Application of the bias aware Ensemble Kalman Filter with Confirming Option (Bias-CEnKF) in groundwater flow data assimilation

Abstract

The Ensemble Kalman Filter (EnKF) has been widely applied for real-time simulation of groundwater flow and solute transport. The majority of previous studies tend to assume no bias in forecast models, therefore ignoring the model uncertainties. This assumption, however, may be invalid when a conceptual model is not accurately generalized. As a result, forecast bias will lead to incorrect estimation of the system parameters or states. In this work, a bias aware Ensemble Kalman Filter with Confirming Option (Bias-CEnKF) is proposed to take into account the forecast bias by the model during the filtering process. The proposed method is tested in a real-time groundwater simulation considering model uncertainties, by setting inaccurate boundary conditions, initial conditions and recharge items. The results show that the standard EnKF may lead to filter divergence and assimilation failure, if the forecast model is not accurately generalized. Instead, Bias-CEnKF not only achieves better performances, but also reduces the inconsistency caused by the nonlinear relationship among the parameters, variables and bias corrections. Four scenarios are investigated, with the results showing the aquifer hydraulic conductivities and heads obtained by Bias-CEnKF are close to the real values, and the prediction results are also more reliable. This study further improves the applicability of the EnKF under the uncertain condition of model generalization, and provides a reliable method for groundwater data assimilation under complex field conditions.

Keywords

• data assimilation
• ensemble kalman filter
• conceptual model uncertainty
• forecast bias
• inconsistency