Integrating Compositional Data Analysis (CoDA) and Random Forest for lithology-specific geochemical baseline...

Area of Science:

• Environmental Geochemistry
• Applied Geology
• Data Science in Earth Sciences

Background:

• Accurate geochemical baselines are vital for distinguishing natural soil composition from anthropogenic contamination in mining-affected regions.
• Geologically complex areas with historical mining pose challenges for establishing reliable background values.
• Existing methods may be insufficient in areas with long-term industrial impact and diverse geological formations.

Purpose of the Study:

• To develop and validate an improved methodology for defining lithology-specific geochemical baselines.
• To accurately distinguish natural geochemical variations from legacy contamination in soils.
• To support regulatory decisions and remediation strategies in complex geological settings.

Main Methods:

• Integration of Compositional Data Analysis (CoDa) with the Random Forest algorithm for robust lithological classification.
• Analysis of over 300 topsoil and subsoil samples for mineralogy and trace elements (Hg, As, Sb, Cr, Cu, Co, V, Ni).
• Application of robust statistical methods (median ± 2MAD) for baseline value computation.

Main Results:

• Successfully estimated geochemical baselines for Hg, As, and Sb, demonstrating strong lithological control across volcanic and sedimentary domains.
• Identified significant variability in baseline values linked to geological settings, aiding in distinguishing natural anomalies from contamination.
• Observed baseline values exceeding Italian legal thresholds for Hg and As, highlighting the need for site-specific regulations.

Conclusions:

• The proposed workflow effectively establishes lithology-specific geochemical baselines in complex, historically mined areas.
• The methodology provides crucial insights into the natural variability and resilience of different geological domains.
• Findings underscore the necessity of tailored regulatory limits and the potential for broader application in similar contaminated regions.