If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Petroleum refineries are vital to the production of the fuels that people rely on for transportation, heating, and other activities. Refineries continually seek to improve efficiency of operations and to reduce environmental impacts. The American Petroleum Institute (API) has stated :
“The members of the American Petroleum Institute are dedicated to continuous efforts to improve the compatibility of our operations with the environment while economically developing energy resources and supplying high quality products and services to consumers. We recognize our responsibility to work with the public, the government, and others to develop and to use natural resources in an environmentally sound manner while protecting the health and safety of our employees and the public.”
Concerns about emissions from numerous stationary sources, including refineries, power plants and various industries, led to passage of the Clean Air Act in 1970 and to its subsequent amendments in 1990, adding more stringent requirements and the development and installation of site-specific air pollution control technologies. As a result, emissions have been heavily regulated for more than 20 years, and the Clean Air Act requirement for USEPA to review and update new source performance standards (NSPS) and national emissions standards for hazardous air pollutants (NESHAP) regulations every 5–8 yr will maintain and potentially augment air emissions controls in the future.
Petroleum refineries account for <2% of the criteria air pollutant emissions and 3% of the air toxics emissions from all industrial sources [2-4]. Nearly 25% percent of all US refineries were closed between 1990 and 2010, but the industry's crude distillation capacity increased through the expansion and modernization of existing refineries. At the same time, crude oil feedstocks trended toward higher density, crude sulfur content fluctuated and product specifications became more stringent.
Now that air emissions regulations have been in place for more than two decades, it is appropriate to consider whether they have effectively controlled refinery emissions despite changes in the numbers of refineries and in crude density or sulfur content. Examining the relationship between changes in crude properties and petroleum refinery air emissions will inform community discussions as crude properties continue to change. This article uses publicly available data on criteria pollutant and hazardous air pollutant emissions, crude density, sulfur content, and refining capacity.
USEPA data on criteria air pollutant (CAP) emissions, including sulfur dioxide (SO2), nitrogen oxides (NOx), volatile organic compounds (VOC), and particulate matter (PM), was obtained from the national emission inventory (NEI) database  and then trended for the period of investigation. This data was obtained for petroleum refineries designated by standard industrial classification (SIC) 2911 or North American Industrial Classification Systems (NAICS) 32411. This data set represents the most comprehensive record of petroleum refinery emissions over the period of 1990–2010.
The USEPA publishes the NEI database every 3 yr and the following publication years were used in this investigation: 1990, 1996, 1999, 2002, 2005, and 2008 (no data was developed for 1993). Between NEI reporting years, USEPA refines and corrects the emissions data, and may update the database several times. USEPA notes that because the estimates originate from a variety of sources (state and local regulatory agencies, tribes, industry, and USEPA) and because estimation methods are used for differing purposes, the estimates will in turn vary in quality, pollutants included, level of detail, and geographic coverage .
CAP emissions of SO2, NOx, VOC, and PM10 (filterable) are of key interest to petroleum refineries. Emission data for filterable particulate emissions with a diameter of 10 μm or less (PM10) is used as a surrogate for total particulate matter emissions. Although there is now a NAAQS standard for ambient particulate matter with a dynamic diameter <2.5 μm (PM2.5), this study did not examine filterable or condensable PM2.5 emissions because of a lack of reliable historical inventory data. Although carbon monoxide (CO) emissions data is available, it is not considered an area of key interest to petroleum refineries, because all areas around US petroleum refineries have attained concentration standards for the CO NAAQS. Greenhouse gases were not included in this article because they were not regulated pollutants during the period of investigation.
The hazardous air pollutant (HAP) emission data used in this study was downloaded from the USEPA toxic release inventory (TRI)  and also trended for the period of investigation. HAPs are those 187 pollutants that are known or suspected to cause cancer or other serious health effects, such as reproductive effects or birth defects, or adverse environmental effects. Examples of hazardous air pollutants include benzene, dioxin, toluene, and metals such as cadmium, mercury, chromium, and lead compounds . TRI data should be interpreted carefully because there are various methodologies for estimating emissions and a corresponding variation within the release estimates . HAP emissions were gathered for petroleum refineries listed under SIC 2911 or NAICS 32411.
