Year 2008 UCD- ITS- RR- 08- 28
Lifecycle Analysis of Air Quality Impacts
of Hydrogen and Gasoline Transportation
Fuel Pathways
September 2008
Guihua Wang
Institute of Transportation Studies ◦ University of California, Davis
One Shields Avenue ◦ Davis, California 95616
PHONE: ( 530) 752- 6548 ◦ FAX: ( 530) 752- 6572
WEB: http:// www. its. ucdavis. edu
Copyright © Guihua Wang
2008
All Rights Reserved.
Acknowledgements
I would like to acknowledge my dissertation committee, Prof. Joan Ogden, Prof. Dan
Sperling, Prof. Dan Chang, and Prof. Pat Mokhtarian, for their insightful guidance and
consistent support. I especially want to thank Dr. Ogden, my primary advisor, for
offering me extensive professional training and insightful supervision throughout my
doctoral study. I highly appreciate her understanding, patience, and encouragement,
which have helped me to better handle the challenges I have faced. Special thanks are due
to Dr. Sperling, my committee co- chair, who led me into the transportation field and has
provided inspiring ideas for my study. Dr. Chang offered many insightful suggestions for
my research, and I really appreciate his timely feedback every time. Dr. Mokhtarian has
been very supportive throughout my graduate study, and she offered many valuable
discussions, comments, and other academic guidance.
For academic help during my graduate study at UC Davis, particular thanks also go to Dr.
Mark Delucchi, Dr. Chris Yang, Dr. Marc Melaina, Prof. Yueyue Fan, Prof. Richard
Sexton, Prof. Julian Alston, and Prof. Arthur Havenner.
This dissertation benefited from discussions with David McCollum, Song Bai, and
Michael Nicholas; their help is highly appreciated. I would like to thank all the members
in Prof. Ogden’s research group ( Ryan McCarthy, Nils Johnson, Stephenie Ritchey,
Nathan Parker, Jonathan Weinert, Brett Williams, Rusty Heffner, Brent Riffel, Are
Gjellan, Anthony Eggert, Joshua Cunningham, et al.). They are great colleagues.
- ii-
I also want to thank my friends: Xinyu Cao, Peng Wu, Zhenhong Lin, Changzheng Liu,
Huaizhu Gao, Ling Li, Jie Zheng, Julia Wang, Xiaoying Zhou, Jonathan Hughes, Reed
Benet, Justin Regnier, and Nic Lutsey. They have made my life in Davis much easier and
more enjoyable.
I am grateful to my family for their love and support.
I would like to thank the Hydrogen Pathways ( H2P) program and the Sustainable
Transportation Energy Pathways ( STEPS) program at the Institute of Transportation
Studies ( ITS) at the University of California, Davis for their support. This research is
partially funded by the Jastro- Shields graduate research award at UC Davis.
- iii-
To the memory of my father, to my mother, and to my wife Gengxin.
- iv-
Guihua Wang
September 2008
Civil and Environmental Engineering
Lifecycle Analysis of Air Quality Impacts of Hydrogen and Gasoline
Transportation Fuel Pathways
Abstract
Hydrogen has been proposed as a low- polluting alternative transportation fuel. This
dissertation analyzes the lifecycle air quality impacts of hydrogen and gasoline use in
light duty vehicles, including impacts from fuel production, delivery, and vehicle use.
The analysis is conducted for various scenarios in Sacramento, California, for four
pollutants: CO, NOx, VOC, and PM10. Three natural gas- based hydrogen supply
pathways are considered: onsite hydrogen production via small- scale steam methane
reforming ( SMR), central SMR production with gaseous hydrogen pipeline delivery, and
central SMR production with liquid hydrogen truck delivery. Four gasoline pathway
scenarios, as compared to hydrogen pathways, are also investigated in the study. A new
method is developed using travel demand model data to estimate air quality impacts of
gasoline fleet operations, regression analysis is used to explore the relationship between
lifecycle precursor emissions and secondary ozone formation for each hydrogen supply
pathway, and a Gaussian atmospheric dispersion model is used to analyze ambient
impacts.
- v-
The centralized/ pipeline hydrogen pathway and the onsite hydrogen production pathway
reduce pollution the most. The centralized hydrogen production with liquid truck delivery
is the least clean option among the three means of hydrogen supply. The examined
gasoline pathway, even with advanced new gasoline vehicles, would lead to much higher
ambient concentrations of pollutants than the hydrogen pathways, producing 273 times
greater CO, 88 times greater VOC, 8 times greater PM10, and 3.5 times greater NOx
concentrations than those caused by the centralized/ pipeline hydrogen pathway, assuming
the same size vehicle population. The study also estimates the potential impacts of the
above hydrogen pathways on secondary ozone air quality. The results indicate that
adding a significant number of hydrogen fuel cell vehicles ( FCVs) to the region would
have a very small impact on secondary ozone pollution; in fact, it does not necessarily
increase the peak ozone concentration, and may even cause it to decrease in some cases.
The results show that advanced gasoline vehicle technologies significantly reduce air
quality impacts of light duty vehicles, but hydrogen vehicle technologies provide still
greater benefits, reducing the contribution of light duty vehicles to ambient air pollutant
concentrations to near- zero.
- vi-
Table of Contents
Title Page ............................................................................................................................ i
Acknowledgements ........................................................................................................... ii
Abstract....................................................................................................................... ...... v
Table of Contents ............................................................................................................ vii
1. Introduction................................................................................................................... 1
1.1. Research background ............................................................................................... 1
1.2. Research objectives.................................................................................................. 3
1.3. Research approach: lifecycle analysis ..................................................................... 5
1.4. Research contributions............................................................................................. 9
1.5. Dissertation organization ....................................................................................... 11
2. Lifecycle Impacts of Hydrogen Supply Pathways on Urban Air Quality of
Primary Pollutants.......................................................................................................... 13
2.1. Introduction............................................................................................................ 13
2.2. Methodology.......................................................................................................... 15
2.2.1. Overview of hydrogen pathway scenarios...................................................... 15
2.2.2. Estimating hydrogen demand ......................................................................... 17
2.2.3. Hydrogen supply schemes .............................................................................. 20
2.2.4. Spatial layout of hydrogen pathway steps ...................................................... 22
2.2.5. Lifecycle emission inventories ....................................................................... 28
2.2.5.1. Hydrogen production infrastructure......................................................... 29
- vii-
2.2.5.2. Electric power plants................................................................................ 30
2.2.5.3. Diesel- fueled delivery trucks ................................................................... 32
2.2.6. Atmospheric transport and urban air quality .................................................. 32
2.2.6.1. Atmospheric transport.............................................................................. 33
2.2.6.2. The ISC model ......................................................................................... 35
2.2.6.3. The TMY2 dataset ................................................................................... 35
2.2.6.4. The Air Quality System ( AQS) for air pollution monitors...................... 36
2.2.6.5. The NAAQS standards and actual measurements ................................... 37
2.3. Results and discussion ........................................................................................... 38
2.3.1. Incremental pollution attributable to hydrogen pathways .............................. 38
2.3.2. Comparison to the current ambient measurements......................................... 46
2.3.3. Further comparison among hydrogen pathways............................................. 47
2.3.4. Source contributions to incremental ambient pollution.................................. 48
2.4. Conclusions............................................................................................................ 52
3. Estimating Changes in Urban Ozone Concentrations Due to Lifecycle Emissions
from Hydrogen Transportation Systems...................................................................... 55
3.1. Introduction............................................................................................................ 55
3.2. Literature review on predictors of ozone formation .............................................. 56
3.3. Methodology.......................................................................................................... 59
3.3.1. Overview of methodology .............................................................................. 59
3.3.2. Hydrogen pathway scenarios and dispersion model applications................... 60
3.3.3. Data and the ozone regression model ............................................................. 62
3.3.4. Applying the regression model ....................................................................... 70
- viii-
3.4. Results and discussion ........................................................................................... 72
3.4.1. Incremental 3- hour average pollution of ozone precursors ............................ 72
3.4.2. Changes in peak ozone concentrations, ΔO3 ( max) ....................................... 77
3.4.3. Percentage changes in peak ozone concentrations, % ( max) 3 ΔO ................... 78
3.4.4. Further discussion on ozone pollution ............................................................ 81
3.5. Conclusions............................................................................................................ 82
4. Investigating Contributions of Gasoline Pathways to Urban Air Pollution Using
Travel Demand Model Data........................................................................................... 86
4.1. Introduction............................................................................................................ 86
4.2. Overview of gasoline pathway scenarios............................................................... 88
4.3. Gasoline fleets considered ..................................................................................... 91
4.4. Methodology.......................................................................................................... 94
4.4.1. Overview of methodology .............................................................................. 94
4.4.2. The EMFAC model......................................................................................... 96
4.4.3. The CONVIRS and IRS models ..................................................................... 99
4.4.4. The SACMET model .................................................................................... 100
4.4.5. The DTIM model .......................................................................................... 102
4.4.5.1. Temporal distribution of emissions ....................................................... 103
4.4.5.2. Spatial distribution of emissions............................................................ 103
4.4.5.3. Determining the size of grid cells .......................................................... 103
4.4.6. The ISC model ( revisited)............................................................................. 104
4.4.7. The TMY2 dataset ( revisited)....................................................................... 106
4.5. Results and discussion ......................................................................................... 107
- ix-
4.5.1. Air pollutant concentrations caused by gasoline fleet operations................. 107
4.5.2. The treatment of gasoline- delivery trucks .................................................... 113
4.5.3. Further discussion on aggregate vehicle emission trends............................. 116
4.6. Conclusions.......................................................................................................... 119
5. Comparing Air Quality Impacts of Hydrogen and Gasoline Supply Pathways . 122
5.1. Introduction.......................................................................................................... 122
5.2. Summary of hydrogen supply pathway scenarios ............................................... 123
5.3. Summary of gasoline supply pathway scenarios ................................................. 128
5.4. Results and discussion ......................................................................................... 131
5.4.1. Incremental concentrations caused by hydrogen pathways .......................... 131
5.4.2. Air pollutant concentrations caused by gasoline pathways .......................... 132
5.4.2.1. Air pollutant concentrations caused by gasoline fleet operations.......... 132
5.4.2.2. Air pollutant concentrations caused by gasoline- delivery trucks .......... 133
5.4.2.3. Summary ................................................................................................ 134
5.4.3. Comparison between hydrogen and gasoline pathways ............................... 135
5.5. Conclusions.......................................................................................................... 139
6. Summary and Conclusions....................................................................................... 143
Bibliography .................................................................................................................. 150
- x-
1
1. Introduction
1.1. Research background
Hydrogen is a compelling alternative transportation fuel. Despite many technical and
economic barriers, a hydrogen economy is quite attractive ( NRC, 2004). Use of hydrogen
in vehicles has many potential benefits ( Sperling and Ogden, 2004). Hydrogen can be
derived from a variety of sources such as natural gas ( NG), coal, biomass, solar, wind,
hydropower, and nuclear power, and could reduce oil supply insecurity ( Ogden, 2002;
Ogden, 1999a; Ogden, 1999b; Ogden et al., 2004). Hydrogen fuel cell vehicles ( FCVs)
produce no tailpipe emissions, and if made from renewables, decarbonized fossil fuels, or
nuclear energy, hydrogen can also be produced and used with no emissions of greenhouse
gases ( GHGs), and could help mitigate global warming ( Ogden, 2002; Ogden, 1999b;
Ogden et al., 2004; Sperling and Ogden, 2004). In addition, fuel cell vehicles running on
hydrogen could offer fuel economy around 2.5 times that of today’s conventional internal
combustion engine ( ICE) vehicles ( Ahluwalia et al., 2004; Farrell and Sperling, 2007;
NRC, 2004).
