Global Mapping of Soil Water Characteristics Parameters— Fusing Curated Data with Machine Learning and Environmental Covariates

Hydrological and climatic modeling of near-surface water energy fluxes is critically dependent on the availability soil hydraulic parameters. Key among these parameters characteristic curve (SWCC), a function relating content (θ) to matric potential (ψ). The direct measurement SWCC laborious, hence, reported values are spatially sparse usually have only small number data pairs (θ, ψ) per sample. Pedotransfer (PTF) models been used correlate with basic properties, but evidence suggests that also shaped by vegetation-promoted structure climate-modified clay minerals. To capture effects in their spatial context, machine learning framework (denoted as Covariate-based GeoTransfer Functions, CoGTFs) was trained using (a) novel comprehensive global dataset (b) maps environmental covariates properties at 1 km resolution. Two CoGTF were developed: one model (CoGTF-1) based predicted because measured not generally available, other (CoGTF-2) cross-validation CoGTF-1 resulted, for van Genuchten parameters, concordance correlation coefficients (CCC) 0.321–0.565. validate resulting compare two pedotransfer functions from literature, contents 0.1 m, 3.3 150 m evaluated. accuracy metrics considerably better than PTF-based maps.