Robust estimation of snare prevalence within a tropical forest context using N-mixture models

Hunting with snares is indiscriminate and wasteful, and this practice is currently one of the gravest threats to terrestrial vertebrates in the tropics. However, as snares are difficult to detect and often dispersed widely across large, inaccessible areas it is problematic to reliably estimate their prevalence and no standard survey methods exist. Conservation managers need reliable, timely, information on the spatio-temporal patterns of hunting and on responses to interventions, and we present an innovative sampling and analysis framework that allows for the rigorous estimation of snare detectability and 'abundance', but which can be feasibly implemented in challenging field contexts. This new approach was used to undertake a large-scale systematic snare survey in Keo Seima Wildlife Sanctuary, in Eastern Cambodia, and the resulting data were analysed using a novel application of N-mixture models. A range of environmental and management factors were examined as potential determinants of snare abundance and detectability, and proximity to the Vietnamese border was shown to be overwhelmingly the most influential factor. Snares were more common in the wet season rather than the dry season, and the detection probability of snares was shown to be low (similar to 0.33), as predicted. No clear relationships between snaring levels, anti-poaching patrol effort and ungulate densities were evident from these data. There was clear evidence that certain factors, such as the percentage of dense forest cover, will exert confounding effects on both detectability and abundance, highlighting the critical need to take account of the imperfect detection when designing threat monitoring systems.