Analysis of the Air Quality Impacts of
the Use of Ethanol in Gasoline
September 28, 1999
(Revised September 29, 1999)
Authors
Paul Allen, P.E.
Richard Bradley
Bart E. Croes, P.E.
John DaMassa
Robert Effa
Mark Fuentes
Annette Hebert
Dongmin Luo, Ph.D.
Richard Vincent
Luis Woodhouse, Ph.D.
Eugene Yang, Ph.D.
California Air Resources Board
2020 L Street
Sacramento, California 95814
iii
Table of Contents
I. Background 1
II. Automotive Emissions 1
A. VOC Speciation Profiles 1
B. Peer Review 3
C. Mass Emissions 3
D. Vehicle Emission Testing 4
III. Photochemical Modeling 4
A. Model Description 4
B. Results 4
IV. Ambient Air Quality 6
A. Findings of Previous Studies 6
1. Denver, Colorado 6
2. Albuquerque, New Mexico 6
3. Brazil 6
B. Current and Future Air Quality 7
1. CO, NO2, Ozone, PM10, PM2.5 7
2. Acetaldehyde, Benzene, 1,3Butadiene, and Formaldehyde 8
3. Ethanol and MTBE 8
4. Alkylates 8
5. PAN and PPN 9
6. Isobutene, Toluene, Xylene Isomers, nHexane 9
V. Recommendations for Additional Studies 11
VI. References 11
Appendix 1A Development of Emission Profiles for CaRFG w/o MTBE
Appendix 1B Organic Gas Profiles
Appendix 2 Photochemical Modeling
Appendix 3 Emission Inventories
Appendix 4 Literature Review
Appendix 5 Methods to Establish Baseline Air Quality Levels and Estimate Future Air Quality
Appendix 6 Ambient Levels of Peroxyacetyl Nitrate in Southern California
Appendix 7 Vehicle Emission Testing and Gasoline Headspace
Analysis Protocols
iii
I. Background
Governor Gray Davis issued Executive Order D599 on March 25, 1999 calling for the removal of methyl tertiarybutyl ether (MTBE) from gasoline at the earliest possible date, but not later than December 31, 2002. Task 10 of the Executive Order states “the California Air Resources Board (ARB) and the State Water Resources Control Board (SWRCB) shall conduct an environmental fate and transport analysis of ethanol in air, surface water, and groundwater. The Office of Environmental Health Hazard Assessment (OEHHA) shall prepare an analysis of the health risks of ethanol in gasoline, the products of incomplete combustion of ethanol in gasoline, and any resulting secondary transformation products. These reports are to be peer reviewed and presented to the Environmental Policy Council by December 31, 1999 for its consideration.”
This report has been prepared in response to the Executive Order D-5-99. To assist OEHHA in its risk assessment, we conducted an analysis to estimate the changes in ambient air concentrations of potentially detrimental contaminants of exhaust and evaporative components and subsequent reaction products that would result from substituting ethanolblended gasoline for gasoline blended with MTBE. We also included nonoxygenated gasoline in our analysis to provide some basis for comparison. The following sections summarize our estimates of volatile organic compound (VOC) emission profiles and emission inventories, modeling of air quality impacts, and data analysis of current and future air quality concentrations. Seven appendices contain detailed information. We evaluated emission and air quality impacts for the following four fuels:
¨ Current MTBEbased California Phase 2 Reformulated Gasoline (CaRFG).
¨ Ethanolbased fully complying fuel (with oxygen content of 3.5wt%).
¨ Ethanolbased fully complying fuel (with oxygen content of 2.0wt%).
¨ A nonoxygenated fully complying fuel.
We focused our analysis on the following air contaminants:
¨ Criteria air pollutants [carbon monoxide (CO), nitrogen dioxide (NO2), ozone, and particulate matter (PM10, PM2.5)].
¨ Toxic air contaminants (acetaldehyde, benzene, 1,3butadiene, and formaldehyde).
¨ Fuel oxygenates (ethanol and MTBE).
¨ Alkylates [branched alkanes such as 2methylpentane, 3methylpentane, methylcyclopentane, and 2,2,4trimethylpentane (“isooctane”)].
¨ Peroxyacetyl nitrate (PAN) and peroxypropionyl nitrate (PPN).
¨ Nitric acid (HNO3).
¨ Additional compounds of interest to OEHHA (isobutene, toluene, xylene isomers, and nhexane).
II. Automotive Emissions
A. VOC Speciation Profiles
We developed VOC speciation profiles for each fuel for the following processes and applied to all gasolinerelated emission inventory categories (e.g., passenger cars, motorcycles, heavyduty vehicles, fuel spillage, etc.):
¨ Start exhaust catalyst and noncatalyst.
¨ Hot stabilized exhaust catalyst and noncatalyst.
¨ Diurnal and resting evaporative.
¨ Hot soak and running evaporative.
¨ Liquid fuel.
For MTBEblended CaRFG, we used VOC speciation profiles previously developed from ARB surveillance data and presented at a public workshop in September 1998.
We reviewed available data on VOC speciation from existing vehicle emission test programs to identify those that best represent the three fully complying nonMTBE fuels. These formed the basis for adjustments that can be applied to the VOC speciation profiles for the CaRFG base fuel. We developed the adjustments by comparing an emission profile for MTBEblended CaRFG with a profile for an ethanolblended or oxygenfree test fuel from the same study that was similar in hydrocarbon composition. The ethanolblended test fuels were made with the same hydrocarbon bases as were the MTBEblended fuels. However, to meet the Reid Vapor Pressure (RVP) limit in the CaRFG regulations, refiners that blend commercial fuels with ethanol will usually use modified hydrocarbon bases; most likely pentane content will decline and alkylates content will increase. (Aromatics and olefins will be constrained by the Predictive Model.) Such changes will not involve highly reactive species; so, data from splashblended test fuels rather than commercial fuels should be adequate here with regard to reactivity.
Two following studies provided emission data contrasting MTBEblended CaRFG with CaRFG blended with ethanol:
¨ The recent ARB testing of an MTBEblended CaRFG and a fuel with high RVP and 10% ethanol.
¨ A test program sponsored by ARB in 1995, “Effects on Exhaust and Evaporative Emissions of Phase 1 and Phase 2 Gasolines”, by Automotive Testing Laboratories (ATL).
In addition, “Auto/Oil” Technical Bulletin 17A provided the data for adjustments to approximate emissions from at least one type of oxygenfree CaRFG. In addition, we input properties of ethanolblended and oxygenfree CaRFGspredicted in a recent linearprogram modeling study by MathPro sponsored by the California Energy Commissioninto the ARB’s Predictive Model for exhaust emissions of benzene and 1,3butadiene and into newly created models for aldehyde emissions and evaporative benzene emissions.
In general, within each emission study, the VOC speciation profiles for the MTBEblended test fuel are similar to those for the ethanolblended or oxygenfree test fuel. In most cases, the only significant differences are the interchange (or removal of) the oxygenate and, for exhaust profiles, the interchange of the major partial combustion products of the oxygenates (e.g., formaldehyde and isobutene from MTBE, acetaldehyde from ethanol). The nonoxygenated fuel predicted by MathPro has a lower aromatic content than does the MTBEblended CaRFG; the octane replacement for MTBE is provided by increased blending of alkylates. These changes result in lower ozoneforming potential, as calculated using the maximum incremental reactivities developed by Carter (1994). However, the exhaust profiles for nonoxygenated fuel have higher contents of C8 and higher aromatic species and higher ozoneforming potential because all nonMTBErelated species were predicted to increase when MTBE is removed from the fuel. Based on vehicle emission test data, this would not be expected. Also, when removing MTBE, sulfur would have to be reduced. Lower sulfur would result in more efficient catalysts and lower exhaust hydrocarbon emissions and generally less reactive emissions. This issue is being further investigated from the perspective of both a further evaluation of engineering principles and a review of test data. This apparent discrepancy will be partially addressed by the gasoline headspace and vehicle emission testing described in Section II.D.
The profiles from the MTBEblended test fuels are usually similar (in some cases, identical) to the current ARB profiles adopted last year. Therefore, the differences between profiles within the test studies can be applied with some confidence to adjust the VOC speciation profile for the CaRFG base fuel. However, each study referenced earlier has limitations that complicate its use. The Auto/Oil work did not measure extended diurnal emissions. None of the studies used noncatalyst vehicles. The evaporative data from ARB’s MTBEethanol test program have excessively high nbutane, due to the manner in which the carbon canisters were prepared. The ATL data do not include alcohols or aldehydes, which are the most important contrasting species between emissions from MTBE and ethanolblended fuels. (However, surrogate aldehyde data are available.) As splashblended test fuels, the ethanolblended fuels do not exactly reflect commercial fuels. Also, they were not true CaRFGs. (The ethanolblended fuels did not meet the RVP limit at 7 psi. Also, they did not completely satisfy the Predictive Model. In particular, the ARB ethanolblended fuel had a high oxygen content that caused a high NOX prediction.) Finally, the ATL work did not include speciation of the gasolines, so that its emission profiles cannot be related to its gasoline compositions.
Although not perfect representations, the compositions of the comparison fuels used in this analysis generally conform to statements from refiners about how they would have to change their gasolines to meet the CaRFG regulations with ethanol. The changes include significant removal of pentanes and an increased use of alkylates.
