This study evaluates the performance of the regional climate model RegCM4, which incorporates the Biosphere–Atmosphere Transfer Scheme (BATS) and Community Land Model (CLM3) land-surface schemes, in simulating the summer precipitation over East Asia. The characteristics of summer precipitation are analysed in terms of mean amount, frequency and intensity of daily precipitation. The results show that the simulation of the summer precipitation is significantly sensitive to the choices of the land-surface schemes. Despite several deficiencies, the simulation of daily precipitation with CLM3 exhibits superior performance to that with BATS. The BATS simulation tends to systematically overestimate both precipitation frequency and intensity, and hence total precipitation across the whole domain. On the other hand, the CLM3 simulation substantially reduces the wet biases produced in the BATS simulation. The difference in performance between the two simulations mainly results from convective precipitation rather than large-scale precipitation. Since excessive convective precipitation tends to suppress large-scale precipitation, the BATS simulation also exhibits a limitation in properly simulating the ratio of convective and large-scale precipitation. Such behaviour can be explained by the influence of soil moisture on convective precipitation. Persistently wetter soil moisture in the BATS land-surface scheme can modulate the partitioning of surface heat fluxes inadequately, leading to overestimation of latent heat flux and underestimation of sensible heat flux over South China, in particular. Consequently, it affects the thermodynamic structure (as described by the stability indices), which in turn affects the atmospheric stability to determine the convective activity. The CLM3 simulation generates a more realistic representation of equivalent potential temperature, convective available potential energy and convective inhibition, and thus improves the characteristics of daily precipitation.