Taking the next step with abundance based distribution models to assess climate change

Background/Question/Methods

After more than a decade of research exploring how climate change may impact biotic communities, there remains a need to push onward to address critical challenges.  There has been considerable progress made in empirical statistical and process based mechanistic models.  We have focused on empirical abundance-based habitat models utilizing the latest statistical techniques (RandomForest) and have generated tools and summaries to better understand potential changes of tree species habitats in the eastern US (www.nrs.fs.fed.us/atlas).  The potential for change and the direction of change will likely be species specific and significant.  Yet, because we can not know how the climate will change precisely, we use multiple climate models and emission scenarios.  While this variability is important, it challenges interpretation.  In addition, the ubiquity, rarity, or spatial configuration of species can limit our ability to generate representative models, thereby also inhibiting modeling of potential future habitat.  Here we address some challenges in projecting potential changes under climate change.  First, we model 134 tree species with 3 climate models, each under 2 emissions scenarios (A1Fi and B1).

Results/Conclusions

The differences in projected changes for the tree species between higher and lower emissions results in similarity in predicting the aerial extent of a species range (mean similarity =85%), but differs substantially in areas of higher habitat importance (mean similarity =60%, SD=20%).  From this assessment, we identify the species and locations of greatest uncertainty based on climate models and emission scenarios.  Next, the challenge of modeling species distribution into uncertain futures is to capture many life history characteristics which will ultimately be responsible for shaping their realized distributions.  Therefore, we consider not only the habitat models, but also quantitatively score the influences of climate, disturbance, and biology for each species to measure how vulnerable a species’ distribution may be to climate change independent of the habitat model.  Species such as Ulmus americana, which is projected to increase, based on our habitat models, by 5 to 25 %, has the biological characteristics (e.g. dispersal, edaphic specificity) associated with gains in area.  However, the strong negative impact of disease and susceptibility to drought limits the species within its current and future range.  In contrast, the maples may be more adapted to withstand climate change than suggested from the models because of their high regeneration capacity. This research addresses multi-model interpretation, and provides a framework to consider how broad-scale species models can incorporate non-modeled biotic and abiotic factors.