An assessment of the benefit of additional physical site characterization for modeling the field-scale transport of solutes in a heterogeneous unsaturated soil.

Abstract

The migration of chemical substances in an unsaturated field soil is a process that occurs in many environmental problems, whose solutions often rely on decisions made based on computer simulations of that phenomenon. The purpose of this study was to investigate the benefit of gathering different types and amounts of information on soil transport properties, using the Las Cruces Trench database, so that the description of their spatial variability in the field can be used to enhance the confidence with which predictions of the migration of a conservative solute through the unsaturated zone are made. The two soil transport properties of interest were the soil water retention curve, theta(psi), and the saturated hydraulic conductivity, <italic> K_s</italic>. The methodology involved the creation of stochastic realizations of these properties through geostatistical simulation, the use of the Miller-Miller and of the Leverett scaling techniques and of a multi-step approach based on a prior categorization of soil water retention curves. These realizations of soil properties were then used in a two-dimensional numerical simulator of unsaturated flow and the simulated plumes were analyzed in terms of differences in their first and second spatial moments as well as for times of arrival. The results indicated that: (1) scaling techniques oversimplify the description of heterogeneity, and, consequently, produced a very narrow space of uncertainty; (2) the use of the multi-step approach produced a more realistic representation of the heterogeneity of soil properties, which increased the uncertainty in the solute plume's moments and in the times of arrival; (3) under Leverett scaling assumptions and log-normal distribution of <italic>K_s</italic>, funds should be allocated to collection of samples and/or measurements of <italic>K_s</italic> only to the extent where its mean and variance are well characterized; (4) for the multi-step approach, a large number of samples needs to be collected to fully characterize the variability in soil water retention curves. If further collection of samples results in the increase in the uncertainty about movement of the solute plume, achieving a reduction in this uncertainty by collecting an even larger number of samples might not be economically feasible; (5) soil initial conditions and the solute infiltration rates have only a secondary role in the prediction of solute migration, compared to that of the description of the soil transport properties. These results indicate that measurements of both water retention curves and saturated hydraulic conductivity need to be collected in the field, and that these properties must be analyzed separately to achieve realistic representations of the subsurface and to obtain conservative measures of uncertainty of metrics such as spatial moments and times of first arrival.