Applied Energy 125 (2014) 206217

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

A parametric study of light-duty natural gas vehicle competitiveness

in the United States through 2050
Meghan B. Peterson , Garrett E. Barter, Todd H. West, Dawn K. Manley
Sandia National Laboratories, P.O. Box 969, Livermore, CA 94551, United States

h i g h l i g h t s

 NGVs are economical, but limited by infrastructure and OEM model availability.
 NGVs compete more with EVs than conventional vehicles.
 By displacing EVs, NGVs offer little or negative GHG reduction benets.
 Public refueling infrastructure is a better investment than home CNG compressors.
 Bi-fuel vehicles can be a bridge technology until infrastructure build-out.

a r t i c l e i n f o a b s t r a c t

Article history: We modeled and conducted a parametric analysis of the US light-duty vehicle (LDV) stock to examine the
Received 4 November 2013 impact of natural gas vehicles (NGVs) as they compete with electric vehicles, hybrids, and conventional
Received in revised form 28 February 2014 powertrains. We nd that low natural gas prices and sufcient public refueling infrastructure are the key
Accepted 22 March 2014
drivers to NGV adoption when matched with availability of compressed natural gas powertrains from
automakers. Due to the time and investment required for the build out of infrastructure and the introduc-
tion of vehicles by original equipment manufacturers, home natural gas compressor sales and bi-fuel
Keywords:
NGVs serve as bridge technologies through 2030. By 2050, however, NGVs could comprise as much as
Natural gas vehicle
Compressed natural gas
20% of annual vehicle sales and 10% of the LDV stock fraction. We also nd that NGVs may displace elec-
Greenhouse gas emission tric vehicles, rather than conventional powertrains, as they both compete for consumers that drive
enough miles such that fuel cost savings offset higher purchase costs. Due to this dynamic, NGVs in
our LDV stock model offer little to no greenhouse gas emissions reduction as they displace lower emis-
sion powertrains. This nding is subject to the uncertainty in efciency technology progression and the
set of powertains and fuels considered.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Recent advances in drilling technologies and new domestic
shale gas discoveries have dramatically increased natural gas
(NG) production in the United States (by 15% from 2008 to
2012). This boom in domestic supply has dropped the price of nat-
ural gas signicantly (by 50% from 2008 to 2012) [1,2]. The attrac-
tiveness of natural gas is due not only to its low price, but also its
lower carbon content relative to other fossil fuels (roughly 55% of
the carbon per unit energy as coal and 73% of petroleum) [3,4].
In the US energy economy, the usage of natural gas in the residen-
tial, commercial, industrial, and electricity generation sectors is
Corresponding author. Tel.: +1 925 294 3201 (work); fax: +1 925 294 3870.
E-mail address: mbpete@sandia.gov (M.B. Peterson).
growing and even nding new niches [1]. One sector of the US
economy where natural gas does not currently play a large role
is transportation (0.1% of total delivered energy in 2011) [1]. The
technology has long existed for natural gas to serve as a transpor-
tation fuel in the form of compressed natural gas (CNG), which can
be stored on the vehicle at high pressure, or in a cooled environ-
ment as liquid natural gas (LNG) [5]. There are also other, less
established pathways for natural gas to serve as a transportation
fuel. Natural gas can be chemically rened into liquid fuels, such
as ethanol, methanol, gasoline, and diesel. However, creating
drop-in replacement fuels from natural gas requires more process-
ing than other fossil fuel feedstocks, meaning it is more expensive
and less efcient [3]. Finally, natural gas can also be processed with
steam methane reforming to produce hydrogen for hydrogen fuel-
cell vehicles [6]. In the scope of this work, however, the only nat-
ural gas derived fuel pathway considered is CNG, as the others

