Identifying habitat correlates of latent occupancy when apparent annual occupancy is confounded with availability for detection

Abstract

Estimating relationships between landscape variables and the presence of a species via occupancy modeling is a common practice for many animal species. However, estimating these relationships can be difficult for species where availability for detection is driven by factors that match the traditional primary sampling period. We present a new hierarchical formulation of the occupancy model to estimate these relationships in the presence of among-year variation in availability for detection for the threatened Mojave desert tortoise (Gopherus agassizii). There were large swings in apparent annual proportion of area occupied ranging from 0.19 to 0.66, with year-to-year changes in the apparent annual proportion ranging from −25.7% to 230.2%. The model estimated the true latent proportion of area occupied was 0.57 (95% Cr.I. 0.51–0.631). The predictive raster surface developed from the novel model formulation validated well using an independent data set (Pearson's r=0.948), with radio-telemetered desert tortoises spending disproportionately more time in higher predicted probability of occurrence portions of the study area. The coefficients from the model, and more specifically the occupancy probability predictive raster surface developed from them, can be used by land managers to guide future survey efforts and to spatially prioritize restoration actions across a 35,000ha conservation reserve. In general, acknowledging the challenge of confounding availability for detection with apparent annual occupancy and using the modeling framework presented here can be used to determine fundamental relationships between landscape configuration and latent occupancy for a variety of heretofore unaddressed species.