Two-dimensional simulation of water flow and solute transport below furrows: model calibration and validation

In this study a two-dimensional numerical flow/transport model (HYDRUS-2D) was calibrated and experimentally validated using data from long furrow irrigation experiments. The model was calibrated using data from an experiment carried out assuming free-draining (FD) outlet conditions, and subsequently validated against data from three experiments assuming blocked-end conditions. The data were analyzed using the Richards' equation for variably saturated flow and either the traditional convection-dispersion equation (CDE) or the physical nonequilibrium mobile-immobile (MIM) model for solute transport. Optimization was accomplished by means of Levenberg-Marquardt optimization in combination with the HYDRUS-2D model. Simultaneous and two-step optimization approaches were used to estimate the soil hydraulic and solute transport parameters near the FD furrow inlet and outlet sites. First, the saturated hydraulic conductivity (K-s) and CDE or MIM solute transport parameters were estimated simultaneously. We also used sequential (two-step) estimation in which we first estimated the soil hydraulic parameters followed by estimation of the solute transport parameters. In the two-step method, the saturated soil water content (theta(s)), the n parameter in van Genuchten's soil water retention model, and K-s values were estimated during the first step, and the CDE or MIM solute transport parameters during the second step. Estimated soil hydraulic and solute transport parameters were found to vary substantially between the inlet and outlet sites. Estimated CDE and MIM transport parameters were very similar for both optimization approaches. The two-step method significantly improved predictions of the soil water content during model calibration, while the solute concentration predictions were nearly the same for both approaches, with both not providing a good description of the observed concentrations. Solute data were also analyzed using horizontal averages to somewhat lessen the effects of spatial variability. Horizontally averaged concentration distributions showed better agreement with the predictions. Soil water contents for the three blocked-end experiments during model validation were well predicted. The two-step method produced slightly better agreement with observed data. However, both optimization approaches produced relatively poor agreement between measured and predicted solute concentrations and deep percolation rates. (C) 2003 Elsevier Ltd. All rights reserved.