In the U. S., the current petroleum- based transportation system emits significant amounts
of carbon monoxide ( CO), nitrogen oxides ( NOx), total organic gases ( TOGs) or volatile
organic compounds ( VOCs), and particulate matter ( PM10, referring to particulates with
an aerodynamic diameter less than 10 μm), as well as carbon dioxide ( CO2). To assure
that transportation investments do not cause undermine efforts to attain ambient air
2
quality standards, the U. S. Clean Air Act requires that transportation plans for highway
and transit projects be consistent with the air quality goals set by a state implementation
plan ( SIP) ( U. S. DOT, 2007; U. S. EPA, 2007).
Although air quality in the U. S., in general, has been improving over the past several
decades, it is still a challenging problem in many regions ( designated to be in non-attainment).
For example, mobile- source VOC and NOx emissions are precursors to
secondary ozone formation and aerosols. Particulates and ozone are the two criteria
pollutants of greatest concern causing human health damage and leading to a social cost
issue ( ExternE, 1998; McCubbin and Delucchi, 1996; Murphy et al., 1999; Delucchi et al.,
2002).
Hydrogen has been proposed as a low- polluting alternative transportation fuel that could
help improve urban air quality. In particular, it has been suggested that hydrogen FCVs
be introduced into the vehicle marketplace where zero emission vehicle ( ZEV) mandates
are enacted, e. g., in California ( CARB, 2008a). To achieve high energy efficiencies and
low overall emissions, the different pathways for producing hydrogen as a transportation
fuel must be carefully examined ( Wang, 2002). This dissertation analyzes the potential
air quality impacts of hydrogen transportation fuel, using a lifecycle analysis ( LCA)
approach.
3
1.2. Research objectives
The overall objective of this study is to address the following questions.
( 1) What would be the impact of hydrogen fuel cell vehicles on criteria pollutant
emissions and air quality, considering all the emissions involved in the full fuel
cycle, including producing, transporting, and using hydrogen?
( 2) What changes in peak ozone pollution would occur if typical hydrogen supply
pathways were introduced in Sacramento, California?
( 3) What hydrogen supply strategy would be environmentally best for a specific
county, region, or air basin in the U. S.? Is it onsite production or centralized
production? Which delivery mode for centralized hydrogen is better, liquid truck
or gaseous pipeline?
( 4) What would the optimal, feasible spatial layout of hydrogen pathway steps be in a
specific region, assuming that a possible hydrogen pathway type has been
determined? and
( 5) How do hydrogen FCVs compare to conventional or advanced gasoline vehicles
in terms of the resulting impacts on emissions and air quality, from a lifecycle
analysis perspective?
4
In this dissertation, a regional lifecycle analysis of air quality impacts is carried out to
explore the hydrogen economy in more depth. We choose Sacramento, California as a
site for our case study. The specific task- oriented objectives of this study are:
( 1) To design typical, promising hydrogen pathways for the specific region,
considering the regional feedstock resource availability;
( 2) To compile emission inventories for several near- term alternative hydrogen
pathways and both current and advanced gasoline fleet operations;
( 3) To investigate the impacts of natural gas- to- hydrogen pathways on urban air
quality of primary pollutants, using a Gaussian atmospheric dispersion model;
( 4) To explore the relationship between secondary ozone formation and lifecycle
precursor emissions from each hydrogen supply pathway, using regression
analysis; and
( 5) To develop a new methodological framework for estimating the air quality
impacts of gasoline fleet operations, using travel demand model data and grid-level
emission inventories.
5
1.3. Research approach: lifecycle analysis
This study uses the approach of lifecycle analysis ( LCA), also written as lifecycle
assessment. The LCA refers to the cradle- to- grave cycle of a product ( Wang, 1999).
More generally, LCA is a technique used to assess all the inputs and outputs of a product,
process, or service ( U. S. EPA, 2008). To assess the environmental aspects or potential
impacts associated with a product, process, or service, the common procedure of LCA
application usually has the following sequence ( U. S. EPA, 2008).
( 1) Lifecycle inventory. This refers to compiling a complete inventory of related
energy and material inputs and outputs;
( 2) Impact assessment. This refers to assessing the potential environmental impacts
associated with inputs and outputs identified; and
( 3) Lifecycle interpretation. This refers to interpreting the results to help people
make a decision based on more comprehensive and complete information.
Lifecycle assessment of hydrogen production via natural gas steam reforming has been
conducted in a number of studies, e. g., Spath and Mann ( 2000). The concept of a full fuel
cycle is illustrated in Figure 1, using the case of hydrogen made from natural gas. A fuel
cycle for a given transportation fuel includes, but is not limited to, the following three
stages ( Wang, 1999):
6
( 1) The feedstock stage. Including feedstock extraction/ production, transportation,
and storage;
( 2) The fuel stage. Including fuel production, transportation, storage, and distribution;
and
( 3) Vehicle operation. Also called downstream activities, including fuel combustion,
evaporation, brake wear, and tire wear.
The full fuel cycle is also called well- to- wheels ( WTW). In contrast, well- to- tank ( WTT)
includes all activities during both the feedstock and fuel stages; i. e., the vehicle operation
stage is not included in WTT.
emissions
Coal extraction Rail transport
emissions
emissions
Power plant Elec transmission
Pipeline distribution
Compression Dispensing FCV operation
emissions
H2 production
Refueling station
emissions
NG extraction
& cleanup
Pipeline transport
Figure 1. Full fuel cycle of hydrogen made from natural gas ( i. e., hydrogen pathway). ( The upper figure
represents the sub- pathway of electricity consumed in steps of the primary fuel pathways.)
7
Several models have been developed for lifecycle analysis of alternative fueled vehicles
( AFVs). To estimate lifecycle energy use and emissions for a transportation fuel, both the
GREET and LEM models work very well.
The Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation ( GREET)
model, developed by the Argonne National Laboratory, can calculate both the WTT and
WTW results ( Wang, 1999). About 100 fuel production pathways and 70 vehicle/ fuel
systems, covering all major vehicle technologies, are included and assessed in GREET
( Wang, 1999).
The Lifecycle Emissions Model ( LEM), developed by the University of California, Davis,
can provide results for up to 30 countries and for the years 1970 to 2050 ( Delucchi, 2003).
The LCA boundary in LEM is defined more broadly, and it takes into account the
following four individual lifecycles ( Delucchi, 2003):
( 1) Lifecycle of fuels and electricity;
( 2) Lifecycle of vehicles. Including materials use, vehicle assembly, vehicle
operation and maintenance, and secondary fuel cycle for transport modes;
( 3) Lifecycle of materials. Including crude- ore recovery, finished- material
manufacture, and the transport of finished materials to end users; and
8
( 4) Lifecycle of infrastructure. Including energy use and materials production
associated with the construction of highways, railways, etc.
Figure 1 shows a representative hydrogen pathway, which is a full fuel cycle for
hydrogen made from natural gas. For a specific hydrogen pathway, the energy
consumption and emissions vary with parameters such as the natural gas pipeline length,
pipeline flow rate, plant capacity factor or plant size, plant conversion efficiency, truck
distribution distance, etc. The emissions of criteria air pollutants ( APs) and GHGs are
closely related to pathway- specific situations and rely heavily on geographic or regional
information.
Assuming that a certain number of hydrogen FCVs are operating in a specific region,
Sacramento, California, this research investigates several possible fuel supply pathways
to satisfy the vehicular demand for hydrogen fuel. Considering regional feedstock
resource availability and commercially available hydrogen production technology, we
focus on steam methane reforming ( SMR) of natural gas in this dissertation.
Gasoline pathways, referred to as the petroleum- based fuel pathways ( including both
gasoline and diesel), are also examined for the purposes of comparison to hydrogen
pathways. We investigate contributions of various current and advanced gasoline/ diesel
pathways to urban air pollution using travel demand model data.
9
Although some researchers have analyzed lifecycle emissions and energy use associated
with hydrogen supply chains ( Wang, 1999; Spath and Mann, 2000; Delucchi, 2003;
Colella et al., 2005; Jacobson et al., 2005), no study has examined the change in air
quality ( i. e., pollutant concentrations rather than emissions only) due to a hydrogen
transportation system. This study could provide the basis to quantitatively compare
hydrogen and gasoline pathways in terms of the resulting air pollutant concentrations,
and the results may be useful to policy makers in thinking about hydrogen.
1.4. Research contributions
The study makes an initial attempt to explore the lifecycle air quality impact of hydrogen
pathways on a regional scale. It investigates the complicated but important relationships
among transportation, energy, and the environment. The results of this research
contribute to development of alternative fuel strategies, with important implications for
policy makers. Below are specific original contributions.
( 1) This study provides a quantitative basis for a better understanding of the air
quality impacts of a hydrogen economy. It offers a more detailed and descriptive
understanding of hydrogen systems than previous studies;
( 2) This study develops new methods to quantitatively explore the air quality impacts
of hydrogen- based transportation systems, for both primary criteria pollutants and
secondary ozone formation. The results of comparison among the examined three
hydrogen supply pathways are useful for determining hydrogen supply strategies;
10
( 3) The methodological framework developed here for estimating the impacts of
hydrogen pathways could be adapted to analyzing other alternative transportation
fuel pathways, such as biofuels or coal- based hydrogen;
( 4) This study depicts the region- specific empirical kinetic modeling approach
( EKMA) property of secondary ozone formation, and points out that ozone
formation is limited by NOx in the summer and by VOC in the fall. This finding is
useful for environmental policy makers to control ozone pollution more
effectively;
( 5) This study estimates the air quality impacts of light duty gasoline vehicles. The
cases chosen span a range of possibilities for gasoline vehicles: from current
technology to advanced technology. This allows us to estimate the air quality
impacts resulting from cleaner, newer gasoline vehicles;
( 6) This study develops a new systematic methodology for estimating contributions
of mobile sources to urban air pollution, and it effectively connects transportation
emissions and air quality. That would be useful for transportation conformity
planning and mobile pollution control;
( 7) The results, when comparing hydrogen/ FCV scenarios with gasoline/ ICE
scenarios, will have meaningful implications for policy makers to evaluate
11
hydrogen, a promising alternative fuel, as compared to rapidly improving, low
emission gasoline vehicles; and
( 8) The results from this study are directly applicable for economists or scientists to
estimate the external cost of air pollution causing human health damage ( due to
either hydrogen pathways or gasoline pathways), following such a methodological
sequence: Source Emissions 􀃆 Ambient Pollution 􀃆 Physical Impacts 􀃆
Economic Evaluation.
1.5. Dissertation organization
The organization of this dissertation is as follows. Chapter 1 provides an introduction to
the dissertation research and its framework. Chapter 2 estimates the lifecycle impacts of
hydrogen supply pathways on urban air quality. Research emphasis is given to primary
pollutants, including CO, NOx, VOC, and PM10, as well as SOx. Going a step further,
Chapter 3 explores the relationship between secondary ozone formation and lifecycle
precursor emissions from hydrogen transportation systems by using regression analysis.
Chapters 2 and 3 focus on hydrogen pathways. With a focus on gasoline fleet operations,
Chapter 4 investigates the contributions of gasoline pathways to urban air pollution using
traditional four- step travel demand model data. All the four important mobile- source
pollutants, namely, CO, NOx, VOC, and PM10, are considered. In Chapter 5, both
hydrogen and gasoline supply pathways are examined together, and their air quality
impacts are compared from a lifecycle emissions perspective. Finally, Chapter 6
12
summarizes all the individual projects and presents the key findings of the entire
dissertation research.
13
2. Lifecycle Impacts of Hydrogen Supply Pathways on Urban
Air Quality of Primary Pollutants
2.1. Introduction
There is growing interest in hydrogen as a transportation fuel. To mitigate concerns about
urban air pollution, zero emission vehicle ( ZEV) mandates have been enacted in
California ( CARB, 2008a) and several other states in the U. S. Hydrogen fuel cell vehicles
( FCVs) are a very promising ZEV option due to a number of potential advantages such as
good performance of the FCV, various production sources of hydrogen fuel, and no CO2
and criteria pollutant emissions throughout the vehicle lifetime ( Ogden et al., 2004; NRC,
2004; Sperling and Ogden, 2004; Jacobson et al., 2005).
One of the key motivations for hydrogen is its potential to reduce emissions of air
pollutants. In contrast, current mobile sources cause significant urban air quality
degradation and damage to human health due to close proximity to people ( Chaaban et al.,
2001). In a European study, the damage costs of automotive air pollution were evaluated
by using the impact pathways approach, and the impacts involved human health,
agricultural crops, and building materials ( Spadaro and Rabl, 2001). Although hydrogen
FCVs emit no tailpipe emissions ( Ogden et al., 1999), hydrogen must be produced from
other sources and delivered to users. These steps can generate air pollutant emissions.
Thus, the entire lifecycle from well to wheels must be considered in an assessment of
hydrogen’s air quality impacts. To achieve high energy efficiencies and low overall
14
emissions, the different pathways for producing hydrogen as a transportation fuel must be
carefully examined ( Wang, 2002). Lifecycle assessment of hydrogen production via
natural gas steam reforming has been conducted in a number of studies, e. g., Spath and
Mann ( 2000). Colella et al. ( 2005) examined the potential change in primary emissions
and energy use from switching from the current U. S. on- road vehicle fleet to a hydrogen
FCV fleet, using a lifecycle analysis ( LCA) of alternative fuel supply chains, which
provides positive evidence supporting the conversion to hydrogen FCVs for
environmental and energy benefits. However, no study has quantitatively examined the
changes in ambient concentrations ( not just emissions) of pollutants resulting from
hydrogen supply pathways on a regional scale.
Clearly, emissions for hydrogen ( and associated environmental impacts) will depend on
how hydrogen is made. Furthermore, air quality is related to emissions in complex ways
that depend on many local factors, such as the mix of emission sources, meteorology, and
geography. In this chapter, we are addressing the following two research questions.
( 1) What would be the impact of hydrogen fuel cell vehicles on ambient
concentrations of primary pollutants ( NOx, VOC, CO, and particulates, as well as
SOx), considering all the emissions involved in the full fuel cycle, including
producing, transporting, and using hydrogen?
( 2) What hydrogen supply strategy would be environmentally best for a specific
county or region in the U. S.? For example, is onsite production at refueling
15
stations preferable to centralized production with delivery? Which delivery mode
for centralized hydrogen is better, liquid truck or gaseous pipeline? and
( 3) What would the optimal, feasible spatial layout of a hydrogen supply system be in
a specific region, assuming that a possible hydrogen pathway type has been
determined?
In this chapter, we develop hydrogen transportation scenarios for Sacramento, California,
and estimate regional air quality impacts for three different hydrogen production and
delivery pathways, based on steam reforming of natural gas ( NG), which is currently the
most common way of making hydrogen. Only primary pollutants that are directly emitted
from emission sources are included, and no secondary atmospheric formation like
secondary particulate matter or ozone ( O3) is considered in this chapter. Using a lifecycle
analysis approach, this research compares these hydrogen supply options, presents the
methodology to link hydrogen pathways to ambient air quality in urban Sacramento,
California, and estimates the increases in ambient pollution corresponding to some key
hydrogen supply chain steps.
2.2. Methodology
2.2.1. Overview of hydrogen pathway scenarios
From a lifecycle analysis ( LCA) perspective, we estimate regional air quality impacts for
three different hydrogen production and delivery pathways, based on steam methane
16
reforming ( SMR) of natural gas, which is a commercially available technology for
producing hydrogen today.
We examine distributed vs. centralized production. Distributed hydrogen is, in general,
economically advantageous in the early hydrogen market, as hydrogen is produced onsite
at the refueling station and no extra delivery is needed ( Ogden, 1999a; Yang and Ogden,
2007; H2A, 2008). We also analyze the two important delivery modes for centralized
hydrogen: gaseous hydrogen pipeline vs. liquid hydrogen truck. Centralized hydrogen
tends to be more economically feasible at higher market penetrations of hydrogen ( Ogden,
1999a; Yang and Ogden, 2007; H2A, 2008). A likely penetration pattern for natural gas-based
hydrogen is: onsite SMR 􀃆 liquid truck SMR 􀃆 gaseous pipeline SMR.
In this chapter, the lifecycle emissions associated with each hydrogen pathway are used
to determine the impact on urban air quality in Sacramento, California. Lifecycle
emissions include all the emissions involved in producing and delivering hydrogen to
vehicles, as well as emissions from electricity generation ( for hydrogen compression or
liquefaction) and petroleum use ( diesel fuel for hydrogen truck delivery). Not only direct
emissions resulting from the primary fuel pathway but also indirect emissions associated
with secondary fuel pathways ( sub- pathways) are taken into account.
To link hydrogen pathways to ambient air quality, we develop a methodological
framework for hydrogen pathway scenarios, as shown in Figure 2 and described further
in later sections.
