A hierarchical approach to assess plant invasions in a region of rapid climate change

Background/Question/Methods Species distributions are controlled by different factors at different spatial scales. We assessed the ability to create accurate predictive models for high-risk invasive plants species using climatic, environmental, and anthropogenic variables at different scales in a region of rapidly changing climate. We selected three highly invasive plant species in Alaska (Cirsium arvense Scop., Melilotus albus  Lam., and Phalaris arundinacea L.) to create predictive models in current and projected future climates at two spatial scales using Maxent. Regional models for coastal Alaska incorporated climatic and topographical variables at a 2-km resolution. Local models for two areas nested within the coastal Alaska region (the Kenai Peninsula and Prince of Wales Island) incorporated additional environmental and anthropogenic variables at a 30-m resolution. We asked the question, “How predictable are current and potential plant species distributions in rapidly changing environments?”

Results/Conclusions  

The regional model for C. arvense performed well (AUC = 0.73), as did the local models for Kenai (AUC = 0.71) and Prince of Wales (AUC = 0.86).  Melilotus albus performed well regionally (AUC = 0.75), but had mixed results locally (Kenai AUC = 0.85; Prince of Wales AUC = 0.63).  The P. arundinacea models performed poorly for the region (AUC =0.59), marginally better locally for Kenai (AUC = 0.74), and poorly for Price of Wales (AUC = 0.57). Regional models predict an increase in area of suitable habitat for all species by 2030; however, the distribution of each species was driven by different climate and topographical variables.  Elevation and distance to right-of-ways are associated with habitat suitability for all three species at the local scale.  In some cases, local models refine the predictions within the area defined as suitable in the regional models, while in others local models included suitable habitat beyond the area defined as suitable in the regional model.  Considering both regional and local models, and including different predictors thought to be important at each scale offers a new and effective tool for addressing potential changes in species distributions.