A novel approach to modeling vegetation distribution through trait-climate relationships
Trait-based approaches have recently elicited much recent interest for the development of the next generation of dynamic global vegetation models (DGVMs). As part of the ongoing development of plant functional trait (FT) hybrid models, a Gaussian mixture model (GMM) was adapted for and applied to predicting the distribution of vegetation in China and investigating the sensitivity of vegetation to changing climate conditions on the basis of trait-climate relationships in China. First, we aggregated data on three keyFTs, including leaf mass per area (LMA), area-based leaf nitrogen (Narea), mass-based leaf nitrogen (Nmass), from the available literature. In addition, one structural trait of plant communities, leaf area index (LAI), was extracted from MODIS products across China. Second, we derived and developed trait-climate relationships and used different trait combinations in a GMM to model vegetation distribution. Finally, the GMM trained by the LMA-Nmass-LAI combination was applied to investigate the climate sensitivity of vegetation under different climate scenarios in China.
The results demonstrated the following: (1) all four traits effectively captured the relationships between climate variables and the traits, as well as effectively predicted vegetation distribution and helped analyze environmental sensitivity; (2) the LMA-Nmass-LAI combination yielded an accuracy of 72.05%, providing more detailed parameter information regarding community structures and ecosystem function, and was therefore selected for training GMMs to predict vegetation distribution; and (3) a sensitivity analysis indicated that increasing temperatures shifted the boundaries of most vegetation northward and westward. Because the forests in these regions are well adapted to growth under rainy conditions, increasing precipitation is predicted to expand the boundaries of forests compared with the baseline vegetation distribution. Although the trait-climate relationship is not the only candidate useful for predicting vegetation distributions and analyzing climatic sensitivity, it sheds new light on developing the next generation of trait-based DGVMs.