http://dx.doi.org/10.1016/j.apenergy.2014.03.062
0306-2619/ 2014 Elsevier Ltd. All rights reserved.
 M.B. Peterson et al. / Applied Energy 125 (2014) 206217
are less well established and must still overcome some technolog-
ical and regulatory hurdles.
The lack of refueling infrastructure, the high incremental price
of NGVs, and limited CNG powertrains offered by automakers have
historically suppressed demand [5]. However, this reality is rapidly
changing for short-range, heavy duty vehicles (HDVs) that are cen-
trally refueled, such as buses, refuse haulers, or delivery trucks. For
these vocations, eet owners can build a private refueling infra-
structure and offset higher vehicle purchase costs with fuel cost
savings in three years or less. There is also interest in strategically
placing CNG or LNG refueling stations along interstate highways to
enable long-haul, publicly refueled HDVs to rely on natural gas as
well [3]. The merits of heavy-duty NGVs, however, are not the fo-
cus of this work, rather we focus on light-duty NGVs.
For light-duty vehicles (LDVs), natural gas is used as a transpor-
tation fuel almost exclusively as CNG, since LNG is impractical for
multiple reasons. This includes the need for a large, thermally insu-
lated fuel storage tank and high-mileage, continuous operation to
minimize the impacts of LNG evaporation [3]. Similar to HDVs,
the early adopters of light-duty NGVs will likely be eet operators
that can establish their own infrastructure, such as government
eets, taxis, and delivery vans [5]. However, for the broader
light-duty personal automobile market, the dearth of refueling
infrastructure is not as easily overcome, and the lower annual
vehicle miles traveled (VMT) compared to HDVs leads to longer
payback periods for consumers (on the order of 10 years versus
12 years) [3,7]. At the same time, LDV consumers are far less
focused on lifetime ownership costs compared to eet owners,
and therefore only consider a few years worth of costs in their pur-
chase decisions [811]. Moreover, light-duty NGVs will have to
compete with improved efciency conventional vehicles (such as
clean diesel or plugless hybrid electric vehicles, HEVs), as well as
other alternative fuel vehicles (AFVs), such as plug-in hybrid
electric vehicles (PHEVs), battery electric vehicles (BEVs), or E85
ex-fuel vehicles (FFVs). Many of these powertrains also face sim-
ilar challenges to NGVs. Thus, in order to gain signicant market
share, NGVs must therefore be an economically viable alternative
to the established and more efcient gasoline-fueled car, and also
a more attractive option than other AFVs.
Earlier surpluses in natural gas combined with high oil prices
and growing concerns about air quality and energy security gar-
nered much interest in NGVs in the 1980s and 1990s [1222]. In
particular, the Department of Energys Alternative Motor Fuels
Act gave rise to a number of reports that examined alternatives
to gasoline and diesel and discussed their costs, displacement po-
tential, and infrastructure challenges [2325]. Many of these stud-
ies highlight the same issues and challenges NGVs face today. A
couple recent studies have also assessed the viability of light-duty
NGVs in the US [3,5]. These studies concur that the fuel price dif-
ferential between oil and natural gas makes NGVs attractive to
consumers. However, they also agree that until CNG refueling
infrastructure becomes more widespread and automakers offer
more CNG models (thereby reducing incremental manufacturing
costs), the high purchase price will remain prohibitive. In these
studies, one limitation of the analysis approaches used is that
NGVs were considered as a single alternative to conventional vehi-
cles, instead of one of many alternative powertrains. One exception
is the National Petroleum Council study [5], which considered a
range of portfolios that included one or more other AFVs. Another
common analysis shortcoming is the limited set of scenarios con-
sidered to understand the thresholds of natural gas price, gasoline
price, technology progress, and consumer behavior for which NGVs
succeed or fail in the marketplace. Our work expands upon these
analyses by considering a broad parametric exploration of this
trade space, in addition to allowing NGVs to compete with all other
types of powertrains available.
207
The path forward for light-duty NGVs in the US must traverse a
number of uncertainties. There are uncertainties associated with
policies that drive the growth of refueling infrastructure, and fu-
ture regulation of carbon emissions [3]. There is also economic
uncertainty given the price volatility of oil and natural gas [5], both
of which determine the fuel cost savings offered by NGVs. Addi-
tionally, there is technological uncertainty in the performance
and cost of future NGVs (as well as other alternatives such as
PHEVs and BEVs), and in the long-term potential for shale gas to
produce signicant natural gas volume in a cost effective manner.
These uncertainties motivate this work to take a parametric ap-
proach to the analysis, similar to [26]. Thus, instead of dissecting
a limited number of future scenarios that differ by one or two
key parameters, we explore a broad trade space to ascertain the
set of conditions that promote or discourage light-duty NGV
adoption in the US. Moreover, this work examines the future pros-
pects of NGVs in the context of the entire portfolio of powertrains
(both alternative and improved efciency). This proves to be a key
point in the analysis, since these technologies will compete with
each other in addition to competing with conventional gasoline
vehicles.
2. Methods
The model tracks the evolution of the light-duty vehicle stock in
the US, its fuel usage, and corresponding demand for raw energy
stocks. The model is broken down into four sub-components, a
vehicle sub-model, a fuel production sub-model, an electricity grid
sub-model, and an energy supply sub-model. The sub-models ex-
change price and demand information for the energy supply stocks
and fuels considered. The model uses system dynamics concepts
(i.e. stocks, ows and feedback loops) to track the changes in var-
ious quantities over time and the interactions between variables.
No predetermined market share targets are assumed, thus technol-
ogies compete directly in the marketplace and are allowed to our-
ish or fail. The LDV stock is segmented by driver demographics
such as population density [27], daily trip distance distributions
[28], and housing type [27], as well as vehicle descriptors such as
state of registration [29], age [30], size [31], and powertrain [32].
Fuel and commodity prices also vary regionally. A detailed descrip-
tion of the model is found in [33], which builds upon an earlier ver-
sion of the model presented in [26]. More detailed extensions of
the previously published model to account for CNG powertrains
and home compressors are found in Appendix A.
Some of the essential model assumptions are reected in the re-
sults, and therefore worth summarizing here. We assume that
vehicles per capita is constant in time, thus the LDV stock grows
at the same rate as population, 0.9%/year [27]. This also implies
