Orange Agricultural Institute
Forest Road, Orange NSW 2800
Phone (02) 6391 3834, Fax (02) 6391 3972
Current management policies allow approximately 15% of rangeland kangaroo populations to be harvested annually in order to mitigate damage to agricultural enterprises. One of several management objectives is to maintain populations throughout their entire natural distribution. The intensity of commercial harvesting is known to be unevenly distributed within management regions, how this impacts on achieving management objectives remains unknown. This paper describes a spatial model constructed to simulate heterogeneity of harvesting intensity within harvested properties. Due to substantial differences between harvesters with respect to search pattern and animal selection the model does not mimic any particular harvester but operates in an optimal manner. Using knowledge of the distribution of kangaroos across a property (derived from field surveys), combined with travel time and search efficiency, the simulated harvester identifies and harvests patches where net yield ($/minute) is maximised. The optimised harvest continues in an iterative manner until the population is reduced to a threshold density below which it is no longer economically viable to continue harvesting. The undulating nature of the resulting threshold density grid illustrates the effects of spatial variation in harvesting intensity due to differences in accessibility, kangaroo density and search efficiency.
It is widely believed that the abundance of kangaroos has increased in pastoral areas since European settlement (Frith, 1964; Newsome, 1975). Current management objectives seek to reduce the impact of kangaroos on agricultural productivity whilst maintaining kangaroos throughout their natural distribution. In New South Wales, Queensland and South Australia kangaroos may be culled for either damage mitigation or commercial utilisation, with the later category comprising the bulk of animals taken. A quota which is set at a fixed percentage of the previous year’s population governs the number of animals that may be taken each year. On the basis of computer simulation, Caughley (1987) suggested that a continued commercial harvest of 10-15 percent of the kangaroo population would reduce the population to 60-70 percent of its unharvested density. However over twenty years to 1994 the commercial harvest has only once achieved the quota, more frequently averaging 65 percent of the annual quota (Eveleigh, 1995). Pastoralists, conservation groups and wildlife management agencies are now questioning if current kangaroo management, via commercial harvesting, has the capacity to deliver the reduction in kangaroo grazing pressure required to achieve agricultural and conservation objectives. The current project aims to evaluate the impacts of the existing commercial harvest on kangaroo populations as well as investigate alternative kangaroo management strategies. Given that both kangaroo density and harvesting pressure vary across any harvested area a model was derived to explicitly incorporate these sources of spatial heterogeneity. The parameters on which the model is based were derived empirically from four commercially harvested field sites. This paper describes the structure of the spatial model constructed and gives an example of the predicted level of reduction in kangaroo density across one of the commercially harvested properties.
The spatial modelling described below was performed with ArcView 3.2 using the Spatial Analyst 2.0 extension.
Analysis of four commercial harvesters revealed significant variation between the strategies adopted during harvesting forays. The choice of search area and minimum body size was determined primarily by commodity type, commodity price, results from previous forays and ‘gut feeling’ amongst other factors. Different harvesting strategies reflect variations in the efficiency with which harvesters remove animals from the population. An optimal harvesting strategy was incorporated into the model so as to predict the maximum reduction in kangaroo density that is economically feasible under current harvesting constraints.
Optimal harvesting may be likened to a predator-prey relationship whereby the predator (harvester) seeks to maximise his intake of prey (kangaroos) while minimising his search effort. Optimal harvesting requires that the harvester have perfect knowledge of the property layout and the distribution of kangaroos so that the area of highest profitability (net yield) may be identified and exploited.
Calculation of Net Yield
Net yield is defined as the net financial return, or profit, that the harvester receives from harvesting. It is determined by the efficiency with which the harvester can acquire carcases after subtracting operating and processing costs. Figure 1 provides a schematic representation of the relationship between parameters and how they are combined to calculate net yield.
Yield was calculated by multiplying the quantity of product acquired (ie. meat or skins) by the commodity price. The commodity price is determined by market demand for kangaroo products and thus fluctuates over time. However it does not vary spatially and is therefore treated as a constant within each simulation. The rate at which animals are harvested and the actual time available for harvesting determine the quantity of product acquired. Both of these components vary spatially.
Harvest rate is the speed at which the harvester procures kangaroo skin or meat. A relationship between skin yield and carcase weight was derived. This allowed harvest rate to be calculated in kilograms per minute for both skin and meat forays.
Location time reflects the time taken to find an animal to harvest; it is a function of the harvester’s search rate and the density of animals in the harvested patch. Both search rate and density were measured empirically for each landsystem within the study site. A landsystem map was rasterised into an array of 0.25Ha cells to produce a landsystem grid suitable for analysis within ArcView-Spatial Analyst. The search rate and density data was then combined with the landsystem grid to produce an input grid for the model that identified the location time for all 0.25Ha cells across the site. After locating an animal the harvester proceeds to acquire (shoot, hang and bleed) and process the carcase. Acquisition and processing times were recorded from forays with commercial harvesters and were averaged per harvester per commodity type. The acquisition and processing constants were then added to the location time grid and the resulting grid was multiplied by mean kangaroo weight to produce the harvest rate grid.
