Predicting biodiversity dynamics in response to climate change is challenging. Processes underlying species’ responses can lead to disequilibrium dynamics such as time delays, memory effects, or alternate unstable states, limiting our ability to forecast biodiversity changes. Here we empirically quantify predictability in plant community responses to climate change in 426 North American plant communities response to temperature and precipitation changes over the last 21,000 years. We used paleo-occurrence data from the Neotoma database and climatic data from the CCSM3 paleoclimate model. Using phase space diagrams, we measure community response to environmental forcings via statistics resuming lag and alternate assemblage states. We also determined whether patterns of predictability were related to six anthropogenic, biotic and abiotic predictors know to influence migration, persistence and stochasticity processes in community response dynamics.
Results/Conclusions
Plant assemblages were generally far from equilibrium with climate (e.g mean absolute temperature deviation of 9.64 ± 3.6 ℃). More than 60% of sites also exhibited low predictability response (more than 1 alternate state). Most of the assemblages (89%) tends to reduce their lag through the study period. These patterns were structured in space with lags and alternate states both increasing with latitude and baseline paleo temperature, while decreasing with human density. Thus, our results indicate that equilibrium and simple lagged assemblage dynamics are not common. Accurate predictions of community responses to environmental forcings may be challenging : detailed knowledge of climate change, community state and composition may often be necessary to predict biodiversity dynamics arising from contemporary global change.