Equitable cleanup of Superfund sites leaving no U.S. community behind

Abstract

Superfund sites are recognized as the most contaminated locations across the U.S. Here we introduce two metrics: (i) the disparity percentage, which quantifies the overrepresentation of vulnerable populations in proximity to Superfund sites, and (ii) the Superfund exposure score, which evaluates the population proportion in a geographical region affected by such proximity. We employ the metrics to develop an Action Priority Matrix (APM) categorizing states and regions into four tiers of cleanup priorities, informed by a spatial analysis of 13,453 Superfund sites across the U.S. About 80% of the U.S. population live within 10 km of at least one Superfund site, with nearly 60% of them residing in areas lacking any cleanup efforts. Asian, Black, and disadvantaged populations are found to be disproportionately overrepresented in Superfund host block groups. Seven states are identified for urgent cleanups using the proposed APM, providing a systematic approach to equitable resource allocation for cleanups.

Introduction

The Superfund Sites program was established in the United States (U.S.) in 1980 as a direct response to the heightened public concern over the health impacts of toxic waste in New Jersey and contamination at Love Canal in New York. Following the Love Canal environmental disaster, the U.S. Congress passed the Comprehensive Environmental Response, Compensation, and Liability Act¹ (CERCLA), laying the legal groundwork for the U.S. Environmental Protection Agency (EPA) to investigate, list, and clean up contaminated sites. In 1986, amendments to the CERCLA, known as the Superfund Amendment and Reauthorization Act² (SARA), stressed the need for permanent solutions and innovative rehabilitation strategies. Several strategies were implemented (e.g., increased stakeholder participation and improved existing tools) to focus on risks to human health. Over time, the EPA’s approach to Superfund cleanup has evolved across different administrations. The first formal emphasis on EJ dates back to 1994 with Executive Order 12898, which directed federal agencies to identify and address disproportionately high and adverse environmental and health impacts on minority and low-income populations³. More recently, between 2021 and 2024, a suite of executive orders has required federal agencies to further integrate environmental justice considerations into policy development and implementation^4,5,6,7,8,9. For example, Executive Order 14008⁶ in January 2021 led to the development of the EPA’s Environmental Justice Action Plan. Superfund remediation efforts incorporated an environmental justice lens underscored in the National Environmental Justice Advisory Council report¹⁰ of May 2021. However, in 2025, Executive Order 14151¹¹ terminated some of these initiatives, including directives to dismantle Diversity, Equity, and Inclusion (DEI) offices and to eliminate screening tools and datasets developed to support environmental justice and equity efforts. Despite these changes, the underlying environmental inequities faced by vulnerable communities remain persistent. This highlights the critical need for independent, data-driven methodologies to guide equitable cleanup and policymaking, regardless of shifting political priorities.

The environmental injustice implications of Superfund sites have always been at the center of attention and controversial debates. These sites pose substantial health and environmental risks to communities, disproportionately impacting historically underserved and socially vulnerable communities (HUSVCs) who may already face additional systemic challenges. Past studies show significantly higher proportions of Black and minority populations adjacent to Superfund Sites^12,13,14. Tract-level analysis of the nation showed that Superfund sites in predominantly Black neighborhoods were less likely to receive National Priority List (NPL) designation¹⁵. The U.S. EPA has implemented several initiatives to address and adhere to environmental justice principles. In this context, EPA, along with other federal agencies and state governments, has developed a range of tools to support environmental justice analysis and inform decision-making related to site remediation, risk screening, and community engagement. Notable tools include EJScreen¹⁶, EnviroAtlas¹⁷, CalEnviroScreen¹⁸, MyEnvironment¹⁹, and Climate Resilience Screening Index²⁰, among others, to protect human health and prioritize HUSVCs. In its latest strategic plan, the EPA delineates its priorities and outlines a roadmap toward addressing climate change and promoting environmental justice²¹.

Funding for the remediation of a Superfund site is made smoother if that site is listed on the NPL or through the Superfund Alternative Approach (SAA) program. The NPL is defined as “the list of sites of national priority among the known releases or threatened releases of hazardous substances, pollutants, or contaminants throughout the U.S. and its territories”²². The EPA primarily prioritizes NPL sites for cleanup funding and timeline. A Superfund site is included in the NPL through three primary methods: (i) Receiving a Hazard Ranking System (HRS) score of 28.5 or higher, (ii) designation of a site by a state, and (iii) issuance of a public health advisory by the Agency for Toxic Substances and Disease Registry (ATSDR)²². The HRS score is a quantitative risk measure based on a scoring framework that considers several factors, including the likelihood that the site has released or potentially will release hazardous substances, the nature of waste, and the exposed people and the environment. There are several gaps in the calculation of the HRS. First, the cutoff score of 28.5 was originally set to cap the initial NPL list at 400 sites²³ but remains the selection threshold today. The efficacy and accuracy of the cutoff score of 28.5 in distinguishing high-priority sites from lower-priority ones in the present context remain unverified²⁴. Second, the HRS does not account for variations in community vulnerabilities and capacities.

