Climate extremes may be more important than climate means when predicting species range shifts
It is well known that temperatures across the globe are rising, but climatic conditions are becoming more variable as well. Forecasts of species range shifts, however, often focus on average climatic changes while ignoring increasing climatic variability. In particular, many species distribution models use space-for-time substitution, which focuses exclusively on the effect of average climatic conditions on the target species across a geographic range, and is blind to the possibility of range-wide population collapse with increasing drought frequency, drought severity, or climate effects on other co-occurring species. Relegated to assessments of broad demographic patterns that ignore underlying biological responses to increasing climatic variability, this prevalent method of distribution forecasting may systematically underpredict climate change impacts. We compare six models of survival and abundance of a subcanopy tree species, Taxus brevifolia, over 40 years of past climate change to disentangle multiple sources of uncertainty: model formulation, scale of climate effect, and level of biological organization. We show that drought extremes increased Taxus individual- and population-scale mortality across a wide geographic climate gradient, precluding detection of a monotonic relationship with average climate. Individual-scale climatic extremes models derived from longitudinal data had the highest predictive accuracy (82%), whereas mean climate models had the lowest accuracy (< 65%). Our results highlight that conclusions drawn from forecasts of average warming alone likely underpredict climate change impacts by ignoring indicators of range-wide population declines for species sensitive to increasing climatic variability.