|
Continental scale modeling of bird diversity using canopy structure metrics of habitat heterogeneity
Scott
J.
Goetz, Woods Hole Research Center, sgoetz@whrc.org
(Presenter)
We assesses and predicted breeding bird species richness across North America using variables describing physical environment, vegetation properties, and a series of vertical structure metrics derived satellite observations. We explored the extent to which the vertical structure metrics compared to other variables, such as vegetation cover and productivity. Bird observations were derived from the North American Breeding Bird Survey (BBS). We used RandomForest, an ensemble regression tree approach, to model the relationship between the predictor variables and species richness, to rank the relative importance of these very different influences on biodiversity patterns, and to create continuous maps of species richness for the continental United States at 1km resolution for different guilds (types) of birds. Overall, the predictors exhibited variable relationships with species richness, with 45% variance explained for all birds but 85% for forest birds. The most important predictors varied by guild. Although the vertical structure metrics did not perform as well as the other variables, this was partly a result of sparse sampling along the bird observation routes, unlike the continuous cover and gridded climate variables. An examination of the vertical structure metrics shows a strong relationship more locally, suggesting that with improved lidar data sets, the relevance of canopy structure would be better resolved. The final richness distribution maps indicate that these techniques can be applied to monitor species richness at continent wide scales, and would be relevant to distribution of specific species – even rare species or those threatened or endangered with extinction. Presentation Type: Poster Session: Science in Support of Decision Making (Wed 10:00 AM) Associated Project(s):
Poster Location ID: 161
|