The USEPA qualifies the TRI data by stating:
“Release estimates alone are not sufficient to determine exposure or to calculate potential adverse effects on human health and the environment. TRI data, in conjunction with other information, can be used as a starting point in evaluating exposures that may result from releases and other waste management activities which involve toxic chemicals. The determination of potential risk depends upon many factors, including the toxicity of the chemical, the fate of the chemical, and the amount and duration of human or other exposure to the chemical after it is released .”
Toxicity hazard potential trends for HAP emissions from petroleum refineries as a group were developed using toxicity criteria (i.e., inhalation unit risk (IUR) and inhalation reference concentration (RfC) values) presented in the USEPA regional screening level table . These values represent the most recent set of criteria compiled by USEPA for use in screening pollutant levels in the environment and include criteria developed by USEPA and other regulatory agencies. In this article, separate analyses were conducted for cancer and noncancer toxicity endpoints. The type of toxicity hazard potential analysis presented in this article constitutes a screening level analysis that is appropriate for prioritizing and investigating general trends over time. It does not constitute a detailed health risk assessment.
To evaluate the overall toxicity hazard potential associated with individual HAP compounds, chemical-specific annual TRI emissions were multiplied by the appropriate toxicity criterion and then summed across all chemicals to yield a total toxicity hazard potential per year. To evaluate cancer effects, the IUR is directly multiplied by the emissions estimates. For evaluating non-cancer effects, the inverse of the RfC was multiplied by the emission estimates (since RfC is inversely proportional to health risk). Because the point of comparison in the analysis is from 1990 data, toxicity hazard potentials for subsequent years were normalized to the 1990 value by dividing 1991 to 2010 annual industry totals by the total toxicity hazard potential calculated based on 1990 emissions. Because emissions and toxicity hazard potentials have decreased over time, the resulting normalized values are fractions that trend downwards from 1 to 0.
The TRI inventory data does not provide sufficient detail for compound specific chemicals for groups of compounds reported in the TRI data, such as “nickel compounds” or “polynuclear aromatic (hydrocarbon) compounds.” The IUR and RfC values provided by USEPA are designed to be applied to specific chemicals or specified groups of chemicals. However, because the composition of the grouped compounds in the TRI (such as PAH) are not specifically identified, it is not feasible to accurately describe the toxicity of these grouped HAPs without speciation of the specific chemicals. Basing the analysis on the individual TRI compounds provides a robust analysis of the trends over time because the individual compounds account for nearly 97% or greater of the total HAP compound emissions for any given year, based on the TRI data.
Data for total petroleum refining industry crude distillation capacity, gross crude feed, capacity utilization, crude oil sulfur content, crude oil density and refined products was obtained from the United States energy information administration (USEIA)  and trended from 1990 to 2010. Where feasible, the data were verified using other industry references including the Oil and Gas Journal Annual Worldwide Survey reports  and the National United States Refining and Storage Capacity Report published by American Fuels and Petrochemical Manufacturers (AFPM), formerly known as National Petrochemical and Refining Association (NPRA) . As with the USEPA data, USEIA information gathered in this study is for petroleum refineries under the NAICS 32411 or SIC 2911.
This article examines the relationships between decreasing CAP and HAP emissions, toxicity hazard potential, regulatory drivers, historical changes in crude oil density and sulfur content, and the refinery production of major types of fuel. CAP and HAP emissions from petroleum refineries have decreased markedly between 1990 and 2010. At the same time, industry's crude distillation capacity increased and crude oil feedstocks trended toward higher density and sulfur content.
Criteria Pollutant Emissions Trends from Petroleum Refining
Total US petroleum refining industry CAP emissions and their 20-yr trend are shown in Figure 1. Annual emissions over the period of investigation show overall reductions of SO2, NOx, VOC and particulate matter (filterable PM10). The historical data shows a marked and continuing downward trend with notable reductions in emissions of SO2.
To attempt to assess the reliability of the total industry emission estimates, the NEI NOx emissions were examined in more detail by comparing NEI report NOx emissions to “boundary” estimates based on USEIA refinery fuels consumption data and EPA AP-42 emission factors . As shown in Figure 2, the NEI emission estimates normalized to total crude throughput are within the independently generated USEIA/AP-42 emissions “boundaries” and in addition, the independent estimates compare favorably with the NEI emission data and with the NOx emission control technology conversions implemented during the period of investigation. This analysis suggests that the NEI data for NOx is reliable and valid. See the Discussion section and Supporting Information for details.
