A data-driven high-accuracy modelling of acidity behavior in heavily contaminated mining environments

Area of Science:

• Environmental Science
• Geochemistry
• Water Quality Assessment

Background:

• Accurate acidity estimation is vital for acid mine drainage (AMD) characterization and remediation.
• Conventional methods like titration and iron speciation models struggle with site-specific geochemical complexities.
• Generalized models often fail in complex AMD environments due to spatial and seasonal hydrochemical variability.

Purpose of the Study:

• To develop a high-accuracy, site-specific empirical model for predicting acidity in AMD-impacted waters.
• To assess the limitations of conventional acidity estimation methods in complex AMD settings.
• To demonstrate the utility of multivariate models calibrated for specific sites.

Main Methods:

• Field data collection at the Trimpancho mining complex (Iberian Pyrite Belt, Spain).
• Application of multiple linear regression (MLR) for model development.
• Comparison of the developed model's performance against the standard Hedin method.

Main Results:

• A robust predictive model for AMD acidity was established using Cu, Al, Mn, Zn, and pH (R² = 99.2%).
• The site-specific MLR model significantly outperformed the Hedin method (13% vs. 29% mean absolute percentage error).
• Strong spatial and seasonal hydrochemical variability was observed, highlighting the need for localized models.

Conclusions:

• Site-calibrated multivariate models offer a practical and accurate approach to acidity prediction in complex AMD systems.
• The developed model provides a superior alternative to generalized methods for the Trimpancho site.
• This research underscores the importance of accounting for site-specific geochemistry in AMD management.