Wastewater contamination detection in urban stormwater systems using multi-sensor data and machine learning

Stormwater is generally released untreated into receiving environments such as rivers, coastal waters, or groundwater. Stormwater networks can act as pathways for contaminants, particularly when affected by misconnections, illicit discharges, overflows, or damaged infrastructure. These issues can significantly degrade water quality, posing risks to both public health and the environment, while also creating costly operational challenges for water and wastewater companies. <br/><br/>Detecting wastewater contamination and tracing its entry points into stormwater systems remains a significant challenge due to various potential sources of incoming wastewater, dilution and dispersion of contaminants by tributary stormwater flows, and significant differences in consistency, regularity, and flow rate of inflows.<br/><br/>We investigated stormwater pipeline in the UK suspected of receiving wastewater from multiple misconnections. The aim was to determine whether the stormwater system was being impacted so that the statutory undertaker could address the contamination issues and improve the quality of the receiving water environment. Before the study, the origins of contamination were uncertain, although it was suspected they could involve domestic properties, small businesses, or both.<br/><br/>The study employed a comprehensive approach combining water sampling for microbiological indicators and an array of chemical analyses. In addition to grab sampling, we deployed a multi-sensor sonde (Proteus Instruments, UK) to monitor parameters such as tryptophan-like fluorescence, chromophoric / fluorescent dissolved organic matter, electrical conductivity, pH, ORP, turbidity, temperature, ammonium, and dissolved oxygen.<br/><br/>Sensor data were used to model microbial indicator concentrations over a period of roughly three weeks. Two modelling approaches were tested. One followed the methodology recommended by Proteus Instruments, while the other applied a Random Forest machine learning technique. The latter approach offers potential advantages in addressing challenges commonly associated with fluorescence-based sensors. The findings demonstrate the potential for enhanced detection of wastewater misconnections, providing a more efficient method for identifying sources of contamination within stormwater systems.