that the new vehicle sales rate and scrapping of older vehicles
are also xed in time. Similarly, we assume that the proportion
of vehicle size classes is xed over time. Furthermore, while miles
traveled by each vehicle (VMT) varies across different descriptors
of the vehicle and the driver, there is no change in VMT over time.
The total mileage driven by all vehicles in the LDV stock neverthe-
less increases as the number of on-road vehicles increases.
The vehicle cost to consumers is a generalized utility encom-
passing purchase costs, fuel costs, incentives, and econometric
penalties for range and infrastructure limitations. This cost is
amortized over a required payback period for consumers (due to
high implicit discount rates [10]) and then normalized by VMT.
We use the term, required payback period, for consumers in this pa-
per to refer to the number of years of ownership costs that con-
sumers consider when purchasing a vehicle. Note that the default
value of 3 years is far shorter than the true time it takes for fuel
cost savings to offset purchase premiums. Due to their inherent
208 M.B. Peterson et al. / Applied Energy 125 (2014) 206217
uncertainties, the valuation of the penalties, the required payback
period, and the choice sensitivity to price are all parameterized.
Consumers choose between 16 different powertrains in a nested,
multi-nomial logit choice model. These include conventional, hy-
brid, and plug-in hybrid (PHEV) variants (one with 10 miles all-
electric range, and another with 40) of gasoline (spark ignition,
SI), diesel (compression ignition, CI), and E85 powertrains. There
are additionally a 150 mile range battery electric powertrain, a
compressed natural gas (CNG) powertrain, a CNG hybrid power-
train, and a CNG-gasoline bi-fuel powertrain (CNGBI). After-market
conversions of one powertrain to another are not tracked. The
model does not include a vehicle systems analysis, but rather relies
on the powertrain templates provided by [34]. Thus individual
powertrain components and engine performance maps are exoge-
nous to the model.
Future commodity prices for oil, coal, and natural gas are taken
from [1]. Conversion efciencies between these and other sources
into vehicle fuels and electricity are provided by [35]. GHG emis-
sions rates, measured in CO2 equivalency, are also taken from the
same source. It should be noted, however, that there is great uncer-
tainty surrounding the measurement of greenhouse gases from
alternative fuels (particularly with respect to biofuels), depending
on the extent of the fuel life-cycle considered [23,36]. The set of pure
fuels in the model is motor gasoline, diesel, ethanol, CNG, and
electricity. The pump-fuels are E10, E85, and B20 blends, as well
as diesel, CNG, and electricity. We do not consider any other fuel,
such as a drop-in biofuel replacement for gasoline or biomethane.
The model tracks CNG infrastructure insofar as it tracks the num-
ber of CNG refueling stations in each region. The ratio of alternative
fuel pump stations to alternative fuel vehicles is denoted by the
Vehicle Refueling Index (VRI) [37], and is a surrogate for infrastruc-
ture growth in our model. The VRI is a policy-driven input parame-
ter that scales with the number of AFVs purchased in a given region
to determine the number of new alternative refueling stations
added. By parametrically adjusting the VRI, the model can explore
the dependency between NGV adoption and refueling infrastructure
growth. Thus, infrastructure investment is not directly captured, but
is assumed to dynamically respond to NGV sales.
The LDV stock is initialized in 2010 with 232 million private
light-duty vehicles in the continental United States. The simulation
evolves the LDV stock using 3rd order RungeKutta time stepping
through 2050.
3. Results and discussion
3.1. Model parameterization and baseline inputs
Since it is not feasible to independently vary every single input
parameter (due to variable inter-dependency and also tractability),
Table 1
Baseline, minimum, and maximum values for sensitivity analysis parameters. Also shown are Spearman rank correlation coefcients for outputs (columns) with respect to inputs
(rows). Output metrics are measured at simulation end, 2050.
Parameter Baseline Min
LDV stock growth rate 0.9% 0.5%
Vehicle sales rate 6.7% 5%
Carbon price ($/MT CO2 -equiv) 0 0
Oil price multiplier 1 0.25
Natural gas price multiplier 1 0.25
Required payback period (years) 3 2
Penalty multiplier 1 0
Battery cost multiplier 1 0.1
Combustion vehicle cost multiplier 1 0.9
Home ller cost reduction [per year] 3% 0%
Combustion powertrain eff multiplier 1 0.6
New stations/1 k new vehicles (VRI) 0.7 0.0
we use a multiplier approach to consider a breadth of parameter
values. For example, instead of varying the crude oil price in
2030 and 2031 independently, all oil prices are scaled by a multi-
plier that varies from 1 in 2010 to a user-specied value in 2050.
This approach is similarly applied to other energy source supply
curves, vehicle cost and efciency data, and consumer choice pen-
alties. Other input variables were parameterized directly. In total,
there are 19 model variables selected for parametric exploration,
many of which are listed in Table 1 along with baseline, minimum,
and maximum values. These values formed a triangular distribu-
tion in the sensitivity studies (with the baseline value equal to
the mode). The parameter ranges do not necessarily reect likely
future values, but rather capture potential bounding cases. Addi-
tionally, all of the input parameters are assumed to be indepen-
dent, which is plausible near the baseline values, but likely less
valid at some extreme values. For example, at high values of oil
price or carbon price, the required payback period of consumers
would likely be longer than the baseline value (consumers consider
more years of ownership costs) [38].
The input parameters selected can be categorized across con-
ceptual labels such as inherent modeling assumptions, economic
forecasting, technological development, and future policy deci-
sions. The carbon price, required payback period, and the penalty
multiplier could all be considered parameters subject to policy
inuence. The carbon price parameter appends an additional cost
to fuels proportional to the emission of GHG and represents a po-
tential carbon tax policy. For instance, the maximum price consid-
ered ($500 per metric ton of CO2 equivalent) corresponds to an
additional price on gasoline of nearly $5 per gallon. Economic
uncertainties are captured by the commodity price multipliers
(oil, natural gas, etc.). Technological performance and cost
uncertainty is incorporated by varying combustion powertrain
efciency, electric powertrain efciency, battery cost, vehicle
cost, and home compressor costs. The home compressor purchase
and installation cost starts at $6500 in 2010 [39] and then de-
creases some percentage each year as specied by the parameter-
ization. Here the combustion efciency and cost parameters
impact all powertrains that burn fuel. The uncertainty associated
with infrastructure development is captured by the refueling sta-
