Machine learning to assess and support safe drinking water supply: a systematic review

Drinking water is essential to public health and socioeconomic growth. Therefore, assessing and supporting safe drinking water supply is a critical task in modern society. Conventional approaches for analyzing and controlling drinking water quality are often labor-intensive, costly, and low in throughput. Machine learning (ML) has emerged as a promising alternative for assessing and supporting safe drinking water supply. Existing reviews have summarized the applications of ML in drinking water supply from different aspects. However, a state-of-theart, comprehensive review is missing that summarizes the applications of ML in analyzing, monitoring, simulating, predicting, and controlling drinking water quality, especially in municipal engineered water systems. This review aims to fill that gap by critically compiling the applications of ML in assessing and supporting drinking water quality in municipal engineered water systems. To be comprehensive, we also cover the applications of ML in other drinking-water-related settings, such as source water and water purification processes. We begin by explaining the basic mechanics and workflows of ML, focusing on its use for assessing and controlling key causal factors in drinking water from the physical, chemical, and microbiological aspects. Those causal factors affect water quality and public health, such as water pipeline failures, disinfection by-products, heavy metals, opportunistic pathogens, biofilms, and antimicrobial resistance genes. We then illustrate the distribution of ML model usage across research topics in safe drinking water supply. Finally, we discuss the challenges and outlooks for applying ML in safe drinking water supply. This review provides a comprehensive and systematic summary of the applications of ML in assessing and supporting water quality in municipal engineered water systems and other settings, thereby advancing drinking water safety.