Data-based modelling and environmental sensitivity of vegetation in China

A process-oriented niche specification (PONS) model was constructed to quantify climatic controls on the distribution of ecosystems, based on the vegetation map of China. PONS uses general hypotheses about bioclimatic controls to provide a "bridge" between statistical niche models and more complex process-based models. Canonical correspondence analysis provided an overview of relationships between the abundances of 55 plant communities in 0.1 degrees grid cells and associated mean values of 20 predictor variables. Of these, GDD(0) (accumulated degree days above 0 degrees C), Cramer-Prentice alpha (an estimate of the ratio of actual to equilibrium evapotranspiration) and mGDD(5) (mean temperature during the period above 5 degrees C) showed the greatest predictive power. These three variables were used to develop generalized linear models for the probability of occurrence of 16 vegetation classes, aggregated from the original 55 types by k-means clustering according to bioclimatic similarity. Each class was hypothesized to possess a unimodal relationship to each bioclimate variable, independently of the other variables. A simple calibration was used to generate vegetation maps from the predicted probabilities of the classes. Modelled and observed vegetation maps showed good to excellent agreement (kappa = 0.745). A sensitivity study examined modelled responses of vegetation distribution to spatially uniform changes in temperature, precipitation and [CO2], the latter included via an offset to alpha (based on an independent, data-based light use efficiency model for forest net primary production). Warming shifted the boundaries of most vegetation classes northward and westward while temperate steppe and desert replaced alpine tundra and steppe in the southeast of the Tibetan Plateau. Increased precipitation expanded mesic vegetation at the expense of xeric vegetation. The effect of [CO2] doubling was roughly equivalent to increasing precipitation by similar to 30 %, favouring woody vegetation types, particularly in northern China. Agricultural zones in northern China responded most strongly to warming, but also benefited from increases in precipitation and [CO2]. These results broadly conform to previously published findings made with the process-based model BIOME4, but they add regional detail and realism and extend the earlier results to include cropping systems. They provide a potential basis for a broad-scale assessment of global change impacts on natural and managed ecosystems.