Steam cracker facilities in the United States: operations, emissions, and sociodemographic patterns of surrounding populations

Background: Production of shale gas in the United States (US) increased more than 10-fold from 2008 to 2021, yielding greater quantities of hydrocarbon feedstocks and incentivizing expansion of petrochemical facilities. Steam crackers (SCs) convert hydrocarbon feedstocks into ethylene and propylene (the building blocks of plastics), while releasing toxic chemicals and greenhouse gases (GHGs). Analyses of environmental health and justice impacts of SCs are limited. Methods: We described SC operations, locations, and emissions, and evaluated sociodemographic characteristics of populations residing near SCs to better understand potential public health hazards and inform future studies. We summarized and described industry-reported emissions from the US Environmental Protection Agency’s Toxic Release Inventory and GHG Reporting Program. We compared population characteristics of US Census block groups ⩽5 km and >5 km from a steam cracker-containing facility (SCF) within the same county. Results: We identified 32 SCFs across five US states, with most in Texas and Louisiana. Toxic chemicals with the greatest reported cumulative air emissions in 1987–2019 were: ethylene, propylene, hydrochloric acid, benzene, n-hexane, 1,3-butadiene, ammonia, toluene, vinyl acetate, and methanol. Reported total annual GHG emissions were 4% higher in 2019 versus 2010, with total GHG emissions of >650 million metric tons (carbon dioxide equivalents) in 2010–2019. We found that 752 465 people live in census block groups ⩽5 km from an SCF, regardless of county. Compared to block groups >5 km away within the same county, block groups closer to SCFs had statistically significantly lower median incomes ($54 843 vs $67 866) and more vacant housing (15% vs 11%), and higher proportions of residents who were non-Hispanic Black (31% vs 19%) and unemployed (8% vs 6%). Conclusion: SCs emit substantial amounts of GHGs and toxic chemicals in locations with historically disadvantaged populations. Future research could further evaluate the accuracy of reported emissions, conduct monitoring in proximate communities, and assess population-level health impacts.


Introduction
Shale gas production, the extraction of natural gas from low-permeable geologic formations using horizontal drilling and high-volume hydraulic fracturing [1,2], increased over 10-fold in the United States (US) between 2008 and 2021 [3]. The abundance of hydrocarbons supplied by this upstream development has incentivized buildout of downstream industrial petrochemical infrastructure. Numerous studies documented Consistent with US Environmental Protection Agency (EPA) terminology, we consider a facility to be 'all buildings, equipment, structures, and other stationary items which are located on a single site or on continuous or adjacent sites and which are owned or operated by the same person [or entity], [meaning] a facility may contain more than one establishment' [45]. Other terms used interchangeably with petrochemical facility in the literature include petrochemical complex or petrochemical industrial park. We consider an individual SC an 'establishment' or unit at a single physical location where industrial operations are performed. An SC may be collocated with other types of industrial units or establishments (e.g. refineries) within a given facility. Therefore, we use the term SC for a specific cracker unit, and SC facility (SCF) for a facility containing an SC. In general, we found data were available at the SCF level, not the SC level.

Identification of SCs
Because there is no centralized, standardized list of all SCs, we compiled a list of possible US-based SCs from: (a) US Energy Information Administration (EIA), (b) Oil & Gas Journal (a petroleum industry journal), (c) non-governmental organization Beyond Plastics, and (d) non-governmental organization Environmental Integrity Project [46][47][48][49][50]. This list also included data such as location, ownership, and emission reporting identifiers (IDs). We visually inspected maps and aerial photos of each facility to confirm presence of SCs [51].
For all SCFs with a confirmed or possible SC, we obtained IDs from the EPA Federal Registry Service, Toxic Release Inventory (TRI), and GHG Reporting Program (GHGRP). We performed several checks to maximize accuracy of IDs assigned to each SCF and standardize IDs across the databases. We conducted EPA Federal Registry Service online queries to find all EPA-reporting facilities with the same or similar company name, site name, and location (city, county, and state) as each SCF on our list. We inspected system facility names by reporting program (TRI and GHGRP), associated facility names and organizations, Standard Industrial Classification Codes, and North American Industry Classification System (NAICS) codes to obtain the most likely IDs pertaining to each SCF on our list [52]. Finally, in cases where the years of operational status of the SCF was unclear, we performed web-based searches of SCF permit applications, company websites, industry publications, and news articles (e.g. about a plant closure/opening) to resolve discrepancies. We performed an additional review of TRI emissions records to check for the presence of ethylene and propylene, which are commonly emitted from these units.

