Evaluation of Aerial Line Transects for Estimating Mule Deer Densities

Publisher Website
Google Scholar
Abstract
We evaluated line transect methodology for estimating mule deer (Odocoileus hemionus hemionus) densities with helicopter surveys in pinyon pine (Pinus edulis)-Utah juniper (Juniperus osteosperma) woodland in Piceance Basin, northwestern Colorado [USA]. Eight sighting distance models and their associated estimators (Burnham et al. 1980. Buckland 1985) were applied to 8 line transect trials flown in enclosures of 194-482 ha with known deer densities (21-76/km2). We used 8 measures of bias and precision to evaluate performance of each estimator. The exponential polynomial (EXPL) and negative exponential (NEXP) estimators generally performed best. The EXPL had small bias (estimated density [.cxa.D] = 0.90D, P = 0.201) and no deviation from the nominal 95% confidence interval coverage with 8 of 8 (P = 0.663) intervals covering true density. Bias was not detected for the NEXP (.cxa.D = 1.02D, P = 0.830), and 8 of 8 confidence intervals covered true density (P = 0.663). The exponential power series (EXPS) estimator was biased (.cxa.D = 0.82D, P = 0.030) and had a mean standard error (SE) 2-3 .times. greater than other estimators. All other estimators had relatively good precision, but were biased 17-25% low (P .ltoreq. 0.020). The most reasonable explanation for negative bias in density estimates is missing animals on the line. Violation of this assumption causes estimators with a shoulder on the sighting function to underestimate density. We propose some guidelines for use of line transects, but more rigorous ones are needed for selecting truncation distance, cut points, and an estimator. Otherwise, there is too much flexibility to massage data to generate an "acceptable" density estimate. With 2 surveys on a deer winter range, line transects produced higher density estimates than quadrat sampling, but precision was similar for equal flying time. Due to smaller bias, line transects should be preferred over quadrat sampling for the deer encountered in this study.

This publication has 5 references indexed in Scilit: