Development of pedotransfer functions for soil moisture prediction using soil and remote sensing data at the watershed scale

Abstract

Soil moisture plays a crucial role in agriculture, hydrology, and erosion control, especially in semi-arid regions. Direct measurement of soil moisture is costly and time-consuming, prompting the use of pedotransfer functions (PTFs) for predicting field capacity (FC) and permanent wilting point (PWP). This study aimed to advance new PTFs, which are models used to estimate soil moisture properties from easily measured soil data, for predicting FC and PWP soil properties and remote sensing data coupled with regression. 100 soil samples from four land uses were analysed for bulk density (BD), texture, organic matter (OM), and calcium carbonate (CaCO3). A GIS was used to extract five spectral indices from Landsat 8 satellite data to improve model predictions. Three scenarios were tested using soil properties (Scenario I), using spectral indices (Scenario II), and combining both soil properties and spectral indices (Scenario III). Strong correlations were found between %clay and FC (r = 0.57) and PWP (r = 0.62), while BD negatively correlated with FC (r = − 0.66) and PWP (r = − 0.54). FC and PWP were also significantly correlated with SAVI (r = 0.46) and NDVI (r = 0.45). Scenario III, integrating soil properties and spectral indices, yielded the most accurate predictions, with R2 of 0.85 for FC and 0.77 for PWP, compared to Scenario I (R2 of 0.82 for FC and 0.70 for PWP) and Scenario II (R2 of 0.54 for FC and 0.63 for PWP). This combined approach enhances soil moisture prediction, aiding sustainable agriculture and land-use planning in semi-arid regions.