High-resolution maps of forest soil bulk density and pH across China at 90-m resolution

The source data comprises 4,356 forest-specific soil profiles combined with 41 high-resolution environmental covariates, including climate, vegetation, and terrain variables. Quantile Regression Forests with forward selection and hyperparameter tuning were employed to generate the predictions.

Model performance was evaluated via 10-fold cross-validation and independent validation, showing strong predictive accuracy (bulk density: 0.78–0.89 in cross-validation, 0.60–0.66 in independent validation; pH: 0.84–0.87 in cross-validation, 0.71–0.81 in independent validation). The spatial patterns reflect varying pedogenic controls by depth and property, with climate influencing bulk density and vegetation–parent material interactions dominating pH.