To examine the sensitivity of cumulus parameterization scheme in climate simulating over the Tibetan Plateau, three cumulus parameterization schemes in RegCM3 (The ICTP Regional Climate Model, V3.0), Anthes-Kuo (AK), Grell-Fritsch&Chappell (GFC), and MIT-Emanuel (MIT) are compared and analyzed in temperature and precipitation simulation over the Tibetan Plateau. The results show that AK, GFC and MIT schemes well reproduce the distribution of temperature and precipitation over the Tibetan Plateau, and the spatial distribution and annual change of the simulated temperature and precipitation with AK scheme are more closed to observations than other schemes. The simulated temperature shows that there is a systematic cold bias over the Tibetan Plateau about the three schemes. There is a cold bias about 4℃ over Tibet and a warm bias about 2℃ over Tarim and Junggar Basin, the cold bias over the south is large than the north over Tibet, and the variability of temperature is slightly smaller than the observed values. Precipitation simulation results indicate that the simulated precipitation is 0－6 mm/d more than observed one, the simulated one is high in the north and low in the south, the variability of precipitation is also slightly smaller than the observed values. In addition, RegCM3 distinctly improved the MICRO3.2 simulation results, the distribution details of temperature and precipitation are better than the MICRO3.2 simulations for complex terrains, such as Tianshan, Tarim Basin, Junggar Basin, Kunlun Mountain, Gangdise Mountain, and southeastern Tibetan Plateau, which implies that the RegCM3 performance is successful in downscaling.