Environmental Drivers and Explainable Modeling to Resolve Trace Metal Dynamics in a Lotic System

Area of Science:

• Environmental Science
• Geochemistry
• Computational Science

Background:

• Trace metal contamination in freshwater systems is highly variable due to hydrology, geochemistry, and human activities.
• Characterizing trace metal exposure in urban and peri-urban streams is challenging.
• System-level controls on dissolved trace metal signatures require integrative analytical approaches.

Purpose of the Study:

• To investigate the environmental structuring and governing controls of dissolved trace metal signatures in a human-impacted stream.
• To apply a system-oriented computational framework to decipher complex trace metal distributions.
• To capture temporal variability using a season-resolved sampling strategy.

Main Methods:

• Implemented a station-based, season-resolved sampling strategy during wet and dry periods.
• Analyzed physicochemical parameters (pH, temperature, dissolved oxygen, conductivity), nitrogen species, phosphorus fractions, and dissolved trace metals (Cr, Cu, Ni, Pb, Cd, Hg, As).
• Utilized regression-based machine learning models (Random Forest, XGBoost) for element-specific sensitivity analysis and prediction.

Main Results:

• Machine learning models achieved high predictive performance for most trace metals (R² > 0.95).
• Random Forest excelled for Cr, Ni, Pb, Cd, As, and Hg; XGBoost was optimal for Cu.
• Explainability revealed heterogeneous, metal-specific controls: Cr by temperature, Ni by NO₂⁻ and redox, Cd by NH₃ and temperature, As by Hg and phosphorus/redox proxies. Pb showed lower predictability.

Conclusions:

• Trace metal distributions in streams are primarily structured by differential environmental sensitivity, not uniform source inputs.
• Integrative computational frameworks are essential for interpreting freshwater contamination under anthropogenic and climatic pressures.
• Understanding metal-specific environmental drivers is crucial for effective management of urban stream contamination.