Hazardous Air Pollutant Emissions Trends from Petroleum Refining
Petroleum refining industry HAP emissions trends using emissions estimated from the USEPA TRI database are shown in Figure 3. All refineries are required to report toxic emissions annually from air, water and waste through the TRI. In this analysis, if the reporting facility left a cell blank in its Form R submission, it was assumed that there were zero emissions of that TRI chemical for the reporting facility for that year. As with CAP emissions, the HAP emissions show a marked reduction during the period of investigation. In general, emissions of organic HAP compounds were reduced as a result of total VOC emission reductions and emissions of inorganic HAP compounds were reduced with PM reductions. Increases in the TRI releases from the period 1994–1996 were due to an increased number of chemicals subject to reporting requirements.
Petroleum refining toxicity hazard potentials were calculated using the TRI data and USEPA toxicity criteria and are shown in Figure 4. Separate trends are presented for cancer and noncancer effects. The reduction in individual HAP compound emissions has resulted in a reduction in the overall toxicity hazard potential of HAP emissions from the petroleum refining industry over time. There is a dramatic reduction in the noncancer toxicity hazard potential trend from 1992 to 1993. Based on the TRI emissions data, this reduction is nearly all related to a significant reduction in chlorine emissions from one refinery.
Trends in Average US Crude Oil Density (API Gravity) and Sulfur Content
Overall petroleum refinery CAP and HAP emissions have substantially decreased during the period of investigation despite processing crude oil with greater density and increasing sulfur content. The density of crude oil is measured using the API gravity, a metric based on the specific gravity (SG) that was developed by the American Petroleum Institute. The API gravity has an inverse relationship to the specific gravity through the following equation : API Gravity = 141.5/SG–131.5.
As reported by the USEIA and shown in Figure 5, the average API gravity for the crude oil used by the US petroleum refining industry has gradually decreased over time. Annual average API gravity has decreased (i.e., density has increased) from ∼31.9° to 30.7° API over the last 20 yr. Energy and crude supply forecasts do not indicate dramatic changes in API gravity of the future US crude slate .
As shown in Figure 5, annual average sulfur content in the US refining industry has consistently ranged from ∼1.1 to 1.4% over the last 20 yr. The USEIA has reported that the average sulfur content of the US crude oil imports increased from 0.9% in 1985 to 1.4% in 2005. Energy and crude supply forecasts do not indicate dramatic changes in sulfur content of the future US crude slate . It is important to note that density and sulfur content are not necessarily related. Some individual crudes are relatively dense with low sulfur content, while others may have low density and high sulfur content.
Trends in US Petroleum Refinery Crude Distillation Capacity and Capacity Utilization
During the period of investigation, there has been modest net growth in US crude distillation capacity due to expansion of existing refineries. On January 1, 1990, the Oil and Gas Journal identified 188 operating refineries with a total crude distillation capacity of ∼2.5 × 106 m3 per calendar day . On January 1, 2010, there were 148 operable petroleum refineries in the United States (excluding Puerto Rico and the US Virgin Islands)  and the USEIA estimated the total crude refining capacity in the United States at 2.8 × 106 m3 per calendar day, or an increase of 15% from 1990. This net modest growth in US crude distillation capacity at existing refineries has compensated for the loss in capacity from refineries that were permanently shutdown (a gross loss of total crude oil distillation capacity of 0.24 × 106 m3 per calendar day during the period of investigation ). Also during the two decade period, only one entirely new petroleum refinery has been constructed, and one refinery has been almost completely reconstructed. Therefore, the increase in refining capacity is mostly related to the expansion of existing refineries.
Based on USEIA definitions, capacity utilization is a comparison of the total gross crude oil (domestic plus foreign) input to crude oil distillation units measured in m3 per calendar day (or barrels per calendar day) against the amount of available operable crude refining capacity. Figure 6 presents the historical annual US petroleum refining industry distillation capacity in m3 per calendar day, gross inputs in m3 per calendar day and resulting percent capacity utilization. During the entire reported period, the capacity utilization has remained relatively stable although a decline is seen in the 2004 to 2009 period.
