Estimating disease survey intensity and wildlife population size from the density of survey devices: Leg-hold traps and the brushtail possum

摘要

Wildlife disease surveillance requires accurate information on the proportion of managed populations sampled or their population density, parameters that are typically expensive to measure. However, these parameters can be estimated using spatially explicit modelling of capture probabilities, based on the distribution and deployment times of capture devices, given accurate information on the relationships between these variables. This approach is used in New Zealand's surveillance programme aimed at confirming areas free of bovine tuberculosis (bTB(1)) in brushtail possums (Trichosurus vulpecula). However, there is uncertainty about the accuracy of the underpinning parameters characterizing possum trappability (g), given the distance between where a trap is placed and the possum home range centre. Sampling intensity (SI: the percentage of the population sampled during a population survey) and sigma (sigma; 95% home range radius/2.45) were measured, using leg-hold traps deployed under a set protocol to standardize survey effort, at four sites containing previously radio- and GPS-collared individuals. Those data were used to derive an estimate of the nightly probability of capture of possums in a trap set at their home range centre (g(0)). Those estimates were compared to the standard assumptions currently used as defaults in the day-to-day approach used by bTB managers. Home-range size (and therefore a) varied widely between sites (range 3.6-49.4 ha), probably largely in response to differences in possum density. Field measured SI also varied widely between sites, and was closely positively correlated with home range size (R-2 = 0.967; P = 0.017); wide-ranging possums were more trappable than sedentary ones. We found that g(0) was inversely related to sigma, but the magnitude of increases in g o with declining a appeared to be insufficient to compensate for the fewer places at which each possum could be trapped when those home ranges were small. SI was, therefore, not constant across sites where a standard survey effort was applied. The assumed relationship between g o and a in the current spatial model may, therefore, need reassessment. The management implication of these result is that the sampling effort required to attain a target sampling intensity is dependant on the target animal density, and for bTB management of possums in New Zealand, is under-estimated by the current default parameters in a model of freedom-from-disease for higher density possum populations.