The performance of the NCAR regional climate model (RegCM) for east Asia, where topography and shoreline are rather complex, is examined through experiments to simulate the climate during 1 month using ECMWF data as lateral boundary conditions, before its application in the nested GCM/RegCM method to predict future climate changes caused by global warming. In this study, June and January climates, in which typical precipitation phenomena in Japan such as bai-u and winter snow are observed, are simulated using the model with lateral resolutions of 50 km (Base case) and 25 km (High case). The resolution effects of the model are also examined using a series of sensitivity studies. The main results are as follows: (1) In June and January, while cyclones passing over the inner region tend to be more intensified in the model (sometimes unrealistically) than those observed, weak cyclones in the outer region are usually not well simulated in the model. This seems to be due to the stronger control by the lateral boundary conditions. Anticyclones are stronger in the simulation than those observed, especially in the inner region, which leads to overestimation of the sea level pressure there. (2) The model yielded a lower surface air temperature than that observed, especially in January, which may depend on the performance of BATS or the radiative transfer scheme. (3) The model tends to overestimate the regional mean precipitation in the inner region and underestimate it in the outer region in June, while it was slightly underestimated in the inner and outer regions in January. Overestimations are caused by overdeveloped simulated cyclones or by the large amount of precipitation on the unrealistic topography such as a steep slope facing a moist tongue. Underestimation is due to stronger control by the lateral boundary conditions that prevent the development of cyclones or fronts. (4) In the high-resolution models, weak cyclones in the outer region and anticyclones in the inner region are more realistically simulated, although they do not greatly improve the model results in the Base case. Precipitation is increased by ~10--15% of that in the Base case in the inner region in June and January due to enhancement of cyclones or fronts, where the high-resolution effect of topography is only ~1/5 of the total high-resolution effect of the model in June. In January, for northwestern Japan, high model resolution contributes to the correction (increase) of precipitation there. Generally, the high-resolution effect of the model on the surface air temperature does not systematically improve the results but does vary locally by the improvement of model topography. To improve the model results, the lateral boundary should be moved outward (the domain of the model should be extended to the west and south) and/or the precipitation scheme should be improved, including the adjustment of parameters. BATS or the radiative transfer scheme should be improved. More recommendations for the improvement of the model are proposed. ¿ 1999 American Geophysical Union