Wednesday, November 15, 2006 - 1:45 PM
278-3

A Multi-Objective Optimization Based Approach to Estimate Soil Hydraulic Parameters for Unsaturated Flow and Transport Modeling.

Navin Kumar Twarakavi1, Hirotaka Saito1, and Jirka Simunek2. (1) University of California-Riverside, 2208 Geology Building, Riverside, CA 92521, (2) University of California at Riverside, Environmental Science, Bourns Hall A135, Riverside, CA 92521

Numerical models of unsaturated water flow and solute transport rely on the robustness of the soil hydraulic parameter estimates. Soil hydraulic parameters used in most standard models include the saturated and residual water contents, so-called shape parameters and the saturated hydraulic conductivity. Numerical models predictions greatly improve when the estimated hydraulic parameters adequately represent the highly nonlinear relationships between the water content, pressure head and hydraulic conductivity. Often, one has to estimate these parameters from experimental data relating the water content and pressure head. One common approach is to estimate the van-Genuchten model parameters (van Genuchten, 1980) from the experimental data by non-linear least square technique using, for example, the RETC program (van Genuchten et al., 1991). It is assumed that fitting the van-Genuchten model to the water retention data allows optimal characterization of the hydraulic conductivity dependence on pressure through the van Genuchten-Mualem model. Considering the high non-linearity of these relationships, it may not be necessarily true. In this paper, we attempt to circumvent this problem by developing a multi-objective optimization (MO) approach to estimate the soil hydraulic parameters. The proposed approach (RETC-MO) considers different facets of the water content-pressure head-hydraulic conductivity relationship during parameter estimation. RETC-MO uses the MOSCEM-UA method which is a general-purpose global multi-objective optimization algorithm designed to infer the "Pareto optimal" parameter sets within a single optimization run.  RETC-MO was analyzed for the following case studies: (1) Retention data for selected soils from UNSODA  (Leij et al., 1996) and, (2) Las Cruces flow and transport experiment (Wierenga et al., 1990). It was observed that RETC-MO gives better estimates of soil hydraulic parameters than RETC for modeling vadose zone flow and solute transport. Apart from sufficiently fitting the water retention curve, RETC-MO shows improvements in representing the highly non-linear relationship between hydraulic conductivity and pressure head.