17
Emission inventories
Atmospheric dispersion
model ( ISCST3)
Typical meteorological
year profile ( TMY2)
Air quality
impact
Hydrogen supply schemes
Regional hydrogen
demand
Emission factors:
extracted from GREET1.7
Spatial layout of hydrogen
pathway steps
Figure 2. Methodological framework for hydrogen pathway scenarios
2.2.2. Estimating hydrogen demand
We use the urbanized Sacramento conventional light duty ( LD) fleet in 2000 as the
baseline. We consider two scenarios, where 9% and 20% of light duty vehicles in
Sacramento are assumed to be hydrogen fuel cell vehicles ( FCVs). We keep the number
of gasoline vehicles constant in both scenarios, and add hydrogen vehicles and hydrogen
18
fuel supply systems to the Sacramento area. Thus, the total vehicle population is the sum
of the year 2000 light duty gasoline fleet and the added hydrogen vehicles. This allows us
to estimate the incremental impact of hydrogen energy systems on ambient pollution
levels in the Sacramento area, without the complexities of simultaneously reducing the
number of gasoline vehicles ( see Section 2.3.1). Table 1 shows demographic data for
Sacramento. From these and hydrogen vehicle assumptions we estimate regional
hydrogen demand for vehicle use for each scenario ( see Table 2).
In scenario 1, we add a number of hydrogen FCVs equal to 10% of the total LD fleet in
urbanized Sacramento in the year 2000. In this study, we are not replacing gasoline
vehicles with hydrogen FCVs; therefore, scenario 1 with 111,400 hydrogen vehicles
would mean hydrogen FCVs operating at a market penetration of 9% ( 10%
1 10%
=
+
). In
scenario 2, we add a number of FCVs equal to 25% of the urbanized Sacramento LD fleet
in the year 2000. Similarly, Scenario 2 with 278,600 hydrogen vehicles would mean a
hydrogen vehicle penetration of 20% ( 25%
1 25%
=
+
). The hydrogen demand is calculated
for each case ( the added 111,400 FCVs in scenario 1 and 278,600 FCVs in scenario 2).
Table 1. Demographic data for Sacramento
Parameters Value
City Population in Sacramento ( in 2000) 1 1.393 million
LD gasoline vehicle ownership 0.8 vehicles/ person
LD gasoline vehicle population ( 2000) 1.114 million
1 Obtained from Population of U. S. Urbanized Areas from the 2000 U. S. Census ( U. S. Census
Bureau, 2006).
19
Table 2. Hydrogen vehicle assumptions and hydrogen demand
Parameters Scenario 1 Scenario 2
Hydrogen FCV market penetration 9% 20%
Number of added hydrogen FCVs 111,400 278,600
Hydrogen fuel demand 78,000 kg/ day 195,000 kg/ day
Number of hydrogen stations 27 66
Fuel economy of the hydrogen FCV 60 miles/ kg hydrogen
Vehicle miles traveled ( VMT) 15,000 miles/ year/ vehicle
Hydrogen consumption 0.7 kg/ day/ vehicle
Hydrogen station size 3,000 kg/ day
Liquid truck capacity 3,000 kg liquid hydrogen
We consider only physical transport of conserved pollutants in this chapter. Thus, the
assumed background ambient pollution levels do not influence the results for the
incremental ambient concentrations due to hydrogen vehicles. Therefore, the above two
market penetration scenarios ( 9% and 20%) can be presented in the following way.
Scenario 1 explores the resulting concentrations of pollutants from a hydrogen system
with 111,400 fuel cell vehicles, a number equal to 10% of the total LD gasoline fleet in
urbanized Sacramento in the year 2000. Similarly, scenario 2 explores the resulting
concentrations of pollutants from 278,600 fuel cell vehicles, a number equal to 25% of
the total gasoline fleet in urbanized Sacramento in 2000.
Note that year 2000 was a somewhat arbitrary choice ( based on available information at
the time of analysis initially conducted) to provide a real- world basis for the number of
hydrogen vehicles to analyze.
20
2.2.3. Hydrogen supply schemes
The delivered cost of hydrogen made from natural gas was examined in detail in many
studies ( Ogden, 1999a; Ogden, 1999b; Yang and Ogden, 2007; Mintz et al., 2006). The
hydrogen supply cost via any pathway depends on many factors, e. g., the scale of supply,
demand level, feedstock cost, and other key inputs ( Leiby et al., 2006). A complete
technical and cost analysis of hydrogen production and delivery is conducted by U. S.
DOE’s H2A program ( H2A, 2008).
For each hydrogen scenario examined in this study, we assume that hydrogen supply
meets a steady state regional hydrogen demand on a daily basis. The following three
hypothetical hydrogen supply pathways, all natural gas- based, are investigated in the
study, as they are likely to be the lowest cost near- term options of supplying hydrogen
over the next few decades ( NRC, 2004).
( 1) The onsite pathway. Figure 3 illustrates onsite hydrogen production at the refueling
station;
( 2) The pipeline pathway. Figure 4 illustrates centralized hydrogen production with
gaseous hydrogen pipeline delivery systems; and
( 3) The truck pathway. Figure 5 illustrates centralized hydrogen production with liquid
hydrogen ( LH2) truck delivery systems.
21
Compression Dispensing FCV operation
emissions
SMR reformer
Onsite reforming facility
emissions
NG extraction
& cleanup
Long- distance pipeline transport
Figure 3. Natural gas- to- hydrogen pathway with onsite production
Pipeline distribution
Compression Dispensing FCV operation
emissions
H2 production
Refueling station
emissions
NG extraction
& cleanup
Pipeline transport
Figure 4. Natural gas- to- hydrogen pathway with pipeline delivery systems
Pipeline transport Truck distribution
Dispensing FCV operation
emissions
Central H2 plant
Liquefaction
emissions
H2 production
emissions
NG extraction
& cleanup
Figure 5. Natural gas- to- hydrogen pathway with liquid hydrogen truck delivery systems
22
For simplicity, emissions associated with electricity used for compressing, liquefying, or
transporting hydrogen are not shown in Figures 3- 5, although in our calculations we do
add these emissions into the total. Further details on the hydrogen supply scenarios, e. g.,
the number of hydrogen stations, are presented in Table 2.
2.2.4. Spatial layout of hydrogen pathway steps
To estimate the environmental impacts of hydrogen vehicles, we consider all emissions
associated with the system, including the following processes: feedstock extraction and
transport; fuel production, storage, distribution, and dispensing; and vehicle operation
( Wang, 1999).
Figure 6 presents the lifecycle of one of the three integrated natural gas- to- hydrogen
pathways considered. The parts of the lifecycle system included in this analysis are
enclosed by the dashed line ( Wang and Delucchi, 2005). The parts of the system outside
the dashed line are assumed to be either remote enough or low- emitting enough to have
little or no impact on air quality in urban Sacramento. As described below, emissions
from the pathways steps outside the “ dashed lines” could impact air quality in regions
outside the Sacramento area, with an attendant impact on human health. Thus, by
focusing on the Sacramento region, our method somewhat underestimates the global
impact of the hydrogen pathway.
23
Refueling station FCV operation
emissions
H2 production
emissions
NG extraction
& cleanup
Pipeline transport
Central H2 plant
emissions
Coal extraction Rail transport
emissions
emissions
Power plant Elec transmission
emissions
Oil extraction Tanker delivery Storage Truck distribution Gas station
emissions
emissions
emissions emissions
emissions
Oil refinery Pipeline
Truck distribution
Liquefaction
emissions
Figure 6. Integrated NG- to- H2 pathways ( liquid hydrogen illustration). ( The upper figures represent the
sub- pathway of electricity consumed in steps of the other fuel pathways. The lower figures depict the sub-pathway
of diesel fuel used by the hydrogen- delivery trucks.)
The spatial locations of emission sources associated with the various hydrogen pathway
supply steps have a strong influence on the regional air pollutant concentrations. In this
study, we assume particular spatial locations for each step of the hydrogen pathway:
natural gas extraction, hydrogen production, hydrogen delivery, and refueling stations, as
well as hydrogen vehicle operation. The spatial layouts of the hypothetical stations and
hydrogen plant are shown in Figures 7 and 8. Emission locations of each pathway step
are described in detail below and summarized in Table 3.
24
Figure 7. Spatial layouts of refueling stations, central plant, and air quality receptors in Sacramento ( 9%
market penetration)
25
Figure 8. Spatial layouts of refueling stations, central plant, and air quality receptors in Sacramento ( 20%
market penetration)
( 1) Natural gas extraction and transport. This pathway step is not included in the
research on air quality. Natural gas fields are located far away from Sacramento, and
therefore the impacts of natural gas extraction and pipeline transport on air quality in
urban Sacramento are neglected;
26
( 2) Centralized hydrogen production. Because of the availability of natural gas, a
central hydrogen production plant is assumed to be near currently existing natural gas-fired
power plants in south Sacramento ( see Figures 7 and 8), and is treated as a point
source of emissions;
( 3) Onsite hydrogen production at refueling stations. Emissions associated with
hydrogen production from small steam reformers at refueling stations occur at the station
sites. They are assumed to be a point source of emissions;
( 4) Electricity for hydrogen liquefaction at the central plant. Actual locations of
power plants in the Sacramento area are used to estimate incremental emissions
associated with hydrogen liquefaction at the hydrogen plant. Although electricity is
consumed at the central plant, the locations of emissions occur at those power plants
which are assumed to be a point source;
( 5) Electricity for hydrogen compression and pipeline delivery at the central plant.
