A narrative review of spatial-temporal data sources for estimating population-level exposures to oil and gas development in the United States

Abstract

Purpose of review:

Oil and gas development is a rapidly expanding industry that may impact population health. However, much of the research to date is conducted state-by-state, partially due to exposure data limitations. New developments related to national-scale oil and gas development data sources offer the opportunity to extend studies beyond single-state analyses. We review the current data options, highlighting their advantages, disadvantages, and ideal use-cases.

Recent findings:

Five data sources suitable for national-scale epidemiologic analyses of oil and gas development were identified. Private sector data offer detailed production information but have limited accessibility. Nongovernmental sources are often specialized, focusing on specific aspects like chemical or methane exposure. Government agency data, while typically less detailed, provide useful linkage tools for cross-industry analysis.

Summary:

This review clarifies the strengths and limitations of these sources, facilitating national-level exposure assessment and broadening the geographic reach of oil and gas development-related epidemiology in the U.S.

INTRODUCTION

Oil and gas development continues to rapidly expand across the United States (U.S.). Natural gas production increased by 48% from 2014 to 2023, while crude oil production increased by 62% in the same period (1). This industry is highly geographically dispersed, with geological reserves yielding resource production across the country (Figure 1). Estimates indicate that about 5% of the U.S. population (18 million people) reside within 1.6 km (1 mile) of this industry (2,3). Given the wide range of environmental exposures from oil and gas extraction near local communities (e.g., air pollution, drinking water contamination, persistent light and noise pollution) and compounding factors (i.e., social stressors that accompany a changing local economy) understanding the influence of oil and gas development on population health is critical for informing health-protective policy decisions (4–10).

Figure 1:

Geographic distribution of U.S. oil and gas geology and production, adapted from the U.S. Energy Information Administration

Residential proximity to oil and gas development has been associated with a wide range of health risks, including adverse pregnancy outcomes, asthma exacerbation, cardiovascular events, depressive symptoms, and mortality (9–40). Concerned communities continue to call for research and action to protect population health (41–48). Measuring oil and gas development exposure is challenging, and the methods used by researchers to measure proximity, intensity, and activity have advanced over time (49–51). Many epidemiologic studies of oil and gas development to date are limited to single-state settings, partially due to the complexities of working with regional or national exposure data. Differences in data reporting, such as missing information on the timing of oil and gas site development or varied reporting standards by state or county, have created problems in generating uniform exposure estimates across state and other administrative boundaries. National-scale studies often facilitate comparison of industry-specific policies across state lines, allow for pooling of health data to investigate rare outcomes, and builds evidence for policy impact on different levels of governance (i.e., comparing federal, state, and local policies) (52–55).

Identifying locations of oil and gas development sites over large geographies can be difficult, creating barriers for exposure assessment in epidemiologic studies. Oil and gas development is generally regulated and permitted at the state level, so national-scale database must be pieced together on a state-by-state basis, a challenging task given substantial differences in reporting and characterization across state lines (56). Despite the geographic dispersion of oil and gas development sites and the variability in data reporting across states, many national-scale efforts are underway to develop integrated databases to identify and track these hazards, each with underexplored advantages and disadvantages. The recent emergence of national-scale oil and gas well databases creates opportunities for population health research on oil and gas development to extend research questions from single-state studies to multi-state or national analyses. However, the availability of multiple oil and gas development databases and the lack of clear guidance on their data sourcing, availability, and relative merits poses a design challenge for researchers.

Using a combination of systematic review and expert opinion, we review the characteristics with the corresponding advantages and disadvantages of available oil and gas development data sources. Our goal is to inform exposure assessment in future U.S. national scale epidemiological research. This review is designed to orient newcomers to the field of important data considerations for research on oil and gas development.

