Comparison of ensemble data assimilation methods for the estimation of time-varying soil hydraulic parameters

Abstract The hydraulic properties of the soil top layer may change during the growth period due to various factors such as wetting and drying cycles, tillage practices, and crop root growth. In this study, the potential of the assimilation method to estimate time-varying soil hydraulic parameters is explored. Four assimilation schemes, including the simultaneous update state augmentation method, the partitioned update state augmentation method, the simultaneous update parameter correction method, and the partitioned update parameter correction method, are compared. The performance of four assimilation schemes on parameter estimations and soil moisture simulations is tested first using the synthetic case. The influence of initial parameter values and a parameter update order on assimilation is also analyzed. Finally, the partitioned update parameter correction method is applied to a real case involving a field drip irrigation experiment. The results show that when the analyzed parameter has either a periodical or a linear variation, there is a time lag between the assimilation value and the true value. The assimilation method can respond immediately to an abrupt change of the parameter value. Using the simultaneous update method leads to an obvious parameter correlation problem. In contrast, the partitioned update method can relieve the parameter correlation problems, thereby improving the accuracy of parameter estimations and pressure head simulations. However, when the initial values of the parameters deviate from their true values to a certain extent, the partitioned update method cannot obtain accurate parameter estimations. Compared with the partitioned update state augmentation method, the partitioned update parameter correction method is not sensitive to the parameter update order. The partitioned update parameter correction method has higher computational efficiency and assimilation stability, and it can obtain more accurate parameter estimations and soil moisture predictions in comparison with the traditional state augmentation method. The partitioned update parameter correction method provides an assimilation tool for improving the predictions of soil moisture by considering the time-varying parameters.