Toxic chemical emissions
EPA's TRI program requires facilities operating in certain industries to annually self-report emissions of toxic chemicals [53], defined as those that cause: (a) cancer or other chronic human health effects, (b) significant adverse acute human health effects, or (c) significant adverse environmental effects [54]. Facilities that manufacture, process, or use TRI-covered chemicals above 25 000 pounds for manufacturing or processing and 10 000 pounds otherwise are required to annually self-report emissions to EPA TRI [55]. EPA TRI maintains these emission estimates in a publicly available database. Because emissions are self-reported and not based on independent monitoring, they are difficult to verify and may be underreported [53]. The list of chemicals covered by the TRI is updated annually, and at the time of data acquisition included reporting years 1987-2019. We assembled a dataset of TRI emissions to air, land, and water for all SCFs.

GHG emissions
EPA's GHGRP requires that certain categories of facilities with GHG emissions above 25 000 carbon dioxide equivalents (CO2e) per year self-report annual GHG emissions and that EPA maintain this information in a publicly available database [56]. The GHGRP requires reporting of carbon dioxide, methane, and nitrous oxide. Facilities have different options in calculating the reported emissions, including monitoring systems, fuel composition data, and default emission factors. We assembled a dataset of GHGRP emissions for all SCFs during reporting years 2010-2019.

Socioeconomic and demographic data for neighboring communities
We obtained sociodemographic variables from the 2019 US Census American Community Survey (5-year estimates ending in 2019) [57]: population educational attainment (percent ⩾ 25 years with ⩽ high school education), race (percent identifying as White, Black/African-American, Asian, other/multiple races), ethnicity (Hispanic/Latino), percent unemployed, percent households receiving public assistance income in past 12 months, median household income in past 12 months, percent population in renter-occupied housing, percent vacant housing units, crowding (percent housing units with >1 person/room), and percent population in professional occupations. Population density was calculated by dividing the block group population estimate by the block group area.

Descriptive and statistical analysis
For Objective 1 (describe steps of SC operations), we developed a figure to illustrate the multi-step processes of steam cracking, accompanied by a narrative description of each step, emphasizing potential emission sources.
For Objective 2 (construct database of US SC locations and emissions), we mapped locations of all SCFs in our database. We graphed annual median and total emissions of TRI toxic chemicals emitted to air in the greatest quantities from 1987 to 2019 and annual median and total SCF GHG annual emissions from 2010 to 2019.
For Objective 3 (describe sociodemographic characteristics of populations living near SCs), we classified 2019 census block groups in counties with an SCF as being 'near' an SCF if a 5 km buffer around the SCF coordinates intersected with any portion of the block group or 'far' from the SCF otherwise. We selected 5 km based on published health studies conducted in other nations that observed increased air concentrations or health risks in populations living 2-10 km from petrochemical complexes [7,25]. We compared the mean and standard deviation of sociodemographic variables between near and far census block groups and tested the statistical significance of differences using the Wilcoxon rank sum test. We conducted a sensitivity analysis in which we redefined 'near' as block groups that intersected with a 5 km radius around the coordinates of the SCF regardless of county and 'far' as block groups >5 km and ≤20 km from an SCF. Finally, we calculated the total population living in a block group intersecting with a 5 km radius around an SCF, regardless of county designation.
All statistical analyses were conducted in SAS (version 9.4; SAS Institute Inc.), and all tests were two-sided with an alpha level of 0.05.

