Generalizability of neotropical bird abundance and richness models

Predicting the consequences of land-cover change on tropical biotas is a pressing task. However, testing the applicability of models developed with data from one region to another region has rarely been done. Bird faunas were sampled along 3.0-km routes in southern Costa Rica (Coto Brus) to develop statistical models to describe the abundance and richness of groups as a function of land-cover characteristics. The relative value of the land-cover models was assessed by comparing them with null models. The generalizability of the models was tested with data from north-western Costa Rica (Monteverde) to determine whether the models were applicable to another area that has undergone significant land-cover change in the last 60 years. The richness and abundance of understory, open-country and edge non-insectivore groups showed clear relationships with land-cover variables, and the land-cover models had lower prediction errors than the null models for Coto Brus. With one exception, useful models for canopy birds, edge insectivores and hummingbirds could not be developed. The land-cover models of abundance of canopy insectivores, understory insectivores and non-insectivores, and edge non-insectivores were generalizable to Monteverde whereas the land-cover models of abundance of open-country birds and species richness for any of the groups were not better than null models for Monteverde. The results indicate that land-cover models that describe the abundance or richness of various bird groups provide useful predictions in the area where the data were collected and that models of abundance of some canopy, understory and edge birds may perform well in areas that are similar in elevation, life zones and land use to the area from which data were collected. Land-cover models of the abundance of other groups, and of the richness of the majority of groups, may be less generalizable to other areas, or it may be difficult to develop models at all.