Calibrating land hydrological models and enhancing their forecasting skills using an ensemble Kalman filter with one-step-ahead smoothing

The abundance of satellite remotely sensed data in the past few decades has provided a great opportunity to improve hydrological models’ simulations and their forecasting skills. This, however, requires an advanced data integration strategy, which is usually implemented using a data assimilation approach. In this study, a recently proposed assimilation method, the Unsupervised Weak Constrained Ensemble Kalman Filter (UWCEnKF), is extended to calibrate model parameters simultaneously with the state. The derivation of the new method is based on a One-Step-Ahead (OSA) smoothing formulation of the standard joint state-parameter filtering problem, which results in a dual-type filtering scheme separately updating the state and parameters using two interactive Ensemble Kalman Filters (EnKFs). The new calibration and assimilation method comprises three main steps at each assimilation cycle: (1) calibrate the parameters based on the observations, (2) update the system state based on the calibrated parameters and observations, and (3) enforce the water budget constraint. Numerical experiments based on assimilating multiple datasets simultaneously into a hydrological model are carried out to assess the performance of the proposed approach over different basins and over two testing periods: calibration and forecasting. Assimilation results suggest that the new filtering algorithm successfully improves the simulated water components during both the calibration and forecasting periods. These improvements are the result of the effective assimilation-calibration procedure introduced by the proposed method.