METHODS

By adapting the SALSA (i.e., Search, Appraisal, Synthesis, Analysis) method (57), we created a review framework to characterize the amount and quality of data sources to study national scale exposure to oil and gas development for epidemiologic research. Briefly, we reviewed the literature to identify spatial-temporal data sources used for oil and gas development exposure measurement in epidemiologic studies in the U.S. Using web searches and consultations with researchers in the field of oil and gas development, we identified sources of oil and gas development data being used in epidemiologic research, appraised the utility of the data using categories developed to be relevant to exposure derivation, synthesized by conceptualizing the pros and cons of each data source, and analyzed by quantifying the use of each data source in the relevant literature (57,58). To facilitate future research in the field, we further detailed key characteristics of oil and gas development data that may be valuable to researchers, including important industry terminology and definitions.

We collected and synthesized information about data sources’ temporal resolution, pollutant measurement, production reporting, and geographic scope across the U.S. We evaluated source-by-source from most comprehensive to least comprehensive, keeping in mind that each source has attributes that may be valuable for different research questions (Table 1). We then compared the data sources to each other by evaluating advantages and disadvantages and provided a list of attributes and characteristics that are shared or that differ.

Table 1:

Attributes of identified U.S. inclusive data sources.

Question	EIA	Enverus	FracFocus	HIFLD	OGIM
How many, and what proportion of U.S. states are included in this source? (n, %, state abbreviations)a	30, 60% (AK, AL, AR, AZ, CA, CO, FL, KS, KY, LA, MD, MI, MS, MT, ND, NE, NM, NV, NY, OH, OK, OR, PA, PA, SD, TN, TX, UT, VA, WV, WY)	32, 64% (AL, AK, AZ, AR, CA, CO, FL, KS, KY, LA, MD, MI, MS, MO, MT, ND, NE, NV, NM, NY, OH, OK, OR, PA, SD, TN, TX, UT, VA, WA, WV, WY)	26, 52% (AL, AK, AR, CA, CO, ID, KS, KY, LA, MI, MS, MT, NC, ND, NE, NV, NM, OH, OK, PA, SD, TN, TX, UT, WV, WY)	34, 68% (AL, AK, AZ, AR, CA, CO, FL, IL, IN, KS, KY, LA, MD, MI, MS, MO, MT, ND, NE, NV, NM, NY, OH, OK, OR, PA, SD, TN TX, UT, VA, WA, WV, WY)	34, 68% (AL, AK, AZ, AR, CA, CO, FL, IL, IN, KS, KY, LA, MD, MI, MS, MO, MT, ND, NE, NV, NM, NY, OH, OK, OR, PA, SD, TN TX, UT, VA, WA, WV, WY)
What type of organization does this source originate from?	Federal, public	Private	Non-governmental, nonprofit, organization	Federal, public	Non-governmental, nonprofit, organization
Does this source utilize API or UWI numerical identifiers?	No	Yes, API & UWI	Yes, API	Yes, API	No
What temporal information is included about the first time the well was drilled?	None	Spud date, first production date, completion date	Job start date is captured for each disclosure	Completion date	Spud date, completion date
What temporal information is included about the last time the well was used for production?	None	Last production date	Job end date is captured for each disclosure	None	None
What does this source include related to production activity of oil and gas?	Oil or gas production volume or status is not captured.	Production is measured using cumulative volume of oil, gas, etc.	Oil or gas production volume or status is not captured.	Categories of Producing, Active, Non-Active, Unknown, Well Development, Storage Well/Maintenance Well/Observation Well Producing, Non-Active Well are documented, but no production volumes are captured.	Categories of Abandoned, Drilling, Inactive, Other, and Producing are captured. Production volumes (oil and gas) are captured for some observations in the “production” layer.
What time-varying information is included related to oil and gas production over the well’s life cycle?	None	A measurement of the production volume during the first month, first 6 months, first 12 months, first 24 months, first 60 months, and last 12 months since the well was drilled is captured. Additionally, a “months produced” variable is captured.	None	None	Production year and production days are captured for some observations
What production type information does this source include?	Wells are classified as oil or gas and are stored in different files respectively.	Includes many categories describing production type, including oil, gas, and oil & gas.	Oil or gas production type is not specified	Includes 23 categories describing production type including oil, gas, and oil & gas.	Oil or gas production type is not specified, but can be inferred for some observations using production volume
What drilling type information does this source include?	None	Drilling types directional, horizontal, vertical, and unknown are included.	This source only includes hydraulic fracturing sites.	Some of this information is embedded in the “production type” variable. Could be parsed but appears incomplete.	Drilling types of deviated, directional, high angle, in perforation, low angle, multilateral, natural drift, slant vertical, and horizontal are included
Does this source include information about chemicals used during drilling?	No	No	Yes	No	No
Does this source include information about associated emissions?	No	No	No	No	No
Does this source include any non-U.S. sites?	No	Yes, Canadian sites are included.	No	Yes, Canadian sites in British Columbia and Manitoba are included if within 100 miles of the U.S. border.	Yes, global oil and gas wells are included where publicly available data exists.
Does this source include any other potentially valuable linking information (other than API/UWI)?	Yes, U.S.GS basin.	Yes, operator name.	Yes, CAS numbers of chemicals disclosed	Yes, NAICS codes, permit number, operator name.	Yes, Facility name, operator name.
In this data source, how many total wells are there?	~500,000 gas wells, and ~600,000 oil wells recorded	~5,000,000 rows in the “wells” set and ~4,000,000 in the “production” set.	~215,000 (represent disclosures, not individual wells)	~1,500,000 rows	~2,500,000 records
Anything not captured by questions above?	Oil and gas data are provided in separate files.	Offers academic licensing for researchers.	Geocoding is not standardized to one projection system.	Likely going behind an access barrier soon.	Designed for examining methane leaks and vents.

