EPA 600/R-05/032
February 2005
Predicted Ground Water, Soil and Soil Gas Impacts
from US Gasolines, 2004
First Analysis of the Autumnal Data
by
James W. Weaver
Ecosystems Research Division
National Exposure Research Laboratory
Athens, Georgia 30605
Lewis Jordan and Daniel B. Hall
Department of Statistics
University of Georgia
Athens, Georgia 30602
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
-------
Notice
The U.S. Environmental Protection Agency through its Office of Research and Development
funded and managed the research described here. It has been subjected to the Agency's peer and
administrative review and has been approved for publication as an EPA document. Mention of
trade names or commercial products does not constitute endorsement or recommendation for use.
Abstract
Ninety Six gasoline samples were collected from around the U.S. in Autumn 2004. The
samples included regular and premium grade fuels, conventional and reformulated gasolines,
high and low elevation samples, and fuel subject to state regulations (especially bans of methyl
tert-butyl ether). A detailed hydrocarbon analysis was performed on each sample resulting in
data set of approximately 300 chemicals per sample. Comparisons were made between several
significant individual parameters. The results showed that benzene was, as required by the Clean
Air Act, below 1 percent in areas that require reformulated gasoline (RFG). Higher benzene
levels were found in some samples outside these areas. Methyl tert-butyl ether (MTBE) was
found in use in RFG areas where there were no state bans in place. Where MTBE was banned,
ethanol was found as the replacement oxygenate. Statistical analyses were performed on the
entire suite of reported chemicals. These analyses were used to determine which gasolines were
similar and if characteristics of the gasolines (conventional/reformulated, grade, elevation, and
MTBE ban status) could be used to segregate the fuels into a decision tree. The statistical
analyses showed that the gasolines were separable by these factors, but that differences due to
state regulations or variability in the fuels make certain gasolines dissimilar from gasolines of
similar characteristics. This feature was evident in the premium grade gasolines, presumably
because refiners use different approaches to boost the fuel's octane number. Estimated
solubilities and vapor pressures of the majority of the chemicals were used to predict the
effective solubilities and gas phase concentrations of each chemical in a gasoline. Three types of
low elevation, regular grade gasolines were included: conventional, MTBE-RFG and ethanol
RFG. Pair-wise comparisons of these fuels showed predicted shifts in aqueous and gas phase
concentrations.
11
-------
Acknowledgments
The authors gratefully acknowledge the contributions of:
Hal White, US EPA, Office of Underground Storage Tanks;
Dr. Patricia Ellis, Delaware Department of Natural Resources and Environmental
Conservation;
Dr. Michel Boufadel, Temple University, Civil Engineering;
Chris Englehart and Joe Haas, New York State Department of Environmental
Conservation;
Dr. John T. Wilson, Eric Kleiner, and Fran Kremer, U.S. EPA, Office of Research and
Development;
Jen Traub and Dr. Randall J. Charbeneau, The University of Texas at Austin, Department
of Civil and Environmental Engineering;
Matt Jones, Levi Hein, Brad Thornhill, and Dr. Ken Rainwater, Texas Tech University,
Department of Civil and Environmental Engineering;
Jeff Kuhn, Montana Department of Environmental Quality;
Chris Dufex and Mahesh Albuquerque, Colorado Division of Oil and Public Safety;
Andy Hess and Wally Moon, U.S. EPA Region 10, Seattle Washington;
Mark Restaino U.S. EPA Region 5, Chicago, Illinois;
Greg Lovato, Dr. Matthew Small, and Steve Linder, U.S. EPA Region 9, San Francisco,
California;
Randall S. Weaver, Statham, Georgia;
Susan Swanson, The Cascade Group, Streetsboro, Ohio
Mike Birke and Robert Legg Southwest Research Institute, San Antonio, Texas
Drs. Fred Tillman, and Jay Choi, National Research Council, Athens, Georgia;
Dr. Drew Ekman, Dr. Mike Cyterski, Carlyn Haley, Pam Gunter, Dr. Said Hillal,
Dr. Susan Richardson, Tim Rowan, David Spidle, and Dr Paul Winget, U.S. EPA
Office of Research and Development, Athens, Georgia,
Sue Davis of U.S. EPA Office of Research and Development, Research Triangle Park,
North Carolina, and
Rachel Claussen of U.S. EPA, Research Triangle Park, North Carolina.
Special thanks to Brandy Manders of U.S. EPA, Office of Research and Development, Athens,
Georgia
in
-------
Foreword
The National Exposure Research Laboratory's Ecosystems Research Division (ERD) in Athens,
Georgia, conducts research on organic and inorganic chemicals, greenhouse gas biogeochemical
cycles, and land use perturbations that create direct and indirect, chemical and non-chemical
stresses, exposures, and potential risks to humans and ecosystems. ERD develops, tests, applies
and provides technical support for exposure and ecosystem response models used for assessing
and managing risks to humans and ecosystems, within a watershed / regional context.
The Regulatory Support Branch (RSB) conducts problem-driven and applied research, develops
technology tools, and provides technical support to customer Program and Regional Offices,
States, Municipalities, and Tribes. Models are distributed and supported via the EPA Center for
Exposure Assessment Modeling (CEAM) and through access to Internet tools
(www.epa.gov/athens/onsite).
ERD undertook a nationwide study of gasoline composition to generate insight and input data for
risk assessment models, such as the Hydrocarbon Spill Screening Model (Weaver et al., 1994)
and the OnSite on-line calculators (Weaver, 2004). Further, from the nationwide scope of the
study, with its intent to include important geographic and regulatory zones, a set of relationships
defining typical gasolines has been developed. This report describes results from the first
analysis of data collected in the autumn of 2004. Future publications will present results from
later sampling and more comprehensive analyses.
Rosemarie C. Russo, Ph.D.
Director
Ecosystems Research Division
Athens, Georgia
IV
-------
Leaking Underground Storage Tank Assessment Report Series
A series of research reports is planned to present data and models for leaking underground
storage tank risk assessments. To date these include:
1. Gasoline Composition
Weaver, James W., Lewis Jordan and Daniel B. Hall, 2005, Predicted Ground Water, Soil and
Soil Gas Impacts from US Gasolines, 2004: First Analysis of the Autumnal Data, United
States Environmental Protection Agency, Washington, D.C., EPA/600/R-05/032.
2. Simulation Models
Weaver, James W., 2004, On-line Tools for Assessing Petroleum Releases, United States
Environmental Protection Agency, Washington, D.C., EPA 600/R-04/101.
As more reports are added to the series, they may be found on EPA's web site at:
http://www.epa.gov/athens/publications.
-------
Table of Contents
Notice ii
Abstract ii
Acknowledgments iii
Foreword iv
Leaking Underground Storage Tank Assessment Report Series v
List of Figures 3
List of Tables 5
Abbreviations and Acronyms 7
Introduction 8
Literature and Purpose 9
Purpose 11
Approach and Methods 12
Results 18
Octane Number 18
Benzene 19
Methyl Tert-Butyl Ether 20
Tert-Butyl Alcohol 22
Tert-Amyl Methyl Ether 23
Oxygen Content 24
Ethanol 27
Brand 29
Similarities and Differences Among Gasolines 30
Cluster Analysis 30
Principal Components Analysis 41
Discriminant Analysis 49
Representative Gasolines 50
Cluster Analysis 52
-------
Predicted Environmental Impacts of the Study Fuels on Water and Air 53
Predicted Aqueous Concentrations 54
Predicted Gas Phase Concentrations 63
Discussion and Conclusions 74
References 76
Appendix: Theoretical Basis of the Principal Components Analysis 79
-------
List of Figures
Figure 1 Benzene data for RFG and conventional gasoline locales 19
Figure 2 MTBE and benzene contents for all fuels 20
Figure 3 MTBE content plotted against octane number 21
Figure 4 Tert-Butyl Alcohol (TEA) content plotted against MTBE content, showing only small
amounts of TEA present in these fuels 22
Figure 5 Tert-Amyl Metyl Ether content plotted against octane number 23
Figure 6 Composite weight of oxygen from all oxygenates and benzene concentrations 24
Figure 7 Fuel oxygen content plotted against octane number for all samples 26
Figure 8 Ethanol content plotted against octane number 27
Figure 9 Ethanol content plotted against MTBE content showing the mutual exclusive nature of
the usage of these chemicals 27
Figure 10 Distribution of benzene by brand 29
Figure 11 Dendrogram showing the cluster analysis' separation of 93 octane (left) from 87
(right) octane Georgia gasolines 31
Figure 12 Dendrogram showing relationship between regular and premium gasolines from
conventional gasoline areas in Georgia, Ohio and Oklahoma 34
Figure 13 Dendrogram showing relationship between high elevation areas in Lubbock, Texas
and Helena, Billings and Great Falls, Montana 36
Figure 14 Dendrogram showing relationships between conventional gasolines from Georgia and
RFG gasolines from Virginia 36
Figure 15 Analysis of three RFG areas without MTBE bans 37
Figure 16 Dendrogram of RFG gasolines from an MTBE-ban state (New York) and two states
without bans 38
Figure 17 Dendrogram of MTBE-ban states with State or Federal RFG requirements 39
Figure 18 Dendrogram showing relationship between two MTBE ban, RFG gasoline areas. . 41
Figure 19 Plot of the cumulative proportion of variation explained by PC number 42
Figure 20 Scree graph of the variance estimate by PC number for the individual gasoline
component data 43
Figure 21 Principal Components plot of the 96 gasoline samples with respect to the first two
principal components labeled by state 46
Figure 22 Principal Components plot of the 96 gasoline samples with respect to the first two
principal components labeled by octane number 46
Figure 23 Plot of the 96 gasoline samples with respect to the first two principal components by
gasoline brand 48
Figure 24 Decision tree for gasolines typical of the data collected in this study. The parenthesis
indicate where the cluster analysis indicated that samples from the various states are
similar enough to group together 50
Figure 25 Cluster analysis performed on the representative gasolines 52
Figure 26 Effective solubility of hydrocarbon groups and oxygenates for the average
conventional, low-elevation, no-MTBE-ban, regular gasolines 55
Figure 27 Effective solubility of hydrocarbon groups and oxygenates for the average
reformulated, low-elevation, no-MTBE-ban regular gasolines 56
-------
Figure 28 Effective Solubility for hydrocarbon groups and oxygenates for the Illinois regular
gasolines (low elevation, MTBE ban, RFG) 59
Figure 29 Gas phase concentration of hydrocarbon groups and oxygenates for the average
conventional, low-elevation, no-MTBE-ban regular gasolines 63
Figure 30 Gas phase concentration of hydrocarbon groups and oxygenates for the average
reformulated, low-elevation, no-MTBE-ban regular gasolines 64
Figure 31 Gas phase concentration for hydrocarbon groups and oxygenates for the Illinois
regular gasolines (low elevation, MTBE ban, RFG) 68
Figure 32 Scatter plot of predicted gas phase concentration versus the effective solubility of the
average conventional, low-elevation, no-MTBE-ban, regular gasoline 72
-------
List of Tables
Table 1 Geographic locations, characteristics, and numbers of samples included in the study.
Elevations less than 2500 ft are judged "low." Fuel requirements 13
Table 2 ASTM D 6729-1 known pairs of co-eluting peaks 15
Table 3 Percent by weight of unidentified compounds in analytical results 16
Table 4 Specified chemicals included in the analysis 16
Table 5 Hydrocarbon groups and numbers of member compounds 16
Table 6 Octane ratings of sampled gasolines 18
Table 7 Oxygen content of each oxygenated additive, based upon the number of carbon,
hydrogen and oxygen atoms per mole of compound 24
Table 8 Cluster centroids in weight % for Georgia 87 octane and 93 octane gasolines defined
by all components present at 0.50 wt % or greater 31
Table 9 Compounds that were present in greater abundance in Georgia premium than regular
grades by at least 0.50 wt % 33
Table 10 Compounds that were present in lesser abundance in Georgia premium gasolines than
in regular grades by at least 0.50 wt % 33
Table 11 Degree of similarity where all fuels of a given grade join in cluster analyses of
individual locales (states) 35
Table 12 Statistics of the first eight PC from the PCA of the individual gasoline components
including variance, proportion of the total variance estimate, and the cumulative
proportion of variation explained 42
Table 13 Gasoline components with the largest loadings in the first eight principal components
including class, mean, standard deviation, and the rank of the standard deviation (from
largest (1) to smallest(8)) 44
Table 14 Discriminant analysis results 49
Table 15 Predicted higher aqueous constituent concentrations (>0.05 mg/L) of RFG (regular
grade, no-MTBE-ban, low-elevation) in comparison to conventional gasoline 57
Table 16 Predicted lower aqueous constituent concentrations (>0.05 mg/L) of RFG (regular
grade, no MTBE ban, low elevation) in comparison to conventional gasoline 59
Table 17 Predicted higher aqueous constituent concentrations (>0.05 mg/L) of MTBE-ban
RFG (regular grade, low elevation) in comparison to non-MTBE ban RFG gasoline. . . 60
Table 18 Predicted lower aqueous constituent concentrations (>0.05 mg/L) of MTBE-ban RFG
(regular grade, low elevation) in comparison to non-MTBE ban RFG gasoline 61
Table 19 Predicted higher gas phase constituent concentrations (>0.05 mg/L) of RFG (regular
grade, low elevation) in comparison to conventional gasoline 65
Table 20 (Continuation) Predicted higher gas phase constituent concentrations (>0.05 mg/L)
of RFG (regular grade, low elevation) in comparison to conventional gasoline 67
Table 21 Predicted lower gas phase constituent concentrations (>0.05 mg/L) of RFG (regular
grade, low elevation) in comparison to conventional gasoline 67
Table 22 Predicted higher gas phase constituent concentrations (>0.05 mg/L) of MTBE-ban
RFG (regular grade, low elevation) in comparison to non MTBE-ban RFG 69
Table 23 Predicted lower gas phase constituent concentrations (>0.05 mg/L) of MTBE-ban
RFG (regular grade, low elevation) in comparison to non MTBE-ban RFG 70
-------
Table 24 (Continuation) Predicted lower gas phase constituent concentrations (>0.05 mg/L) of
MTBE-ban RFG (regular grade, low elevation) in comparison to non MTBE-ban RFG.
