Large scale assessment of ungulate populations via nocturnal distance sampling. Do survey designs based on random footpaths selection provide reliable estimates?

Conventional Distance Sampling (CDS) recently emerged as an advantageous technique to avoid the assumption of constant detectability implicit in many monitoring surveys of ungulate abundances carried on in large study areas. CDS can be applied for large mammals with relatively high detectability and, compared to capture-mark-recapture, it has the advantage of providing an estimate of detection probability without catching animals. In spite of this, when the ungulate population of interest occupies a large, forested study area, the terrain roughness and the lack of an adequate road networks may prevent CDS application. Or, at least, the environmental constraints may not allow an appropriate placement of the samplers across the study area, thus introducing systematic biases in the survey design, resulting also in an uneven coverage probability. In nocturnal surveys where animals are detected by thermal imagining, a common solution is to collect data along existing footpaths but, till now, nobody has investigated whether this approach provides reliable estimates. In this study, we consider two sampling designs for CDS along footpaths: i) random selection of footpaths, and ii) two-stage sampling selection of footpaths. Adopting a simulation approach, we apply them to arbitrary-distributed populations and we evaluate the performance of the estimators in terms of accuracy and precision, showing that the two-stage sampling designs with few blocks may emerge as a cost-effective design to improve the estimate of ungulate abundances at a landscape scale. We demonstrate our ideas by applying this method to a red deer (Cervus elaphus) population in the Italian Apennine.