In order to investigate the impact of hydrological data assimilation scenarios on streamflow simulations of hydrological models, a data assimilation framework was proposed based on the Xin'anjiang model and the ensemble Kalman filter, in which remotely sensed evapotranspiration products and measured streamflow were used as observation data. Based on this framework, four assimilation scenarios including DA-ET, DA-ET(K), DA-ET-Q, DA-ET-Q(K) and a comparative scenario (i.e., OL) were designed and tested in Ganjiang River Basin, aiming to evaluate the impact of the time resolution of remote sensing evapotranspiration products, whether the time-parameters related to model evapotranspiration and multi-source data assimilation on runoff simulation after hydrological data assimilation. The results showed that assimilation of two evapotranspiration products with different time resolutions could improve the accuracy of model's streamflow simulations in the DA-ET scenario, and model performed better by assimilating the products with higher temporal resolutions. On the basis of DA-ET scenario, adding measured streamflow into assimilation could improve the accuracy of model's streamflow simulations, and the reduction of the relative error obtained by DA-ET(K) and DA-ET-Q(K) scenarios was more than 20%. The results indicated that the accuracy of streamflow simulations was improved through assimilation of ET products, especially when the model evapotranspiration parameter was treated to be time-varying. Compared to the OL scenario, four assimilation scenarios could improve the accuracy of model's the 10% highest streamflow simulations in different degrees, but the other scenarios with no significant difference was better than DA-ET-Q(K) scenarios. Our results could further understand the differences of streamflow simulation in different data assimilation scenarios and provided scientific basis for the efficient utilization and management of water resources.