Actual harvest time was calculated by reducing the total time available for harvesting by the time taken to travel to a harvesting patch. Global positioning systems were placed within the harvesters’ vehicles when they were harvesting the field sites. This allowed the speed of movement to be recorded whilst traversing a number of tracks as well as off-track driving within each of the harvested landsystems. This data was summarised and then combined with maps containing fences and other physical barriers to construct an impedence grid covering the site. This grid identified the time taken to traverse each 0.25Ha cell.
The harvester always commenced his foray at the entrance to the property so this was used as the source when calculating travel time. The impedence grid was combined with the source location and analysed with the CostDistance function in ArcView to produce an accumulated time grid. This grid summarised the time taken to travel from the source to any location within the site. Values in the accumulated time grid were subtracted from the total available time to produce the actual harvest time grid (figure 1).
The vehicle operating cost comprises the registration, insurance, depreciation and running costs for a commercial 4WD fitted out for kangaroo harvesting. Switala’s (1997) estimate of annual running costs was converted to dollars per minute so that it could be subtracted from yield as part of the net yield calculation. Vehicle operating cost is incorporated in the model as a constant with no spatial variation.
The processing cost is determined by the number of carcases harvested and the cost of processing each carcase. The latter includes the cost incurred during the dispatch and processing of a kangaroo. It is calculated by averaging the cost of rifle, ammunition, royalty tags, knives and sundries on a per head basis. Once calculated, the processing cost does not vary temporally or spatially within the model and is thus included as a constant.
The number of carcases harvested (kangaroos/minute) was calculated by dividing the harvest rate grid (kg/minute), described above, by the mean carcase weight (kg/roo), which was derived empirically from field sampling. The processing cost grid ($/minute) was calculated simply as the product of the number of kangaroos harvested (kangaroos/minute) and the carcase cost ($/kangaroo).
Finally, the net yield grid was calculated by subtracting the vehicle operating cost and the processing cost grid from the yield grid. The undulating nature of the resulting net yield grid can be attributed to spatial variation in accessibility, kangaroo density and search efficiency.
Calculating threshold density
Under optimal harvesting areas of maximum net yield are harvested until the resulting reduction in density causes other areas to become more lucrative. The focus of harvesting is then moved to these new areas until they too are reduced to a lower level of profitability. The cycle continues in an iterative manner until the reduction in density due to harvesting causes the net yield to drop to $0/minute across all areas of the site that were initially profitable. At this point the costs of harvesting equal the yield and the harvester can extract no further financial gain from harvesting. The resulting density grid illustrates the maximum reduction in density that is economically feasible under the current commercial harvesting regime. It’s this threshold density grid that is of interest.
Instead of calculating threshold density in an iterative manner, as outlined above, it can be calculated more quickly by rearranging the equation used to calculate net yield. The equation is reconfigured to isolate density as the output variable. Substituting a net yield value of zero into the reconfigured equation produces the same threshold density grid described above.
The results presented below comprise outputs of the model for one of the four field sites.
As the name suggests, commercial harvesting implies that the harvester derive a profit from his labour. The level of profit required by a harvester determines the density at which he will cease operating in a given area. Figure 2a indicates that only the most accessible areas of the property would be harvested if the harvesters’ profitability threshold was set at $0.85 per minute. As the profitability level is relaxed, more distant areas become viable and the density in the most accessible areas is reduced further (figure 2b and 2c).
For this particular study site, the harvester stipulated that he would continue harvesting only while he was able to achieve a net profit of $120 during a six hour foray. This equates to a net yield of $0.30/minute (= $120 / 360 minutes). Figure 2c illustrates the density to which kangaroos would be reduced if the harvester ceased operating at the $0.30/minute breakpoint. This result highlights a marked difference in harvested density between landsystem types. The lighter regions are landsystems that are relatively open and therefore easily traversable. The harvester’s search rate is high in these areas, which promotes an increased harvest rate and net yield. This equates to high harvest intensity and heavy population reduction. Conversely, the darker patches comprise landsystems dominated by gidgee woodlands and turpentine shrublands, which impede the harvester’s progress. At the $0.30/minute level of profitability harvesting of these landsystems is prohibitive except along cleared tracks where search rate is higher. This is noticeable as striations in the harvested density of some heavily timbered landsystems (figure 2d and 2e).
