Context Federal Superfund sites are some of the most polluted in the United States. This dataset contains a multifaceted view of Superfunds, including free-form text descriptions, geography, demographics and socioeconomics. Content The core data was scraped from the [National Priorities List (NPL)](https://www.epa.gov/superfund/national-priorities-list-npl-sites-state) provided by the U.S. Environmental Protection Agency (EPA). This table provides basic information such as site name, site score, date added, and links to a site description and current status. Apache Tika was used to extract text from the site description pdfs. The addresses were scraped from site status pages, and used to geocode to latitude and longitude and Census block group. The block group assignment was used to join with the Census Bureau's [planning database](https://www.census.gov/research/data/planning_database/2015/), a rich source of nationwide demographic and socioeconomic data. The full source code used to generate the data can be found [here, on github](https://github.com/4d4stra/Federal_Superfunds). I have provided three separate downloads to explore: - priorities_list_full.json: the NPL containing all geographic, site information, text descriptions, and Census Bureau data from the relevant block groups. - pdb_tract.csv: the planning database aggregated on the tract level with an additional indicator (has_superfund) noting whether or not the tract contains the address of a Superfund site. - pdb_block_group.csv: the planning database aggregated on the block group level with an additional indicator (has_superfund) noting whether or not the block group contains the address of a Superfund site. Some caveats: 1. The planning database contains 300+ columns. For a full description of these columns, please see the documentation [here](https://www.census.gov/research/data/planning_database/2015/). 2. Since the Google geocoder is relatively aggressive in providing address matches, geocoding was done through a hierarchy of queries (full address, city-state-zip, and zipcode only) to prevent gross errors. The address string used to geocode is noted through the 'geocode_source' column. 3. While this data is linked to demographic and socioeconomic data based on either the block group (tract for pdb_tract.csv), the impacts of a particular site's pollution may extend beyond these geographic regions. Acknowledgements I would like to thank the EPA and the Census Bureau for making such detailed information publicly available. For relevant academic work, please see [Burwell-Naney et al. (2013)](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4228303/) and references, both to and therein. Please let me know if you have any suggestions for improving the dataset!