Petroleum Refining Trends Summary
Table 1 summarizes trends in US refining processing and CAP and HAP emissions over the period of investigation. Even though US petroleum refining crude distillation capacity has increased 15% and the associated crudes have increased in density and sulfur content, industry SO2 emissions have decreased 79%, total HAP emissions have decreased 64%, and the normalized toxicity hazard potential has decreased 69% for the cancer HAP emissions and 85% for noncancer HAP emissions during this same period. These reductions in emissions and associated toxicity hazard potential are closely related to industry compliance with federal and state air pollution control rules.
Table 1. Summary of US petroleum refining statistics for 1990 versus 2010
US refining industry
Number of operating refineries
106 m3 day−1
106 m3 day−1
Crude API gravity
Crude sulfur content
Criteria air pollutant emissions
Hazardous air pollutants emissions
Normalized total cancer HAP toxicity hazard potential
Normalized total noncancer HAP toxicity hazard potential
The Clean Air Act requirement for USEPA to review and update NSPS and NESHAP regulations will maintain and potentially augment air emissions controls in the future as noted earlier. Therefore, emissions have actually been inversely related to increases in distillation capacity and crude oil density and sulfur content over the last 20 yr. Figure 7 provides a comparison of the criteria pollutants versus annual production rates of the major petroleum refining product types. Based on the comparisons, the US petroleum refining industry has been able to increase production of finished motor gasoline, kerosene-type jet fuel, and distillate fuel oil, while reducing emissions of criteria pollutants. As refineries install equipment to process heavier crudes and/or increase sulfur processing capacity, emissions are likely to continue to decline due to both technology improvements and increased stringency of existing regulations.
Assessment of Regional CAP Emissions and Crude Qualities
To explore any relationships between historical CAP emissions, crude sulfur content, and API gravity, capacity-weighted normalized CAP emissions were calculated using the NEI refinery specific data and plotted for each PADD (petroleum administration defense district). The comparison of the capacity-weighted normalized NOx emissions and API gravity is shown in Figure 8, and the comparison of normalized SO2 emissions and crude sulfur content (as weight percent sulfur) is shown in Figure 9. As can be seen in Figure 8, the overall petroleum refining industry normalized NOx emission rates go down as a group year by year, but the PADD averaged normalized NOx emissions within any given year are not correlated to the average API gravity. Over the period of on this analysis, the range of the average API gravity found in the five PADDs tends to become smaller from 1990 to 2008. Similar results are seen for the SO2 emissions in Figure 9; however, the overall range of PADD average crude sulfur content tends to broaden over the same time period. Based on these PADD average comparisons, no relationship can be established between the crude characteristics (API gravity and crude sulfur content) and the normalized criteria air pollutant emissions. This would suggest that refinery air pollutant emissions for NOx and SO2 are independent of crude sulfur content and density. See Supporting Information for more detail.
Petroleum Refining Processes and Air Emissions Trends
Over the past 20 yr, the US crude slate has become slightly heavier and crudes with slightly higher sulfur content have been used. Over the next 10–20 yr, global production of light- and medium-density crude oils is expected to remain flat while heavy oil and nonconventional crude production are expected to grow slowly . Increased density increases crude oil processing complexity because more processing is needed to convert heavier crude to lighter products. In addition, if crudes have more sulfur and/or product sulfur specifications are raised, as mandated by the EPA, more processing is needed. Based on the observed trends, if modifications are made to refineries, CAP and HAP emissions may continue to decrease as refineries install modern equipment within the existing regulatory framework.
The analysis in this article has shown a significant reduction in CAP and HAP emissions by the petroleum refining industry despite changes in crude oil properties and refinery capacity increases. As a result of existing regulation, the petroleum refining industry and associated support industries have developed and implemented effective air pollution control technology. This has resulted in higher removal efficiencies and lower outlet concentration levels that will likely drive future best available control technology (BACT) determinations.
In addition, this study has documented a reduction in HAP emissions in the petroleum refining industry, which suggests a reduction in the overall toxicity hazard potential of HAP emissions. Approximately 3% of TRI emissions are grouped mixtures of compounds and were not included in the analysis. These grouped mixtures of compounds include metal compounds that TRI defines as including any unique chemical substance that contains the named metal (e.g., antimony, nickel, etc.) as part of that chemical's structure. There is no way of knowing the impact of these compounds without accurate speciation data.
