Abstract: To assess the effects of spatial resolutions of regional climate models that are used for the simulation of summer surface air temperatures (SATs) over China, the regional climate model RegCM3 is employed for a 5year (2000-2004) long simulation of summer SATs over a fixed domain within China by choosing 3 horizontal resolutions (30, 60 and 90 km) and 3 vertical resolutions (14, 18, and 23 layers). The assessment is carried out in terms of several parameters, e.g., simulated bias, root mean square error, and spatial correlation coefficient, and major conclusions are obtained as follows: (1) Simulated 5 year mean summer SATs generally show a sensitivity to spatial resolutions for the whole China, where the subareas of North China and Northwest China are most affected with maximum resolutioninduced differences of over 1℃, due mainly to the arid climate. (2) Increasing resolutions lead to an increasing sensitivity of simulated SATs to resolutions for the subareas above. (3) With higher horizontal resolutions, the model gives more detailed information for SATs, and better simulations of the temperatures at the regions with steep topography, and thus improves, as a whole, the SAT distributions over the subareas of China. However, increasing vertical resolution shows little effect on the SAT distributions. (4) SAT simulations do not always get closer to observations as resolutions increase. Proper configuration of horizontal and vertical resolutions is very important for a better simulation. In addition, increasing resolutions to quite a high level is necessary even for the regional climate research at the scale of South China and Northwest China. (5) Effects of resolutions on simulated temperatures do not show complete consistency with those on simulated precipitations. Although the influence of model resolutions may not be larger than that of model physics, this study still reveals the importance of choosing appropriate model resolutions as well as some uncertainties when a regional climate model is applied for the SAT simulation.