Random forest of Classification and Regression Tree (CART) in the estimation of SWC based on meteorological inputs and hydrodynamics behind

This study employs the random forest algorithm of Classification and Regression Trees (CART) to estimate soil water content (SWC) at shallow depths in a grassland terrain site. Leveraging meteorological parameters, the random forest model demonstrates efficient and effective SWC estimation in a 12-folds time series cross-validation. The results reveal distinct strategies employed by the CART model for different SWC depths, addressing seasonal variations and SWC sensitivity to precipitation. Additionally, the study highlights limitations of CART in extrapolating beyond training data, leading to misfit in certain scenarios. These findings contribute valuable insights for improved SWC estimation and agricultural practices.