Scientists use machine learning to spot wastewater leakages
Spillages of sewage and wastewater are thankfully uncommon, but when they do happen, it’s useful to have the tools to spot it before it can do lasting damage.
There are over 200,000km worth of watercourses which flow through the UK. That might not sound like much at first, but it’s actually several times the distance you would have to travel, in order to travel the entire diameter of the Earth at the Equator, which would equal 12,756km.
We hope the results will help to improve [waste water treatment plant] management and compliance oversight and, ultimately, contribute to a reduction in the discharge of untreated sewage to rivers and coastal waters.– Professor Peter Hammond, published findings in Clean Water
Each stream, canal or river can be teeming with life, but these fragile ecosystems can be damaged greatly by spillages of untreated sewage and wastewater. As with many environmental concerns, such damage must be stopped in its tracks as soon as possible, in this case with the support of the latest state-of-the-art machine learning.
AI lends a hand to water
Thankfully, such help is at hand, according to a paper published in the Clean Water journal. The responsibility for ensuring water health is with the Environmental Agency, it is tasked with monitoring and regulating discharges of wastewater pollution in English water bodies, while incidents of leakage must be reported by operators.
Many times, however, it falls to members of the public to spot spills which might have gone unnoticed. In 2018, 400 water pollution incidents were reported in this way, and machine learning could help us catch even more of them before they go too far.
Professor Peter Hammond published findings in Clean Water, claiming that artificial intelligence software was able to detect as many as 926 incidents of spillages from the storm overflows of two unidentified water treatment plants between 2009-18. These ‘dark discharges’ had been recorded by the AI as having occurred during dates which had supposedly not seen a single leak, according to operators.
The AI noted that such incidents had occurred despite a lack of unexceptional rainfall. This suggests that the plants were both making a number of non-compliant discharges of wastewater over the period monitored by the software, as operators would have only been permitted to make such discharges during periods of extreme rainfall, according to the European Commission.
Tracing back to the source
The findings could be pivotal in improving the health of British rivers and streams, as they suggest a significant under-reporting of water pollution incidents over a significant period of time. Just one day of concentrated sewage spilling into waterways can do great damage, and unless it is picked up at the source as soon as possible, irreversible damage can be done to the natural environment.
AI is a growing field of discovery for scientists, and in the case of Professor Hammond’s work, it was used to generate no less than 20 variations of algorithms to work out how a spill could occur, based on known spill events. With this knowledge in mind, it helped generate reconstructed data of past leaks. It might be seem little more than a bit of hindsight, but by finding the weak spots in the two plants’ waste water treatment systems, this software could help prevent future damage.
In conclusion, the paper added: “Our experience and analysis methodology might be of use to the sewerage industry and regulatory authorities. We hope the results will help to improve [waste water treatment plant] management and compliance oversight and, ultimately, contribute to a reduction in the discharge of untreated sewage to rivers and coastal waters.”