tion growth rate parameter, VRI. The remaining parameters are
notable model assumptions with the potential for inuencing out-
put metrics.
A graphical depiction of some key input parameters, their base-
line values, and parameterization is shown in Fig. 1. Fig. 1a illus-
trates the range of oil and natural gas prices considered in the
analysis (in 2010 dollars), with both prices doubling over the sim-
ulation time. The shaded region highlights the parametrization
range using the multiplier approach described above. As the simu-
lation time progresses, this range gets larger, emphasizing the
growing economic uncertainty of these prices far into the future.
Max CNG mileage fraction Home fueling rate
2.0% 0.03 0.02
9% 0.09 0.04
500 0.49 0.29
3 0.22 0.10
3 0.44 0.19
11 0.22 0.22
1 0.12 0.39
2 0.23 0.10
1.5 0.21 0.11
40% 0.06 0.44
3 0.25 0.17
1.75 0.11 0.31
 M.B. Peterson et al. / Applied Energy 125 (2014) 206217
Fig. 1. Baseline projections and parameterization range for commodity prices, vehicle costs and LDV stock efciency.
Fig. 1b depicts the change in compact car price over the course of
the simulation (which is representative of all car sizes). By 2015,
most natural gas (CNG-BI, CNG) and gasoline (PHEV10-SI, HEV-SI,
and SI) vehicle prices atten out and stay fairly constant for the
next 35 years, while HEVCNG vehicles see a steady decrease in
price up to 2030 before attening out. In contrast, EVs (both
plug-in hybrids, PHEVs, and battery electric vehicles, BEVs) see
sharp decreases in price in the rst 20 years and steady decreases
thereafter. This is indicative of the technology advances projected
for electric vehicles, in particular battery technology. Fig. 1b also
shows the parameterization for just SI vehicles (other combus-
tion-based powertrains scale similarly), given the cost uncertain-
ties to comply with increasingly stringent emission standards.
Fig. 1c shows the parameterization range around the baseline bat-
tery price, a key technological uncertainty and strong inuence on
the affordability of EVs. The average LDV stock fuel economy is
projected to improve signicantly from 2010 to 2050 (Fig. 1d).
The parameterization range is again highlighted, with the lower
bound roughly corresponding to the CAFE regulation. Note that
the efciency of individual powertrains are input values subject
to parameterization via the multiplier approach, but the LDV stock
average economy is subject to consumer choice and can be consid-
ered a model output.
209
3.2. Sensitivity analysis
We performed a global sensitivity analysis to verify expected
model behavior and to reveal the most signicant drivers of vari-
ability in the output metrics. Specically, we sampled the input tri-
angular distributions more than 2000 times in a Monte Carlo
simulation with Latin hypercube sampling. Spearman rank correla-
tion coefcients were then computed between the output metrics
and input parameters. The magnitude of the correlation coefcient
relates the degree to which a given input parameter variance is sta-
tistically associated with the variance of an output. The sign of the
correlation coefcient signies whether the output metric is posi-
tively or negatively correlated with the input parameter. A coef-
cient value of 1 or 1 represents a perfect positive or negative
correlation, respectively.
The results of the sensitivity analysis are presented in Table 1.
The magnitudes of the correlation coefcients show that CNG mile-
age fraction (the fraction of all LDV stock miles fueled by CNG) is
most sensitive to any carbon price, and the price of natural gas. A
number of variables, including the required payback period, com-
bustion efciency, the price of oil, the purchase price of combus-
tion vehicles, and the price of batteries, show secondary levels of
correlation. Note that CNG mileage fraction captures both the
210 M.B. Peterson et al. / Applied Energy 125 (2014) 206217
market penetration of NGVs, as well as the CNG usage intensity by
the bi-fuel vehicle drivers. Intuitively, low natural gas prices, low
carbon prices, and higher combustion efciencies make NGVs more
attractive relative to other powertrains, especially EVs. Based on
the negative sign of the required payback period correlation
coefcient, however, consumers start to move away from NGVs
when considering extended ownership costs, suggesting they are
adopting alternative powertrains. By the prominence of the battery
price correlation coefcient, NGVs may be in competition with EVs.
This is an interesting result that will be explored in greater depth
below.
Another output metric germane to NGVs, home compressor
adoption rate (or home fueling rate), is chiey correlated with the
cost of the compressor technology, a carbon price, infrastructure
growth, and consumers sensitivity to infrastructure limitations
(through the penalty multiplier). The signs of the coefcients sug-
gest that as public refueling infrastructure becomes more available,
or if NGV drivers nd limited infrastructure less of an inconve-
nience, then home compressor purchases will decrease. This too will
be explored in greater detail below.
3.3. Baseline model predictions
Before investigating the parameter ranges and tradeoffs, it is
helpful to understand the baseline model state about which varia-
tions are taken. Output metrics of interest for the baseline model
are shown in Figs. 2ad. Fig. 2a shows the LDV stock breakdown
Fig. 2. Baseline LDV stock fractions, fuel mileage fractions, and GHG emissions. Fuel blends include E10, B20, and E85.
by powertrain over time. In this projection, ICEs (consisting of gas-
oline SI, diesel CI, E85, gasoline HEV-SI, diesel HEV-CI, and HEV-E85
powertrains) are about 55% of the LDV stock in 2050. NGVs account
for 10% with the remaining 35% made up of EVs. A similar story is
seen in Figs. 2b and c. Sales fractions of ICEs are just 30% by 2050,
with NGVs comprising 20% of all vehicle sales (Fig. 2b). The shift
from PHEV sales to HEV sales around 2017 affect the expiration
of credits from the American Recovery and Reinvestment Act of
2009, and thus illustrates the impact of incentives upon sales. In
addition, more than half of the mileage driven is using a fuel other
than E10 by 2050 (Fig. 2c), primarily split between CNG, E85, and
electricity. These alternative fuels start to take hold after 2030,
where public infrastructure and most vehicle offerings start to be-
come ubiquitous. Due to the rise in LDV stock fuel efciency
(Fig. 1c), and the increasing market share of AFVs (many of which
have lower or no emissions), total GHG emissions decrease com-
pared to 2010 levels. As shown in Fig. 2d, the model projects a
baseline reduction of a about 50% by 2050.
3.4. NGV adoption
As seen in the sensitivity analysis, a number of factors inuence
NGV adoption including commodity prices, required payback per-
iod, infrastructure growth, and policy. Intuitively NGV adoption is
driven by low natural gas prices and correspondingly high oil
prices. This is seen in Fig. 3a, which shows the highest NGV pene-
tration with the lowest natural gas price and the highest oil price.
 M.B. Peterson et al. / Applied Energy 125 (2014) 206217 211

