AI helps Toledo get the lead out
The Environmental Protection Agency has awarded Toledo, Ohio, $200,000 to use artificial intelligence to identify lead pipes that are endangering the safety of drinking water.
Using funds from EPA’s State Environmental Justice Collaborative Problem-Solving Cooperative Agreement (SEJCA) Program, Toledo will develop a machine learning model that will predict the likelihood of a home having lead pipes. That information will allow the city to identify and prioritize those homes facing serious health risks from pipes that must be replaced.
Working with water infrastructure analytics consultant BlueConduit, the University of Toledo, the Toledo-Lucas County Health Department and local partners, the city aims to reduce lead exposure through well-tested, data-driven prioritization techniques, according to the project summary.
The partners will assess the probability that a home’s water pipes are lead based on existing parcel and neighborhood-level data and a representative sample of water service lines in the city. A predictive algorithm will more accurately pinpoint the location of lead service lines without having to dig up pipes to determine whether they are copper or lead.
It can cost between $3,000 and $10,000 per home to replace lead pipes, with part of the cost coming from the trial and error usually involved in accurately locating lead service lines, according to BlueConduit. This cost savings makes using the technology a priority.
With the predictions in hand, Toledo officials will be able to prioritize remediation efforts, guiding decisions on whether homes should receive targeted education, water filters or replacement of their lead pipes.
BlueConduit also worked with officials in Flint, Mich., in 2016 and 2107 to deploy a predictive model to more accurately locate homes with lead service lines. Using decades-old handwritten notes, annotated maps and service records for homes in Flint, the company’s founders described how they cross checked that information with data on the age, value and location of homes to build a predictive model to identify lead pipes. “Leveraging new algorithmic and statistical tools, we are able to produce a significantly more complete picture of the risks and challenges in Flint,” they wrote in 2016.
The model hit an 80% accuracy rate, the company said, but the project was derailed over objections from members of the public who thought the AI-based model was unfairly ignoring their homes. After city officials stopped using the algorithm and started digging up whole blocks looking for lead service lines, only 15% of the excavated pipes were found to be lead, slowing the replacement program and adding costs.
“This project will reduce lead exposure risks for Toledo’s most vulnerable residents by using historical data and technology to target lead service line replacements,” said EPA Region 5 Administrator Kurt Thiede. “We are excited to fund such a worthy project, one that could serve as a model for cities around the country.”
Through the SEJCA program, EPA is providing grants over a two-year period to advance collaborative work with communities facing environmental justice challenges. The goal is to further understand, promote and integrate approaches to provide meaningful and measurable improvements for public health and the environment.
Connect with the GCN staff on Twitter @GCNtech.