Cost-effective nocturnal distance sampling for landscape monitoring of ungulate populations

Abstract

Estimating the size of ungulate populations dwelling forested habitats is technically difficult and expensive. In particular, population assessment via conventional distance sampling (CDS) at landscape scales is challenging and often discarded for its high costs. The development of a framework for its methodological optimization is mandatory. To tackle this issue, we used survey simulations. For arbitrary-distributed populations, we compared the following: (i) systematic random transect allocation, (ii) random selection of footpaths, and (iii) two-stage sampling selection of footpaths (2, 5, 10 blocks). The performance of two-stage sampling and random footpath selection estimators was similar. Then, we applied previous results to estimate the size of a red deer (Cervus elaphus) population in the Italian Apennines. Using data from a pilot survey, we estimated density via CDS and density surface modeling (DSM), and we quantified survey costs per unit effort. Considering our deer distribution, we finally simulated and evaluated the cost-effectiveness of the abovementioned designs for a range of realistic efforts (25–65 transects). CDS produced a negatively biased and less precise estimate than the corresponding DSM. For an effort of 65 transects, design (ii) estimates were unbiased (coefficient of variation = 0.31), while design (iii) provided negatively biased estimates (coefficient of variation = 0.27). Two-stage sampling designs with few blocks were less expensive than other designs in attaining the same level of precision, and they emerged as a cost-effective survey design. Our simulation approach thus provided managers a readily available tool to improve the estimate of ungulate abundances at a landscape scale.