Operational overview
The steam cracking system is a complex and energy-intensive process with the overall purpose of converting hydrocarbons (paraffins) to light olefins (ethylene and propylene), the chemical intermediates necessary for plastic production. Historically, most such establishments have been feedstock flexible, but the majority of new investments have been in expanding ethane-specific feedstock capacities [58,59]. The establishments' design varies based on numerous factors including age built and feedstock composition; however, the general process can be divided into four main steps (figure 1): (1) pyrolysis (cracking), (2) quenching, (3) compression, cooling, and drying, and (4) fractionation (distillation) [60]. Cleaning and maintenance, though not directly part of the chemical process, is also an important process in terms of overall operational emissions.
Pyrolysis begins by feeding petroleum hydrocarbons (feedstock)-primarily ethane, propane, butane, and naphtha-into furnaces, also referred to as heaters (figure 1, Step 1). SCs can often accommodate a mixture of feedstocks, and the composition is usually determined by resource availability and market conditions [47,48,61]. During this energy-intensive step, feedstocks are heated to high temperatures in the presence of steam to alter (i.e. 'crack') their molecular structure. During quenching, the cracked gas leaves the furnaces at high temperatures where it is cooled in an oil-or water-based quenching column (figure 1, Step 2). This simultaneously preserves composition of the gases while reducing presence of undesirable side reactions. Next, the cracked gas undergoes several rounds of compression via a turbine-driven centrifugal compressor coupled with additional cooling and drying to remove water introduced prior to and during compression (figure 1, Step 3). During fractionation the compressed cracked gas is fed into distillation columns that separate the gas into different components, also referred to as fractions (figure 1, Step 4). Various distillation columns are used including a demethanizer to separate methane and hydrogen, a deethanizer to separate ethane and ethylene, and a C3 splitter to separate propylene [60]. Finally, flaring is a high-temperature oxidation process that combusts excess hydrocarbon emissions to relieve pressure (figure 1, Step 5).
There are several emission sources throughout an SC's multistep process. Cracking furnaces and supporting infrastructure used in pyrolysis are potentially one of the largest sources of toxic chemical and GHG emissions (figure 1, Step 1). During normal operating procedures cracking furnaces are permitted to emit a range of pollutants and toxic chemicals including particulate matter with aerodynamic diameter ⩽2.5 µms (PM 2.5 ) and ⩽10 µms (PM 10 ), nitrogen oxides, carbon monoxide, sulfur dioxide, and volatile organic compounds, and GHGs including methane, carbon dioxide, and nitrous oxide. Cracking furnaces also emit similar pollutants, toxic chemicals, and GHGs during the decoking process, during which built up residue on the furnace coils is combusted and released to the atmosphere. Facility-wide fugitive emissions that often result from system inefficiencies include ammonia, 1,3-butadiene, and benzene. Other sources of emissions include flares that combust excess hydrocarbon emissions to relieve pressure, as well as startup and shutdown procedures. Emissions can also be attributed to supporting infrastructure including backup generators and firewater pump engines [62].

Identification of SCs
Our initial searches identified 44 facilities potentially containing an SC. After record-by-record review, we excluded two facilities that were not operational during our study period (1987-2019) and ten whose location and operational status could not be confirmed, yielding a final dataset of 32 US facilities verified to contain an operational SC during our study period (table 1).
Most identified SCFs were in southeastern Texas (n = 19) and southern Louisiana (n = 10), with the remaining three located in Iowa, Illinois, and Kentucky (figure 2). Four companies own half of all identified SCFs: LyondellBasell (n = 6), Dow (n = 4), Chevron Phillips Chemical (n = 3), and ExxonMobil (n = 3) (table 1). The facilities are classified under various NAICS codes, most commonly an NAICS code of 325 110 corresponding to petrochemical manufacturing (n = 17) or 325 199 corresponding to all other basic organic chemical manufacturing (n = 11).

SCF emissions 3.2.1. Toxic emissions
Most reported emissions of toxic chemicals were to air, with lesser contributions to water or land. While many TRI toxic chemicals were emitted to the air in median amounts exceeding 1000 pounds yr −1 , three chemicals had non-zero median emissions to water (sodium sulfate, nitrate compounds, ammonia), and median releases to land were zero for all TRI toxic chemicals. Across all SCFs and reporting years, the ten toxic chemicals with the largest mass emitted to air were: ethylene, propylene, hydrochloric acid, benzene, n-hexane, 1,3-butadiene, ammonia, toluene, vinyl acetate, and methanol (figure 3). Ethylene and propylene consistently remained the most dominant toxic chemicals emitted to air across all reporting years. These toxic chemicals act as asphyxiants at high exposure levels, an important health and safety issue in enclosed occupational settings, but are not generally toxic at environmental concentrations. However, the top ten list also includes known human carcinogens (e.g. benzene, 1,3-butadiene), chemicals impacting the nervous system (e.g. n-hexane, vinyl acetate), and chemicals affecting the respiratory system (e.g. toluene, hydrochloric acid) (supplemental table S1). Annual median SCF emissions of the top ten toxic chemicals declined from 1987 through 2007 and then remained relatively stable though 2019. Total annual emissions followed a similar pattern (supplemental figure S1).