Superfund sites not on the NPL may potentially receive funding for cleanup using the SAA. Superfund remediation funding is primarily sourced through four mechanisms: (i) funds provided by the Potentially Responsible Party (PRP), who is held liable for the cleanup²⁵; (ii) funds provided by the Superfund when the PRP is unable to pay; (iii) federal appropriations, which contribute to cleanups²⁶; and (iv) state government funding, which may include revenues from the excise tax on chemical manufacturers. It is important to note that listing a site on the NPL or addressing it through the SAA opens a pathway for federal involvement and potential access to cleanup resources. However, this designation does not guarantee immediate or full funding. The process of identifying and securing contributions from potentially responsible parties (PRPs) is often lengthy and legally complex. For sites without viable PRPs, federal funding becomes essential. Although the Superfund tax was reinstated to support such efforts, recent reports indicate that revenues have been lower than expected, and funding for orphan sites remains constrained²⁷. These limitations contribute to delays in remediation and prolonged exposure risks for nearby communities.

Environmental justice concerns also arise in the processes of selecting and remediating Superfund sites. For example, some studies²⁴ highlight that sites with cost-effective remediation measures may be excluded from the NPL, while less cost-effective sites are prioritized for inclusion. This exclusion can delay remediation efforts for certain sites, leaving HUSVCs near these locations exposed to hazardous conditions for prolonged periods. Such disparities underscore the need for more equitable prioritization frameworks that consider the unique vulnerabilities of affected communities, irrespective of their NPL status.

Despite continued efforts over the past few decades, critical gaps remain in current practices regarding a practical and standardized system that enables operationalizing environmental justice for listing and cleaning up Superfund sites while accounting for environmental and socioeconomic metrics. The allocation of an additional $3.5 billion to the Superfund through the Infrastructure Investments and Jobs Act⁷ offers an opportunity to clean up more sites. However, while the current screening approaches and scoring systems are comprehensive in assessing environmental risks, they fall short in considering the different capacities of vulnerable populations to cope with the risks they face. Here, we propose integrating environmental justice metrics to facilitate informed decision-making. Such integration can inform decision-making frameworks for enlisting Superfund sites in the NPL by prioritizing sites that directly affect HUSVCs.

We conducted spatial analyses of 13,453 Superfund sites across the U.S. to address the following key questions: (1) To what degree are disadvantaged communities overrepresented in blocks hosting Superfund sites relative to non-host blocks at the same geographical scale? (2) Does the extent of overrepresentation of disadvantaged communities in proximity to Superfund sites vary based on the cleanup status of these sites? (3) How can we integrate environmental justice metrics to provide opportunities for sites that do not have adequate funding (not listed in the NPL), allowing for the equitable distribution of resources? Answers to these questions guide the study toward its primary objective, which is to develop the Action Priority Matrix (APM) to support equitable decision-making for allocating resources for cleanup activities.

Results

We spatially analyzed 1336 sites identified as NPL sites, 439 sub-sites within NPL locations, and 66 sites on the SAA program. In addition, we examined 11,612 sites, which comprised both non-NPL and proposed NPL sites. We classified the sites into two distinct categories. The sites listed in the NPL, part of an NPL site, or part of the SAA program were categorized under “Active cleanup” based on their cleanup status. Sites not included on the NPL or proposed to be on the NPL were classified as “No cleanup.” The categories were established to investigate whether disparities vary based on cleanup status. While acknowledging the potential existence of cleanup protocols at non-NPL sites, it is assumed that NPL sites generally receive a substantial portion of the Superfund’s resources.

The results from our analysis are organized into three principal sections. The initial section delineates the composition of communities residing in Superfund host blocks at the national, regional, and state levels. The second section presents the results of the nationwide disparity analysis, where we estimate the overrepresentation of HUSVCs in Superfund host blocks at national, regional, and state geographical levels. Third, we present our proposed APM to facilitate decision-making for prioritizing cleanup and Superfund resource allocation.

Composition of communities residing in Superfund host sites

An estimated 254 million individuals, constituting nearly 80% of the U.S. population in 2015–2019, resided within a 10 km buffer around at least one Superfund site. Almost 60% of these people (148 million individuals or 45% of the US population) lived in blocks where cleanup activities were lacking. These population estimates were derived using a 50% areal containment method, where a block was included if at least 50% of its area fell within the 10 km buffer surrounding a Superfund site. We compared the sociodemographic and socioeconomic composition of populations residing in host blocks to those in non-host blocks at the EPA regional and national levels. The EPA consists of 10 regions, each responsible for the oversight and execution of the EPA’s programs. The regions with their corresponding states are shown in Fig. 1a.

Fig. 1: The composition of communities based on cleanup status.

a The ten EPA regions with their corresponding states, b Block level comparison of population proportions at the national and regional levels in Superfund host vs. non-host areas by educational attainment, poverty status, Black populations, and Hispanic populations. The red color indicates active cleanup or NPL (National Priorities List), and blue indicates no cleanup sites (non-NPL). Box plots show the median (center line), 25th and 75th percentiles (box edges), and whiskers extending to Q1 − 1.5 × IQR and Q3 + 1.5 × IQR (minimum and maximum), respectively. Sample sizes for each boxplot are given in Table S1, c The proportions of AIANNH (American Indian/Alaska Native/Native Hawaiian), Black, Asian, and Hispanic populations in blocks with no cleanup activities (non-NPL) versus those with active cleanup activities (NPL) across EPA Regions 1 through 10. The red color indicates active cleanup, and blue indicates no cleanup sites. Shapefiles were sourced from the U.S. Census Bureau TIGER/Line® dataset.