^aNote that some states are not located on viable geology for oil and gas development. Therefore, 100% state coverage is not expected nor realistic.

Search and Identification of Data Sources

Our search began with oil and gas development epidemiologic studies identified in a scoping review by Deziel et al., published in 2020, which identified a list of published journal articles that used oil and gas development data sources to evaluate population health (n = 29) (59). We then completed forward citation chaining used Google Scholar to identify epidemiologic studies that cited the Deziel et al. review and incorporated those subsequently published articles on the health effects of oil and gas development (n = 18) (34,60–76). We appended this search with additional studies from the updated Aker et al. 2024 review (n = additional 10 studies) (10,36,77–85). We used this literature sample (n = 57 studies) to identify which data sources researchers used to define exposure to oil and gas development.

Since the absence of a dataset from the literature necessarily does not preclude its utility, we supplemented our literature search by leveraging our coauthor team’s domain expertise in the public health impacts of oil and gas development to identify additional national-scale datasets with potential utility. This step is particularly important in case any issues related to publication bias have occurred in the research process (86), potentially leading to certain data sources being preferred over others.

Appraisal of Data Sources

We generated an a priori list of interrelated questions to evaluate the utility of each data set for exposure assessment in an epidemiologic framework:

1. How many, and what proportion of states are included in this source?

2. What temporal information is included about the timing of drilling and production?

3. What drilling type information does this source include?

4. What production type information does this source include?

5. What does this source include related to production activity and volume of oil and gas?

6. What temporal information is included about the last time the well was used for production?

7. What temporal information is included related to oil and gas production over the well’s life cycle?

8. How does this source include information about chemicals used during drilling?

9. How does this source include information about associated emissions?

10. How accessible is this source (e.g., open access, academic license, paid contract use)?

These questions were designed to extract key information about each data source. In spatial epidemiology, boundaries of space and time are valuable for all study designs: questions A, B, and F parse how each source compared along those attributes (52). Different types of drilling (e.g., vertical, horizontal, directional) and different types of production (e.g., oil, gas) may pose different risks to nearby populations, and differentiating by drilling type, production type, and potential hazards allows epidemiological studies to be more pointed; questions C, D, H, and I show how each data source compares across drilling type, production type, and potential hazard types. Production volume may be important to researchers since not all oil and gas development sites produce the same volume of output, which is likely correlated with volume of and pollution (85); questions E and G are designed to better understand how each data source accounts for oil and gas development site production volumes. Finally, question J reveals the accessibility and practicality of each data source, a key piece of information for future research.

