Influence of point count length and repeated visits on habitat model performance

Point counts are commonly used to monitor bird populations, and a substantial amount of research has investigated how conducting counts for different lengths of time affects the accuracy of these counts and the subsequent ability to monitor changes in population trends. However, little work has been done to assess how changes in count duration affect bird-habitat models developed from point count data. In this paper, we present an empirical comparison of the performance of bird-habitat models, which were developed via logistic regression analyses based on point count data from 3-, 5-, 10-, and 20-min counts. We also investigated the effect of the number of visits to each survey point on model performance. We assessed model performance on the basis of R²-values and percent concordant pairs. A positive relation between model performance and count duration was most apparent for species with relatively low detection probabilities, whereas performance of models for species with relatively high detectability was fairly consistent or even decreased as count duration increased. Our results suggest that while some improvement in bird-habitat models for species with low detection rates can be achieved via longer point counts, the modest gains in model performance should be weighed against the increased time and effort required to conduct longer counts. Models based on data from a single visit to each point did not performed as well as models based on multiple visits. However, we found little or no improvement in model performance when the number of visits per point increased from 2 to 3. We suggest that current recommendations on point count durations (5 or 10 min) will provide adequate data for modeling bird-habitat relations.