First, we examined sociodemographic attributes such as Black, Hispanic or Latino, Asian, and AIANNH (American Indian/Alaska Native/Native Hawaiian) populations. The median population of Black and Hispanic people was consistently higher in Superfund host blocks nationally and across most regions (Fig. 1b and Supplementary Table S1). Notably, we observed higher medians of the Black population at the national level and in EPA Regions 2, 3, 4, 5, and 6. Similarly, the median of the Hispanic population in host blocks exceeded that of non-host blocks in all regions, with the largest regional difference observed in Region 9. In terms of socioeconomic characteristics, the median proportion of the population with low educational attainment was also higher in Superfund host blocks, particularly in Region 9. Due to data sparsity (i.e., blocks with zero values), comparative analysis for Asian and AIANNH populations was limited in several regions.

We examined the racial composition based on cleanup status, that is, NPL (active cleanup) or non-NPL (no cleanup) blocks. The results (Fig. 1c) show a notably higher mean proportion of AIANNH in blocks lacking cleanup activities. Specifically, in Regions 6, 8, 9, and 10, the proportions are more than twice as high as those of blocks with ongoing cleanup activities. Similarly, the proportion of the Black population was also observed to be higher in blocks without cleanup activities across EPA regions. A similar trend is observed for the Asian population across all EPA regions.

Disparities in the exposure of vulnerable populations

We examined disparities in the exposure of vulnerable populations using two metrics: the percent difference disparity metric (D_p) and the ratio disparity metric (D_r). As explained in the “Methods” section, D_p is defined as the absolute difference in the proportion of a population group (e.g., low-income, racial/ethnic minority) between host blocks and the reference geography (e.g., state, region, or nation), expressed as a percentage (Eq. 1). This metric helps indicate the magnitude of over- or under-representation. The disparity ratio (D_r) is calculated as the ratio of the proportion of a population group in host areas to that in reference to the population (Eq. 2). A ratio greater than 1 indicates overrepresentation, while a ratio less than 1 indicates underrepresentation. While D_p and D_r are often correlated, each offers distinct interpretive value, and we report both throughout the analysis to provide a more nuanced understanding of disparity patterns. Disparities were assessed across three categories: (1) all Superfund sites, (2) NPL sites (active cleanup), and (3) non-NPL sites (no cleanup).

Our findings (Supplementary Tables S2–S6) reveal notable disparities in the racial and ethnic composition of Superfund host communities at both national and regional levels. At the national scale, Asian populations exhibit the greatest overrepresentation around all Superfund sites (D_p = 198.6%, D_r = 2.99), followed by the Black population (D_p = 100%, D_r = 2.00). Moving down to the EPA-region level, disparities for the Black population were even more pronounced in comparison to other racial groups. At this EPA regional level, our findings show a pronounced overrepresentation of the Black population in Region 1 (D_p = 857.5%, D_r = 9.58), Region 5 (D_p = 514.6%, D_r = 6.15), and Region 7 (D_p = 428.8%, D_r = 5.29). Meanwhile, regional disparities were also identified among Asian and Hispanic populations. At the state level, overrepresentation was found again to be highest among the Asian population, Particularly in Massachusetts (D_p = 1029.3%, D_r = 11.29), New Jersey (D_p = 872.9%, D_r = 9.73), New York (D_p = 685.4%, D_r = 7.85), and Oregon (D_p = 455.1%, D_r = 5.55). Some notable state-level disparities include the Black population in Massachusetts (D_p = 950.7%, D_r = 10.51), Connecticut (D_p= 606%, D_r = 7.06), and Nebraska (D_p = 596.3%, D_r = 6.96). Notable disparities were also observed among the Hispanic population.

Focusing only on NPL (active cleanup) sites produces a similar hierarchy. At the national scale, the Asian population (D_p = 267.8%, D_r = 3.68) remains the most overrepresented race/ethnicity group, followed by the Black population (D_p= 112.3%, D_r = 2.12) and Hispanic population (D_p= 68.7%, D_r = 1.69). No disparities were observed for AIANNH populations. Disparities at the EPA regional levels were pronounced for Black and Asian populations. Our findings show a pronounced overrepresentation of Black populations in Region 1 (D_p = 929.8%, D_r = 10.30), Region 7 (D_p = 579.7%, D_r = 6.80), and Region 5 (D_p = 547.4%, D_r = 6.47). Regional disparities among Hispanic, Black, and Asian populations were largest in regions 1 and 2. State-level disparities where most pronounced overrepresentation was found among the Black population in Massachusetts (D_p = 1110.7%, D_r = 12.11), Nebraska (D_p = 983.4%, D_r = 10.83), Connecticut (D_p = 593.7%, D_r = 6.94), and Minnesota (D_p = 583.8%, D_r = 6.84). Some notable state-level disparities include the Asian population in Oregon (D_p = 683.3%, D_r = 7.83), and Utah (D_p= 435.2%, D_r = 5.35), Hispanic population in New York (D_p= 412.4%, D_r = 5.12), and AIANNH population in Indiana (D_p = 95%, D_r = 1.95).

By contrast, non-NPL (no-cleanup) sites reveal a different emphasis: the Asian population are still nationally overrepresented (D_p = 207.9%, D_r = 3.08), yet the most striking regional disparity is for the Black population in Region 1 (D_p = 1269.1%, D_r = 13.69) and the Hispanic population in the same region (D_p = 864.1%, D_r = 9.64). Region 10 is the lone area where the Asian population overrepresentation is not observed.