Synthesis and Analysis of Exposure Data

We examined our appraisal results, focusing on each data source’s advantages and disadvantages. The overarching principle that guided our appraisal was that more variables and more completeness of these variables would indicate a higher quality source, provided that the additional variables were of sufficient quality based on our team’s domain expertise (e.g., data missingness, regional coverage). We placed the advantages and disadvantages of each source into context: discussing, including its utility for generating exposure metrics, described how each source might be leveraged in different research contexts, and identified any contexts not captured in the prior steps.

RESULTS

Data Quality and Key Considerations

When working with data related to oil and gas development, there are several industry-specific terms and concepts that clarify the characteristics of each site. To orient newcomers to this field, we described the definition and importance of key terms related to the oil and gas industry, specifically highlighting the points that guide our evaluation of each data source. This list of terms is designed to provide standardized definitions for key variables and concepts, as opposed to a comprehensive glossary of oil and gas development terminology.

API/UWI Identifiers and Data Linkages

Originally developed by the American Petroleum Institute (API), the API (sometimes referred to as “U.S. Well Numbers”) identifier is an up-to-14-digit number assigned to each petroleum industry well in the U.S., no matter the operator or owner. The identification of wells (i.e., an oil or gas extraction site) with a single number allows comprehensive data management. Digits 1–10 refer to the state, county, and well in that order, while digits 11–14 vary in meaning across states but can refer to a wellbore, completion, or conversion of well type. We recommend referring to a specific state’s regulatory materials if interested in using digits 11–14. Today, the Professional Petroleum Data Management Association owns and issues the API number standard; however, not all regulatory agencies choose to use this standardized identifying system (e.g., Indiana does not issue API numbers, Texas uses only the first 10 digits) (87–89).

The use of API numbers in a data source is valuable. API numbers allow for direct data linkage, rather than relying on spatial matching or other fuzzy methods, and allow for the extraction of additional information about a drilling site, such as a conversion of well type or completion. API numbers also allow for data verification, comparing the locating digits to the supplied spatial coordinates to ensure an accurate measure of location. This feature is particularly valuable when evaluating the quality of spatial exposure data. Additionally, the inclusion of API numbers allows data sources to be merged with each other. For example, if a combination of data fields across two sources were valuable to a researcher, API/UWI identifiers are a valuable tool to systematically generate a combined dataset representing all fields of interest.

Temporal Measurements

Important dates in oil and gas development production include spud date, last production date, first production date, and completion date. Spud date is the date when the drill bit enters the ground (90). First production date is the first recorded date that a site has extracted oil or gas (91). In hydraulic fracturing, wells must be “completed” after spudding to produce hydrocarbons (i.e., oil or gas); the first date a well can produce hydrocarbons is the completion date (91). Last production date is the last recorded date that a site extracted oil or gas (90). These dates are key for epidemiologists to construct time-bound exposure measurements. However, it is worth considering that the data reveal some uncertainties: production dates are almost always reported on the first of the month in any data source, which is unlikely to be the case in reality (31).

Spatial Measurements

Considerations related to spatial measurements are important to researchers using proximity as an exposure metric. One example is the potential difference in locations between surface holes and bottom holes in oil and gas development sites. Surface hole locations are where a well is drilled using machinery on the surface well pad, whereas bottom hole locations may be hundreds of meters away laterally from the surface hole (92). Additionally, a single surface hole may have multiple bottom holes; this feature may or may not be captured by the final digits of that site’s API number, depending on the state (92).

