Estimation of the Soil Moisture Content in a Desert Steppe on the Mongolian Plateau Based on Ground-Penetrating Radar

Desert grasslands are a crucial component of terrestrial ecosystems that play vital roles in regional and global hydrological cycling, climate change, and ecosystem balance through variations in their soil moisture content (SMC). Despite this, current research on the SMC of desert grasslands remains insufficient, with many areas remaining underexplored. In this study, we focused on a typical desert grassland located in the northern foothills of the Yinshan Mountains. Ground-penetrating radar (GPR) exploration and soil sampling were used to test existing mixed-media models, and a new mixed-media model was calibrated using cross-validation methods. Among the three general mixed-media models, the Topp and Roth models yielded more accurate SMC estimates for the study area, with root mean square errors of 0.0091 g/cm3 and 0.0054 g/cm3, respectively, and mean absolute percentage errors of 25.86% and 19.01%, respectively, demonstrating their high precision. A comparison of the calibrated and original mixed-media models revealed that the estimation accuracy was significantly improved after parameter calibration. After parameter calibration, the Ferre model achieved an accuracy comparable to that of the Topp model. Parameter-calibrated models can be used to estimate the SMC using GPR data, offering a higher precision than general models and possessing greater suitability for the study area. The soil in the study area is primarily composed of sand particles and is therefore more compatible with the parameters of the Topp model, whereas the Ferre model requires further parameter calibration to achieve effective application.