We illustrated these sociodemographic disparities based on cleanup status (Fig. S2). The scatter plot illustrates the disparity percentage for each geographical region on two axes: NPL (active cleanup) on the x-axis and non-NPL (no cleanup) on the y-axis. The 1:1 line of equality serves as a benchmark for parity; in an ideal scenario, observations would cluster at or near this line, signifying negligible differences between the two cleanup statuses. Deviation from the line, therefore, indicates inequity, with points situated above it revealing regions in which disparities are more pronounced around non-NPL sites, an outcome representing the most acute environmental justice concern. Our results show that most points (geographical regions) cluster well above the diagonal for AIANNH populations. Among the Hispanic population, a majority of points are also above the line but spread more evenly, except for a few points, such as Region 2 and Maryland. Among the Black communities, most points sit below the line of equality except for some notable geographical regions (e.g., South Dakota, Region 10, Wyoming, Kentucky, and Idaho). Finally, for Asian populations, most points cluster along the line of equality.

We also examine the disparities in the representation of populations identified as “disadvantaged” between Superfund host blocks and non-host blocks. Our findings (Fig. 2 and Supplementary Tables S7–S11) reveal significant disparities in the presence of disadvantaged communities within Superfund host blocks at the state level.

Fig. 2: Disparity percentage of disadvantaged populations between Superfund host and non-host areas.

Disparities are shown for three categories: a all sites, b sites listed in the National Priorities List (NPL; active cleanup sites), and c sites not listed in the NPL (no cleanup sites). Values are capped at 100% and indicate the percentage of disadvantaged populations overrepresented in Superfund host blocks relative to non-host blocks within the same geographic level. A red color gradient denotes the magnitude of disparities, with lighter shades indicating lower disparity and darker shades indicating higher disparity. The designation of “disadvantaged” is based on the Climate and Economic Justice Screening Tool (CEJST). Shapefiles were obtained from the U.S. Census Bureau TIGER/Line® dataset.

Nationally, the analyses identified no overrepresentation of disadvantaged populations in all Superfund categories. At the EPA regional levels, the disparity (all) was particularly pronounced in Region 4 (D_p = 52.7%, D_r = 1.53) and Region 1 (D_p = 42.6%, D_r = 1.43). Substantial disparities at the state level were found in Nevada (D_p = 251.4%, D_r = 3.51) and Vermont (D_p = 183.1%, D_r = 2.83).

The disparity analysis for NPL (active cleanup) also showed no overrepresentation of disadvantaged populations nationally. At the EPA regional levels, the most significant disparities were found in Region 5 (D_p = 55.3%, D_r = 1.55), followed by Region 2 (D_p = 50%, D_r = 1.50) and Region 9 (D_p= 37.5%, D_r = 1.38). Notable state-level disparities for this category were found in Maryland (D_p= 241.1%, D_r = 3.41), Vermont (D_p= 214.3%, D_r = 3.14), Nebraska (D_p= 166.8%, D_r = 2.67), Indiana (D_p= 146.5%, D_r = 2.47), and South Dakota (D_p= 102.4%, D_r = 2.02).

Lastly, disparity analysis for non-NPL (no cleanup) was performed. At the national level, the disadvantaged population was overrepresented by 5.8% (D_p = 5.8%, D_r = 1.06). Notable disparities at the regional level were found in Region 5 (D_p = 59.2%, D_r = 1.59), followed by Region 9 (D_p = 19.1%, D_r = 1.19), and Region 7 (D_p = 10.8%, D_r = 1.11). The largest state-level disparities were found in Nevada (D_p = 251.4%, D_r = 3.51) and Minnesota (D_p= 137%, D_r = 2.37).

Proposed action priority matrix (APM) for allocation of cleanup resources

We adopted the Eisenhower (importance-urgency) matrix to develop a multi-attribute APM (Fig. 3) that incorporates environmental justice into decision-making for Superfund site cleanup. Madani^28,29 proposed using this matrix for public policy agenda setting and identifying the most prioritized policy actions and interventions with respect to the resource (e.g., financial, time, political capital, public support) limitations in the real world. By considering the disproportionate risk that disadvantaged communities are facing, our proposed APM facilitates a more equitable agenda-setting regarding the cleanup of Superfund sites. Each geographic unit (e.g., state, region) is placed into one of the four distinct cleanup classes based on the Superfund Exposure Score (SES) it receives and disparity, i.e., the degree of the overrepresentation of disadvantaged populations in blocks with Superfund sites relative to those without Superfund sites at the same geographical scale (disparity). These classes are as follows (Fig. 3): (1) balanced cleanup (Low-Low: low superfund exposure score and low disparity percentage), (2) focused cleanup (Low-High), (3) broad-based cleanup (High-Low), and (4) urgent cleanup (High-High). The quadrant boundaries are designed to be dynamic to ensure the cleanup strategy is adaptive, responding to the nuanced differences across geographical units. This flexibility allows policymakers to calibrate the matrix based on the scale of analyses, the level of resources available, and the socioeconomic and sociodemographic characteristics of the targeted geographic scale and community engagement. Considering these factors in delineating quadrant boundaries allows for targeted strategies for cleanup. Several studies from a wide range of sciences, including geography, health, and climate adaptation, have adopted the Eisenhower matrix, specifying their own approaches to dividing quadrants, providing decision-makers with a framework to effectively prioritize tasks^{30,31,32,33,34}. In this study, states are included within the matrix as an initial step to present an overarching framework. However, for targeted actions, it is essential to conduct the analyses at a higher resolution (i.e., county level) by recalculating both SESs and disparity percentage accordingly. Including counties within the matrix allows for more granular assessments that may be utilized to ensure resources are allocated to local areas where Superfund site cleanup is needed most.

