Assessing Efficacy of Random Encounter Models for Low Density Populations Using Unmarked Mule Deer and Coyotes

Cullen Anderson

doi:10.7273/000005173

Estimating population parameters is an essential yet challenging process for wildlife conservation and management globally. Local, national, and international policies often depend on population trend data derived from density estimates. However, logistical and financial limitations may constrain regular assessment of wildlife densities using traditional mark-recapture based techniques. Random Encounter Models (REM) represent a simple, cost-effective method for estimating unmarked species densities using camera traps, and numerous studies have demonstrated REMs are robust compared to traditional methodologies. However, REM performance seems to be weakest for low density populations, and literature sources disagree about its use in these situations. Uncertainty in REM input variables (e.g., species movement rates, camera trap detectable area) may further decrease the utility of applying REM estimation methods for low density populations. Here, we examined the efficacy of REM for low density populations of two species with contrasting movement rates: a less mobile, large herbivore, mule deer (Odocoileus hemionus), and more mobile mesocarnivore, coyotes (Canis latrans). We deployed an array of camera traps across a ~101-km2, low permeability study area in northeastern Oregon and estimated mule deer and coyote densities using REMs from image detections. We then compared REM estimates between species and to existing density estimates for those species in the same study area derived from modern modifications of traditional mark-recapture models (e.g., Sight Mark Resight, Spatially Explicit Capture Recapture). To assess effects of variation in movement rate and camera trap detectable area on REM estimates, we simulated and visualized plausible mule deer, coyote, and camera trap characteristics. Comparisons of REM estimates to previous density estimates suggested REMs underestimated mule deer densities but adequately estimated coyote densities relative to mark-recapture methods. Our simulations and visualizations revealed variation in movement rates created the greatest degree of uncertainty in REM estimates for both species, whereas variation in camera trap detectable area had relatively little influence. However, environmental conditions may have decreased mule deer population densities between original and REM estimates, and our REM coyote estimates were likely biased high by underestimated movement rates. Our study demonstrated potential applicability of REM models to low density wildlife populations while illustrating the need for critical evaluation of parameter uncertainty before interpreting results.

Assessing Efficacy of Random Encounter Models for Low Density Populations Using Unmarked Mule Deer and Coyotes

Files and links (1)

Abstract

Metrics

Details