Continuing the sequence, figure 2d shows the reduction in density likely to be achieved if the harvester were to relax his profitability level below his stated $0.30/minute threshold. Here the open landsystems have been harvested to the point that it becomes equally profitable for the harvester to venture into previously unharvested areas within the timbered landsystems. At net yield = 0 (figure 2e) the harvester derives no profit from any further harvesting. Given the current economic constraints it is unlikely that commercial harvesting would attain this level of reduction unless the harvest was subsidised from outside sources.
The results presented above illustrate how variation in the level of profitability accepted by the harvester effects the density to which kangaroos can be reduced. Amongst the suite of input parameters there are several other variables that significantly effect the model outputs. Table 1 lists the percentage deviation in net yield as a result of halving each of the input variables individually. It should be noted that halving some variables caused net yield to increase while others caused it to decrease. However, for simplicity only the absolute percentage deviation is listed in Table 1.
Not surprisingly, commodity price is the single most important variable affecting the net yield (profitability) of the harvester. Interestingly, acquisition time is the second most important variable, this is due to the magnitude of this parameter. Total harvest duration can be broken down into time spent travelling, searching, acquiring and processing. Table 1 indicates that acquisition time is the largest of these four components and therefore a proportional reduction in the time taken to acquire carcases will increase profitability more so than proportional reductions in search, travel or processing time.
Table 1 Sensitivity of Net yield to variation (50% reduction) in parameter values
Mean size function (slope)
Mean size function (y int.)
Vehicle operating cost
The density grids produced by the model at current economic thresholds will be analysed with respect to the objectives of kangaroo management put forward by stakeholder groups. Pastoralists aim for kangaroo density to be reduced below a target level through commercial harvesting, whilst conservationists seek to establish non-harvest areas or refugia. Government wildlife management agencies aim to reduce kangaroo density whilst maintaining populations throughout their natural distribution. Evaluation of the model outputs will indicate the extent to which commercial harvesting achieves each of these objectives at the property scale.
Upon completion of the property scale model it is planned to apply the same methodology to a regional scale. The regional model will address the same issue of spatial variation in harvest intensity on a broader scale. A requirement of kangaroo harvesting for meat production is that all carcases be returned to a central point (chiller) at the completion of each foray. This imposes a substantial limitation on the spatial distribution of harvesting intensity. The proposed regional model will be calibrated to harvester return data from a sample of chillers so that the results may be extrapolated to other chiller locations throughout the western region of New South Wales. The regional model outputs will be combined with knowledge of the distribution of all chillers across the state to allow harvesting intensity and the resultant population reduction to be mapped for the entire commercial harvesting area. Thus allowing the kangaroo management objectives listed above to be assessed on a statewide scale.
The model presented in this paper improves our ability to manage harvested populations. Commercial harvesting of any species is constrained by economic factors that are variable spatially and temporally. By explicitly combining spatial data summarising availability and distribution of resources with current economic parameters, this optimisation model allows harvest intensity and the resultant density reduction to be calculated spatially. Threshold density maps produced by this technique facilitate objective analysis of current commercial harvesting regimes and therefore promote improved management of harvested species.
Thanks to Steve McLeod for his significant contribution to the formulation of the model. Thanks also to all who assisted with field data collection. NSW Agriculture and the Murray Darling Basin Commission funded this research.
Caughley, G. (1987). Ecological relationships. In Kangaroos: their ecology and management in the sheep rangelands of Australia (Caughley, N. Shepherd and J. Short, eds.). Cambridge University Press, Cambridge. pp. 159-87
Eveleigh, J. (1995). The current New South Wales kangaroo management program. G. In Conservation through Sustainable use of Wildlife (Grigg, P. Hale and D. Lunney, eds.). Centre for Conservation Biology, The University of Queensland, Brisbane. pp. 186-8.
Frith, H.J. (1964). Mobility of the red kangaroo, Megalia rufa. CSIRO Wildlife Research, 9, 1-19.
Newsome, A.E. (1975). An ecological comparison of the two arid-zone kangaroos of Australia and their anomalous prosperity since the introduction of ruminant stock to their environment. The Quarterly Review of Biology, 50, 389-424.
Switala, J. (1997). Economics of kangaroo farming in the mulgalands of Queensland. In Farming in the information age. Proceedings of the Australian Farming Management Society’s 23rd National Conference. University of Southern Queensland, Toowoomba.
Figure 1 Stages in calculation of net yield showing relationship between input grids and variables
Figure 2 Reduction in kangaroo density as the result of commercial harvesting constrained to profitability thresholds of a)$0.85, b)$0.75 and c)$0.45 per minute
Figure 2 Cont: Reduction in kangaroo density as the result of commercial harvesting constrained to profitability thresholds of d)$0.30, e)$0.15 and f)$0.00 per minute