Fig. 3: The proposed action priority matrix (APM) framework for guiding equitable Superfund cleanup decisions across U.S.

The matrix uses two metrics: Superfund Exposure Score (SES) on the x-axis, representing the proportion of a state’s population within 10 km of a Superfund site, and disparity percentage (D_p) on the y-axis, capturing the overrepresentation of disadvantaged populations in host areas. The APM defines four cleanup priority groups (Balanced, Focused, Broad-Based, and Urgent) based on combinations of high and low SES and D_p values. Quadrant boundaries can be determined using national or regional medians, user-defined thresholds, or policy-specific benchmarks tailored to the scale and objectives of the analysis.

In addition, we propose action items for each classification within our matrix at the local, corporate, academic, and federal levels (Table 1). This multi-level approach is designed to emphasize the importance of engaging multiple actors in addressing environmental justice concerns in Superfund remediation. While we highlight the potential value of collaboration, we acknowledge that Superfund cleanups are frequently marked by stakeholder conflict and competing priorities. As case study research has shown³¹, PRPs often have financial incentives to contest liability, delay remedial action, or challenge stringent cleanup standards, which can limit the speed and scope of remediation. Despite these constraints, the action items in our matrix are intended to provide a structured starting point for discussion and policy coordination, even in settings where consensus is difficult to achieve. Geographical regions in the urgent cleanup quadrant require the highest level of prioritization within the APM. At the local level, once the EPA investigation reports are disseminated to the state and communities, the CERCLA mandates an opportunity for public participation, particularly in the planning and selection of remedial actions³⁵. In addition, state and local government officials are encouraged to bolster preparedness and provide venues for education to ensure informed feedback. Finally, two-way communication should be established, and education on contaminants and remediation methods should be provided through partnerships with different organizations. Specifically, partnerships between researchers and community members can enhance the effectiveness of educational workshops and support community-driven research. At the corporate level, private companies in charge of remediation should be incentivized to hire locally. Finally, at the federal level, transparency should be a primary focus to share progress regarding cleanup efforts and streamline funding and permitting processes.

Table 1 Potential action items across each cleanup quadrant at the community, corporate/industry, academic, and regional/federal levels

In the analysis, we defined the boundaries of the quadrants based on the national medians of the disparity percentage (59.7%) and the SES (0.74%). Figure 4 illustrates the relationship between disparity percentage and SESs aggregated at the state level across the U.S. The findings reveal seven states within the urgent cleanup quadrant. Among these, Nevada (SES = 0.82, D_p = 251.4%), Maryland (SES = 0.84, D_p = 141.1%), Illinois (SES = 0.87, D_p = 71.7%), and Indiana (SES = 0.80, D_p = 73.5%) yield the highest disparity percentage and SESs. These high SES values reflect the extensive presence of Superfund sites near population centers. New York, Illinois, Wisconsin, and Indiana were found to be the most populous states within this quadrant. The same states were found to be home to a substantial proportion of the disadvantaged population in Superfund host blocks. Our results suggest that prioritizing sites in these states could help address racial disparities through the equitable allocation of resources. For instance, in Nevada, the total population characterized as disadvantaged is approximately 1 million individuals, or 30.6% of Nevada’s population. In Maryland, approximately 0.9 million individuals, or 14.4% of Maryland’s population, are identified as disadvantaged. New Jersey yields a disadvantaged population of 2.1 million, constituting 22.1% of New Jersey’s population.

Fig. 4: State-level application of the action priority matrix (APM).

a Shows the placement of states using Superfund Exposure Score (SES) and disparity percentage (D_p) values, with bubble size representing the total disadvantaged population and color hue indicating the share of that population in Superfund host blocks. National median thresholds (SES = 0.74, D_p = 59.7%) are used to define the quadrant boundaries. b Zooms in on the high-priority region outlined in (a) to provide a detailed view of states approaching the upper limits of SES and D_p. State labels use official two-letter postal abbreviations.

Within the focused cleanup group, Vermont (SES = 0.41, D_p = 182.1%) yields the highest disparity percentage and SESs. However, the disadvantaged population within 10 km of a Superfund site in this state was relatively low, with 0.09 million individuals. Wisconsin and Iowa emerged as the most populous states within this quadrant, home to 0.94 million (16.2% of Wisconsin’s population) and 0.49 million (15.6% of Iowa’s population) disadvantaged individuals, respectively.

