Harnessing Machine Learning to Mitigate Water Pollution in Support of Climate Action

Abstract

Wastewater treatment plants (WWTPs) are crucial in protecting public health and the environment by reducing pollutants before discharge into water bodies. This research presents a data-driven approach to enhance wastewater monitoring, ensuring compliance with environmental regulations by evaluating the predictive accuracy of several machine learning models in assessing effluent quality and categorizing effluent threats. In the first task, regression models such as Decision Tree, Random Forest, AdaBoost, and Support Vector Machine (SVM) were applied to predict Biochemical Oxygen Demand (BOD) and Chemical Oxygen Demand (COD), with Mean Absolute Error (MAE) and R-squared (R2) used as evaluation metrics. In the second task, the same models were utilized to categorize effluent threat levels, and their performance was measured through accuracy, precision, recall, and F1-score. The results demonstrate that Gradient Boosting Regressor (GBR) and AdaBoost performed well in COD prediction, achieving the lowest MAE of 6.11 and the highest R2 of 0.81. At the same time, Random Forest obtained the lowest MAE of 1.61 for BOD prediction. In the classification task, the Gradient Boosting Classifier (GBC) and AdaBoost achieved superior precision, recall, and F1 Scores, with all models attaining an overall accuracy of 97%. According to these results, machine learning methods, particularly GBC and AdaBoost, can significantly enhance prediction and classification accuracy for effluent quality, thereby improving WWTP management. This study contributes to climate resilience and sustainability by applying AI to minimize wastewater pollution, supporting SDG 6 (Clean Water and Sanitation), SDG 13 (Climate Action), SDG 14 (Life Below Water), and SDG 9 (Industry, Innovation, and Infrastructure).