Thursday, August 5, 2010 - 4:40 PM

OOS 44-10: Historically calibrated predictions of butterfly species range shift

Jeremy Kerr, University of Ottawa

Background/Question/Methods

Climate and land use changes have elicited widespread biological responses and are poised to accelerate extinction rates. Predicting how species will respond to the challenges posed by global change is likely to be a necessary component in strategies intended to minimize negative outcomes. Considerable effort has focused on the development of species distribution models to predict future responses to anticipated climate changes, although these models are known to be susceptible to an array of errors. Here, we test whether historical calibration of species distribution models can influence their predictions regarding how species are affected by climate change in particular. Using an extensive assembly of 20th century butterfly observations from Canada, which covers an extensive area in which anthropogenic climate change began relatively early, aspatial models linking each species observation against its contemporary environment was used to project changes in spatial distribution in an array of time periods. Maximum Entropy was used for this purpose.

Results/Conclusions

It is clear that the historical calibration of species' responses to observed, modern-day climate changes can reliably predict temporal shifts in the distribution of species: species have begun to shift in directions predicted by models. These shifts include areas of apparent range collapse as well as range expansion and are not simply poleward. It is equally clear, however, that some species have responded to the advent of non-analog climates in surprising ways, risking ecological extinction. Climate change impacts on species distributions are demonstrated quantitatively: model predictions fit direct species observations closely across space and through time in the recent past. This historical calibration of species distribution models may assist in predicting species' future responses, but the advent of no-analog climates presents a dangerous risk that models will sometimes fail when projected into the future.