Mark‐Recapture Distance Sampling for Aerial Surveys of Ungulates on Rangelands

Authors:
Publisher website
Google Scholar
Abstract
Aerial surveys are an efficient technique for counting animals over large geographic areas such as rangelands. In southwestern rangelands, aerial surveys are routinely conducted for ungulates, with the implicit understanding that abundance estimates represent an undercount. Distance sampling can correct for visibility bias, but assumes perfect detection on the survey line, a condition often violated in aerial surveys of ungulates. The incorporation of mark‐resight methods into the distance‐sampling framework, termed mark‐recapture distance sampling (MRDS), corrects for both visibility bias and imperfect detection on the survey line. However, the use of MRDS introduces logistical and technical constraints that may not be practical for aerial surveys. We conducted aerial surveys of ungulates on rangelands in south Texas, USA, to evaluate the feasibility and performance of MRDS relative to conventional distance sampling (CDS) and multiple covariate distance sampling (MCDS). We conducted surveys prior to and after leaf‐fall in 2013–2015 to test the hypothesis that distance sampling corrected for changes in visibility bias. We surveyed white‐tailed deer ( Odocoileus virginianus ), nilgai ( Boselaphus tragocamelus ), collared peccary ( Pecari tajacu ), and feral swine ( Sus scrofa ) on 4 sites. Each site was surveyed seasonally for 2 years; twice both prior to (Nov) and after leaf‐fall (Feb). Probability of detection on survey lines for each species was high (range = 0.82–0.97) and the average for each species was similar between seasons (0.89 and 0.92 during pre and post leaf‐fall, respectively). The MRDS density estimates often were only ~10% greater than CDS and MCDS estimates; all population estimates had overlapping 95% confidence intervals. Further, CDS and MCDS estimates were similar indicating that measured covariates (seat position, vegetation type, and cluster size) contributed little towards detection probabilities. Despite similar average probability of detection for all species before and after leaf‐fall (0.47 and 0.51, respectively), deer and nilgai population estimates were 22–59% lower during fall surveys than winter surveys. The discrepancy between consistent probability of detections with different population estimates suggests that availability bias, which cannot be addressed with distance sampling, was an issue. Overall, the MRDS technique addressed imperfect detection on the survey line and generated probabilities of detection in the survey area consistent with previous studies done in Texas. However, the extra costs (~US$13,000) and logistical hurdles to preserve observer independence for a small increase in precision of population estimates may not be justifiable. © 2020 The Wildlife Society.

This publication has 38 references indexed in Scilit: