Loading...
HomeMy WebLinkAboutLetter submitted by Richard Klein: Gas Station Public Health Safety Zones 05-02-21COMMUNITY & ENVIRONMENTAL DEFENSE SERVICES Richard D. Klein, President (410) 654-3021 24 Greenshire Lane E-Mail Rklein@ceds.org Owings Mills, Maryland 21117 Web Page: www.ceds.org May 2, 2021 Commissioner Brad Grabow Commissioner Jeff Hill Commissioner Carrie Holle Commissioner Nick Kestner Commissioner Joshua Kirsh Commissioner Alan Potasnik Commissioner Kevin Rider Commissioner Susan Westermeier Commissioner Christine Zoccola City of Carmel Plan Commission One Civic Square Carmel, Indiana 46032 RE: Public Health & Safety Zones Needed for New Gas Stations in Carmel Dear Commissioners Grabow, Hill, Holle, Kestner, Kirsh, Potasnik, Rider, Westermeier & Zoccola: On behalf of the Carmel Healthy Neighborhoods Alliance and the 324 Carmel residents who have signed the Alliance petition, I respectfully request that the Plan Commission recommend approval of the proposed Unified Development Ordinance amendment increasing the public health safety zone for new gas stations to a minimum of 500 feet. Over the last decade a number of scientific studies have shown that the public health impact of gas station emissions are far greater than previously thought. These impacts result from benzene and other harmful compounds released to the air from underground storage tank vents and during refueling at the pump. The studies are summarized in this letter and the most relevant papers are attached at the end. Unfortunately, the Indiana Air Regulations for Area Source Gasoline Distribution Dispensing Facilities do not require measures that will resolve the public health impact. The only measure which can minimize adverse health effects is to guide new gas stations to locations where the site is at least 500 feet from residentially zoned properties, schools, and other locations where people regularly spend extended periods of time, especially the young and old. As explained in the last section of this letter, the proposed 500-foot public health safety zone will not preclude new gas stations with Carmel; just guide them to locations where public health will not be jeopardized. 2 SCIENTIFIC FINDINGS & GAS STATION HEALTH EFFECTS A number of compounds injurious to human health are released from gas stations during vehicle fueling and from underground storage tank vents: benzene, toluene, ethyl benzene, and xylene. Of these, benzene is the gasoline constituent most harmful to human health. Adverse health effects of benzene include cancer, anemia, increased susceptibility to infections, and low birth weight. According to the World Health Organization Guidelines for Indoor Air Quality, there is no safe level for benzene. As explained later in this letter, measures to reliably resolve these adverse health effects are not employed at new gas stations in Indiana. In 2005, the California Air Resources Board became the first agency in the U.S. to recommend a minimum public health safety zone between new gas stations and "sensitive land uses such as residences, schools, daycare centers, playgrounds, or medical facilities." The recommendation appeared in Air Quality and Land Use Handbook: A Community Health Perspective. The State of California is widely recognized as having some of the most effective air pollution control requirements in the nation. Yet even with California controls a minimum separation is still required to protect public health. The U.S. Environmental Protection Agency echoed concerns about the health risk associated with gas station emissions in their School Siting Guidelines. The USEPA recommended screening - but not excluding - school sites for potential health risk when located within 1,000 feet of a gas station. The safety zone distances were prompted by the growing body of research showing that adverse health effects extend further and further from gas stations. A seminal 2015 study contained the following summary regarding the health implications of living, working, or learning near a gas station: "Particularly affected are residents nearby gas stations who spend significant amounts of time at home as compared to those who leave their home for work because of the longer period of exposure. Similarly affected are individuals who spend time close to a gas station, e.g., in close by businesses or in the gas station itself. Of particular concern are children who, for example, live nearby, play nearby, or attend nearby schools, because children are more vulnerable to hydrocarbon exposure." A 2019 study of U.S. gas stations found that benzene emissions from underground gasoline storage tank vents were sufficiently high to constitute a health concern at a distance of up to 518-feet. Also, the researchers noted: "emissions were 10 times higher than estimates used in setback regulations [like that in the California handbook] used to determine how close schools, playgrounds, and parks can be situated to the facilities [gas stations]." 3 Prior to the 2019 study it was thought that most of the benzene was released at the pump during fueling. The 2015, 2019, and a 2020 study referenced, in the next section, accompany this letter. If you wish we can arrange for Carmel officials to speak with the scientists who performed these studies. CONTROL MEASURES WILL NOT RESOLVE HEALTH THREAT The two most common control measures are Stage II Vapor Recovery and Onboard Refueling Vapor Recovery (ORVR). A decade ago most gas pump nozzles were designed to capture vapors released during refueling. The vapors were then sent to the 10,000- to 20,000-gallon underground tanks where gasoline is stored. These Stage II vapor recovery systems were phased out beginning in 2012 as a result of the widespread use of Onboard Refueling Vapor Recovery (ORVR) systems. As the name implies, Onboard Refueling Vapor Recovery systems are built into new cars. The system captures vapors during refueling which are then stored in canisters within the vehicle. A study published in February, 2020, examined the effectiveness of Onboard Refueling Vapor Recovery systems. The researchers found that 88% of vehicles monitored released vapors during refueling despite the presence of Onboard Refueling Vapor Recovery systems. The 2019 study cited previously in this letter addressed the release of benzene from underground gasoline storage tank vents. The 2019 study documented that the amount of benzene released was substantial and could be detected at a distance of up to 518 feet. The unfortunate conclusion from these studies is that we cannot rely upon controls required for new vehicles and new gas stations to resolve the health and safety threat to those who live, learn, or work in the vicinity. As stated earlier in this letter, the Indiana Air Regulations for Area Source Gasoline Distribution/Dispensing Facilities do not require measures that will resolve the public health impact. The only measure which can minimize the public health impact is to guide new gas stations to locations where the site is at least 500 feet from any residentially zoned property, school, or other location where people - especially the young and old - regularly spend extended periods of time. NEW CARMEL GAS STATIONS NOT PRECLUDED BY HEALTH SAFETY ZONE The proposed 500-Foot Public Health Safety Zone will not preclude new gas stations within the City of Carmel. We know this because we analyzed the effect of the 500-foot setback on the 14 existing gas stations in Carmel. Of course, our proposed setback will only apply to new stations. The 14 stations are shown in the aerial photo on the next page. 4 For reasons described in the following video, new gas stations tend to locate near existing gas stations: https://www.youtube.com/watch?v=u4cKzGj58q4. Therefore, if most existing gas stations are more than 500 feet from the nearest home, then it's likely there are a number of locations where future gas stations could be built without posing a threat to the health of area residents. The red and green rings in the aerial above are located 500-feet from the outer edge of each gas station site. The stations with green rings are more than 500 feet from the nearest home. The red ringed stations are within 500 feet of a home. Of the 14 stations, nine have green rings. Assuming that future gas station siting follows past patterns, this exercise indicates that two- thirds of future sites would meet the 500-foot minimum public health safety zone. What the proposed Unified Development Ordinance amendment will accomplish is to guide new stations to sites where we can gain the benefits of these retail establishments without jeopardizing the health of our Carmel neighbors. It is for these reasons that the Carmel Healthy Neighborhoods Alliance urges the City of Carmel Plan Commission to recommend approval of the Unified Development Ordinance amendment which would require a minimum 500-foot public health safety zone for all new gas stations. Sincerely, Richard Klein Existing Carmel Gas Stations & Proximity to Homes Green circle = No homes within 500 feet; Red circle = Homes within 500 feet 4 mi N➤➤N © 2021 Google © 2021 Google © 2021 Google 9 2015 Study Hydrocarbon Release During Fuel Storage and Transfer at Gas Stations: Environmental and Health Effects AIR POLLUTION AND HEALTH (JD KAUFMAN AND SD ADAR, SECTION EDITORS) Hydrocarbon Release During Fuel Storage and Transfer at Gas Stations: Environmental and Health Effects Markus Hilpert 1 &Bernat Adria Mora 1 &Jian Ni 2 &Ana M. Rule 1 &Keeve E. Nachman 1 Published online: 5 October 2015 #Springer International Publishing AG 2015 Abstract At gas stations, fuel is stored and transferred be- tween tanker trucks, storage tanks, and vehicle tanks. During both storage and transfer, a small fraction of unburned fuel is typically released to the environment unless pollution preven- tion technology is used. While the fraction may be small, the cumulative release can be substantial because of the large quantities of fuel sold. The cumulative release of unburned fuel is a public health concern because gas stations are widely distributed in residential areas and because fuel contains toxic and carcinogenic chemicals. We review the pathways through which gasoline is chronically released to atmospheric, aque- ous, and subsurface environments, and how these releases may adversely affect human health. Adoption of suitable pol- lution prevention technology should not only be based on equipment and maintenance cost but also on energy- and health care-saving benefits. Keywords Gasstations .Vaporemissions .Fuelspills . Adversehealtheffects .Pollutionprevention Introduction Theprimaryfunctionofgasstationsistoprovidegasolineand dieselfueltocustomers,whorefillvehicletanksandcanisters. Operating a gas station requires receiving and storing a suffi- cient amount of fuel in storage tanks and then dispensing the fuel to customers. During delivery, storage, and dispensing of fuel at gas stations, unburned fuel can be released to the envi- ronment in either liquid or vapor form. Fuel is a complex mixture of chemicals, several of them toxic and carcinogenic [1]. Of these chemicals, the health consequences of chronic benzene exposure are best understood. Occupational studies have linked benzene exposures to numerous blood cancers, including acute myeloid leukemia and acute non- lymphocytic leukemia [2]. Concerns have been raised that gasoline vapor exposures incurred by gas station attendants [3] and tanker truck drivers [4] may result in health risks. The potential for fuel released to the environment at gas stations, in the form of liquid spills or vapor losses, to elicit adverse health outcomes could be substantial due to the wide- spread distribution of gas stations across communities and the intensive usage of vehicle fuel in industrialized nations. For example, the USA consumed about 137 billion gallons of gasoline, or about 430 gallons per US citizen, in 2014 [5]. If only a small fraction of this gasoline was to be released to the environment in the form of unburned fuel, for instance 0.1 %, then about 1.6 L of gasoline would be released per capita per yearintheUSA.InCanada,astudyestimatedthatevaporative losses at gas stations in 2009 amounted to 58,300,000 L [6]. Withapopulationofabout34million,weestimatedthatabout 1.7 L of gasoline was released per capita per year in Canada from evaporative losses, without counting the liquid spills. While personalintakeofthisquantityofgasolinewould result in serious adverse health effects, environmental dilution can decrease personal exposure. An overarching question is under which conditions dilution in the aqueous and atmospheric environments can limit personal exposures to acceptable levels. For example, cumulative adverse health effects could be more pronounced inmetropolitan areas where more people This article is part of the Topical Collection on Air Pollution and Health *Markus Hilpert markus_hilpert@jhu.edu 1 DepartmentofEnvironmentalHealthSciences,BloombergSchoolof Public Health, Johns Hopkins University, Baltimore, MD, USA 2 Carey School of Business, Johns Hopkins University, Baltimore, MD, USA Curr Envir Health Rpt (2015) 2:412–422 DOI 10.1007/s40572-015-0074-8 areexposedandwherethedensityofgasstationsislargerthan in rural areas. Engineers and regulators have paid a lot of attention to leaking underground storage tanks (LUSTs) and leaky piping betweenstoragetanksandgasoline-dispensingstations,which canresultincatastrophicfuelreleasetothesubsurface[7].For instance, double-walled tanks have become standard in order to minimize accidental release of liquid hydrocarbon. Tech- nologies that prevent pollution due to non-catastrophic and unreportedreleasesofhydrocarbonthatoccurduringfuelstor- age and transfer (hereafter referred to as Bchronic releases^), however, have not been uniformly implemented within the developed world. The state of California in the USA has the strictest policies to minimize chronic releases, either in liquid or in vapor form. Other US states and industrialized nations, however, have not uniformly adopted California’sstandards, potentially because comprehensive economic and public health analyses to inform policy making are not available. This paper focuses on chronic hydrocarbon releases at gas stations (including both liquid spills and vapor losses), their contributions to human exposures and potential health risks, and factors that influence the adoption of suitable pollution prevention technology. Chemical Composition of Fuel Fuels have historically contained significant fractions of harmful chemicals, some of which have been documented as contributing to morbidity and mortality in exposed persons. Crude oil, from which fuels have historically been refined, already contains toxic chemicals such as benzene [8]. Fuel additives includinganti-knockingagentsandoxygenateshave historically also been a health concern [9]. Fuel composition has changed over time, primarily due to environmental and health concerns [9]. Fuel composition also depends on geo- graphic location and fuel type (e.g., conventional versus reformulated gasoline) [10]. In the 1920s, lead was added to gasoline as an anti-knocking agent to replace added benzene because of its carcinogenicity [11 ]. Due to the massive release of lead to the environment and its neurotoxicity[12], lead was replaced in the 1970s by less toxic anti-knocking agents in- cluding methyl tert-butyl ether (MTBE) [13]. To reduce for- mation of ground-level ozone and associated adverse respira- tory health effects [14], cleaner burning of fuel was sought in the 1990s by adding oxygenates to gasoline. This was accom- plishedbyincreasingtheconcentrationsofMTBE,whichacts asan oxygenate [9].However, MTBE accidentally released to the subsurface [15] contaminated downstream drinking water wells relatively quickly, moving almost with the speed of groundwater, because MTBE is hydrophilic and poorly bio- degradable [16]. MTBE was later on identified as a potential human carcinogen [16]. In the USA, MTBE was therefore phased out in the 1990s; at the same time, refineries began supplementing fuel with ethanol as an oxygenate [17]. In current gasoline formulations, benzene, toluene, ethyl- benzene, and xylene (BTEX) and particularly benzene are the most studied chemicals and are currently believed to be of greatest health concern [18]. Table 1 shows that fuels have historicallycontainedlargefractionsoftoxicandcarcinogenic chemicals. In many countries, lead and MTBE are no longer used. Benzene levels in gasoline are currently much lower in most countries (e.g., on average 0.62 % by volume in the USA), though the chronic health effects of benzene and other BTEX chemicals at relevant exposure levels are not well understood. Chronic Release and Environmental Transport of Contaminants from Fuel At gas stations, fuel can be released in both liquid and vapor phases during delivery, storage, and dispensing. Direct vapor releaseisusuallyassociatedwithatmosphericpollution,while liquid spillage is commonly associated with soil and ground- water contamination. However, spilled liquid fuel also evap- orates into the atmosphere. Hypothetically, hydrocarbon va- pors can also condense back into liquid form; however, this appears to be unlikely due to quick dilution in a typically turbulent atmosphere. Figure 1 depicts how releases of un- burned fuel contaminate the atmospheric, subsurface, and sur- face water environments (omitting LUSTand leaky piping as well as marine gas stations which may release fuel directly to surface water). Liquid Fuel Spills Liquid fuel spills at the nozzle have received less attention than liquid releases due to LUSTs. These fuel spills occur when the dispensing nozzle is moved from the dispensing station to the vehicle tank and vice versa, when the automatic shutoff valve fails, due to spitback from the vehicle tank after the shutoff has beenactivated,andwhenthe customertopsoff the tank. Table 1 Historical content of non-negligible amounts of toxic and carcinogenic chemicals in fuel Chemical of concern Fraction Health effects Benzene Up to 5 % [75]Carcinogenic [2] Lead Up to 2 g per gallon [76]Central nervous system [12] MTBE Up to 15 % [77]Potential human carcinogen [78] Curr Envir Health Rpt (2015) 2:412–422 413 In a study quantifying fuelspill frequencies and amounts at gasstationsinCalifornia,about6Lofgasolinewasspilledper 16,200 gallons of gasoline dispensed at gas stations without stage II vapor recovery compared to 3.6 L at gas stations per 14,043gallons ofgasolinedispensedatgas stationswithstage II vapor recovery (at the nozzle) [19]. This would mean that about 0.007 and 0.01 % of dispensed gasoline are spilled in liquid form during vehicle refueling at gas station with and without stage II recovery (numbers calculated using the as- sumedfueldensityof6.2pounds/gallon).Ontheotherhand,a study sponsored by the American Petroleum Institute found that more spills occurred at gas stations with stage II recovery [20]. We have recently performed laboratory experiments to ex- amine the fate of liquid spill droplets. Following our previous protocol [21•], we spilled fuel droplets onto small concrete samples and measured the mass added to the concrete as a function of time. This added mass is the sum of the masses of the sessile fuel droplet and the infiltrated fuel. Figure 2 shows results for diesel and gasoline. After a certain period of