top of page

AI in Water Management

The integration of machine learning with environmental monitoring, as demonstrated in the recent study by Shyu et al. (2023), shows that by monitoring multiple parameters and employing advanced machine learning algorithms, we can better understand freshwater quality dynamics to enable more effective interventions and improved freshwater health.

Limitations of traditional methods 

Traditional methods of analysing water quality, focusing on parameters like chemical oxygen demand (COD), total suspended solids (TSS), and pathogens such as Escherichia coli (E. coli), have relied on manual samples of single parameters, followed by laboratory analysis. 

These methods are time-consuming and do not offer real-time insights into water quality, which pose challenges in promptly detecting contamination and treatment responses. Water quality is also influenced by a complex interplay of chemical, physical, and biological processes. Focusing on a single parameter provides a limited view of water quality and can lead to misleading conclusions about the presence of contaminants like E. coli, nitrates, and heavy metals.

Leveraging machine learning across multiple water parameters 

The study by Shyu et al. highlights the development of real-time monitoring systems for onsite wastewater treatment systems (OWTS) using machine learning algorithms to predict key water quality parameters from in-line sensor data. This approach shows the importance of monitoring multiple water quality parameters — such as pH, dissolved oxygen (DO), conductivity, turbidity, and temperature—rather than relying on a single parameter.

Machine learning for water quality monitoring

The study demonstrates the application and potential for machine learning algorithms to develop soft sensors for predicting chemical oxygen demand (COD) total suspended solids (TSS), and E. coli concentrations from real-time, measurable parameters like turbidity, pH, and electrical conductivity.

Each parameter contributes unique insights into water quality conditions and potential contamination sources. For instance, turbidity and colour can indicate the presence of particulate matter and organic compounds, while conductivity and pH levels can signal changes in water chemistry that may affect contaminant behaviour and treatment processes. 

“By analysing multiple parameters collectively, machine learning algorithms can detect patterns and correlations that would be missed when focusing on a single parameter,” says Abi Croutear-Foy, Managing Director of AquaWatch. “Monitoring multiple water quality parameters can also improve the predictive power of machine learning”.

Algorithms trained on water quality data 

The study's use of machine learning algorithms, including:

  • partial least square regression (PLS), 

  • support vector regression (SVR), 

  • cubist regression (CUB), and 

  • quantile regression neural network (QRNN), 

showcases the potential of data-driven models in environmental monitoring. 

These algorithms, which were trained on water quality data from an advanced OWTS, demonstrated that even with limited data points, machine learning-based soft sensors could provide accurate predictions of COD and TSS. While the prediction of E. coli proved challenging, the study highlights the need for further research and model refinement.

Machine learning across multiple parameters has revolutionised water management

Recent advancements in continuous, real-time monitoring, sensor technology and machine learning algorithms have revolutionised water quality monitoring and management. For environmental scientists, policymakers, and water treatment operators, embracing these technological advancements means moving towards more efficient, accurate, and timely water quality monitoring systems. 

“These findings and advancements are supported by the cutting-edge water monitoring solutions offered by AquaWatch,” says Croutear-Foy. “By adopting a holistic data approach and leveraging machine learning, we can significantly improve our effectiveness in water quality management.”

The insights gained from machine learning models based on comprehensive parameter monitoring can inform better decision-making, enhance treatment processes, and ultimately protect public health and the environment. 

To find out more about the latest advances and how we can revolutionise your water quality monitoring, please contact us here



bottom of page