Aerial surveys of multiple species: critical assumptions and sources of bias in distance and mark–recapture estimators

Recent developments in the application of line-transect models to aerial surveys have used double-observer sampling to account for undercounting on the transect line, a crucial step in obtaining correct population estimates. This method is commonly called the mark–recapture line-transect sampling method and estimates the detection probability at zero distance to correct line-transect estimates of abundance. An alternative approach, which uses the same methodology during data collection, is to use a range of covariates, including distance from the transect, in a mark–recapture model. This approach overcomes the implicit assumption of uniform distribution of distances in line-transect estimators. In this paper, we use three alternative approaches (a multiple-covariates distance method, a distance method incorporating adjustment for incomplete detection on the transect line using mark–recapture sampling, and a mark–recapture method with distance as a covariate) to estimate the abundance of several medium-sized mammals in semiarid ecosystems. Densities determined with the three estimators varied considerably within species and sites. In some cases distance estimates were larger than mark–recapture estimates and vice versa. Despite large numbers of observations, distance uniformity was not observed for any species at any site, nor for any species where sites were combined. Possible reasons, which include sampling variability, movement in response to the aircraft and failure of the mark–recapture independence assumption, are discussed in detail.