A soil-adaptive depth function with a standardized K parameter for high-resolution 3D soil mapping

We present a soil-adaptive depth function (SADF) framework that integrates eight fundamental depth-function forms and standardizes the key parameter K for 30 m resolution 3-D soil mapping of the geomorphologically complex black-soil region in Northeast China. The study uses 380 typical soil profiles (1,534 genetic horizons) and multi-source environmental covariates processed on Google Earth Engine. Nine soil attributes—pH, SOC, TN, TP, TK, CEC, clay, sand and silt—were modelled with four machine-learning algorithms across 0–200 cm depth. SADF automatically selects the optimal depth function for each property, while the standardized K parameter quantifies the pedogenic-factor effect under different land-use types. Spatial-block cross-validation yields good to moderate accuracy at all depth intervals, with factor importance varying by property and depth. This record contains the complete Python source code and Google Earth Engine (GEE) scripts, allowing full reproduction of all figures and statistical results presented in the paper.