Improving soil pH prediction and mapping using anthropogenic variables and machine learning models

This study evaluates the impact of anthropogenic activities on soil pH prediction in China's Huang-Huai-Hai Plain using four machine learning models (RF, LightGBM, XGBoost, SVM). By incorporating five anthropogenic variables (fertilization, population density, heat flux, urbanization, road density) alongside 24 environmental factors, model performance improved significantly (R² up to 0.56 for RF). Key findings: (1) Fertilization and population density were the most influential human factors; (2) Precipitation remained the dominant predictor overall. The results highlight human activities' substantial role in soil pH variation, supporting precision agriculture and sustainable land management in intensively cultivated regions.