Near-Surface Soil Moisture Assimilation to Quantify Effective Soil Hydraulic Properties Using Genetic Algorithm: A Field Scale Numerical and Validation Study

Most large-scale soil-vegetation-atmosphere-transfer (SVAT) models rely on the use of soil hydraulic functions to describe the behavior of soil moisture in the unsaturated zone. The question of whether insitu or remotely sensed near-surface soil moisture temporal data is capable of quantifying the effective soil hydraulic properties is yet to be answered. In this paper, we present a near-surface soil moisture assimilation technique that can be possibly used to estimate the effective soil hydraulic parameters of the root zone. The procedure is based on inverse modeling using genetic algorithm to invert the soil moisture movement equation. Two major case studies were conducted: case (1) a field scale numerical study, and case (2) a field scale validation study. Case 1 is important for this kind of analysis because it can serve as a benchmark for further analysis. Under an error-free scenario, the only sources of uncertainties are the inverse procedure itself and parameter correlations and sensitivities. We found that the identifiability of parameters increases as we approach the outer ranges of the soil textural class. We also found that parameter identifiability is higher when the soil is predominantly drying than when being dominated by upward flux from a shallow water table. When the contribution of the upward flux exceeds 50 percent of the seasonal evapotranspiration (ET), the near-surface soil moisture assimilation fails. For Case 2, we used in situ datasets from SGP97 and SMEX02 hydrology campaigns. In real-world conditions, uncertainties in measurement, model, boundary conditions, etc. could influence the outcomes of the experiments. We examined some of these sources of uncertainties in our analyses and observed how the solutions behaved. Generally, the uncertainties pertaining to initial and bottom boundary conditions impact significantly the available mass of water in the soil profile and tend to underestimate soil moisture in the subsurface layers even if the simulated near-surface soil moisture fitted well with the observed data. Likewise, if they are not well represented, root length and density impacted the simulated sub-surface soil moisture. Our results demonstrate that using the right combination of conditions to define the effective modeling domain is important in implementing near-surface soil moisture assimilation for real-world conditions.