Currently, short wave radiation at the ground surface (GSW) is calculated under the assumption of a horizontal surface. This method of estimating the GSW may lead to considerable errors when the model resolution becomes higher and the model terrain becomes steeper. In this paper, to improve the short wave solar radiation simulations, a terrain slope and orientation parameterization has been implemented into the non-hydrostatic mesoscale model GRAPES (Global/Regional Assimilation and Prediction System). The effects of the terrain slope and orientation on different short range weather processes in China under differentmodel resolutions are simulated and discussed. In the simulations, topography height is taken from NCEP (National Centers for Environmental Prediction) with a resolution of 1 kin, and the slope and orientation of terrain are calculated using different staggering schemes and under different weather conditions. The results show that when the model resolution is low (30 and 60 kin) and the slope of terrain is not large, the influence of the slope and orientation of terrain on the GSW is not evident; otherwise, however, it is not negligible.Under high model resolutions (3 and 6 kin), the increase (decrease) of simulated precipitation corresponds to the decrease (increase) of the GSW induced by the slope effect, and the variations of precipitation are usually ranged between -5 and 5 mm. Under the high resolution, the surface temperature and heat fluxes are strongly correlated to each other and the high correlation exists mostly in the complex terrain regions. The changes of the GSW, precipitation, surface temperature, and heat fluxes induced by the effects of the terrain slope and orientation are more obvious in mountainous regions, due to the alternations in the atmospheric circulation. It is found as well that under the weather condition of less cloud and less precipitation, the effects of the terrain slope and orientation can be more realistically seen. Therefore, the terrain slope and orientation can usually be neglected in numerical models when the horizontal model resolution is low and the slopes are moderate, but should be taken into account when the model resolution becomes high and the terrain is steep and undulating.