Regional Geochemical Background Field Simulation and Anomaly Separation Based on Random Forest and Multi-Source Environmental Factors

Geochemical exploration is vital for mineral prospecting as it provides direct evidence of mineralization. Regionally, stream sediment surveys are commonly used to detect geochemical anomalies, however, traditional anomaly detection methods often fail to separate subtle anomalies due to complex geology, high-dimensional data, and strong spatial heterogeneity. This study introduces a new method that uses a Random Forest Regression (RFR) model integrated with multi-source environmental factors to simulate the geochemical background and identify anomalies. The RFR model demonstrated high accuracy for the Cu background field, achieving 78.4% explained variance on the test set. Compared to traditional methods, the new approach significantly enhances the spatial consistency between anomalies and known deposits, raising the anomaly verification effectiveness rate (EI) from 45.3% to 67.8%. Furthermore, feature importance analysis indicates that Cu distribution is mainly influenced by lithological background and temperature. This study demonstrates that Random Forest Regression (RFR) effectively integrates multi-source environmental factors to accurately simulate the spatial variation of geochemical background fields. It provides a powerful solution to the challenge of distinguishing anomalies from background in the presence of lithological variations and topographic dilution.