Soil moisture content (SMC) is an important factor in agricultural productivity; it has an impact on crop growth, water use efficiency, and soil health. However, accurately predicting SMC, especially at deeper soil layers, remains challenging due to high variability and limited spatiotemporal data resolution. This study developed and evaluated a Random Forest Regression (RFR) model to predict SMC in loam soil at five different depths (5, 20, 40, 60, and 80 cm) utilizing meteorological data (temperature, humidity, precipitation, wind speed, and solar radiation) and vegetation indices: the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Moisture Index (NDMI). Data were collected during the maize vegetation season in 2023 in Mosonmagyaróvár, Hungary. The results showed that the mean SMC ranged from 12.61% to 16.19%. Correlation analysis demonstrated that precipitation and NDMI had the strongest positive correlation with SMC, especially at shallower depths r = 0.78 at 5 cm depth, Solar radiation had a moderate correlation with SMC, especially at the deeper depths. The RFR model performed well at all depths, achieving an R² of 0.86 at 5 cm depth; the model accuracy enhanced at deeper layers, achieving R² values of 0.91 and 0.94 at 60 and 80 cm depths, respectively. The most significant predictors according to the feature importance analysis were precipitation, humidity, and NDMI, with NDMI playing a crucial role in subsurface moisture retention at deeper depths. These findings highlight the potential for machine learning algorithms to optimize irrigation approaches and improve water management in precision agriculture.
Read full abstract