Dominant Mode of Climate Variability, Intermodel Diversity, and Projected Future Changes over the Summertime Western North Pacific Simulated in the CMIP3 Models

A set of multimodel twentieth-century climate simulations for phase 3 of the Coupled Model Inter-comparison Project (CMIP3) is analyzed to assess the model reproducibility of the Pacific-Japan (PJ) teleconnection pattern. It is the dominant low-frequency anomaly pattern over the summertime western North Pacific (WNP), characterized by a meridional dipole of zonally elongated vorticity anomalies in the lower troposphere and by anomalous precipitation over the tropical WNP. Most of the models can reproduce the PJ pattern reasonably well as one of the leading anomaly patterns or their combination. The model reproducibility of the pattern tends to be higher for those models in which the climatological-mean state over the WNP is better reproduced. Furthermore, intermodel diversity in the summertime climatological-mean fields over the WNP, especially in the lower troposphere, is found to be large and projected most strongly onto the observed PJ pattern. Nevertheless, the multimodel ensemble (MME) mean of these climatological-mean states is close to the observations. Projected future changes in the summertime climatological-mean state under the Intergovernmental Panel on Climate Change's (IPCC) Special Report on Emission Scenarios (SRES) A1B also bear certain similarities with the PJ pattern, in a manner consistent with the aforementioned sensitivity of the model climate to that pattern. The MME projection indicates an overall increase in precipitation over the entire tropics, but it is overwhelmed locally by the effects of the enhanced tropospheric stratification over the tropical WNP. A resultant local reduction of the mean ascent is dynamically consistent with the anticyclonic projection around the Philippines and the cyclonic projection around Japan in MME, as in the observed anomalous dipole associated with the PJ pattern. However, the polarity and magnitude of the PJ-like projected change vary substantially from one model to another.