Abstract: This study presents the testing of two unsupervised classification methods for their ability to accurately identify unknown individual tigers, Panthera tigris, and snow leopards, Panthera uncia, from their footprints. A neural-network based method, the Kohonen self-organizing map (SOM), and a Bayesian method, AutoClass, were assessed using hind footprints taken from captive animals under standardized conditions. AutoClass successfully discriminated individuals of both species from their footprints. Classification accuracy was greatest for tigers, with more misclassification of individuals occurring for snow leopards. Examination of variable influence on class formations failed to identify consistently influential measurements for either species. The self-organizing map did not provide accurate classification of individuals for either species. Results were not substantially improved by altering map dimensions nor by using principal components derived from the original data. The interpretation of resulting classifications and the importance of using such techniques in the study of wild animal populations are discussed. The need for further testing in the field is highlighted.

Abstract: We examined predation activity throughout Bhutan by tiger (Panthera tigris), common leopard (Panthera pardus), snow leopard (Uncia uncia) and Himalayan black bear (Ursus thibetanus) on a variety of livestock types using data gathered over the first two years (2003-2005) of a compensation scheme for livestock losses. One thousand three hundred and seventy five kills were documented, with leopards killing significantly more livestock (70% of all kills),
than tigers (19%), bears (8%) and snow leopards (2%). About 50% of livestock killing were of cattle, and about 33% were of horses, with tigers, leopards and snow leopards killing a significantly greater proportion of horses than predicted from availability. Examination of cattle kills showed that leopards killed a significantly greater proportion of smaller prey (e.g., calves), whereas tigers killed a significantly greater proportion of larger prey (e.g., bulls). Overall, livestock predation was greatest in summer and autumn which corresponded with a peak in cropping agriculture; livestock are turned out to pasture and forest during the cropping season, and subsequently, are less well guarded than at other times. Across Bhutan, high horse density and low cattle and yak density were associated with high rates of livestock attack, but no relationship was found with forest cover or human population density. Several northern districts were identified as 'predation hotspots', where proportions of livestock lost to predation were considerable, and the ratio of reported kills to relative abundance of livestock was high. Implications of our findings for mitigating livestock losses and for conserving large carnivores in Bhutan are discussed.