72
-------
Abbreviations and Acronyms
BTEX Benzene, Toluene, Ethylbenzene, Xylenes
CAAA Clean Air Act Amendments
CG Conventional gasoline
DHA Detailed hydrocarbon analysis
DIPE Di-Isopropyl Ether
DA Discriminant Analysis
ETBE Ethyl Tert-Butyl Ether
GC-MS Gas chromatography/mass spectroscopy
HCA Hierarchical Cluster Analysis
IUPAC International Union of Pure and Applied Chemistry
LUST Leaking underground storage tank
MTBE Methyl Tert-Butyl Ether
M/Z Mass to charge ratio
ON Octane number
PC Principal Component
PCA Principal Components Analysis
RFG Reformulated Gasoline
SPARC SPARC Performs Automated Reasoning in Chemistry
TAEE Tert Amyl-Ethyl Ether
TAME Tert Amyl-Methyl Ether
TAA Tert Amyl Alcohol
TEA Tert Butyl Alcohol
WO Winter oxygenate
-------
Introduction
Gasoline consists of numerous petroleum-derived chemicals and additives. At a
fundamental level the composition determines the physical and operational properties of the fuel.
Gasoline as a product, though, is defined in terms of these properties that include among others
the octane number and the vapor pressure. The concentration of certain individual components
remains important though as they may be regulated or mandated. The most notable of these are
benzene and oxygenated additives. As a potential source of environmental contamination,
however, the major consideration is the composition. Composition data are relatively scarce as
historically the greatest concerns for chemical contamination of ground water at leaking
underground storage tank (LUST) sites have been placed on benzene, methyl tert-butyl ether
(MTBE) and other aromatics (toluene, ethylbenzene and the xylenes1). The logic of this
approach appears to follow from several factors: the status of benzene as a carcinogen, the low
taste and odor threshold of MTBE, the relatively high concentrations in fuel of BTEX and, in
some cases MTBE, and the solubilities and volatilities of these chemicals that allow them to
form ground water plumes and/or vapor clouds in the subsurface.
A more fully multicomponent approach may be called for in certain circumstances. The
non-MTBE oxygenates are becoming more prominent because of some states' bans on MTBE
use, while the Federal oxygenate mandate remains in place. Ethyl Tert-Butyl Ether (ETBE),
Tert Amyl-Methyl Ether (TAME), Tert Amyl Ethyl Ether (TAEE) and Di-Isopropyl Ether
(DIPE) are possible ether oxygenates. Ethanol and Tert Butyl Alcohol (TEA) are possible
alcohol oxygenates. As shifts are made away from MTBE as the dominant oxygenate, other
changes in gasoline composition may occur. This might result in differing levels of various
other contaminants in soil gas or ground water. Knowledge of shifts in composition can provide
decision-makers with foreknowledge of potential impacts. A second reason for a
multicomponent approach is that biodegradation is generally agreed upon to be electron-acceptor
limited. Thus the pool of electron acceptors is available only for a finite mass of contaminants at
most sites. The loading of all species contributes to usage of electron acceptors and may need to
be considered. Multicomponent analysis may be useful for distinguishing among different
gasolines (Stout et al., 2003) as part of a environmental forensics investigation. A forth reason
for studying multicomponent gasoline composition is that detailed characterization of gasolines
may provide a means to predict the types of potential impacts that may occur. This need results
from the very nature of leaking underground storage tanks: leaks occur unseen and undetected
for years in many cases. It is clearly impossible to go back in time and measure the composition
of the leaked fuel. Knowledge of composition is needed to model or estimate the risks
associated with the release.
Benzene, toluene, ethylbenzene and xylene, taken together, are known as BTEX.
-------
Literature and Purpose
The Clean Air Act Amendments (CAAA) of 1990 required the US EPA to develop
oxygenated fuel regulations for cities that failed to meet carbon monoxide ambient air standards.
Beginning in 1992 gasolines sold in these areas were required to contain a minimum of 2.7%
oxygen by weight (Strikkers, 2002). Reformulated gasolines (RFG) which contain oxygenates
and meet other requirements were also mandated by the CAAA for certain ozone non-attainment
areas after January 1, 1995. EPA regulations require that the benzene content of reformulated
fuels be 1.0 % by volume or less when determined on a per gallon basis, or 0.95 % volume or
less on average with a 1.3 % volume maximum per gallon (40 CFR 80, Subpart 41). The oxygen
content must be 2.0 % by weight or more when determined on a per gallon basis, or 2.1 %
weight or more on average with a 1.5 % by weight minimum per gallon (40 CFR 80 Subpart 41).
A so-called anti-dumping provision mandates compliance in areas using conventional gasoline
with baseline gasoline composition and properties based on 1990 gasoline production. State
requirements, including MTBE bans are given in state laws for California (California Air
Resources Board, 2003), Colorado (Colorado, 2002), and Washington (Washington State, 1996).
Data on gasoline composition have been collected by several groups. The Canadian
Petroleum Products Institute (1994) performed a survey of Canadian gasolines during winter and
summer of 1993. Their analysis included 44 compounds, which represented a fairly high
fraction of most of the samples. Their study was divided by Canadian provinces, which gives
the study an East/West geographical separation, but they did not report specific octane number
nor elevation of the regular, mid and premium grade samples included in the 128 samples per
season. Only two of the samples contained a significant amount of MTBE.
Other, less intensive, sources of data include the generalized composition reported by
Gustafson et al. (1997), and Environment Canada (Jokuty et al., 1999), which contains six
examples of detailed chemical composition of gasoline among 450 petroleum products and crude
oils in total. Some states perform and publish surveys of gasoline composition which usually
focus on benzene, BTEX and oxygenates (e.g., Maine Department of Environmental Protection,
2005)
U.S. EPA collects certain data on reformulated gasoline including the amount of benzene
and oxygenates (US EPA, 2005). EPA collates data generated by industry collected in summer
and winter surveys. Summaries of the data are available on the EPA web site for the years 1995
to 2003. The surveys may contain as many as 10,000 gasoline retail samples that are collected
during one-week survey periods by the RFG Survey Association (U.S. EPA, 2005). Prior to
1998, the surveys reported only total oxygen and oxygenate content, benzene content, aromatics
content and Reid vapor pressure. In 1998 and subsequent years, the surveys also reported sulfur,
olefins and certain distillation properties. Separately EPA compiles compliance data from
producers and importers. Average values of parameters are available for summer and winter
composition for the years 1997-2002 for both conventional and reformulated gasolines as
national averages.
-------
A voluntary industry consortium produces semiannual Petroleum Product Survey reports
and is based in Bartlesville, Oklahoma. This group collects data on benzene, oxygenates, the
boiling point distribution, vapor pressure, octane number and other properties in annual winter
and summer surveys of U.S. gasolines (e.g., Dickson 2004a, 2004b). These surveys have been
conducted since the 1930s and each may contain data from as many as 800 to 900 samples.
Winter and summer sampling periods last from December through February for winter samples
and June through August for the summer samples.
The octane number (ON) of a fuel is an important measure of its quality (Strikkers,
2002). The typical octane number of gasolines produced in the 1920s was 40-60. It has
increased since in order to meet performance needs of modern engines, namely to allow higher
compression ratios without pre-ignition of fuel and the resultant engine knock. The octane
number is determined by measuring the combustion efficiency of a fuel in comparison to a
prescribed mixture of isooctane (2,2,4-trimethylpentane) and n-heptane. Of these, the first has
an ON of 100, and the second of 0, because of their ability or lack of ability, respectively, to
prevent fuel pre-ignition. Thus the relative combination of these represents chemicals either
suppressing or enhancing engine knock. The octane number posted at gasoline stations in the
U.S. is an average of the Research Octane Number and the Motor Octane Number, although
many other variations on Octane Number exist and are used for various purposes (Chevron
U.S.A., Inc, 2005). Higher ON is provided to fuel by increased amounts of aromatics, olefins
and iso-paraffms (Strikkers, 2002)2. In general, branched hydrocarbons and aromatics are more
knock resistant than unbranched paraffins (ASTM, 1958, Meusinger and Moros, 2001). This
characteristic is illustrated in the archetypical octane enhancer 2,2,4-trimethylpentane or
isooctane.
The environmental impacts of fuels depend, in part, on their chemical properties. Data
on chemical properties are available from a variety of sources. There are no "complete" data
sets that include all hydrocarbons of interest. Further, the majority of property data are collected
at temperatures of 25 °C and data for shallow subsurface temperatures (5 °C to 25 °C) are
unavailable. The International Union of Pure and Applied Chemistry (IUPAC) published two
series of hydrocarbon solubility data (Shaw, 1989a, 1989b) that include hydrocarbons from C5 to
C7 and C8 to C36, respectively. For some well-studied compounds such as benzene, there were
multiple sets of solubilities reported over a range of temperatures. Depending upon the amount
and quality of the available data, IUPAC evaluated the data and provided recommended, best or
reported values.
Gustafson et al. (1997) developed parameter correlations for solubility, vapor pressure
and other properties based upon the equivalent carbon number of the compound. The
correlations were developed for aromatic and aliphatic constituents of fuels. Use of the
correlations requires estimation of the equivalent carbon number via the boiling point of the
2The oil industry uses the terms paraffins to refer to what are otherwise known as alkanes,
olefins for alkenes and alkynes, and naphthenes for cycloalkanes.
10
-------
chemical.
Because recommended values are not available for all desired properties of all chemicals
at all temperatures of interest, estimated properties were used instead. The SPARC Performs
Automated Reasoning in Chemistry (SPARC3) property estimator (Karikhoff et al., 1991, Hillal
et al., 2003) uses artificial intelligence, molecular structure and large training datasets (8000 data
points) to generate property estimates.
Purpose
The purpose of this study was to generate data on gasoline composition that could be
used as inputs to risk assessment models for ground water or indoor air contamination. The
approach taken was to characterize the chemical composition of gasolines from various locations
around the U.S, using as detailed a method as practical. Unlike the established EPA/Industry
consortia, the purpose here was to collect a much data as possible on the individual chemicals
composing the fuels. This approach was chosen because environmental impacts from fuels
depend on the specifics of their composition and the desire was to evaluate the impacts from the
greatest number of hydrocarbons as possible.
3SPARC is available at http://ibmlc2.chem.uga.edu/sparc.
11
-------
Approach and Methods
Gasolines were collected by sampling directly from retail pumps using the following
general protocol. A list of specified brands was used to choose two to four gas stations within a
1 mile to 2 mile radius of each other. Select brands were used because of the possible, but
unlikely, prospect that the petroleum-derived fractions of gasoline were brand specific. The
chosen stations were to be "clean" in appearance to minimize the potential for inclusion of off-
spec product in the study. Multiple stations were selected to provide close geographic
clustering and, where possible, stations across the street from each other were included. These
were included specifically to test for similarity among different brand names. To avoid mixing
of product in delivery lines, the stations were required to have three separate hoses for
dispensing gasoline.
The geographic locations that were chosen for the study were intended to include a
mixture of locales using conventional or reformulated gasoline (RFG), locales where ethanol
usage was likely, states with MTBE bans, low and high elevations, rural and urban areas, and a
balance of East and West coast samples (Table 1). The designation of RFG areas was based on
information from EPA that was updated as of February 23, 2004 (US EPA, 2005) and applies to
the specific town or city included (as Federal RFG requirements are not state-wide). A state
winter oxygenate requirement was in effect in Colorado (Colorado, 2002). Other requirements
for gasoline include limitations on Reid vapor pressure and sulfur content in some areas of the
country. Since these two were not included in the measured data, there is no further discussion.
The varying requirements may, however, cause changes in other components of gasoline that are
included in this study. Information on MTBE bans was provided by the American Petroleum
Institute (API, 2004).
Summer and winter samples were to be taken. The number of samples was limited by the
available project funding that meant roughly 120 samples in both winter and summer sampling
seasons would be analyzed. Because of several reasons, sampling was delayed and the data
presented in this report were collected between September and November 2004 and thus likely
contain both winter and summer samples.4 This "autumnal" sample contains 96 samples, most
of which are likely to be winter samples.
Regular and Premium gasoline were dispensed into separate 1 gallon gasoline cans and
transferred to 20 ml scintillation vials. The vials were filled to the bottom of the threading: (1) to
prevent gasoline from flowing over the outside of the vial, (2) to prevent breakage of vials due to
thermal expansion, and (3) to comply with one of the requirements of the small quantity
exceptions to dangerous good shipping under 49 CFR 173.4. The vials were then overnight
shipped in chilled containers for storage in Athens, Georgia. In Athens, samples were then
selected for analysis and shipped to The Cascade Group (TCG) for analysis. All analyses were
4Planned future sampling will provide true summer and winter samples and the data
reported here will be combined with the appropriate summer or winter data.
12
-------
completed by TCG within two weeks of receiving the samples.
The samples were analyzed by ASTM D 6729-1, a high resolution, gas chromatograph
method. The method uses a 100 meter capillary column and flame ionization detector. This
method was developed for the determination of individual hydrocarbon components of spark-
ignition engine fuels, including alcohol and ether oxygenated additives, all with boiling points
below 225 °C. Since a typical gasoline is a mixture of over 400 components, it would be
impractical if not impossible to impose data quality indicators on each analyte of interest.
Therefore, one component from each of the functional groups was tracked to assess the overall
quality of the analytical performance. The laboratory routinely monitored the repeatability and
reproducibility of its analysis. The repeatability was monitored through the use of laboratory
replicates at the rate of one per 10 samples. Reproducibility was monitored through the use of a
quality control sample analyzed at the rate at least one per 15 samples. Potential matrix effects
were monitored by the use of a spiked sample. Every 20th sample was spiked with 1% each of the
following components: Ethlybenzene, MTBE and 2,2,4-Trimethylpentane. The quality control
samples were plotted on individual control charts and the upper and lower control limits were
determined in accordance with laboratory protocols.
Using this method to identify gasoline components relies on retention time. As a result,
there is a possibility of co-elution of some peaks. Although a majority of the hydrocarbons can
be clearly resolved, there is a possibility of co-elution of compounds with olefms above C7.