States that fall within the Broad-based Cleanup quadrant exhibit high SES but lower D_p. This combination suggests that while disadvantaged populations may not be disproportionately affected in relative terms, a significant segment of the overall population, including both disadvantaged and non-disadvantaged groups, is exposed to potential environmental hazards. These are states where exposure is widespread and systemic, requiring broad, coordinated, and sustained intervention strategies. For example, California, Florida, Ohio, Pennsylvania, and New Hampshire all fall within this category. California has an SES of 0.89 and a D_p of 7.1%, while Florida has an SES of 0.82 and a D_p of 29.2%. These states also have some of the largest absolute numbers of disadvantaged individuals living near these sites: 14.7 million people in California, which represents 37.4% of the state’s population, and 8.4 million in Florida, accounting for 40% of its population. Although the relative disparities may be moderate or low, the scale of exposure highlights the urgency of broad-based action. States in this quadrant would benefit from strategies that combine technical cleanup with inclusive engagement and education. Expanding voluntary cleanup programs and providing incentives for local governments and private stakeholders to take proactive roles can accelerate progress. Incorporating environmental justice education into public school curricula can raise awareness and foster long-term advocacy. Enhancing local emergency preparedness efforts is crucial, particularly in regions where Superfund sites are vulnerable to additional risks from natural hazards, such as flooding or wildfires. Transparent, accessible communication platforms can help ensure that affected residents are kept informed and involved throughout the remediation process. Public-private-academic partnerships may also be especially valuable in these states, where scale and complexity demand both technical expertise and grassroots collaboration.

In contrast, the Balanced Cleanup quadrant comprises states with both low SES and low D_P values. This indicates relatively minimal exposure and low disparity. States like Montana (SES = 0.57, D_p = 0), Arkansas (SES = 0.35, D_p = 0), and Wyoming (SES = 0.42, D_p = 0) fall within this group. Although these states may not face immediate or severe environmental justice challenges related to Superfund sites, it is critical that this does not lead to complacency. These areas can take proactive steps to preserve environmental quality and promote long-term community resilience. Recommended actions in these states include maintaining robust environmental monitoring systems to detect any emerging risks. Proactive outreach and community engagement, despite the current low levels of exposure, can foster awareness and build trust, laying the groundwork for rapid response if future issues arise. Supporting community-led initiatives, such as citizen science programs or local advisory committees, can strengthen local capacity and foster environmental stewardship. Finally, these states should work to integrate Superfund site information into broader planning efforts, such as climate resilience strategies or hazard mitigation plans, to ensure long-term sustainability and equity.

Discussion

The disproportionate exposure of HUSVCs to Superfund sites is often the result of a combination of historical, social, economic, and political factors. These factors include historical discrimination and segregation practices such as redlining and discriminatory zoning laws that have systematically pushed HUSVCs to less desirable areas, often in proximity to Superfund sites³⁶. The limited economic capacities of members of HUSVCs often restrict their housing options, reducing their ability to relocate and increasing their likelihood of residing in areas with low property values³⁷. In addition, members of HUSVCs often lack the political power and representation necessary to influence decision-making processes. Although many Superfund sites originated from historical leaks, spills, or illegal dumping that predated modern hazardous-waste siting regulations, these legacy releases continue to have long-lasting impacts on surrounding communities.

Our study aimed to develop an APM that provides a robust and practical basis for equitable decision-making in allocating resources for cleanup activities. We propose that the initial step towards equitable distribution of resources for Superfund site remediation involves identifying HUSVCs that may benefit the most from cleanup activities. To achieve this goal, we calculated disparity percentages and SESs to position U.S. states within our matrix.

We not only highlight the prevalent environmental justice issues but also underscore the need for an equitable approach to resource allocation for remediation. The development of the APM, as proposed in the study, is a step toward advancing such an approach. From the findings of this study and similar studies, we raise a critical question: why has the overrepresentation of AIANNH, Black, and disadvantaged populations’ proximity to Superfund sites remained unaddressed in resource allocation? Traditional approaches, where NPL designation determines funding allocation, are insufficient in addressing challenges faced by HUSVCs. The APM we propose here offers a practical mechanism for addressing these existing issues by integrating environmental justice and equity attributes. We provide an equity-based tool for the just and equitable allocation of resources for Superfund remediation.

The study’s findings can inform a wide range of policies aimed at preventing future forms of inequity. For instance, our study findings can enhance the effectiveness of several EPA equity priority actions³⁸ by providing data-driven insights that can inform decision-making. By integrating the disparity percentage and superfund saturation scores, agencies can better assess community impacts, target capacity-building efforts, and engage with HUSVCs more effectively. Additionally, our metrics can inform civil rights compliance and strengthen community science initiatives.

The findings from this study provide actionable insights for developing policies aimed at reducing environmental inequities. Agencies could utilize the metrics along with the APM proposed in this study to design policies that identify and prioritize communities disproportionately burdened by Superfund sites. These metrics and APM could be used to determine areas requiring immediate remediation or to inform preventive measures that address systemic disparities. This data could be integral in shaping equitable resource allocation strategies within federal programs, including Justice40, as well as at the state and local levels.

In addition to policy development, the findings can enhance decision-making processes for entities involved in NPL listing and remediation. By integrating the disparity percentage and SES into resource allocation frameworks, agencies can ensure remediation efforts are directed toward communities with the greatest need, based on both environmental and socioeconomic vulnerabilities. Our metrics can also facilitate the development of new HRS scores that are scaled for the varying vulnerabilities and capacities of communities. Adjustments to the existing HRS scores could incorporate these metrics to better reflect cumulative impacts and the unique challenges faced by underserved communities. Practical implementation could involve the development of geospatial tools or dashboards that visualize these metrics, thereby supporting transparent and equitable decision-making.

The results also provide a framework for equity-focused community engagement. By leveraging these metrics and the APM, agencies can better align their capacity-building efforts and funding with the specific needs of disadvantaged communities. For instance, targeted outreach programs could involve HUSVCs in the decision-making process, empowering them to advocate for their priorities. Strengthening partnerships with local organizations and stakeholders would ensure that interventions are both culturally sensitive and community-informed, fostering trust and collaboration between communities and agencies.

