Waterways like this feed into larger river systems including ones affected by PFAS contamination. (Photo courtesy of Jachan DeVol/Unsplash)
Most people have never even heard of PFAS, but it’s already surrounding us. Most commonly known as polyfluoroalkyl substances; PFAS are a group of man made chemicals that don’t break down in the environment. North Carolina happens to be one of the states with the largest PFAS problem in the country.
In 2017 it was found that DuPont had been dumping PFAS from its Cumberland County facility into the Cape Fear River for more than 40 years. In this current era, drinking water for an estimated 3.5 million North Carolinians sits above the health standards for these chemicals.
The Cape Fear Watershed by itself is over 9,000 square miles and supplies water to roughly 2 million people, but the problem isn’t only for one singular river. These chemicals have turned up in water systems and private wells across the whole state, from industrial discharge, firefighting foam or even biosolids spread on farmland.
The good news is, it is possible to remove PFAS from water, although the process is quite pricey. Testing each and every well in the state of North Carolina for the many different types of PFAS is another task that requires much time and money. That is a large reason why many wells are left untested, and there is a large lack of information on where the contamination is at its worst.
This is where machine learning can step in. Instead of testing each and every well, many scientists have been building up models to predict which wells are most likely to contain high levels of PFAS.
One study using groundwater data built a random forest model that could identify high-PFAS wells with 91% accuracy, without needing to test them all first. A separate large-scale analysis found that similar models can achieve over 90% accuracy in both predicting contamination levels and identifying likely pollution sources.
In theory, this could allow environmental regulators to target certain wells or water systems that have the highest risk of containing PFAS. This would save time and resources instead of trying to test every well they suspect may have contamination.
For a state like North Carolina, which has a large problem with PFAS in its water, but much less funding for water safety, this method could discover many contaminated wells before they are even tested.
