Code
# Load libraries and data
library(tidyverse)
sfdata <- read_csv("data/SF_NPL_data.csv")
glimpse(sfdata)
fire_counties <- read_excel("data/fire_dataset.xlsx", sheet = "Counties")
glimpse(fire_counties)Wildfire Risk at Superfund Sites in the United States
Introduction
As climate change shifts weather patterns around the globe, many parts of the United States are at increased risk from natural disasters, including wildfires. (1) In addition to contributing to air pollution and causing property destruction, wildfires may play a role in mobilizing preexisting contamination in the environment, particularly at Superfund sites containing hazardous waste. (2) The United States Environmental Protection Agency’s Inspector General has recommended that more attention should be paid to natural disaster risk at Superfund sites. From a preliminary study of 157 federal facilities around the country (Superfund sites owned by the federal government), they found that 20% of this subset of sites were at risk from wildfires, and only a third of those sites factored wildfire preparedness into their decision documents (3). Nearby communities to Superfund sites already face exposures to legacy pollution, and without proper preparedness, these exposure pathways could be exacerbated by damage to site control mechanisms. By expanding beyond federal facilities to consider the full universe of Superfund sites in the nation, we build on this identified risk management gap.
Data
Our first dataset is the EPA’s Superfund National Priorities List (NPL). (4) This is a CSV file that includes the site, location, EPA region, contaminated media, and the contaminants found at the site. This was originally in the long format with each row showing the contaminant. This was cleaned by removing the contaminant type for this analysis, collapsing the media column, and concatenating the contaminated media type by site into one string. This set was processed into wide format so there is one observation per site and each site could be observed as one point on the map. For the bar chart visualization, media types were expanded into wide format to obtain frequency counts.
Our second dataset was the USDA Forest Service: Wildfire Risk to Communities. (5) Wildfire risk only at the county level was extracted, but this set contained risk on the state, community, and tribal area levels. This data contained the GEOID of the county so could be merged with the tigris state boundaries shapefiles in order to make our choropleth map and assess risk by EPA region. It was a more complex process to merge this data file with the Superfund file, which did not have a GEOID, because many county names had to be cleaned in order for the join to work. One data gap in this set was the fact that Puerto Rico, the Virgin Islands, and the Pacific Islands did not have fire risk data, so while their Superfund sites are presented in the visualization, fire risk is not.
The risk value used in this data set was the national RPS percentile within the US. The RPS value denotes Risk to Potential Structures, which measures the hypothetical danger to a structure on any given plot of land. (5) We selected this value to be able to do nationwide comparisons, and because the potential damage to infrastructure is a particularly relevant measure for the type of risks that we are evaluating.
Visualization 1: [Title Here]
Describe what this plot shows, and why it’s important.
Code
# your code hereVisualization 2: [Title Here]
Describe what this plot shows, and why it’s important.
Code
# your code hereVisualization 3: [Title Here]
Describe what this plot shows, and why it’s important.
Code
# your code hereOR Shiny app: Wildfires & Waste: Exploring Compound Environmental Hazards
We created a Shiny App to allow users to explore the spatial relationships between wildfire risk and Superfund site locations. Because this is a specific and fairly technical issue, our main audience for this app is the community of environmental professionals who are involved in Superfund site cleanup and risk management. In particular, personnel in the EPA’s Superfund program may find this tool valuable for identifying vulnerable sites and planning more protective remedies. State and local governments, as well as nonprofit stakeholders and community groups, may find the tool valuable for setting location-specific priorities and allocating resources for disaster preparedness. Public users may also find the app valuable for exploring potential risks in their immediate geographic vicinity. Additionally, researchers may find this tool valuable for setting future investigation priorities.
The first visualization in our app shows a map of the United States that allows the user to explore the map with zoom and pop-up features. We have layered blue dots showing Superfund site locations on top of a choropleth map showing a yellow - to - red color gradient to depict the relative wildfire risk of each county. By clicking each county, users can reveal fire risk scores, while clicking each site reveals the type of media that is contaminated.
Our second visualization is a reactive boxplot that allows users to select EPA regions, which highlights both that region in the boxplot, and the states included within it in the map. This visualizes the distribution of fire risk across EPA regions, while also allowing the user to see how county risk is distributed within that region. By adding jittered dots, we make it easy to detect outlying counties, and by sorting regions in order of increasing fire risk, we can better visualize differences. Each dot represents a county and is sized based on how many Superfund sites are located there, enabling the user to easily see outliers and explore the total Superfund count through a tooltip.
