Performance of distance sampling estimators: a simulation study for designs based on footpaths

Recently, distance sampling emerged as an advantageous technique to estimate the abundance of many animal populations, including ungulates. Its basic design involves the random selection of several samplers (transects or points) within the population range, and a Horvitz-Thompson-like estimator is then applied to estimate the population abundance while correcting for animal detectability. Ensuring even coverage probability is essential for subsequent inference on the population size, but it may not be achievable because of limited access to parts of the population range. Moreover, in several environmental conditions, a random selection of samplers may induce very high survey costs because it does not minimize the displacement time of the observer(s) between successive samplers. We thus tested whether two-stage designs - based on the random selection of points and then of nearby samplers - could be more cost-effective, for a given population size and when even area coverage cannot be guaranteed. Here, we further extend our analyses to assess the performance of two-stage designs under varying animal densities.