Beyond their immediate application to Superfund remediation, these metrics and the APM have broader implications for addressing systemic environmental justice issues. These metrics and the APM could inform climate adaptation strategies, helping to address the cumulative impacts of environmental hazards and climate change in disadvantaged communities. They may also advance civil rights compliance by identifying areas where potential systemic discrimination exists and supporting the development of targeted strategies to address inequities in environmental decision-making.

Applying the proposed APM could lead to several measurable outcomes for HUSVCs. By identifying areas with the most pressing needs for Superfund remediation, the matrix can potentially mitigate health disparities resulting from prolonged exposure to hazardous substances. It can inform strategies to address the cumulative impacts of climate change, particularly in communities already facing vulnerabilities due to their proximity to Superfund sites. The matrix also supports targeted resource allocation, enabling decision-makers to prioritize interventions that provide the greatest benefit to disadvantaged populations.

Future research should address the practical challenges of implementing the APM, including the complexities of intergovernmental coordination necessary for Superfund remediation on Tribal lands. These challenges encompass jurisdictional conflicts, cultural considerations, and resource constraints, which require a nuanced understanding and collaborative approaches to overcome. We recommend future research to refine the APM for site-by-site decision-making, enabling more precise prioritization at finer spatial scales. This approach could enhance its utility for addressing localized disparities and tailoring remediation strategies to specific community needs. Further studies should also explore barriers to implementation, such as logistical, institutional, and policy-level constraints, and identify strategies to overcome them.

Although our results are informative, they are not without limitations. The presence of military bases among Superfund sites potentially complicates substantive interpretations of our results becuase the active-duty military population and civilian population in a given area may differ with respect to their demographic attributes. Our analyses do not consider the spatial distribution of Superfund sites, which could be located away from population centers, and our results may be sensitive to the buffer distance used in the study. To maximize consistency with the measurement approach used in EJScreen, we applied a 10 km buffer distance in our analysis. Future research should assess the sensitivity of our results, which are based on high-quality publicly available data sources, when alternative buffer distances are applied. However, it is important to note that the consistency between our results and those found in previous studies suggests that our substantive results would be robust to alternative buffer distance specifications.

Methods

Input data

The 2015–2019 American Community Survey (ACS) data was used to examine demographic compositions at the Census block group geographical level. In this study, “blocks” will be used as a shorthand throughout the paper to refer to Census block groups. Census data extracted from the ACS includes the proportion of the population in each respective block that is non-Hispanic Black (Black), non-Hispanic American Indians, Alaskan Natives, and Native Hawaiians (AIANNH), non-Hispanic Asian (Asian), Hispanic (any race) (Hispanic), low-income (personal income 200% below the federal poverty threshold in the last 12 months), and without a high school education (persons ages 25+ without a high school diploma). Here, we consider racial minority groups, including Blacks, Asians, Hispanics (any race), and AIANNH, as HUSVCs. The HUSVC indicators align with those used in previous literature. They are available via various screening tools used to assess social vulnerability, such as the Social Vulnerability Index by the Centers for Disease Control and Prevention³⁹. The analysis included a total of 217,289 blocks within the contiguous U.S.

In addition to census data, we employed the Climate and Economic Justice Screening Tool (CEJST)⁴⁰ to identify “disadvantaged communities”. This tool assesses regions nationwide, evaluating significant burdens across categories, including climate change, energy, health, housing, legacy pollution, transportation, wastewater, and workforce development. The CEJST employs a ranking system based on percentile thresholds, where the 90^th percentile is a cutoff for most burdens. According to the CEJST, a census tract is designated “disadvantaged” if it meets one of the following criteria: (1) resides in a tract that surpasses thresholds for at least one burden, or (2) is in land within federally recognized tribal boundaries. CEJST employs a ranking system based on percentage thresholds to identify disadvantaged communities. Specifically, communities are considered disadvantaged if they are located in a census tract that meets the thresholds for at least one of the tool’s categories of burden. The threshold is typically set at the 90^th percentile for most burdens, meaning the top 10% of census tracts for each burden are designated as disadvantaged. Each category of burden in the CEJST tool (e.g., air pollution, water pollution, energy burden) has its own distinct threshold criteria based on available data. For example, in the case of air pollution, the threshold is based on the 90^th percentile of particulate matter concentration in census tracts. For energy burden, it is the top 10% of census tracts with the highest percentage of household income spent on energy costs. The CEJST disadvantaged designation is assigned at the census tract level. To apply this designation at a more granular level, we assumed that all blocks within a disadvantaged Census tract are considered disadvantaged. While this assumption introduces some generalization, it enables a meaningful comparison and spatial overlay between social disadvantage indicators and the exposure-based metrics used in APM. Moreover, communities located within federally recognized tribal boundaries are automatically designated as disadvantaged, regardless of whether they surpass the thresholds for individual burdens. While these automatic designations are aligned with federal priorities and commitments, it is important to consider the potential implications of bias within the analysis. Automatically designating tribal lands as disadvantaged could result in skewing results by inflating the representation of burdens or masking heterogeneity within tribal lands. For instance, some tribal areas may not meet the same burden thresholds as other disadvantaged tracts but are still included in the analysis due to the automatic designation. Moreover, a study argues that current federal screening tools, including CEJST, often fail to capture Indigenous-specific environmental justice concerns adequately⁴¹. Conversely, the automatic designation does not differentiate between levels of burden among tribal lands, potentially obscuring areas that require greater intervention. Nevertheless, this automatic designation is a recognition of systemic inequities faced by tribal lands.