(a) (b)

(c) (d)

Fig. 3. Contours of stock fractions for natural gas vehicles. Baseline represents default values in the model for the x and y variables.

Fig. 3b shows the effects of natural gas price and required payback
period on NGV stock fraction. At low natural gas prices a longer re-
quired payback period only makes NGVs look more attractive. At
higher natural gas prices, however, longer required payback peri-
ods do not signicantly impact NGV stock fraction, indicating that
other powertrains are preferred. This again highlights the competi-
tion between NGVs and other powertrains, which is addressed in
Section 3.5. The angle of the contour lines in Fig. 3c, where the
tradeoff between the cost of combustion-based powertrains and
batteries for EVs is shown, suggests that the competition is be-
tween NGVs and EVs. Note that both axes in Fig. 3c impact the
price of PHEVs. Fig. 3d depicts the tradeoff between natural gas
price and station growth. It is interesting to note how the slope
of the contour lines change in the plot. At low natural gas prices,
station growth parameter values greater than 1 have less of an ef-
fect on NGV adoption, as there are enough CNG refueling stations
to eliminate the refueling infrastructure penalty to consumers. This
is consistent with the ndings of [37], who observed that countries Fig. 4. Demographic segmentation and NGV adoption rates. Dashed lines are NGV
bi-fuel vehicles, solid lines are all NGVs (including bi-fuel and NGV hybrids).
successful at NGV promotion had station growth of approximately
Simulation done at 2050 oil prices of $300/barrel and 2050 natural gas prices of
1 per thousand vehicles. At high natural gas prices, however, this $4.40/Kcf.
saturation point is not reached in the span of infrastructure growth
parameter values considered.
Fig. 4 breaks down the NGV adoption by demographic segmen- The more miles traveled, the greater the cost savings by switching
tation to observe which consumers are adopting NGVs. When nat- to natural gas and the more rapidly initial purchase cost differ-
ural gas is notably cheaper than petroleum, such as in Fig. 4, the ences are paid off. In fact, there is about a 12 percentage point dif-
drivers with highest VMT demonstrate the highest adoption rates. ference between the NGV adoption rates of those with the lowest
212 M.B. Peterson et al. / Applied Energy 125 (2014) 206217
Fig. 5. LDV stock fraction variations with natural gas and oil prices. Parameters were taken from Table 1 with the required payback period set to 9 years.
VMT to the highest in this case. Furthermore, Fig. 4 also shows that
with a more than 2:1 ratio, most of the NGVs in this pricing envi-
ronment are pure-CNG powertrains. In this setting there is little
rationale for using a bi-fuel vehicle when gasoline is expensive
and high NGV adoption rates stimulate infrastructure growth.
With extensive CNG refueling infrastructure growth, there is also
no clear advantage to owning a home compressor. In fact, the
NGV adoption rate for those that live in single family homes is low-
er than those that do not because those with single family homes
are also well suited to owning electric vehicles.
Before 2030, Fig. 4 shows that most of the NGVs are bi-fuel to
overcome any infrastructure limitations (the solid and dotted lines
overlap). After 2030, the solid curves break away from the dotted
curves, indicating that bi-fuel NGVs make up a smaller and smaller
fraction of total NGVs. This occurs as infrastructure and availability
issues are solved and bi-fuel exibility is no longer needed. Thus
bi-fuel vehicles can be a bridge powertrain until infrastructure is
built and consumers are acclimated to using NGVs.
3.5. NGV competition with EVs
Fig. 5 shows the competition between ICEs, EVs, and NGVs
across variations in natural gas and oil prices (when natural gas
price was varied, oil price was held constant, and vice versa). In
these plots, the required payback period was increased to nine
years to emphasize the competition. With shorter required pay-
back periods, the trends are identical, but the market share of ICEs
is higher so the uctuations in NGVs and EVs are less dramatic.
As natural gas prices rise, the NGV stock fraction decreases by
about 35 percentage points while the ICE stock fraction increases
by less than 10. The remaining 25 percentage points are made up
with an increase in the EV stock fraction. EVs gain much more than
ICEs with the reduction of NGVs showing that there is strong com-
petition between EVs and NGVs as the natural gas price uctuates.
This competition is consistent with the observations in Fig. 4. There
is a limited set of high VMT drivers for whom higher upfront pur-
chase costs can be offset by fuel costs savings over a nite time
period. Just as NGVs have higher initial purchase costs, but cheaper
fuel costs compared to conventional gasoline vehicles, the same is
true for EVs. At baseline natural gas prices, NGVs are the most cost
effective alternative to conventional vehicles. However, if natural
gas prices rise and the fuel cost savings of NGVs are diminished,
the higher VMT drivers will shift towards EVs. Alternatively, if oil
prices rise while natural gas prices remain xed, the NGV stock
fraction increases by about 20 percentage points while the ICE
stock fraction decreases by about 25 and the EV stock fraction
changes minimally. In this context, the fuel cost savings of NGVs
becomes even more pronounced for more and more drivers, thus
NGV market share rises even further. It should also be noted that
the model does not include natural gas PHEVs. If these were in-
cluded, the NGV-EV competition might be less of an issue as there
would be overlap between the two categories.
To ensure that this key nding is not an artifact of our chosen
nest structure (Fig. 9a), where NGVs and EVs compete at the same
nest level, the same analysis was performed for two other nest
hierarchies. Both a nest hierarchy with NGVs embedded within
the ICE branch (Fig. 9b) and an entirely at nest where all power-
trains compete directly (Fig. 9c) were evaluated. The results for
these verication studies demonstrate that the competition
dynamics are robust and independent of the chosen nest. In both
of these perturbation nests, the same behavior is observed when
natural gas prices rise while oil prices are xed and also when oil
prices rise for xed natural gas prices.
3.6. Environmental impact of NGVs
CNG is often touted for its lower CO2 emissions compared to
gasoline due to its lower carbon:hydrogen ratio [3]. It can reduce
tailpipe GHG emissions by 2030% when compared to diesel and
gasoline [15,4043], however many of these estimates do not ac-
count for system-wide CH4 (another potent greenhouse gas) leak-
age [44]. As can been seen by the near-vertical curves in Fig. 6a, the
factor that best promotes NGV adoption, natural gas price, has al-
most no inuence on absolute GHG reductions. This is because
NGVs still produce emissions, and are not necessarily lower emit-
ting than EVs, which they displace from the LDV stock. Instead,
GHG emissions are more affected by technology advances that im-
prove the efciency of all combustion-based powertrains, includ-
ing NGVs. To emphasize this point, we analyzed the GHG
emissions and petroleum consumption of the LDV stock both with
and without NGVs available to consumers to determine their over-
all benet. Fig. 6b shows this comparison at baseline conditions.
The top panel of Fig. 6b shows that slightly greater GHG reduction
is achieved without NGVs (green1 dashed curve) than with NGVs
(blue solid curve). This is due to the fact that NGVs displace lower
1
For interpretation of color in Figs. 6 and 7, the reader is referred to the web
version of this article.
 M.B. Peterson et al. / Applied Energy 125 (2014) 206217 213