The API and AFPM have stated that:
“Metals and sulfur partition as they pass through the refinery. The metals typically end up in the heavy oil fractions. The sulfur proportions among streams and then is removed from some streams and remains in other liquid streams. As a result, there is no correlation between sulfur and metals in crude and air emissions .”
Some USEPA tools have attempted to put TRI release data (including mixtures of metal compounds) into a chronic health context. In doing this, these tools assign all metal compounds the same toxicity weight as the parent metal compound associated with the highest chronic toxicity weight , even though the chronic toxicity of some metal compounds is actually lower and often much lower than the assumed toxicity weight. This approach introduces a great deal of uncertainty into the analysis, does not capture the actual toxicity of the mixtures of compounds, and was not conducted for purposes of this article. Because of limitations in the TRI emissions reporting of these compound categories, data is not available to determine the precise chemical makeup of these mixtures on an industry scale and to assign accurate IUR and RfC toxicity estimates.
Assessment of the CAP Emissions Estimates
Although NEI NOx inventories were verified through AP-42 emissions factors and industry fuel usage, similar analysis cannot be performed for SO2, VOC, and PM10 because these emissions result from numerous sources and are controlled through many different control technologies. However, it is assumed that these data sets are as reliable as the NOx data because the emission estimates were done in similar manners using emission factors, stack test data and continuous emissions monitors (CEMS). In the US petroleum refining industry, there has been increased integration of CEMS for the measurement of SO2 due to USEPA petroleum refining initiative. Similarly, added stack testing and monitoring of particulate matter emissions are likely resulting in more and more accurate emissions estimates.
For routine annual VOC emission estimates, petroleum refiners often rely on AP-42 emission factors. For VOC losses from piping components, the refinery may use their leak detection and repair (LDAR) programs and associated data bases and AP-42 emission factors. Component leak estimates can also be done using component-specific leak equations that use the measured leak concentration to estimate the emission rate . AP-42 VOC emission factors also exist for point sources of VOC emissions including fluidized catalytic cracking unit vents, cooling towers, blowdown systems, and vacuum distillation vents. Finally, AP-42 provides detailed and rigorous emission calculations for storage tanks (found in Ap-42 Chapter 7, Section 7.1), and organic loading and unloading operations (found in AP-42 Chapter 5, Section 5.2) are used in the calculation of annual VOC emissions. The NEI database uses annual emission estimates and does not specifically identify episodic emission events such as upsets, startups, or shutdowns.
Although some regulators have suggested that there might be a bias in the annual VOC emission estimates at petroleum refineries in the US, measurement data indicates that emissions estimates using emission factors such as those found in AP-42 are comparable to measured emission results. In 2007, USEPA stated concerns about US annual VOC emissions estimates based on VOC measurement studies performed in Europe and Canada using differential absorption light detection and ranging systems (DIAL) factors . According to the studies, fugitive emission at petroleum handling facilities in Europe and Canada that were estimated using VOC emission factors apparently did not include upsets, malfunctions, or certain VOC sources including “large VOC sources from unexpected sources ”. From this work, it was recommended that additional similar studies needed to be completed to determine if there are differences between measured and estimated emissions for US facilities.
CONCAWE, the European oil industry's organization for environment, health, and safety, conducted a study in 2008 to compare optical gas imaging (OGI) techniques such as DIAL measurements with flux measurements and AP-42 equations for making annual emission estimates . The study showed that DIAL measurements were slightly lower than CONCAWE's flux measurements, and that the AP-42 equations compared well on an hourly basis . It was further noted in this study that OGI measurements can only provide a short-term emission measurement and that use in an annual emission estimate can lead to large errors due to temporal variation of refinery emissions. In the United States, the US EPA's “critical review of DIAL emission test data for BP petroleum refinery in Texas City, Texas” (2010)  found the DIAL results cannot be used to assess the validity of default assumptions in AP-42 procedures nor how well the AP-42 procedures estimate short-term emission rates because the DIAL fluxes often included an unknown quantity of emissions from upwind sources. Therefore, NEI data remains the most reliable and representative profile of the petroleum refining industry's past and current annual emissions.
