Near-surface soil moisture assimilation for quantifying effective soil hydraulic properties using genetic algorithm: 1. Conceptual modeling

We used a genetic algorithm (GA) to identify soil water retention theta(h) and hydraulic conductivity K(h) functions by inverting a soil-water-atmosphere- plant (SWAP) model using observed near-surface soil moisture (0-5 cm) as search criterion. Uncertainties of parameter estimates were estimated using multipopulations in GA and considering data and modeling errors. Three hydrologic cases were considered: (1) homogenous free-draining soil column, (2) homogenous soil column with shallow water table, and (3) heterogeneous soil column under free-drainage condition, considering three different rainfall patterns in northern Texas. Results (cases 1 and 2) showed the identifiability of soil hydraulic parameters improving at coarse and fine scales of the soil textural class. Medium-textured soils posed identifiability problems when the soil is dry. Nonlinearity in theta(h) and K(h) is greater at drier conditions, and some parameters are less sensitive for estimation. Flow regimes controlled by upward fluxes were found less successful, as the information content of observed near-surface data may no longer influence the hydrologic processes in the subsurface. The identifiability of soil hydraulic parameters was found better when the soil profile is predominantly draining. In case 3, top soil layer hydraulic properties were defined using near-surface data alone as criterion. Adding evapotranspiration (ET) improved identification of the second soil layer, although not all parameters were identifiable. Under uncertainties, theta(h) was found to be well defined while K(h) is more uncertain. Finally, we applied the method to a validation site in Little Washita watershed, Oklahoma, where derived effective soil hydraulic properties closely matched the measured ones at the field site.