(a) (b)

Fig. 6. Contours of GHG/mile reductions and overall benet of NGVs on total GHG emissions and petroleum consumption. Baseline represents default values in the model for
the x and y variables.

(a) (b)

(c)

Fig. 7. Factors that impact home compressor sales for candidate NGV owners and the impact of infrastructure on LDV stock mileage from CNG.

emission EVs from the LDV stock, as described in Section 3.5. If the would expect, NGVs are indeed successful in reducing overall LDV
GHG reductions were to include CH4 leakage [44], it is expected stock petroleum consumption, by as much as 510 percentage
this result would be even more dramatic. Nevertheless, as one points.
214 M.B. Peterson et al. / Applied Energy 125 (2014) 206217
(a)
Fig. 8. Impact of growth in NGV model offerings by automakers on NGV adoption.
3.7. CNG refueling infrastructure
NGV owners that live in a single family house with natural gas
have the option of refueling at public fueling stations or using the
natural gas piped into their home in combination with a compres-
sor. Fig. 7a shows how the home compressor purchase and instal-
lation cost changes with different annual cost reduction rates
(dened as reducing the home compressor cost x% per year). With
no reduction rate, the cost stays constant at about $6500 (blue)
[39]. A large annual reduction rate of 50% decreases the compres-
sor cost to nearly $0 in 10 years (purple). Both of these extremes
are equally unlikely, but serve as bounding cases in the parameter-
ization. Fig. 7b shows how the set of reduction rates translate to
annual compressor sales. Here, the compressor sales are those pur-
chased by NGV owners who live in single family homes with nat-
ural gas. Public refueling infrastructure is growing concurrently, and
once signicant numbers come on-line after 1015 years, the com-
pressor sales growth rate diminishes over time. If the home com-
pressor annual cost reduction rate is negligible, annual compressor
sales will only reach approximately 40,000 units per year by 2050.
Higher annual cost reduction rates of 10% and 50% lead to sales of
almost an order of magnitude higher. Thus, reducing the home com-
pressor purchase and installation costs to less than $1000 or even
$500 would entail noteworthy increases in purchase rates.
In addition to home compressor purchase rates, the growth of
public CNG refueling infrastructure can impact NGV adoption in
general and also CNG usage rates for bi-fuel vehicle drivers. This
dynamic is shown in Fig. 7c, which depicts variations of CNG mile-
age fraction in 2050 across refueling station growth rates and an-
nual cost reduction rates for home compressors. Greater gains in
CNG mileage fraction are achieved by increasing the station
growth rate than by reducing the annual cost of home compressors
from current price points to inconsequential price points
($500$1000). This implies that public infrastructure has a greater
potential impact on NGV sales than home refueling infrastructure,
which makes sense because it can service a wider demographic.
Just as limited refueling infrastructure can stymie NGV adop-
tion, so too can the availability of CNG-powered vehicles in auto-
makers eets restrict NGV growth. This point is captured in
Fig. 8. Each colored curve in Fig. 8b represents a different growth
rate of refueling infrastructure. By increasing the VRI from 0 to
1.5 (solid blue and solid teal curves, respectively), the NGV stock
fraction increases by about 10 percentage points. These solid lines
depict NGV stock fractions when NGV models are introduced
gradually over time, as depicted in Fig. 8a, our default assumption.
(b)
However, if NGV models were offered ubiquitously starting today
(meaning a CNG-powered option would be available for every cur-
rent car model), the NGV stock fraction could reach 15% as soon as
2030 (dotted curves). Fig. 8b shows that in the next 25 years, the
number of NGV models available can have a dramatic impact on
NGV adoption. In fact, this impact outweighs any limitations im-
posed by infrastructure growth rates. However, in the long term
there will likely be plenty of NGV models available, even if they
are introduced gradually, thus the solid and dashed lines begin to
converge.
4. Conclusions
The model presented in this paper supported a parametric trade
space analysis of the drivers and tradeoffs of NGV adoption and the
broader impact of NGVs in the LDV stock. Since model conclusions
are highly dependent on the assumption set used, we employed a
parametric approach to understand the sensitivity of the model to
these assumptions and explored a wide range of values rather then
a few scenarios. The major driver of NGV adoption is low natural
gas prices and correspondingly high oil prices. This fuel cost sav-
ings allows drivers to more quickly pay off initial purchase cost dif-
ferences (with even quicker payoffs seen by high VMT drivers).
However, other alternative powertrains, such as EVs, also offer fuel
cost savings despite higher purchase costs. Thus, NGVs compete
more with EVs than with conventional vehicles. Consequently, by
displacing some of the lowest emitting vehicles in the LDV stock,
NGVs do not necessarily reduce absolute GHG emissions and offer
little incremental GHG benet. Regardless whether NGVs are in-
cluded as part of the LDV stock or not, GHG emissions per mile
can be reduced by as much as 50% compared to 2010 levels.
Accounting for CH4 leakage, however, may further increase GHG
emissions.
Infrastructure also plays a role in NGV adoption, with greater
investment needed to promote NGV adoption at higher natural
gas prices. In addition, there is a tradeoff between public and home
infrastructure growth. As public CNG refueling expands and ser-
vices a greater segment of the population, compressors for home
fueling become less necessary and purchase rates decrease. In
the interim though, bi-fuel NGVs can serve as a bridge until public
infrastructure is built and consumers are acclimated to using
NGVs. Delays in NGV adoption are also attributable to the time it
takes for automakers to offer CNG-powered models in their eets.
More rapid deployment by automakers can spur both sales and
infrastructure investment.
 M.B. Peterson et al. / Applied Energy 125 (2014) 206217
Acknowledgments
Financial support was provided by the United States Depart-
ment of Energy, Vehicle Technologies Ofce. The authors would
also like to thank Akhil Reddy for his contributions to the analysis
and to the reviewers for their helpful comments.
Sandia National Laboratories is a multi-program laboratory
managed and operated by Sandia Corporation, a wholly owned
subsidiary of Lockheed Martin Corporation, for the U.S. Depart-
ment of Energys National Nuclear Security Administration under
Contract DE-AC04-94AL85000.
Appendix A. Additional model detail
This section provides additional model detail not supplied in
the main article body or found in previous publications. Thus, the
emphasis is on the extension of the existing model description
found in [33] to NGVs and their fueling infrastructure.
The LDV stock is segmented by seven attributes, denoted by
subscripts, a age, r region, s vehicle size, n powertrain, p population
density, d driving intensity, and h housing type. The elements of
each attribute are,
1. Age: 046 years.
2. Region: The 48 continental states and the District of Columbia.
3. Size: Compact, midsize, small SUV, large SUV, and light pickup
trucks.
4. Powertrain: 16 in total. The rst 12 powertrains are conven-
tional, hybrid (HEV), plug-in hybrid electric vehicle with a 10
mile all-electric range (PHEV10), and PHEV40 variants of
gasoline-fueled Spark Ignition (SI), dieselfueled Compression
Fig. 9. Nested consumer choice diagrams of the default nest, a perturbation with NGVs within the conventional vehicle branch, and a at nest.
215
Ignition (CI), and E85 ex-fuel. Next is a battery-electric vehicle
with 150 mile all-electric range (BEV150). The nal three
powertrains are NGVs, one which runs on CNG only, one is a
CNG-gasoline bi-fuel vehicle, and the nal one is a CNG hybrid
vehicle.
5. Population density: Urban, suburban, and rural.
6. Driving intensity: High, medium, and low based on binning
average daily trip distance.
7. Housing type: Single family house with natural gas, single family
house without natural gas, and other.
Consumers choose between 16 different powertrains in a
nested, multi-nomial logit choice model. After-market powertrain
conversions are not tracked. Used car sales are also not explicitly
tracked as the vehicles remain in the stock and are assumed to
operate similar to other vehicles of the same size and age. Sales
fractions are calculated using a nested powertrain choice hierarchy
[45], shown in Fig. 9a. The nested choice has different price sensi-
tivity at each level, where the bottom branches have the highest
price elasticity and the top branches have the lowest. Meaning,
in the default nest hierarchy, consumers rst choose whether to
purchase a conventional, electric, or natural gas vehicle. This
choice is evaluated with a logit choice exponent of 9. For those
that choose a conventional vehicle, the next decision is to buy a
hybrid or not, and the nal choice is gasoline SI, diesel CI, or E85
powertrain variants. The nal choice is done at a logit choice
exponent of 15. Due to the central place of the nest in the entire
model, it is parameterized to explore uncertainties in true values of
consumer price sensitivity and as a modeling approach in general.
The logit choice exponents in each nest are uniformly scaled up or
down. The nest structure itself is also changed to the options
216 M.B. Peterson et al. / Applied Energy 125 (2014) 206217

