Use of multiple LIDAR-derived digital terrain indices and machine learning for high-resolution national-scale soil moisture mapping of the Swedish forest landscape

Spatially extensive high-resolution soil moisture mapping is valuable in practical forestry and land management, but challenging. Here we present a novel technique involving use of LIDAR-derived terrain indices machine learning (ML) algorithms capable accurately modeling at 2 m spatial resolution across the entire Swedish forest landscape. We used field data from about 20,000 sites Sweden to train evaluate multiple ML models. The predictor features (variables) included suite generated national LIDAR digital elevation model ancillary environmental features, including surficial geology, climate use, enabling adjustment class maps regional or local conditions. Extreme gradient boosting (XGBoost) provided better performance for 2-class model, manifested by Cohen’s Kappa Matthews Correlation Coefficient (MCC) values 0.69 0.68, respectively, than other tested methods: Artificial Neural Network, Random Forest, Support Vector Machine, Naïve Bayes classification. depth water index, topographic wetness ‘wetland’ categorization derived property were most important predictors all presented enabled generation 3-class with MCC 0.58. In addition classified maps, investigated technique’s potential producing continuous maps. argue that probability pixel being as wet can be 0–100% index (dry wet) moisture, resulting could provide more information management