Inverse estimation of hydraulic parameters of soils with rock fragments

Abstract

Listen

Translate article

The application of inverse modeling to determine the hydraulic properties of layered soils with rock fragments (RF > 2 mm) is associated with complex challenges such as selecting suitable hydraulic functions and defining a strategy to optimize many parameters. These two issues were addressed in this study by performing field drainage experiments in layered soil profiles (three layers) with different particle size distributions and different percentages of RF. Water contents (θ) and pressure heads (h) were monitored at different soil depths during drainage experiments (h ≥ −200 cm). Data of θ and h for a drier range (h < −5000 cm) were determined with a dew point potentiometer on disturbed soil samples. The unimodal and bimodal van Genuchten-Mualem functions were evaluated, and their parameters were optimized using the Hydrus-1D “Inverse solution”. The RETC curve fitting software was used to estimate the second set of parameters of the bimodal function. The parameters for the uni- and bimodal models (15 and 27 parameters, respectively) were repeatedly optimized by fixing some parameters while estimating others. The unimodal van Genuchten-Mualem function provided a satisfactory fit of θ and h measurements only when either drainage or dew point potentiometer measurements were used. On the other hand, the bimodal van Genuchten-Mualem function provided a satisfactory fit of θ and h measurements when both sets of data (drainage and dew point potentiometer measurements) were used simultaneously. Furthermore, the optimized parameters were physically consistent except for a few high saturated hydraulic conductivity values. Therefore, bimodal functions should be considered to represent the hydraulic properties of soils with RF. The problem of many calibrated parameters when using bimodal functions can be managed using sequential optimization, which allowed us to estimate 27 parameters for layered soils with RF successfully.