time, the sessile droplet vanishes and the measured mass levels off. The remaining mass represents the infiltrated por- tion. The evaporated mass can be obtained by subtracting the infiltrated mass from the initial droplet mass m0. Evaporation is greater for gasoline, while infiltration is greater for diesel spills. This is because gasoline is more volatile than diesel. Diesel has therefore a higher potential for soil contamination because of the higher infiltrated mass. Spilled fuel may move downward in liquid or vapor phase and potentially reach the groundwater table. The physical mechanisms that govern subsurface movement of spilled fuel are the same as for fuel released due to LUST, except that spilled fuel must first penetrate relatively impermeable pave- ment underneath fuel-dispensing stations. Gasoline and diesel will not penetrate the groundwater table as a liquid, because they have densities lower than that of water. Released fuel may also evaporate within the sediment, and a portion of it will move downward as a vapor and potentially reach the groundwatertable[22].Whetherthefuelreachesgroundwater in liquid or vapor form, the fuel will then partition into groundwater and become a dissolved chemical that is carried away by molecular diffusion and groundwater flow and asso- ciated hydrodynamicdispersion [23]. Therefore, the spills can contaminate downstream drinking water wells [24]. Biodeg- radation can decrease contaminant concentrations significant- ly; however, its efficiency depends on many factors including the chemical composition of the fuel and the presence of suit- able microbial species that can metabolize a given contami- nant, bioavailability, and electron acceptor availability [25]. Partitioning of the contaminant into other phases will cause Fig. 1 Gas stations are embedded into the natural environment and can consequently release pollutants to the atmosphere, the subsurface including soil and groundwater, and surface water 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 04812162024m/m0t (h) Diesel Gasoline Fig. 2 Results from laboratory experiments, in which we spilled a mass m0 =1 g of diesel or gasoline onto concrete samples. The measured mass m represents the masses of the sessile droplet and infiltrated liquid 414 Curr Envir Health Rpt (2015) 2:412–422 retarded transport of the contaminant within groundwater. For instance, hydrophobic contaminants such as benzene tend to sorb to the sediment. For this reason, large-scale contamina- tion of aquifers and associated adverse health effects due to the ingestion of contaminated drinking water from these aqui- fers are often considered a lesser concern for hydrophobic contaminants [16]. Stocking et al. [26] evaluated the potential of groundwater contamination due to small one-time releases of liquid gaso- line. Ina casestudy, theyassumed a spill volumemuchbigger than the ones typically measured by the study of gas stations in California [19], i.e., 0.5 L, and they concluded the risk to groundwater to be small. This analysis, however, did not in- clude consideration of a key mechanism for fuel spillage; namely, that much smaller droplets are typically released dur- ing vehicle refueling [19]. To address this question, Hilpert and Breysse [21•] calculated cumulative spill volumes due to repeated small spillages that occur at gasoline-dispensing fa- cilitiesandestimatedthatagasstationsellingabout400,000L of gasoline per month would spill at least 150 L each year. They also developed a model that shows that the fraction of spilled gasoline that infiltrates into the pavement increases as the droplet size decreases. Therefore, repeated small spills could be of greater concern for groundwater contamination than an instantaneous release of the cumulative spill volume; thus, a risk to groundwater may not be as small as previously estimated. Laboratory experiments and modeling have shown that gasoline from small-volume spills can infiltrate into the con- crete that usually covers the ground underneath gasoline- dispensing stations—despite the low permeability of concrete and the high vapor pressure of gasoline [21•]. It is unlikely that liquid fuel fully penetrates a concrete slab to contaminate the underlying natural subsurface due to the low permeability of concrete [27],although preferential pathways for fluid flow such as cracks and faulty joints between concrete slabs can allow for such liquid penetration. It has been hypothesized that evaporation of infiltrated gasoline and subsequent down- ward migration of the vapor through the concrete may lead to contamination of underlying sediment and groundwater [21•]. Consistent with these two proposed pathways of subsurface contamination, soil/sediment underneath concrete pads of a gas station in Maryland was contaminated by diesel oil and gasoline (leaky piping could have also contributed to the con- tamination) [28]. Runoff water that flows over pavement can also get con- taminated with hydrocarbons spilled onto the pavement [29–31], and such contamination has specifically been linked to gas stations [32–34]. If a spill occurs while runoff occurs, the hydrocarbon can be expected to float on top of the water sheet, because gasoline, diesel oil, and lubricants are typically less dense than water (light non-aqueous phase liquids or LNAPLs). While runoff water is not directly ingested, it is funneled into the stormwater drainage system, and may be released to natural water bodies, often without treatment. Whereas volatilization decreases contaminant levels in the stormwater within hours depending on the exact environmen- tal conditions [35], and biodegradation will further decrease levels, significantly contaminated stormwater might be re- leasedto naturalwater bodiesif they are close by. Finally, fuel spilled at marine gas stations may directly enter natural water bodies. Vapor Fuel Releases Fuel evaporative losses have received more attention than liquid fuel spills (even though they are related) [6,36]. These losses are related to the fact that the headspace above liquid fuel in vehicle and storage tanks tends to approach thermody- namic equilibrium with the liquid. Consequently, almost sat- uratedgasolinevaporscanbereleasedtotheatmospherewhen tanks are refueled, unless a suitable vapor recovery system is in place. Since saturated gasoline vapors have a density that is three to four times larger than the one of air, i.e., 4 kg/m 3,and the density of liquid gasoline is about 720 kg/m 3 [37], about 0.5 % of liquid gasoline dispensed to a tank is released to the atmosphere if the entire headspace is in equilibrium with the liquid fuel. This is true for any type of tank, whether it is a vehicle tank, a canister, an underground storage tank (UST), or an above-storage tank. The percentage loss is less if a tank received clean air relatively recently, e.g., when the fuel level in a storage tank drops because of gasoline-fuel dispensing. It is important to note that vapor recovery at the nozzle can cause vapor releases at the storage tank, because vapors re- covered at the nozzle are typically directed into the storage tank. The storage tank, in turn, can Bbreathe^and potentially releaserecoveredvaporsimmediatelyoratalatertime.Atank sucks in relatively uncontaminated air as the liquid fuel level drops in the tank due to vehicle refueling, and it releases va- pors through the vent pipe into the atmosphere if the gas pressure increases and exceeds the cracking pressure of the pressure/vacuum valve, when fuel evaporates into unequilibrated gas in the headspace. As discussed in the BLiquid Fuel Spills^section above, we note that liquid spills also contribute to air pollution because spilled droplets form sessile droplets on pavement that can then evaporate into the atmosphere. On concrete, most of spilled gasoline droplets evaporate into the atmosphere (Fig.2). This, however, does not mean that the small fraction that infiltrates into the concrete is not of concern. Exposure and Risks to Human Populations Gas stations exist as part of the built environment and are widely distributed across communities. As a result, they may be surrounded by residential dwellings, businesses, and other Curr Envir Health Rpt (2015) 2:412–422 415 buildings such as schools. Operation of gas stations may thus create opportunities for a variety of human populations to be exposedtovaporsduringstationtankfillingandvehicle refueling. These human populations can be broadly grouped into three groups: populations exposed occupationally as a result of employment in various capacities at the service sta- tion; those exposed as customers engaging in vehicle refueling; and those passively exposed either by residing, at- tending school, or working near the refueling station. The exposures to benzene and other components of refueling va- pors and spills experienced by these populations vary based on a number of factors, including the size and capacity of the refueling station, spatial variation in pollutant concentrations in ambient air, climate, meteorological conditions, time spent at varying locations of the service station, changing on-site activity patterns, physiological characteristics, and the use of vapor recovery and other pollution prevention technologies. Employees at service stations (such as pump attendants, on-site mechanics, and garage workers) are among those with greatestexposuretobenzene originating fromgas stations [3]. These receptors spend the most time on site (potentially reflecting approximately 40 h per week, for decades) and in- termittently spend time where vapors from the pump are at their highest concentrations, with benzene concentrations measuring between 30 and 230 ppb in the breathing zone [38–40]. Gas station patrons can also be exposed to vapors when refueling. Compared to station employees, their expo- sures are brief and transient. A Finnish study reported a me- dian time spent refueling of approximately 1 min, whereas 3 min was the median duration inthe USA [41,42]. The same USstudyreportedanaveragebenzenepersonalexposurecon- centration at the pump of 910 ppb, with the strongest predic- tors of benzene levels being fuel octane grade, duration of exposure, and season [42]. Those occupying residences, businesses, and other struc- tures neighboring gas stations can also be exposed to fuel vapors originating in the gas station, though typically at lower concentrations than those measured at the pump. While vapor concentrations will drop as the distance from the service sta- tionincreases,exhaustfumesfromwaitingcustomersandfuel delivery trucks can also contribute to vapors in proximity to gas stations. A small number of studies have examined ben- zene concentrations at the fenceline of the service station and beyond. A study published by the Canadian petroleum indus- try found average benzene concentrations of 146 and 461 ppb at the gas station property boundary in summer and winter, respectively [43]. A South Korean study examined outdoor and indoor benzene concentrations at numerous residences within 30 m and between 60 and 100 m of gas stations and found median outdoor benzene concentrations of 9.9 and 6.0 μg/m3 (about 3.1 and 1.9 ppb), respectively. Median in- door concentrations at these locations were higher, reaching 13.1 and 16.5 μg/m3 (about 4.1 and 5.2 ppb), respectively [44]. Another study found median ambient benzene levels of 1.9 ppb in houses both <50 and >100 m from a service station [45]. Yet, another study [46] found that benzene and other gasoline vapor releases from service stations can be discerned from traffic emissions as far as 75 m from service stations and that the contribution of service stations to ambient benzene is less important in areas of high traffic density. This is because vehicle exhaust is usually the most abundant volatile organic compound (VOC) in urban areas, often followed by gasoline vapor emissions from fuel handling and vehicle operation [47]. Beyond contact with surface-level gasoline vapors, fuel releases may result in other exposure pathways. Soil and groundwater contamination is common at gas stations. Drink- ing water wells proximate to gas stations, which in rural areas are often the only drinking water source, can become contam- inated, potentially exposing well users to benzene and other chemicals [48,49]. In addition, runoff from rain and other weather events can carry spilled hydrocarbons, which can contaminate surface waters; those using surface waters, either recreationally or for other purposes, may be exposed to these contaminants through dermal contact or incidental ingestion. In the USA, the Environmental Protection Agency (EPA) regulates releases of benzene under the Clean Air Act as a hazardous air pollutant, and benzene is listed as number 6 on the 2005 priority list of hazardous substances under the Com- prehensive Environmental Response, Compensation, and Li- ability Act and any release greater than 10 pounds triggers a reporting requirement. Different quantitative toxicity metrics exist for benzene inhalation. The EPA Integrated Risk Infor- mation System (IRIS) has published a reference concentration of 0.03 mg/m 3 (about 9.4 ppb), corresponding to decreased lymphocyte counts [50], whereas the NIOSH recommended exposure limit (REL) is a time-weighted average concentra- tion (for up to a 10-hour workday during a 40-hour work- week) of 0.319 mg/m 3 (about 100 ppb) [51]. While research attention has been paid to measurement of gasoline vapor constituent concentrations in air at and near service stations, less is known about the health consequences faced by those that are exposed to gasoline vapors. Of the limited literature examining these exposures, service station workers have received the greatest attention, and exposure is often assessed as a function of job title, rather than specific measurements of vapor constituent concentrations. An older study looking broadly at leukemia incidence in Portland, Or- egon, found that gas station workers were at significantly in- creased risk for lymphocytic leukemia [52]. A proportionate mortality ratio analysis of all deaths recorded in New Hamp- shire among white men from 1975 to 1985 found elevated leukemia mortality in service station workers and auto me- chanics [53]. The type of leukemia was not specified. An Italian occupational cohort study of refilling attendants that examinedrisksamongworkersatsmallergasstationsreported 416 Curr Envir Health Rpt (2015) 2:412–422 non-significant increases in mortality for non-Hodgkin’slym- phoma and significantly elevated mortality for esophageal cancer in men, as well as increased brain cancer mortality in both sexes [54]. A different cohort of 19,000 service station workersinDenmark,Norway,Sweden,andFinlandexamined an array of cancer end points and found increased incidence for multiple sites (nasal, kidney, pharyngeal, laryngeal, and lung) among workers estimated to be occupationally exposed tobenzeneintherangeof0.5–1 μg/m3 (0.16-0.31ppb).Non- significant increased incidence was found for acute myeloid leukemia in men and for leukemia different from acute mye- loid leukemia and chronic lymphocytic leukemia in women [55]. A case–control study of multiple occupations including subjectsfromtheUSAandCanadafoundsignificantincreases in rates of total leukemia and acute myeloid leukemia but not acute lymphocytic leukemia in gas station attendants [56]. A 2015 review of studies examining potential relationships be- tween benzene exposures and hematopoietic and lymphatic cancers among vehicle mechanics yielded inconclusive re- sults, although it suggested that if an effect was to exist, it would be small and difficult to rigorously ascertain with existing epidemiologic methods [57]. The health consequencesof nearbyresidents of gas stations have not been studied. However, it is known that contaminat- ed groundwater can affect large numbers of people if the groundwater is used as drinking water, as was the case in Camp Lejeune (North Carolina, USA) where thousands were exposed to a range of chemicals including gasoline released from LUSTs [58]. A study of Pennsylvania residents residing incloseproximity to alargegasolinespill fromaLUST found evidence of increased leukemia risks [49,59••]. The health consequences of chronic fuel releases at gas stations that can, for example, occur due to ingestion of contaminated groundwater, fuel vapor intrusion from contaminated soil and groundwater into dwellings [60], and atmospheric vapor releasesduringfueltransferandstoragehavenotbeenstudied. While limited measurements of ambient concentrations of va- por constituents in communities were identified, literature searches did not identify studies of the health consequences of inhalation exposures to gasoline vapors among community residents [61]. Pollution Prevention Pollution prevention technologies have been developed that can efficiently reduce the releases of unburned fuel to the environment that routinely occur during fuel storage and transfer (see Fig.3): 1.Stage I vapor recovery collects vapors that would be ex- pelledfromUSTs duringfueldelivery[62].Withoutstage I vapor recovery, about 80 kg of gasoline vapor would be released from a 40 m 3 UST if one assumes a saturated vapordensityof4kg/m3 [37]andvaporsintheheadspace Fig. 3 There are several sources of chronic release of unburned fuel at gas stations that occur due to fuel storage and dispensing: vapor release through the vent pipe of the storage tank, vapor release from the vehicle tank during refueling, leaky dispensing hoses, liquid spills during vehicle refueling,andvaporemissionsthroughevaporationofthisspilledfuel.As indicated, suitable pollution prevention technology can minimize the releases. Onboard refueling vapor recovery (ORVR) Curr Envir Health Rpt (2015) 2:412–422 417 to be at half saturation. Stage I vapor recovery can thus prevent substantial fuel vapor releases that would occur within a short period of time. Such releases might expose tanker truck drivers and persons in the proximity of a gas station to significant doses of fuel vapors. Stage I vapor recovery is accomplished by establishing a closed loop between the UST and the tanker truck. Through a fuel delivery hose, liquid fuel is pumped into the UST, while a vapor recovery hose directs vapors displaced from the UST into the headspace of the tanker truck. Stage I vapor recoveryiscurrentlyrequiredforhigh-throughputgassta- tions in all states in the USA and in most countries. 2.Stage II vapor recovery technology can efficiently collect vapors expelled from vehicle tanks during refueling, there- by minimizing personal exposure of customers and workers to fuel vapors during dispensing of gas [63]. Re- covered vapors are directed into the UST. Two technolo- gies for stage II vapor recovery have been developed, the vaccum-assist method and the balance method. In the vacuum-assist method, contaminant-laden air is actively removed/pumped from the nozzle into the UST. In the balance method, displaced vapors are passively withdrawn by connecting the vapor recovery hose to the inlet of the vehicletankviaanairtightseal.Thepressureincreaseinthe headspace of the vehicle tank provides a driving force that seeks to push the vapors into the storage tank. Stage II vapor recovery has been required in many states of the USA and in other countries, although there is currently an effort todecommissionstageII vaporrecovery(seebelow). 3.Technology development at the hose and nozzle level can also contribute to reduced fuel releases. Low-permeation hoses, for instance, limit the release of gasoline vapors through the wall of the refueling hoses [64]. Dripless noz- zles have been developed to minimize liquid spills that occur when the nozzle is moved between the fill pipe and the dispensing unit. 4.Passenger vehicles and trucks can be equipped with on- board refueling vapor recovery (ORVR) systems which di- rect vapors that, during vehicle refueling, would be released to the atmosphere into an activated carbon-filled canister in the vehicle [65,66]. Collected vapors are later reintroduced into the vehicle’s fuel system. However, canisters, motorcy- cles, and boats are not equipped with ORVR. 5.Impermeable liners underneath the concrete pads can re- duce the risk of soil and groundwater contamination once environmentalfuelreleases,inliquidorvaporphase,have occurred. However, this technology might eventually re- sult in air pollution, because liquid fuel that is hindered from moving downward in the concrete pad will tend to saturate the pavement and eventually evaporate into the atmosphere. 6.Finally,unburnedfuelvapor can bereleased fromanUST when the tank pressure exceeds the cracking pressure of the pressure/vacuum valve and itcan beprevented bytwo pressure management techniques, burning or separation of air and fuel vapors. Released air/fuel vapors can be burned, however, which results in the release of combustion-related pollutants into the atmosphere. Alter- natively, a semi-permeable membrane can be used to sep- arate the air from the fuel vapors. Depressurization of the tank is then achieved by releasing the relatively clean air through the pressure/vacuum valve to the atmosphere. When it comes to evaluating the efficiency of vapor recov- ery during liquid transfer between tanks, it is of upmost im- portance to consider potential releases from all tanks; they form a system. Otherwise, the overall efficiency of stage II vapor recovery cannot be understood. For instance, stage II vapor recovery based on the vacuum-assist method can nega- tively interfere with ORVR. In that case, no vapors are re- leased from the vehicle tank and the stage II pump draws relatively clean air from the atmosphere into the storage tank. IntheUST,thisairwillbecomesaturatedwithfuelvaporsthat evaporatefromthestoredfuel.Thisresultsinpressurizationof the USTand release of contaminant-laden air if the tank pres- sure exceeds the cracking pressure of the pressure/vacuum valve of the UST. This might occur immediately or at a later point in time. However, there are stage II systems that do not negatively interfere with ORVR including the balance method. Estimates for the efficiency of pollution technologies are usually provided by the manufacturers. However, adoption of these technologies by gas station owners usually relies on the certification and quantification of efficiencies by independent parties. In the USA, the Cal- ifornia Air Resources Board and EPA typically assume this role [36]. Consultants and environmental agencies have used these estimates to determine current releases of unburned fuel to the environment and to evaluate the effects of pollution prevention technology [67]. While many studies have found health benefits from pol- lution prevention technology intended to minimize chronic gasoline spills, these studies typically do not quantify overall financial benefits and costs. Instead, only equipment and maintenance cost are typically considered [68]. Adopting the new equipment can reduce fuel losses and reduce environ- mental cost and health risks. However, this new equipment comes with non-trivial upfront costs. It is therefore a concern that the related policy-making process of chronic fuel spills relies only on non-comprehensive cost estimates. Studies are needed that account for health care cost due to released pol- lutantsandenergy-savingbenefitsduetopollutionprevention. Such econometric studies have, for example, been performed in the context of pollutant emissions from coal-fired power plant and commercial real estate development [69••,70]. At times, there is also the perception that pollution prevention 418 Curr Envir Health Rpt (2015) 2:412–422 costsareonlycarriedbythespecificindustry[71]. Adoption of the environmentally friendly technology could be slow when the firms have long equipment replacementcyclesorwhen the firms do not have sufficient information to evaluate whether or not a switch to an environmentally friendly technology is in theirprivateinterests.Itis,however,notclearthatthisapparent investment,intheformofpreventioncost,mightalsobepartly shouldered by customers and that this apparent cost might actually (at least in the long run) be beneficial to customers, gas station workers, nearby residents, and other populations that spend significant amounts of times in the proximity of gas stations (e.g., school children in nearby schools). Policy intervention is often expected to expedite the adoption of such environmental friendly technologies, in order to reduce the difference in the private and social values of adoption. Efforts are currently underway that could potentially allow decommissioning stage II vapor recovery in the USA due to the widespread use of ORVR in the motor vehicle fleet [68]. However, the remaining legacy fleet without ORVR and all motorcycles and boats (lacking ORVR) can produce signifi- cant emissions during vehicle refueling, emissions that could be avoided by stage II vapor recovery. For the State of Mary- land, it has been estimated that fuel consumption of non- ORVR-equipped vehicles was about 10 % in 2015 (Table 4 in [67]). These emissions can result in direct hydrocarbon exposures among vehicle owners during vehicle refueling as well as in passive exposure of other populations. A compre- hensive cost analysis of the decommissioning of stage II re- covery represents an opportunity to inform policy makers on their recommendation with regards to stage II recovery. Conclusions Even if only a small fraction of unburned fuel is lost during vehicle refueling and fuel storage, the cumulative release of fuel to the environment can be large if large total amounts of fuel are dispensed at gas stations. For instance, about 0.01 % of fuel can be spilled during the refueling process and up to about 0.5 % can be lost in vapor form if equilibrated gasoline vapors are released from a tank to the atmosphere during refueling (worst-case scenario). For a medium-size gas sta- tion, which sells 400,000 L of gasoline per month, this results in 480 L of spilled gasoline and in 24,000 L of liquid gasoline that is anually released in vapor form to the environment. Even though dilution can reduce concentrations of released contamination, research is needed to assess whether such re- leases represent an environmental health concern. The potential for pollution prevention, moreover, is sub- stantial. Technology has already been developed and partially employed that can efficiently decrease vapor losses and liquid spills. Particularly, when it comes to vapor losses, it is crucial to consider not only vapor recovery at the vehicle tank/nozzle but also at the storage tank, since vapors recovered at the nozzle are directed into the storage tank, from which they might be potentially released. While California has imple- mented the strictest regulations when it comes to preventing chronic hydrocarbon releases at gas stations, other highly in- dustrialized states and nations do not employ the same stan- dards for different reasons. For instance, pressure/vacuum valves on vent pipes of fuel storage tanks are not common in Canada, because they might freeze in the wintertime, poten- tially causing a tank implosion [6]. Relatively little research has been done on potential soil and groundwater contamination due to chronic releases of liquid fuel during vehicle refueling. Unlike catastrophic re- leases, such as LUST, chronic spills are not reported. Limited field investigations suggest that spilled fuel may penetrate concrete underneath dispensing pads to contaminate underly- ing sediment. However, it is possible that such soil contami- nation occurs routinely over the life span of a gas station and that this contamination pathway is masked or erroneously ex- plainedbyleaksinthepipingfromtheUSTstothedispensers. Overall, large-scale soil and groundwater contamination by fuel appears to be a lesser problem, because many of the toxic compounds in fuel are hydrophobic (including BTEX) and can therefore be expected not to travel too far in groundwater. However, customers, gas station workers, and nearby resi- dents may get exposed to the hydrocarbons if groundwater is used as a drinking water supply or if fuel vapor intrusion in dwellings occurs. Health effects of living near gas stations are not well un- derstood. Adverse health impacts may be expected to be higher in metropolitan areas that are densely populated. Par- ticularly affected are residents nearby gas stations who spend significantamountsoftimeathomeascomparedtothosewho leave their home for work because of the longer period of exposure. Similarly affected are individuals who spend time close to a gas station, e.g., in close by businesses or in the gas station itself. Of particular concern are children who, for ex- ample, live nearby, play nearby, or attend nearby schools, because children are more vulnerable to hydrocarbon expo- sure [72]. Potential future changes in fuel composition might pose new environmental health challenges as there is a history of adding even large amounts of toxic substances to fuel (Table 1). Changes in fuel composition could occur due to an increasing usage of biofuels, or to comply with air quality standards, which might also change over time. Chemicals newly added to fuel or changes in chemical concentrations can have unforeseen ramifications. One could argue that fu- ture fuel composition changes will be performed with more care; however, it was only in the 1990s, decades after the Safe Drinking Water Act (SDWA) was passed in 1974, that MTBE was added to gasoline without critically evaluating its trans- port behavior in groundwater and toxicity, a mistake which Curr Envir Health Rpt (2015) 2:412–422 419 nowadays is considered avoidable [73]. Interestingly, ethanol, which has largely replaced MTBE, can inhibit biodegradation of BTEX, which is not the case for MTBE [74]. Given the complexitiesofchemicalfateandtransportintheenvironment and the potential for insufficient toxicity testing, using appro- priate pollution prevention technology that minimizes release of unburned chemicals with known and unknown adverse health effects during fuel storage and transfer seems a wise, long-term, and cost effective idea given ever-changing fuel compositions. Finally, employing efficient pollution prevention technolo- gy might be economically advantageous. The evaluation of economic benefits of pollution prevention technology needs to account not only for the cost of implementation and main- tenance of such technology but also for public health burdens due to released pollutants and energy-saving benefits due to valuable hydrocarbons not wastefully released to the environment. Acknowledgments This work was funded by a seed grant from the Environment, Energy, Sustainability and Health Institute at Johns Hop- kins University. Compliance with Ethics Guidelines Conflict of Interest Markus Hilpert, Bernat Adria Mora, Jian Ni, Ana Rule, and Keeve Nachman declare that they have no conflict of interest. Human and Animal Rights and Informed Consent This article does not contain any studies with human or animal subjects performed by any of the authors. References Papers of particular interest, published recently, have been highlighted as: •Of importance ••Of major importance 1. Wang Z et al. Characteristics of spilled oils, fuels, and petroleum products: 1. composition and properties of selected oils. United States Environmental Protection Agency. Report No.: EPA/600/R- 03/072, 2003. 2. IARC,IARCmonographsontheevaluationofcarcinogenicrisksto humans. Vol. 100F. 2012. 3. Karakitsios SP et al. Assessment and prediction of exposure to benzene of filling station employees. Atmospheric Environment. 2007;41(40):9555–69. 4. Javelaud B et al. Benzene exposure in car mechanics and road tanker drivers. International Archives of Occupational and Environmental Health. 1998;71(4):277–83. 5. U.S. Energy Information Administration. How much gasoline does the United States consume? 2015 [cited 2015 July 20]; Available from:http://www.eia.gov/tools/faqs/faq.cfm?id=23&t=10. 6. Statistics Canada. Gasoline evaporative losses from retail gasoline outlets across Canada 2009. Environment Accounts and Statistics Analytical and Technical Paper Series, 2012. 7. Dowd RM. Leaking underground-storage tanks. Environmental Science & Technology. 1984;18(10):A309–9. 8. Centers for Disease Control and Prevention. Facts about benzene. 2015 Available from:http://www.bt.cdc.gov/agent/benzene/basics/ facts.asp. 9. Nadim F et al. United States experience with gasoline additives. Energy Policy. 2001;29(1):1–5. 10. Weaver JW, Exum LR , Prieto LM. Gasoline composition regula- tions affecting LUSTsites. U.S. Environmental Protection Agency Office of Research and Development Washington, DC 20460, 2010. Report No.: EPA 600/R-10/001. 11. Needleman HL. The removal of lead from gasoline: historical and personal reflections. Environmental Research. 2000;84(1):20–35. 12. Grandjean P, Landrigan PJ. Developmental neurotoxicity of indus- trial chemicals. Lancet. 2006;368(9553):2167–78. 13. Thomas VM. The elimination of lead in gasoline. Annual Review of Energy and the Environment. 1995;20:301–24. 14. U.S. Environmental Protection Agency. National Ambient Air Quality Standards for ozone; final rule. Fed Reg. 2008;73:16436– 514. 15. U.S. Environmental Protection Agency, Achieving clean air and clean water: The report of the blue ribbon panel on oxygenates in gasoline, 1999. 16. SquillacePJetal.Reviewoftheenvironmentalbehaviorandfateof methyl tert-butyl ether. Environmental Toxicology and Chemistry. 1997;16(9):1836–44. 17. U.S. Energy Information Administration, Eliminating MTBE in gasoline in 2006. 2006. 18. U.S. Environmental Protection Agency, Fuel oxygenates (MTBE, TBA, and ethanol). 2015. 19. Morgester JJ, Fricker RL, Jordan GH. Comparisonof spill frequen- cies and amounts at vapor recovery and conventional service sta- tions in California. Journal of the Air & Waste Management Association. 1992;42(3):284–9. 20. MuellerEA.Asurveyandanalysisofliquidgasolinereleasedtothe environment during vehicle refueling at service stations. Washington, DC: American Petroleum Institute; 1989. 21.•Hilpert M, Breysse PN. Infiltrationand evaporation of small hydro- carbon spills at gas stations. Journal of Contaminant Hydrology. 2014;170:39–52.This study examines for the first time the fate and transport of small fuel droplets spilled during vehicle refueling. 22. Dakhel N et al. Small-volume releases of gasoline in the vadose zone: impact of the additives MTBE and ethanol on groundwater quality. Environmental Science & Technology. 2003;37(10):2127– 33. 23. Charbeneau R.J., Groundwater hydraulics and pollutant transport 2006: Waveland Press, Inc. 24. Grady, S. and G. Casey, Occurrence and distribution of methyl tert- butyl ether and other volatile organic compounds in drinking water in the northeast and mid-Atlantic regions of the United States, 1993–98. Water Resources Investigations Report WRIR 00–4228. Geological Survey, U.S., 2001. 25. Leahy JG, Colwell RR. Microbial-degradation of hydrocarbons in the environment. Microbiological Reviews. 1990;54(3):305–15. 26. Stocking, A.S., et al., Evaluation of fate and transport of methyl tertiary butyl ether (MTBE) in gasoline following a small spill. In: Stanley, Anita, (eds.) Petroleum hydrocarbons and organic chemicals in ground water—prevention, detection, and remedia- tion, Houston, Tex., Nov. 17–19, 1999. Proceedings.: National Ground Water Association, and American Petroleum Institute, 1999: p. 229–246. 27. Jacobs P. Permeabilität und Porengefüge Zementgebundener Werkstoffe. ETH Zürich: Switzerland; 1994. 28. Aria Environmental Inc., Underground Storage Tank Closure Report Chesapeake House –Exxon Facility (Northern Service 420 Curr Envir Health Rpt (2015) 2:412–422 Station) I-95 Travel Plaza, North East, Cecil County, Maryland, 2014, Maryland Transportation Authority. 29. Latimer JS et al. Sources of petroleum-hydrocarbons in urban run- off. Water Air and Soil Pollution. 1990;52(1–2):1–21. 30. Ohe T, Watanabe T, Wakabayashi K. Mutagens in surface waters: a review. Mutation Research-Reviews in Mutation Research. 2004;567(2–3):109–49. 31. Hoffman EJ et al. Urban runoff as a source of polycyclic aromatic- hydrocarbons to coastal waters. Environmental Science & Technology. 1984;18(8):580–7. 32. Borden RC, Black DC, McBlief KV. MTBE and aromatic hydro- carbons in North Carolina stormwater runoff. Environmental Pollution. 2002;118(1):141–52. 33. Garcia MR et al. Assessment of polycyclic aromatic hydrocarbon influxandsedimentcontaminationinanurbanizedestuary.Environ Monit Assess. 2010;168(1–4):269–76. 34. Khan E,Virojnagud W, RatpukdiT. Use ofbiomasssorbentsforoil removal from gas station runoff. Chemosphere. 2004;57(7):681–9. 35. U.S. Environmental Protection Agency. Technical factsheet on: Benzene. 2015 Available from:http://www.epa.gov/ogwdw/pdfs/ factsheets/voc/tech/benzene.pdf. 36. CARB. Vapor Recovery Program. 2015 Available from:http:// www.arb.ca.gov/vapor/vapor.htm. 37. International Chemical Safety Cards (ICSC). ICSC #: 1400. 2015 Available from:http://www.cdc.gov/niosh/ipcsneng/neng1400. html. 38. van Wijngaarden E, Stewart PA. Critical literature review of determinants and levels of occupational benzene exposure for United States community-based case–control studies. Applied Occupational and Environmental Hygiene. 2003;18(9):678–93. 39. Hartle R. Exposure to methyl tert-butyl ether and benzene among service station attendants and operators. Environmental Health Perspectives. 1993;101 Suppl 6:23–6. 40. Periago JF, Zambudio A, Prado C. Evaluation of environmental levels of aromatic hydrocarbons in gasoline service stations by gas chromatography. Journal of Chromatography A. 1997;778(1– 2):263–8. 41. Vainiotalo S et al. Customer exposure to MTBE, TAME, C6 alkyl methyl ethers, and benzene during gasoline refueling. Environmental Health Perspectives. 1999;107(2):133–40. 42. Egeghy PP, Tornero-Velez R, Rappaport SM. Environmental and biological monitoring of benzene during self-service automobile refueling. Environmental Health Perspectives. 2000;108(12): 1195–202. 43. Akland GG. Exposure of the general population to gasoline. Environmental Health Perspectives. 1993;101 Suppl 6:27–32. 44. Jo W-K, Moon K-C. Housewives’exposure to volatile organic compounds relative to proximity to roadside service stations. Atmospheric Environment. 1999;33(18):2921–8. 45. Jo W-K, Oh J-W. Exposure to methyl tertiary butyl ether and ben- zene in close proximity to service stations. Journal of the Air & Waste Management Association. 2001;51(8):1122–8. 46. Terrés IMM et al. Assessing the impact of petrol stations on their immediate surroundings. Journal of Environmental Management. 2010;91(12):2754–62. 47. Watson JG, Chow JC, Fujita EM. Review of volatile organic com- pound source apportionment by chemical mass balance. Atmospheric Environment. 2001;35(9):1567–84. 48. Wallace LA. The exposure of the general population to benzene. Cell Biology and Toxicology. 1989;5(3):297–314. 49. Patel AS et al. Risk of cancer as a result of community exposure to gasoline vapors. Archives of Environmental Health. 2004;59(10): 497–503. 50. U.S. Environmental Protection Agency. Benzene (CASRN 71-43- 2). 2015 Available from:http://www.epa.gov/iris/subst/0276.htm. 51. National Institute for Occupational Safety and Health. NIOSH PocketGuidetoChemicalHazards:Benzene.2015Availablefrom: http://www.cdc.gov/niosh/npg/npgd0049.html. 52. MortonW,MarjanovicD.Leukemiaincidencebyoccupationinthe Portland‐Vancouver metropolitan area. American Journal of Industrial Medicine. 1984;6(3):185–205. 53. Schwartz E. Proportionate mortality ratio analysis of automobile mechanics and gasoline service station workers in New Hampshire. American Journal of Industrial Medicine. 1987;12(1): 91–9. 54. Lagorio, S., et al. Mortality of filling station attendants. Scand Journal Work Environ Health 1994: 331–338. 55. Lynge E et al. Risk of cancer and exposure to gasoline vapors. American Journal of Epidemiology. 1997;145(5):449–58. 56. Terry PD et al. Occupation, hobbies, and acute leukemia in adults. Leukemia Research. 2005;29(10):1117–30. 57. Hotz P, Lauwerys RR. Hematopoietic and lymphatic malignancies in vehicle mechanics. Critical Reviews in Toxicology. 1997;27(5): 443–94. 58. Savitz, D.A.e.a., Contaminated water supplies at Camp Lejeune: assessing potential health effects 2009: National Academies Press. 59.••Talbott EO et al. Risk of leukemia as a result of community expo- sure to gasoline vapors: a follow-up study. Environmental Research. 2011;111(4):597–602.This study suggests a possible association between chronic low-level benzene exposure due to a leaking underground storage tank and increased risk of leu- kemia among residents that live nearby a gas station. 60. Sanders PF, Hers I. Vapor intrusion in homes over gasoline- contaminated ground water in Stafford, New Jersey. Ground Water Monitoring and Remediation. 2006;26(1):63–72. 61. Caprino L, Togna GI. Potential health effects of gasoline and its constituents: a review of current literature (1990–1997) on toxicological data. Environmental Health Perspectives. 1998;106(3):115. 62. U.S. Environmental Protection Agency, Design criteria for Stage 1 vapor control systems—gasoline service stations. 1975. 63. U.S. Environmental Protection Agency, Technical guidance— stage II vapor recovery systems for control of vehicle refueling emissions at gasoline dispensing facilities, Volume I: Chapters, 1991. 64. McPhee, J., Gasoline dispensing facility (GDF) balance hose per- meation study, 2008, California Air Resources Board (CARB) 65. U.S. Environmental Protection Agency. Commonly asked ques- tions about ORVR. Available from:http://www.epa.gov/otaq/regs/ ld-hwy/onboard/orvrq-a.txt. 66. Musser, G. and H. Shannon, Onboard control of refueling emis- sions. 1986: p. SAE Technical Paper 861560. 67. Meszler Engineering Services, Stage II emission reduction benefits. Report to the Maryland Department of the Environment. 2012. 68. Federal Register, Air quality: widespread use for onboard refueling vapor recovery and stage II waiver. Final Rule by US Environmental Protection Agency. Federal Register 2012. 77(95). 69.••Currie J et al. Environmental health risks and housing values: evi- dence from 1,600 toxic plant openings and closings. American Economic Review. 2015;105(2):678–709.This study provides an empirical framework for understanding the effects and the health cost of toxic atmospheric emissions. The research design could be applied to pollution prevention at gas stations. Such analysis could provide important policy recommendation in or- der to mitigate this type of environmental risk. 70. Ni J. Environmental cost and economic benefit of commercial real estate development. Working Paper: Johns Hopkins University; 2015. Curr Envir Health Rpt (2015) 2:412–422 421 71. Fong M et al. California dry cleaning industry technical assessment report. State of California Air Resources Board: Technical report; 2006. 72. Irigaray P et al. Lifestyle-related factors and environmental agents causing cancer: an overview. Biomedicine & Pharmacotherapy. 2007;61(10):640–58. 73. McGarityTO.MTBE:aprecautionarytale.HarvardEnvironmental Law Review. 2004;28(2):281–342. 74. Powers SE et al. The transport and fate of ethanol and BTEX in groundwater contaminated by gasohol. Critical Reviews in Environmental Science and Technology. 2001;31(1):79–123. 75. Jakobsson R et al. Acute myeloid-leukemia among petrol station attendants. Archives of Environmental Health. 1993;48(4):255–9. 76. U.S. Environmental Protection Agency, EPA requires phase-out of lead in all grades of gasoline. EPA press release –November 28, 1973. 1973. 77. U.S. Energy Information Administration (EIA), MTBE, oxygen- ates, and motor gasoline. 2000. 78. U.S. Environmental Protection Agency. Assessment of potential health risks of gasoline oxygenated with Methyl Tertiary Butyl Ether (MTBE). Washington,DC:Office of Research and Development, U.S. EPA. 1993. EPA/600/R-93/206. 422 Curr Envir Health Rpt (2015) 2:412–422 10 2019 Study Vent pipe emissions from storage tanks at gas stations: Implications for setback distances This page is intentionally blank Vent pipe emissions from storage tanks at gas stations: Implications for setback distances Markus Hilpert a,⁎, Ana Maria Rule b, Bernat Adria-Mora a, Tedmund Tiberi c a Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, NY 10032, United States of America b Department of Environmental Health and Engineering, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD 21205, United States of America c ARID Technologies, Inc., Wheaton, IL 60187, United States of America HIGHLIGHTS •At gas stations, fuel vapors are released from storage tanks through vent pipes. •We measured vent pipe flow rates and tank pressure at high temporal resolu- tion. •Vent emission factors were N10 times higher than previous estimates. •Modeling was used to examine exceed- ance of benzene short-term exposure limits. GRAPHICAL ABSTRACT abstractarticle info Article history: Received 3 July 2018 Received in revised form 11 September 2018 Accepted 23 September 2018 Available online 24 September 2018 Editor: Pavlos Kassomenos At gas stations, fuel vapors are released into the atmosphere from storage tanks through vent pipes. Little is known about when releases occur, their magnitude, and their potential health consequences. Our goals were toquantify ventpipe releasesand examineexceedance ofshort-term exposure limitsto benzene around gas sta- tions.AttwoUSgasstations,wemeasuredvolumetricventpipe flowratesandpressureinthestoragetankhead- space at high temporal resolution for approximately three weeks. Based on the measured vent emission and meteorological data, we performed air dispersion modeling to obtain hourly atmospheric benzene levels. For the two gas stations, average vent emission factors were 0.17 and 0.21 kg of gasoline per 1000 L dispensed. Modeling suggests that at one gas station, a 1-hour Reference Exposure Level (REL) for benzene for the general population (8 ppb) was exceeded only closer than 50 m from the station's center. At the other gas station, the REL was exceeded on two different days and up to 160 m from the center, likely due to non-compliant bulk fuel deliveries. A minimum risk level for intermediate duration (N14–364 days) benzene exposure (6 ppb) was exceeded at the elevation of the vent pipe opening up to 7 and 8 m from the two gas stations. Recorded vent emission factors were N10 times higher than estimates used to derive setback distances for gas stations. Setback distances should be revisited to address temporal variability and pollution controls in vent emissions. © 2018 Elsevier B.V. All rights reserved. Keywords: Gas stations Benzene emissions Setback distances Air pollution modeling Measurements Science of the Total Environment 650 (2019) 2239–2250 ⁎Corresponding author at: Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, 722 West 168th St., New York, NY 10032, United States of America. E-mail address:mh3632@columbia.edu (M. Hilpert). https://doi.org/10.1016/j.scitotenv.2018.09.303 0048-9697/© 2018 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Science of the Total Environment journal homepage:www.elsevier.com/locate/scitotenv 1. Introduction In the US, approximately 143 billion gal (541 billion L) of gasoline were dispensed in 2016 at gas stations (EIA, 2017) resulting in release of unburned fuel to the environment in the form of vapor or liquid (Hilpert et al., 2015). This is a public health concern, as unburned fuel chemicals such as benzene, toluene, ethyl-benzene, and xylenes (BTEX) are harmful to humans (ATSDR, 2004). Benzene is of special concern because it is causally associated with different types of cancer (IARC, 2012). Truck drivers delivering gasoline and workers dispensing fuel have among the highest exposures to fuel releases (IARC, 2012). However,peoplelivingnearorworkinginretailatgasstations,andchil- dren in schools and on playgrounds can also be exposed, with distance to the gas stations significantly affecting exposure levels (Terres et al., 2010;Jo & Oh, 2001;Jo & Moon, 1999;Hajizadeh et al., 2018). A meta- analysis (Infante, 2017) of three case-control studies (Steffen et al., 2004;Brosselin et al., 2009;Harrison et al., 1999) suggests that child- hood leukemia is associated with residential proximity to gas stations. Sources of unburned fuel releases at gas stations include leaks from storage tanks, accidental spills from the nozzles of gas dispensers (Hilpert & Breysse, 2014;Adria-Mora & Hilpert, 2017;Morgester et al., 1992),fugitivevaporemissionsthroughleakypipesand fittings, vehicle tank vapor releases when refueling, and leaky hoses, all of which can contribute to subsurface and air pollution (Hilpert et al., 2015). Routine fuel releases also occur through vent pipes of fuel storage tanks but are less noticeable because the pipes are typically tall, e.g., 4 m. These vent pipesareputinplacetoequilibratepressuresinthetanksandcanbelo- cated as close as a few meters from residential buildings in dense urban settings (Fig. 1). Unburned fuel can be released from storage tanks into the environ- ment through “working”and “breathing”losses (Yerushalmi & Rastan, 2014). A working loss occurs when liquid is pumped into or out of a tank.Forastoragetank,thiscanhappenwhenitisrefilledfromatanker truck or when fuel is dispensed to refuel vehicles (Statistics Canada, 2009) if the pressure in the storage tank exceeds the relief pressure of thepressure/vacuum(P/V)valve (EPA, 2008).P/Vvalvethreshold pres- sures are typically set to around +3 and -8 in. of water column (iwc) (7.5 and -20 hPa). However, P/V valves are not always used, particu- larly in cold climates, as valves may fail under cold weather conditions (Statistics Canada, 2009). Breathing losses occur when no liquid is pumped into or out of a tank because of vapor expansion and contraction due to temperature and barometric pressure changes or because pressure in the storage tank may increase when fuel in the tank evaporates (Yerushalmi & Rastan, 2014;EPA, 2008). Although delayed or redirected by the P/V valve, breathing emissions can be significant and represent an environ- mental and health concern (Yerushalmi & Rastan, 2014). Stage I vapor recovery systems, put in place to prevent working losses while delivering fuel to a station, collect the vapors displaced while loading a storage tank, redirecting them into the delivery truck. Stage II vapor recovery systems minimize working losses while deliver- inggasfromthestoragetanktothecustomer'scar.DuringStageIIvapor recovery, gasoline vapors can be released through the vent pipe, if the sumof the flowratesofthereturnedvolumeandofthefuelevaporating withinthestoragetankisgreaterthanthevolumeofliquidgasolinedis- pensed (Statistics Canada, 2009). We refer to this scenario as pressure while dispensing (PWD). In theory, a properly designed Stage II vapor recovery system should not have working losses, although in practice this is not typically the case (McEntire, 2000). Regulations on setback distances for gas stations are based on life- time cancer risk estimates. Several studies have assessed benzene can- cer risk near gas stations (Atabi & Mirzahosseini, 2013;Correa et al., 2012;Cruz et al., 2007;Edokpolo et al., 2015;Edokpolo et al., 2014; Karakitsios et al., 2007). Based on cancer risk estimations, the California Air Resources Board (CARB) recommended that schools, day cares, and othersensitivelandusesshouldnotbelocatedwithin300ft.(91m)ofa large gas station (defined as a facility with an annual sales volume of 3.6 million gal = 13.6 million L or greater) (CalEPA/CARB, 2005). This CARB recommendation has not been adopted by all US states, and within states setback distances can depend on local government. Nota- bly, CARB regulationsdo not accountfor short term exposure limits and health effects. An important limitation of existing regulations is the use of average gasolineemission rates estimated in the 90s that do not con- sider excursions (CAPCOA, 1997). The main objective of this study is to evaluate fuel vapor releases through vent pipes of storage tanks at gas stations based on vent emis- sion measurements conducted attwo gasstations in theUSin 2009and 2015, including the characterization of excursions at a high temporal resolution (~minutes) and meteorological conditions at an hourly tem- poral resolution.In addition,we performed hourly simulationsof atmo- spheric transport of emitted fuel vapors to inform regulations on setback distances between gas stations and adjacent sensitive land uses by comparing modeled benzene concentrations to four 60-min benzene exposure limits: an acute Reference Exposure Level (REL) for infrequent (once per month or less) exposure (WHO, 2010)and Emergency Response Planning Guidelines ERPG-1, ERPG-2 and ERPG- 3(AIHA,2016).FinallywecomparedsimulatedbenzenelevelstoaMin- imal Risk Level (MRL) for benzene for intermediate exposure duration (14 to364 days)(ATSDR,2018) becausethatdurationwindowincludes our duration of data collection. See Table 1 for the various benzene ex- posure limits and issuing agencies. 2. Methods Although we provide SI unit conversions, we report some measures in English engineering units (ft, gal, and lb) as regulatory agencies such as CARB use these units. 2.1. Sites Data for this study were obtained from vent release measurements conducted at two gas stations as part of technical assistance to the gas stations to quantify fuel vapor losses through the vent pipes of their storage tanks. A motivation for conducting the measurements was to perform a cost-benefit analysis to compare the economic losses due to the lost fuel versus the cost of technologies that reduce the emissions. The exact location of the two gas stations is not revealed for confidenti- ality reasons. The gas station managers and staff who authorized the Fig. 1.The three vent pipes (enclosed by the red ellipse) on the right side of the convenience store of a gas station are b10 m away from the residential building. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 2240 M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 collection and analysis of these data have not been involved in the cur- rent manuscript. The firstgas station,“GS-MW,”waslocated intheUSMidwestand is a 24-hour operation. The study was conducted from December 2014 to January2015for20fulldays,andfuelsales _V sales wereabout450,000gal (1.7 million L) per month. Fuel deliveries to the gas station usually took place during the nighttime. The second gas station,“GS-NW,”was lo- cated on theUSNorthwestcoast and closed at night. Hours of operation were between 6:00 am and 9:30 pm on weekdays and between 7 am and 7 pm on weekends. That study was conducted in October 2009 for 18 full days, and fuel sales were _V sales ~700,000 gal (2.6 million L) per month. Both gas stations are considered to be high-volume, because they dispense N3.6 million gal of gasoline (both regular and premium) per year (CalEPA/CARB, 2005), and fuel was stored in underground storage tanks (USTs), which is typical in the US. Both gas stations had Stage II vapor recovery installed using the vacuum-assist method. In that method, gasoline vapors, which would be ejected into the atmosphere as a working loss during refueling of customer vehicle tanks, are col- lected atthe vehicle/nozzle interface bya vacuum pump. The recovered vapors are then directed via a coaxial hose backinto the combined stor- age tank ullage (head space) of the gas station. Stage I vapor recovery was also used at both gas stations during fuel deliveries. Both sites had a 3-inch diameter (7.5 cm) single above-grade vent pipe with below- grade manifold that connected the vent lines from several USTs; the cracking pressures of the P/V valves were set to +3 and -8 iwc (+7.5 and -20 hPa). 2.2. Vent emission measurements Toquantifyevaporativefuelreleasesthroughtheventpipeof a stor- age tank, the volumetric flow of the mixture of gasoline vapor and air was measured in the vent pipe. A dry gas diaphragm flow meter (American Meter Company, Model AC-250) was used. For each cubic foot (28 L) of gas flowing through the meter, a digital pulse was gener- ated. Every minute, the number of pulses was read out and stored to- gether with date and time on a data logger. Gas flow meters were obtained from a distributor calibrated and equipped with temperature compensation and a pulse meter. To determine the time-dependent volumetric flow rate Q(t)ofthe gasoline vapor/air mixture through the vent pipe, the time series of measured flow volumes were integrated over an averaging period (15 or 60 min) and divided by the duration of that period. I.e.,Q(t)is given by the number of pulses registered by the gas flow meter in a time window multiplied by 1 cubic foot and divided by the averaging time. The 15-minute averaging time was chosen to visualize time- dependent data, while the 60-minute averaging time was chosen be- cause air pollution simulations were performed at that resolution. Gas pressure p in the ullage of the storage tank was measured to as- sess vent emission patterns. For instance, releases can occur when the pressure exceeds the cracking pressure of the P/V valve in the vent pipe (the dry gas flow meter was fitted with a P/V valve on the outlet). Pressure was measured with a differential pressure sensor (Cerabar PMC 41, Endress + Hauser) every 4 s, and 2-minute average values were stored. The sensor range was scaled from -15 to +15 iwc (-37 to +37 hPa), with a full scale accuracy of 0.20%. We also obtained 15- and 60-minute averaged tank pressure data p(t) where averages repre- sent the means of the 2-minute average pressure measurements taken during each time window. 2.3. Descriptive analysis For the 60-minute flow rate, we calculated medians and inter quar- tile ranges (IQRs). To illustrate diurnal fluctuations in vapor emissions, we created box plots for the 60-minute flow rate distribution that oc- curred during each hour of the day. Spearman correlation coefficients between the time series for pressure and flow rate were calculated to evaluate whether pressure can be used to infer vent emissions. To estimate the mass flow rate of gasoline _mgas that is released through the vent pipe in the form of a mixture of gasoline vapors and freshair,weassumed,followingtheprotocolofastudybytheCalifornia Air Pollution Control Officers Association (CAPCOA) that assessed risks from fuel emissions from gas station (Appendix D-2 (CAPCOA, 1997)), that the density of gasoline vapors in this mixture is given by ρgas (v)= 0.3 × 65 lb / 379 ft 3 =0.824kg/m3, i.e., the molar percentages of gaso- line and air were 30% and 70%, respectively. Then the volumetric flow rate Q can be converted into a mass flow rate of the vaporized gasoline: _mgas ¼ρ vðÞ gas Q ð1Þ To arrive at ventemission factors, we first calculated themeanvolu- metric flow rate Q , and then the mean mass flow rate _mgas ¼ρðvÞ gas Q . From the latter, one can calculate the vent emission factor EFvent ¼_mgas =_V sales ð2Þ For EFvent, CARB uses units of pounds of emitted gasoline vapors (also calledtotalorganicgases(TOG))per1000galdispensed,ormorebriefly lb/kgal where kgal stands for kilogallons. Aswewerenotabletomeasurebenzenelevelsinthetankullage,we assumed like the CAPCOA study (Section C) that the density of the mix- ture of gasoline vapors and fresh air was ρmix (v)=1.05 lb/ft 3 = 1.682 kg/m 3 and that the emitted gasoline vapor/air mixture contained 0.3% of benzene by weight (CAPCOA, 1997). Therefore, the mass flow rate of benzene through the vent pipe was estimated as follows: _mbenz ¼0:003 ρ vðÞ mix Q ð3Þ 2.4. Air pollution modeling We used the AERMOD Modeling System developed by the US Envi- ronmental Protection Agency (EPA) to model thedispersion of benzene vaporsreleasedintotheenvironmentthroughventpipesof fuelstorage tanks and from other sources (Cimorelli et al., 2005). AERMOD simu- lates atmospheric pollutant transport at a 1-hour temporal resolution. 3Dpolargridswerecreatedwiththegasstationintheoriginandpoten- tial receptors at differentradial distances (up to170 m) and angles (10° Table 1 Benzene exposure limits, to which we compared simulation results. For unit conversion, we assumed a temperature of 25 °C, i.e., 1 ppm = 3194 μg/m3 (CAPCOA, 1997). Agency Name Value (ppb) Value (μg/m3)Exposure duration California Office of Environmental Health Hazard Assessment (OEHHA)REL 8 26 1 h American Industrial Hygiene Association (AIHA)ERPG-1 50 159,700 1 h AIHA ERPG-2 150 479,100 1 h AIHA ERPG-3 1000 3,194,000 1 h Agency for Toxic Substances and Disease Registry (ATSDR)MRL 6 19 14 to 364 days ERPG = Emergency Response Planning Guidelines. The primary focusof ERPGs isto provide guidelines for short-term exposuresto airborne concentrations of acutelytoxic, high-priority chemicals. 2241M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 increments).Thegrids were placed at thegroundlevel(z=0m),inthe breathing zone (z = 2 m), and at the 2nd floor level (z = 4 m) where the vent pipe emissions were assumed to occur. The topography was simplified for modeling purposes consistent with the CAPCOA study (CAPCOA, 1997), i.e., the terrain was assumed to be flat with no build- ings present. Vent pipe emissions were modeled as a capped point source. Chemical reactions of benzene were not modeled, as residence times of atmospheric benzene are on the order of hours or even days (ATSDR, 2007), i.e. much longer than the travel time of benzene vapors across the 340-m diameter model domain. For the period of time when vent emission measurements were made, we obtained meteorological data at a 1-hour temporal resolution that are representative for the geographic locations of the two gas sta- tions. Table SI-1 provides descriptive statistics of that data. The time se- ries were used in AERMOD to model the transport of benzene in the temporally varying turbulent atmosphere. We also used the 1-hour av- erage time series of benzene emission rates (Eq.(3)) as an input into AERMOD. Toevaluate at each grid pointwhether OEHHA's acute REL or AIHA's ERPG levels were exceeded at least once, we determined maximum 1- hour average benzene concentrations that were simulated for about three weeks. To evaluate how often the OEHHA REL was exceeded at each grid point in the breathing zone, we created plots indicating the number of exceedances and the day when the maximum benzene level was observed. To facilitate comparison to published benzene measurements around gas stations, we determined for each simulated radial distance from a gas station the mean of the average concentrations simulated for each ten degree increment on the radius around the gas station. 3. Results: vent releases 3.1. Times series of tank pressure and flow rate Fig.2 showsthetime-seriesdataforthevolumetric flowrate Q ofthe gasoline vapor/air mixture through the vent pipe and tank pressure p that we collected at the two gas stations. At GS-MW, little vapor was typically released in the late night and in the very early morning, whilereleasesweregenerallymuchhigherduringthedaytimeandeve- nings, presumably when more fuel was dispensed (Fig. 2a). Occasion- ally, no vapor releases occurred for several hours. While we do not have access to time of fuel delivery records,field visits indicate that timeperiods withnoreleasescoincide withfueldeliveries.Forinstance, fuel delivery likely occurred on January 6 at 7 pm(see Fig. 3a;an ampli- fication of data shown in Fig. 2a). As a result, the UST pressure dropped by about 10 hPa, far below the cracking pressure of the P/V valve. The decreased gas pressure in the ullage increased until the cracking pres- sure of the P/V valve was reached. A very small vapor release (~2 L/min) was observed briefly on the next day at 2 am. The vapor flow rate becomes relatively large again, ~12 L/min, only after 6 am, i.e., 11 h after fuel delivery. Fig. 3bamplifies a major vapor release at GS-MW. The UST pressure significantly exceeded the cracking pressure of the P/V valve and rose rapidly up to 37 hPa, which coincides with vapors being released at a high flow rate (15-min average) of about 470 L/min. AtGS-NW,vaporreleasesfollowedaquitedifferentpattern(Fig.2b). Contrary to GS-MW, vapor releases occurred in a cyclical pattern, and tended to be higher in the late night and in the very early morning when the gas station was closed. 3.2. Statistics of vapor emissions The average volumetric flow rate Q throughtheventpipefortheen- tire period of time during which measurements were taken was Q = 7.9 L/min for GS-MW and Q = 15.4 L/min for GS-NW, which is consistent with the higher sales volume _V sales of GS-NW. These emis- sions consist of a mixture of gasoline vapors and air. Using Eq.(1),the volumetric flow rates were converted into average mass flow rates of gasoline:_mgas = 0.39 kg/h for GS-MW and _mgas = 0.76 kg/h for GS- NW. Using Eq.(2), we determined a vent emission factor EFvent= 0.17 kg per 1000 L = 1.4 lb/kgal for GS-MW and EFvent= 0.21 kg per 1000 L = 1.7 lb/kgal for GS-NW. The medians (IQRs) for the 60-minute averaged flow rate Q (L/min) were 6.1 (1.9, 10.9) for GS-MWand 16.0 (12.7, 18.4) for GS-NW. For GS- MW, the mean is larger than the median, indicating a more skewed dis- tribution of flow rates when compared to GS-NW. Also the first quartile ismuchlowerthanthemedianforGS-MW,indicatingthattherearepe- riods of time during which little emissions occurred. Conversely, GS- NW was releasing emissions more consistently. Fig. 4a shows boxplots illustrating the distribution of flow rate Q for eachhourofthedayatGS-MW.Lessvaporwasreleasedbetween10pm and 4 am, even though the gas station was in operation, albeit at lower activity levels. The flow rate Q at GS-NW (Fig. 4b) had fewer outliers, and the highest outlier was an order of magnitude lower than the highest one at GS-MW. Emissions were highest between 1 and 3 am, when the gas station was closed. The Spearman correlation coefficients between tank pressure p and vent flow rate Q were r = 0.58 for GS-MW and r = 0.85 for GS-NW. Thus, vent releases are moderately and strongly correlated with tank pressure, respectively.Table 2 summarizes statistical properties of vent emissions at the two gas stations. 4. Results: air pollution modeling 4.1. Emission sources and rates Vent pipe emissions of benzene were modeled at a 1-hour temporal resolution as described in Section 2.4. However, they are not the sole source of gasoline emissions at gas stations. Accidental spills from noz- zles regularly occur near the dispensers,“refueling losses”can occur when gasoline vapors are released from the vehicle tank during refueling due to the rising liquid levels in the tanks, fuel vapors are re- leased from permeable dispensing hoses, and “fugitive”or leakage emissions occur with driving force derived from storage tank pressure. In Section A of Supportingmaterial, we detail how these other emission sources were modeled.Table 3 summarizes estimated mean emission rates. Note that the vent pipe losses are much greater than other losses. 4.2. Predicted benzene levels Fig. 5 shows for both gas stations and at each grid point the maxi- mum1-houraveragebenzeneconcentrationobservedduringthesimu- lated periods in time. Benzene levels depend significantly on elevation within a 50-meter radius around the centers of the gas stations. Close to the centers of the gas stations, benzene levels are higher at the 4-m elevation and at ground level due to vent pipe emissions, which repre- sent the largest emission source (Table 3). Further than 50 m away fromthecenter,theverticalconcentrationdifferencesbecomelessobvi- ous due to dispersion causing vertical mixing of benzene vapors. At GS-MW, the 1-hour acute REL of 26 μ g/m3 was exceeded 160 m away from the center of the gas station, at the location (x =158m,y = 28 m) both at ground level and in the breathing zone. At grid points with a distance N50 m from the center of the gas station, the REL was exceeded at most once (Fig. SI-1a). How- ever, the exceedance at different grid points did not occur on the same day (Fig. SI-1b). Within the 20 days during the measure- ment campaign, exceedances occurred on the 4th and 13th of January. At GS-NW, the furthest REL exceedance occurred at 50 m from the center of the gas station at the grid point (x =-38 m,y =32m)as 2242 M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 Fig. 2.Time series of ullage pressurep(left ordinate) and volumetricflow rateQ(right ordinate) for (a) GS-MW and (b) GS-NW. Horizontal tick marks indicate midnights. The vertical dashed and thick solid gray lines enclose weekends.2243M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 Fig.3.Amplificationsoftimeseriesdata(15-minuteaverages)forGS-MW.(a)Tankpressure p becamenegativeafterfueldelivery.Asaresult,ventemissionceasedforseveralhours.(b)A major vapor release (burst) likely occurred when the cracking pressure of the P/V valve was significantly exceeded at around 9 pm during a non-compliant bulk fuel delivery. (a) (b) Fig.4.Distributionofventemissions Q observedforeachhourofthedayat(a)GS-MW[insertshowstheIQRsof Q]and(b)GS-NWgasstations.In(a),outliersmakeitdifficulttorecognize variations in median hourly emissions. We therefore plotted in the inset only the IQRs. Boxes indicate median and IQR, whiskers values within 1.5 the IQR, and asterisks outliers. 2244 M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 shown in Fig. SI-2a. At a distance of 40 m, the REL was exceeded three times at one grid point (260° angle), and at 35 m four times at two grid points (250° and 260° angles) (Fig. SI-2b). At a distance of 20 m, theRELwasexceededat30(outof36)gridpoints,andonninedifferent days. Average benzenelevels are shown in Fig. 6 for both gas stations. The MRL is exceeded at the elevation of the vent pipe opening,z =4m,up to 7 m away from for GS-MW and up to 8 m from GS-NW.Fig. 7 shows theaveragebenzeneconcentration asa functionof distanceataneleva- tion of 2 m. Close to the center, benzene levels first increase and then decrease. 5. Discussion 5.1. Vent emission factors We present unique data on vent emissions from USTs at two gas sta- tions. Emissions can be compared to vent losses assumed by CAPCOA (CAPCOA, 1997). For a gas station with StageI and II vapor recovery tech- nology and a P/V valve on the vent pipe of the UST (Scenario 6B), the CAPCOA study assumed loading losses of 0.084 and breathing losses of 0.025 lb/kgal dispensed. The total loss of gasoline through the vent pipe is the sum of the two and amounts to a vent emission factor EFvent= 0.109 lb/kgal. Based on actual measurements in two fully functioning US gas stations, we obtained EFvent values of 1.4 lb/kgal for GS-MW and 1.7 lb/kgal for GS-NW, more than one order of magnitude higher than the CAPCOA estimate. While the difference between our measurements and the CAPCOA estimates may appear surprising, it is important to con- sider that the CAPCOA estimates are based on relatively few measure- ments and some unsupported assumptions (Aerovironment, 1994), particularlywithregardtouncontrolledemissionsduetoequipmentfail- ures or defects (Appendix A-5 (CAPCOA, 1997)). 5.2. Pressure measurements Tank ullage pressure p was moderately to strongly positively cor- related with vent flowrate Q, likely because exceedance of the crack- ing pressure of the P/V valve causes a vent release. Thus pressure measurements can be used to infer vent releases. Real-time detec- tion of equipment failures and leaks via so-called in-station diagnos- tics systems is based on our observed correlations between p and Q. 5.3. Diurnal fluctuations in vent emissions Diurnal vent emissions were quite different at the two gas stations. At GS-MW, a 24-hour operation, vent emissions were high during the daytime, presumably due to PWD. Emissions ceased at night, likely be- cause less gasoline was dispensed and fuel deliveries with relatively cool product were frequent. Evaporative losses could also have been lower at nightbecause thecoolerdelivered fuel would cause slightcon- tractionoftheliquidphasewithcorrespondinggrowthintheullagevol- ume while at the same time lowering the vapor pressure of gasoline in the UST. At GS-NW, vent pipe releases occurred most of the time, during the daytimewhenfuelwasdispensed(PWD)andatnightwhenthegassta- tion was closed. Vent releases were higher when the gas station was closed, suggesting that during the day-time Stage II vapor recovery re- sulted in the injection of vapors into the storage tank that were not completely equilibrated with the liquid gasoline. During night-time, the gradual equilibration of unsaturated air in the ullage of the UST withgasolinevaporscouldthenhavecausedexceedanceofthecracking pressure of the P/V valve and consequently vapor release. It seems counterintuitive that less nighttime emissions occurred at the gas sta- tion where fuel was dispensed. However, while fuel is being dispensed, the outgoing liquid creates additional ullage volume, and depending on excess air ingestion rate, a negative pressure could result that lowers vent pipe emissions. Dispensing fuel to customer vehicles and the associated Stage II vapor recovery system interact with vent emissions and can even cause vent emission during PWD, because the vacuum-assist method can negatively interfere with Onboard Refueling Vapor Recovery (ORVR) installed in customer vehicles (EPA, 2004). However, Stage II vapor recovery is not obsolete. It can be used in conjunction with ORVR to minimize exposure of gas station customers and workers to benzene due to working losses (Cruz-Nunez et al., 2003), particularly when customer vehicles are not equipped with ORVR (e.g., older vehi- cles, boats, motorcycles) or small volume gasoline containers are refueled. Enhanced Stage II vapor recovery technology can significantly reduce vapor emissions both at the nozzle and from UST vent pipes (CARB, 2013). 5.4. Fuel deliveries and accidental vent releases Based on observations and interpretation of time series of the tank pressure data, it is likely that the peak vent emissions (e.g.,Fig. 3b) were partly due to non-compliant bulk fuel drops where the Stage I vapor recovery system either was not correctly hooked up by the deliv- ery driver or to hardware problems with piping and/or valves. This Table 2 Summary of gas station characteristics and vent emissions. GS-MW GS-NW Units Sales volume _V sales 450,000 700,000 gal/month Volumetric flow rates (of gasoline vapor/air mixture) MeanQ 7.9 15.4 L/min Median (IQR) of 60-min average 6.1 (1.9, 10.9) 16.0 (12.7, 18.4) L/min Maximum of 60-min average 250 32.1 L/min Vent emission factor EFvent 1.4 1.7 lb/kgal Mass flow rates of gasoline (w/o air) Mean _mgas 0.39 0.76 kg/h Maximum of 60-min average 12.3 1.6 kg/h Correlation coefficient Between Q and p 0.58 0.85 – Table 3 Mean benzene emission rates _mbenz for the two gas stations. Emission source Benzene emissions (mg/s) Gas station GS-MW GS-NW Vent pipe 0.80 1.55 Spillage 0.39 0.65 Refueling 0.41 0.69 Hose permeation 0.06 0.10 Total 1.67 2.90 2245M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 conjecture is consistent with typical US storage tank volumes (~10,000 to 30,000 gal). Assuming that Phase I vapor recovery did not work at all and that 10,000 gal (~38,000 L) of fuel were delivered, the working loss (volume of gasoline vapor/air mixture released to the atmosphere through the vent pipe) is 38,000 L. It is also reasonable to assume that delivery lasted less than 1 h. Accordingto Table 2, the maximum hourly flow rate through thevent pipe was250 L/minatGS-MW, whichwould result in a maximum cumulative vapor release of 15,000 L within this hour. The measured maximum cumulative release underestimates the assumed working loss of 38,000 L. This could be due to a fuel delivery, which involved dropping fuel from multiple compartments of a tanker truck, with the vapor return hose not being correctly hooked up for only some of the emptied compartments. At GS-MW, UST pressure decreased after fuel delivery (causing vent emissions to cease for several hours) during the climatic conditions prevalent during the observation period, behavior not observed at GS- NW.Inpractice,itispossibletoobservebothpositiveandnegativepres- sureexcursions,evenduringthesamefueldelivery(whenmultiplefuel Fig. 5.Modeled maximum benzene concentrations for GS-MW and GS-NW at three different elevations z.Thex-andy-axes indicate horizontal coordinates in meters. The color indicates benzene levels inunits of μg/m3. Left column: time series of benzene emission rateswere used. Rightcolumn:average benzene emission rate was used in the modeling. The white isoline indicates OEHHA's acute REL of 26 μg/m3 = 8 ppb. 2246 M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 compartments of tanker trucks are unloaded), when Stage I vapor re- covery is in place (personal observation by TT). 5.5. Exceedance of 1-hour exposure limits AERMOD air pollution modeling suggests that at GS-MW the 1- hour acute REL was exceeded at one grid point 160 m (525 ft) from the center of the gas station once in 20 days (Fig. 5). This distance is larger than the 300-ft (91 m) setback distance recommended by CARB for a large gasoline dispensing facility (CalEPA/CARB, 2005). Assuming the gas station's fence line is b225 ft. (69 m) from its cen- ter (where the vent pipe was assumed to be located), our study shows that sensitive land uses at a distance further than 300 ft from the fence line of the gas station would represent a health con- cern despite compliance with the CARB guidelines because of non- compliance with the acute REL. Fig. 6.Modeled average benzene concentrations for GS-MW and GS-NW at three different elevations z.Thex-andy-axes indicate horizontal coordinates in meters. The color indicates benzene levels in μg/m3 and the white isoline the MRL of 19 μg/m3 =6ppb. 2247M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 At any location further than 50 m from the gas station's center, the REL was exceeded at most once during the 20-day measurement cam- paign (Fig. SI-1a). However, exceedance occurred at several locations, and on two different days (Fig. SI-1b). E.g., at a distance of 120 m from the center, the REL was exceeded at three grid points, and the number of grid points increased with closer proximity to the gas station. This suggests that it was not just a single worst-case scenario or a single ac- cidental vapor release that led to REL exceedance; rather exceedance may occur more frequently than is anticipated. Prevalent wind direc- tions during the measurement campaign explained the directional pat- terns of exceedances (see the wind rose in Fig. SI-3a). At GS-NW, despite its higher sales volume, the REL was exceeded only closer than 50 m from the gas station's center. However, exceed- ance occurred much more frequently (Fig. SI-2), likely because of the higher sales volume of GS-NW. Again, the wind rose for GS-NW (Fig. SI-3b) explains spatial patterns of REL exceedance. NoneofAIHA'sthree ERPGlevels were exceeded, meaningthat indi- viduals, except perhaps sensitive members of the public, would not have experienced more than mild, transient adverse health effects. 5.6. Average benzene levels The initial increase in average benzene levels when moving away from the gas stations' centers (Fig. 7) is likely due to the vent emissions (at 4 m)whichrepresentthelargestbenzenesource, andwhichrequire a certain transport distance until they reach the 2-m level through dis- persion.Furtherawayfromthegasstation,benzenelevelsarehigherfor GS-NWthanforGS-MWlikelybecauseofthehighersalesvolumeofGS- NW. However, close to the center, benzene levels are higher at GS-MW. This can beattributed to thehigher windspeeds at GS-NW (Table SI-1), which result in greater initial dilution of emitted pollutants in the in- coming airstream and also in greater subsequent pollutant dispersion. Modeled average benzene concentrations are generally lower (~10 μg/m3 or less) than those measured in the surroundings of gas stations, likely because our simulations do not account for traffic-related air pol- lution (TRAP). For instance, a study published by the Canadian petro- leum industry found average benzene concentrations of 146 and 461 ppb (466 and 1473 μg/m3) at the gas station property boundary in summer and winter, respectively (Akland, 1993), values orders of magnitudes higher than ours. A South Korean study examined outdoor and indoor benzene concentrations at numerous residences within 30 m and between 60 and 100 m of gas stations and found median out- door benzene concentrations of 9.9 and 6.0 μg/m3, respectively (Jo & Moon, 1999), while we simulated benzene levels on the order of 1 μg/ m3 (Fig. 7). In a study on atmospheric BTEX levels in an urban area in Iran, the three highest BTEX levels were measured near gas stations (~150 m away); the measured benzene levels (64 ± 36, 31 ± 28, 52 ±26μg/m3) were again much higher than ours simulated at that dis- tance, likely due to TRAP. Our modeled average benzene levels at a dis- tanceofabout50mareonthesameorderasbackgroundbenzenelevels of 1.0 μg/m3 that were measured in 2010 in the National Air Toxics Trend Sites (NATTS) network of 27 stations located in most major urban areas in the US (Strum & Scheffe, 2016). However, our modeled levels at a distance of 170 m were 0.07 at GS-MW and 0.12 at GS-NW, a non-negligible addition to urban background levels. At both gas stations, the MRL was exceeded at the level of the vent pipe opening in the vicinity of the gas stations, up to 7 m away from the vent pipe at GS-MW and 8 m at GS-NW. Therefore there might be an appreciable risk of adverse noncancer health effects for individuals living at the 2nd-floor level relatively close to high-volume gas stations such as GS-MW and GS-NW. 5.7. Limitations A limitation of our study is that data were collected only in fall and winter. Results cannot be easily extrapolated to other seasons, because ventpipeemissionsareseasonallydependent,e.g.,duetoseasonallyde- pendent gasoline formulations and meteorological conditions. How- ever, modeled exceedance of the OEHHA acute REL in the winter season is already of concern, because that REL was developed for once per month or less exposures. Another limitation is that we did not directly measure benzene levelsintheventpipe,andinsteadmadeassumptionsaboutvaporcom- position that were also made in the CAPCOA study (CAPCOA, 1997)of gas station emissions. In practice it may be difficulttoobtainpermission from gas station owners to measure benzene levels directly. In part because we did not want to reveal the locations of the gas stations, we did not use site-specific topography information in the air dispersionmodelingandinsteadassumed flatterrain.Whilethissimpli- ficationresultsinlessaccurateairpollutionpredictionsforthetwosites, usinga “generic”gasstationisperhapsmorerepresentativeofothergas stationsites,andisconsistentwithanapproachusedinapreviousstudy (CAPCOA, 1997). Finally, our study did not predict benzene levels in indoor environ- ments. Even though indoor air pollution levels may substantially differ fromoutdoorlevelsduetoindoorsources(e.g.,smoking,photocopying) (El-Hashemy & Ali, 2018), our study can still inform exposure levels in indoor environments as outdoor sources may be the main contributors toindoorairpollution,e.g.,inbuildingssituatedinurbanareasandclose to industrialzones orstreets withheavy traffic(Jones, 1999).This is rel- evant to workers and customers in C-stores or other fast-food/gasoline station combination facilities. 6. Conclusions Our study is to the best of our knowledge the first one to (1) report hourly vent emission data for gasoline storage tanks in the peer- reviewed literature and (2) use these data in hourly simulations of at- mospheric benzene vapor transport. This allowed us to examine poten- tial exceedance of short-term exposure limits for benzene. Prior studies including CAPCOA's (CAPCOA, 1997) could not do so as average emis- sion rates were used (only meteorological data was used at an hourly resolution). Our findingssupporttheneedtorevisitsetbackdistancesforgassta- tions, which are based on N2-decade old estimates of vent emissions (Aerovironment, 1994). Also, CARB setback distances are based on a bi- nary decision, related to whether the gasoline sales volume _V sales is N3.6 million gal per year. Our data support, however, that setback Fig. 7.Mean benzene concentrations as a function of distance from the center of the gas stations. 2248 M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 distances shouldbea continuous function ofsales volume _V sales and also include the type of controls installed at the facility. Setback distances should also address health outcomes other than cancer. OEHHA's acuteRELforbenzenecouldbeusedtoinformsetbackdistancesasitac- counts for non-cancer adverse health effects of benzene and its metab- olites (Budroe,2014). ATSDR's MRL could also be considered since it is a health-based limit. We note that CARB recommended their setback distances in 2005, presumably assuming pollution prevention technology yielding a 90% reduction in benzene emissions (CalEPA/CARB, 2005). Since then, CARB further promoted use of second-generation vapor recovery tech- nology (Enhanced Vapor Recovery, EVR) to reduce emissions further. EVR includes technology that is supposed to prevent fuel vapors in overpressurized tanks from being expelled into the atmosphere (CARB, 2017). To that end,“bladder tanks”have been proposed, into which the gasoline vapor/air mixture is directed as the pressure in the combined ullage space of the storage tank increases, and from which the mixture is redirected into the fuel storage tanks if the ullage pres- sure becomes negative (when fuel is dispensed). The challenge with suchasystemistoensurethatthebladdertankcapacityisnotexceeded by the fuel evaporation rate. Alternatively, fuel vapor release can be re- duced by processing the fuel/air mixture through either a semi- permeable membrane which selectively exhausts clean air and returns enriched fuel vapor (Semenova, 2004) or an activated carbon filter which adsorbs hydrocarbons (and water vapor) and exhausts air into the atmosphere, or by combusting the fuel/air mixture which would otherwise be released through the P/V valve. Therefore, current CARB setback distances might be adequate for gas stations in California but less so for the other 49 US states, and other countries—depending on pollution prevention technology requirements. The larger areal extentof modeled REL exceedance at GS-MW is due to “accidental”releases of gasoline vapors. Even though regulations ap- pear generally not to be driven by accidental releases, at GS-NW such releases likely led on two differentdaysto REL exceedances atdistances beyond CARB's recommended setback distances. Policies should ad- dressaccidentalfuelvaporreleasesthatdependingonpollutionpreven- tiontechnology(hereStageIvaporrecovery)anditsproperfunctioning can occur on a frequent basis (twice at GS-MW within about three weeks). Infuturework,potentialexceedanceofothershorter-termexposure limits should be examined, e.g., the 15-minute short-term exposure limits (STELs) and the 8-hour time-weighted averages (TWAs) used for occupational exposures. Acknowledgements This work was supported by NIH grant P30 ES009089 and the Envi- ronment, Energy, Sustainability and Health Institute at Johns Hopkins University. Competing financial interest declaration TT directs a company (ARID), which develops technologies for re- ducing fuel emissions from gasoline-handling operations. AMR, BAM and MH have no conflicts of interests to declare. Appendix A. Supplementary data Supplementary data tothis article canbefoundonlineat https://doi. org/10.1016/j.scitotenv.2018.09.303. References Adria-Mora, B., Hilpert, M., 2017. Differences in infiltration and evaporation of diesel and gasoline droplets spilled onto concrete pavement. Sustainability 9 (7).https://doi. org/10.3390/su9071271. Aerovironment, 1994.I. Underground Storage Tank Vent Line Emissions form Retail Gas- oline Outlets. Prepared for WSPA (AV-FR-92-01-204R2). AIHA, 2016.ERPG/WEEL Handbook. Current ERPG® Values (2016). American Industrial Hygiene Association, p. 2016. Akland, G.G., 1993.Exposure of the general population to gasoline. Environ. Health Perspect. 101 (Suppl. 6), 27–32 (Epub 1993/12/01. PubMed PMID: 8020446; PMCID: PMC1520004). Atabi, F., Mirzahosseini, S.A., 2013. GIS-based assessment of cancer risk due to benzene in Tehran ambient air. Int. J. Occup. Med. Environ. Health 26 (5), 770–779.https://doi. org/10.2478/s13382-013-0157-4 (Epub 2014/01/28, PubMed PMID: 24464541). ATSDR, 2004.Interaction Profile for: Benzene, Toluene, Ethylbenzene, and Xylenes (BTEX). Agency for Toxic Substances and Disease Registry. ATSDR, 2007.Toxicological Profile for Benzene. Agency for Toxic Substances and Disease Registry (CAS#: 71-43-2). ATSDR, 2018. Minimal Risk Levels (MRLs): Agency for Toxic Substances and Disease Reg- istry. Available from:https://www.atsdr.cdc.gov/mrls/index.asp (May 24, 2018). Brosselin, P., Rudant, J., Orsi, L., Leverger, G., Baruchel, A., Bertrand, Y., et al., 2009.Acute childhood leukaemia and residence next to petrol stations and automotive repair ga- rages: the ESCALE study (SFCE). Occup. Environ. Med. 66 (9), 598–606. Budroe,J.,2014.Noticeofadoptionofrevisedreferenceexposurelevelsforbenzene:Office of Environmental Health Hazard Assessment (California, US). Available from:https:// oehha.ca.gov/air/crnr/notice-adoption-revised-reference-exposure-levels-benzene. CalEPA/CARB, 2005.Air Quality and Land Use Handbook: A Community Health Perspec- tive: California Environmental Protection Agency & California Air Resources Board. CAPCOA, 1997.Gasoline Service Station Industrywide Risk Assessment Guidelines. Toxics Committee of the California Air Pollution Control Officers Association (CAPCOA). CARB, 2013.Revised Emission Factors for Gasoline Marketing Operations at California GasolineDispensingFacilities. California Air Resources Board,Monitoringand Labora- tory Division. CARB, 2017. Public workshop to discuss: overpressure conditions at gasoline dispensing facilities equipped with underground storage tanks and phase ii enhanced vapor re- covery including in-station diagnostic systems. December 12–13, 2017 Diamond Bar & Sacramento,CACaliforniaAir Resources Board.Availablefrom:https://www.arb.ca. gov/vapor/op/wrkshps/dec2017op_vr_pres.pdf. Cimorelli, A.J., Perry, S.G., Venkatram, A., Weil, J.C., Paine, R.J., Wilson, R.B., et al., 2005. AERMOD: a dispersion model for industrial source applications. Part I: general model formulation and boundary layer characterization. J. Appl. Meteorol. 44 (5), 682–693. Correa, S.M., Arbilla, G., Marques, M.R.C., Oliveira, K.M.P.G., 2012.The impact of BTEX emissions from gas stations into the atmosphere. Atmos. Pollut. Res. 3 (2), 163–169. Cruz, L., Alves, L., Santos, A., Esteves, M., Gomes, Í., Nunes, L., 2007.Assessment of BTEX concentrations in air ambient of gas stations using passive sampling and the health risks for workers. J. Environ. Prot. 8, 12–25. Cruz-Nunez, X., Hernandez-Solis, J.M., Ruiz-Suarez, L.G., 2003. Evaluation of vapor recov- ery systems efficiency and personal exposure in service stations in Mexico City. Sci. Total Environ. 309 (1–3), 59–68.https://doi.org/10.1016/s0048-9697(03)00048-2. Edokpolo, B., Yu, Q.J., Connell, D., 2014.Health risk assessment of ambient air concentra- tions of benzene, toluene and xylene (BTX) in service station environments. Int. J.Environ. Res. Public Health 11 (6), 6354–6374 (PubMed PMID: PMC4078583). Edokpolo, B., Yu, Q.J., Connell, D., 2015.Health risk characterization for exposure to ben- zene in service stations and petroleum refineries environments using human adverse response data. Toxicol. Rep. 2, 917–927. EIA, 2017. U.S. product supplied of finished motor gasoline: U.S. Energy Information Ad- ministration. Available from:http://www.eia.gov/dnav/pet/hist/LeafHandler.ashx? n=pet&s=mgfupus1&f=m. El-Hashemy, M.A., Ali, H.M., 2018.Characterization of BTEX group of VOCs and inhalation risks in indoor microenvironments at small enterprises. Sci. Total Environ. 645, 974–983. EPA, 2004.Stage II Vapor Recovery Systems Issues Paper. U.S. EPA. Office of Air Quality Planning and Standards. Emissions Monitoring and Analysis Division. Emissions Fac- tors and Policy Applications Group (D243-02). EPA, 2008.Transportation and marketing of petroleum liquids. Environmental Protection Agency. Petroleum Industry vol. I (Chapter V, AP 42). Hajizadeh, Y., Mokhtari, M., Faraji, M., Mohammadi, A., Nemati, S., Ghanbari, R., et al., 2018.Trends of BTEX in the central urban area of Iran: a preliminary study of photo- chemical ozone pollution and health risk assessment. Atmos. Pollut. Res. 9 (2), 220–229. Harrison,R.M.,Leung,P.L.,Somervaille,L.,Smith,R.,Gilman,E.,1999.Analysisof incidence of childhood cancer in the West Midlands of the United Kingdom in relation to prox- imity to main roads and petrol stations. Occup. Environ. Med. 56 (11), 774–780. Hilpert, M., Breysse, P.N., 2014.Infiltration and evaporation of small hydrocarbon spills at gas stations. J. Contam. Hydrol. 170, 39–52. Hilpert, M., Mora, B.A., Ni, J., Rule, A.M., Nachman, K.E., 2015. Hydrocarbon release during fuel storage and transfer at gas stations: environmental and health effects. Curr. En- viron. Health Rep. 2 (4), 412–422.https://doi.org/10.1007/s40572-015-0074-8. IARC, 2012. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. vol. 100F Available from:.http://monographs.iarc.fr/ENG/Monographs/vol100F/(Decem- ber 24, 2017). Infante, P.F., 2017.Residential proximity to gasoline stations and risk of childhood leuke- mia. Am. J. Epidemiol. 185 (1), 1–4. Jo, W.K., Moon, K.C., 1999. Housewives' exposure to volatile organic compounds relative to proximity to roadside service stations. Atmos. Environ. 33 (18), 2921–2928. https://doi.org/10.1016/s1352-2310(99)00097-7. Jo, W.K., Oh, J.W., 2001. Exposure to methyl tertiary butyl ether and benzene in close proximity to service stations. J. Air Waste Manage. Assoc. 51 (8), 1122–1128. https://doi.org/10.1080/10473289.2001.10464339. 2249M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 Jones, A.P., 1999.Indoor air quality and health. Atmos. Environ. 33 (28), 4535–4564. Karakitsios, S.P., Delis, V.K., Kassomenos, P.A., Pilidis, G.A., 2007.Contribution to ambient benzene concentrations in the vicinity of petrol stations: estimation of the associated health risk. Atmos. Environ. 41 (9), 1889–1902. McEntire, B.R., 2000.Performance of Balance Vapor Recovery Systems at Gasoline Dis- pensing Facilities. San Diego Air Pollution Control District. Morgester, J.J., Fricker, R.L., Jordan, G.H., 1992.Comparison of spill frequencies and amounts at vapor recovery and conventional service stations in California. J. Air Waste Manage. Assoc. 42 (3), 284–289. Semenova, S.I., 2004.Polymer membranes for hydrocarbon separation and removal. J. Membr. Sci. 231 (1–2), 189–207. Statistics Canada, 2009. Gasoline evaporative losses from retail gasoline outlets across Canada: environment accounts and statistics analytical and technical paper series. Available from:http://www.statcan.gc.ca/pub/16-001-m/2012015/part-partie1-eng. htm. Steffen, C., Auclerc, M.F., Auvrignon, A., Baruchel, A., Kebaili, K., Lambilliotte, A., et al., 2004. Acute childhood leukaemia and environmental exposure to potential sources of benzene and other hydrocarbons; a case-control study. Occup. Environ. Med. 61 (9), 773–778.https://doi.org/10.1136/oem.2003.010868. Strum,M.,Scheffe,R.,2016.Nationalreviewofambientairtoxicsobservations.J.AirWaste Manage. Assoc. 66 (2), 120–133.https://doi.org/10.1080/10962247.2015.1076538 (1995, PubMed PMID: 26230369, Epub 2015/08/01). Terres, I.M.M., Minarro, M.D., Ferradas, E.G., Caracena, A.B., Rico, J.B., 2010. Assessing the impact of petrol stations on their immediate surroundings. J. Environ. Manag. 91 (12), 2754–2762.https://doi.org/10.1016/j.jenvman.2010.08.009. WHO, 2010.WHO Guidelines for Indoor Air Quality: Selected Pollutants. World Health Organization, Geneva. Yerushalmi, L., Rastan, S., 2014.Evaporative losses from retail gasoline outlets and their potential impact on ambient and indoor air quality. In: Li, A., Zhu, Y., Li, Y. (Eds.), Pro- ceedings of the 8th International Symposium on Heating, Ventilation and Air Condi- tioning. Indoor and Outdoor Environment Vol. 1. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 13–21. 2250 M. Hilpert et al. / Science of the Total Environment 650 (2019) 2239–2250 2020 Study Gasoline Vapor Emissions During Vehicle Refueling Events in a Vehicle Fleet Saturated With Onboard Refueling Vapor Recovery Systems - Need for an Exposure Assessment BRIEF RESEARCH REPORT published:07 February 2020 doi: 10.3389/fpubh.2020.00018 Frontiers in Public Health | www.frontiersin.org 1 February 2020 | Volume 8 | Article 18 Edited by: Efstratios Vogiannis, National Observatory of Athens, Greece Reviewed by: Yue-Wern Huang, Missouri University of Science and Technology, United States Rich Baldauf, United States Environmental Protection Agency, United States *Correspondence: Jenni A. Shearston js5431@cumc.columbia.edu Specialty section: This article was submitted to Environmental Health, a section of the journal Frontiers in Public Health Received:30 August 2019 Accepted:20 January 2020 Published:07 February 2020 Citation: Shearston JA and Hilpert M (2020) Gasoline Vapor Emissions During Vehicle Refueling Events in a Vehicle Fleet Saturated With Onboard Refueling Vapor Recovery Systems: Need for an Exposure Assessment. Front. Public Health 8:18. doi: 10.3389/fpubh.2020.00018 Gasoline Vapor Emissions During Vehicle Refueling Events in a Vehicle Fleet Saturated With Onboard Refueling Vapor Recovery Systems: Need for an Exposure Assessment Jenni A. Shearston *and Markus Hilpert Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, NY, United States Background:Gasoline contains large proportions of harmful chemicals, which can be released during vehicle refueling. Onboard Refueling Vapor Recovery (ORVR) can reduce these emissions, but there is limited research on the system’s efficacy over time in an actual vehicle fleet. The aims of this study are: (1) determine the feasibility of using an infrared camera to view vapor emissions from refueling; (2) examine the magnitude of refueling-related emissions in an ORVR-saturated fleet, to determine need for an exposure-assessment. Methods:Using an infrared camera optimized for optical gas imaging of volatile organic chemicals, refueling was recorded for 16 vehicles at six gas stations. Pumps were inspectedfordamage,refuelingshut-offvalvefunctioning,andpresenceofStageIIVapor Recovery. Vehicle make/model and age were recorded or estimated. Results:Vapor emissions were observed for 14 of 16 vehicles at each station, with severity varying substantially by vehicle make/model and age. Use of an infrared camera allowed for identification of vapor sources and timing of release, and for visualizing vapor trajectories. Discussion:Notably emissions occurred not only at the beginning and end of refueling but also throughout, in contrast to a prior study which did not detect increases in atmospheric hydrocarbon levels mid-refueling. Future studies are vitally needed to determinetherisktoindividualsduringtypicalrefuelinginanORVRsaturatedvehiclefleet. We recommend comprehensive exposure-assessment including real-time monitoring of emitted volatile organic compounds paired with infrared gas-imaging and measurement of internal dose and health effects of gas station customers. Keywords: gasoline, environmental exposure, vehicle refueling, volatile organic compounds, gas station Shearston and Hilpert Gasoline Vapor During Vehicle Refueling INTRODUCTION Gasoline is a complex mixture of many chemicals, several of which are known to adversely aect human health. Of particular concern are volatile aromatic hydrocarbons, including benzene, toluene, ethylbenzene, and xylene (BTEX group), which may be released during vehicle refueling (1,2). For example,benzeneisaknownhumancarcinogenandisassociated with multiple health problems, including respiratory, nervous system, and immunological conditions (3). In addition, studies evaluating non-cancer outcomes have found decreased red blood cell counts, hemoglobin, and hematocrit levels in gas station attendants (4). While some studies have evaluated exposures to gasoline from vehicle refueling specifically (5–7), to our knowledge, few have been completed in the past decade. It is essential that such studies are repeated frequently and in varied geographic locations, as fuel composition, weather, climate, and pollution control strategies all impact individual exposures and can change over time. In the United States (US), changes in regulations outlining gasoline vapor recovery during vehicle refueling have made this an especially pressing question. During refueling, gasoline vapor in a vehicle’s tank is pushed into the atmosphere by the rising liquid gasoline level in the tank—unless a vapor recovery system is in place. From 1998 to 2006, the US Environmental Protection Agency (EPA) rolled out a requirement that nearly all newlymanufacturedvehiclesbeequippedwithonboardrefueling vapor recovery (ORVR) systems (8), which function by directing vaporized gasoline into a canister on the vehicle, thereby substantially reducing escape of vapors into the atmosphere. Briefly, this requirement was rolled out in stages, first for light duty vehicles (1998: 40% of new vehicles, 1999: 80%, 2000: 100%), then for light duty trucks and vans (2001: 40%, 2002: 80%, 2003: 100%), and finally for heavier light duty trucks (2004: 40%, 2005: 80%, 2006: 100%) and trucks with a >10,000 pounds gross vehicle weight rating (100% by 2006). By 2006, nearly all new gas-powered vehicles with <14,000 pound gross vehicle weight rating were required to have ORVR systems (8). In contrast, Stage II vapor recovery systems, which are used on gasoline pumps themselves, direct vaporized gasoline into gas station underground storage tanks through systems on the pumps. In 2012, the EPA determined that the US vehicle fleet was suciently saturated with ORVR that states could allow the removal of Stage II systems (8), thus making vapor recovery during refueling primarily dependent on ORVR systems. Despite this change in regulations, limited information on the eciency of ORVR systems is available, although the US EPA suggests they are 98% ecient and require minimal maintenance(8).AGermanstudyfoundnomeasurableincreases in atmospheric hydrocarbon concentrations in a Sealed Housing for Emissions Determination (SHED) in which an ORVR- equipped vehicle was placed during refueling, although increases were detected at the beginning and end of refueling (9). Even though a study of presumably non-ORVR equipped vehicles in Mexico found older vehicles to have more evaporative emissions than newer ones (10), to the best of our knowledge, no assessment of the continuous functioning of ORVR systems to reduce emissions during vehicle refueling over the course of a vehicle’s lifetime, within the conditions of an actual vehicle fleet, has been completed. It is possible that as vehicles age, hoses, seals, and other parts of the gas tank and ORVR system degrade, resulting in increased vapor emissions during refueling. Additionally, while some studies (6,7) evaluated exposure to gasoline vapors during vehicle refueling in the US, finding evidence of benzene in blood and exhaled breath samples, those studies were completed before saturation of the US vehicle fleet with ORVR systems, and are thus likely over-estimates of exposures that may occur with ORVR systems. It is not currently known whether the amount of vapors today’s population is exposed to would have similar, if any, eects. Past studies assessing exposure from vehicle refueling used aluminum tubes as passive samplers (7) and sorbent tubes attached to pumps (6) to quantify exposure to gasoline vapors, positioned in the breathing zone of participants. However, such methods may not be able to detect the lower levels of exposure anticipatedfromavehiclefleetwitha98%ecientORVRsystem. Additionally, while these methods quantify environmental exposure to vapors during refueling, they are not easily used for source identification or to capture the dispersion and movement of vapors at the station. It is also not possible to use these devices to determine when during a refueling event vapors are more likely to be released (i.e., at the end vs. throughout), information which can help determine the cause of vapor release. Use of other technologies, such as an infrared camera optimized for visualizing compounds present in petroleum products, is needed to determine the sources of vapors during refueling (i.e., from exhaust, the vehicle tank, or the pump nozzle) and how they move through space. Such cameras are also fine-tuned to detect very small amounts of vapors, and thus may be invaluable in determining if exposure to gasoline vapors is occurring from ORVR equipped vehicles, warranting a more involved exposure-assessment. Research on the functioning of ORVR in the actual US vehicle fleet over time, and thus an understanding of the quantity of vapors individuals may still be exposed to, is limited. Additionally, the tools traditionally used to assess exposure to vapors during vehicle refueling do not give a complete picture, as they lack the ability to determine vapor sources and movement. With this pilot study, we aim to determine the plausibility and usefulness of conducting a full exposure-assessment for exposures to gasoline vapors during vehicle refueling, in a vehicle fleet dependent on ORVR for vapor recovery. The objectives of this pilot study are to (1) determine the feasibility of qualitatively capturing fuel vapor emissions from vehicle refueling events in New York City (NYC) using a FLIR infrared camera designed specifically to detect volatile organic compounds present in petroleum products, and to (2) examine the magnitude of fuel vapor emissions over a range of dierent vehicle/ORVR system ages as a precursor to assessing the continuous functioning of ORVR systems over the lifetime of a vehicle in the actual US vehicle fleet. Frontiers in Public Health | www.frontiersin.org 2 February 2020 | Volume 8 | Article 18 Shearston and Hilpert Gasoline Vapor During Vehicle Refueling MATERIALS AND METHODS Study Overview A convenience sample of gas stations in Northern Manhattan, NYC, was selected for vapor release monitoring. At each gas station, a study member approached individuals just before they began refueling their vehicles and asked for verbal permission to recordtheirvehicletanksasthevehiclewasrefueled.Thisstudyis nothumansubjectsresearch,asnoinformationaboutindividuals was obtained, and is thus not subject to IRB oversight. A total of six gas stations were visited over the course of a single winter day. Three vehicle refueling events were recorded at eachstation,withtheexceptionofonestationwhereanattendant was present. For this station, only one vehicle refueling event was recorded. In total,n =16 refueling events were recorded. Data Collection An infrared camera optimized for optical gas imaging of volatile organic chemicals (FLIR model GF320; described below) and frequently used to detect leaks in petroleum refining operations, was used to record the fuel pump nozzle and external vehicle fuel tank filler pipe during each refueling session. In addition, researchers visually inspected gasoline pumps for hose damage, refueling shut-o valve functioning, and presence of Stage II Vapor Recovery systems. Researchers recorded the make and model of the vehicle when it was visible on the outside of the automobile, while year was estimated using photographs of the vehicle. Year was estimated by searching for images of the vehicle make and model, and comparing dierent years, especially the front and rear bumpers and headlight shape, to those shown in the photographs. When researchers could not definitively determine the year of the vehicle, the midpoint of the plausible year range was used. Vehicles were assigned a type based on the EPA Vehicle Classification system. Overview of FLIR Infrared Camera The FLIR model GF320 infrared camera can detect 20 gases, including: 1-pentene, benzene, butane, ethane, ethanol, ethylbenzene, ethylene, heptane, hexane, isoprene, m-xylene, methane,methanol,methylethylketone,MIBK,octane,pentane, propane, propylene, and toluene (FLIR Systems Inc., 2017). The camera is tuned to detect very small spectral ranges, so that it can selectively visualize specific compounds that absorb or emit electromagnetic energy at that spectral range. A narrow bandpass filter is used to ensure that only gases with a strong signal in the specified infrared range are detected, and other components of the camera are built to emit very little energy, to reduce the signal-to-noise ratio. The manufacturer does not provide estimates of limits of detection of their camera, but we found that the GF320 can detect quite small vapor leakage rates, e.g., gas emissions from an unignited pocket lighter in outdoor atmosphericenvironmentsimagedfromadistanceofatleast2m. Qualitative and Statistical Analysis To determine how representative our convenience sample is of New York State and New York City vehicle fleet ORVR saturation, we used New York State’s publicly available Vehicle, Snowmobile, and Boat Registrations database to calculate the proportion of registered vehicles in both the state and city that were gasoline powered and manufactured in 2006 or later (out of all gasoline powered vehicles), the year the EPA suggests that “essentially all” new gas-powered vehicles <14,000 pounds were manufactured with ORVR systems (8). We compared this to the proportion of ORVR equipped vehicles in our sample. In addition, we compared the median vehicle manufacturing age in our sample to that of registered vehicles in New York State and City. Each infrared video was reviewed to identify the presence and magnitude of vaporized gasoline emitted during a refueling session. An overall qualitative description of each video was created, and patterns of vapor emission were identified and assigned to each session. Vapor origin (i.e., ambient vapors vs. vapors from the vehicle fuel tank) and the timing of vapor release was reviewed in all sessions. Representative video frames of “typical” emissions for each vehicle were extracted from the middle and end of each refueling session. The vapor plume was delineated using the brush feature in Microsoft Paint based on repeated observations of the videos, and not just a single frame, as it is dicult to identify the plume from a static image. Exploratory statistical analysis was conducted in R version 3.5.1 (11). A logistic model was fit to obtain an association between estimated vehicle age and presence of vapor release duringthemiddleofvehiclerefueling,operationalizedasabinary variable.Duetothesmallsamplesizenocovariateswereincluded in the model. Figures were created with the tidyverse package in R (12), as well as with Inkscape (www.inkscape.org) and MATLAB (The MathWorks Inc., 2010). RESULTS A total of 16 refueling events at six gas stations were recorded. Our convenience sample was fairly representative of the estimated ORVR penetration proportion in New York State and City vehicles: according to EPA regulations 94% of our sample should have been equipped with ORVR, while for both New York State and City, we estimate that at least 81% of registered vehicles should have been equipped with ORVR. The median manufacturing year of our sample was 2013, the same as that for New York State and City. Table 1 provides details about gas stations and vehicles. Of the six stations, only one had a Stage II vapor recovery system, and fourhadliquidgasolineleakingaroundthehosejoints.Estimated vehicleagerangedfrom1to32years(manufacturingyears1987– 2018), and several vehicle types (e.g., SUV, mid-size car) were represented in the sample. For 15 out of 16 vehicles, vehicle age and type combination indicated they were required to contain ORVR systems. The average refueling length was 86s. Ambient temperature ranged from 33 to 41◦F (0.5–5◦C). The infrared camera was able to detect gasoline vapors during vehiclerefueling.Inaddition,evaluationofthevideofilesallowed researchers to identify vapor sources, pinpoint the time of vapor release during each video, and to see how the vapors moved after being emitted. Frontiers in Public Health | www.frontiersin.org 3 February 2020 | Volume 8 | Article 18 Shearston and Hilpert Gasoline Vapor During Vehicle Refueling TABLE 1 |Characteristics of gas stations and vehicle refueling events. Gas station ID Stage II vapor recovery system Hose joints Vehicle ID EPA vehicle size classification Estimated model year ORVR mandate* Length of refueling (s) 2 None No leakage 29 Minicompact car 2014 Yes 66 30 Midsize car 2005 Yes 88 32 Standard sport utility vehicle 2013 Yes 88 3 None Leakage 33 Midsize car 2006 Yes 76 34 Mid-size car 2018 Yes 78 35 Small sport utility vehicle 2013 Yes 84 4 None Leakage 36 Mid-size car 2008 Yes 131 37 Standard sport utility vehicle 2018 Yes 133 38 Standard sport utility vehicle 2015 Yes 71 8 Vacuum assist Leakage 41 Compact car 2005 Yes 72 42 Midsize car 2016 Yes 122 43 Midsize car 2008 Yes 66 9 None Leakage 44 Standard sport utility vehicle 2004 Yes 56 45 Large car 1987 No 110 46 Midsize car 2015 Yes 106 7 None No leakage 47 Minivan 2013 Yes 32 *Indicates whether 100% of new vehicles were required to have included ORVR systems for the specific manufacturing year and vehicle type (i.e., light duty vehicle, light duty truck, and van, heavier light duty trucks, etc.). Fuel vapor emissions were observed for 14 out of 16 vehicles and at every gas station. The single vehicle older than ORVR manufacturing mandates in the US clearly had much larger refueling vapor emissions than the newer vehicles. However, the majority of newer vehicles also had substantial fuel vapor emissions, particularly at the end of refueling. Qualitative descriptions of each refueling event are provided in Table 2. Six overall patterns of vapor emission were identified: no vapor release (one vehicle), ambient vapors only (one vehicle), release toward the end of refueling (two vehicles), release when nozzle was withdrawn (three vehicles), release toward the end of refueling and after nozzle was withdrawn (six vehicles), and near continuous vapor release (three vehicles).Figure 1 shows the number of vehicles in each category, and the years of the vehicles’ manufacture. The three vehicles with near continuous vapor release were estimated to be 5, 11, and 32 years old. Of note, all vehicles that emitted vapors at any point during the refueling session also did so at the end of the refueling session. Representative video frames from the middle and end of each refueling session are available in the Supplementary Material (two frames per vehicle). In Figure 2, examples from each of the six vapor emission patterns are shown, with gasoline vapor plumes delineated in blue in each frame, and vehicle IDs in the top right corner. For example, for the “release when nozzle withdrawn” category, the representative screenshot during the middle of the refueling session does not show any vapors, however, at the end of the session, vapors can be seen spilling out around the pump nozzle and the vehicle fuel tank opening. The range of emission magnitude can be seen from the various sample frames. Full video recordings for each refueling event are available at the following link: https://github.com/jenni- shearston/Vehicle_Refueling_Videos. Results from the exploratory logistic regression were not significant, as there were not enough observations to detect an association (n =16; yes release [n =3]/no release [n = 13]). The model suggested that a 1 year increase in estimated vehicle age was associated with a 1.15 increase in likelihood of emitting vapors during the middle of refueling (95% CI =0.97, 1.51), but this result is likely driven by the results for the 32 years old vehicle, which was much older than the rest of the vehicle population. DISCUSSION This work highlights the value of using an infrared camera to compliment more traditional methods of exposure measurement for determining potential health risks