J. B. Dunham
J. Peterson
2003
<p><span>Effective </span>conservation<span> efforts for at-risk species require knowledge of the locations of existing populations. Species presence can be estimated directly by conducting field-sampling surveys or alternatively by developing predictive </span>models<span>. Direct surveys can be expensive and inefficient, particularly for rare and difficult-to-sample species, and </span>models<span> of species presence may produce biased predictions. We present a Bayesian approach that combines sampling and </span>model<span>-based </span>inferences<span> for estimating species presence. The accuracy and cost-effectiveness of this approach were compared to those of sampling surveys and predictive </span>models<span> for estimating the presence of the </span>threatened<span> </span>bull<span> </span>trout<span> (Salvelinus confluentus) via simulation with existing </span>models<span> and empirical sampling data. Simulations indicated that a sampling-only approach would be the most effective and would result in the lowest presence and absence misclassification error rates for three thresholds of detection probability. When sampling effort was considered, however, the combined approach resulted in the lowest error rates per unit of sampling effort. Hence, lower probability-of-detection thresholds can be specified with the combined approach, resulting in lower misclassification error rates and improved cost-effectiveness.</span></p>
application/pdf
10.1046/j.1523-1739.2003.01579.x
en
Society for Conservation Biology
Combining inferences from models of capture efficiency, detectability, and suitable habitat to classify landscapes for conservation of threatened bull trout
article