Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
As pollution shrinks the amount of freshwater that people across the world can use, tackling the problem with new treatments and upgraded infrastructure is urgent. However, having the most useful information possible to act on is essential. This is why innovators are looking to artificial intelligence — which can provide accurate real-time analysis — to identify contaminants and enable faster, focused clean-ups.
The sources of water pollution are complex, as they can include industrial chemicals, plastic waste, the contaminants picked up by floodwater and the toxins and ash created by wildfires.
There is also increasing run-off of chemicals from agricultural land, while untreated urban wastewater entering rivers and lakes is resulting in more algal blooms, which create “dead zones” where plants and animals cannot survive, as well as looking and smelling bad.
The analytical power of AI has the potential to help develop treatment technologies and inform decisions on where to locate water facilities. But it needs data to assess why, where and when contamination is taking place.
A key focus of attention is ageing sewerage and clean water infrastructure, much of it underground and in constant need of repair to prevent leaks and overflows polluting freshwater.
While cameras, robots and sensors have long been used to inspect out-of-reach areas, the data retrieved on water quality or the state of pipes has been reviewed by humans. AI can assume this task by analysing video and images accurately and quickly.
CCTV inspections are well established, says Amy Bunszel, who leads on architecture, engineering and construction at US software company Autodesk. “Now we can take that data and leverage AI to get to a rapid answer.”
Autodesk is using AI image analysis tech to enhance its Info360 Asset tool, which automates inspection reviews for utilities and municipalities. It was developed by Vapar, an Australian pipeline maintenance software company, which raised A$5mn in funding last year from investors, including Autodesk.
Vapar’s tech enables field crews to upload and automatically review CCTV footage from water pipes and sewers, saving time and increasing accuracy. “Humans miss about 40 per cent of the problems. AI learns and will get to a much higher success rate,” Bunszel predicts.
More stories from this report
As well as detection, AI could help prevent contamination. In agriculture, it can use data from sensors and satellites to generate insights about crop growth and soil quality to aid farmers in avoiding unnecessary use of fertilisers and pesticides. The result is less chemical run-off leaching into water supplies.
“You can optimise nutrient addition when it’s required on a crop so you’re not over-fertilising,” says Alexander Crowell, partner at Amsterdam-based water-tech venture capital investment firm PureTerra Ventures. One of the companies it has backed is Spain’s Hemav, which last year raised €8mn to develop AI-powered crop prediction models.
So far, satellites’ usefulness is limited to detecting general changes in large bodies of water. This is where ground-based sensors come in, because they can collect the level of detail AI needs to generate the best insights, says Scott Bryan, president of Imagine H2O, an accelerator that has backed more than 200 water-focused start-ups across the world.
Bryan cites a pilot project in Pakistan run by a London-based start-up backed by Imagine H2O. With World Bank funding, AquAffirm’s disposable digital sensors are used to detect arsenic in 250 wells in Punjab province, where levels of this highly toxic element are above safety thresholds in 20 per cent to 30 per cent of water supplies.
Combining disposable test strips with web-connected analytics software on a mobile app makes the tech easier to access. “Instead of having one agency collecting the data, farmers or other employees of a government entity can capture and collect that data,” says Bryan.
However clever the AI-based tech is, the underlying data is always key. Granular information is essential for applying analytics to water pollution, says Orianna Bretschger, whose US-based start-up Aquacycl uses microbial tech to cut the cost and carbon footprint of industrial wastewater treatment. “AI is only as good as the data entered into it.”
Without a full understanding of pollution levels, their sources and how they fluctuate, she sees dangers in rushing into use of AI. “If you’re applying AI with limited data sets, you’ll get bad answers and make the wrong decisions,” she warns.