Assessing model state and forecasts variation in hydrologic data assimilation

Data assimilation (DA) has been widely used in hydrological models to improve model state and subsequent streamflow estimates. However, for poor or non-existent state observations, the state estimation in hydrological DA can be problematic, leading to inaccurate streamflow updates. This study evaluates the soil moisture and flow variations and forecasts by assimilating streamflow and soil moisture. Three approaches of Ensemble Kalman Filter (EnKF) with dual state-parameter estimation are applied: (1) streamflow assimilation, (2) soil moistue assimilation, and (3) combined assimilation of soil moisture and streamflow. The assimilation approaches are evaluated using the Sacramento Soil Moisture Accounting (SAC-SMA) model in the Spencer Creek catchment in southern Ontario, Canada. The results show that there are significant differences in soil moisture variations and streamflow estimates when the three assimilation approaches were applied. In the streamflow assimilation, soil moisture states were markedly distorted, particularly soil moisture of lower soil layer; whereas, in the soil moisture assimilation, streamflow estimates are inaccurate. The combined assimilation of streamflow and soil moisture provides more accurate forecasts of both soil moisture and streamflow, particularly for shorter lead times. The combined approach has the flexibility to account for model adjustment through the time variation of parameters together with state variables when soil moisture and streamflow observations are integrated into the assimilation procedure. This evaluation is important for the application of DA methods to simultaneously estimate soil moisture states and watershed response and forecasts.