World’s first deployment of AI powered water quality sensors in coastal waters at Poole Harbour concludes successfully.
Poole Harbour is one of the world’s largest natural harbours and has been recognised internationally as an important area for nature conservation and is designated a Special Protection Area. The harbour supports extensive wild and aquaculture shellfish beds, and water quality is vitally important for this key local industry.
Like many harbours and coastal sites, Poole Harbour faces challenges to maintaining healthy water quality from events such as bacterial build up, sewage discharges, fertiliser runoff and industrial discharges, which can be harmful for aquatic ecosystems. Traditionally, the only way to monitor and model the water quality at a coastal site like Poole Harbour has been through expensive and time consuming laboratory tests. BCP Council wanted a better way.
BCP’s ground-breaking innovation project with UK SME UnifAI Technology was a two year pilot which aimed at taking tools like artificial intelligence and machine learning out of the lab and demonstrating their utility in the real world. In this case, in the harsh real-world environment of the ocean. Low cost wireless sensors were integrated with UnifAI’s cutting edge AI capabilities to provide a real-time monitoring solution unlike anything in use at the time.
Two years on and the initial project has come to a successful end with the innovative approach to monitoring showing that low cost sensors with AI have the potential to be a game changer for the water industry.
Background and Timeline
In 2018, Bournemouth, Christchurch and Poole (BCP) Council was selected as a partner on the Interreg 2 Seas project SPEED (Smart Ports Entrepreneurial Ecosystem Development), a European Interreg project aiming to build an ecosystem for smart port applications development in Belgium, France, the Netherlands and the UK, bridging the gap between the worlds of European ports and the nascent data science – IoT market.
In March 2020, BCP Council held a competition to identify an innovative and cost-effective water quality monitoring capability for Poole Harbour as part of the SPEED project. UnifAI Technology won the competition for a project to trial the use of low-cost water quality sensors with artificial intelligence (AI) in Poole Harbour and Poole Park Lagoon.
UnifAI Technology and BCP Council worked with the Poole Harbour Commissioners to identify suitable locations for monitoring water quality, and in October 2020 six sensors were deployed on existing buoys and pontoons in the harbour and the lagoon. At the same time an initial set of AI algorithms were trained and deployed using historic Environment Agency data.
In October 2020, Bournemouth University conduced an initial assessment of AI outputs for a limited number of parameters against a calibrated TSI Pro Plus Multi-Probe and laboratory testing. The target was for AI predictions to be within 15%-20% of laboratory results. The results of the first comparisons exceeded this expectation, with the AI predicting levels of Nitrites for example within 8% of the laboratory results.
In the summer of 2021, Bournemouth University conducted a series of laboratory tests in Poole Park Lagoon to assess how well the sensors were performing after a year in the ocean with no maintenance. The results again exceeded expectations.
There were challenges along the way. During the first winter some sensors were damaged in storms and were replaced, and over time some sensors saw a build up of marine growth that affected their readings. Part of the project aims was to learn how such sensors performed over time in seawater and to experiment with ways to mitigate these issues and extend the useful life of the sensors, such as encasing the sensor probes in copper to inhibit plant growth.
In June 2021 the project was a finalist for two awards at the Water Industry Awards 2021, and was Highly Commended in the Data Analytics, Cloud and AI Project of the Year category.
As a result, what began as a small project in Poole Harbour gained the attention of one of the more innovative water utilities. In summer 2021 Wessex Water partnered with UnifAI Technology to begin their own pilot, testing the same approach for monitoring water quality at a popular inland bathing site at Warleigh Weir. The success of that project, which uses AI to predict harmful levels of e.coli and enterococci bacteria, was subsequently recognised as the Digitalisation Project of the Year at the Water Industry Awards 2022.
The BCP Council-UnifAI Technology project at Poole Harbour delivered a remarkable amount of value in testing hardware, AI algorithms and an innovative approach to water quality monitoring, for a tiny budget. All parties invested time and resources beyond the monetary budget, and the project can be said to have pioneered the use of AI for water quality monitoring in a way that is now being adopted more widely by parts of the water industry.
Cllr Anderson, Portfolio Holder for Environment and Place at BCP, said “I am really pleased that following our successful trials in Poole harbour measuring pollution that we plan to develop and introduce a real-world application that will provide residents and bathers with real time data on whether the water around Boscombe is safe, rather than relying on the EA predicted alerts which are not well understood. I believe this is one of the first schemes in the country if not the world where AI is used to show when pollution is present and this will in the long term allow us to develop a Green flag scheme.”
The low-cost sensors performed remarkably well in the harsh salt-water environment. The final phase of deployment, with copper encasing the probes, has so far proved to be successful in extending the useful life of those sensors.
The AI algorithms produced good results when compared with laboratory samples, given the limitations of training data and the uncertainties with sensor calibration over time. The solution provides an ability to monitor changes in real-time and to understand and predict events through identifying patterns and anomalies in the data.
The project demonstrated a strong case for using low-cost sensors with AI for identifying patterns and events in water as an early warning system, a capability that is particularly valuable at a time when the quality of water in our rivers and beaches is more in the public eye than it has ever been.