Table 2 lists the potentially co-eluting compounds that are identified in ASTM D 6729-1. The
method notes that the list is not exhaustive. Although co-elution raises a potential problem in
interpreting the results, two factors mitigate this problem. First is that ASTM D 6729-1, Table 4,
indicates that one compound is predominant. Of the compounds listed in Table 4, most are not
of specific interest, with the exception of toluene. The second mitigating factor is that the co-
eluting compounds listed in Table Al.l of ASTM D 6729-1, are usually present at no more than
1000 ppm (0.1% by mass). Thus reported concentrations of MTBE and ETBE that themselves
are above 0.1 represent likely masses of these chemicals and not the co-eluting chemicals. A
compound that co-eluted with ETBE, Methylcyclopentane, was found to interfere with the ETBE
results from ASTM D 6829-1. Despite any mitigating factors, the nature of the ETBE results, in
particular, requires that confirmation by other means be performed. Confirmation is planned for
future samples and thus the ETBE data were omitted from this report.
13
-------
Table 1 Geographic locations, characteristics, and numbers of samples included in the study.
Elevations less than 2500 ft are judged "low." Fuel requirements for reformulated gasoline (RFG),
winter oxygenates (WO), and conventional gasoline (CG) are provided on a location by location
basis.
State
California
Colorado
Georgia
Delaware
Illinois
Montana
New York
Ohio
Oklahoma
Pennsylvania/New
Jersey
Texas
Virginia
Washington
Locale(s)
Los Angeles (Huntington Beach,
Irvine)
Denver
Athens (Bogart, Watkinsville)
Dover
Chicago (Batavia)
Billings, Great Falls, Helena
Long Island (Centereach, Merrick,
Riverhead)
Cincinnati
Ada
Philadelphia (Bala Cynwyd, PA,
Pennsehawken, NJ)
High Plains (Lubbock, Wolfforth)
Arlington
Seattle (Port Orchard)
Elevation
Low
High
Low
Low
Low
High
Low
Low
Low
Low
High
Low
Low
Fuel Requirements
RFG
WO
CG
RFG
RFG
CG
RFG
CG
CG
RFG
CG
RFG
GG
ctf
X>
W
B
s
Ye
s
Ye
s
No
No
Ye
s
No
Ye
s
No
No
No
No
Ye
s
Ye
s
Total
No. of
Samples
Js
00
<2
P*H
5
3
4
3
3
6
6
3
3
2
3
4
o
J
48
Premium
3
3
5
3
3
6
6
3
3
3
3
4
3
48
14
-------
Table 2 ASTM D 6729-1 known pairs of co-eluting peaks.
Predominant compound
Coeluting Compound(s)
Chemical
Notes on Co-eluting compound
Co-eluting pairs listed in Table 4 ofD 6729
3 ,3 -dimethylpentane
2-methylhexane
2,5-dimethylhexane
3 ,3 -dimethylhexane
toluene
1 , 1 ,2-trimethy Icy clopentane
C8-diolefin
4-methyloctane
1 ,2,3,4-tetramethylbenzene
5-methyl- 1 -hexane
C7-olefin
C8-olefin
C8-olefin
2,3 ,3 -trimethylpentane
C7-triolefin
C8-paraffin
C9-olefin
Cn-aromatic
if 2,3,3-trimethylpentane > 5 times the
toluene concentration
Co-eluting pairs listed only in notes to table Al.l ofD 6729
n-propanol
MTBE
MSBE (Methyl sec-butyl ether)
ETBE
isobutanol
3 -methyl- 1 -pentene
2,3 -Dimethyl- 1 -butene
1 -hexene
2,3 -Dimethyl- 1,3-
butadiene
4,4-Dimethy 1- 1 -pentene
usually < 1 000 ppm
usually < 1 000 ppm
usually < 1 000 ppm
usually < 1000 ppm
usually < 1 000 ppm
The results included identification of 312 chemicals in gasoline, of these 27 were
identified to only an isomer (i.e., Cn-Aromatic, C10-Iso-Paraffin, etc.). Unknown chemicals
were grouped into five categories: unidentified, unidentified-aromatic, unidentified-Iso-Paraffin,
unidentified-naphthenes, and unidentified-olefins. Table 3 identifies the minimum, average and
maximum amounts of each of these categories of unknowns. Of the remaining chemicals in
gasoline, two sets of chemicals were specifically included in the analysis: the first group
included BTEX, trimethylbenzenes and oxygenates, the second included the 44 chemicals used
15
-------
in the Canadian Petroleum Products Institute study of 1993 (Table 4). The remainder of
chemicals were selected from the analyte list of ASTM D 6729-1. In addition the chemicals
were classified by hydrocarbon group (Table 5).
Table 3 Percent by weight of unidentified compounds in analytical
results.
Group
Unidentified
Unidentified-aromatics
Unidentified-ISO-Paraffins
Unidentified-naphthenes
Unidentified-olefins
Minimum
1.87
0.004
0.088
0.08
0.021
Average
3.933
0.049
0.333
0.267
0.082
Maximum
9.987
0.094
0.584
0.801
0.202
Table 4 Specified chemicals included in the analysis.
Aromatics and Oxygenates
Additional CPPI (1994) Study Chemicals
Benzene
Toluene
Ethylbenzene
o-, m-, p-xylene
1,3,5-Trimethylbenzene
1,2,4-Trimethylbenzene
Methanol
Ethanol
t-Butanol
Methyl tert-butyl ether
Di-isopropyl ether
Ethyl-tert butyl ether
Tert-amyl alcohol
Tert-amyl methyl ether
Tert-amyl ethyl ether
1,3 Butadiene
iso-Butane
iso-Pentane
2,2-Dimethylbutane
2,3-Dimethylbutane
2-Methylpentane
3 -Methylpentane
2,4-Dimethylpentane
2-Methylhexane
3-Methylhexane
2,2,4-Trimethylpentane
2,4-Dimethylhexane
2,3,4-Trimethylpentane
2,3,3 -Trimethylpentane
2,3-Dimethylhexane
3 -Methylheptane
2-Methylbutene-l
2-Methylbutene-2
1-Methyl-3-ethylbenzene
1 -Methyl-4-ethylbenzene
1 -Methyl-2-ethylbenzene
Cyclopentane
Methylcyclopentane
Cyclohexane
Methyl cy cl ohexane
Naphthalene
2-Methylnaphthalene
1 -Methylnaphthalene
trans-Butene-2
cis-Butene-2
trans-Pentene-2
cis-Pentene-2
n-Butane
n-Pentane
n-Hexane
n-Heptane
16
-------
Table 5 Hydrocarbon groups and numbers of member compounds.
Group
Di/Bicyclo
Naphthenes
Di-Olefins
Indenes
I so -Paraffins
Iso-Olefins
Mono-Aromatics
Member
s
1
19
1
76
53
48
Group
Naphthenes
Napheheno/Olefino-
Benzenes
Naphtheno-Olefins
n-Olefins
Oxygenates
Paraffins
Member
s
4
o
J
4
37
12
12
SPARC was used to generate estimates of the boiling points, and, for temperatures from
0°C to 25°C in 5 °C increments, the vapor pressures, infinite dilution activity coefficients, and
solubilities of the chemicals reported in the gasoline samples. Of the 285 chemicals identified in
the analysis, SPARC was able to determine the properties for 281. These values were used in
the assessment of effective solubility and gas phase concentration that follows.
17
-------
Results
The data were evaluated in two ways. The first method was through comparisons of
concentration of single components against the octane number or another component. In almost
all cases scatter plots were used to get a sense of the abundance of various components under
various circumstances. The second set of analyses were performed using the entire suite of
chemicals reported for each sample. Because of the large number of gasoline components,
statistical methods were used: hierarchical cluster analysis (HCA), discriminant analysis (DA),
and principal components analysis (PCA). The results from these analyses provided assessment
of the variation among the gasolines and formed the basis for the decision tree presented below.
Octane Number
Data collection was designed to obtain samples of regular and premium gasolines. The
octane ratings of these fuels vary, largely because of elevation. Lower octane fuels are sold at
higher elevations. Because of larger numbers of low elevation samples, the most common
regular and premium gasoline octane numbers were 87 and 93, respectively (Table 6).
Table 6 Octane ratings of sampled gasolines.
Grade
Regular
Premium
Octane
Number
85.0
85.5
86
87
87.5
89
90
91
92
93
Totals
Conventional
0
4
2
15
1
0
2
9
3
9
45
Reformulate
d
3
0
0
22
0
1
0
5
1
19
51
Total
3
4
2
37
1
1
2
14
4
28
96
18
-------
Benzene
The benzene data showed variation at all octane levels (Figure 1). Because of mandates
of the Clean Air Act (CAA), benzene levels in RFG areas were approximately 1% or less5. In
conventional gasolines benzene varied at levels that were generally between 0.5% and 3% by
weight. The higher levels (greater than 2%) were found over the whole octane range (85.5 to
93). The triangles indicating winter oxygenate and high benzene were from the Colorado
samples, where an wintertime oxygenate requirement is imposed (Colorado, 2002).
3 i
2
CD
C
CD
N
CD
DO
1
A
A
A
O
O
84
86
90
92
94
Octane Number
Q Conventional
O Reformulated
A Winter Oxygenate
Figure 1 Benzene data for RFG and conventional gasoline
locales.
5The federal FRG mandate on benzene is given in volume percent, as opposed to the
weight percent used for these data.
19
-------
Methyl Tert-Butyl Ether
The MTBE concentrations plotted against the benzene concentrations are shown in
Figure 2. The fuels with high MTBE concentrations all occur with benzene concentrations less
than 1%. This follows from requirements of the RFG program where benzene concentrations
must be less than one percent by volume. Five conventional gasoline samples had MTBE
concentrations of 2.8% to 4.2%. These concentrations could be the result of MTBE use as an
octane enhancer, mixing of fuel types, or other reasons. All of these samples, however, were
premium gasolines obtained in Georgia, suggesting the deliberate use of the oxygenate as an
octane enhancer. At benzene concentrations above 1%, all the MTBE concentrations are low
and most of the samples come from conventional gasoline areas.
16 i
12
LJJ
DQ
4
o
o
n
D
1 2
Benzene (wt %)
D
O
A
Conventional
Reformulated
Winter Oxygenate
Figure 2 MTBE and benzene contents for all fuels.
When compared with the octane number (Figure 3), the MTBE concentration appeared to
depend on RFG requirements more than octane number. MTBE is high where RFG
requirements are in place and there is no MTBE ban. Otherwise the MTBE content is fairly low,
with the exception of the Georgia premium samples.
20
-------
16 |
12
LU
DO
4
84
90
ra
92
94
Octane Number
D
O
A
Conventional
Reformulated
Winter Oxygenate
Figure 3 MTBE content plotted against octane number.
21
-------
Tert-Butyl Alcohol
Figure 4 shows that the Tert-Butyl Alcohol (TEA) concentrations plotted against MTBE
content of the fuels. All TEA concentrations were about 0.1 % or less. The higher values
occurred with higher MTBE content, indicating some tendency for TEA to increase with MTBE.
0.8
0.6
<
GO
0.4
0.2
Q QO'
12
16
MTBE (wt %)
Figure 4 Tert-Butyl Alcohol (TEA) content plotted
against MTBE content, showing only small amounts of
TEA present in these fuels.
22
-------
Tert-Amyl Methyl Ether
Tert-Methyl Amyl Ether (TAME) was found at only low concentrations (< 1%) in the
fuel samples (Figure 5).
0.8
^ 0.6 |
-I*
LU
< 0.4
0.2
84
©
0
90
92
94
Octane Number
Q Conventional
O Reformulated
A Winter Oxygenate
Figure 5 Tert-Amyl Metyl Ether content plotted against
octane number.
23
-------
Oxygen Content
The oxygen content of the fuels was supplied by several compounds including MTBE,
ethanol and ETBE. Small amounts of TAME and DIPE were found in some samples. The total
oxygen content of each sample was calculated by summing the contributions from each
oxygenate. Table 7 lists the fraction of oxygen supplied by each. Because the weight fraction of
oxygen in each additive varies, the amount of the compound needed to meet the RFG
requirements also varies. Because the ETBE data were in doubt, ETBE was omitted from the
results presented in Figure 6. The actual oxygen content of the fuels may be higher than
indicated because this omission.
Table 7 Oxygen content of each oxygenated additive, based upon the number of
carbon, hydrogen and oxygen atoms per mole of compound.
Oxygenate
Methanol
Ethanol
Tert-Butyl Alcohol
Methyl Tert-Butyl Ether
Di-Isopropyl ether
Ethyl Tert-Butyl Ether
Tert-Amyl Alcohol
Tert-Amyl Methyl Ether
Tert-Amyl Ethyl Ether
Number of Atoms
Carbon
12 g/mole
1
2
4
5
6
6
5
6
7
Hydrogen
1 g/mole
4
6
10
12
14
14
12
14
16
Oxygen
16 g/mole
1
1
1
1
1
1
1
1
1
Weight
Fraction of
Oxygen
0.50
0.35
0.22
0.18
0.16
0.16
0.18
0.16
0.14
The oxygen content of these additive were summed and plotted against the benzene
concentration in Figure 6. RFG fuels show oxygen contents above 2 %. The majority of these
had benzene concentrations less than one. Five samples had total oxygen about 3% and had high
benzene concentration (above 1%). These were collected in Denver where state requirements
mandate the oxygen content at this level. Conventional gasoline samples showed oxygen
content below 1%, but not zero.
24
-------
3 i
2
"c
o
O
c
0
en
>,
x
O
O
o
A A
A
A
D
D
Benzene (wt %)
Q| Conventional
O Reformulated
A Winter Oxygenate
Figure 6 Composite weight of oxygen from all oxygenates
and benzene concentrations.
25
-------
Figure 7 shows the oxygen content (wt %) plotted against the octane number for both
reformulated and conventional gasoline locations. This plot also shows that oxygen content is
high in RFG areas and that lesser amounts (< 1%) were found in conventional gasolines.
3 i
2
O
O
c
0
en
O
A
A
A
A
O
O
O
D
-Q^
-Q-
84
86
88 90
Octane Number
92
94
O
A
Conventional
Reformulated
Winter Oxygenate
Figure 7 Fuel oxygen content plotted against octane
number for all samples.
26
-------
Ethanol
Ethanol has replaced MTBE as the principal oxygenate in three states included in the
Study: New York, Illinois, and California. Figure 8 shows the ethanol content in the fuels. The
ethanol content was high only in RFG locales, indicating little incidental presence where an
oxgenate was not required. Figure 9 shows that MTBE and ethanol use were mutually exclusive,
as the presence of one was associated with the absence of the other.
