The invention discloses an evapotranspiration data assimilation method based on a distributed time varying gain hydrological model. The method includes the following steps that first, data preparation is conducted, and daily evapotranspiration is obtained based on a remote sensing model and flux observation data to be used for 'observation'; second, an initial ambient field is generated and model simulation is driven; third, when 'observation' data occur, an observation field and a prediction field are obtained; fourth, an analysis field is acquired in an assimilation mode through an ensemble Kalman filtering algorithm; fifth, the ambient field is updated. According to the method, through the data assimilation technology, an evapotranspiration result with high precision is used as observation information, the hydrological model is adjusted to operate, error accumulation is reduced, and an evapotranspiration sequence with high precision and continuous time is simulated. Compared with the prior art, the evapotranspiration can be assimilated directly through the method, the model nondeterminacy can be lowered, operating is easy, a water cycle physical mechanism is realized, and the method can be widely applied to accurately estimating and continuously acquiring region evapotranspiration.