Estimating prey abundance and distribution from camera trap data using binomial mixture models

Measures of absolute animal abundance may be estimated by capture-recapture, removal, or distance sampling methods. We investigate the usage of binomial mixture models to estimate local group abundance of major prey species that is frequently used as a surrogate for prey abundance to study predator or prey-mediated ecological interactions such as predator-prey relationships. We evaluate mixture models using data from a camera-trapping survey intended for a tiger Panthera tigris census in Chitwan National Park, Nepal, where the entire park was surveyed in 361 4-km2 quadrats. We chose four prey species (chital Axis axis, sambar Rusa unicolor, muntjac Muntiacus muntjac, and wild boar Sus scrofa) that collectively account for > 75% of prey biomass consumed by tigers. Abundance of prey group was modeled as a random variable with a Poisson or a negative binomial distribution, with the mean abundance affected by distance from water sources, elevation, and normalized difference vegetation index (NDVI). Except for wild boar, the top models for all other species included the hypothesized covariates while the null model was the most parsimonious model for the wild boar. The most parsimonious chital model included effects of distance from water sources (−) and elevation (−). The sambar model supported all three covariates: distance from water sources (−), elevation (+), and NDVI (+). Only distance from water sources (−) was supported by the most parsimonious muntjac model. Our abundance estimates also conformed to the results obtained from recently conducted labor-intensive distance sampling procedure. We conclude that camera-trapping survey data can be effectively utilized adopting the binomial mixture model framework to understand animal abundance-habitat relationships and estimate abundance of animal that are not identifiable individually.