10 i
6
03
-^ 4
-t-t ^
LJJ
2
A
A
©00-
-m-
90
92
94
Octane Number
n
o
A
Conventional
Reformulated
Winter Oxygenate
Figure 8 Ethanol content plotted against octane number.
27
-------
10 |
o- 6 _
03
-^ 4
-t-t ^
LU
12
0 4 8
MTBE (wt %)
Figure 9 Ethanol content plotted against MTBE content
showing the mutual exclusive nature of the usage of these
chemicals.
16
28
-------
Brand
The gasolines sampled were sold under seventeen brand names. There is little evident
distinction among brands as indicated in Figure 10. This plot indicates that the benzene content
varies over roughly the same range regardless of brand. Other indicators of the lack of
significance of brand were generated by the cluster analysis and the principal components
analysis presented later.
3 i
2
A
A
0
0
N
0
DQ
1
D
D
A D
A
n
n
n
n
A
o
a
o
o
n
n
n
n
n n
n
n
O
n
n
o
n n
n
o
ooQ
o
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Brand (1 to 17)
Q Conventional
O Reformulated
A Winter Oxygenate
Figure 10 Distribution of benzene by brand.
29
-------
Similarities and Differences Among Gasolines
Cluster Analysis
The chemical analysis determined the composition of the gasolines in terms of
approximately 300 compounds. Since all of these compounds vary from one gasoline to the
next, a cluster analysis using all the available concentration data used to determine which
gasolines were similar. Figure 11 shows the results of a cluster analysis that included only the
samples from Georgia. By including gasolines from only one geographic location, variation due
to elevation and differing regulations are eliminated. Differences may remain due to fuel source
or vendor, and octane number among others.
The results presented in the figure are drawn as a dendrogram. On the horizontal axis the
sample identification is given. These samples are labeled GA for Georgia, 1,2,3, or 4 for the
station number, 87 or 93 for the octane number and 1,2,3, or 4 for the replicate number. Notice
especially that the samples are separated into two groups differing by octane number, where the
93 octane premium gasolines appear to the left of the center. The vertical axis represents the
degree of similarity between two samples. If two samples were identical then the similarity
would be 100% and they would be connected along the horizontal axis. For these data, no two
samples were exactly the same, so all of the connections occurred at lower levels of similarity.
These lower levels are plotted inversely on the vertical axis, so that the further a connection is
away from the horizontal axis, the more dissimilar were the samples. Since the samples are all
gasolines, there is ultimately a connection made between all the samples, albeit at a low level of
similarity.
The cluster analysis grouped the fuels into two clusters which differ by octane number.
This result was generated by the clustering algorithm acting only on the compositional data and
not the octane data. Two samples (GA-1-93-2 and GA-F-1-93-1) differed only by the fill level
of the sample vials and are nearly identical. The analysis grouped these two samples closest
together of any of the nine at a similarity level of 97.5%. The other three 93 octane samples
were included in the grouping as the similarity level dropped to 77.7%. The 87 octane gasolines
grouped with only each other although generally at lower similarity levels (64.7%).
Table 8 gives a definition of the clusters in terms of the most prevalent compounds
present at 0.50 wt % or more. In either case a different, but overlapping set, of compounds
defines the centroid of the cluster. The centroid is analogous to that of a physical object, but
here the number of dimensions (chemicals) is much greater than the three dimensions of space.
Although defining the cluster, this analysis does not determine which compounds differentiate
the clusters.
One noticeable difference between the regular and premium grades was the MTBE
content. In the regular grade fuel MTBE was present at an average mass percent of 0.52%, while
in the premium fuel MTBE was present at an average mass percent of 3.33%. This indicates that
the MTBE may be purposefully present in these Georgia gasolines as an octane enhancer. A
30
-------
similar result is found in industry data for their Southeast region6, Florida and a few other
scattered samples (Dickson, 2004a, 2004b).
The increased abundance of chemicals of at least 0.50% weight in Georgia premium
grade gasolines compared to regular grades is shown in Table 9. These data show that the
compounds that increase are predominantly iso-paraffins, oxygenates and aromatics. The pure
compound octane numbers from ASTM (1958) are given in the table. Since many of these
compounds are octane enhancers, their octane numbers are generally high (i.e., above that of the
regular grade fuel - 87). Similarly, Table 10 shows the compounds that decreased. These
include paraffins, oxygenates, aromatics and olefins, generally with lower octane numbers.
Conventional Gasoline: Georgia
31.45-
> 54,30-
2
1
en
77.15-
100,00
CO
cr.
LL
<
O
Cr.
CJi
Cri
Cr.
OO
QO
QO
OO
Observations
Figure 11 Dendrogram showing the cluster analysis' separation of 93 octane (left) from 87
(right) octane Georgia gasolines.
6The region includes North Carolina, South Carolina, Georgia, Tennessee, Alabama,
Mississippi, Arkansas, and Louisiana.
31
-------
Table 8 Cluster centroids in weight % for Georgia 87 octane and 93 octane gasolines
defined by all components present at 0.50 wt % or greater.
87 Octane
Toluene
i-Pentane
m-Xylene
n-Pentane
2,2,4-Trimethylpentane
2-Methylpentane
Unidentified
1,2,4-Trimethylbenzene
Ethyl-Tert-Butyl-Ethe and
Methy [cyclopropane
2-Methylhexane
o-Xylene
n-Hexane
Ethylbenzene
3-Methylpentane
1 -Methyl-3-ethylbenzene
p-Xylene
3-Methylhexane
2,3,4-Trimethylpentane
Benzene
n-Heptane
n-Butane
2-Methylbutene-2
2,3-Dimethylbutane
Methylcyclohexane
1,3,5-Trimethylbenzene
1 -Methyl-4-ethylbenzene
t-Pentene-2
1 -Methyl-2-ethylbenzene
3-Methylheptane
1,2,3-Trimethylbenzene
2,4-Dimethylpentane
n-Propylbenzene
Cluster
10.3725
6.4405
4.3295
4.075
3.477
3.3808
3.2733
2.814
2.534
2.493
2.3658
2.3655
2.236
2.2058
2.0488
1.8975
1.7058
1.6695
1.4538
1.3763
1.2463
1.1935
1.0755
0.9983
0.9268
0.9265
0.8038
0.7798
0.7228
0.7148
0.7038
0.701
l,2-Dimethyl-4-ethylbenzene 0.664
n-Octane
2-Methylheptane
Cyclohexane
2-Methylbutene-l
1 -Methyl-3-n-propy Ibenzene
2,4-Dimethylhexane
2,3-Dihydroindene
2,3-Dimethylhexane
Methyl-t-butylether
0.6523
0.6495
0.6065
0.5715
0.5595
0.5475
0.5383
0.5208
0.5188
93 Octane
Toluene
2,2,4-Trimethylpentane
i-Pentane
1,2,4-Trimethylbenzene
2,3,4-Trimethylpentane
Methyl-t-butylether
m-Xylene
1 -Methyl-3-ethylbenzene
Unidentified
n-Pentane
2-Methylhexane
2-Methylpentane
o-Xylene
Ethylbenzene
p-Xylene
1,3,5-Trimethylbenzene
2,3-Dimethylbutane
1 -Methyl-4-ethylbenzene
3-Methylhexane
3-Methylpentane
2,4-Dimethylhexane
2,2,5-Trimethylhexane
2,4-Dimethylpentane
2,5-Dimethylhexane
2,2-Dimethylbutane
n-Butane
2,3-Dimethylhexane
1,2,3-Trimethylbenzene
1 -Methyl-2-ethylbenzene
n-Hexane
Ethyl-Tert-Butyl-Ether and
Methy [cyclopropane
n-Heptane
Benzene
2-Methylbutene-2
n-Propylbenzene
1 ,2-Dimethyl-4-ethylbenzene
1 -Methyl-3-n-propylbenzene
t-Pentene-2
1 ,3-Dimethyl-5-ethylbenzene
TAME
Cluster
14.2292
8.4866
6.108
4.2254
3.5538
3.3262
3.246
2.608
2.5376
2.2486
2.2142
2.1668
1.8414
1.53
1.5
1.346
1.287
1.2142
1.1952
1.1754
1.0838
1.0666
1.0514
1.032
1.001
0.9968
0.9952
0.9942
0.9716
0.903
0.8538
0.8476
0.822
0.8168
0.8078
0.678
0.6678
0.6048
0.5464
0.5064
32
-------
Table 9 Compounds that were present in greater abundance in Georgia
premium than regular grades by at least 0.50 wt %. Octane Number data
were taken from ASTM (1958).
Chemical
2,2,4-Trimethylpentane
Toluene
Methyl-t-butylether
2,3,4-Trimethylpentane
1 ,2,4-Trimethylbenzene
2,2,5-Trimethylhexane
2,2-Dimethylbutane
1 -Methyl-3-ethylbenzene
2,4-Dimethylhexane
2,5-Dimethylhexane
Pure Chemical
Octane Number
( R + M )/2
100
118
147
99.5
136
89.5
93
150
69
-
Increase in wt %
Regular to
Premium
5.0096
3.8567
2.8074
1 .8843
1.4114
0.5923
0.5725
0.5592
0.5363
0.5175
Table 10 Compounds that were present in lesser abundance in Georgia premium gasolines than
in regular grades by at least 0.50 wt %. Octane Number data were taken from ASTM (1958).
Chemical
n-Pentane
Ethyl-Tert-Butyl-Ether and
Methylcyclopropane
n-Hexane
2-Methylpentane
m-Xylene
3-Methylpentane
Unidentified
Ethylbenzene
Methylcyclohexane
Benzene
n-Heptane
o-Xylene
Pure Chemical
Octane Number
( R + M )/2
64
20.5
80
134.5
83
-
115.5
94
94
0
111
Decrease in wt%
Regular to
Premium
-1 .8264
-1.6802
-1 .4625
-1.214
-1.0835
-1.0304
-0.7357
-0.706
-0.6909
-0.6318
-0.5287
-0.5244
The similarity of three conventional, low elevation, gasolines were tested by performing
a cluster analysis of Georgia, Ohio and Oklahoma gasolines (Figure 12). These samples are
designated by state code, station number (1,2,3,4), octane number (87 or 93) and replicate
sample number (1,2,3,4). The 87 octanes are clustered with each other on the left most side of
Figure 12. The regular grade samples all cluster at 80.05% similarity, while the two main
subclusters (samples GA-1-87-2, GA-2-87-3 and GA-3-87-3 form the leftmost cluster) both join
at a similarity level of around 85%. The premium grade samples group into three clusters, the
33
-------
first contains the Georgia samples only. The second and third, while containing one location
each, also show lower levels of similarity, 71% and 82%, respectively. This indicates that there
is more variability among the premiums than among the regulars, because the premiums fail to
cluster together. The poor clustering of the premium samples is partly due to the appreciable
MTBE contained in the Georgia premium samples which is absent in those from Ohio and
Oklahoma. The MTBE content is the most obvious difference between the fuels but others exist.
The Ohio and Oklahoma samples themselves display variation because of their low level of
similarity where they cluster. The complete separation by grade in Figure 12, however, shows
that the 87 octane and 93 octane gasolines are more similar to each other than fuel of the
opposite grade from the same location.
Low Elevation, Conventional Gasoline: Georgia, Ohio, Oklahoma
59.77 -
> 73.18-
_2
__
86,59 -
inn nn -
r
i
r-.
CO
i
iH
1
i
r
_l_
^^^^^
p_
i-IOnOJi-li-li-li-li-
i
r^
nn
i
r\l
i
r-
n
- r-
1 0
1 1
^ r^- [~~
3 00 0
i i
1
- r-
-i n
i
ID ^ I~H on i-
i
i
i
- r-
i r
^ ir
D ^.
rp--"-"-"-"-"-"---"-"-"-"
H OJ OJ T-I on IT
i r-
- o
- r
1 r
III
T IT) 01 IT
Ji CTi CTI G
iii
HmOOOJiH-rH'^riT
1
i i
<<<
-------
three locales (California, Colorado, and Texas High Plans), there was not good separation
between the regular and premium grades. In all other cases the two grades formed distinct
clusters.7 On average the regular grade gasolines tended to cluster at higher degrees of similarity
than did the premiums, meaning that the regular grade fuels were more similar to each other than
were the premiums. This result suggests that refiners may be using a variety of approaches to
boosting the octane numbers of the premium fuels.
Table 11 Degree of similarity where all fuels of a given grade join in cluster analyses of individual
locales (states).
Locale
California
Colorado
Delaware
Georgia
Illinois
Montana
New York
Ohio
Oklahoma
Pennsylvania
Texas High Plains
Virginia
Washington
Average
Standard Deviation
Degree of Similarity
at Complete Joining
Regular
(%)
55.72
47.98
56.94
64.70
79.35
87.44
60.51
66.96
84.69
84.10
32.60
38.77
38.77
64.44
17.86
Premium
(%)
43.87
43.59
82.42
77.76
61.91
62.36
67.72
48.37
80.58
21.94
32.38
50.04
37.59
54.66
19.18
Premium Samples
Joined to Regular
Clusters
2 of 3
2 of 3
--
--
--
Iof6
--
--
--
--
2 of 3
--
--
7The regular grade and premium grade samples from one of the Montana gas stations
joined the cluster of the opposite grade. It is possible that these samples were mislabeled when
sampled or that wrong fuel had been delivered to the tanks at the station.
35
-------
High elevation samples from Montana and the Texas High Plains are shown in Figure 13.
The two states' samples do not join in common clusters. Further, four of the six Texas samples
cluster together despite differences in grade, while one premium sample (TX-1-90-2) remains
separate from all other samples until the lowest degree of similarity is attained.
High Elevation, Conventional Gasoline: Texas-Montana
63.65 -
> 75.77 -
Jl
1
S7.88 -
100.00
in
in
oo
OJ
in
in
o
OJ
iH
CTi
n
oo
in
in
o
in
in
op
in
rxi rxi
oo
iH
X
oo
r\i
X
OJ
CD
CTi
CM
X
ro
CTi
01
X
CM
rv
oo
m
X
CJi
CM
iH
CJi
in
CTi
i i
CTi 00
i i
oj n
i i
CTi
OJ
cb
X
Observations
Figure 13 Dendrogram showing relationship between high elevation areas in Lubbock, Texas
and Helena, Billings and Great Falls, Montana.
