Numerical Simulations of Summer Precipitation in China during 1951-2000 Using a Regional Climate Model

In order to test the capability of the high-resolution regional climate model RegCM3, simulations of summer precipitation in China between 1951 and 2000 are carried out. In these simulations, NCEP/NCAR (National Center for Environmental Prediction and National Center for Atmospheric Research) reanalysis data are used as initial and boundary conditions. The model domain covers the whole China and surrounding areas, and centers at 32.5°N, 95°E. It has 160 and 95 grids in zonal and meridional directions, respectively, with a grid space of 60km. The Fritsch Chapper of Grell scheme is employed. Monfthly sea surface temperatures used are GIS2.3b from the Hadley Center. Simulations are carried out from May 1 to September 1 for each year. Analysis of precipitation focuses on the period of June-July-August. In order to test results, simulated results and observations are converted into grid data with 1° latitude multiply by 1° longitude. Seasonal and total rainfalls are compared with observations from 160 stations by the National Climate Center (NCC) to validate the model's performance in simulating the climatological mean and interannual variations of the summer precipitation. Main results are: (1) for the climatological mean precipitation over China, the simulated rainfall is close to observations, much better than that from NCEP/NCAR reanalysis data. (2) for the spatial distribution of precipitation anomalies, the simulated results over the Northeastern part of China are better than over other parts, and the simulated precipitations in the valley of the Yellow River and the Yangtze are the worst; (3) for the temporal evolution, the simulated results display decadal variations, and the simulated results over the 1960's and 1990's are better and stable, however, the results over the 1970's and 1980's show large variations; (4) simulated precipitation over the 1970's and 1980's is poor, heavier than observations, suggesting that the model is unable to generate observed decadal variations.