Comparison of Aerial Survey Methods for Elk in Arizona

• Published in: Wildlife Society bulletin 2019-03, Vol.43 (1), p.77-92
• Main Authors: BRISTOW, KIRBY D., CLEMENT, MATTHEW J., CRABB, MICHELLE L., HARDING, LARISA E., RUBIN, ESTHER S.
• Format: Article
• Language: English
• Subjects: Cervus canadensis; detection probability; double‐observer; elk abundance; hybrid model; mark–recapture; Original Article; population; sightability
• ISSN: 1938-5463; 2328-5540; 1938-5463
• DOI: 10.1002/wsb.940

Elk (Cervus canadensis) populations in Arizona, USA, have historically been managed using estimates of relative abundance. Recent questions related to the influence of large wildfires, habitat alterations, and predator–prey relations have increased the demand for absolute estimates of elk abundance. Between 2014 and 2016, we conducted experimental helicopter surveys of selected areas to compare several methods designed to model elk detection and estimate abundance. We conducted annual autumn helicopter surveys in 3 areas that contained radiocollared elk, and recorded information on covariates affecting both detection (i.e., vegetative cover, vegetation type, burn category, group size, activity, and ambient light) and observer bias (i.e., observer position, pilot experience). We used information theory to rank a set of candidate a priori models to determine which covariates affected detection and select the most parsimonious models among sightability, double-observer, and hybrid modeling methods. We then used the top model from each method, as well as a simultaneous double-count method, to calculate annual site-specific elk abundance estimates for comparison with concurrent mark–recapture abundance estimates, which we assumed best represented the true population size. The best supported models included all detection covariates, with the influence of covariates on elk detection generally adhering to expectations. Relative to mark–recapture estimates, the best performing hybrid model generally provided abundance estimates that were more accurate than double-observer models and more precise than sightability models. The most economical methods to implement were the simultaneous double-count and double-observer methods because these methods did not require model development or presence of marked animals, but they occasionally appeared to provide biased and overly precise abundance estimates. Although the hybrid model required substantial cost to develop, our results suggest that our hybrid model now provides the most cost-efficient and accurate implementation option for estimating elk abundance in similar habitats and under similar conditions in Arizona.