Improving population analysis using indirect count data: A case study of chimpanzees and elephants

Abstract

Estimating spatiotemporal patterns of population density is a primary objective of wildlife monitoring programs. However, estimating density is challenging for species that are elusive and/or occur in habitats with limited visibility. In such situations, indirect measures (e.g., nests, dung) can serve as proxies for counts of individuals. Scientists have developed approaches to estimate population density using these “indirect count” data, although current methods do not adequately account for variation in sign production and spatial patterns of animal density. In this study, we describe a modified hierarchical distance sampling model that maximizes the information content of indirect count data using Bayesian inference. We apply our model to assess the status of chimpanzee and elephant populations using counts of nests and dung, respectively, which were collected along transects in 2007 and 2021 in western Uganda. Compared with conventional methods, our modeling framework produced more precise estimates of covariate effects on expected animal density by accounting for both long-term and recent variations in animal abundance and enabled the estimation of the number of days that animal signs remained visible. We estimated a 0.98 probability that chimpanzee density in the region had declined by at least 10% and a 0.99 probability that elephant density had increased by 50% from 2007 to 2021. We recommend applying our modified hierarchical distance sampling model in the analysis of indirect count data to account for spatial variation in animal density, assess population change between survey periods, estimate the decay rate of animal signs, and obtain more precise density estimates than achievable with traditional methods.