Abstract: A new evapotranspiration data assimilation system was proposed by using the ensemble Kalman filter to assimilate the remote sensing evapotranspiration and simulated evapotranspiration by the Xin'anjiang model. The assimilated evapotranspiration was then used to estimate the soil moisture content in Xin'anjiang model via Particle Swarm Optimization (PSO) to improve the accuracy of runoff simulation. The Hanjiang River basin in China was used as a case study. The remote sensing evapotranspiration (ET_SEBS) based on the surface energy balance system model (SEBS) was validated by the water balance evapotranspiration (ET_GRACE) calculated through the water balance equation based on the GRACE water storage anomaly data, and compared with the other evapotranspiration productions (i.e., ET_GLDAS, ET_Zhang and ET_MODIS). The results showed that ET_SEBS outperformed the other evapotranspiration productions when ET_GRACE was used as the reference evapotranspiration, with three statistical criterion (R, E_RMSand B values of 0.93, 11.93 mm/month and -3.47 mm/month, respectively. Furthermore, the proposed assimilation system was applied to the Xunhe River basin, a tributary of Hanjiang River. The results indicated that during the period 2005-2007, the Nash-Sutcliffe efficiency coefficient (E_NS) was 0.85, which was higher than the E_NS value of 0.81 without assimilation, and the evapotranspiration assimilation system improved the accuracy for runoff simulation with a slightly improvement during the drought period and a remarkable improvement during wet period, particularly for the peak values.