Trait-Based Climate Change Predictions of Vegetation Sensitivity and Distribution in China

Dynamic global vegetation models (DGVMs) suffer insufficiencies in tracking biochemical cycles and ecosystem fluxes. One important reason for these insufficiencies is that DGVMs use fixed parameters (mostly traits) to distinguish attributes and functions of plant functional types (PFTs); however, these traits vary under different climatic conditions. Therefore, it is urgent to quantify trait covariations, including those among specific leaf area (SLA), area-based leaf nitrogen (N(area)), and leaf area index (LAI) (in 580 species across 218 sites in this study), and explore new classification methods that can be applied to model vegetation dynamics under future climate change scenarios. We use a redundancy analysis (RDA) to derive trait–climate relationships and employ a Gaussian mixture model (GMM) to project vegetation distributions under different climate scenarios. The results show that (1) the three climatic variables, mean annual temperature (MAT), mean annual precipitation (MAP), and monthly photosynthetically active radiation (mPAR) could capture 65% of the covariations of three functional traits; (2) tropical, subtropical and temperate forest complexes expand while boreal forest, temperate steppe, temperate scrub and tundra shrink under future climate change scenarios; and (3) the GMM classification based on trait covariations should be a powerful candidate for building new generation of DGVM, especially predicting the response of vegetation to future climate changes. This study provides a promising route toward developing reliable, robust and realistic vegetation models and can address a series of limitations in current models.