Production Types and Volume Measurements

In the context of epidemiologic research, production type typically refers to the product being extracted, drilling type refers to the method of extraction, and production volume quantifies the amount of oil, gas, or both being extracted. oil and gas development sites in the U.S. can produce oil, natural gas, or both (90). Which product oil and gas development produces and at what volume depends largely on geology, which varies by region in the U.S. Hydraulic fracturing (colloquially known as “fracking”) refers to using high-pressure liquid to extract oil or gas from layered deposits within rock formations, though specific regulatory definitions of fracking may vary by state (93). This process is often paired with horizontal or directional drilling, the combination of which is referred to as unconventional drilling. Conventional drilling refers to the extraction of oil or natural gas from larger pools, typically using a vertical drilling technique (i.e., straight down into the ground, more or less). When extracted, oil is typically measured in barrels (bbl) of 42 gallons, and gas is measured in either cubic feet or barrel of oil equivalents (BOE) (94). There are many additional nuances to the engineering techniques related to the production process that are beyond the scope of this review; however, these terms cover what most epidemiologic researchers will typically encounter.

Description of Exposure Datasets

Our review identified five U.S. national-level oil and gas development datasets: Enverus and FracFocus from the published epidemiologic literature (95,96) and Energy Information Administration (EIA), Homeland Infrastructure Foundation-Level Data (HIFLD), and Oil and Gas Infrastructure Mapping (OGIM) from expert input among our team (97–99). We also observed that many studies leveraged within-state data sources such as the Carnegie Museum of Natural History Well Database (Pennsylvania) (100), Railroad Commission of Texas (101), or the California Geologic Energy Management Division (CalGEM) (102). For the national-scale data sources, we reviewed the parameters, strengths, and limitations for each data source as outlined in our guiding questions. In Table 1, we documented the answers to our a-priori questions for each identified source.

In the next sections, we provide a narrative review of our findings, including notes on subtle data issues that are specific to accurate exposure assessment for oil and gas development.

Energy Information Administration (EIA)

The Energy Information Administration (EIA) is a federal agency that is independent of the Department of Energy. This agency is the primary federal authority on energy statistics and analyses, and they publish and disseminate a multitude of data products. Specifically, we describe their “Oil Wells” and “Natural Gas Wells” layers of the U.S. Energy Atlas, a library of data and interactive maps documenting U.S. energy infrastructure.

We find that the EIA data source for oil and gas development is limited in scope (Table 1). It contains records for 30 states, classifies wells as oil or gas, and does not contain potentially valuable linking variables, such as consistent identifiers aside from U.S.GS basin, which could be used to link site presence to geological features of the surrounding area (97). These data may be valuable for a screening-level ecological study where the temporality is not as substantial as a concern. Additionally, this data source is a piece of the larger EIA Energy Atlas, which can be leveraged to generate a more comprehensive understanding of individual or population energy industry exposure since it includes many additional infrastructure types.

Homeland Infrastructure Foundation-Level Data (HIFLD)

The Homeland Infrastructure Foundation-Level Data (HIFLD) subcommittee is a division of the U.S. Department of Homeland Security that provides geospatial data publicly to support research and disaster preparedness. The open data platform contains an “Oil and Natural Gas Wells” shapefile with information on oil and gas wells drilled across the U.S. and some provinces of Canada, specifically, within 100 miles of the U.S. border. These data are derived by the Oak Ridge National Laboratory from individual state and province public data (98).

We found that HIFLD covers 34 states, encompassing all states with historical or current drilling (Table 1). The dataset included the completion date, and variables describing production type, status, and drilling type. Categorizations, such as “producing,” “active,” “storage,” could be used by researchers to estimate exposure in lieu of information on production volumes, which HIFLD does not include. HIFLD also reports North American Industry Classification System (NAICS) codes, which are used by federal agencies to classify business establishments by type. The inclusion of NAICS codes could be valuable to a researcher looking to link oil and gas development sites to information about the operating business. NAICS groups business establishments into industries and is a tool used to link economic statistics to business records across the U.S., Canada, and Mexico (103). HIFLD provides open access to their oil and gas development data with a free account.

Due to the nature of energy data’s importance for national security, the policies around data access are evolving at HIFLD. In particular, our coauthor team has noted that some oil and gas development data have more restricted access at certain times.

