Monitoring ambient water quality using machine learning and IoT: A review and recommendations for advancing SDG indicator 6.3.2

This review examines the current state of ambient water quality monitoring systems (AWQMS) in relation to Sustainable Development Goal (SDG) indicator 6.3.2, which focuses on assessing water quality in natural water bodies, independent of specific human usage. This approach underscores the significance of evaluating water quality in rivers, lakes, and groundwater concerning their natural state. On a global scale, poor ambient water quality is primarily driven by weak regulatory oversight of industrial discharges, agricultural runoff, unsustainable farming practices, and inadequate wastewater treatment infrastructure. Real-time monitoring enabled by machine learning (ML) models and Internet of Things (IoT) technologies offers a promising solution to these challenges. In alignment with SDG 6.3.2, this review analyzes the capabilities of ambient water quality monitoring systems (AWQMS), focusing on SDG 6.3.2 Level 1 parameters, model types, performance evaluations using the REFORMS checklist, monitored water body categories, IoT-based AWQMS comparisons, and prototyping insights drawn from 42 studies published between 2000 and 2024. Key findings reveal (1) the need for further refinement of ML models, (2) limited monitoring of nitrogen, phosphorus, and total oxidized nitrogen within Level 1 parameters, (3) insufficient application of the REFORMS checklist for model evaluations, (4) minimal focus on groundwater monitoring, (5) inadequate model prototyping, (6) heavy reliance on battery-powered sensors with limited investigation into power-harvesting technologies, and (7) restricted open access to ambient water quality data. This review aims to guide future research and policy initiatives, driving meaningful progress towards achieving SDG 6.3.2. • In-situ monitoring of N, P, COD, and NOx is significantly lacking in many studies. • The REFORMS checklist ensures rigor and reliability in ML-based water quality research. • Groundwater monitoring is lacking in many countries, resulting in insufficient data. • Poor nations struggle to monitor ambient water frequently and to acceptable standards. • IoT prototyping is vital for testing and validating before full-scale deployment.