Tropical cyclone (TC) plays an important role in climate research.However,due to the relatively low model resolution,there remain challenges in simulating the distribution of TC by using global climate models.Dynamical downscaling has been a useful tool for TC research.In this paper,a regional climate model RegCM3 was used to dynamically downscale the historical TC simulation of FGOALS-g2(grid-point version 2 of Flexible Global Ocean-Atmosphere-Land System model),a global climate system model developed by LASG/IAP.A tropical cyclone track detection method was applied to reveal the added value of dynamical downscaling in simulating track distribution,intensity and rainfall of TCs.The result of dynamical downscaling shows distinct improvements in simulating the TC distribution,intensity and associated rainfall.The pattern correlation coefficient of TC track distribution is improved from 0.53 to 0.74 by the regional model.The underestimated TC intensity from FGOALS-g2 simulation is enhanced in RegCM3(Version 3 of the Regional Climate Model) by 20 m s⁻¹.In addition,the total rainfall contribution and intensity are enhanced by 10% and 4.7 mm d⁻¹ in RegCM3 compared to that in the global model over the western North Pacific region.In addition to the influence of increasing model resolution,large-scale variables such as increased moisture,enhanced relative vorticity and ascending motion,and weakened vertical wind shear are responsible for the improvements of TC simulating skill in RegCM3 as a result of better reproduction of CISK (Conditional Instability of the Second Kind) mechanism.