Abstract: This study evaluates the performance of FGOALS (Flexible Global Ocean-Atmosphere-Land surface-Sea ice coupled model) with ocean assimilation in the simulation of summer rainfall-SST relationship during 1979-2005 in the western North Pacific (WNP), and compares the results with corresponding simulations forced by observed sea surface temperature and FGOALS historical simulation. Results show that the FGOALS with ocean assimilation well captures the interannual variability of summer SST over the WNP except that over east of the Philippines. For the interannual variability of precipitation, it barely demonstrates any skill over Asian summer monsoon region, which is comparable to the AMIP (Atmospheric Model Intercomparison Project) simulation. However, for the summer rainfall-SST relationship, the observed negative correlations over South China Sea and east of Philippines are partly reproduced in the FGOALS with ocean assimilation, in particular when the precipitation leads SST by one month and is concurrent with SST. The simulated skill is better than the AMIP simulation, but is inferior to the historical simulation. Based on observations, anomalous convection and circulation in the summer over the WNP are primarily driven by SST anomalies over the area near the dateline and the eastern Indian Ocean-Maritime Continent. The induced anomalous convections affect solar radiation reaching the sea surface, which contributes significantly to local SST anomalies and leads to negative SST-rainfall correlation and SST tendency-rainfall correlation. In the AMIP simulation, the anomalous circulation over the WNP driven by the remote forcing is underestimated. Since the AMIP simulation is forced by observed SST, the anomalous convection and circulation are forced by underlying SST over some places of the WNP, leading to positive rainfall-SST correlation. Although the anomalous circulations over the WNP driven by the remote forcing are also underestimated in both FGOALS with ocean assimilation and historical simulation, weaker than observed negative SST-rainfall correlations are produced since local air-sea coupling is included. In addition, the historical simulation tends to overestimate the forcing from SST anomalies over the WNP south of 20°N, which leads to better simulated SST-rainfall correlation than the FGOALS with ocean assimilation over South China Sea and south of Japan islands.