干旱区生态环境知识资源中心

Population estimates derived from aerial surveys of ungulates are biased by imperfect detection, where probability of sighting groups is influenced by variables specific to terrain features and vegetation communities. Therefore, methods for bias-correction must be validated for the region to which they will be applied. Our objectives were to quantify factors affecting detection probability of mule deer (Odocoileus hemionus) during helicopter surveys in Texas, USA, rangelands, and develop a detection probability model to reduce bias in deer population estimates. We placed global positioning system (GPS) collars on 215 deer on 6 sites representative of mule deer range in the southern Great Plains and the Chihuahuan Desert during 2008-2010. We collected data during aerial surveys in January-March and fit logistic regression models to predict detection probability of mule deer based on ecological and behavioral covariates. We evaluated the model using independent estimates of population size derived from a mark-resight procedure. Detection of mule deer was negatively related to distance from the transect, increasing brush cover, sunlight, and increasing terrain ruggedness (P< 0.01). Probability of detection in brush cover was greater if deer were active (P = 0.02). Population estimates corrected for visibility bias using our detection probability model or mark-resight models averaged 2.1 +/- 0.49 (SD; n = 50) and 2.2 +/- 0.62 times larger, respectively, than uncorrected counts. Estimates of population size derived from the detection probability model averaged 101 +/- 26% of mark-resight estimates. However, the detection probability model did not improve precision of population estimates, probably because of unmodeled variation in availability of deer during surveys. Our detection probability model is a simple and effective means to reduce bias in estimates of mule deer population size in southwestern rangelands. Availability bias may be a persistent issue for surveys of mule deer in the Southwest, and appears to be a primary influence of variance of population estimates. (C) 2016 The Wildlife Society.