Assessing and predicting shifts in mountain forest composition across 25 years of climate change. Issue 5 (9th March 2017)

• Record Type: Journal Article
• Authors: Scherrer, Daniel; Massy, Stéphanie; Meier, Sylvain; Vittoz, Pascal; Guisan, Antoine
• Editors: Serra‐Diaz, Josep
• Is Part Of: Diversity & distributions. Volume 23:Issue 5(2017)
• Page Start: 517
• Page End: 528
• Publication Date: 2017-03-09
• Subjects: community predictions -- ecological indicator values -- global warming -- monitoring -- plant community -- species distribution models -- Switzerland; Biodiversity -- Periodicals; Biodiversity conservation -- Periodicals; 577
• DOI: 10.1111/ddi.12548
• Languages: English
• ISSNs: 1366-9516

Abstract

Abstract: Aim: Spatial predictions of future communities under climate change can be obtained by stacking species distribution models (S‐SDM), but proper evaluation of community S‐SDM predictions across time with fully independent data has rarely been carried out. The aim of this study was to evaluate the predictive abilities of S‐SDMs for whole forest communities across the last 25 years in a mountain region. Location: The western Swiss Alps. Methods: We used past vegetation surveys (2, 984 plots) and environmental data from the 1990s to calibrate SDMs for 364 plant species and predict changes in forest composition under contemporary conditions. These projections were then evaluated by resurveying a random subset of 92 forest plots in summer 2014. Results: Species distribution models showed the same accuracy in the past (calibration data) and present (evaluation data). The S‐SDMs correctly predicted the general trends in species richness and shift of ecological conditions (i.e., temperature, moisture) at the regional level. However, it proved more difficult to identify precisely which forest communities or areas are most or least affected by climate change. Main conclusion: Our results show that, across a period of a few decades, S‐SDMs can usefully predict trends in macroecological properties such as richness or average ecological conditions, but fail to accurately predict changes in composition. This is likely due to the combined effects of the stochasticity of local colonization and extinction events, dispersal limitations, community assembly rules (e.g., competition), observer bias, model and location errors and interannual variation. Furthermore, these models cannot account for potential species adaptations leading to persistence in sites predicted unsuitable. This highlights the need for developing more accurate forest community predictions as support to help prioritizing conservation actions.