Performance of 'Gamma-Shaped' Detection Functions in Aerial-Line Transect Surveys of Large Terrestrial Mammals
Joel H. Reynolds1, Anna-Marie Benson2
1Western Alaska Landscape Conservation Cooperative, Anchorage, AK, United States; 2Division of Realty and Natural Resources, U.S. Fish and Wildlife Service, Fairbanks, AK, United States

Using aerial surveys and distance sampling to estimate the density of large terrestrial mammals can result in detection functions where the maximum probability of detection does not occur along the transect line or inner edge of the search strip. This prompted development of a distance sampling approach using 'gamma distribution'-shaped detection functions. We report two investigations of estimator bias and precision for this approach. Both used simulated surveys of a brown bear (Ursus arctos) population in southwest Alaska, USA. The first study showed density estimates were unbiased, but estimates of some detection-function parameters were highly biased. Likely causes of the bias are discussed (model fitting with nonignorable selection weights) and estimator modifications proposed to correct the issue. Further, standard errors were overestimated; estimator modifications are proposed to reduce that bias. The second study demonstrated underestimation bias in density estimates when observers were not truly independent, a problem gaining recognition in the distance sampling literature. The underlying problem is described and potential estimator modifications discussed. The feasibility of the modified approach, for monitoring large terrestrial mammals in Alaska, is discussed in light of these results and other possible survey methods.

Keywords: Distance sampling; Estimator bias; Induced dependence; Wildlife management

Biography: Joel Reynolds is an applied statistician working in natural resource management, ecology, and environmental science. For the last decade he has focused predominantly on sampling and monitoring problems in wildlife management in Alaska. He has worked on applications ranging from oil spills to wildfire burn severity to climate change effects on vegetation communities, and with species ranging from seabirds to salmon to brown bears. Currently most of his time seems to be spent encouraging agency managers to develop well-thought out monitoring objectives and employ greater usage of statistical planning tools. He recovers from these efforts by skiing, hiking, rafting, or kayaking in the wildlands and waters of Alaska.