Damien A. Fordham1, Frédérik Saltré2
1 The Environment Institute and School of Biological Sciences, The University of Adelaide, SA, 5005, firstname.lastname@example.org,@DamienFordham
2 The Environment Institute and School of Biological Sciences, The University of Adelaide, SA, 5005, email@example.com,@FredSaltre
Species distribution models (SDMs) are the primary tools used to forecast shifts in geographical range limits and by extension extinction risk. They assume that model predictions of habitat suitability and species occurrence are transferable through time. To better establish confidence in forecasts of range movement resulting from climate change, we assessed the accuracy of > 5000 SDM hindcasts of species range shifts using independent validation data. Hindcasts were for species range shifts in the northern hemisphere over the last 20,000 years and included plants, invertebrates and vertebrates. We show that SDMs tend to produce reliable hindcasts of observed 20th century range movement, however, their predictive capacity decreases with time from the present. Hindcasts from the present-day to the late Pleistocene (20−10kya) were rarely able to accurately predict range movements inferred from the fossil record. We show that the extent of observed global and regional climate change during the 20th century is small compared to what was observed over the last 20,000 years and what is forecast for the end of the 21st century. These results suggest that validating hindcasts using only 20th century resurvey data could lead to false optimism in the capacity of SDMs to forecast changes in geographical range limits, because the extent of observed climate change during this period is small compared to what is likely for the near future. We conclude that long-term retrospective analyses are needed to establish a more robust framework for predicting the likely effects of climate change on future range shifts.