Field-validated species distribution model of Canada Warbler (Cardellina canadensis) in Northwestern Ontario

We ran different models also to test the accuracy of the models using only desktop data and a datasete enhances with field-collected data. The SDM’s environmental covariates included a bare soil index (BASI), a normalized water index (NDWI) as an indicator of deciduos vegetation, an enhanced vegetation index (EVI), a digital elevation model (DEM), years since disturbance (DISTURB [usually by logging] 1-20 years since last disturbance happened, 21 value represent undisturbed or no disturbed more than 20 years ago), distance to mature coniferous forest (D_CONIF), tree canopy height (CAN) and distance to water (WATER) as indicator of riparian zones.

The models that used field-collected data showed a moderate performance for both training and test data (AUC 0.7) while the model that used only desktop dataset showed a poor performance (AUC 0.6); NDWI, WATER, EVI and D_CONIF were the most influential covariates indicating high association of CAWA to deciduous vegetation, riparian areas, shrub cover and importance of coniferous stands. CAWA occurrence probability was high in undisturbed areas, but also it has a high predicted probability (>0.6) in areas within six years since disturbance; CAWA may take advantage of regenerated forest depending on shrub density and retention of old-growth forest structure ( CAWA had a high prediction of occurrence areas with canopies higher than 10m tall).