A comparison of reformulated and conventional gasolines is given in Figure 14 by
comparing the Georgia gasolines with RFG gasolines from Northern Virginia. Clustering of the
Georgia gasolines is the same as shown previously in Figure 11. Four distinct clusters were
found in the data. In addition to the two Georgia clusters, the Virginia gasolines formed two
grade-dependent clusters. Thus the two different clusters of Virginia gasoline are more similar
to each other than they are to the Georgia gasolines. The 93 octane Virginia cluster has a sample
(VA-4-93-4) that is unlike the others and only joins the cluster at a low level of similarity
(59.0%).
36
-------
Conventional:Reformulated Gasolines: Georgia: Virginia
38.78-
> 59.19-
79,59-
100,00
1
oo
oo
OJ
00
oo
oo
OJ
oo
oj
00
rh
CTi
rh
CTi
CJi
(J (J (J (J
CO
CTi
OJ
<
(J
00
CO
OJ
CO
n
oo
oo
ro
CJi
OJ
OJ
rh
CTi
oo
oo
CJi
Observations
Figure 14 Dendrogram showing relationships between conventional gasolines from Georgia and
RFG gasolines from Virginia.
Figure 15 compares gasolines from low elevation, RFG, non-MTBE ban areas in
Arlington, Virginia, Dover, Delaware, and Philadelphia, Pennsylvania. These gasolines separate
along grade lines first8. The regular gasolines form two distinct clusters: one composed of the
Virginia gasolines and the other composed of the Dover, Delaware and Philadelphia gasolines.
Notably Dover and Philadelphia are close geographically and may consequentially receive
similar gasolines. The premium gasolines show similar behavior with a lessened tendency for
clustering. The samples VA-4-93-4 and PA-3-93-4 join together but were not sold under the
same brand name.
retailer.
8The Delaware 89 octane sample (DE-1-89-2) may represent fuel mislabeled by the
37
-------
RFC: Virginia-Delaware-Pennsylvania
54.15-
69.43 -
.2
1
84.72 -
100.00
00
oo
00
oo
0\l
0\l
000000000000000000
oS
cr.
Ovl 00
I I
> >
<
<
o\i
oo
LLJ
Q
0\l
oh
cr.
o\i
oh
cr.
oh
oh
cr.
oo
cr.
oo
- Q > > > >
Observations
oo
or.
LLJ
Q
oh
cr.
OJ
oo
or.
00
lil
Q
00
CTi
oh
cr.
Figure 15 Analysis of three RFG areas without MTBE bans.
Three Federal RFG states in the study have banned MTBE: New York, Illinois and
California. To compare gasolines from MTBE-ban states with others, a cluster analysis was
performed on the data from Virginia, Pennsylvania and New York. Figure 16 shows that the
New York data is separate from the Pennsylvania and Virginia data. Beginning from the left of
Figure 16, the New York 87 octane gasolines cluster, followed by the New York 93 octane
gasolines. These two clusters join together at a similarity of 71.0%, before they join with the
clusters from Virginia and Pennsylvania. The data from these two states tend to cluster with
themselves before joining the other state. Even so they join each other at higher levels of
similarity, before joining with the corresponding gasoline grades from New York.
38
-------
Reformulated Gasolines: Virginia, Pennsylvania, New York
36.15-
100.00
Observations
Figure 16 Dendrogram of RFG gasolines from an MTBE-ban state (New York) and two states
without bans.
The MTBE-ban states and locales with RFG requirements (state or Federal) are
compared in Figure 17. The highest similarity between samples occurred for two California
gasolines (CA-3-87-2 and CA-4-87-2). Other California gasolines tended to cluster, but two
high octane gasolines were included with the 87 octane fuels. A high degree of similarity
occurred with the New York gasolines for both 87 octane and 93 octane gasolines. Some of the
Illinois and Colorado samples were grouped with the 87 octane New York gasolines. The
Illinois 93 octane samples were similar at a fairly low level (69.5%). Two of the Colorado high
octane samples (CO-2-91-1 and CO-3-91-1) grouped with the composited 87 and 93 octane
clusters formed largely by the New York samples. Since the Colorado samples were taken from
high altitude, and Colorado has specific state RFG requirements, the samples' lack of similarity
with the other locales should not be unexpected.
39
-------
RFG (State or Federal), MTBE Ban States: CA, CO, IL, NY
60.31 -
> 73.54-
.2
1
36.77 -
100.00
TrrvJfXI^l-rvjrjrVJ-i-li-l-i-li-l-i-li-l-i-l-i-li-li-li-l-i-li-l-i-li-l-i-li-l-i-li-irXI
I I I I I I I I I I I I I I I I I I I I I I I I I I I
ooooooaioooocnoooooooocoooooooooooooaiaicnaiaiQiaiooaiaiaicricricri
LLLLLLLLLLLLLLO'O'^'
Observations
Figure 17 Dendrogram of MTBE-ban states with State or Federal RFG requirements.
Figure 18 shows a cluster analysis of New York and Illinois samples. These both come
from states with MTBE bans, low elevation, and RFG requirements. The New York samples
form a cluster, that at lowered levels of similarity, are joined by the Illinois samples. The
premium gasolines from New York cluster with these regular gasolines at higher levels of
similarity than they do with the Illinois premiums. Again this result suggests a low level of
similarity among the premium grade samples.
40
-------
Reformulated Gasoline, MTBE ban New York, Illinois
1.84-
> 69.23 -
«
84.61 -
100.00
OO
1
OO
OOOO
OO
OO
OO
1
OO
1
OO
1
CTiCri
1 1
Cri
1
Cri
1
Qi
1
cn
1
en
OJ
cn
CO
n
en
OJ
Observations
Figure 18 Dendrogram showing relationship between two MTBE ban, RFG gasoline areas.
Principal Components Analysis
The basic goal of principal component analysis (PCA) is to reduce the dimensionality of
a multivariate data set consisting of a large number of variables, while capturing as much of the
variation as possible in the data set. PCA is a mathematical technique that transforms a number
of potentially correlated variables into a small number of uncorrelated variables called principal
components (PC). The first PC accounts for as much of the variability in the data as possible,
and each succeeding PC accounts for as much of the remaining variability as possible (Kendall
1975, Jolliffe 2002). Further background information on PCA is given in an Appendix (page
79). The PCA performed on the gasoline data served two purposes. First it complimented the
results from the cluster analysis by indicating which groups of samples were similar to each
other. Second, it gave information on the components that had the biggest influence on the
groupings.
Results of the PCA on the individual gasoline components are presented in Table 12.
Only 8 PC were required to account for 95 percent of the variation in the gasoline samples. The
total variance estimate of the gasoline components was found to be 87.53. The first two PC
accounted for almost 60 percent of the variation of the gasoline components. A plot of the
cumulative proportion of variation explained by PC number is given in Figure 19. The figure
41
-------
shows that between 99 and 100 percent of the variation in the gasoline components can be
explained by 16 and 33 PC, respectively. A scree diagram is a plot of the variance of each PC
against the PC number, p (the rank importance of the PC, see page 79 for more information).
Figure 20 contains the scree plot of the variance estimates by PC number which shows that the
variance estimates decrease rapidly until approximately the tenth PC. Beyond this point there
was little change in the variance estimates.
Table 12 Statistics of the first eight PC from the PC A of the individual
gasoline components including variance, proportion of the total variance
estimate, and the cumulative proportion of variation explained.
PC
1
2
O
4
5
6
7
8
Varianc
e
29.493
20.652
16.408
8.574
3.189
2.522
1.376
1.169
Proportio
n
0.337
0.236
0.188
0.098
0.036
0.029
0.016
0.013
Cumulative
0.337
0.573
0.760
0.858
0.895
0.924
0.939
0.953
42
-------
1 .0 :
0.9 :
0.8 :
I O.Gd
10
30
40
50
PC Number
GO
70
80
90
100
Figure 19 Plot of the cumulative proportion of variation explained by PC number.
25.0 :
I
10.0 :
0 10 20 30 40 50 GO 70 80 90 100
PC Number
Figure 20 Scree graph of the variance estimate by PC number for the individual gasoline
component data.
43
-------
The largest loadings from the eight PC indicate which components contribute the most to
the variation (Table 13). Table 13 shows that the gasoline component corresponding to the
largest absolute loading in the first PC was 2,2,4-Trimethylpentane. This component also had
the largest standard deviation (rank =1) among all of the gasoline components analyzed. From
Table 13 we can also see that those gasoline components with the largest coefficients
consistently had the largest standard deviations. Thus, the eight gasoline components presented
in Table 13, represent those components which contributed the most to the variation among the
gasoline samples. As discussed below, the first two principal components were dominated by
2,2,4-Trimethylpentane (isooctane) and MTBE, representing the influence of octane number and
oxygenated additives, respectively.
Table 13 Gasoline components with the largest loadings in the first eight principal components
including class, mean, standard deviation, and the rank of the standard deviation (from largest
(1) to smallest(8)).
PC
1
2
o
J
4
5
6
7
8
Gasoline Component
2,2,4-Trimethylpentane
Methyl-t-butylether
Toluene
Ethanol
2-Methylhexane
Unidentified 1
n-Butane
n-Pentane
Class
Iso-Paraffins
Oxygenates
Mono-Aromatics
Oxygenates
Iso-Paraffins
Paraffin
Paraffin
Mean
5.62
2.31
8.15
2.28
3.56
3.93
2.50
2.74
Std. Dev.
4.60
4.07
3.35
3.36
2.34
1.60
1.11
1.40
Rank Std.
Dev.
1
2
4
3
5
7
10
8
A plot of the first two principal components labeled by state is given in Figure 21. From
the figure it can be seen that marked differences exist among the states, with the data falling into
roughly five clusters. Samples from Virginia, Pennsylvania/New Jersey, and Delaware were
concentrated at the bottom of the graph. These fuels are all low elevation RFGs relying on
MTBE for their oxygen content. Georgia premiums lay in the middle of the graph. As
previously noted, these conventional gasolines contained MTBE presumably as an octane
enhancer. At a level of the 2nd principal component of about 0 to 2, a large group of samples
form a thin band. These included the majority of the conventional gasolines. At a level of 3 to
5, a thin band was distinguished by the four locations with RFG requirements and MTBE bans.
These gasolines shared the common characteristic of low MTBE concentration and ethanol use
to meet their oxygen requirement. The last separate cluster consisted of gasolines from
Colorado, that were under an oxygenate requirement. The 2nd principal component can be
identified with the the two dominant components that supply oxygen in the fuels: MTBE which
is at a maximum at the bottom and ethanol which is at its maximum at the top of Figure 21.
44
-------
Marked differences appeared between the octane levels (Figure 22). With a few
exceptions, the octane numbers fell into two classes, 87.5 or lower and 90 or above. The regular
gasolines (87.5 or below) were grouped to the right of the value of about 3 on the horizontal axis,
while the premium gasolines (90 or above) levels fell to the left of this value. The first principal
component was most highly influenced along its negative axis (toward high octane fuels) by
2,2,4-Trimethylpentane (isooctane) and other iso-paraffms. In the positive direction (toward low
octane fuels) it was influenced most strongly by paraffins and a variety of compounds of the
other classes. Isooctane had the strongest influence of any chemical on this principal
component. The mean values 2.78 and 8.46 % were for the 87.5 or lower and 90 or above octane
groups, respectively. The dominance of isooctane explains the ability of this principal
component to separate the octane classes. Isooctane is neither the only chemical that influences
the octane number, nor the principal component. Thus there was not a direct relationship
between the principal component and the octane numbers.
A plot of the first two principal components labeled by gasoline brand is given in Figure
23. No obvious patterns can be detected among the 17 brands under which the gasolines were
sold.
45
-------
E
S
31
1
0
_]
-4
-S
-6
-9:
-1 0 :
-11
High
Ethanol
t
'High
MTBE
Winter Oxygenate
FG, MTBE ban
Conventional
MTBE
(for Octane
RFC No
MTBE ban
-10
10
RwtPC
Figure 21 Principal Components plot of the 96 gasoline samples with respect to the first two principal components labeled by state.
46
-------
6:
4:
3:
2:
1 :
0:
-1 :
-2:
-3:
-4:
-5:
-6:
-7:
-8:
-9:
-10:
-11
-20
1
T
10
Figure 22 Principal Components plot of the 96 gasoline samples with respect to the first two principal components labeled by octane
number.
47
-------
o
a.
o
o
-------
Discriminant Analysis
A step-wise discriminant analysis was used to determine which chemical components
were most powerful in distinguishing among the samples (see Rencher, 1995). The analysis tests
the statistical level of significance of each variable and determines if it contributes to
discriminating among the samples. The results of this analysis are shown in Table 14. The
significance level was high for RFG/Conventional gasoline, MTBE ban, grade, elevation and
altitude. It was low for all factors taken together and brand. Thus gasolines having the same
values for the characteristics and a high significance level are expected to separate well based on
composition. Two clear compositional indications emerged: for distinguishing an MTBE ban
gasoline, the most important chemical was ethanol. This follows from the substitution of ethanol
for MTBE in CA, CO, IL and NY. Ethanol and MTBE were most important for distinguishing
RFG from conventional gasolines, clearly because of their usage as oxygenates.
Table 14 Discriminant analysis results.
Group Tested
All
RFG/Conventional
gasoline
MTBE Ban (y/n)
Brand (17)
Grade (<89, >90)
Altitude (H/L)
Significance
Level
0.59
0.99
0.96
0.40
0.95
0.94
Number of
Components
Required
18
19
20
22
23
23
Five Most Significant
Components
Methyl-t-butylether
Ethanol
n-Hexane
3t-Ethylmethylcyclopentane
n-Pentadecane
Ethanol
Methyl-t-butylether
1 c,2t,4-Trimethylcyclopentane
C9-01efm4
o-Xylene
Ethanol
n-Pentadecane
3t-Ethylmethylcyclopentane
C10-Isoparaffin2
3,4-Dimethylpentene- 1
3,5,5-Trimethylhexene-l
l-Methyl-3-n-propylbenzene
lt-M-2-(4-MP)cyclopentane
n-Pentadecane
2,3,3-Trimethylbutene- 1
Ethyl-Tert-Butyl-Ether and
Methylcyclopropane
4-Methylnonane
n-Hexane
2-Methylindan
lt,3-Pentadiene
C 1 1 -Isoparaffin2
i-Propylbenzene
2,2-Dimethylpropane
Benzene
2-Methylbutene-l
49
-------
Representative Gasolines
Figure 24 shows a decision tree that represents four characteristics of the fuels: grade,
sale elevation, MTBE ban status, and legal requirements. The latter include the federal
reformulated gasoline mandate, state requirements (as the California RFG requirement, or
Colorado winter oxygenate requirement) and, alternatively, conventional gasoline. These
provide a potential means for dividing or grouping samples. For a given grade of gasoline there
are 12 branches which may represent 12 or more types of gasoline because of the number of state
requirements. Seven of the branches were represented in this study. In all cases more data
would be desirable to increase the number of samples in each category.