FracFocus

FracFocus is a nongovernmental organization that collects voluntary reports of chemical use in hydraulic fracturing (one type of oil and gas extraction) (104). As of 2022, the database received reports from over 1,600 companies across 26 states—this source lists a planned update including reporting water use by source and chemicals in hydraulic fracturing sites. Additionally, FracFocus has been leveraged by researchers at FracTracker, another nongovernmental organization, to create public dashboards and maps to analyze and communicate the risks of oil and gas development (105).

We found that FracFocus reported temporal information, including fracturing job start and end date, but does not specify production type between oil, gas, or both (Table 1). This data source’s limit to exclusively hydraulic fracturing sites means that drilling type is implied by inclusion in the data. FracFocus presents the names, chemical abstracts service (CAS) registry numbers, and proportions of “ingredients” to hydraulic fracturing fluid when disclosed. Additionally, FracFocus data has been used by a Wetherbee et al to create “WellExplorer,” an open database that connects chemical use reported in FracFocus to human hormonal pathways and toxicity (106). With input from other fields such as hydrology, these data have the potential to be incorporated into epidemiologic studies to better understand the influence of frac chemicals on human health.

Oil and Gas Infrastructure Mapping (OGIM)

The Oil and Gas Infrastructure Mapping (OGIM) database is a global source of information on oil and gas infrastructure originally designed for examination of methane leaks and vents. OGIM was created by Environmental Defense Fund and its subsidiary, MethaneSAT, a non-governmental organization. The data contains a layer, “Wells,” which documents the location of oil and gas drilling sites. The OGIM database is generated from the acquisition and processing of over 450 public-domain location-specific data sources for oil and gas infrastructure globally (99).

While no epidemiologic studies have utilized the OGIM database to date, the database characterizes all U.S. states with spatial data and contains the spud date and completion date when provided by the state in the underlying dataset (Table 1). Additionally, the database includes descriptors of drilling and production types, facility and operator names, and production days. These data appear to be high quality. A unique contribution of the OGIM database is its inclusion of global information beyond the U.S. However, limitations of the data by country are subject to the scope of the original administrative data collected: for this reason, some areas have a higher missing-data rate than others. We did not evaluate the non-U.S. data, as those data are outside the scope of our review.

Enverus (formerly DrillingInfo)

Enverus is a private and commercial provider of data on the oil and gas sector, and this company has often provided academic access to researchers for epidemiology. These data include many of the important for exposure assessment (e.g., spud date, production volume). However, completeness is an intermittent concern depending on which state or region is of interest. Another potential concern when using the Enverus data is that some states with known current or historical drilling are not necessarily included in all data products, likely due to state-specific data limitations.

We found that Enverus covers 32 states, utilizes API identifying numbers, and includes temporal variables (Table 1). Enverus reports production in three main ways: cumulative measurements (e.g., production over the lifetime of a well), temporally linked measurements (e.g., production in the first 6 months of extraction activity), and monthly volumes reported by API number. Cumulative measurements can provide imperfect insights into the productivity of a given well, while the monthly volumes are a complete time-series of production, which would allow detailed examination of exposure timing. Drilling and production type are documented as separate variables.

Note that other private data sources could be used for similar purposes (e.g., Rystad, WellDatabase) (107,108). However, Enverus remains the only one used in epidemiologic analyses from what we have found to date. To our knowledge, the data products provided by Rystad and WellDatabase are nearly identical to what is available in Enverus. We focus on Enverus as it is used widely in epidemiology, and this company functions as an intermediary between the EIA and state agency data sources (109). We also note that the limited accessibility of private data sources yields concerns related to reproducibility and open science efforts (110), as data access is dependent on providers’ willingness to grant academic licensure.

