9th International Deer Biology Congress

AUTHORS: Lisa J. Koetke, Floyd W. Weckerly – Texas State University

ABSTRACT: Mathematical models can predict how species respond to their environment, and they are a critical tool for global concerns including climate change, population management, pest control, and species conservation. The Gompertz and Ricker state-space models are both commonly used to estimate population parameters of large mammals using population counts, but few studies have compared the utility of these models while accounting for survey imprecision, or observer error. We used elk population abundance data collected by multiple sources in the U.S. Department of Energy’s Hanford site in southcentral Washington between 1983 and 1999 [1, 2, 3]. This dataset was useful since it includes observed counts over a wide range of abundances. We estimated the parameters of the Gompertz and Ricker hierarchical state-space models using Bayesian statistics in the RJAGS package in R. The Gompertz model estimated observer error following a log normal distribution, whereas the Ricker model estimated observer error following a Poisson distribution. When taking into account survey imprecision, the root mean square error of the Ricker hierarchical state-space model was almost half that of the Gompertz model. Furthermore, the projected 95% credible intervals of the Ricker model encompassed all observed counts, whereas 3 of 17 counts lied outside of the projected 95% credible intervals of the Gompertz model. Finally, the Ricker model, but not the Gompertz model, estimated carrying capacity. Therefore, the Ricker model appears to more precisely model population dynamics of this elk population than the Gompertz model when density dependent factors influence fluctuations in abundance.