Amount awarded: £5,080,719
Led by: Northumbrian Water
Partners: ADAS, Anglian Water, Cognizant, Dŵr Cymru Welsh Water, Northern Ireland Water, South West Water, Stream, The Rivers Trust, Tidal, Google LLC., Uisce Éireann, Water Research Centre Limited, Wessex Water, Xylem Inc.
We will develop open-source, scalable, digital models to inform effective action to tackle waterbody pollution. Our novel use of Machine Learning will efficiently analyse existing data and new diverse data inputs. This will unlock new insights into the complex factors impacting waterbodies, bringing deeper understanding and accelerating positive change.
“The River Deep Mountain AI (RDMAI) project is set to revolutionise the way we gather insights and data on waterbody pollution – and will accelerate real positive environmental change across our regions. We are really excited to launch this exciting project at our very own Innovation Festival in July.” – Nigel Watson, Group Information Services Director, Northumbrian Water Group
—
The five building blocks below reflect progress across modelling efforts, focused on pollution detection: BB1A: E.coli Model Development: The first draft of the E.coli Model v1.0 has been completed, with performance evaluated across multiple metrics. Sensitivity analysis will be conducted to assess the model’s robustness, and GIS landcover data is being prepared for integration. BB1B: Flow Estimation: The initial version of the flow estimation model (LSTM 1.0) and its code has been finalized. Data from multiple sources (Estreams, EA, and CALS) has been collected, and reviews of the Neural Hydrology Library have been completed. Further work includes improving model performance through additional layers and compiling observational data for refinement. BB1C: Phosphate Analysis: A framework to retrospectively assess phosphate data has been developed by condensing 2959 determinants to 900 and integrating 25 years of rainfall data. The implementation of the framework is underway, with standardized metrics being introduced to minimize bias and enhance data reliability. BB2A: Spill and Diffuse Pollution Sources: Models are being built to identify spill hotspots in pilot catchments and detect agricultural pollution sources using remote sensing images. BB2D: Monitoring Station Optimization: Criteria identified to build a model for determining the optimal locations of monitoring stations
—
Resources
For more information on River Deep Mountain AI, take a look at the following resources:
