Lead-Free Water Challenge: What We Learned About How To Prepare Your Inventory Without Knowing Everything Yet

Dunrie Greiling and Ian Robinson, BlueConduit

One year has passed since the launch of the Lead-Free Water Challenge, EPIC’s pilot technical assistance program seeking to boost lead service line replacement programs in six municipalities in partnership with Center for Geospatial Solutions, Blue Conduit, and WaterPIO

Lead exposure through drinking water is, unfortunately, a risk many families and communities face. The water crisis in Flint, Michigan was an early indication of the nationwide issue of lead pipes in the drinking water system. 

The EPA’s recent Lead and Copper Rule Revision (LCRR), announced in 2020 and codified December 2021, requires communities to inventory the materials of the pipes (specifically the service lines - the pipes that carry water to a home). Where the pipe is made of lead/galvanized or where the pipe material is unknown, utilities must create water sampling, pipe replacement replacement, and communications plans by October 16, 2024. 

October 2024 is coming quickly.

Inventory Drives LCRR Compliance

An accurate service line inventory is the first step to creating the following components to comply with the following LCRR requirements: 

  • Replacement plan

  • Water sampling plan

  • Testing at schools and child cares

  • Annual communication and notifications to  homes served by lead/galvanized pipes or pipes of unknown materials

Yet Utility Records are Often Incomplete, Inaccurate, and Unorganized

Index card records on service lines. Photo taken at Detroit Water and Sewerage Department. Note: their records have been digitized.

In our work with communities across the nation and working with EPIC on the Lead-Free Water Challenge, we have seen all kinds of things: rooms full of undigitized index cards, paper records destroyed by water, handwritten and unreadable records, unique descriptions, and water systems that say they have no records at all. 

Peering at a scribbled note on a paper card, it may be hard to discern an “L” for lead from a “C” for copper. We recently heard from a utility director that an inspector consistently wrote “lead” to mean “pipe” or “line” not as a material for the pipe.       

Perhaps most frustrating of all is that these records, where they even exist and are legible, do not necessarily reflect reality. Although the records may have been accurate at the time the data was recorded, the “ground truth” may have changed since - for instance, if work has been done either by the utility or the homeowner without any update to the records. Sometimes, as many as half of the records we test are inaccurate.

If an accurate inventory unlocks compliance, what can utilities do without good records?

Predictions Fill in the Gaps

Utilities cannot budget and plan without understanding the scope of the issue. 

We don’t have to rely on outdated and incomplete records to start getting a better understanding of how many and where lead lines are. Using machine learning, we have shown that we can make accurate predictions about pipes we haven’t yet seen from information we do know. Blue Conduit started this work in Flint, Michigan, and have now used predictive analysis in over fifty communities. 

Here are two example maps from recent work in Toledo, Ohio - the first showing what was known about verified materials at the start of their lead pipe replacement program, the second shows a map filled in with predictions.

A map of service line materials in Toledo, Ohio showing lots of dark blue “no record” locations.

A map of service line materials in Toledo. Ohio showing predictions of service line materials: the golden yellow is high likelihood of lead.

To learn more about this assessment in Toledo, and how it fits into national efforts to fund lead pipe replacements, read Blue Conduit’s latest storymap.

Toledo then shared this data in a public-facing map that enables residents to see address-level data for both service line segments across neighborhoods and the schedule (prioritized by risk) for lead line replacement efforts, at https://toledo.oh.gov/residents/water/lead-service-lines/city-side

The schedule for lead service line replacements in Toledo, from their website: toledo.oh.gov/lead-lines

Toledo’s public-facing map, available on their website: toledo.oh.gov/lead-lines

Toledo’s map allows visitors to zoom in and see the service line materials (verified or predicted) at individual home addresses. It also shows the status of the lead line replacements, including this home which had its line replaced in April. 

The predictions in their service line inventory helped Toledo budget, plan, and ultimately cut years off of their lead pipe replacement plan (from a projected end date of 2050 to a now projected end of 2027).

Are Predictions Acceptable? 