Division of the data by the characteristics is supported to some degree by the statistical
analyses. These provide a scientific basis for what appear to be logical groupings. The
discriminant analysis supported division by grade, elevation, MTBE ban and RFG requirement
as these were found to be factors that provided a high degree of reliability of discriminating
between samples. The principal components analysis supported division by grade and combined
MTBE ban status and RFG requirements. The cluster analysis generally supports division by
grade, RFG requirement, and MTBE ban status, though not elevation. Because it was performed
on subsets of the data, the cluster analysis showed that the premium gasolines were more
variable than the regular gasolines.
Figure 24 indicates where the data are likely to be best grouped together. At the ends of
the decision tree, parenthesis around state names indicate that the cluster analysis indicated
similarity. All of these are not clear cut, and other judgements could be made to separate or join
data together. One argument in favor of grouping data together at the end of each branch of the
decision tree is that the variability will be maximized by this procedure. Risk assessments
relying on typical data, such as these, should include the maximum range of variability that
might occur.
50
-------
Grade
Premium
Regular
Elevation
> 3000 ft
< 3000 ft
> 3000 ft
< 3000 ft
MTBE Ban
Status
ban
no ban
ban
no ban
ban
no ban
ban
no ban
Legal Requirements
Federal RFG
State Requirement CO
Conventional
Federal RFG
State Requirement
Conventional (TX), (MT)
Federal RFG (NY), (IL)
State Requirement CA
Conventional WA
Federal RFG (VA), (DE, PA)
State Requirement
Conventional (GA), (OH), (OK)
Federal RFG
State Requirement CO
Conventional
Federal RFG
State Requirement
Conventional (TX), (MT)
Federal RFG (NY, IL)
State Requirement CA
Conventional WA
Federal RFG (VA, DE, PA)
State Requirement
Conventional (GA, OH, OK)
Figure 24 Decision tree for gasolines typical of the data collected in this study. The parenthesis
indicate where the cluster analysis indicated that samples from the various states are similar
enough to group together.
51
-------
Cluster Analysis
The degree-of-similarity of the representative gasolines was tested by performing a
cluster analysis. Figure 25 shows the results from averaging all samples at each branch of the
decision tree. For the maximum differentiation between the representative gasolines, ideally the
clustering would be at low levels of similarity. The identification of these data are: RFG -
N/Y/Y-State, MTBE ban - Y/N, grade R/P, elevation H/L. A few clear indications emerged
from this figure. Beginning on the right hand side, the low elevation, RFG, grades join at a low
level of similarity (Y-N-R-L and Y-N-P-L). Some RFG gasolines do not cluster well with
others: YS-Y-P-H and Y-Y-P-L. The remaining categories showed a fairly low level of
similarity for the most part, with the exception of the N-N-R-H and N-Y-R-L samples which join
at a level of similarity (85%) that was judged to be favorable for individual samples.
Analysis of Decision Tree End Points
38.88 -
> 59.26 -
"Z
1
79.63 H
100.00
C
D_
Observations
Figure 25 Cluster analysis performed on the representative gasolines.
52
-------
Predicted Environmental Impacts of the Study Fuels on Water and Air
From an environmental contamination perspective, fuel composition is most relevant for
soil contamination where the fuel itself is contained within the sample. Ground water and soil
gas contamination are equally important, but the relevance of each compound depends on its
abundance in the fuel and its properties.
Raoult's Law partitioning relationships were used to determine the equilibrium
concentration of each chemical in air and water at 15 °C (Schwarzenbach et al., 2003).
Set = XiSi (1)
where Sei is the effective solubility of chemical i resulting from the gasoline mixture [mg/L], x; is
the mole fraction of chemical i in the mixture [moles/moles], and S; is the solubility of chemical i
[mg/L]. Similarly the effective vapor pressure due to the presence of the chemical in the mixture
are given by Raoult's Law
P., = ^VPi (2)
where Pei is the partial pressure of chemical i resulting from the gasoline mixture [atm], X; is the
mole fraction of chemical i in the mixture [moles/moles], and VP; is the vapor pressure of
chemical i [atm].
Because solubility and vapor pressure data are not available for all hydrocarbons at all
temperatures of interest and because there exists considerable variability in literature values, the
SPARC Performs Automated Reasoning in Chemistry (SPARC) calculator was used to estimate
all properties used in the analysis. SPARC estimates are generally viewed to be within one-third
to one-half log unit of true values. Thus the SPARC estimates may differ from commonly used
estimates of some parameters. SPARC recognized 292 of the 296 chemicals with CAS numbers
in the data set.9 The disadvantages of SPARC were outweighed by the advantage of having a
consistent set of parameters for all chemicals at all temperatures of interest (0 °C to 25 °C). The
mole fractions, x;, were calculated from the mass fractions, m;, by
gmWj
(3)
E-^
gmw.
9In addition to the specific chemicals, 5 groups of unknowns and 30 groups of
generalized composition were reported.
53
-------
and using the gram-molecular weights, gmw; [g/mole]. Equation 3 was used by assuming a
sample size of Ig of gasoline. From this mass the number of moles of each identifiable chemical
was determined, summed and used to calculate the mole fractions. Chemicals that were
unidentified were omitted from the calculation, so the results gave an approximation of the true
mole fractions.
Gas phase concentrations, Ca, that would be associated directly with the fuel were
calculated from equations 2 and 3, and the ideal gas law:
P
e
Ca = ^gmw{ (4)
where R is the gas constant, 8.205 x 10"5 atm - m3/mole °K, and T is the temperature in Kelvin.
Predicted Aqueous Concentrations
Figure 26 shows a plot of the effective solubility of each chemical, labeled by group
(oxygenate, BTEX, parafins, etc), and plotted against the weight % in the sample. The data were
the average conventional gasoline, no MTBE ban, low elevation, regular grade samples. Only
solubilities above 0.001 mg/L were plotted as any lower values would be below typical detection
limits. The plotting position on this graph indicates the partitioning behavior of the compound.
For example, compounds plotting at the lower right (high weight % and low effective solubility)
are present in large quantities in the fuel but have little impact on water quality. Conversely, the
plotting in the upper right are compounds with high effective solubility and abundance in the
fuel. These are the major components of fuel that do potentially impact water quality.
The highest effective solubilities (10 mg/L to 100 mg/L) occurred for oxygenates and
BTEX (Figure 26). In the next lower group (1 mg/L to 10 mg/L) were representatives of all the
remaining groups. Many hydrocarbons plot lower in effective solubility (less than 1 mg/L) and
also occur at low weight percent.
Similar results for the average no-MTBE-ban, low-elevation, regular-grade, reformulated
gasoline are shown in Figure 27. Here MTBE had the highest effective solubility, again
followed by BTEX and oxygenates in the 10 mg/L to 100 mg/L range. Table 15 shows predicted
concentration increases greater than 0.05 mg/L between conventional gasoline and RFG. The
compounds with the highest increases are all oxygenates, followed by naphthenes and olefins.
Table 16 shows the corresponding decreases in concentration. The compounds with the largest
decreases are BTEX, and mono-aromatics, followed by a few paraffins, an iso-paraffin and di-
olefins.
If an MTBE ban were introduced, gasoline composition might be similar to the data
obtained from the Illinois samples. The average effective solubility of chemicals in the average
54
-------
Illinois sample are plotted in Figure 28. The highest effective solubility was found for ethanol,
followed by a similar pattern for the other fuels. For this potential change in fuel, there are a
smaller number of chemicals with increased effective solubility (Table 17), including the
alcohols and benzene along with two paraffins and an iso-paraffm. Compounds with decreased
effective solubility (Table 18) are the ether oxygenates, BTEX other than benzene, and a suite of
other compounds including n-olefins, iso-olefms and other compounds.
Conventional, No MTBE Ban, Low Elevation, Regular Grade
100000
10000
1000
D)
E
o
LU
100 -=
^ 10
o
CO
CD
>
1 -=
0.1
0.01
0.001
V
O
V
1 IS. t>
'v
A
^
0 ^
'V
^ [^
^
D
O
A
V
>
3
C
e
Oxygenates
BTEX
Other Aromatics
Naphthenes (all)
Paraffins
Iso-Paraffins
n Olefins
Iso Olefins
Di-Olefins
048
Mass Percent in Fuel
Figure 26 Effective solubility of hydrocarbon groups and oxygenates for the average
conventional, low-elevation, no-MTBE-ban, regular gasolines.
12
55
-------
RFC, No MTBE Ban, Low Elevation, Regular Grade
D)
E
100000
10000
1000
100
UJ
j=i 10
o
C/}
CD
> 1
ts
0.1
0.01
0.001
o
A
V
>
3
c
e
Oxygenates
BTEX
Other Aromatics
Naphthenes (all)
Paraffins
Iso-Paraffins
n-Olefins
Iso-Olefins
Di-Olefins
048
Mass Percent in Fuel
Figure 27 Effective solubility of hydrocarbon groups and oxygenates for the average
reformulated, low-elevation, no-MTBE-ban regular gasolines.
12
56
-------
Table 15 Predicted higher aqueous constituent concentrations (>0.05 mg/L)
of RFG (regular grade, no-MTBE-ban, low-elevation) in comparison to
conventional gasoline.
Class
Oxygenates
Oxygenates
Oxygenates
Oxygenates
Oxygenates
Oxygenates
Oxygenates
Oxygenates
Naphtheno-Olefins
Naphtheno-Olefins
n-Olefins
[so-Olefins
n-Olefins
n-Olefins
Mono-Naphthenes
n-Olefins
n-Olefins
Naphtheno-Olefins
n-Olefins
[so-Olefins
Chemical
Methyl-t-butylether
t-Butanol
TAME
2-Butanol
n-Butanol
i-Propanol
n-Propanol
i-Butanol
1 -Methylcyclopentene
Cyclopentene
t-Butene-2
2-Methylbutene-2
t-Pentene-2
c-Butene-2
Cyclopentane
c-Pentene-2
Butene-1
Cyclohexene
t-Hexene-2
2 -Methy Ipentene -2
Change in Concentration (mg/L)
1580
42.2
4.89
4.70
3.80
3.78
2 92
1.59
0.50
0.48
0.33
0.27
0.26
0.25
0.15
0.11
0.09
0.07
0.06
0.06
57
-------
Table 16 Predicted lower aqueous constituent concentrations (>0.05 mg/L)
of RFG (regular grade, no MTBE ban, low elevation) in comparison to
conventional gasoline.
Class
Mono-Aromatics
Mono-Aromatics
Mono-Aromatics
Mono-Aromatics
Mono-Aromatics
Mono-Aromatics
Paraffin
[so -Paraffin
Mono-Aromatics
Mono-Aromatics
Paraffin
Mono-Aromatics
Mono-Aromatics
Di-Olefms
Mono-Aromatics
Mono-Aromatics
Mono-Aromatics
Di-Olefms
Di-Olefins
Chemical
Toluene
Benzene
m-Xylene
o-Xylene
Ethylbenzene
p-Xylene
n-Butane
i-Pentane
1 ,2,4-Trimethylbenzene
1 -Methyl-3-ethylbenzene
n-Pentane
1 ,2,3-Trimethylbenzene
1 -Methyl-2-ethylbenzene
2-Methyl- 1 ,3 -Butadiene
1 ,3,5-Trimethylbenzene
1 -Methyl-4-ethylbenzene
n-Propylbenzene
lc/t,4-Hexadiene
lt,3-Pentadiene
Change in Concentration (mg/L)
-17.80
-17.80
-2.16
-1.58
-1.00
-0.90
-0.83
-0.74
-0.52
-0.25
-0.19
-0.16
-0.16
-0.13
-0.12
-0.11
-0.09
-0.07
-0.06
58
-------
100000
10000
1000
RFG, MTBE Ban, Low Elevation, Regular Grade
UJ
100
D)
>.
-i»
!5
-^ 10
o
(f)
CD
O
0.1
0.01
0.001
V
a
o
A
V
>
3
Oxygenates
BTEX
Other Aromatics
Naphthenes
Paraffins
Iso-Paraffins
Olefms
Iso-Olefins
Di-Olefms
048
Mass Percent in Fuel
Figure 28 Effective Solubility for hydrocarbon groups and oxygenates for the Illinois
regular gasolines (low elevation, MTBE ban, RFG).
12
59
-------
Table 17 Predicted higher aqueous constituent concentrations (>0.05
mg/L) of MTBE-ban RFG (regular grade, low elevation) in comparison
to non-MTBE ban RFG gasoline.
Class
Oxygenates
Oxygenates
Mono-Aromatics
Paraffin
[so -Paraffins
Paraffin
Chemical
Ethanol
i-Butanol
Benzene
n-Butane
i-Pentane
n-Pentane
Change in Concentration (mg/L)
61080
2.69
2.57
1.47
0.60
0.39
60
-------
Table 18 Predicted lower aqueous constituent concentrations (>0.05 mg/L) of
MTBE-ban RFG (regular grade, low elevation) in comparison to non-MTBE ban
RFG gasoline.
