Ecological risk and machine learning based source analyses of trace metals in typical surface water

Area of Science:

• Environmental Science
• Analytical Chemistry
• Data Science

Background:

• Surface water quality is compromised by trace metal pollution from human activities.
• Accurate identification of pollution sources is crucial for mitigating ecological risks.

Purpose of the Study:

• To classify potential emission sources of trace metals in the Elbe River using machine learning.
• To evaluate the effectiveness of different machine learning algorithms for source apportionment.

Main Methods:

• Applied shallow and deep learning algorithms trained on environmental discharge data.
• Utilized Random Forest for source apportionment and compared its performance with other algorithms.
• Analyzed trace metal concentrations (Zn, Ni, Cu, As, Pb, Cr, Cd) in river tributaries and mainstream.

Main Results:

• Identified wastewater disposal and metal industrial emissions as primary sources in the Triebisch tributary and river segments.
• Detected mineral industry emissions as an additional source in upstream and downstream segments.
• Random Forest demonstrated superior performance with high Kappa median (0.59) and low Hamming loss (0.22).
• The tributary Triebisch exhibited the highest risk quotient (>86).

Conclusions:

• Random Forest is an effective tool for identifying trace metal pollutants in aquatic environments.
• The findings support source-oriented adaptive management strategies for the Elbe River.
• Machine learning approaches offer a robust method for environmental pollution source identification.