Estimating Field-Scale Hydraulic Parameters of Heterogeneous Soils Using A Combination of Parameter Scaling and Inverse Methods

Abstract

As the Hanford Site transitions into remediation of contaminated soil waste sites and tank farm closure, more information is needed about the transport of contaminants as they move through the vadose zone to the underlying water table. The hydraulic properties must be characterized for accurate simulation of flow and transport. This characterization includes the determination of soil texture types, their three-dimensional distribution, and the parameterization of each soil texture. This document describes a method to estimate the soil hydraulic parameter using the parameter scaling concept (Zhang et al. 2002) and inverse techniques. To this end, the Groundwater Protection Program Science and Technology Project funded vadose zone transport field studies, including analysis of the results to estimate field-scale hydraulic parameters for modeling. Parameter scaling is a new method to scale hydraulic parameters. The method relates the hydraulic-parameter values measured at different spatial scales for different soil textures. Parameter scaling factors relevant to a reference texture are determined using these local-scale parameter values, e.g., those measured in the lab using small soil cores. After parameter scaling is applied, the total number of unknown variables in hydraulic parameters is reduced by a factor equal to the number of soil textures. The field-scale values of the unknown variables can then be estimated using inverse techniques and a well-designed field experiment. Finally, parameters for individual textures are obtained through inverse scaling of the reference values using an a priori relationship between reference parameter values and the specific values for each texture. Inverse methods have the benefits of 1) calculating parameter values that produce the best-fit between observed and simulated values, 2) quantifying the confidence limits in parameter estimates and the predictions, 3) providing diagnostic statistics that quantify the quality of calibration and data shortcomings and needs, and 4) not restricting the initial and boundary-flow conditions, the constitutive relationships, or the treatment of heterogeneity. On this project, inverse modeling was performed using the combination of two computer models, one for forward flow modeling and the other for nonlinear regression. The forward model used to simulate water flow was the Subsurface Transport Over Multiple Phases (STOMP) numerical simulator (White and Oostrom 2000). STOMP was designed to solve a variety of nonlinear, multiple-phase, flow and transport problems for unsaturated porous media. The Universal CODE (UCODE) model (Poeter and Hill 1998) was used to perform inverse modeling posed as a parameter-estimation problem using nonlinear regression. Inverse techniques were applied to two cases of one-dimensional flow in layered soils and one case of three-dimensional flow in a heterogeneous soil. The results show that the simulation errors were significantly reduced after applying parameter scaling and inverse modeling. When compared to the use of local-scale parameters, parameter scaling reduced the sum of squared weighted residue by 93 to 96% for the relatively smaller scale (~2 m [~6.6 ft]) one-dimensional flow and 59% for the more complex Sisson and Lu site, which has the spatial scale of about 18 m (60 ft). This parameter estimation method will be applied to analyze the first two years of field experiments completed at the Sisson and Lu site.