Bias and precision of predicted densities of overabundant kangaroo populations

Abstract

Validated predictions of wildlife population density would be very useful for managers of overabundant wildlife and their effects on biodiversity. In this study such predictions were generated by modelling population dynamics of 18 non-culled populations of Eastern Grey Kangaroos (Macropus giganteus) across small biodiversity conservation reserves in Canberra. Predictions were validated using three analyses of independent, out-of-sample, data from 11 populations which were non-culled or culled. Association (analysis 1) showed observed and predicted densities were significantly positively correlated (R² = 0.79, P = 0.045, n = 5) with unbiased slope and y intercept for non-culled populations, though observed and predicted densities of culled populations were unrelated (R² = 0.32, P = 0.24, n = 6). Coverage (analysis 2) showed predicted densities were within the 95% confidence interval of observed densities in five of five non-culled populations and four of six culled populations, with one underestimate and one overestimate in the latter group. Bias (analysis 3) showed the mean bias (=observed – predicted) was 0.18 (±0.23 SE) kangaroos/ha for non-culled and 0.08 (±0.23 SE) for culled populations. The results have been used to adjust kangaroo management approaches in Canberra as part of adaptive management.