Comparing methods to estimate feral burro abundance

Abstract

Abstract Obtaining precise and unbiased estimates of feral burro ( Equus asinus ) abundance in the western United States is challenging due to their cryptic pelage and the rugged terrain they inhabit. Management agencies employ helicopter‐based, simultaneous double‐observer sightability surveys (hereafter denoted as DOS) to estimate abundance of burros; but the DOS method routinely produces negatively biased estimates due to residual heterogeneity in detection probability. Consequently, testing alternative methods to improve upon current procedures is warranted. Residual heterogeneity in DOS surveys can be minimized by including radio‐collared individuals in the population. Alternatively, if distance measurements are recorded, residual heterogeneity can also be reduced via a mark‐recapture distance sampling (MRDS) approach. Aerial infrared (IR) surveys offer a safer alternative than helicopter‐based surveys because they can be flown at a higher altitude and require fewer observers in the aircraft. Further, IR surveys using a distance sampling approach have been shown to generate accurate and precise estimates of feral horse ( E. caballus ) populations. Accordingly, we compared results of surveys using aerial IR distance sampling, the standard DOS survey, a DOS survey incorporating detections of radio‐collared individuals, and an MRDS analysis of a feral burro population with a known minimum population size in central Utah, winter 2015–2016 and spring 2016. The minimum number of burros known alive during the winter and spring surveys were 236 and 136, respectively. The average detection probability of IR surveys was P = 0.88 (SE = 0.16) and distance models produced estimates of 127 burros (95% CIs = 99–175) for the winter survey, and 94 burros (CIs = 72–134) for the spring survey. Mean detection probability of the standard DOS surveys was P = 0.78 (SE = 0.09), and model‐generated abundance estimates were 155 burros (CIs = 133–227) in winter, and 92 burros (CIs = 79–139) in spring. Incorporating detections of radio‐collared individuals in the DOS survey resulted in a decreased detection probability ( P = 0.46; SE = 0.06) and increased abundance estimates to 267 (CIs = 169–571) and 155 (CIs = 128–263) for winter and spring, respectively. Mark‐recapture distance sampling produced a mean detection probability of P = 0.48 (SE = 0.12) and resulted in estimates of 282 (CIs = 178–385) and 169 (CIs = 73–310) burros in winter and spring, respectively. Our study demonstrated that aerial IR surveys conducted using standard distance sampling can produce precise estimates of burro population sizes; however, estimates were negatively biased relative to the known population size. Small sample size limits generalization of our results, but the IR‐based distance approach did not improve upon DOS surveys. Accounting for residual heterogeneity through use of radio‐collars and mark‐recapture distance sampling eliminated the negative bias from the standard DOS survey but decreased survey precision. Managers will need to decide whether unbiased but less precise abundance estimates are preferable compared to a more precise, but biased, estimate.