Animal Density Estiamtion Using a Trapping Web Design: Field Validation Experiments

Abstract

A recent developed trapping web sampling design (Anderson et al. 1983) and a series of four density estimation models were evaluated using known field populations of ground—dwelling darkling beetles (Tenebrionidae: Eleodes spp.) in a shrub—steppe ecosystem in southwestern Wyoming. Known densities (range 794—11 793 beetles/ha) of marked beetles were established annually in two circular enclosures (415.5 m2 each) from 1984 to 1987. After releasing the beetles, we sampled the marked populations in each enclosure, using 141 pitfall traps arranged in a web design. The resulting data were analyzed with four density estimation models (Fourier series, exponential power series, negative exponential, and half normal) using the program TRANSECT. We then evaluated each model's ability to estimate the true beetle density (D). We found that the negative exponential model was most reliable, estimating D "correctly" (i.e., the 95% confidence interval of D included D) in 100% of the trials (n = 11), and having a mean coefficient of variation (CV) of 20% and no detectable bias. The exponential power series and Fourier series also performed well, providing confidence interval coverage of D in 91 and 73% of the trials, respectively. However, the precision of the exponential power series model was extremely low, resulting in very wide confidence intervals. The half normal model consistently underestimated the true densities, and provided only 18% confidence interval coverage. Tests of the sampling design's assumptions (regarding animal capture probabilities and movements) using actual field data showed that, in most cases, all assumptions were satisfied; even when assumptions were violated, the there successful models continued to function well. An improved procedure for setting confidence intervals is also provided. Based on these analytical results, and considering the logistical and theoretical advantages of trapping webs over trapping grids, we recommend the use of the web design and distance measure analyses for estimating population densities of mobile small animals.