Journal of Hydrology, 2011, Vol.403(1), pp.141-156
► A surface-subsurface flow model with multi-objective global optimization is presented. ► The model performance was evaluated using bench-scale flow experiments. ► Inverse parameter estimation required observations at different spatial positions. ► The Pareto trade-off and model mismatch suggest lateral flow in soil layers. ► The model system is versatile for studying soil water and overland flow. A comprehensive description of water flow in environmental and agricultural systems requires an account of both surface and subsurface pathways. We present a new model which combines a 1D overland flow model and the 2D subsurface flow HYDRUS-2D model, and uses the multi-objective global search method AMALGAM for inverse parameter estimation. Furthermore, we present data from bench-scale flow experiments which were conducted with two 5-m long replicate soil channels. While rainfall was applied, surface runoff was recorded at the downstream end of the soil channel, subsurface drainage waters were sampled at three positions equally spaced along the channels, and pressure heads were recorded at five depths. The experimental observations were used to evaluate the performance of our modeling system. The complexity of the modeling approach was increased in three steps. First, only runoff and total drainage were simulated, then drainage flows from individual compartments were additionally evaluated, and finally a surface crust and immobile soil water were also considered. The results showed that a good match between measured and observed surface runoff and total drainage does not guarantee accurate representation of the flow process. An inspection of the Pareto results of different multiobjective calibration runs revealed a significant trade-off between individual objectives, showing that no single solution existed to match spatial variability in the flow. In spite of the observed crust formation, its consideration in the more complex model structure did not significantly improve the fit between the model and measurements. Accounting for immobile water regions only slightly improved the fit for one of the two replicate soil channels. Discrepancies between relatively complex model simulations and seemingly simple soil channel experiments suggest the presence of additional unknowns, such as heterogeneity of the soil hydraulic properties. Nevertheless, with its versatile subsurface options and powerful inverse method, the model system shows promise for studying hillslope flow problems involving both surface runoff and subsurface flow.
Overland Flow ; Surface Runoff ; Multi-Objective Global Parameter Optimization ; Mobile–Immobile Model ; Simulation ; Flow Channel Experiment ; Geography
View record in ScienceDirect (Access to full text may be restricted)