Implementation of evapotranspiration data assimilation with catchment scale distributed hydrological model via an ensemble Kalman Filter

Abstract

errestrial actual evapotranspiration (ETa) is a crucial component of terrestrial water cycles. The most common methods of estimating catchment-scale ETa are remote sensing and hydrological models. These methods have limitations when applied at the catchment scale due to coarse resolutions of data and uncertainties in model predictions. Data assimilation techniques that combine complementary information from hydrological models and observed data can overcome some of the limitations. These techniques have been used in many hydrological modeling studies, but few have applied data assimilation to ETa within a distributed precipitation-runoff catchment modeling framework. This paper proposes a catchment scale ETa data assimilation technique (termed ET-DA) that assimilates remotely sensed ETa data into a distributed time-variant gain hydrological model (DTVGM-ET) for improving hydrological model simulations. The DTVGM-ET improved the ETa computation on the basis of a nonlinear soil water availability function to establish an explicit time response relationship between ETa and soil moisture for implementing the ETa assimilation. The proposed ET-DA system was tested in the Upper Huai River Basin (UHRB), China using data from 2000 to 2012. Through synthetic simulation experiments, the capability of ET-DA for obtaining accurate, continuous time series of ETa estimates and achieving assimilation feedback on soil moisture and streamflow was examined. The results demonstrated that ET-DA provided improved regional ETa monitoring capability, and assimilation of ETa into the hydrological model led to improved model predictions of soil moisture and streamflow.