DISCUSSION

Among the five datasets, we found substantial variation in content and data quality. We hypothesize that this result is due to differences in motivation for assembling the dataset and the mission of each agency or organization involved in data curation. Many of these differences have important implications for population health researchers interested in investigating exposures to oil and gas development, posing tradeoffs among domains, such as the degree of detail per well (an issue in many government and non-governmental organization data sources) or the extent with which the data set is openly accessible (an issue with private companies). Depending on the question and context, epidemiologic researchers may want to prioritize certain data features over others.

When examining data sources from different sectors, we found that Enverus has more complete information regarding production and temporal information but is less accessible given the paywall. We expect that this trend would extend to other private data sources as well. Nongovernmental organization sources often had a more limited scope but also tended to provide more specialized oil and gas development exposure data to examine a particular piece of oil and gas development exposure (e.g., chemical exposure, methane exposure). Finally, we found that government agency sources tended to present more granular information about production and timing than non-governmental sources and less information than private sources, but also often contained helpful linking identifiers that could be valuable to the researcher (Table 1). Given the lack of consistent data from federal sources, researchers have often relied on private data sources, including many members of our team. If public data sources were to contain more of the fields that we described as necessary for epidemiologic work (e.g., temporal fields to determine timing of exposure, production fields to estimate intensity of exposure), we anticipate that the field would efficiently pivot away from private data.

When granular production information is needed to answer the research question, private data sources such as Enverus are best suited. We found that Enverus does contain some near-duplicate entries (i.e., multiple rows using the same API number but with different values for other variables); these entries are likely due to multiple permits for one well, and we recommend that researchers are careful not to overcount exposure to oil and gas development when using this source. Furthermore, we note that Enverus represents a trend towards the privatization of public data sources, as many states use Enverus as the conduit for their oil and gas reports to EIA (111). The decision of government agencies to rely on the private sector to manage and provide what is arguably public data may stem from concerns including costs, convenience, the perception of improved efficiency of the private sector in delivering a service, and the availability of needed expertise in the private sector (111,112). However, there is a danger that escalating reliance on the private sector, without proper management, could result in formerly public goods being privatized as a resource to be profited from. This may potentially jeopardize other agency values such as openness, transparency, and accountability to the public (111,112).

Despite increasing availability of national scale data, characterizing a single site can still pose a challenge. All sources contain spatial coordinates to estimate proximity and intensity, meaning that all sources can at least be linked to each other or other data sources via spatial merge. For example, some sites will extract both oil and gas, but some data sources report dichotomized production types without a third option for “both” or “combined.” This challenge extends to other oil and gas development site characteristics and is in part due to varying regulatory definitions by state (93). When national sources combine state sources, they need to harmonize and align differences, which can lead to lost or unclear information in the final data. Information on hydrocarbons produced can be useful in characterizing the specific exposures that nearby communities experience (113). Additionally, there are other well types that can exist in oil and gas production areas, such as wastewater disposal wells, injection wells, storage wells, or non-producing abandoned wells. These wells can be co-located with active oil and gas production sites but can also exist in former oil and gas production sites, depleted aquifers, and other areas, and potentially contribute to community exposures. Only FracFocus contained rich information regarding chemical disclosures specific to hydraulic fracturing. Across all well types, we did not formally evaluate information on production volumes across time, largely due to slightly differing scope than our question at hand. Furthermore, this information is stored in separate records and data structures and requires much more intensive computing to generate an estimate of production volume per oil and gas development site.

There are also research questions about where individual state resources will better suit the researcher’s needs. Since oil and gas development is largely regulated at the state level in the U.S., each state collects and shares data on oil and gas development sites through their relevant agency (e.g., Pennsylvania Department of Environmental Protection, Railroad Commission of Texas, Colorado Energy and Carbon Management Commission). These state data sources ultimately serve as the underlying data for all the data sources presented in this review, but sometimes going back to the source is important to the research question. For example, if completing a research project on hydraulic fracturing in Pennsylvania, the use of the Carnegie dataset (a compilation of eight data sources from the including location, reports on production, waste, and compliance, and spud date) makes sense since the source is highly validated, and may contain additional fields which national data sources might omit (100). Such fields might be omitted if only a few states report values, which is likely given the high variation of data collection styles state-by-state, therefore avoiding the creation of a column in a national dataset with high missingness.