GHG emissions
Annual median reported GHG emissions of SCFs ranged from 1.3 to 1.6 million metric tons of CO2e across the years 2010-2019; median emissions were 25% higher in 2019 versus 2010 (figure 3). Over 99% of emissions were attributable to CO 2 , with <1% due to reported releases of methane and nitrous oxide. In 2019, self-reported GHG emissions from all US SCFs totaled 68 million metric tons CO 2 e, equivalent to the annual GHG gas emissions of >14.8 million US passenger vehicles [63]. Total emissions in 2019 were 4% higher than 2010. Total SCF GHG emissions summed over 2010-2019 were >650 million metric tons CO 2 e.

Sociodemographic characteristics of neighboring communities
SCFs were present in 16 counties across five states. Within these 16 counties, 484 census block groups were within a 5 km radius of an SCF, and 3110 were beyond the 5 km radius (table 2). A total of 752 465 people lived in block groups within 5 km of an SCF, regardless of county designation.
Block groups closer to SCFs differed from those further from SCFs with respect to several sociodemographic factors indicating greater exposure potential in communities experiencing other burdens (table 2). Compared to block groups >5 km of an SCF, block groups ⩽5 km had a statistically significantly greater proportion of non-Hispanic Black residents (31% versus 19%), higher proportions of residents who were unemployed (8% versus 6%), lower median income ($54 843 versus $67 866), greater proportion of vacant housing (15% versus 11%), lower percentage of individuals in professional occupations (25% versus 33%), and greater proportion of individuals with maximum educational attainment of high school education or less (53% versus 45%). Block groups closer to SCFs had a statistically significantly smaller proportion of people identifying as Hispanic, lower population density, and lower percentage of renter-occupied housing. Block groups did not differ with respect to percent population identifying as non-Hispanic White, percentage receiving public assistance, and percentage of crowded housing arrangements.
We observed similar results using a different 'near' vs. 'far' categorization (supplemental table S2).

Discussion
Despite current and planned expansions of SCs in the US, there is limited public health research concerning their placement, operations, emissions, and potential public health hazards. We present new analyses synthesizing the process, locations, toxic chemical emissions, GHG emissions, and distributive environmental justice issues in relation to 32 SCFs in the US. Previous epidemiologic studies of health of populations residing near petrochemical facilities focused on hematological and respiratory outcomes, particularly cancers. The toxic chemicals emitted in the greatest quantities from SCFs exhibit a range of toxic and carcinogenic properties consistent with these observed endpoints (supplemental table S1). The toxic chemicals most emitted from SCFs also include those with nervous system, immune, urinary, and developmental effects, which are less well-studied in relation to residential proximity to petrochemical sites [25].
Few community health studies of petrochemical facilities in the US have been conducted, particularly those using more recent data from the 2000s onward. Our finding that more than 750 000 individuals live in block groups ⩽5 km of SCFs raises concerns about potential exposures and health impacts from emissions from SCFs and suggests that further research is needed. Changes in feedstock composition and technological changes could mean differences in the composition or quantity of emissions in more recent years. Epidemiologic and exposure studies in the US, where regulations, meteorology, demographics, and potential confounders may differ from those of other countries, could contribute new information to facilitate cross-cohort comparisons, triangulate evidence, and enhance generalizability [64].
Although increased health risks are documented in populations living near petrochemical complexes in various countries, studies specific to SCs are few. While our goal was to focus on SCs specifically, we were limited in our ability to isolate emissions data from SC units because emissions data were for facilities in which SC units were located, not for individual SC units (table 3). This highlights the need for more granular data on SC emissions.
Current and planned expansions of SCFs for plastic production has significant implications for climate change. The reported GHG emissions from SCFs were slightly greater in 2019 compared to 2010, highlighting the contribution to climate change [65]. Furthermore, climate change is driving more frequent SC startups and shutdowns (e.g. freezing temperature or hurricanes in Texas). SCs emit significant toxic chemical and GHG emissions during these processes, which could negatively affect GHG emission targets [32] and have negative health consequences for surrounding communities [18].
Observed associations between sociodemographic vulnerabilities and locations of SCFs in the current work is consistent with decades of distributive environmental justice research that demonstrates inequitable There are gaps in the documentation of planned SCs.
Without a consistent record of SC investments and planned projects, research is likely to underestimate overall SC emissions and potentially exposed populations.
A publicly available database of planned and under construction SCs could be made available and maintained with current information by local, state, and federal government bodies.

