This month, the U.S. Environmental Protection Agency (EPA) released a new drinking water dataset that has now been incorporated into the Environmental Justice Screening and Mapping Tool (EJScreen). For the first time, environmental justice communities will be able to link Safe Drinking Water Act (SDWA) violations to affected populations. Prior to this release, there was no publicly available national data source that enabled researchers, policymakers, and advocates to determine who was drinking polluted water.

For environmental justice communities known to be at higher risk of exposure to industrial pollutants, living with poor water infrastructure, and facing long-standing public health threats related to water quality, community members and advocates now have a tool available to better assess drinking water quality and the cumulative burden of pollutants. Residents will be able to monitor SDWA violations in their community, take action to protect their health and safety, demand corrective action against polluters, and document needed infrastructure improvements.

Background

“Community water systems” are public water systems defined under the SDWA. There are about 50,000 community water systems across the country. In Fiscal Year 2022, about 43% of community water systems violated one or more SDWA standards, and 7% had health-based SDWA violations assessed by the EPA. Until now, however, the violations of drinking water standards by community water system providers could not be connected to the local residents drinking their water.  

There have been some datasets available prior to this, but they have varied widely in availability by state. In the majority of states, there was no known publicly available, centralized dataset of community water system service area boundaries. Only 11 states provided service area boundaries for over 90% of their community water systems. To gauge households affected by water violations, service area boundaries had to be approximated based on jurisdictional boundaries that might incorrectly include some households or exclude others. 

Over the past couple years, the U.S. Geological Survey and a collaboration of the Environmental Policy Innovation Center (EPIC), the Internet of Water, and SimpleLab each undertook efforts to map community water system service boundaries. They were limited, however, to publicly available state data and did not include small drinking water systems.

This year, the EPA launched an effort to build upon this previous work and to expand it nationwide. A particular strength is that the EPA dataset includes small water systems that account for about 50% of all community water systems. These systems, about 12% of which are in mobile home parks, provide water to fewer than 500 people, reportedly account for about 80% of SDWA violations, and are more likely to serve low-income and disadvantaged populations.
 
The new dataset was released in July 2024, and a Drinking Water Noncompliance Score has now been incorporated into Version 2.3 of EJScreen. The dataset provides information about community water systems with violations and affected households down to the Census block level. 

Methodology

The EPA dataset compiles available state data and supplements that with data modeling. About 40% of service area boundaries are based on available state or municipal data; about 60% are modeled.¹ A first step in developing the dataset was to distinguish public systems from private systems. The EPA relied upon three states with particularly strong service area data – California, Connecticut, and New Jersey – and multiple variables to “train” the model through machine learning. If a census block is partially public, it is defined as “public.” After identifying those community water systems that are public, the EPA mapped those specific systems to Census blocks. 

The EPA found that the modeled data is quite reliable when compared to available state data. About 95% of modeled census blocks identified as being served by public water systems are correct when compared to actual state data. When model-generated service area boundaries do not completely match the actual boundaries, they do capture most of the population served by the water system. 

In developing the EJScreen Drinking Water Non-compliance Score, the EPA considered the severity of violations and how long they lasted. Acute violations that are chronic (e.g., open for five years) receive a higher score on the EJScreen. As with other indicators within the EJScreen, communities are ranked by percentile based on the severity of their exposure to pollutants.

Limitations

Where the dataset and the associated EJScreen Drinking Water Non-Compliance Score is based on modeling, it is not 100% accurate. But modeled data do correlate strongly with actual service boundary data, as described above. The data also only pertain to public water systems. They do not include private wells, and the dataset cannot predict household level water quality. In other words, it can reliably predict whether the water provided to a household line is contaminated or not, but it cannot inform water quality potentially affected by lead pipes in the home. 

Significance

The drinking water dataset provides valuable information to assess the cumulative burden of environmental health exposures. Using the mapping tools, which are linked to EPA enforcement and compliance data, can help environmental justice communities not only identify those households within a census block that have likely been exposed to drinking water contaminants, but they can also identify which contaminants are involved. In the future, there is an opportunity to continue to enhance the dataset and to incorporate it into other screening tools, such as the Climate and Environmental Justice Screening Tool (CEJST). 

What’s Next 

EPA staff and the Environmental Policy Innovation Center emphasize that while this dataset represents a significant step forward in the ability to identify households and communities directly affected by drinking water violations, there is more work to do. 

• Environmental justice advocates are encouraged to review the map, particularly in those states that rely upon modeled data, to confirm the accuracy of the EPA map. Inaccuracies can be reported via email to the EPA. 
• Advocates can also work with states without publicly available service boundary data to encourage the availability of such data and enhance the accuracy of service boundary mapping.
• To learn more about the EJScreen, advocates can attend EPA office hours on August 21 at 12:00 EST.

1. The states with service area boundaries that rely entirely on modeling are Alabama, Alaska, Colorado, Delaware, Georgia, Hawaii, Idaho, Illinois, Indiana, Iowa, Louisiana, Maine, Maryland, Massachusetts, Michigan, Minnesota, Missouri, Montana, Nebraska, Nevada, North Dakota, Ohio, Oklahoma, Oregon, South Carolina, South Dakota, Vermont, Virginia, Wisconsin, and Wyoming. ↩︎