been increasing, but then so has the quota as a proportion of the population (Figure 15f). As a result the proportion of the quota taken each year has fluctuated around 80% (Figures 15gand 15h). Inadvertently, the pattern of harvesting kangaroos more heavily during drought approaches Stocker and Walters' (1984) optimal strategy, which was based upon Caughley's (1976) interactive model (see Chapter 2).