Reference: Goovaerts, “Geostatistical prediction of water lead levels in Flint, Michigan: A multivariate approach.” Science of the Total Environment. 10. Jan 2019. DOI: https://doi.org/10.1016/j.scitotenv.2018.07.459
The people of Flint, Michigan knew their water was dangerous long before official confirmation. Descriptions of tap water back in the summer of 2014 were documented as dark shades of brown and orange, with a foul smell and metallic bite. The Michigan Department of Environmental Quality dismissed the complaints and assuaged all public complaints that Flint’s water supply was polluted. Eighteen months later, researchers discovered that the number of children with elevated lead levels in their bloodstream had doubled. It turned out that an estimated 140,000 people were exposed to lead from their drinking water. A class-action lawsuit was filed by citizens of Flint for complications of lead poisoning, which included accounts of autoimmune disorders, skin lesions, and cognitive decline.
Understanding the Flint Water Crisis
Lead levels in water are measured in terms of “parts per billion” (ppb). As an example: 1 gram of lead dissolved in 1 billion grams of water has a concentration of 1 ppb. The US Environmental Protection Agency (EPA) states that water containing over 15 ppb of lead is unsafe to drink. Collected tap water around Flint had levels of lead as high as 13,200 ppb- a level so high that it qualified as toxic waste. Many attribute Flint’s lead crisis to their drinking water source, the Flint River. But in reality, the high lead levels came from the very pipes used to carry water to its people. Highly acidic river water caused lead to leach directly out of the pipes and into the water supply. Flint, like many US cities, has old infrastructure pipes containing lead. These legacy pipes are treated with an inner corrosion-control coating. And so long as this coating stays in place, lead also stays put, and does not leached from the pipes. Flint was previously using Lake Huron and the Detroit River as water supply without any issues. The trouble began in 2014, when Flint switched their drinking water source to the Flint River. Drastic changes in water chemistry between these sources triggered lead leaching and ultimately lead poisoning in the citizens of Flint.
A hard truth to swallow is that wide-scale lead contamination is a possibility in many cities in the United States. The EPA estimates that roughly 10 million homes and buildings in the US receive water from pipes containing lead. Limited data on where these pipes remain and scarce funding for infrastructure makes the removal of legacy pipes difficult and costly, and with the proper anti-corrosive water treatment, the remaining lead pipes aren’t likely to leach lead. But as seen in Flint, changes in water chemistry or infrastructure failure can trigger releases of lead out of pipes and into the water.
At the height of the Flint water crisis, DIY at-home water testing kits became freely available to homeowners. Interested residents could pick up these kits, and follow enclosed instructions on sampling water from their own faucets. This yielded 18,760 samples collected for over a year at 10,341 unique sites. Data collected from “Citizen Science” projects like these are often criticized for their margin of inaccuracy and sampling variability. There is no way to know that every person monitoring their drinking water was taking the same exact steps, which could alter the data. But datasets this large are hard to come by and valuable for data analysts. Dr. Pierre Goovaerts, Chief Scientist at BioMedware Inc. in Ann Arbor, MI was able to combine these data with existing datasets to create predictive models to assess the risk of lead contamination in individual homes and neighborhoods. Existing datasets such as historical records and city infrastructure data was combined with the residential water tests to develop modeling of lead levels in the public water system. Lead levels could then be predicted for each of the sampling sites in Flint using a series of statistical models that develop geospatial patterns.
Models such as the ones developed by Goovaerts are used to estimate the probability of dangerous lead levels in drinking water on a home-by-home basis. This is significant when one considers how wide a range there was in reported lead water levels during the Flint Crisis. Despite the city’s shared water source, factors such as pipe infrastructure, proximity, and water stagnancy influenced lead readings that ranged from 1 to 13,000 ppb. Predicative modeling could help determine which homes are at higher risk of lead poisoning- and why. Goovaerts was able to model how factors like water temperature, location of the faucet, and stagnancy could change lead readings in the same house. Distrust of city officials leaves many people in Flint skeptical that their water will ever be safe enough to drink. Understanding the factors that caused variation in lead readings could help the city begin to regain the trust of its community.
Predictive modeling is only as powerful as the data used to calibrate the model. While many scientists may express concern for using “citizen science” data, there are many benefits to harnessing large datasets such as these to help shape predictions for varying water quality scenarios. Predictive modeling relies on large datasets that are often unfeasible for state or local governments to obtain. Lead pipes are expected to be removed from Flint by 2020 but it is unlikely that all lead-baring pipes will be removed from US infrastructure anytime soon. In the future, more cities could implement similar studies to predict at-risk waterlines and faucets. The damage caused to the City of Flint is irreversible. But a push for better data and more geostatistical modeling could make a world of difference in how we think about lead in drinking water, and how we mitigate future drinking water conflicts.