Digital soil mapping (DSM) using machine learning algorithms

Digital mapping of soil types and properties (soil organic carbon (SOC), soil organic matter (SOM), clay, sand, silt) is generated using machine learning algorithms to identify relationships between soil properties and several covariates across different landscapes. One of the simplest machine learning methods for finding relationships between several covariates and a target variable is multiple linear regression (MLR). With MLR, two or more covariates are used to predict the outcome of the target variable by fitting a linear equation with quantitative coefficients (model parameters). MLR algorithm was compared on the basis of five factors: (1) target variable, (2) sample size, (3) sampling design, (4) covariate set, and (5) interpretation of the resulting model: Cross validation and Independent validation (R2, RMSE).