Decoding distribution data: extracting actionable information from drinking water distribution system water quality time series

Advancements in continuous in-situ water quality monitoring provide a unique opportunity to enhance our understanding of water distribution systems, ensuring safe and clean drinking water for all while mitigating contamination risks. However, this kind of monitoring is relatively new, with its true value yet to be fully demonstrated. This research developed novel analytical routines to extract actionable information from diverse real-world datasets provided by five different water service providers. A vital first step to maximising the value of these datasets was taken with the development of a data quality assessment framework, specifically tailored to address the challenge of water quality sensors' sensitivity to errors when deployed in drinking water distribution systems. Next, an investigation into the optimal analysis of in-network discolouration events using turbidity time series was conducted. This was informed by an innovative crowd-sourced event labelling exercise, integrating the perspectives of 48 domain experts and employing time series forecasting to devise a turbidity event scale. This scale effectively distinguishes between advisory (<2 NTU), alert (2-4 NTU), and alarm (>4 NTU) events, enabling reactive and proactive analysis of network events. An overarching finding of this research was the demonstration that the level of insight obtainable when moving from single parameter single sensor to multiple parameters and sensors increases in a multiplicative fashion, as evidenced by the application of developed approaches to multiple real-world examples. This research paves a clear path towards enhanced intelligent utilisation of water quality sensor networks to improve network management capabilities. This digitalisation-driven approach, in the face of increasing climate change related challenges, promises to provide the resilience required to safeguard these vital public health assets.