Reducing effort while improving inference: Estimating Dall's sheep abundance and composition in small areas

Abstract

Recent work has demonstrated that aerial distance sampling surveys are more efficient and effective than unadjusted minimum count surveys for estimating Dall's sheep (Ovis dalli) abundance, although large sample size requirements (e.g., 150-200 detections) may discourage implementation of these methods in small (<2,500 km 2 ) or low density areas. However, a Bayesian analytical approach using informed priors and borrowing detection information across surveys can increase precision and decrease required sample sizes. Using these methods, we conducted distance sampling surveys across a majority of the Dall's sheep habitat within National Park Service units in Alaska during 2010-2011. We compared 4 analytical scenarios using increasing amounts of detection information to demonstrate the increases in efficiency that can be gained over time through the use of this approach. Based on our analysis using all available survey information in the estimation of the detection function, we estimated that 2,252 (1,871- 2,765), 2,809 (2,361-3,379), 1,669 (1,339-2,120), and 12,428 (10,780-14,470) sheep occurred in Denali National Park and Preserve (DENA), the Western Arctic National Parklands (WEAR), the Itkillik preserve subarea of Gates of the Arctic National Park and Preserve, and Wrangell-St. Elias National Park and Preserve (WRST), respectively. These estimates were achieved with relatively small numbers of group detections in DENA (n ¼ 57), the Itkillik preserve area (n ¼ 48), and WEAR (n ¼ 100), suggesting that sample size requirements for Dall's sheep distance sampling surveys can be reduced by an additional 50-75% over previously recommended levels when adequate prior information is available. In addition, we describe a formal approach for estimating the size of individual composition classes (i.e., lambs, ewe-like sheep, <full- curl rams, � full-curl rams) and sex and age ratios, corrected for incomplete detection. We implemented the composition analysis within the distance sampling analytical framework as part of the abundance estimation process.This approach to theestimation ofpopulation composition could replace commonly usedindices and provides more detailed and rigorous estimates that are directly comparable among survey areas and years. We found that ratios offull curl rams:100 ewe-like sheep in WEAR and Itkillik preserve subarea populations were less than in the DENA population, whereas lamb:ewe-like ratios were similar across all 4 survey areas, suggesting relatively consistent productivity. We recommend that aerial distance sampling survey methods using prior information, combined with direct estimation of population composition, be used to increase the effectiveness of Dall's sheep population monitoring and management throughout their range. 2013 The Wildlife Society.