This is similar to electricity use for hydrogen liquefaction ( see above);
( 6) Liquid hydrogen truck delivery. Heavy heavy- duty diesel- fueled trucks ( HHDTs)
delivering liquid hydrogen are a mobile source of emissions. Liquid hydrogen trucks are
assumed to travel on real- world highways and the actual route that each truck follows
from the central hydrogen plant to the station is determined using geographic information
system ( GIS) data on a minimum travel time basis. The number of truck trips is estimated
27
based on the assumed station size and truck capacity ( see Table 2). At the steady state,
the road segments of the truck routes are treated as a thin- and- long area source of
emissions;
( 7) Refueling stations. We choose sites for hydrogen stations from among existing
gasoline station locations in Sacramento. Hydrogen station sites are selected to minimize
the average travel time from home to the closest station for all customers, given a certain
number of stations. Customer locations are approximated using traffic analysis zones
( TAZs). The method employs GIS data and optimization techniques and is described in
detail by Nicholas ( 2004). The locations of stations in our study are shown in Figures 7
and 8. They are assumed to be a point source of emissions. Figure 7 corresponds to the
scenario of a 9% market penetration and 27 refueling stations. Similarly, Figure 8
corresponds to the scenario of a 20% market penetration and 66 refueling stations;
( 8) Electricity for hydrogen compression at refueling stations. To efficiently store and
dispense gaseous hydrogen, some electricity is consumed at refueling stations. The
treatment of emissions associated with electricity consumption at refueling stations is
similar to electricity use for hydrogen liquefaction at the central plant ( see above); and
( 9) Vehicle operation. Hydrogen fuel cell vehicle operation is assumed to emit none of
the air pollutants examined in this study ( Ogden et al., 1999). Therefore, vehicle locations
are not important for the analysis. Note that water vapor is the only emission if hydrogen
28
is used with fuel cells; in contrast, there are traces of NOx generated if hydrogen is
burned in air ( Ogden, 1999b).
Table 3. Description of hydrogen pathway steps, locations, and emissions
Hydrogen pathways Pathway steps included in the research1
Onsite hydrogen production 3( 7) 􀃆 8 􀃆 9
In this case, steps 3 and 7 are essentially the same.
Centralized hydrogen production with
pipeline delivery
2 􀃆 5 􀃆 7 􀃆 8 􀃆 9
Centralized hydrogen production with
liquid hydrogen truck delivery
2 􀃆 4 􀃆 6 􀃆 7 􀃆 9
1 The numbers in Table 3 refer to the pathway steps listed above in the text.
2.2.5. Lifecycle emission inventories
By using both emission rates and emission locations, we can develop spatially
deterministic emission inventories which are one of the most important inputs to the
subsequent air quality model. Only increases in air pollution due to emissions of primary
criteria pollutants and ozone precursors are estimated; i. e. the focus is on the following
directly emitted pollutants: carbon monoxide ( CO), nitrogen oxides ( NOx, referring to
both NO and NO2), volatile organic compounds ( VOCs, in some cases also called non-methane
organic carbon, or NMOC), and particulate matter ( PM10, referring to
particulates with an aerodynamic diameter less than 10 μm), as well as sulfur oxides ( SOx;
this refers roughly to SO2 here in the study). We do not account for re- entrainment of
PM10 and particulates from tire wear and brake wear, too. ( In the next chapter, we will
29
consider ozone production from its precursor emissions from hydrogen supply pathways
examined in the study.)
Based on location information and emission source assumptions in the above sections, we
consider the following direct emission sources: hydrogen plant or onsite production
stations, electric power plants, and diesel- fueled delivery trucks ( see Figure 6).
2.2.5.1. Hydrogen production infrastructure
To assess energy consumption and emissions of each hydrogen pathway step, models of
emissions factors and hydrogen infrastructure engineering/ economic designs are used. A
full fuel cycle energy use and emissions model, GREET1.7, which is developed and
maintained by the Argonne National Laboratory, is the source of those data on emission
factors and energy consumption of hydrogen infrastructure such as the hydrogen plant or
onsite production stations ( GREET1.7, 2006; Wang, 1999). The technologies making up
the hydrogen energy supply ( e. g., hydrogen production, compression, or liquefaction) are
assumed to have efficiency and emissions levels corresponding to current ( year 2005)
technologies ( see Table 4 for efficiencies) ( GREET1.7, 2006).
Because hydrogen systems will not be commercialized until a future year around 2020,
we here compare the technology efficiency levels of hydrogen supply for 2005 vs. 2020,
shown in Table 4 ( GREET1.7, 2006). Table 4 indicates that most technology efficiencies
will slightly increase over time. Therefore, it is conservative that we estimate the air
30
quality impacts of each hydrogen pathway using currently available energy supply
technologies which correspond to the scenario year 2005 represented in GREET1.7.
Table 4. Hydrogen supply technology efficiencies for 2005 vs. 2020, on a lower heating value ( LHV) basis
2005 2020
Key technology efficiency Onsite
pathway
Pipeline
pathway
Truck
pathway
Onsite
pathway
Pipeline
pathway
Truck
pathway
Conversion efficiency 69.0% 71.5% 71.5% 70.5% 73.0% 73.0%
Compression efficiency1 94.0% 92.5% N. A. 94.0% 92.5% N. A.
Liquefaction efficiency N. A. N. A. 70.5% N. A. N. A. 72.0%
1 Electric compressors apply to both the central plant and onsite stations.
2 Efficiency data are extracted from GREET ( GREET1.7, 2006).
2.2.5.2. Electric power plants
In this case study, we have chosen to neglect the impact of spatially distant pathway steps
( such as natural gas extraction and oil refining) on air quality in Sacramento. However,
we do consider emissions from the electricity used in hydrogen pathways steps. The
emission factors for electricity consumption are extracted from GREET1.7.
Electricity consumed in both the primary hydrogen pathway and the sub- pathways is
assumed to come from the average power mix for Sacramento. The electric generation
mix in Sacramento is derived from the U. S. Department of Energy’s eGRID2002 dataset
for the year 2000 ( eGRID, 2006). The power control area ( PCA) of interest is specified as
the Sacramento municipal utility district. There are 17 power plants serving the region
and their profiles are shown in Table 5. The electric generation mix ( i. e., percentage of
31
each kWh of electricity generated in 2000) by fuel type is summarized in Table 6. Clean
renewables ( i. e., solar, wind, and hydro) in Sacramento accounts for more than 42% of
electric generation, which means the electric grid is less polluting than other regions
when a large amount of electricity is consumed to compress or liquefy hydrogen.
Table 5. Sacramento PCA power plant profiles in 2000
Plant name County name Primary
fuel
Generator
capacity
( MW)
Annual net
generation ( MWh)
CAMINO EL DORADO Hydro 154 429969
CAMP FAR WEST PLACER Hydro 7 31560
CARSON ICE CG SACRAMENTO NG 126 556594
HEDGE PV SACRAMENTO Solar 0.2 362
JAYBIRD EL DORADO Hydro 154 612984
JONES FORK EL DORADO Hydro 12 22297
KIEFER LF SACRAMENTO Biomass 9 74731
LOON LAKE EL DORADO Hydro 82 98011
MCCLELLAN SACRAMENTO NG, Oil 74 15743 ( NG), 7 ( Oil)
PVUSA YOLO Solar 1 253
ROBBS PEAK EL DORADO Hydro 30 49464
SCA SACRAMENTO NG 150 649213
SOLANO WIND SOLANO Wind 7 6774
SOLAR SACRAMENTO Solar 2 1887
SPA SACRAMENTO NG 174 1404149
UNION VALLEY EL DORADO Hydro 47 139504
WHITE ROCK EL DORADO Hydro 230 592124
PCA total 1257 4685626
32
Table 6. Sacramento PCA electric generation resource mix in 2000
Power plant type Generation mix
Oil 0.0001%
Biomass 1.59%
NG 56.04%
Coal 0.00%
Nuclear 0.00%
Solar 0.05%
Wind 0.14%
Hydro 42.17%
Total 100.00%
2.2.5.3. Diesel- fueled delivery trucks
Heavy heavy- duty diesel- fueled trucks ( HHDTs) delivering liquid hydrogen are
considered as a mobile source of emissions. Liquid hydrogen trucks are assumed to travel
along a fixed route from the central plant and arrive at a refueling station, and then come
back along the same truck route. The emission factors of delivery trucks are from
GREET1.7. For simplicity, the truck routes, which are determined by using a GIS- based
optimization algorithm, are treated as a line source of vehicle exhaust. The number of
truck trips is estimated based on the assumed station size and truckload capacity ( see
Table 2).
2.2.6. Atmospheric transport and urban air quality
We employ a complicated simulation model ( ISCST3, introduced in later sections) for
atmospheric transport of pollutants to estimate increases in pollutant concentrations in the
33
Sacramento area for each hydrogen supply case. No chemical transformation of
pollutants is involved. We employ the spatial layouts in Figures 7 and 8 for the 9% and
20% market penetration scenarios, respectively. We estimate incremental concentrations
at nine “ receptor sites” in Sacramento, which are actual locations of air pollution
monitors in EPA’s monitoring network. These nine stations are shown in Figures 7 and 8
as triangles numbered 1 to 9. This allows us to compare the incremental changes in
ambient concentrations due to hydrogen against actual measured ambient concentrations.
2.2.6.1. Atmospheric transport
We assume that each emission source in a hydrogen pathway emits pollutants at a
constant rate. We further assume that pollutants disperse on an urban or regional scale,
and the distance from an emission source to any air quality monitor of concern is less
than 100 km, which assures that the above pollutants can be considered as conserved
pollutants ( ExternE, 2005). Studies by other researchers show that incremental annual
concentrations are of much more interest than hourly or daily fluctuations, as they are
more feasible and simpler to use to estimate yearly external costs associated with human
exposure to ambient pollution ( ExternE, 2005; Delucchi and McCubbin, 2004; McCubbin
and Delucchi, 1996).
