Collaborative innovation undertaking River Deep Mountain AI (RDMAI) has introduced the open-source launch of a set of synthetic intelligence and machine studying (AI/ML) fashions that it says are set to remodel the best way water high quality knowledge is collected and used.
Funded by Ofwat’s Water Breakthrough Problem and led by Northumbrian Water, with Spring Innovation because the knowledge-sharing associate, RDMAI is a cross-sector initiative constructing open-source, scalable AI instruments to sort out waterbody air pollution and enhance river well being. Information from a variety of sources, together with citizen science and satellites, has been used to construct the fashions.
The discharge of AI/ML and remote-sensing fashions on the open-source platform GitHub is the undertaking’s first main milestone, following completion of the event and preliminary testing phases. All through this era, the undertaking crew collated datasets from inside and outdoors the sector, run experiments with AI/ML fashions and held co-creation classes with companions and stakeholders.
The ensuing fashions and datasets goal to assist:
- River circulation predictions
- Air pollution supply monitoring
- Air pollution hotspot mapping
Suggestions is invited at this stage to assist refine and improve the fashions because the undertaking progresses.
The UK’s water surroundings is underneath strain from inhabitants progress, local weather change, air pollution from a number of sources and nutrient overload. Simply 14% of English rivers are assembly Water Framework Directive requirements for good ecological standing.
Launched in July 2024, River Deep Mountain AI goals to handle this problem by growing open-source, scalable AI/ML fashions to uncover air pollution patterns and unlock actionable insights for safeguarding waterbodies.
Northumbrian Water’s undertaking companions are: ADAS, Anglian Water, Cognizant, Northern Eire Water, South West Water, Stream, The Rivers Belief, Google, WRc, Wessex Water and Xylem.
George Gerring, undertaking lead, Northumbrian Water, mentioned, “We’ve got constructed a set of capabilities that use synthetic intelligence, machine studying, generative AI and distant sensing to grasp and predict completely different variables impacting waterbodies well being.
“The open-source launch of those fashions on GitHub means they’re obtainable for residents, researchers, water organisations and NGOs to make use of. Any suggestions on the early releases will assist us refine and construct on what we’ve achieved to date.”
Angela MacOscar, head of innovation, Northumbrian Water, mentioned: “Useable knowledge on waterbody well being is disparate and laborious to entry, which is why the RDMAI crew is working to squeeze as a lot actionable data out of current knowledge as doable.
“By integrating knowledge from varied sources, together with environmental sensors, satellite tv for pc imagery and citizen science, the undertaking is bridging the info gaps in waterbody well being and empowering higher, quicker and simpler interventions. Open-sourcing these fashions marks a significant shift in how we collaborate to sort out environmental challenges.”
Stig Martin, international head of ocean, Cognizant, mentioned: “This undertaking is a testomony to the ability of analysis and growth and daring to make use of know-how to resolve complicated, large-scale environmental issues.
“We imagine in transparency and are proud that this undertaking is open-source, permitting everybody to see how the system is constructed and the way it generates its insights. It has been extremely rewarding to be a part of a collaboration that’s not simply speaking about change however is actively constructing the instruments to make it occur.”
Part three of the programme, now underway, will give attention to mannequin enchancment, validation in new catchments and evaluating the potential to scale throughout the UK. The refined variations of the fashions are set to be launched in November.
The GitHub web page for RDMAI might be considered at https://github.com/Cognizant-RDMAI
