Fitting process-dependence performance of the van Genuchten soil water retention model to simulate the soil water flow

Abstract

Determining the parameters of the soil water retention curve (SWRC) is essential to model the water flow in soil and investigate the water and nutrients transport around the roots of plants and ensure optimal management of irrigation. The unweighted least squares regression (ULS) is the most common approach applied to fit the SWRC functions to the observed data-points to optimize their parameters. The main purpose of this study was to evaluate the ULS and the weighted least squares regression (WLS) fitting process of SWRC and their impact on simulations of soil water flow. In this regard, the measured SWRC data in six parallel samples from given depth were fitted to the SWRC equations to optimize their parameters, through either the WLS or the ULS. In the WLS approach, the weights were calculated as the inverse of the variance in each pressure head of different iterations (six repetitions in this study). The results showed that although WLS generally resulted in an increased error in estimating the SWRC in the van Genuchten-Mualem model (RMSE = 0.027 and 0.043 cm3 cm−3 in the ULS and WLS methods, respectively), it improved the accuracy of the estimations at lower water contents (dry-end), as compared to the ULS. However, the efficiency of the ULS and WLS in estimating the SWRC was different from that in simulating the soil water flow. The water flow in the soil was more accurately simulated using the hydraulic parameters obtained by the WLS (with RMSE = 0.029 and 0.026 cm3 cm-3 at x = 0 and x = 20 cm, respectively, in the WLS, and RMSE = 0.062 and 0.065 cm3 cm-3 at x = 0 and x = 20 cm, respectively, in the ULS), especially at the low water range. While there was an important difference between the ULS and WLS in terms of estimating the SWRC and water flow simulations, other dispersion scenarios should be evaluated in future studies to compute the error of the fitting processes at low-pressure heads.