Photographs gathered by the Friends of Bradford’s Becks community group have been used to train a new AI model to spot visual markers of river health in what is believed to be a world first, with hopes that the tool can be rolled out across the UK.
With funding from Natural England, Water Research Centre (WRc), in collaboration with Rain++ and students at the National Taiwan University (NTU), defined visual markers of river health, including the presence of wildlife, outfalls discharging into the river, rubbish, obstructions to fish passage, channel modifications and discoloration. A subset of photos that contained these markers was used to test and train AI models to recognise the signs of river health.
WRc Technical Lead Dr Amy Jones said: “We took 15,700 photos, submitted by members of the Friends of Bradford’s Becks community group, and used them to train and test the AI model. Using image recognition models to tackle environmental challenges is an approach that is starting to gain momentum but, as far as we are aware, using it to identify pollution in rivers from citizen photos is a world first.”
River health monitoring is normally carried out with chemical and physical equipment by the Environment Agency and water industry professionals, according to Amy. This equipment is expensive and complicated to set up and maintain. Camera monitoring could offer a cheaper, more accessible way to monitor river health, supplementing traditional methods.
Amy said: “The AI model can be used to see what citizen photos tell us about the health of Bradford’s Becks. Maps of the results can then be used to identify hotspots, outfall locations or areas close to sewage litter, which could then be a priority for further investigation by the water company or local authority. Last year’s Big River Watch encouraged almost 6000 people across the UK to take photos of their local waterbody, illustrating widespread potential for this model. When applied alongside conventional monitoring, the model has the potential to be really powerful.”
The number of photographs required to train the AI model has made the analysis challenging. Amy said that the six-month funding from Natural England has enabled the project team to build the initial model and with the renewed funding for 2024, more manually labelled photos can be incorporated into the data set, improving the model’s accuracy. Alongside NTU and Rain++, WRc is also investigating how it can be applied to video footage and cameras will be installed on Bradford’s Becks to test the model’s ability to continually monitor for pollution events.
Although it is still under development, the model and its documentation have already been made freely available on GitHub, an online developer platform for storing and sharing code.
Natural England senior advisor Dr Rachel Palfrey said: “Clean water is essential for life. This innovative project, using technology to help us protect, restore and enhance water quality within the Bradford District forms part of our wider Nature Recovery Project across Bradford and the South Pennines. It also has the potential to revolutionise how we monitor river quality, making it easier to see what is happening, when it is happening and where, so we can act accordingly.”
