Assessing bias in aerial surveys for cetaceans: Results from experiments conducted with the franciscana dolphin

Abstract

Line transect aerial surveys are widely used for estimating abundance of biological populations, including threatened species. However, estimates obtained with data collected from aircraft are often underestimated because of visibility bias and bias in estimating group sizes from a fast-moving platform. An assessment of multiple sources of bias in aerial surveys were carried out in Brazilian coastal waters by experiments on multiple survey platforms (i.e., boat, airplane and helicopter). These studies focused on evaluating visibility bias (perception and availability bias) and potential differences in the estimation of group sizes from different types of platforms used in franciscana (Pontoporia blainvillei) abundance surveys. The ultimate goal was to develop correction factors to improve accuracy of estimates of density and population size for this threatened dolphin. Estimates of density and group sizes computed from boats were assumed to be unbiased and were compared to estimates of these quantities obtained from an airplane in the same area and period. In addition, helicopter surveys were conducted in two areas where water turbidity differed (clear vs. murky waters) to determine surfacing-diving intervals of franciscana groups and to estimate availability for aerial platforms. Abundance computed from the aerial survey data underestimated the true abundance by about 4-5 times, with ~70% of the total bias resulting from visibility bias (~80% from availability bias and ~20% from perception bias) and ~30% from bias in estimates of group size. The use of multiple survey platforms in contrasting habitats provided the opportunity to compute correction factors that can be used to refine range wide abundance estimates of the threatened franciscana given certain assumptions are met. Visibility bias and group size bias were substantial and clearly indicate the importance for accounting for such correction factors to produce unequivocal population assessment based on aerial survey data.