Assessment of HAP Emission Estimates
Because refineries process crude oil, the main HAP emissions are organic HAPs, including benzene, toluene, xylenes, ethylbenzene, 1,2,4-trimethyl benzene, and n-hexane. Sources of these HAPs include fugitive emissions from storage tanks and components, typically from equipment where volatile organic streams such as crude, reformates, and gasoline products are handled. Refinery emissions also include inorganic HAPs, such as nickel particulate matter from fluidized catalytic cracking units (FCCU) regeneration; and other HAP, compounds including diethanolamine from amine scrubbers used for the removal of sulfur compounds in fuel gases, polycyclic aromatic compounds and formaldehyde from high temperature combustion devices, and various other metals and organics used in the petroleum refining processes.
Although there is USEPA guidance on emission estimating, there are no required methods for estimating HAP emissions. As a result, TRI data should be carefully interpreted because variations between facilities can result from the use of different estimation methodologies . Also, the increases in the TRI releases from the period 1994–1996 were due to an increased number of chemicals being reported. In 1993, the TRI list of chemicals was increased with the addition of certain resource conservation and recovery act (RCRA) chemicals and certain hydrochlorofluorocarbons (HCFCs); and in 1994, a Chemical Expansion Final Rule (59 FR61431, November 30, 1994) was promulgated expanding TRI by 286 new chemicals and categories. The final TRI list report was increased to over 600 chemicals and categories.
Assessment of Air Pollution Control Technology
In general, US petroleum refineries are complex in that they are designed with several processing steps to produce relatively complex final products, mostly fuels. Refineries are dynamic operations and must adapt to changes in product demand, crude supply, and state and local mandates regarding fuel quality. At the same time, refineries must continue to comply with existing standards for air emissions. As new refinery products are added—even if driven by mandatory fuel quality standards—compliance with Clean Air Act regulations may trigger the need for additional air pollution controls.
To meet the ever more stringent emission limitations, the petroleum refining industry has invested in new, state of the art air pollution control technology while developing rigorous work practices, such as leak detection and repair (LDAR) programs and flare event minimization. Many of these technologies were developed and implemented for the first time in response to federal and state air quality regulations. Going forward, there may be incremental improvements to various technologies, as industry continues to seek more effective emission controls with improved operability, lower cost and increased reliability. While the technologies have been found to be cost-effective for achieving the desired emission reductions, and will continue to be employed to achieve comparable percentage levels of reductions, future technology developments and reductions will likely be less dramatic simply because refining emissions are already well-controlled.
Furthermore, a key lesson learned in air pollution control technology implementation is the need to avoid a “one size fits all” approach. Because each refinery and each emission source is unique, technology requirements need to be flexible and tailored to the site. As noted on USEPA's RACT/BACT/LAER Clearinghouse (RBLC) , numerous technologies exist for the same sources and species. This is an indicator of the variety of site-specific approaches that are necessary to effectively minimize air emissions. Finally, there are inherent limits to technology effectiveness. The fundamental laws of thermodynamics dictate that separation of pollutants from process effluents becomes more difficult as the concentration of the pollutant decreases. Even under the most optimistic assumptions for technology improvements, it is not possible to get to zero emissions, and controls become much more costly as lower concentrations are targeted.
A more detailed examination of the evolution of air pollution control technologies over the last 20 yr is beyond the scope of this article. See this article's Supporting Information for additional details.
CAP and HAP emissions from petroleum refineries in the United States have decreased over the last 20 yr despite increasing crude oil density, changes in sulfur concentrations and increasingly stringent product specifications. Emissions of CAPs for the total refining industry have decreased as much as 80% from 1990 to 2010. Emissions of HAPs and their associated toxicity hazard potential have decreased nearly 70%. Even though petroleum refineries have historically not been the most significant industrial emission source category for either CAP or HAP emissions in the United States, the refining industry has nevertheless significantly reduced air emissions. The methodologies for emissions estimation and measurement have continued to evolve since reporting of CAP and HAP emissions was initiated over 20 yr ago. As improved estimation techniques were developed and instrumentation employed to assist in estimation, the resultant emissions estimates have likely become more accurate. Examination of CAP emissions normalized to refinery crude feed rates has not shown any relationship to crude sulfur content and API gravity. Continued application of existing Clean Air Act regulations will continue to drive reductions in air emissions from refineries.
The author thanks Derick Kopp of Sage Environmental Consulting for assembling much of the data used in this article. Chevron USA provided funding for this article.