Recounting bias can affect abundance estimates from intensive helicopter surveys of feral goats

Abstract

Context Aerial surveys are widely used for estimating the abundance of wildlife over large areas. The failure to count all animals within survey transects is commonly acknowledged and there are many techniques to measure and correct for underestimation. However, the possibility of animals being counted more than once in intensive surveys, which leads to overestimation, is rarely examined. Animals can move in response to observers or vehicles, and bias can occur when animals move before or after detection. Movement of animals immediately prior to and associated with observation processes is methodologically accommodated in distance sampling but bias attributable to responsive movement after observation platforms have passed requires investigation. Aims We sought to investigate potential biases caused by animal movement during intensive helicopter surveys of feral goats, and to quantify the probability that animals are available for recounting because of their responsive movements. Methods Using ground-based behavioural studies simultaneous with intensive helicopter strip surveys of feral goats, we measured the extent of responsive movement, distances and directions moved, and sampling design parameters, and contrasted those with random movements. Key results Feral goats did not move randomly in response to helicopters. Animals within the transect strips, and therefore potentially visible from the aircraft, were more likely to move than those outside the transect. Considerable responsive movement (flushing) occurred between transects and more animals (64%, n = 448) moved towards unsampled transects than towards transects already sampled. Because of the spatial separation of transects, 21% of goats were available for recounting in adjacent transects, leading to potential overestimation. Conclusions Although most extensive surveys of macropods and other wildlife in Australia account for overestimation in their design, surveys that sample intensively and apply valid corrections for undercounting are likely to produce positively biased estimates of abundance where flushing occurs. Likewise, intensive thermal surveys could be subject to positive bias for animals prone to flushing. This is routinely ignored in wildlife management and research where close transects are used to estimate abundance. Implications Responsive movement requires consideration when designing intensive aerial surveys of wildlife. Randomised transects without replacement or larger distances between transects will counteract recounting bias.