With all data sources analyzed, measurement error has potential to impact epidemiological analysis, largely due to the differential data quality by state. In terms of spatial coordinates, we are confident that there is not a substantial degree of measurement error across data source. However, when it comes to temporal variables measuring when a site is active or not, the variability of data quality on a state level becomes an upstream concern for all data sources analyzed, which all compile and harmonize state level data.

Key Considerations for Future Researchers

Researchers evaluating the utility of different oil and gas data sets for epidemiological research should consider factors including:

1. At what spatial scale do outcome data exist for the research question?

2. How important is precise time-bound exposure to the research question?

3. Is production volume information important to the exposure assessment protocol necessary to address the research question?

4. What types of oil and gas development are included in this exposure assessment (e.g., hydraulic fracturing, only oil, only gas, all types)?

5. Is information regarding specific chemicals injected, used, or stored during hydraulic fracturing or other well operations, or other specific emission types important to the analysis and interpretation of findings?

6. Are the type of oil and gas production wells or the oil and gas development production activity contributing the most relevant exposures?

7. What additional information, if any, will be linked to the exposure data? How will that linkage occur (e.g., merge based on shared variable, spatial match)?

We observed similarities between all data sources. This result is expected given their related overarching goals to collect spatially explicit data related to oil and gas development. It is possible that many of these data sources leverage similar, or even the same, underlying data source to create their products, which would contribute to this similarity. However, tracing the underlying data sources has been challenging given limited disclosures and documentation.

Priorities for Future Data Development

We propose the elements for a more comprehensive dataset of national oil and gas development exposures. These data should include detailed and complete information in six critical categories: location, production, engineering, all chemicals used, all emission types, and scheduling and operations. Location includes the coordinates of each well, including all well types on site. It could be enriched with information about how large the oil and gas development site is, presence of wastewater pits, presence of flaring, and spatial boundaries of areas designated for oil and gas production. For production, the gold standard would include production volumes and production types over time at regular intervals, such as monthly. Engineering details would include information on drill type, secondary boreholes, well reworks, fluid use disclosures, and methods used for plugging postproduction wells. Scheduling and operations details would include time-linkages for all production and engineering details, including start date, end date, spud (first drill) date, completion date, plug date, abandonment date, refrac/restimulation dates, changes to well configuration and purpose, and time-bound production and fluid use in addition to accurate point location information and site ownership. This dataset would continue to follow wells taken out of service as oil and gas production wells are repurposed as underground natural gas storage injection or withdrawal wells, wastewater disposal wells, or for other purposes, including storage of carbon dioxide. This dataset would also follow wells used for novel purposes, such as hydrogen production or storage, or for storage of carbon dioxide. Additionally, a gold standard dataset would be continuously updated and accessible via downloadable formats and direct access through an application programming interface (API) that is freely available to the public.

CONCLUSIONS

Our evaluation of oil and gas development datasets available provides information that can help researchers decide which national scale datasets to use for assigning exposures for public health studies. The information on oil and gas development from the five sources evaluated in this paper can help to translate data and population health findings to policy action. There is precedent for the epidemiologic literature being a key component of oil and gas regulations, such as the initial moratorium on high-volume hydraulic fracturing in New York and the case of setback distances in California (114,115). We find that each source has trade-offs between granular temporal information, production information, and chemical disclosure information, lending each source to different research questions.

ABBREVIATIONS:

U.S.

United States

UWI

Universal well identifier

BBL

Barrel of crude oil

BOE

Barrels of oil equivalent

EIA

Energy information administration

HIFLD

Homeland infrastructure foundation level data

OGIM

Oil and gas infrastructure mapping

CalGEM

California Geologic Energy Management Division

CAS

Chemical abstracts service

NAICS

North American Industrial Classification Standards

SB

Senate bill

SALSA

Search, Appraisal, Synthesis, Analysis