Data on active Superfund sites in 2023 was sourced from the Superfund Enterprise Management System (SEMS) Database⁴², an official repository of Superfund site data. The SEMS database consists of the locations of all sites, given by either their address or latitude and longitude. They were geocoded based on available addresses using Geocodio⁴³ for sites consisting of only addresses. The generated corresponding coordinates were then plotted onto a map using the XY table-to-point tool in ArcGIS Pro (version 3.1.3). From the database, we analyzed a total of 13,453 sites consisting of 1,336 sites identified as NPL sites, 439 as sub-sites within NPL locations, 66 sites on the SAA program, and 11,612 sites, which comprised both non-NPL and proposed NPL sites.

Host block selection

Three distance-based approaches are widely used in environmental justice scholarship to approximate the number of people living in close proximity to environmental burdens. These include (i) the 50% areal containment method, (ii) the Boundary intersection method, and (iii) the areal apportionment method^37,44,45. These methods use the locations of facilities to assess the sociodemographic and socioeconomic characteristics of all geographical units within a uniform specified distance. However, each method has its own limitations. In this study, we apply the 50% areal containment method to classify blocks as either a host or a non-host (Fig. S1). Specifically, blocks in which at least 50% of the area is intersected within a 10 km buffer are designated as hosts. While the selection of buffer distances is inherently subjective, we use a 10 km radius, aligning with the proximity scoring approach used by EPA’s EJScreen 14. All spatial analyses were performed using ArcGIS Pro (version 3.1.3).

Disparity analysis

We analyzed block-level data for superfund host blocks at the state, EPA regional, and national levels and compared them with non-host blocks on the same geographical scales. We used EPA regional levels in our analyses to inform the EPA regional offices regarding disparities found in their respective regions.

The disparity percentage calculations used in the study are consistent with the method previously employed by Vahedifard et al.⁴⁶. In our analyses, the disparity percentage (D_p) represents, in percentage terms, the extent to which each vulnerable group of interest is disproportionately overrepresented in host blocks relative to non-host blocks within the same geographical scale. For instance, to calculate the disparity of the disadvantaged population in State A, we use the following equation:

$${D}_{p}(\%)={\left[\frac{\%\,{{\rm{Disadvantaged}}}\; {{\rm{pop}}}.{{\rm{in}}}\,\left({{\rm{host}}}\right)-\%\,{{\rm{Disadvantaged}}}\; {{\rm{pop}}}.{{\rm{in}}}\,({{\rm{non}}}-{{\rm{host}}})}{\%\,{{\rm{Disadvantaged}}}\; {{\rm{pop}}}.{{\rm{in}}}\,({{\rm{non}}}-{{\rm{host}}})}\right]}_{{{\rm{State}}}\; {{\rm{A}}}} \times 100$$

(1)

Equation 1 quantifies the disparity in disadvantaged population proportion between host and non-host blocks, within a specific state (State A). The disparity is calculated by dividing the difference in the percentage of the population in the host that is disadvantaged and the percentage of the population in the non-host that is disadvantaged within a given state (State A) by the percentage in the non-host that is disadvantaged within the same state. This approach was used to calculate D_p at each geographical level used in the study. Here, we calculate the disparity for two additional categories of Superfund sites. Disparities for blocks hosting NPL sites relative to non-host blocks (Active cleanup) and disparity for blocks hosting non-NPL sites to non-host blocks (No cleanup).

In addition to quantifying percentage differences between host and non-host blocks, we also calculate a disparity ratio. This ratio compares the percentage of a population in host blocks to the percentage in non-host blocks within a given geographical region, as defined by:

$${D}_{r}={\left[\frac{\%\,{{\rm{Disadvantaged}}}\; {{\rm{pop}}}.{{\rm{in}}}\,\left({{\rm{host}}}\right)}{\%{{\rm{Disadvantaged}}}\; {{\rm{pop}}}.{{\rm{in}}}\,({{\rm{non}}}-{{\rm{host}}})}\right]}_{{{\rm{State}}}\; {{\rm{A}}}}$$

(2)

Superfund exposure score

The study introduces the SES, a metric designed to assess the proportion of the population in a given geographical region that resides within the combined buffer zones when multiple Superfund sites are present. A higher SES indicates a great level of exposure for the population relative to the total population of the region, thereby providing an equitable measure to compare regions with potentially heightened exposure. The SES for state A is calculated using the following equation:

$${SES}={\left[\frac{{\sum }_{i=1}^{{{n}}}{{\rm{Exposed}}}\; {{\rm{population}}}\; {{\rm{within}}}\; {{\rm{buffer}}}\,i}{{\sum }_{j=1}^{m}{{\rm{Total}}}\; {{\rm{population}}}}\right]}_{{{\rm{State}}}\; {{\rm{A}}}}$$

(3)

where n denotes the number of buffer zones surrounding Superfund sites in State A, and m represents the total number of population units (e.g., census blocks) within State A.

This equation can be applied to each geographical region to determine their scores. The study uses this measure and disparity percentage to facilitate targeted cleanup activities and Superfund resource allocation.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.