This paper presents the result of a cooperative study under the Integrated Research Project of FORMOSAT-3/COSMIC and SoWMEX (Southwest Monsoon Experiment) in Taiwan. We focus on the comparison and impact studies of FORMOSAT-3 GPS radio occultation and dropsonde data on regional weather prediction. The WRF model and its 3DVAR component, WRF Var, are used for evaluation on the forecast of heavy rainfall events in Taiwan during an 11-day period of the 2007 Mei-yu season. The result shows that the GPS data tend to improve the simulations for longer integration, e.g., after 36 h. The best performance is on correcting the under-prediction of geopotential height. The dropsonde data help to improve the simulation only at earlier forecast time and at higher levels (e.g., 500 and 300 hPa). With both the GPS and the dropsonde data assimilated together, the simulation shows even more improvement. At 12-24 h, there is no impact of GPS and dropsonde data on rainfall forecasts. However, when the integration time getting longer, the GPS and dropsonde data starts to help the rainfall simulation, especially for small rain thresholds.