Only physical transport of the pollutants is taken into account, without considering
chemical transformation or decaying of pollutants in the atmosphere. The equation below
predicts the time- average concentrations downwind of an elevated point source,
34
accounting for superposition due to reflection from the ground ( ExternE, 2005; Seinfeld
and Pandis, 1998; Heath et al., 2005), shown as
2 2 2
2 2
( , , ; ) exp exp ( ) exp ( )
2 2 2 2
E E
E
y z y z z
C x y z H Q y z H z H
π uσ σ σ σ σ
⎡ ⎤
2
= ⎢− ⎥⎧⎨⎪ ⎡⎢− − ⎤⎥+ ⎡⎢− + ⎤⎥⎪⎫⎬
⎣⎢ ⎦⎥⎪⎩ ⎣ ⎦ ⎣ ⎦⎭⎪
,
( Equation 1)
where
E H : effective stack height. E = physical stack height ( ) + plume rise ( H h ΔH );
( , , ; ) E C x y z H : concentration of the pollutant at a receptor location ( x, y, z) ;
Q: steady- state mass emission rate of the pollutant;
u : mean wind speed at the effective stack height. u = x / t , where t is the travel time of
the pollutant from the release point to the location ( x, y, z) ;
y σ : transverse dispersion parameter. This is the standard deviation of the transverse
concentration distribution at the downwind distance x ; and
z σ : vertical dispersion parameter. This is the standard deviation of the vertical
concentration distribution at the downwind distance x .
35
2.2.6.2. The ISC model
To estimate atmospheric concentrations of pollutants, we run a steady state Gaussian
plume dispersion model, Industrial Source Complex Short Term ( ISCST3), maintained
by U. S. EPA ( ISCST3, 2006; U. S. EPA, 1995). It works directly for point, area, volume,
and open pit sources of pollution, and by approximation to a sequence of long, thin area
sources or volume sources, a line source of pollution can be simulated as well ( U. S. EPA,
1995). It also can be used to assess air pollution from a variety of sources simultaneously.
We use this model to estimate air quality at the receptor locations.
2.2.6.3. The TMY2 dataset
Like most air quality models, ISCST3 needs an annual cycle of local or regional
meteorological information to predict the pollutant dispersion. The Typical
Meteorological Year ( TMY2), developed by the National Renewable Energy Laboratory
( NREL), is a complete annual cycle of hourly meteorological data extracted from the 30-
year period spanning 1961- 1990 to represent a typical long- term meteorological
condition in a specific region ( TMY2, 2006). To run ISCST3, the hourly meteorological
data such as the hour of day, wind direction, wind speed, ambient temperature,
atmospheric stability class, rural mixing height, and urban mixing height are needed. The
TMY2 dataset for Sacramento County is adopted in this research to predict changes in
ambient air pollution under a historically representative meteorological condition rather
than a worst- case condition ( TMY2, 2006; Heath et al., 2005; Heath, 2005).
36
Figure 9 illustrates the Sacramento windrose for 2005 ( including wind speeds and
directions) ( WRCC, 2008), and this windrose pattern is very typical of this region
although it is not derived based on the TMY2 dataset. Note that the regional prevailing
wind direction is from southwest to northeast, shown in Figure 9.
Figure 9. Sacramento windrose for 2005 ( including wind speeds and directions)
2.2.6.4. The Air Quality System ( AQS) for air pollution monitors
The Air Quality System ( AQS) maintained by U. S. EPA contains ambient air pollution
data and profiles of thousands of air quality monitoring stations throughout the country;
37
states, local and tribal agencies submit their data directly to AQS, and EPA itself also
collects data ( AQS, 2006). There are nine appropriate air monitoring stations officially
maintained within or close to urban Sacramento based on the AQS system. These stations
serve as receptors of pollutants in the research, and their profiles are shown in Table 7
( AQS, 2006). Figures 7 and 8 present their spatial layouts in Sacramento. The individual
incremental concentrations at these receptors and their average values represent the
ambient pollution level attributable to each of hydrogen pathways. Note, however, that a
receptor is not necessarily a typical representative of urban air quality when it happens to
be located very close to a truck route or a refueling station.
Table 7. Air quality monitors in urban Sacramento ( i. e., receptors of pollutants)
Monitor Name and address
1 Sacramento- 3801 Airport Road
2 West Sacramento- 15th Street
3 Sacramento- T Street
4 Sacramento- Health Dept Stockton Blvd
5 Folsom- Natoma Street
6 Sacramento- Branch Center Road
7 Sacramento- El Camino
8 North Highlands- Blackfoot Way
9 Sacramento- Del Paso Manor
2.2.6.5. The NAAQS standards and actual measurements
The Clean Air Act, last amended in 1990, requires U. S. EPA to set the National Ambient
Air Quality Standards ( NAAQS) to protect public health, and Table 8 shows NAAQS
38
primary standards ( NAAQS, 2006; CARB, 2006). Table 8 also presents the actual
measurements of pollution level in Sacramento in 2000, which is calculated based on the
AQS dataset above. It is important to keep these ambient “ baseline” concentrations in
mind, as we discuss the incremental concentrations due to additional large numbers of
hydrogen vehicles.
Table 8. NAAQS and ambient measurements in Sacramento in 2000
NAAQS ( EPA, 1990)
Pollutant
Primary standards Averaging times
Sac. 2000 annual
aver. conc. ( μg/ m3) 1
CO 9 ppm ( 10 mg/ m3) 8- hour 639.69
35 ppm ( 40 mg/ m3) 1- hour
VOC No standards N. A. 74.80 ( NMOC)
NO2 0.053 ppm ( 100μg/ m3) Annual ( Arith. Mean) 56.64
PM10 50 μg/ m3 Annual ( Arith. Mean) 2 22.45
150 μg/ m3 24- hour
SOx 0.03 ppm ( 80 μg/ m3) Annual ( Arith. Mean) 7.92
0.14 ppm ( 365 μg/ m3) 24- hour
1 The Sacramento 2000 annual average ambient measurements are calculated based on data from U. S.
EPA’s Air Quality System ( AQS).
2 EPA revoked the annual PM10 standard in 2006 ( effective December 17, 2006).
2.3. Results and discussion
2.3.1. Incremental pollution attributable to hydrogen pathways
We use the ISCST3 program to estimate the additional pollution at a receptor for each of
our three hypothetical hydrogen pathways, at each of two market penetrations, 9% and
20%, respectively. Figures 10- 19 present the magnitudes of incremental annual average
concentrations of conserved pollutants due to existence of hydrogen pathways. There are
39
three pathways ( i. e., the onsite pathway, the pipeline pathway, and the truck pathway),
five pollutants ( i. e., CO, VOC, NOx, PM10, and SOx), and nine pollution receptors
( denoted by R1 through R9). It is easy to see that environmental impacts vary with
receptor site, which reflects the location variations and geographic factors, even when
they are attributable to the same hydrogen pathway.
The first thing to note is that all three hydrogen supply pathways result in very small
incremental amounts of pollution compared to ambient pollution levels, on the order of
0.1% increase in concentrations at the 20% market penetration, and often much less.
This is in contrast to recent results for natural gas- based distributed generation of
electricity in California, which resulted in more air pollution than central power plants
( Heath et al., 2005).
For the truck pathway, emissions tend to be higher than for the other two supply
pathways, though they are still small. As shown in Figures 20- 24, most of the emissions
for the liquid truck pathway are due to diesel truck emissions resulting from the delivery
of the liquid hydrogen and to the electricity used to liquefy the product hydrogen.
For the onsite scenario, there are no hydrogen delivery emissions since all the hydrogen
fuel is produced and dispensed onsite at the refueling stations. Also, the emissions are
distributed throughout the metropolitan area so the wind direction has little impact on the
average air pollution at receptors.
40
It can be seen in the following graphs that the change in air quality due to the onsite
scenario is comparable to that caused by the central hydrogen pathway with pipeline
systems, and both are very clean. The truck pathway also results in very low incremental
pollution levels, but higher than concentrations resulting from the other two pathways.
Meteorological conditions, especially wind directions, have a large impact on the effect
of emissions from the central plant. The prevailing wind in Sacramento is from southwest
to northeast and seldom from east to west. The site of the central plant can be
strategically located so as to minimize the effect on urban air quality. In our example, the
site is somewhat advantageous in that it is only occasionally upwind of the urban area.
The site for the central plant could be further improved by placing it east of the
metropolitan area since this location is almost always downwind of the urban region.
Furthermore, it would be meaningful to carry out a sensitivity analysis regarding the
central hydrogen plant siting, although it is beyond the scope of this research. Again,
geographic conditions have a significant effect on the impact of hydrogen production on
urban air quality.
