Improving the temporal transferability of species distribution models under climate change by incorporating historical species-climate relationship

Xin Chen, Daijiang Li

Understanding how species will respond to climate change is one of the current key challenges in ecology and nature conservation. The tempo-spatial variations of climate makes it more challenging to predict species responses to climate change across their entire ranges. Species distribution models have been widely used for identifying how species distributions respond to climatic drivers. Despite the spatial transferability of SDMs has been largely tested, validation of their temporal transferability is still rare. In addition, there is still no consensus regarding how integration of temporally independent data can improve the reliability of model predictions under different time. Here, we modelled the distribution of white oak (Quercus alba) across its entire range in the eastern United States for the 1900s and the current. We compared predictive performance, predicted suitability and distributions, and predictor’s importance of SDMs accordingly, and quantified climate novelty between the two time periods to assess the temporal transferability and predictive accuracy of SDMs. We found that the SDM fitted with the 1900s climate outperformed that calibrated with the current climate, suggesting its higher transferability from the 1900s to the current. Such difference in transferability between SDMs fitted with the 1900s and current climates may be attributed to the climate novelty and change in limiting factors for white oak distributions during model transfer. Our results demonstrate the temporal transferability of SDMs on predicting temporally independent species distribution across its entire range. With the increased availability of historical species occurrences, incorporating such data into SDMs will increase the reliability of model predictions under future climate change.