We used point-count data collected as part of Bird Studies Canada's Boreal Forest Bird Program to validate habitat suitability models for 22 forest bird species in Ontario's Habitat Suitability Matrix. We found that many of the species'models performed relatively poorly in discriminating between occupied and unoccupied sites, primarily due to the high error of commission rates (false positive predictions). Since species presence and abundance were assessed by single, five minute point counts, insufficient sampling was at least partly responsible for some of the observed over-prediction. Results suggested that model parameters for at least nine of the species tested (hairy woodpecker [Picoides villosus], blueheaded vireo [Vireo solitarius], red-eyed vireo [Vireo olivaceus], red-breasted nuthatch [Sitta canadensis], Swainson's thrush [Catharus ustulatus], hermit thrush [Catharus guttatus], Tennessee warbler [Vermivora peregrina], Blackburnian warbler [Dendroica fusca] and dark-eyed junco [Junco hyemalis]) should be reviewed to improve the predictive capability of the models and to ensure appropriate consideration of the habitat needs of these species during forest management planning. Key words: boreal forest, forest birds, discrimination capacity, habitat models, habitat suitability matrix, model accuracy, model validation, relative operating characteristic curve, ROC
Habitat suitability modelling to improve conservation status of two critically endangered endemic Afromontane forest bird species in Taita Hills, Kenya
