Wildfires increasingly threaten oil and gas wells in the western United States with disproportionate impacts on marginalized populations
Data files
May 28, 2024 version files 25.86 MB
Abstract
The western United States is home to most of the nation’s oil and gas production and, increasingly, wildfires. We examined historical threats of wildfires for oil and gas wells, the extent to which wildfires are projected to threaten wells as climate change progresses, and the exposure of human populations to these wells. From 1984–2019, we found that cumulatively 102,882 wells were located in wildfire burn areas and 348,853 people were exposed (resided ≤ 1 km). During this period, we observed a five-fold increase in the number of wells in wildfire burn areas and a doubling of the population within 1 km of these wells. These trends are projected to increase by the late century, likely threatening human health. Approximately 2.9 million people reside within 1 km of wells in areas with high wildfire risk, and Asian, Black, Hispanic, and Native American people have disproportionately high exposure to wildfire-threatened wells.
README: Wildfires increasingly threaten oil and gas wells in the western United States with disproportionate impacts on marginalized populations
Authors
David J.X. González*, Rachel Morello-Frosch, Zehua Liu, Mary D. Willis, Yan Feng, Lisa M. McKenzie, Benjamin B. Steiger, Jiali Wang, Nicole C. Deziel, and Joan A. Casey
*Correspondence to djxgonz@berkeley.edu
Summary
The dataset deposited here is associated with a peer-reviewed study published in One Earth in June 2024. Recent increases in wildfire activity in the Western United States have coincided with the proliferation of oil and gas development and substantial population growth in the wildland-urban interface. From 1984–2019, we observed a five-fold increase in the number of wells in wildfire burn areas and a doubling of the population within 1 km of these wells. These trends are projected to increase by the late century, likely threatening human health, with disproportionate impacts on racially marginalized populations.
We obtained geospatial data on the locations and operation dates of oil and gas wells from Enverus DrillingInfo. We intersected each well in the study region with a geospatial dataset on wildfires in the United States, drawing from both Monitoring Trends in Burn Severity (MTBS) and the National Interagency Fire Center (NIFC). Once we identified areas where wildfires and wells intersected, we used gridded population data from SocScape-30 to estimate how many people were exposed to these wells. To estimate future risks, we incorporated a gridded dataset of the Keetch-Byram Drought Index (KDBI), an indicator of wildfire risk, made available by the authors of Brown et al. (2021). Links to these datasets are provided below.
Data and file structure
Interim
wells_wildfire_intersection_state_year (folder)
Set of geospatial data (sf objects saved as R data files), each of which includes polygons for the area of intersection between wildfire burn areas and 1 km boundaries around oil and gas wells. People living within this intersection zone would be considered exposed to these wells.
wells_wildfire_intersection_state_year (.csv)
Tabular data
- state - Two-letter abbreviation for U.S. state
- year - Year of the observation
- intersection_area_km2 - Total area (in square km) of overlap between wildfires and a 1 km buffer around oil/gas wells
Processed
wells_wildfire_state_year
Tabular data counts of the wells in wildfire burn areas by state and year.
- state - Two-letter abbreviation for U.S. state
- year - Year of the observation
- n_wells - Count of oil and gas wells in wildfire burn areas
- n_wells_buffer_1km - Count of oil and gas wells within 1 km of wildfire burn areas
pop_exposed_state_year
Tabular data with estimates of the population exposed to wells in wildfire burn areas by state and year.
- state - Two-letter abbreviation for U.S. state
- year - Year of the observation
- pop_exposed_n - Estimate of population exposed to wells in wildfire burn areas (residing ≤ 1 km)
wells_individual_wildfires
Tabular data with
- wildfire_id - Identifier for the wildfire
- wildfire_name - Name for the wildfire provided by the reporting agency
- year - Year wildfire started
- state - State the wildfire started in
- data_source - Source of data, either MTBS or NIFC
- n_wells - Count of oil and gas wells in wildfire burn area
- n_wells_buffer_1km - Count of oil and gas wells within 1 km of the wildfire burn area
- n_wells_dates - Count of oil and gas wells with at least one operational date in wildfire burn area
- n_wells_dates_buffer_1km - Count of oil and gas wells with at least one operational date within 1 km of the wildfire burn area
- wildfire_area_km2 - Total area of the wildfire (in square km)
wells_kbdi
Tabular data with, for each well in the dataset, the maximum Keetch-Byram Drought Index (KBDI) value for each of the three time periods considered. These KBDI estimates are derived from data products provided by Brown et al. (2021).
- api_number - Unique identifier for each well
- kbdi_max_2017 - The assessed KBDI value for each well for 2017
- kbdi_max_2050 - The assessed KBDI value for each well for mid-century (2046-2054)
- kbdi_max_2090 - The assessed KBDI value for each well for late century (2086-2094)
- well_on_federal_land - Indicator (1 = yes, 0 = no) for whether the well is located on federal land
Sharing and Access Information
Publicly Available Datasets
We obtained shapefile data on wildfires from Monitoring Trends in Burn Severity (MTBS) and the National Interagency Fire Center (NIFC).
Gridded population for the United States, including data disaggregated by race/ethnicity, were provided by SocScape.
Other Datasets
We obtained data on oil and gas wells from Enverus DrillingInfo, a private data aggregation service that makes data available to researchers upon request.
Gridded historical and projected future KBDI estimates were requested from and provided by the authors of Brown et al. (2021).
Code
The codebase for this study has been deposited on Zenodo. Analyses for this project were conducted using R.