Our third visualization shows a bar chart that depicts the types of environmental media found at contaminated sites across the country. This chart is reactive to the regional and state selection menus, allowing the user to isolate individual states and see what contaminated environmental media predominate at Superfund sites in that state. This allows the user to explore contamination trends with more granularity. It also gives some indication of what environmental media might be best to prioritize in future research on Superfund site vulnerability to wildfires.
Lastly, the app contains several call-out boxes with information on our conclusions. This helps direct the user to important takeaways from this relatively data-heavy application.
Interpretation
Through our analysis of Superfund site wildfire risk, we are able to build on the EPA Inspector General’s previous assessment of wildfire risk to federal facility sites. We found that 24% of Superfund sites nationwide are located in high fire risk counties, defined as being above the 80th percentile nationwide. While this tool cannot be used to predict exact risk at individual sites, it reiterates the EPA IG’s conclusions that more attention should be allocated to managing fire risk at Superfund sites.
We also found that there are substantial differences in risk between different areas of the country. High wildfire risk areas are concentrated largely in the Western United States, and especially in EPA Regions 8, 9 and 10. High risk areas were also demonstrated on the east coast in Florida and in New Jersey, which also has the greatest number of Superfund sites nationwide. This indicates that these regions may benefit most from proactive planning around Superfund site vulnerability.
The distribution of media across the states also gives a sense of what exposure pathways researchers and risk managers might want to prioritize. Across most states, groundwater was the most frequently contaminated environmental medium, indicating that particular attention might be warranted to how wildfires are anticipated to impact groundwater treatment systems. The states did differ somewhat in their media profiles, and future analysis that expands the “Other” category which dominated some states is warranted.
Limitations
Though several Superfund sites exist in Puerto Rico, the Virgin Islands, and the Pacific Islands, the wildfire data set was not complete for these areas. This is a limitation in projecting risk for island communities. In the future, we are interested in exploring specific classes of contaminants that are present at each site, and assess comparative vulnerabilities based on contaminant type. We are also interested in further exploring the population and demographic variables that relate to both fire risk and Superfund proximity, applying an environmental justice lens to assess areas of particular vulnerability.
The biggest limitation of our app came in the deployment phase. Because of the data intensivity of loading county-level shapefiles, the app does not run for very long before it crashes due to running out of memory. We tried to mitigate this by simplifying county geography and saving our csv files as RDS files locally to avoid processing large spreadsheets during app deployment, which is the best we can do without upgrading to a more expensive tier of Shiny.
Conclusion
By connecting these two datasets, we were able to effectively visualize which Superfund sites are highly threatened by wildfires. With global temperatures rising and wildfires becoming more prevalent, risk assessment needs to account for the increasing threat of unpredictable natural disasters such as wildfires on preexisting contamination sources. While more research is necessary to better understand risks and potential interventions, our dashboard contributes to highlighting the importance of mitigating chemical contamination to communities in high wildfire risk zones.
References
Content and Data Sets: 1. NASA. (n.d.) Wildfires and Climate Change Website. https://science.nasa.gov/earth/explore/wildfires-and-climate-change/
US EPA Office of the Inspector General (2026). Review of Federal Superfund Site Risks due to Natural Disasters and Extreme Weather. https://storymaps.arcgis.com/stories/340c1111e4d041da97d6ee23588c7b77
EPA Office of the Inspector General (2026). Evaluation of Risks to Federal Facility Superfund Site Remedies. https://www.epa.gov/office-inspector-general/report-evaluation-risks-federal-facility-superfund-site-remedies-wildfires
US EPA Superfund Data Reports (2026). Contaminant of Concern Data by Decision Document. https://www.epa.gov/superfund/superfund-data-and-reports
USDA Forest Service (2025). Wildfire Risk to Communities Conditional Risk To Potential Structures https://data-usfs.hub.arcgis.com/datasets/usfs::wildfire-risk-to-communities-conditional-risk-to-potential-structures-image-service/about
Chart Style and App Development:
For the Boxplot with graduated dot sizes, we were inspired by an application shared by Dr. Hu: https://climawatch.climate.mathematica.org/
We consulted R-Graph-Gallery.com for exploring different visualization types: https://r-graph-gallery.com/
For coding assistance, we used Shiny Assistant for help with developing the frame of the application, and Claude and ChatGPT for help with debugging code issues
https://gallery.shinyapps.io/assistant/#
https://claude.ai/new
https://chatgpt.com/