from vehicle refueling, and visually highlights the sometimes large amounts of fuel vapor emissionsthatoccurevenwithinanORVRsaturatedvehiclefleet. A FLIR camera allowed us to identify the source of the vapors; for example, in one video (Vehicle ID 44) vapors can be seen, but they do not originate from the pump nozzle or the vehicle tank. Of note, we observed leaking gasoline around the hose joints at this station (Station 9). For all other videos, vapors are clearly seen coming out of the pump nozzle, vehicle tank, or both. This allows for the dierentiation of sources of vapor exposure, crucial information needed to intervene on exposures at gas stations generally, or to determine how eective ORVR is at minimizing vapor outflow. In addition, use of the infrared camera allowed us to confirm that vapors were emitted in a location where an individual filling up their gas tank might breathe them in (the “breathing zone”), and to visualize the dispersion and movement of the vapors. The infrared camera also made it possible to pinpoint when during a refueling session Frontiers in Public Health | www.frontiersin.org 4 February 2020 | Volume 8 | Article 18 Shearston and Hilpert Gasoline Vapor During Vehicle Refueling TABLE 2 |Qualitative description and overall patterns of vehicle refueling events. Vehicle ID Qualitative description Overall pattern 29 Some gasoline vapor can be seen escaping into the atmosphere from the beginning of the refueling event, continuing throughout the duration of refueling. At around 0:00:41, a larger amount of vapor is seen escaping from the vehicle tank, generally increasing in amount until the end of the refueling session Near continuous vapor release 30 No vapors are seen escaping into the atmosphere until more than a minute of refueling has passed (0:01:13), after which a large amount of vapor escapes as the vehicle tank presumably reaches full Release toward end of refueling 32 Minimal vapor was released into the atmosphere throughout the duration of the refueling event. At the very end of refueling, as the pump is removed from the tank, a small amount of vapor can be seen escaping Release toward end of refueling and after nozzle withdrawn 33 No vapors are seen escaping from the vehicle tank until the end of refueling, around 0:01:13, after which a large amount of vapor escapes, presumably as the tank reaches full. After the pump is withdrawn from the tank, fuel vapor continues to escape into the atmosphere in substantial quantities Release toward end of refueling and after nozzle withdrawn 34 No vapor is seen escaping until the end of the refueling session, around 0:01:11, after which a substantial amount of fuel escapes into the atmosphere, continuing to escape even after the pump is withdrawn from the vehicle Release toward end of refueling and after nozzle withdrawn 35 No vapor is seen escaping from the vehicle tank until the end of refueling. Vapors escape when the pump handle is partially withdrawn (0:01:12) and the tank is presumably topped off, and continue to escape even after the pump is fully withdrawn Release toward end of refueling and after nozzle withdrawn 36 Although the pump is inserted into the vehicle from the beginning of the video, it appears that fuel is not dispensed until around 0:00:43 when the individual’s hand squeezes the pump handle. As dispensing begins, large amounts of vapors can be seen escaping from the tank. Of note, the individual refueling does not fully insert the pump into the tank. Vapors escape nearly continuously throughout refueling, sometimes in large amounts. Toward the end of the session another large amount of vapor escapes, as the pump is pulled further out of the vehicle (0:01:55). Substantial amounts of vapor continue to escape until the end of refueling, including after the pump is fully withdrawn (0:02:49) Near continuous vapor release 37 No vapor release observed No vapor release 38 No vapor is observed until around 0:00:51, after which vapor is released nearly continuously. Vapor is observed escaping from the tank after the pump is withdrawn Release toward end of refueling and after nozzle withdrawn 41 Some vapor is released at the beginning of the refueling session (0:00:14), but no more is observed until toward the end of refueling around (0:01:08). After this time, vapor is observed in substantial quantities until the pump is withdrawn (0:01:21), after which only minimal vapors are observed escaping Release toward end of refueling 42 No vapors are observed until the very end of refueling, when the pump is withdrawn (0:01:59). Vapor continues to be released from the tank until it is capped Release when nozzle withdrawn 43 No vapor release observed during refueling; a small amount of vapor may be released after pump is withdrawn (0:01:08) Release when nozzle withdrawn 44 Poor video focus makes vapor observation difficult; however, ambient vapors appear to be present (upper right, 0:00:35, 0:00:40, 0:00:54) Ambient vapors only 45 Substantial vapor release observed as cap is removed from tank, and continuously throughout refueling Near continuous vapor release 46 No vapor release observed during refueling; a slight amount of release from pump observed as it was removed from tank (0:01:57) Release when nozzle withdrawn 47 Slight amount of vapor release observed at start of refueling (0:00:03), and then again at end of refueling (0:00:24). Vapor continues to be released after pump removed Release toward end of refueling and after nozzle withdrawn vapors were released. Sorbent tubes attached to pumps, passive samplers, and real-time monitors are not able to do this because the amount of vapor measured is averaged over a time period, so it is challenging to determine when the vapor is released, or if it is released continuously. Information about the timing of vapor releases is particularly useful because it can help researchers determine why vapors are being released. For example, ORVR systems with “liquid seals” are known to release some vapors at the end of refueling (13), because as the flow of gasoline into the vehicle tank decreases, the air gradient into the tank created by the moving gasoline decreases, allowing vapors to flow both into the tank and out of it (and thus into the atmosphere) (9). Release at the end of vehicle refueling was indeed one of our most common observations. However, vapor releases occurring in the middle of therefuelingsession,orthroughoutthesession,bothofwhichwe observed in multiple refueling events, may suggest a breakdown in functioning of the ORVR system. These findings appear to be inconsistent with the ones by Tumbrink who did not observe measurableemissionsduringrefueling(9).RenandHaoinChina did find measurable emissions throughout refueling, but at low levels,withvaporconcentrationincreasingovertimeandranging from 0 to 4.5 mg/m3 (13). Emissions could be the result of a leak in part of the vehicle’s fuel system, aging of the activation sites or oversaturation of the charcoal filter used in the ORVR, or a malfunctioning mechanical seal. It is also possible that that Frontiers in Public Health | www.frontiersin.org 5 February 2020 | Volume 8 | Article 18 Shearston and Hilpert Gasoline Vapor During Vehicle Refueling the pump nozzle itself is damaged, resulting in vapor release. In addition, Ren and Hao found that ambient temperature, fuel temperature, filling flow, and filling pipe diameter all have FIGURE 1 |Dotplot depicting the number of vehicles in each vapor release category (each dot represents one vehicle), with year delineated by color. an impact on the time to liquid seal formation and on vapor emissions (13). Emissions were increased when either ambient or fuel temperature was higher (13). As our study was conducted at cold ambient temperatures (0.5–5◦C), we expect that emissions during Spring, Summer, and Fall would be greater than what we observed. Our study found an average refueling time of 86s (1.43min), similar to the 1.13min found by Vainiotalo et al. (5) in Finland and less than that found by Egeghy et al. (7) in North Carolina (median of 3min). These studies, and others, included various biomarkers and measures of exposure: internal dose (blood) (6), exhaled breath (7), and breathing zone air (5–7), all of which suggested individuals were exposed to benzene, a known human carcinogen, during refueling. As all studies were conducted before widespread adoption of ORVR and only at gas stations without Stage II vapor recovery, their results are likely not representative of the typical exposure today. Somewhat concerningly, however, our study suggests that despite extensive use of ORVR, individual exposures at similar magnitudes to those experienced before ORVR requirements were implemented may still occur—two of the three refueling FIGURE 2 |Two sample frames for each of the 6 identified patterns of vapor release during refueling: one during the middle of the refueling session, and one at the end. Vehicle ID and an indicator for middle (“Mid”) or end (“End”) of the video are included in the upper right corner of each photo. Gasoline vapors, when present, are outlined by a blue line. Frontiers in Public Health | www.frontiersin.org 6 February 2020 | Volume 8 | Article 18 Shearston and Hilpert Gasoline Vapor During Vehicle Refueling events categorized as “near continuous vapor release” happened in vehicles manufactured after the rollout of ORVR. Without Stage II vapor recovery, the population is not protected from emissions arising from the so-called legacy fleet without ORVR, vehicleswithdeterioratingORVR,ormotorcyclesandboats,both of which do not have ORVR. Of particular importance for public health and policy is the ability of ORVR systems to (1) reduce exposure to gasoline vapors during refueling to a safe level, and (2) continue to function at a high level over the lifetime of a vehicle. This is important for two reasons. First, volatile organic compounds (VOCs) released during refueling can chemically react in the atmosphere,contributingtoozoneandothersecondarypollutant formation, which can harm human health directly through cardiovascular pathways (14). ORVR systems are intended to reduce this potential, by preventing VOCs from escaping into the atmosphere where they can react with other species. Second, as previously discussed, exposure to primary VOCs, such as those in gasoline can also negatively impact health directly, from exposure during vehicle refueling. However, limited work has been conducted to test the assumption that ORVR reduces exposure to a “safe” level during vehicle refueling. In fact, it is unclear what a “safe” level of exposure to gasoline vapors is, particularly as there is not a standardized formula for gasoline. Numerous studies have been conducted (15,16) to characterize the potential harms of gasoline with specific formulas or additives, but these reports typically compare dierent formulas of gasoline rather than comparing exposure to no exposure. Evidence suggests that while exposure during refueling is likely, health eects from gasoline at infrequent low-levels may be small, although individual components are carcinogenic (15,16). Conversely, evidence from occupational studies has shown that individuals chronically exposed to lower levels of gasoline vapors, for example gas station attendants, are at higher risk for certain cancers (17,18). Despite this evidence, we do not fully understand what risk gasoline vapors pose to the general public during typical vehicle refueling, or the cumulative impact of such exposure over an individual’s lifetime,particularly in today’s regulatory environment. Our findings highlight, in a visually compelling manner, that individuals can be exposed to substantial amounts of gasoline vapors during refueling, even in a vehicle fleet saturated with ORVR. Future studies are vitally needed to determine the risk to individuals during typical refueling sessions in a vehicle fleet saturated with ORVR, especially because exposure to gasoline is ubiquitous and occurs throughout the lifetime. We recommend comprehensive exposure assessments that estimate exposure, internal dose, and health eects, as well as real-time monitoring ofvolatileorganiccompounds,potentiallyusingaportableSHED (19) deployed at a gas station and paired with an infrared camera optimized for gas imaging. In addition, we recommend future work to develop an algorithm for estimating the amount or concentration of vapors shown in video from an infrared camera, to provide a better understanding of the concentration of vapors dispersing around a station. This pilot study has several limitations. First, a convenience sample of stations and vehicles were used, and thus may not be representative of the true vehicle fleet in NYC. However, ORVR saturation in our sample was fairly close to an estimate for all registered vehicles in New York State and City (94 vs. 81%). It is additionally reassuring that both these estimates are above the EPA estimate of 71% for ORVR saturation in the older 2012 US fleet (8) and that the saturation in our convenience sample is above New York State’s modeled estimate of 85% or greater for the older 2013 fleet (20). The median manufacturing year of our sample was consistent with that for New York State and City’s registered vehicles (median =2013). Second, the small sample size does not provide ample power for statistical tests. Third, vehicle make, model, and age were estimated by researchers and thereforethereispotentialformisclassification.Finally,real-time estimates of VOC concentrations were not obtained. CONCLUSIONS In an ORVR saturated vehicle fleet, use of an infrared camera optimized for VOC imaging allowed for the identification of vapor sources, viewing vapor trajectory and dispersion, and identifying the timing of vapor release during refueling. In this pilot study, 14 out of 16 observed refueling events resulted in vapor emissions, with severity varying substantially by vehicle make/model and age. A full exposure-assessment incorporating infrared cameras, quantitative monitors, and biologic samples is needed to understand exposure to and health eects of fuel vapor at gas stations, in an ORVR saturated vehicle fleet. DATA AVAILABILITY STATEMENT All datasets generated for this study are included in the article/Supplementary Material. AUTHOR CONTRIBUTIONS MH and JS conceptualized the study and completed data collection. JS wrote the first manuscript draft and completed initial data analysis. MH supervised and reviewed all the data analysis and edited the manuscript. All authors agree to be accountable for the content of this work. FUNDING MH was supported by NIEHS grant P30 ES009089. JS was supported by NIEHS grant T32 ES007322. The funding sources had no involvement in study design, collection, analysis, and interpretation of data, manuscript development, or publication. ACKNOWLEDGMENTS The authors would like to thank Mark Fishburn of FLIR Systems for facilitating the use of the FLIR GF320 infrared camera for a short test period and Sebastian Rowland for his help reviewing the manuscript. SUPPLEMENTARY MATERIAL The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpubh. 2020.00018/full#supplementary-material Frontiers in Public Health | www.frontiersin.org 7 February 2020 | Volume 8 | Article 18 Shearston and Hilpert Gasoline Vapor During Vehicle Refueling REFERENCES 1.Hilpert M, Mora BA, Ni J, Rule AM, Nachman KE. Hydrocarbon release during fuel storage and transfer at gas stations: environmental and health eects.J Curr Environ Health Rep.(2015) 2:412–22. doi: 10.1007/s40572-015-0074-8 2. Hilpert M, Rule AM, Adria-Mora B, Tiberi T. Vent pipe emissions from storage tanks at gas stations: implications for setback distances.Sci Total Environ.(2019) 650:2239–50. doi: 10.1016/j.scitotenv.2018.09.303 3. ATSDR.Toxicological Profile for Benzene.Agency for Toxic Substances and Disease Registry. Atlanta, GA: US Department of Health and Human Services (2007). 4. Abou-ElWafa HS, Albadry AA, El-Gilany AH, Bazeed FB. Some biochemical and hematological parameters among petrol station attendants: a comparative study.Biomed Res Int. (2015) 2015:418724. doi: 10.1155/2015/ 418724 5. Vainiotalo S, Peltonen Y, Ruonakangas A, Pfäi P. Customer exposure to MTBE, TAME, C6 alkyl methyl ethers, and benzene during gasoline refueling.Environ Health Perspect.(1999) 107:133–40. doi: 10.1289/ehp.991 07133 6. Backer LC, Egeland GM, Ashley DL, Lawryk NJ, Weisel CP, White MC, et al. Exposure to regular gasoline and ethanol oxyfuel during refueling in Alaska. Environ Health Perspect.(1997) 105:850–5. doi: 10.1289/ehp.97105850 7. Egeghy PP, Tornero-Velez R, Rappaport SM. Environmental and biological monitoring of benzene during self-service automobile refueling.Environ Health Perspect.(2000) 108:1195–202. doi: 10.1289/ehp.001081195 8. EPA. 40 CFR Part 51: Air Quality: Widespread Use for Onboard Refueling Vapor Recovery and Stage II Waiver.Environmental Protection Agency. Federal Register. (2012) 77:28772–82. 9. Tumbrink, M.Filtersysteme im Automobil: Innovative Lösungsansätze für die Automobilindustrie [Filtration Systems in the Car: Innovative Solution Approaches for Car Manufacturers]: Haus de Technik–Fachbuchreihe (Tübingen) (2002). 10. Schifter I, Díaz L, Rodríguez R, González-Macías C. The contribution of evaporative emissions from gasoline vehicles to the volatiel organic compound inventory in Mexico City.Environ Monit Assess.(2014) 186:3969– 83. doi: 10.1007/s10661-014-3672-2 11. R Core Team.R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing (2018). 12. Wickham H.Tidyverse: Easily Install and Load the ‘Tidyverse’. R package version 1.2.1 (2017). 13. He R, Ding H. Refueling experiment of on-board refueling vapor recovery. China J Highw Transp.(2017) 30:142–50. 14. Zhao R, Chen S, Wang W, Huang J, Wang K, Liu L, et al. The impact of short-term exposure to air pollutants on the onset of out-of-hospital cardiac arrest: a systematic review and meta-analysis.Int J Cardiol.(2017) 226:110–7. doi: 10.1016/j.ijcard.2016.10.053 15. ATSDR.Toxicological Profile for Gasoline. Agency for Toxic Substances and Disease Registry. Atlanta, GA: US Department of Health and Human Services (1995). 16. NSCAUMATC.EvaluationoftheHealthEectsFromExposuretoGasolineand Gasoline Vapors. Northeast States for Coordinated Air Use Management Air Toxics Committee (1989). 17. Morton W, Marjanovic D. Leukemia incidence by occupation in the Portland-Vancouver metropolitan area.Am J Ind Med.(1984) 6:185–205. doi: 10.1002/ajim.4700060304 18. Schwartz E. Proportionate mortality ratio analysis of automobile mechanics andgasolineservicestationworkersinNewHampshire.AmJIndMed.(1987) 12:91–9. doi: 10.1002/ajim.4700120110 19. Eastern Research Group.Denver Summer 2008 Pilot Study at Lipan Street Station–Report Version 5. Assessment and Standards Division, Oce of Transportation and Air Quality and United States Environmental Protection Agency (2008). 20. NYDEC.Stage II Vapor Collection System Enforcement Discretion Directive. New York, NY: State Department of Environmental Conservation (2011). Available online at: https://www.dec.ny.gov/regulations/74990.html (accessed November 30, 2019). Conflict of Interest:The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Copyright © 2020 Shearston and Hilpert. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Frontiers in Public Health | www.frontiersin.org 8 February 2020 | Volume 8 | Article 18