The BAM team predicted current avian distributions from our observational data, along with mapped climate variables and landcover information, using a powerful species-distribution modelling tool called Maxent. We produced individual distribution maps for 96 boreal bird species (for which sufficient data existed to develop good models). These maps are posted under individual species’ names in RESULTS.
Our models make use of the BAM project's avian dataset (including point count, BBS and breeding bird atlas data), augmented by additional BBS data collected south of the boreal region (to encompass the southern range limits of many species).
The main outputs of the Maxent models are maps that estimate the probability that a species will be present at a given location. The maps we present were transformed into a cumulative or percentile form, where each pixel shows the proportion of all other pixels in the study region having less suitable habitat. These “class rankings” facilitate comparisons among species.
The Maxent models are examples of bioclimatic or niche models that are meant to describe the current potential distribution of species as constrained by climate and vegetation. They are more reliable than spatial interpolation models because they incorporate environmental conditions in addition to purely spatial factors such as location and distance from known species’ occurrences. Our models do not account for physiographic barriers such as the northern extent of the Rocky Mountains that may prevent colonisation of otherwise suitable habitat. Therefore actual species’ distributions may be over-estimated in certain regions, particularly in Alaska.
The BAM study area does not include the entire ranges of most species. Many bird species breed in the boreal region, migrating to the United States, Central and/or South America to winter. Therefore, our analyses will be most useful in improving the northern range limits for species that spend part or all of their life cycle in the boreal region. We did not model the distributions of species across their entire ranges, as this would require a continental-scale dataset.