41
Incremental Annual CO Conc. ( μg/ m^ 3):
9% Scenario
0.0E+ 00
2.0E- 03
4.0E- 03
6.0E- 03
8.0E- 03
1.0E- 02
1.2E- 02
1.4E- 02
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 10. Incremental annual average concentrations of CO by receptor ( 9% scenario)
Incremental Annual CO Conc. ( μg/ m^ 3):
20% Scenario
0.0E+ 00
5.0E- 03
1.0E- 02
1.5E- 02
2.0E- 02
2.5E- 02
3.0E- 02
3.5E- 02
4.0E- 02
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 11. Incremental annual average concentrations of CO by receptor ( 20% scenario)
42
Incremental Annual VOC Conc. ( μg/ m^ 3):
9% Scenario
0.0E+ 00
2.0E- 04
4.0E- 04
6.0E- 04
8.0E- 04
1.0E- 03
1.2E- 03
1.4E- 03
1.6E- 03
1.8E- 03
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 12. Incremental annual average concentrations of VOC by receptor ( 9% scenario)
Incremental Annual VOC Conc. ( μg/ m^ 3):
20% Scenario
0.0E+ 00
1.0E- 03
2.0E- 03
3.0E- 03
4.0E- 03
5.0E- 03
6.0E- 03
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 13. Incremental annual average concentrations of VOC by receptor ( 20% scenario)
43
Incremental Annual NOx Conc. ( μg/ m^ 3):
9% Scenario
0.0E+ 00
5.0E- 03
1.0E- 02
1.5E- 02
2.0E- 02
2.5E- 02
3.0E- 02
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 14. Incremental annual average concentrations of NOx by receptor ( 9% scenario)
Incremental Annual NOx Conc. ( μg/ m^ 3):
20% Scenario
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 15. Incremental annual average concentrations of NOx by receptor ( 20% scenario)
44
Incremental Annual PM10 Conc. ( μg/ m^ 3):
9% Scenario
0.0E+ 00
2.0E- 04
4.0E- 04
6.0E- 04
8.0E- 04
1.0E- 03
1.2E- 03
1.4E- 03
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 16. Incremental annual average concentrations of PM10 by receptor ( 9% scenario)
Incremental Annual PM10 Conc. ( μg/ m^ 3):
20% Scenario
0.0E+ 00
5.0E- 04
1.0E- 03
1.5E- 03
2.0E- 03
2.5E- 03
3.0E- 03
3.5E- 03
4.0E- 03
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 17. Incremental annual average concentrations of PM10 by receptor ( 20% scenario)
45
Incremental Annual SOx Conc. ( μg/ m^ 3):
9% Scenario
0.0E+ 00
2.0E- 04
4.0E- 04
6.0E- 04
8.0E- 04
1.0E- 03
1.2E- 03
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 18. Incremental annual average concentrations of SOx by receptor ( 9% scenario)
Incremental Annual SOx Conc. ( μg/ m^ 3):
20% Scenario
0.0E+ 00
5.0E- 04
1.0E- 03
1.5E- 03
2.0E- 03
2.5E- 03
3.0E- 03
3.5E- 03
4.0E- 03
Onsite Pipeline Truck
R1
R2
R3
R4
R5
R6
R7
R8
R9
Figure 19. Incremental annual average concentrations of SOx by receptor ( 20% scenario)
46
2.3.2. Comparison to the current ambient measurements
Table 9 compares the estimated incremental pollution, resulting from adding large
numbers of hydrogen vehicles, to the actual measured mean concentration of pollutants in
Sacramento, averaging over all the nine receptors. For the 9% market penetration
scenario, the onsite pathway leads to incremental pollution fractions ranging from
0.0006% ( SOx, or approximately CO) to 0.0116% ( NOx) of current ambient
concentrations. The pipeline pathway leads to pollution fractions ranging from 0.0005%
( CO) to 0.0158% ( NOx), and the truck pathway leads to pollution fractions ranging from
0.0013% ( CO) to 0.0337% ( NOx). For the 20% market penetration scenario, the onsite
pathway leads to pollution fractions ranging from 0.0015% ( SOx, or approximately CO)
to 0.0347% ( NOx), the pipeline pathway leads to pollution fractions ranging from
0.0012% ( CO) to 0.0396% ( NOx), and the truck pathway leads to pollution fractions
ranging from 0.0036% ( CO) to 0.0952% ( NOx).
Relatively speaking, hydrogen pathways contribute the least fractions to ambient CO and
the most fractions to ambient NOx. This makes sense because most CO is released from
on- road mobile sources ( as compared to hydrogen supply pathways involving mainly
stationary emission sources), whereas NOx tends to be released from large stationary
sources resulting from combustion, e. g., to produce the high temperature steam which is
used in the central hydrogen plant or in a power plant.
47
In summary, for all scenarios, the incremental pollution due to adding hydrogen cars at a
9% or 20% market penetration is negligible. Truck pathways contribute more than onsite
or central/ pipeline pathways, but all lead to extremely low air pollution.
Table 9. Comparison of estimated incremental ambient pollution due to hydrogen pathways and ambient
measurements
Onsite pathway Pipeline pathway Truck pathway
Pollutant Market
penetration Mean conc.
( μg/ m3)
Pollution
fraction
Mean conc.
( μg/ m3)
Pollution
fraction
Mean conc.
( μg/ m3)
Pollution
fraction
CO 9% 0.00473 0.0007% 0.00316 0.0005% 0.00848 0.0013%
20% 0.01423 0.0022% 0.00791 0.0012% 0.02331 0.0036%
VOC 9% 0.00043 0.0006% 0.00043 0.0006% 0.00114 0.0015%
20% 0.00126 0.0017% 0.00107 0.0014% 0.00326 0.0044%
NOx 9% 0.00654 0.0116% 0.00896 0.0158% 0.01909 0.0337%
20% 0.01967 0.0347% 0.02241 0.0396% 0.05394 0.0952%
PM10 9% 0.00046 0.0021% 0.00048 0.0021% 0.00087 0.0039%
20% 0.00146 0.0065% 0.00119 0.0053% 0.00229 0.0102%
SOx 9% 0.00005 0.0006% 0.00004 0.0006% 0.00063 0.0079%
20% 0.00012 0.0015% 0.00011 0.0014% 0.00198 0.0250%
2.3.3. Further comparison among hydrogen pathways
Table 9 also shows a comparison of pathways in terms of resulting regionwide mean
pollution. The truck pathway results in more pollution especially for SOx, with
concentrations more than an order of magnitude higher than those from the other
pathways. Liquid hydrogen trucks fueled with sulfur- containing diesel make the biggest
contribution to ambient SOx concentrations. This is due to several factors: the trucks run
on U. S. conventional diesel with an estimated sulfur mixing ratio of 200 ppm by mass
48
( GREET1.7, 2006); steam reforming of natural gas is very clean in terms of sulfur-containing
emissions; and electric generation is relatively clean in Sacramento because
renewables account for a very large share of production ( see Table 6). The onsite
pathway and the pipeline pathway result in very similar pollution levels, especially in
terms of VOC, PM10, and SOx. However, the onsite pathway leads to more CO and less
NOx pollution than the pipeline pathway.
The incremental pollution due to each of hydrogen pathways, with the exception of the
pipeline pathway, is not exactly proportional to the regional hydrogen demand,
represented by the different hydrogen FCV market penetrations in this study ( see Table
9). When the added hydrogen vehicle population increases by 2.5 times ( from 10% of the
year 2000 light duty fleet up to 25% of the fleet), the pollution ratio increases by slightly
more than 2.5 times, with the exception of the onsite pathway, whose SOx pollution is
slightly lower than 2.5 times. For the pipeline pathway, it is 2.5 times greater because it is
assumed that the NG to hydrogen conversion efficiency remains the same as hydrogen
demand goes up, holding the electric generation mix constant.
2.3.4. Source contributions to incremental ambient pollution
Based on the locations of emissions, the sources of ambient pollution are categorized into
the following groups ( ignoring the other emission sources that are spatially far away from
urban Sacramento).
49
( 1) Hydrogen plant. This group includes the central hydrogen plant or onsite
hydrogen production stations. Only emissions directly released at these locations are
taken into consideration, and electricity consumed in a hydrogen plant is traced back
to power plants that are referred to as another source contributor to ambient pollution;
( 2) Power plant. This group includes all the 17 power plants that contribute to the
electric generation mix in Sacramento; in fact, only 5 power plants contribute to the
urban air quality since the other 12 power plants, accounting for 42.36% of power
mix, are on a clean energy basis ( i. e., solar, wind, and hydro). For simplicity, only
emissions directly released at the power plants are taken into consideration, i. e.
ignoring emissions upstream of power plants; and
( 3) Truck route. This group only applies to the hydrogen pathway with liquid
hydrogen truck delivery systems. The direct emissions are mainly diesel truck tailpipe
emissions.
The source contributions to ambient pollution averaged over nine urban receptors of
interest are presented in Figures 20- 24. For the pipeline pathway, the hydrogen plant
accounts for the largest share of pollution, and its contributions ( for all the nine air
quality monitors) are typically larger than 70%, and in some cases even larger than 80%.
The exception is SOx pollution, which is almost 100% from power plants.
50
For the onsite pathway, the hydrogen production stations account for the largest share,
typically more than 70%. And again, SOx pollution is the exception as power plants
account for almost all of the SOx pollution. Some receptors are affected by onsite stations
much more severely, especially receptors that are next to and downwind from one or
more stations. On average, hydrogen stations contribute around 90% of incremental
pollution at receptors.
For the truck pathway, there are mainly three pollution components: the truck routes,
hydrogen plant, and power plants. For all the five pollutants, truck routes and power
plants are very important. The hydrogen plant contributes the smallest share, around 10%
to 30%, and essentially 0% in terms of SOx. Truck routes contribute 20% to 40% of
pollution at a receptor, and particularly lead up to 70% in terms of SOx pollution. Power
plants contribute around 30% of pollution at a receptor for NOx and SOx, and they
contribute around 50% pollution for the other three pollutants.