Yes! The EPA has recognized the use of statistical prediction in its Guidance for Developing and Maintaining a Service Line Inventory and statistical prediction is an option in its Template inventory spreadsheet. We outline all of the details in our post on statistical prediction in the EPA Service Line Inventory Guidance. When using predictive modeling, it is essential to follow best practices in data science to make sure that the predictions are accurate.

How to Make a Good Prediction

You may recall from science or statistics classes that how you sample data affects your conclusions. This is as true of service line sampling and predictions as it is anywhere else. 

We use information we have about the properties, historical information on local plumbing standards, information on other work done to the water infrastructure, and most importantly, the ground truth discovered from inspections (or replacements). A full explanation is outside the scope of this article, but you can find more about our predictive model in the “how it works” page on our website.

A key piece of information we need to know is the actual pipe materials at a representative sample of homes in the system. This is the needed ground truth for the predictions.

How to Prepare Your Inventory in the Absence of Perfect Information

  1. Collect the records you do have. Salient information on preparing your data is summarized in this post by one of the other Lead-Free Water Challenge partners.

  2. Do not accept your records as truth.

  3. Collect ground truth: sample at a representative set of homes.

  4. Use statistical prediction to fill in the gaps. The accuracy of your predictions will depend on the representativeness of your ground truth data.

  5. Incorporate new information on service line materials into your inventory as you conduct replacements or do other water infrastructure work.

Our modeling approach and other data considerations are explained in more detail in a joint whitepaper we authored with ASDWA: Principles of Data Science for Lead Service Line Inventories and Replacement Programs.

Where are the Lead Pipes in the US?

We now know that the risk of lead pipes is not spread evenly for all, some states and some communities within states have more risk, based on when the housing stock was built, local regulations around construction and plumbing materials, and many other factors. 

The NRDC has published state-level lead pipe estimates showing which states have the most lead pipes total and the most lead pipes per capita. In 2021, they also published county-level lead risk analysis

We have seen in our work in over fifty communities that the lead pipes are not spread evenly within an affected city. The lead tends to be concentrated in areas with older housing, lower income levels, and more racially diverse neighborhoods

This map shows Toledo neighborhoods with lead service pipes. In Toledo as in many other cities, the lead is concentrated in the older urban neighborhoods. Screenshot from our storymap on funding lead pipe replacements.

What to Do Next 

The problem of lead in our water system developed over time. With current funding, the US government and cities across the US are committed to change that as quickly as possible. For example, the bipartisan Infrastructure Investment and Jobs Act (IIJA) includes $2.9 billion in FY 2022 to replace lead service lines and $15 billion total over the next five years. 

BlueConduit’s national estimate of water-system level lead service line predictions. To get a prediction, you will need the public water system ID (PWSID) for your water system (you can look up PWSIDs here).

For Utilities and Regulators

Utilities should act quickly to access this funding and others as the October 2024 inventory deadline will come quickly. Let’s get the lead out of the ground so that we don’t go any longer with the risk of lead exposure through drinking water. 

For Families and Households

Communities across the nation are working on this issue. Until all of the lead pipes are gone, it’s important to assess the risk for your family - and take measures, such as checking your service line for lead, and if necessary,  using lead-certified point of use filters.

  • Your utility may provide you a water test free of charge. Research options in your community. If not, here are water test options

  • Your utility may be distributing water filters free of charge. If not, you may want to look into a point-of-use filter for your water. Simple Lab is doing independent tap water filter reviews, at the time of this writing they have published two: Brita filter review and PUR filter review

  • What to do if you have lead in your water (NRDC)

    The Environmental Policy Innovation Center (EPIC) would like to thank Dunrie Greiling and Ian Robinson and the outstanding teams at BlueConduit for their collaboration on the Lead-Free Water Challenge and their partnership on the broader goal of replacing toxic lead pipes faster, more efficiently, and equitably across the country.

Previous
Previous

It is possible to work quickly, iteratively and in community to tackle gaps in water data.

Next
Next

What We Learned From Evaluating Drinking Water Funding & Why It’s Currently Too Hard to Effectively Track