Specificity of NAICS categorization
SCs are collocated in large industrial complexes with many other facilities. Therefore, the classification of the complex as a whole may not align with how SCs themselves would be classified.
The EPA could mandate more granular facility emissions reporting (i.e. sub facility or unit level).

Isolating Emissions
Geographic specificity SCs are often collocated in large complexes among other industrial plants The EPA could mandate location and emissions data reporting at a more granular level.
Companies self-report calculated yearly emissions Self-reported yearly emissions may not reflect real-world conditions.
The EPA could mandate emissions reporting at a more temporally granular level.
Year placement of hazardous industries predominantly in low-income communities, communities of color, and other communities that experience structural discrimination [35,39,[66][67][68]. Several SCFs we identified are located in the Louisiana industrial corridor, commonly referred to as 'Cancer Alley' [69,70], where a recent EPA investigation found longstanding racial discrimination in the allowable emissions of toxic air pollutants and consequent adverse impacts to the health of Black residents [71]. The distributive justice issues observed for SCFs are consistent with those observed for other phases of the oil and gas production lifecycle across multiple geographic contexts [68,72]. A recent study reported an association between redlining (racist lending and housing policies) and locations of fossil fuel-based electricity generation plants [73]. Another documented correlation between social vulnerability and density of pipelines [38], and county-level analysis observed an inverse correlation between per-capita income and oil refinery emissions [39]. As new SCs are proposed, developers and policymakers should consider environmental justice issues and cumulative burdens as a critical aspect of the permitting process [74]. Our analysis identified important needs in terms of data availability (table 3). The lack of a centralized database for SCFs necessitated consultation of multiple incomplete sources, and our analyses were limited to SCFs that were confirmed by visual inspection and emission records. We found that the gray literature, including reports from non-governmental organizations, which may employ staff on the ground to verify information, provided important data to supplement government databases. Limitations of relying on disparate, non-academic, and non-governmental sources include possibilities that facilities are missed, inconsistencies in reporting, differences in the type and quality of information, and lack of validation.
Our analyses were based on EPA programs relying on industry-reported emissions, which are often based on estimates and algorithms not empirical data, and tend to underestimate measured emissions [53,75]. These reporting programs are restricted to annual emissions, although short-term, peak toxic chemical emissions may be relevant for health. Further, future work is needed to understand other sources, hazards, and pathways through which such facilities could affect health, such as via water contamination, noise, and psychosocial stress. Another limitation is that our focus on emissions and not ambient concentrations does not consider fate, transport, and transformation of chemicals, or individual-level activity patterns relevant for exposure. While quantity of emissions is useful for prioritizing chemicals, it does not necessarily correspond to toxicity, as many chemicals are toxic at low levels. Future work could compare the toxicity of SC emissions to those of other petroleum facilities, such as refineries; however, because the emission profiles differ substantially, a risk assessment type approach would be needed to facilitate an appropriate head-to-head comparison. There are some limitations associated with using socioeconomic data from the 2019 ACS, as ACS estimates can have large margins of error [76]. Future research could evaluate ambient concentrations of highly emitted chemicals in areas surrounding SCs or conduct individual-level monitoring, and evaluate the subsequent health effects, including which subpopulations are most vulnerable. In addition, future studies could evaluate sociodemographic characteristics of communities targeted for future SCFs.

Conclusions
We leveraged publicly available data to synthesize information about the locations, toxic chemical air emissions, GHG emissions, and communities for 32 SCFs in the US. Industry-reported emissions from SCFs demonstrate chemical releases of known human carcinogens and reproductive toxicants. Annual SCF GHG emissions in 2019 totaled 68 million metric tons CO 2 e, equivalent to the annual GHG emissions of over 14.8 million US passenger vehicles. We found that 752 465 people live in census block groups ⩽5 km from an SCF, and that these communities have increased socioeconomic disadvantages, compared to those residing farther away. These findings underscore the need to consider public health and environmental justice impacts in proposed development of new SCs and for more US-based exposure and health studies of SCs.

Data availability statement
No new data were created or analysed in this study.

Funding
This research was funded by a grant from the High Tide Foundation. NP was funded by the Yale University Summer Experience Award.