Methods for investigating measurement error in cetacean line-transect surveys

Abstract

Line-transect estimates of abundance generally assume that distances and angles to animals are measured without error. Biased distance measurements will lead to corresponding bias in abundance estimates. Efforts to address this have been made, but measuring distance to cetaceans at sea remains problematic. Four distance-estimation experiments were conducted to explore relationships between estimated and measured distance. Heteroscedasticity was found in all four cases. Preliminary evidence suggested that an observer differed in the ability to judge distance to fixed, continuously-visible cues and ephemeral, cetacean cues, which calls into the question the common practice of using marker buoys as cetacean proxies in distance-estimation experiments. Two studies found visual estimates to be positively biased, and two studies found reticule measurements to be negatively biased. Suitable correction factors were developed to address systematic bias from visual estimates, but these varied widely among observers and were sometimes non-linear. Errors in three studies showed positive skew, suggesting that ranges were overestimated to a larger degree than they were underestimated. If reticule and photogrammetric measurements yield log-normally distributed errors generally, then a least-squares regression will always overestimate the correction factor, underestimate range, and overestimate abundance. Photogrammetric methods to measure range to cetaceans performed well, and their use is encouraged. When measurements cannot be made to all sightings, however, it is recommended that experiments be conducted that generate sufficient sample size (of the target species, across typical survey conditions, and beyond the maximum range that most estimates will be made during the survey) to assess error distributions, examine evidence for non-linearity, and to consider inter-observer differences. Distance experiments, and training on survey protocols, can be conducted using ships of opportunity prior to the beginning of a dedicated survey, which would improve observer efficiency while reducing the costs of dedicated vessel time.