AI for ALG
Artificial Intelligence for Algal Monitoring
Transforming how the water industry identifies, counts and predicts risks from algae and cyanobacteria
AI for ALG is developing an artificial-intelligence-based system that can identify and count algae from microscope images of water samples. The project aims to replace slow, labour-intensive monitoring with a rapid, accurate and scalable approach, producing the high-resolution data needed to understand and predict algal water-quality risks.
September 2023 – February 2027
A 3.5-year research and innovation programme
Now in year three
Operational implementation, testing and benchmarking
Open, sector-wide tools
An accessible application, image datasets and supporting resources
Why AI for algal monitoring?
Algae and cyanobacteria are natural components of aquatic ecosystems, but dense blooms and toxin-, taste- or odour-producing species can create serious challenges for drinking-water treatment. These pressures are expected to intensify as the climate warms.
Current monitoring relies heavily on experts identifying and counting cells by microscopy. This is time-consuming, limits the number and volume of samples that laboratories can process, and can delay results by around a week. The resulting datasets are often too sparse for robust predictive modelling. AI for ALG is designed to reduce analysis from weeks to minutes while increasing monitoring throughput and data resolution.
What we are developing
- A UK-wide training dataset of accurately annotated algal microscope images spanning geographic, seasonal and taxonomic diversity.
- Convolutional neural networks optimised to detect, identify and count algae and cyanobacteria in water samples.
- A simple analysis application that allows water-industry users to upload microscope images and rapidly generate algal community data.
- Standardised imaging and verification methods, including comparison with expert microscopy and complementary eDNA community profiling.
- High-resolution datasets for predictive analytics, supporting earlier warning and better management of algal-related risks.
The project programme
Years 1-2
Training data and AI development
Image and annotate water samples collected by partner water companies, verify identifications using expert assessment and eDNA, and develop and optimise the neural-network pipeline and user dashboard.
Year 3 – current phase
Real-time implementation and benchmarking
The system is being implemented and tested under operational conditions in Welsh Water laboratories. Its accuracy, usability, turnaround time and cost are being benchmarked against traditional microscopy and eDNA methods, while high-resolution environmental datasets are assembled for modelling.
Final phase
Predictive analytics and dissemination
High-resolution community data will be used to test improved prediction of algal-related water-quality risks. The project will then share its application, training datasets, procedures and learning with partners and the wider water industry.
Expected impact
- Faster, higher-throughput and more cost-effective algal monitoring.
- Earlier identification of potentially harmful blooms and more timely operational responses.
- Better data for predicting toxins, taste and odour events and other treatment risks.
- Potential for simple field and remote monitoring, including future citizen-science applications.
- Open tools and reference images that support water companies and train the next generation of algologists.
Project team and partners
The project is led by Dŵr Cymru Welsh Water in partnership with MicroLab@Bristol at the University of Bristol, Cardiff University, Anglian Water, United Utilities and Wessex Water, with contributions from a wider network of water-industry and international collaborators.
At MicroLab, Dr Chris Williamson leads the AI monitoring research. Sonia Giulietti leads production and curation of the freshwater algal image dataset, while PhD researchers Holly Liken and Jenni Smith work on neural-network development and complementary eDNA-based community verification.
Open innovation
The project is designed to produce a sector-wide legacy. The final AI application, annotated image resources, trained models and supporting procedures will be shared openly, alongside training and dissemination activities to help water companies and other users adopt and extend the approach.
AI for ALG is funded through the Ofwat Innovation Fund’s Water Breakthrough Challenge.