Friday, August 7, 2009 - 8:40 AM

OOS 46-3: Multistate patch occupancy models to infer movement of organisms: Case study of African elephants

Julien Martin1, Simon Chamaillé-Jammes2, James D. Nichols3, Hervé Fritz4, James E. Hines3, Christopher J. Fonnesbeck5, Darryl I. MacKenzie6, and Larissa L. Bailey7. (1) University of Florida, (2) University of Cape Town, (3) USGS Patuxent Wildlife Research Center, (4) Université Lyon 1; CNRS, (5) University of Otago, (6) Proteus Wildlife Research Consultants, (7) Colorado State University

Background/Question/Methods
The accumulation of monitoring data of organisms at large temporal and spatial scales provides exciting opportunities to evaluate hypotheses related to animal movement. In this presentation we discuss the application of multistate patch occupancy models to make inference about movement from occupancy and count data of organisms at large spatial scales. We demonstrate how multistate patch occupancy models can be used to infer movement of African elephants in relation to waterholes based on count data in Hwange National Park (HNP), Zimbabwe. More specifically, we estimate transition probabilities among three dry-season states for waterholes: (1) unsuitable state (dry waterholes with no elephants); (2) suitable state (waterhole with water) with low abundance of elephants; and (3) suitable state with high abundance of elephants; and evaluate the influence of factors such as annual rainfall and the number of neighboring waterholes on the distribution and abundance of elephants.
Results/Conclusions
Based on our analyses we found that annual rainfall and the number of neighboring waterholes are important factors affecting the distribution and abundance of elephants. There was a positive relationship between annual rainfall and transition probabilities from states of high abundance to states of low abundance. The observed pattern supports the hypothesis that elephants in HNP dispersed to less utilized areas when annual rainfall increased.  Because, the design of the elephant monitoring in HNP did not allow for the estimation of detection probabilities directly, we assigned counts to two large classes of abundance (high versus low abundance) such that it was unlikely for the survey data to be incorrect at that scale. However, we note that the models and software that we used can readily be framed to estimate detection when data collection is based on a study design that allow for the estimation of detection probabilities; we briefly describe such designs. The modeling framework that we present should be applicable to a wide range of existing datasets and should help address important hypotheses related to animal movement for a variety of organisms.