X
shown in Figs. 9b and c to explore the competition dynamics be- CFrsnd PFrhf Na0;rsnpdhf 1  Q t dM
nhf Na0;rsnpdhf ;
tween powertrains. The sales fractions are scaled by an OEM avail- f
ability fraction, j, which represents the penetration of an Narsnpdh;elec mnhf Narsnpdh;CNG ;
alternative powertrain across model years, 8
< 1:905 kWh=gge; 0:79=gge; NGVs in single family
>
k1 mnhf ; dMnhf with natural gas
jn  ; >
:
1 1000 exp k2 t  m0n 0; else
8
< 1;t  j0n
k1 ; k2 0:3=year where PF is the fuel price, N is the fuel consumption, dM is the ac-
: crued maintenance cost per unit energy of compressed CNG, and
0;t < m0n
m is the electricity required to operate the compressor for the same
where j0 is the mass market introduction year of a given power- unit. The values of these constants were taken from the specica-
train. The value of j0 for most powertrains is listed in [33], and tions of existing technology [47].
for CNG the value is 2013 (Honda Civic GX), CNG-BI is 2014 (Chevy
Silverado and GMC Sierra), HEVCNG is 2020 (estimated). A.1. Fuel choice for ex fuel vehicles
Vehicle purchase costs for all powertrains except the NGVs
are calculated using estimates for advanced technology from The E85 powertrains, including their conventional, hybrid, and
[34]. NGV costs were estimated by comparing existing cars with PHEV variants, are treated as ex-fuel powertrains that can run
both CNG and gasoline models. In 2010, dedicated CNG power- on E85 fuel as an alternative to the conventional E10 available.
trains were thus assessed to be $6000 more expensive than gas- The CNG bi-fuel vehicle can similarly operate on either E10 or
oline SI powertrains of a similar size and CNG bi-fuel pickups CNG, but has separate fuel tanks, unlike the FFVs. Thus, drivers of
were assessed to be $10,000 more expensive [5]. Thus, the price E85 and CNG bi-fuel powertrain vehicles face a choice at the pump
premium to consumers varies with vehicle size, but is also ex- of which fuel to use, assuming the alternative fuels are available.
pected to decrease over time as production ramps up. In the This choice is based on a willingness-to-pay distribution for the
end, the cost scaling with size and projected manufacturing price premium of the alternative fuel. Willingness-to-pay is ex-
improvements is done in the same manner as the PHEV10 pow- pressed by a logistic distribution with a parameter value, h, cali-
ertrains. The additional cost of HEVCNG powertrains over dedi- brated such that a price premium of 0:10/gge results in a
cated CNG powertrains was assumed to be the same as the 10:5% change in the alternative fuel demand [48,49]. For the
additional cost of HEV-SI powertrains over dedicated gasoline CNG bi-fuel powertrains this is expressed as,
SI powertrains.
Vehicle owners with the appropriate housing type can recharge 1
rFrh;CNGbi 1  ; x PFrh;CNGbi  PFrh;E10 ;
electric vehicles at home or rell CNG tanks at home. In these 1 ex=h
cases, the purchase cost of the home fueling system, CH , is added where rF is the willingness-to-pay logistic cumulative distribution
to each vehicle, function based on price premium. An identical choice model is used
CHnh dHnh 1  Qt ; for E85 ex fuel powertrains, substituting in the appropriate fuel
8 options. It should be noted that the E85 PHEVs are still assumed
> 0; Level 1 : PHEV10 in single family house to use nightly recharging, but have a fuel choice for the miles trav-
>
>
>
>
>
< 1500; Level 2 : PHEV40; BEV in single family house eled beyond their all-electric range. The nal fuel use rate for a gi-
dHnh ; ven vehicle is a product of this choice function and the probability
>
> 6500; NGVs in single family house with natural gas of visiting a station with an alternative fuel pump in a given region.
>
>
>
> The rate of change of the number of alternative fuel pumps, U, is a
:
0; else function of vehicle sales, dV=dt, in a given region,