B. Peer Review
Professor Robert Harley of the University of California at Berkeley peer reviewed the VOC speciation profiles, and calculated headspace vapors from the liquid fuel speciation profiles as a check on the ones developed inhouse. Professor Harley suggested that the headspace evaporative VOC emission profiles (used to represent diurnal evaporative emissions) developed by ARB staff for the two ethanolblended gasolines (2.0 and 3.5wt% oxygen) may be too high in ethanol emissions and, as a result, too low in emissions of other species. Also, he suggested that the liquid gasoline composition be used as an alternate representation for hot soak vapors. The liquid gasoline has much higher alkane content than ARB's hot soak emissions. This results in lower content of all other gasoline components, especially toluene and ethanol. Thus, the overall result of using Professor Harley's profiles are to reduce the amount of ethanol emissions. We implemented these recommendations by performing photochemical grid model simulations for both the ARB evaporative profiles and those recommended by Professor Harley. His recommended profiles are designated with a trailing “H” in subsequent tables. This approach serves to bracket the range of uncertainty in the ethanol content of the VOC speciation profiles.
C. Mass Emissions
We estimated total mass emissions of VOC, oxides of nitrogen (NOX), and CO for CaRFG using the current mobile source emissions model, MVEI7G. Stationary source emissions were assumed to be the same for all scenarios. For the three fully complying nonMTBE fuels, the Predictive Model constrains the total mass emissions of VOC and NOX.
Based on several vehicle emission test programs and ambient air studies (Dolislager, 1997), CO emissions decrease with increasing fuel oxygen content. For the purpose of this evaluation, we used the same motor vehicle CO inventory for the MTBE and ethanol fuel scenarios with 2% oxygen content, and increased CO emissions by 5% for the non-oxygenated scenario and decreased CO emissions by 15% for the ethanol 3.5% oxygen scenario. The CO emission inventory has a small impact on concentrations of ozone and other secondary pollutants. Since CO is not considered a toxic compound in the context of this study, the CO concentrations simulated for each scenario are not part of any further analysis.
It should be noted that a substantial portion of the CO reductions that can be attributed to CaFRG comes from properties of the fuel other than the oxygen content. This means that addition or removal of oxygen in CaRFG is likely to have less impact on CO emission than its use in non-RFG wintertime blends.
The resulting emission inventories for the South Coast Air Basin (SoCAB) are shown in Tables 1 to 3 for CO, NOX, reactive organic gases (ROG), benzene, 1,3butadiene, acetaldehyde, formaldehyde, ethanol, and MTBE. Table 1 summarizes the total emissions (mobile, area, stationary, and natural sources) for an average ozone episode day for the seven fuel scenarios. Table 2 and 3 present emission changes relative to the 1987 and 2003 baselines, respectively.
D. Vehicle Emission Testing
In order to determine if the VOC speciation profiles for the three fully complying nonMTBE fuels are reasonable, we are conducting complete exhaust VOC speciation tests for multiple vehicles using two prototype nonMTBE regulargrade gasolines currently being produced and sold commercially in California. One is an nonoxygenated gasoline produced by Chevron and the other is a gasoline blended by Tosco with ethanol at a 2.05wt% oxygen content. Both fuels fully comply with California's current CaRFG regulations. We are also speciating the liquid fuels and headspace vapors. We anticipate that fullscale production of fully complying nonMTBE gasoline will result in blends different from those being produced today. However, we expect the results from these tests to yield valuable insights into the directional effects of using ethanolblended and nonoxygenated gasolines. This will provide a reality check for the profiles prepared inhouse and by Professor Harley using limited test data.
III. Photochemical Modeling
A. Model Description
We applied the Urban Airshed Model with the Flexible Chemical Mechanism interface (UAMFCM) for the August 2628, 1987 ozone episode in the SoCAB. Input files for winds, temperature, and diffusion break were developed using special air quality and meteorological data collected during the 1987 Southern California Air Quality Study (Lawson 1990). We simulated initial and boundary conditions, together with emission inventories for calendar years 1997 and 2003 using the meteorology from the 1987 episode. Fixing the meteorological conditions in this way allows the effects of fuel changes to be directly calculated. We used an extended version of the SAPRC97 photochemical mechanism (Carter et al., 1997) to simulate atmospheric chemical transformations. The mechanism includes explicit treatments of the chemistry for the criteria pollutants, acetaldehyde, benzene, 1,3butadiene, ethanol, formaldehyde, HNO3, MTBE, PAN, and PPN. The mechanism tracks secondary formation of acetaldehyde and formaldehyde separately from the contribution of direct emissions. SAPRC97 lumps the alkylates, isobutene, toluene, xylene isomers, and nhexane with similarly reactive compounds, so we estimated upperlimit concentrations based on analysis of ambient air quality measurements (see Section IV.B.6). OEHHA determined that the upperlimit concentrations for isobutene, toluene, xylene isomers, and nhexane were an order of magnitude below any level of concern, so there was no need to treat these compounds explicitly in the photochemical modeling. OEHHA is still evaluating levels of concern for the alkylates.