Class
Oxygenates
Oxygenates
Mono-Aromatics
Oxygenates
Oxygenates
Oxygenates
Oxygenates
Oxygenates
Mono-Aromatics
Mono-Aromatics
Mono-Aromatics
\/lono-Aromatics
Naphtheno-Olefins
n-Olefins
n-Olefins
n-Olefins
so-Olefins
Naphtheno-Olefins
Mono-Aromatics
\/lono-Aromatics
n-Olefins
Mono-Naphthenes
n-Olefins
Iso-Olefins
\/lono-Aromatics
Indenes
Mono-Aromatics
\/lono-Aromatics
Di-Olefins
Mono-Aromatics
Iso-Paraffins
Iso-Paraffins
Iso-Olefins
n-Olefins
Di-Olefins
Chemical
Methyl-t-butylether
-Butanol
Toluene
TAME
2-Butanol
n-Butanol
-Propanol
n-Propanol
m-Xylene
o-Xylene
Ethylbenzene
D-Xylene
Cyclopentene
c-Butene-2
-Butene-2
-Pentene-2
2-Methylbutene-2
1 -Methylcyclopentene
1 ,2,4-Trimethylbenzene
1 -Methyl-3-ethylbenzene
c-Pentene-2
Cyclopentane
3entene-1
2-Methylbutene-1
1 -Methyl-2-ethylbenzene
2,3-Dihydroindene
1 ,2,3-Trimethylbenzene
1 -Methyl-4-ethylbenzene
2-Methyl-1 ,3-Butadiene
1 ,3,5-Trimethylbenzene
3-Methylpentane
2-Methylpentane
2-Methylpentene-2
-Hexene-2
1t,3-Pentadiene
Change in
Concentration (mg/L)
-1606
-44.28
-10.26
-5.36
-4.70
-3.94
-3.78
-2.92
-1.73
-1.16
-1.05
-0.73
-0.45
-0.44
-0.41
-0.33
-0.32
-0.28
-0.26
-0.20
-0.17
-0.16
-0.15
-0.14
-0.11
-0.10
-0.09
-0.07
-0.07
-0.07
-0.07
-0.06
-0.06
-0.06
-0.06
61
-------
Predicted Gas Phase Concentrations
The predicted gas phase concentrations resulting from direct contact between the fuel and
air for the average conventional, no-MTBE-ban, low-elevation, regular-grade gasoline are shown
in Figure 29. The highest concentrations (greater than 100 mg/L) were found from paraffins and
iso-paraffins. In the next lower category (10 mg/L to 100 mg/L) are more paraffins, n-olefins,
and iso-olefms. BTEX and other classes of compounds appear between 1 mg/L and 10 mg/L.
Figure 30 shows the average gas phase concentrations from the similar RFG, no-MTBE-
ban samples. The main, obvious difference in the plots is the presence of MTBE at high weight
% and high concentration. Tables 19, 20 , and 21 show the increased and decreased gas phase
concentrations for a shift from conventional to RFG fuel. MTBE, and a variety of iso-olefms
have increased concentration, while various aromatics and others decrease.
Concentrations for Illinois samples (MTBE ban) are shown in Figure 31. The general
pattern shown is similar to the other fuels. Ethanol has a lowered concentration compared to
MTBE. Concentrations that increase (Table 22) are ethanol and several iso-paraffins, indicating
a need to increase the octane number, among others. Decreasing (Tables 23 and 24)
concentrations were noted for MTBE and a variety of other compounds.
62
-------
Conventional, No MTBE Ban, Low Elevation, Regular Grade
D)
£
C
o
"ro
"c
0
o
c
o
O
0
0)
CO
CO
O
1000 -^
100
V
0.001
0.01
V
>
O
A
V
>
3
e
Oxgenates
BTEX
Other Aromatics
Naphthenes (all)
Paraffins
Iso-Parafins
n Olefins
Iso-Olefins
Di-Olefins
048
Mass Percent in Fuel
Figure 29 Gas phase concentration of hydrocarbon groups and oxygenates for the
average conventional, low-elevation, no-MTBE-ban regular gasolines.
12
63
-------
O)
o
'-I
TO
-------
Table 19 Predicted higher gas phase constituent concentrations (>0.05 mg/L) of
RFG (regular grade, low elevation) in comparison to conventional gasoline.
Class
Oxygenates
Iso-Paraffins
n-Olefins
Paraffin
n-Olefins
n-Olefins
n-Olefins
Iso-Olefins
n-Olefins
Iso-Paraffins
n-Olefins
Naphtheno-Olefins
Mono-Naphthenes
Oxygenates
Iso-Olefins
n-Olefins
Naphtheno-Olefins
Iso-Olefins
n-Olefins
Iso-Olefins
Iso-Olefins
n-Olefins
Mono-Naphthenes
Iso-Olefins
Iso-Olefins
Iso-Olefins
Iso-Paraffins
Iso-Olefins
n-Olefins
Mono-Naphthenes
Iso-Olefins
n-Olefins
Iso-Olefins
Iso-Olefins
n-Olefins
n-Olefins
Iso-Paraffins
n-Olefins
Mono-Naphthenes
Iso-Paraffins
Iso-Olefins
Iso-Olefins
Iso-Olefins
Iso-Olefins
Chemical
Methyl-t-butylether
-Butane
-Butene-2
n-Hexane
c-Butene-2
-Pentene-2
Butene-1
2-Methylbutene-2
c-Pentene-2
3-Methylpentane
-Hexene-2
1 -Methylcyclopentene
Cyclopentane
TAME
2-Methylpentene-1
Pentene-1
Cyclopentene
2-Methylpentene-2
-Hexene-3
3,3-Dimethylpentene-1
3-Methyl-c-pentene-2
c-Hexene-2
Methylcyclohexane
4-Methyl-t-pentene-2
2-Methyl-2-hexene
3-Methyl-t-hexene-3
3-Methylhexane
3-Methylbutene-1
Hexene-1
1 c,3-Dimethylcyclopentane
5-Methyl-c-hexene-2
c-Heptene-2
4-Methyl-t/c-hexene-2
4-Methylpentene-1
-Heptene-3
c-Heptene-3
2,2,5-Trimethylhexane
-Heptene-2
Ethylcyclopentane
2-Methylhexane
3-Methyl-t-hexene-2
2-Methyl-t-hexene-3
3-Methyl-c-hexene-2
1 ,5-DM-Cyclopentene
Change in Concentration
(mg/L)
152.5
4.97
3.72
3.18
2.82
2.50
1.72
1.37
1.06
1.03
0.82
0.78
0.68
0.67
0.60
0.58
0.55
0.50
0.50
0.48
0.41
0.34
0.28
0.26
0.24
0.23
0.22
0.21
0.21
0.15
0.13
0.13
0.13
0.12
0.12
0.12
0.12
0.11
0.11
0.11
0.11
0.10
0.10
0.09
65
-------
Table 20 (Continuation) Predicted higher gas phase constituent concentrations
(>0.05 mg/L) of RFG (regular grade, low elevation) in comparison to conventional
gasoline.
Class
Naphtheno-Olefins
Iso-Olefins
Mono-Naphthenes
Iso-Olefins
Mono-Naphthenes
Mono-Naphthenes
Oxygenates
so-Olefins
Iso-Olefins
Chemical
Cyclohexene
4-Methyl-c-pentene-2
1t,3-Dimethylcyclopentane
2-Ethyl-3-methylbutene-1
1t,2-Dimethylcyclopentane
1 ,1 ,2-Trimethylcyclopentane
t-Butanol
3-Methyl-c-hexene-3
3-Ethylpentene-2
Change in Concentration
(mg/L)
0.09
0.09
0.08
0.08
0.07
0.06
0.05
0.05
0.05
66
-------
Table 21 Predicted lower gas phase constituent concentrations (>0.05 mg/L) of RFG
(regular grade, low elevation) in comparison to conventional gasoline.
Class
Paraffin
so-Paraffins
Paraffin
Mono-Aromatics
Mono-Aromatics
Iso-Paraffins
so-Paraffins
Iso-Paraffins
Iso-Paraffins
so-Olefins
Mono-Aromatics
Iso-Paraffins
\/lono-Aromatics
Mono-Aromatics
\/lono-Aromatics
Di-Olefins
Iso-paraffins
Di-Olefins
Iso-paraffins
Mono-Aromatics
Iso-paraffins
Di-Olefins
\/lono-Aromatics
Iso-paraffins
Iso-paraffins
3araffin
Chemical
n-Butane
i-Pentane
n-Pentane
Toluene
Benzene
2,2-Dimethylbutane
2,2,4-Trimethylpentane
2,3-Dimethylbutane
2,2-Dimethylpropane
2-Methylbutene-1
m-Xylene
2,3,4-Trimethylpentane
Ethylbenzene
o-Xylene
p-Xylene
1c/t,4-Hexadiene
2,4-Dimethylpentane
2-Methyl-1 ,3-Butadiene
2,4-Dimethylhexane
1 ,2,4-Trimethylbenzene
2,3-Dimethylhexane
1t,3-Pentadiene
1 -Methyl-3-ethylbenzene
3-Methylheptane
2-Methylpentane
n-Octane
Change in Concentration
(mg/L)
-62.91
-45.31
-9.36
-3.46
-2.98
-2.22
-1.43
-1.05
-0.47
-0.43
-0.41
-0.27
-0.25
-0.22
-0.17
-0.16
-0.14
-0.12
-0.11
-0.07
-0.07
-0.07
-0.06
-0.06
-0.06
-0.05
67
-------
RFG, MTBE Ban, Low Elevation, Regular Grade
D)
E,
c
o
'-I
05
0
o
c
o
o
0
0)
05
(0
O
1000
V
100 -=
10
0.1
0.01
0.001
V
O
a
o
A
V
>
3
C
6
Oxygenates
BTEX
Other Aromatics
Naphthenes
Paraffins
Iso-Paraffins
Olefins
Iso-Olefins
Di-Olefins
048
Mass Percent in Fuel
Figure 31 Gas phase concentration for hydrocarbon groups and oxygenates for the
Illinois regular gasolines (low elevation, MTBE ban, RFG).
12
68
-------
Table 22 Predicted higher gas phase constituent concentrations (>0.05 mg/L) of
MTBE-ban RFG (regular grade, low elevation) in comparison to non MTBE-ban
RFG.
Class
Paraffin
Iso-paraffins
3araffin
Oxygenates
Iso-paraffins
so-paraffins
Iso-paraffins
so-Olefins
Iso-paraffins
Iso-paraffins
so-paraffins
Mono-Aromatics
Iso-paraffins
Iso-paraffins
Mono-Naphthenes
\/lono-Naphthenes
Mono-Naphthenes
Mono-Naphthenes
so-paraffins
Mono-Naphthenes
Iso-Olefins
Chemical
n-Butane
-Pentane
n-Pentane
Ethanol
2,2,4-Trimethylpentane
2,2-Dimethylpropane
2,4-Dimethylpentane
3-Methylbutene-1
2-Methylhexane
2,3,4-Trimethylpentane
2-Methyl-3-ethylpentane
Benzene
2,3-Dimethylbutane
2,4-Dimethylhexane
1 c,2c,3-Trimethylcyclopentane
1t,2-Dimethylcyclopentane
Cyclohexane
1t,4-Dimethylcyclohexane
2,5-Dimethylhexane
3c-Ethylmethylcyclopentane
2-Methyl-c-hexene-3
Change in Concentration
(mg/L)
111.13
36.77
19.22
13.04
3.45
1.68
1.33
0.90
0.62
0.57
0.48
0.43
0.34
0.15
0.13
0.12
0.11
0.10
0.09
0.06
0.06
69
-------
Table 23 Predicted lower gas phase constituent concentrations (>0.05 mg/L) of
MTBE-ban RFG (regular grade, low elevation) in comparison to non MTBE-ban
RFG.
Class
Oxygenates
so-paraffins
Iso-paraffins
n-Olefins
n-Olefins
Iso-paraffins
n-Olefins
Paraffin
Iso-paraffins
\/lono-Aromatics
n-Olefins
n-Olefins
so-Olefins
Iso-Olefins
Iso-paraffins
n-Olefins
Mono-Naphthenes
Oxygenates
n-Olefins
n-Olefins
so-Olefins
Naphtheno-Olefins
Iso-Olefins
Naphtheno-Olefins
Iso-Olefins
so-Olefins
n-Olefins
Mono-Aromatics
n-Olefins
Iso-Olefins
Mono-Aromatics
Iso-paraffins
Iso-Olefins
Iso-Olefins
Mono-Aromatics
Iso-Olefins
3araffin
Mono-Aromatics
n-Olefins
Iso-Olefins
n-Olefins
so-Olefins
n-Olefins
Iso-Olefins
so-Olefins
Chemical
Methyl-t-butylether
-Butane
2-Methylpentane
c-Butene-2
t-Butene-2
3-Methylpentane
t-Pentene-2
n-Hexane
2,2-Dimethylbutane
Toluene
Pentene-1
c-Pentene-2
2-Methylbutene-2
2-Methylbutene-1
3-Methylhexane
-Hexene-2
Cyclopentane
TAME
Butene-1
t-Hexene-3
2-M et.hy I pe nte n e-2
Cyclopentene
3,3-Dimethylpentene-1
1 -Methylcyclopentene
3-Methyl-c-pentene-2
2-Methylpentene-1
c-Hexene-2
m-Xylene
Hexene-1
4-Methyl-t-pentene-2
Ethylbenzene
3,3-Dimethylpentane
2-Methyl-2-hexene
4-Methylpentene-1
o-Xylene
5-Methyl-c-hexene-2
n-Heptane
p-Xylene
c-Heptene-3
3-Methyl-t-hexene-2
c-Heptene-2
4-M ethy l-t/c-h ex e n e-2
t-Heptene-2
3-Methyl-c-hexene-2
4-Methyl-c-pentene-2
Change in Concentration
(mg/L)
-155.16
-7.11
-5.17
-5.00
-4.66
-4.20
-3.15
-3.14
-2.99
-1.99
-1.92
-1.65
-1.62
-1.36
-1.08
-0.82
-0.75
-0.74
-0.72
-0.56
-0.54
-0.52
-0.48
-0.45
-0.45
-0.44
-0.33
-0.33
-0.33
-0.31
-0.26
-0.24
-0.22
-0.20
-0.16
-0.16
-0.15
-0.14
-0.13
-0.13
-0.13
-0.13
-0.12
-0.11
-0.11
70
-------
Table 24 (Continuation) Predicted lower gas phase constituent
concentrations (>0.05 mg/L) of MTBE-ban RFG (regular grade, low
elevation) in comparison to non MTBE-ban RFG.
