SENSITIVITY OF GLOBAL VEGETATION MODELS TO PRESENT-DAY CLIMATE SIMULATED BY GLOBAL CLIMATE MODELS

This paper provides an evaluation of the reliability of several climat e models with respect to ecosystem modeling and describes a methodolog y for undertaking such an evaluation. The global climate models (GCM) and global vegetation models used in this study are linked in a one wa y mode (i.e., no feedback from the vegetation model to the GCM is allo wed), and the vegetation distributions are compared with those obtaine d using observed climatology. The aim of this study is to identify whi ch simulated ecosystems are sensitive to the biases of the climate sim ulations. Two global static vegetation models, BIOME-1 and a version o f the Holdridge scheme, are used in conjunction with several present-d ay climate simulations. The climate simulations employed come from the GCMs participating in the Model Evaluation Consortium for Climate Ass essment project. The results indicate that the overall performance of coarse resolution GCMs with respect to vegetation prediction is poor. The discrepancies between vegetation distributions computed from obser ved and simulated climatologies represent more than 50% of land area. The comparison of vegetation distributions shows that there are some c ommon tendencies amongst these GCMs to induce the overprediction or un derprediction of certain biomes. For example, the biomes belonging to dry climate regions are underpredicted, and the woodlands and temperat e/cold forests are overpredicted. The climatic variables responsible f or the discrepancies between vegetation predictions are identified, an d it appears that the differences in vegetation predictions are overal l due to the overestimation of the soil moisture index and precipitati on, to the overestimation of growing degree days, and to the underesti mation of the annual minimum temperatures. In summary, this research h as shown that the prediction of biomes using simulated climatologies i s not yet fully satisfactory; however, it is possible to increase our level of confidence in the prediction of vegetation by carefully evalu ating the performance of the vegetation models driven by simulated cli matologies and by identifying the causes of the biases.