A data-driven approach to arsenic classification in groundwater in geothermal Systems: Meta-Analysis and machine learning applications in Western Anatolia, Turkiye
Physics and Chemistry of the Earth, vol.139, 2025 (SCI-Expanded, Scopus)
- Publication Type: Article / Article
- Volume: 139
- Publication Date: 2025
- Doi Number: 10.1016/j.pce.2025.103966
- Journal Name: Physics and Chemistry of the Earth
- Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Aquatic Science & Fisheries Abstracts (ASFA), Artic & Antarctic Regions, Chimica, Communication Abstracts, Compendex, Geobase, INSPEC, Metadex, Pollution Abstracts, Civil Engineering Abstracts
- Keywords: Geothermal Energy, Arsenic, Contamination, Meta-analysis, Machine learning, Western Anatolia, Turkiye
- Dokuz Eylül University Affiliated: Yes
Abstract
Western Anatolia, Türkiye, is renowned for its diverse geothermal resources, encompassing high, medium, and low enthalpy systems. While these systems are valuable for energy production and economic development, they are also associated with significant environmental challenges, particularly high concentration arsenic and boron contamination. This study highlights critical hotspots, including Sandıklı (27 mg/L) and Banaz-Hamamboğazı (95.64 mg/L), with arsenic levels far exceeding the World Health Organization's (WHO) maximum permissible limit of 10 ppb. Such contamination poses significant risks to water quality, agriculture, and public health, especially in major agricultural provinces like Aydın and Manisa. To address these challenges, machine learning models were applied to classify arsenic concentrations. Ensemble methods, including AdaBoost (ABC) and Extra Trees (ETC) classifiers, consistently outperformed others, showing high accuracy of about 97 % in distinguishing geochemical signatures and predicting arsenic levels. In contrast, the k-Nearest Neighbors Classifier (KNNC) proved less effective, with frequent misclassifications. The combination of machine learning and meta-analysis provided a robust framework for identifying spatial and temporal patterns of contamination, offering valuable insights for environmental monitoring. This approach not only enhanced the understanding of arsenic distribution in geothermal systems but also provided actionable insights for mitigating contamination risks. The findings underscore the importance of combining computational techniques with environmental geochemistry to improve the management of geothermal wastewater. Future research should expand these methodologies to other regions and contaminants, leveraging machine learning to develop more effective environmental protection strategies. This study demonstrates the potential of data-driven approaches to address critical environmental issues and supports sustainable development in geothermal-rich areas.