Evaluation of desert bighorn sheep abundance surveys, southwestern Arizona, USA

Abstract

ABSTRACTDesigning wildlife surveys requires biologists to identify their objectives and the accuracy (i.e., bias and precision) required to inform them. In southwestern Arizona, USA, abundance estimates and trends for desert bighorn sheep (Ovis canadensis) rely on detection‐corrected aerial surveys (i.e., group‐size estimator) to inform harvest and assess management actions. The accuracy requirements and trend resolution remain undefined, rendering the surveys ability to address management needs uncertain. We used simulations based on historical surveys to optimize survey efficiency, estimate the accuracy of alternative sampling designs, and evaluate if the accuracy produced by this survey meets management needs. Simulations varied by the amount of area surveyed and temporal frequency. Given annual surveys, we examined trend detection with alternative significance levels (α spanning 0.05–0.30). Our simulations, which accepted the survey assumptions, identified many designs producing unbiased and precise estimates (CV <20%). Alternatives exist for optimizing this survey. For instance, surveying >70% of a Game Management Unit (GMU) provides similar precision to 100% coverage. Population declines ≥12%/year are detectable with annual surveys over a 10‐year period when α = 0.05. Setting α = 0.3 enables detecting declines ≥14%/year within 3 years. This survey assumes constant detection, calculated in 1 GMU 2 decades ago, and applied to 10 other GMUs since. We tested this assumption by estimating group‐size detection from field data in another GMU for 4 years. Bighorn sheep detection varied across GMUs and survey periods. Abundance estimates using the new detection rates were approximately 40% lower than current estimates. Our survey evaluation revealed that differences in abundance of approximately 50% are often not detectable, precision is insufficient to detect large trends in a timely manner (i.e., 40% decline in 7 years), and assuming a fixed detection process remains unfounded. Alternative sampling designs that estimate detectability concomitant with the survey, combined with targeted studies, would better inform management objectives for these desert bighorn sheep. Our assessment demonstrates the problems that occur when survey requirements are vague or mismatch survey design, and that monitoring designs incapable of capturing spatial and temporal variation in detectability will risk misrepresenting animals’ population sizes and trends. © 2018 The Wildlife Society.