The use of species distribution models to evaluate potential responses to climate change is important as we consider management options.  However before we can begin to bring climate change impacts into the complex framework of management priorities, we must evaluate model uncertainties.  We have focused on empirical abundance-based habitat models utilizing decision-tree based techniques (RandomForest) to better understand potential changes of 134 tree species habitats in the eastern US (www.nrs.fs.fed.us/atlas).  Within the broader region surrounding the Chequamegon-Nicolet National Forest, we identify 76 species with habitat presently and/or potentially occurring within northern Wisconsin.  From this list, we focus on identifying the variables driving the species models at broad and regional scales of its distribution, by dissecting the predictive model to identify the weighted node location for each variable and then partitioning the variable importance scores of climate and soil features.  In addition, we describe and rank life history and ecophysiological characteristics that will influence how a species may respond to climate change, variables that cannot be reasonably addressed through our habitat models.  This approach provides a more comprehensive picture of the potential changing tree species pool within northern Wisconsin.  

Results/Conclusions

Species habitats were grouped into classes ranging from potential extirpation, decreasing, no change, increasing and new shifting habitat into northern Wisconsin (1, 18, 7, 21, 29 species, respectively).  The species models projecting losses in habitat generally were much more influenced by broad-scale climate influences (median = 71% weighted variable importance) as compared to climate influences at more local scales (median = 13%).  Next, we considered 9 biological factors reflecting innate characteristics like competition for light and edaphic specificity. Twelve disturbance characteristics address the general species response to events such as drought, insect pests, and fire.  This information draws distinction between species likely to be more tolerant (or sensitive) to environmental changes than the habitat models alone suggest.  For example, red maple is projected to decline in habitat in northern Wisconsin but is the highest ranked species in terms of positive biological response and cumulative disturbance tolerance.  We believe these scores can provide additional interpretive and practical value to habitat model projections.  The combination of modeling habitat, identifying environmental drivers, and compiling life history information provides further insight into how species may respond to climate change, and should help decisions makers evaluate various management options in the face of climate change in northern Wisconsin.