Class
n-Olefins
so-Olefins
Iso-Olefins
Iso-Olefins
so-Olefins
Iso-Olefins
Mono-Naphthenes
Di-Olefins
Di-Olefins
Naphtheno-Olefins
so-Olefins
Oxygenates
Mono-Naphthenes
Iso-Olefins
Mono-Aromatics
Chemical
t-Heptene-3
2-Methyl-t-hexene-3
1 ,5-DM-Cyclopentene
3-Methyl-t-hexene-3
2-Methyl-c-hexene-3
2-Ethyl-3-methylbutene-1
1 ,1 ,2-Trimethylcyclopentane
2-Methyl-1 ,3-Butadiene
1t,3-Pentadiene
Cyclohexene
3-Methyl-c-hexene-3
t-Butanol
Ethylcyclopentane
3-Ethylpentene-2
1 -Methyl-3-ethylbenzene
Change in Concentration
(mg/L)
-0.10
-0.09
-0.09
-0.08
-0.08
-0.07
-0.07
-0.06
-0.06
-0.06
-0.06
-0.06
-0.06
-0.05
-0.05
Figure 32 shows one final example where the predicted gas phase concentration of the
average conventional, no MTBE ban, low elevation, regular grade gasoline components were
plotted against their effective solubilities. For reference a line of equal gas phase concentration
and effective solubility was plotted. The gas phase results are based upon vapor pressure (i.e.,
direct volatilization from the fuel) and the oxygenates plotted below the equal-concentration line,
as did many naphthenes and aromatics. The BTEX compounds, paraffins, Iso-Paraffins and all
three types of olefins plotted above the line, indicating that their presence is greater in the gas
phase than in the water phase. Notably, in the preceding figures of gas phase concentration
(Figures 29, 30, and 31) there was a shift upward of paraffins and Iso-Paraffins relative to the
plots for the effective solubility (Figures 26, 27 and 28).
71
-------
Conventional gasoline, No MTBE, Low Elevation, Regular Grade
D)
E,
O
"CD
"c
CD
O
O
O
CD
CD
.c
Q_
CD
1000 -^
100
10
0.1
0.01
0.001
-
0.0001 -=
E
1E-005 -=
E
1E-006 -=
=
1 P nnv
I Q-UU /
>^Av ^f^ o
/
/
/
/
/
/
O
o
A
V
>
3
e
Oxygenates
BTEX
Other Aromatics
Naphthenes
Paraffins
Iso-paraffins
Olefins
Iso-Olefins
Di-Olefins
/
i mm i i mill i i mill i i mm i i mm i i mm i i mm i i inn
i nun i i inn
1E-007 1E-006 1E-005 0.0001 0.001 0.01
0.1
10
100 1000
Effective Solubility (mg/L)
Figure 32 Scatter plot of predicted gas phase concentration versus the effective solubility
of the average conventional, low-elevation, no-MTBE-ban, regular gasoline.
72
-------
Discussion and Conclusions
Data were collected on 312 components of gasoline based on 96 samples collected from
across the United States. The sample locations were selected to include RFG and conventional
gasoline, high and low elevation, regular and premium grades, likely ethanol-using locations and
MTBE-ban states. Evaluation of single compound concentrations lead to several findings:
1) The benzene concentration of Federal RFG fuels were consistently below 1% by weight. This
result follows from federal requirements and agrees with data collected by EPA for compliance
(US EPA, 2005). In conventional fuels, higher benzene concentrations of roughly 0.5% to 3%
were found. These values generally agree with data generated by industry (Dickson, 2004a,
2004b). Data on conventional fuels available from EPA show the average benzene concentration
to range from 1.11 % to 1.17% by volume for summer and 1.08% to 1.15% by volume for winter
for the years from 1997 to 2002. These data are not necessarily in conflict with the current study
data as the averages are based on many more samples and include more locations across the
country. The EPA data show that benzene ranged from 0.01 % to about 5.0% by volume for
summer and winter samples in 2002 and 2003 (John Weihrauch, U.S. EPA, Office of
Transportation and Air Quality, 2005, personal communication).
A) In Colorado the state oxygenate requirement did not limit the amount of benzene in
the fuel while it required an oxygen content of 3.1 %, which is higher than the federal
requirement.
2) Where MTBE bans are in place, ethanol has replaced MTBE as the main oxygenate in use.
Ethanol was not found in other samples suggesting it was used only where needed to meet an
oxygenate requirement.
3) Only low concentrations of TEA were found in the fuels and were associated with MTBE.
The gasoline data generated for this study represent seven combinations of elevation, RFG
requirements, and MTBE ban status. The regular gasolines showed more similarity given a
particular combination of these factors, than did the premiums. This result suggests that refiners
use more approaches to raise the octane rating of the fuels. Statistical analysis of the data support
the division of the samples by grade, elevation, MTBE ban status and RFG requirements for some
combinations of the factors. Differences due to state RFG requirements, regional use of
oxygenates for octane enhancement and lack of commonality among high elevation fuels, require
specific local data to achieve the goal of prediction of release composition. The need to predict
release composition follows from the lack of detailed characterization of most fuels and lag time
between release and field sampling. Both of these are expected to continue in the future.
The environmental impact of these fuels depends on the exposure pathway. When contact
with contaminated soils is a significant pathway, the components of the fuel are themselves
directly important. If exposure occurs through ground water or soil gas, then the concentrations
in these media indicate the importance of the various components of the fuel. Effective
73
-------
solubilities and gas phase concentrations were approximated from estimates of solubility vapor
pressure and the individual gasoline component concentration data. In several examples, these
showed that a smaller number of components result in high (say greater than 1 mg/L)
concentrations in these media. For ground water the highest concentrations were found for the
oxygenates and BTEX. For the gas phase concentration, the highest were found for paraffins and
oxygenates, with the BTEX occurring at lower concentration. These only provide a rough guide,
because the toxicity of the compound and the exposure ultimately determines the significance of
these chemicals.
74
-------
References
40 CFR 80, Code of Federal Regulations, Title 40 Protection of Environment, Part 80 -
Regulations of Fuels and Fuel Additives
American Petroleum Institute, 2004, State MTBE Activity Laws, 6/14/04, Unpublished
document.
American Society for Testing and Materials, 1958, Knocking Characteristics of Pure
Hydrocarbons, ASTM Special Publication No. 225, Philadelphia, 96 pp.
American Society for Testing and Materials, 2002, Standard Test Method for Determination of
Individual Components in Spark Ignition Engine Fuels by 100 Meter Capillary High
Resolution Gas Chromatography, ASTM International, West Conshohocken,
Pennsylvania, D 6729-1.
California Air Resources Board, 2003, The California Reformulated Gasoline Regulations, Title
13, California Code of Regulations, Sections 2250-2273.5, May 1, 2003.
Canadian Petroleum Products Institute, 1994, Composition of Canadian Summer and Winter
Gasolines 1993, CPPI Report No. 94-5.
Chevron U.S.A. Inc, 2005, Motor Gasolines Technical Review, www.chevron.com.
Colorado, 2002, Regulation 13, The Reduction of Carbon Monoxide Emissions from Gasoline
Powered Motor Vehicles through the use of Oxygenated Gasolines.
Dickson, C. L., 2004, Motor Gasolines, Winter 2003-04, Northrup Grumman Mission Systems,
Bartlesville, Oklahoma, NGMS-235 PPS 2004/3, 67pp.
Dickson, C. L., 2004, Motor Gasolines, Summer 2003, Northrup Grumman Mission Systems,
Bartlesville, Oklahoma, NGMS-233 PPS 2004/1, 67pp.
Gustafson, J.B., J.G. Tell, and D. Orem, 1997, Selection of Representative TPH Fractions Based
on Fate and Transport Considerations, Total Petroleum Hydrocarbons Criteria Working
Group Series Volume 3, Amherst Scientific Publishers, Amherst Massachusetts, 102pp.
Hilal, S.H., S.W. Karickhoff and L.A. Carreira 2003, Verification and Validation of the SPARC
Model, US EPA, Washington, D.C, EPA/600/R-03/033.
Jokuty, P., S. Whiticar, Z. Wang, M. Fingas, B. Fieldhouse, P. Lambert and J. Mullin, Properties
of Crude Oils and Oil Products, Manuscript Report EE-165, Environmental Protection
Service, Environment Canada, Ottawa, Ontario, 1999. Available at http://www.etc-
cte.ec.gc.ca/databases/OilProperties
75
-------
Jolliffee, IT. 2002. Principle component analysis. Springer-Verlag, New York. 487 p.
Jolliffee, IT. 1970. Redundant variables in multivariate analysis. Unpublished PhD.
Dissertation. University of Sussex.
Jolliffee, IT. 1972. Discarding variables in principal component analysis 1: Artificial data.
Appl. Statist (21): 160-173 p.
Jolliffee, IT. 1972. Discarding variables in principal component analysis 2: Real data. Appl.
Statist (22):21-31 p.
Karickhoff, S.W., V.K. McDaniel, C. Melton, A. N. Vellino, D. E. Nute, L.A. Camera, 1991,
Predicting chemical reactivity by computer, Environmental Toxicology and Chemistry,
10, 1405-1416.
Kendall, M. 1975. Multivariate analysis. Hafner Press, New York. 210 p.
Khattree, R., and D. N. Naik. 2000. Multivariate data reduction and discrimination with sas
software. SAS Institute, Cary, NC. 558 p.
Maine Department of Environmental Protection, 2005, 2004 Maine Fuels Report, Agusta, Maine.
Meusinger, R and R. Moros, 2001, Determination of octane numbers of gasoline compounds from
their chemical structure by 13C NMR spectroscopy and neural networks, Fuel, 80, 613-
621.
Rencher, A.C., 1995, Methods of Multivariate Analysis, Wiley, New York, 627pp.
Schwarzenbach, R.P, P.M. Gschwend, and D.M. Imboden, 2003, Environmental Organic
Chemistry, 2nd ed., Wiley, New York, 1313pp.
Shaw, D.G, 1989, Solubility Data Series, Hydrocarbons with Water and Sea Water, Part 1:
Hydrocarbons C5 to C7, Volume 37, International Union of Pure and Applied Chemistry,
Pergammon Press.
Shaw, D.G, 1989, Solubility Data Series, Hydrocarbons with Water and Sea Water, Part 2:
Hydrocarbons C8 to C36, Volume 38, International Union of Pure and Applied Chemistry,
Pergammon Press.
Stout, S.A., A.D. Uhler, R.M. Uhler, E.M. Healey, and K.J. McCarthy, 2003, Detailed chemical
fingerprinting of gasoline for environmental forensic investigations: Part 3. Applications
to environmental forensics investigations, Soil, Sediment and Water, March/April, 16-18.
Strikkers, D.E., 2002, Octane and the environment, The Science of the Total Environment, 299,
76
-------
37-56.
United States Environmental Protection Agency, 2005, Office of Transportation and Air Quality,
http ://www. epa. gov/otaq.
United States Environmental Protection Agency, 2005b, Where You Live Map of Current RFG
Areas, Office of Transportation and Air Quality.
Washington State, 1996, Washington Administrative Code: Chapter 173-492 WAC: Motor Fuel
Specifications for Oxygenated Gasoline.
Weaver, J.W., RJ. Charbeneau, J.D. Tauxe, B.K. Lien, and J.B. Provost, 1994, The
Hydrocarbon Spill Screening Model (HSSM) Volume 1: User's Guide, US EPA,
EPA/600/R-94/039a.
Weaver, James W., 2004, On-line Tools for Assessing Petroleum Releases, United States
Environmental Protection Agency, Washington, D.C., EPA 600/R-04/101.
77
-------
Appendix: Theoretical Basis of the Principal Components Analysis
Following the theory presented by Jolliffe (2002), let X be a vector of p random variables,
such that the variances of the/? random variables and the structure of the covariances or
correlations between the/? variables are of interest. Unless/? is small, it is often unbeneficial to
look at the/? variances and all of the 0.5/?(p-l) correlations or covariances. An alternative
approach is to find a few (less than/?) variables that preserve most of the information given by
these variances and correlations or covariances. The initial step is to find a linear function c^1 X
of the elements of X having maximum variance, where ccj is a vector of/? constants an, a12,..., alp,
and T denotes the transpose, such that
iy*/ (5)
Next, find a linear function «ST X which is uncorrelated with c^1 X having maximum variance,
and continuing repeating this sequence so that at the Mi stage a linear function akT X is found that
has maximum variance subject to being uncorrelated with c^1 X, CCST X, ... ,cckT X. Up top PC
could be found, however it is hoped that the majority of the variation in X will be accounted for
by m PC, where m is less than p.
The first step in PCA is deciding whether to use the correlation matrix or covariance
matrix. Use of the correlation matrix forces all of the variables to have equal variance and this in
turn may defeat the purpose of identifying those variables that contribute more significantly to the
total variability (Khattree and Naik, 2000). The PC of a PCA based on covariance matrices are
also insensitive to units of measurement on the same scale. Since all of the gasoline variables are
measured on a percent weight basis, ensures the PC will be interpretable and unbiased by those
variables having large variance. PCA in this work was done utilizing the covariance matrix of the
variables.
There are two commonly used methods for determining the number of PC to select. The
first, and most commonly used, is based on the cumulative proportion of total variance. An
appropriate minimum percentage of total variation desired to be explained by the PCA is
prespecified, and the smallest number of principal components that satisfies this criterion is
selected. The prespecified percentage is usually taken to be 95 percent variation explained. The
second method is graphical and uses what is called a scree diagram. A scree diagram is a plot of
the variances against the/? PC. From the plot, the number of m PC to be selected is determined in
such a way that the slope of the graph is steep to the left of m, but at the same time not steep to
the right. The idea being that the number of PC to be selected is such that the differences between
consecutive variances are becoming increasingly smaller. Both of the techniques described above
will be used in determining an appropriate number of PC.
When the number of variables/? is large, it is often the case that a subset of m variables,
where m
-------
been cited that if Xcan successfully be described by m PC, then Xcan be replaced by a subset of
m variables (Jolliffe 1970, 1972, 1973). The variable corresponding to the maximum of the
absolute values of the coefficients in the first PC is selected first. The second variable selected is
that variable (if not already selected) whose coefficient in the second PC has the maximum
absolute value. If the variable corresponding to the maximum coefficient was already selected
then we choose the variable corresponding to the next maximum. Proceeding this way, we select
a subset of required size from the list of original variables.
Given the data have/? variables the observations can be plotted as points in/?-dimensional
space. Two-dimensional plots are of use in detecting patterns in the data. If the data lie close to a
two dimensional subspace, plots of the m
------- |