全球变化知识资源中心

In this paper, the fidelity of 28 models under Coupled Model Inter-comparison Project Phase-5 is examined for the Indian summer monsoon for the historical period from 1975 to 2005. It is found that all models simulate the spatial distribution of the seasonal mean surface air temperatures (T-as) quite well (pattern correlation >0.75), whereas the simulation of precipitation is found to be relatively poor (correlation 0.1-0.7). Most models underestimate the T-as with more bias during winter and less bias during summer. In regard to precipitation, most models fail to capture the observed contribution ratio of convective and large-scale precipitation (LSP) and simulate more convective precipitation as compared to the LSP. Extremely large wet (dry) biases are noted in convective (large-scale) precipitation. The total precipitation is also noted to have a large dry bias in most models, which is mainly due to the large dry bias in the LSP. Contrary to the notion that better simulation of the contribution ratio would lead to better simulation of total precipitation or vice-versa, our results show that both of these notions are not valid for most models. In observations, the LSP dominates the annual cycle of the total precipitation, whereas in models, the convective component dominates. In few models, the annual cycle in the individual precipitation component is either weak or completely missing. None of the models are found to simulate the observed trend in precipitation and temperature. The model with the highest resolution, MIROC-4h, simulates many of the observed features better than the other models, thereby emphasizing the usefulness of finer resolutions in better simulation of Indian monsoon. A comprehensive list of models has been prepared on the basis of their capability in simulating various features of Indian summer monsoon. The multimodel mean of the better models identified in this study is expected to produce more reliable projections of the Indian monsoon.