Derivation of soil hydraulic properties (SHPs) using a Physics-Based inverse calibration method and International soil moisture network database

Abstract

This study used extensive soil moisture records to estimate “inverse-calibrated Soil Hydraulic Properties (SHPs)” using a multi-processing technique via high-performance computing clusters. Within this objective, a mass conservative numerical model was developed to solve the one-dimensional Richards Equation incorporating two different soil hydraulic models: the well-known van Genuchten Mualem (VGM) model and the relatively new Fredlund-Xing-Wang (FXW). A multiprocessing version of the Differential Evolution Algorithm (DEA) optimization technique was used for inverse calibration of the soil hydraulic parameters. For FXW, calibration statistics were calculated as means of the KGE’ (0.89 ± 0.1 and 0.83 ± 0.23), R (0.89 ± 0.1 and 0.85 ± 0.21) and ubRMSE (0.017 ± 0.01 and 0.015 ± 0.02) for the depths 50 and 100 cm, respectively. For VGM, calibration statistics were found as means of the KGE’ (0.87 ± 0.11 and 0.78 ± 0.22), R (0.90 ± 0.08 and 0.86 ± 0.17) and ubRMSE (0.019 ± 0.01 and 0.017 ± 0.01) for the same depths, respectively. The employed methodology had highly promising statistical performance for both FXW and VGM to derive SHPs. A comprehensive validation methodology was used to evaluate the reliability of derived SHPs. Correlation analysis showed that derived SHPs strongly correlated with the soil properties and environmental variables. Further, as a validation procedure, initial investigations were also conducted to explore the spatial transferability of the parameters. Despite the use of basic k-means clustering, the resulting soil hydraulic datasets showed statistical similarity or even improvement to hyper-resolution maps used in the literature. While the simulation model of the methodology has certain assumptions and limitations, this study proves that the ISMN database can be used to derive soil hydraulic properties and transfer these parameters to locations other than the calibration points. This study shows that FXW is a promising hydraulic model for the determination of soil moisture at root zone within the complete moisture range. The methodology can also be readily extended to other established soil moisture monitoring networks and potentially extended versions of “inverse-calibrated SHPs” and trained pedotransfer functions are considered to be valuable tools to estimate soil moisture profiles at the root zone.