where Q is the parameterized rate of price decline per year and t
is the years since 2010. The home CNG compressor cost includes
both purchase and installation [39]. Due to the high price point
of home CNG compressors, NGV owners with a single family house
and utility natural gas access are given an additional option to
purchase the compressor or to rely on public refueling (as the
other housing type segments do). This home compressor purchase
rate, rH , is,
0 !
U Vrsnpd;SFGAS 0 CGrsnpdh
r H
P ; 8n 2 NGV; U Vrsnpdh exp b H
; parametric approach, where the focus is placed on the competing
rsnpd V0
h2SF U rsnpdh
hCGrsnpdh ih
0
where U V is the generalized utility, hiih is an averaging over the
h-dimension, and bH 15, the same price sensitivity as the lowest
nest level.
Future vehicle efciency data for all powertrains and all fuel
sources are taken from [34], which also comply with the CAFE reg-
ulatory targets [46]. NGVs are assumed to achieve the same perfor-
mance as other SI powertrains. Due to the focus of this analysis on
NGVs, home compressor electricity use and maintenance require-
ments were considered additional, recurring fuel-like costs. For
NGV owners in a single family house with natural gas, the fuel cost,
CF , is expressed as,
dUrpf X dV a0;rsnpdh
VRI
dt sdh;n2f
dt
A.2. Model verication
In general, predictive models are difcult to validate until hind-
sight data is available many years in the future. This difculty in
validation was part of the motivation for this study to employ a
inuences of multiple factors, instead of drawing conclusions
based on a few single-scenario runs. Nevertheless, model develop-
ment has been conducted using defensible assumptions and refer-
enced input values. The baseline values for models input have
been drawn from the literature or best-available government pub-
lished data wherever possible, and heavily parameterized when
this is not feasible. The nested-choice function price sensitivity
was calibrated with available sales data from 20102013. Where
uncertainty in the input data (e.g. commodity supply curves) or
modeling assumptions (e.g. S-curve powertrain availability) was
high, we relied on sensitivity analysis and parameter exploration
to assess the implications on our results. Thus, we have relied on
model verication, more than validation, where a perturbation in
 M.B. Peterson et al. / Applied Energy 125 (2014) 206217
one or more parameters leads to expected changes in output
metrics.
References
[1] US Energy Information Administration. Annual energy outlook 2012: with
projections to 2035. Tech rep DOE/EIA-0383; 2012.
[2] US Energy Information Administration. Natural gas annual 2012. Washington,
DC; 2012.
[3] Moniz EJ, Jacoby HD, Meggs AJ, editors. The future of natural gas. Cambridge
(MA): Massachusetts Institute of Technology; 2011.
[4] US Environmental Protection Agency. Inventory of US greenhouse gas
emissions and sinks: 19902011. Washington, DC; 2013.
[5] National Petroleum Council. Advancing technology for Americas
transportation future. Washington, DC; 2012.
[6] Ogden JM. Developing an infrastructure for hydrogen vehicles: a Southern
California case study. Int J Hydrogen Energy 1999;24(8):70930. http://
dx.doi.org/10.1016/S0360-3199(98)00131-1.
[7] Whitney G, Behrens CE. Energy: natural gas: the production and use of natural
gas, natural gas imports and exports, EPAct project, Liqueed Natural Gas
(LNG) import terminals and infrastructure security. Alexandria
(VA): TheCapitol.Net, Inc.; 2010.
[8] Greene DL. Uncertainty, loss aversion, and market for energy efciency. Energy
Econ 2011;23(4):60816.
[9] Train K. Discount rates in consumers energy-related decisions: a review of the
literature. Energy Int J 1985;10(12):124353.
[10] Calfee JE. Estimating the demand for electric automobiles using fully
disaggregated probabilistic choice analysis. Trans Res Part
1985;19(4):287301.
[11] Eppstein MJ, Grover DK, Marshall JS, Rizzo DM. An agent-based model to study
market penetration of plug-in hybrid electric vehicles. Energy Policy
2011;39(6):3789802.
[12] Robbins AH, Braun CH. Automotive natural-gas a case-study. In: Proceedings
of the 26th intersociety energy conversion engineering conference, vols. 16;
1991. p. D2505.
[13] Okken PA. A case for alternative transport fuels. Energy Policy
1991;19(4):4005.
[14] Garrett D. Choosing an alternative fuel for the motor-vehicle. In: Proceedings
of the 25th intersociety energy conversion engineering conference, vols. 16;
1990. p. 2219.
[15] Klausmeier RF, Billick IF. Comparative-analysis of the environmental-impact of
alternative transportation fuels. Energy Fuels 1993;7(1):2732.
[16] Bunch DS, Bradley M, Golob TF, Kitamura R, Occhiuzzo GP. Demand for clean-
fuel vehicles in california a discrete-choice stated preference pilot project.
Trans Res Part a Policy Practice 1993;27(3):23753.
[17] Cheema J. Electric vehicles and other clean fuel alternatives: a comparative
analysis. In: Symposium proceedings EVS-11 11th international electric
vehicle symposium electric vehicles: the environment-friendly mobility, vol.
2; 1992. p. 111.
[18] Liss WE, Thrasher WH. Natural gas as a stationary engine and vehicular fuel.
Soc Automot Eng 1989:478.
[19] Weaver CS. Natural gas vehicles a review of the state of the art. Soc Automot
Eng 1989:3555.
[20] Decarvalho AV. Natural-gas and other alternative fuels for transportation
purposes. Energy 1985;10(2):187215.
[21] Golob TF, Kitamura R, Bradley M, Bunch DS. Predicting the market penetration
of electric and clean-fuel vehicles. Sci Total Environ 1993;134(13):37181.
[22] Gandhidasan P, Ertas A, Anderson EE. Review of methanol and compressed
natural-gas (cng) as alternative for transportation fuels. J Energy Resour
Technol-Trans Asme 1991;113(2):1017.
[23] US Energy Information Administration. Alternatives to traditional
transportation fuels: an overview; 1994.
[24] US Department of Energy. Assessment of costs and benets of exible and
alternative fuel use in the us transportation sector. Tech rep DOE/PE-0095P.
Ofce of Policy, Planning, and Analysis; 1990.
[25] US Department of Energy. Assessment of costs and benets of exible and
alternative fuel use in the us transportation sector: progress report two: the
international experience. Tech rep DOE/PE-0085. Ofce of Policy, Planning and
Analysis; 1988.
[26] Barter GE, Reichmuth D, Westbrook J, Malczynski LA, West TH, Manley DK,
et al. Parametric analysis of technology and policy tradeoffs for conventional
and electric light-duty vehicles. Energy Policy 2012;46:47388.
[27] US Census Bureau. 2010 Census of population and housing; 2010.
[28] Greene DL, Lin Z. Who will more likely buy PHEV; a detailed market
segmentation analysis. The 25th world battery, hybrid and fuel cell electric
vehicle symposium & exhibition; 2010.
217
[29] US Department of Transportation Federal Highway Administration. Highway
statistics 2009, state motor-vehicle registrations. Tech rep MV-1, MV-9; 2010.
[30] US Department of Transportation National Highway Trafc Safety
Administration. Vehicle survivability and travel mileage schedules. Tech rep
DOT HS 809 952. Washington, DC; 2006.
[31] US Department of Transportation Federal Highway Administration. 2009
National household travel survey; 2011. <http://nhts.ornl.gov>.
[32] Davis SC, Diegel SW, Boundy RG. Transportation energy data book: edition 31.
ORNL-6987. Oak Ridge, Tennessee: Oak Ridge National Laboratory; 2012.
[33] Barter GE, Reichmuth D, West TH, Manley DK. The future adoption and benet
of electric vehicles: a parametric assessment. SAE Int J Alt Power 2013;6(1).
http://dx.doi.org/10.4271/2013-01-0502.
[34] Moawad A, Sharer P, Rousseau A. Light-duty vehicle fuel consumption
displacement potential up to 2045. Argonne National Laboratory. Argonne,
IL; 2011.
[35] Wang M. GREET 1 2012 rev1. Argonne National Laboratory; Argonne, IL; 2012.
<http://greet.es.anl.gov>.
[36] Arteconi A, Brandoni C, Evangelista D, Polonara F. Life-cycle greenhouse gas
analysis of LNG as a heavy vehicle fuel in Europe. Appl Energy
2010;87(6):200513.
[37] Yeh S. An empirical analysis on the adoption of alternative fuel vehicles: the
case of natural gas vehicles. Energy Policy 2007;35(11):586575.
[38] Puller SL, Greening LA. Household adjustment to gasoline price change: an
analysis using 9 years of US survey data. Energy Econ 1999;21(1):3752.
<http://dx.doi.org/10.1016/S0140-9883(98)00006-1>.
[39] CNG California. About CNG; 2012. <http://www.cngcalifornia.com/
compressed-natural-gas.asp>.
[40] Emission Factors for Greenhouse Gas Inventories. Environmental Protection
B Agency; 2011. <http://www.epa.gov/climateleadership>.
[41] US Energy Information Administration. Carbon dioxide emissions coefcients;
2013. <http://www.eia.gov/environment>.
[42] Clean Air Task Force. Clean diesel versus cng buses: cost, air quality, & climate
impacts. Tech rep; 2012.
[43] Bielaczyc P, Woodburn J, Szczotka A. An assessment of regulated emissions
and CO2 emissions from a European light-duty CNG-fueled vehicle in the
context of Euro 6 emissions regulations. Appl Energy 2014;117(0):13441.
[44] Brandt AR, Heath GA, Kort EA, OSullivan F, Ptron G, Jordaan SM, et al. Methane
leaks from north american natural gas systems. Science
2014;343(6172):7335.
[45] Lin Z, Greene DL. A plug-in hybrid consumer choice model with detailed
market segmentation. Transportation research board 10-1698; 2010.
[46] US Environmental Protection Agency and US Department of Transportation.
20172025 model year light-duty vehicle ghg emissions and cafe standards:
supplemental notice of intent. Tech rep 40 CFR Parts 85, 86, and 600; 2011.
[47] FuelMaker. Vehicle refueling appliance operating instructions. Toronto,
Canada; 2007.
[48] Anderson ST. The demand for ethanol as a gasoline substitute. J Environ Econ
Manage 2012;63(2):15168. http://dx.doi.org/10.1016/j.jeem.2011.08.002.
[49] Salvo A, Huse C. Build it, but will they come? Evidence from consumer choice
between gasoline and sugarcane ethanol; 2013. Working paper.
Glossary
AFV: alternative fuel vehicle
BEV: battery electric vehicle
BXX: XX% biodiesel by volume
CAFE: Corporate Average Fuel Economy
CI: dieselfueled compression ignition
CNG: compressed natural gas
CNGBI: bi-fuel CNG/gasoline vehicle
EV: electric vehicle
EXX: XX% ethanol by volume
FFV: ex fuel vehicle
GGE: Gallon gasoline equivalent
GHG: greenhouse gas
HEV: hybrid electric vehicle
ICE: internal combustion engine
LDV: light duty vehicle
NG: natural gas
NGV: natural gas vehicle
PHEV: plug-in hybrid electric vehicle
RFS2: Renewable Fuel Standard (2007)
SI: gasoline-fueled spark ignition
VMT: vehicle miles traveled (per vehicle)