0%
20%
40%
60%
80%
100%
Pipeline, 9% Onsite, 9% Truck, 9% Pipeline, 20% Onsite, 20% Truck, 20% CO emssion source contribution
Truck route
Power plant
H2 plant
Figure 20. Source pollution shares averaged over all receptors ( CO)
51
0%
20%
40%
60%
80%
100%
Pipeline, 9% Onsite, 9% Truck, 9% Pipeline, 20% Onsite, 20% Truck, 20%
VOC emssion source contribution
Truck route
Power plant
H2 plant
Figure 21. Source pollution shares averaged over all receptors ( VOC)
0%
20%
40%
60%
80%
100%
Pipeline, 9% Onsite, 9% Truck, 9% Pipeline, 20% Onsite, 20% Truck, 20%
NOx emssion source contribution
Truck route
Power plant
H2 plant
Figure 22. Source pollution shares averaged over all receptors ( NOx)
52
0%
20%
40%
60%
80%
100%
Pipeline, 9% Onsite, 9% Truck, 9% Pipeline, 20% Onsite, 20% Truck, 20%
PM10 emssion source contribution
Truck route
Power plant
H2 plant
Figure 23. Source pollution shares averaged over all receptors ( PM10)
0%
20%
40%
60%
80%
100%
Pipeline, 9% Onsite, 9% Truck, 9% Pipeline, 20% Onsite, 20% Truck, 20%
SOx emssion source contribution
Truck route
Power plant
H2 plant
Figure 24. Source pollution shares averaged over all receptors ( SOx)
2.4. Conclusions
We have examined the potential regional air quality impacts of hydrogen transportation
fuel from a lifecycle analysis ( LCA) perspective, including impacts from fuel production,
53
delivery, and vehicle use. The analysis is a case study for a specific region, Sacramento,
California. We considered two levels of market penetration where 9% or 20% of the light
duty fleet are hydrogen fuel cell vehicles. Three natural gas- based hydrogen supply
pathways were considered: onsite production via small- scale steam methane reformer
( SMR), large central SMR with liquid truck delivery, and large central SMR with
pipeline delivery.
The contributions of each hydrogen pathway to ambient air pollution were estimated
using a physical transport model for primary air pollutants NOx, CO, VOC, particulates,
and SOx. ( We used a Gaussian plume dispersion model for the atmospheric transport of
pollutants. However, it does not include chemical production of secondary pollutants
such as O3 formed by the precursors VOC and NOx in the presence of sunlight. Therefore,
these results are not a complete air quality impact assessment of potential hydrogen
pathways.)
The pollution levels associated with each of the hydrogen scenarios are dependent upon
the location of emitters and receptors, regional meteorological conditions, and geographic
factors. The spatial layout of pathway steps therefore plays an important role in
determining ambient pollution levels at air quality monitoring stations. We find that all
the hydrogen pathways considered are associated with extremely low pollution levels
relative to current ambient air concentrations of NOx, CO, VOC, particulates, and SOx.
For the 9- 20% hydrogen scenarios examined, the results are typically less than 0.1% of
the current ambient pollution.
54
Among the hydrogen supply options, it is found that central SMR with pipeline delivery
is the least polluting option available provided the plant is located to avoid transport of
pollutants into the city via prevailing winds. The onsite hydrogen pathway is comparable
to the central hydrogen pathway with pipeline systems in terms of the resulting air
pollution. The pathway with liquid hydrogen trucks has a greater impact on air quality
relative to the other pathways due to emissions associated with diesel trucks and
electricity consumption to liquefy hydrogen. The truck pathway causes more pollution
than the onsite pathway and the pipeline pathway. The truck pathway causes around 15
times more SOx, and around 3 times more of the other pollutants, than the other two
pathways. For the pipeline or onsite pathway, hydrogen production accounts for the
largest share of pollution. For the electricity- intensive liquid hydrogen truck pathway,
emissions from diesel- truck delivery and electric generation at power plants are much
more important than hydrogen plant emissions in terms of the resulting pollution. Again,
compared to measured ambient concentrations, all the three hydrogen pathways result in
negligible air pollution in the region.
In summary, the results in this chapter show that emissions from near- term hydrogen
production and delivery systems would make negligible contribution to ambient urban air
pollution. In later chapters, we compare the emissions from hydrogen systems to those
from advanced gasoline vehicles.
55
3. Estimating Changes in Urban Ozone Concentrations Due to
Lifecycle Emissions from Hydrogen Transportation Systems
3.1. Introduction
This chapter builds on methods and results from Chapter 2 to examine the potential
impact of introducing a hydrogen- based transportation system on urban ambient ozone
concentrations.
Hydrogen has been proposed as a low- polluting alternative transportation fuel that could
help improve urban air quality. The tailpipe emissions of a hydrogen fuel cell vehicle
( FCV) are strictly zero under all operating conditions ( Ogden et al., 1999), but there
would be some emissions related to hydrogen considering the full fuel cycle, including
hydrogen production and delivery. These processes directly emit primary criteria
pollutants and precursors to secondary ozone ( O3), including nitrogen oxides ( NOx, i. e.,
NO2 and NO) and volatile organic compounds ( VOCs, also called non- methane organic
carbon, or NMOC). Ozone is of great concern because it is harmful to human health and
agricultural crops and thus can become a social cost issue ( ExternE, 1998; McCubbin and
Delucchi, 1996; Delucchi et al., 1998; Murphy et al., 1999; Delucchi and McCubbin,
2004). In this chapter, we are addressing the following two research questions.
( 1) What changes in peak ozone pollution would occur if typical hydrogen supply
pathways were introduced in Sacramento, California, considering all the
56
emissions involved in the full fuel cycle, including producing, transporting, and
using hydrogen? and
( 2) What hydrogen supply strategy would be the best for the specific region,
Sacramento, California, in terms of the resulting secondary ozone pollution?
In the previous chapter the three common natural gas- to- hydrogen pathways were
examined, and the incremental pollution of NOx, VOC, carbon monoxide ( CO),
particulate matter ( PM10), and sulfur oxides ( SOx, referring roughly to SO2 here) was
quantified based on the atmospheric physical transport of directly emitted pollutants.
However, no atmospheric chemical reactions like ozone formation were considered.
Going a step further, this chapter develops a region- specific regression model to predict
atmospheric ozone formation. From a lifecycle analysis ( LCA) perspective, the same
three hydrogen pathways are compared in terms of the resulting changes in peak ozone
pollution in urban Sacramento. Predictions of the potential ozone pollution caused by
each of the hydrogen pathways are compared to the current ambient pollution levels.
3.2. Literature review on predictors of ozone formation
Ozone pollution and episodes mainly occur during the daylight hours of the summer
months ( NRC, 1991). In summary the “ high ozone days” are likely affected by such
parameters as the ground- level temperature, upper air temperature, dew point temperature,
wind speed, wind direction, solar radiation or cloud cover, and relative humidity or
precipitation ( NRC, 1991). In addition to the meteorological conditions that lead to
57
ozone episodes, the characteristics and chemical composition of VOC have an impact on
ozone formation, and different species of VOC differ in their photochemical ozone
creation potential ( Derwent et al., 1996). Prediction of ozone formation has improved
over the decades with the development of three- dimensional photochemical transport
models, but they often include a hundred or more coupled reactions just to describe gas
phase changes along with detailed meteorology, and yet may only yield results accurate
to about 25%. For the purposes of this study, a more efficient region- specific method of
estimating the magnitude of the effects of different hydrogen pathways on ozone
production was sought.
One such method might be to generate region- specific ozone isopleth data. Ideally, ozone
isopleth diagrams can be produced by smog chamber experiments. In practice, the
empirical kinetic modeling approach ( EKMA) developed by the U. S. EPA relates the
maximum hourly average ozone concentrations with the 6: 00- 9: 00am average of
precursor concentrations in a region, and both standard and city- specific ozone isopleths
can be generated ( Kinoslan, 1982). Figure 25 shows a typical pattern of ozone isopleths
used in EPA's EKMA ( NRC, 1991, citing Dodge's work). The NOx- limited region in the
ozone isopleth diagram is typical of locations downwind of urban and suburban areas,
and in contrast the VOC- limited region is typical of highly polluted urban areas ( NRC,
1991). From such studies it is recognized that the VOC and NOx precursors to
atmospheric ozone formation often yield a peak ozone when the VOC/ NOx ratio is
around 7- 10 ( Chang et al., 1989), and that ozone formation is retarded by additional NOx
emissions when the VOC/ NOx ratio is less than 5.5 ( Seinfeld and Pandis, 1998).
58
O3 ( ppm)
NOx ( ppm)
VOC ( ppmC)
VOC
NOx
Figure 25. Typical pattern of ozone isopleths used in EPA's EKMA ( NRC, 1991, citing Dodge's work).
( Note that the values in the figure are not necessarily true in reality for all cities, as ozone isopleths are
usually developed empirically for a specific region.)
Regression modeling approaches yield useful region- specific information when sufficient
measurements are available. Based on daily air quality monitoring results for the
downtown Los Angeles station for the months of August, September, and October, and
using the 3- hour ( 6- 9am) averages of total hydrocarbon and NOx concentrations and the
maximum hourly average oxidant concentration occurring on that day, an empirical
model of ozone production was derived ( Merz et al., 1972). Multiple regression modeling
was also conducted to simulate the peak ozone produced by Los Angeles air in outdoor
smog chambers, using the independent variables HC ( i. e., hydrocarbon), NOx, and the
59
average daily temperature ( Kelly and Gunst, 1990). That study was the first published to
quantify the effect of temperature on peak ozone formation in captive air studies ( Kelly
and Gunst, 1990).
3.3. Methodology
3.3.1. Overview of methodology
This research develops a region- specific regression model to predict atmospheric ozone
formation associated with a hydrogen transportation system. Our methodology has
several steps.
First, we estimate the emissions of ozone precursors from the various steps ( and locations)
along selected hydrogen supply pathways. We then use an atmospheric dispersion model
ISCST3 to find the precursor concentrations throughout the Sacramento area, using data
for typical meteorological conditions. For more details, see chapter 2. Next, we use the
meteorological information and air quality data from 2004 to derive the relationship
between ozone formation and its precursor concentrations by using regression analysis.
Finally we use the regression- based model to estimate the incremental ozone
concentrations due to hydrogen pathways.
60
This methodology should give a reasonable estimate of ozone concentrations in future
years when hydrogen might be widely used, say in 2025 or beyond. 1
3.3.2. Hydrogen pathway scenarios and dispersion model applications
Consistent with earlier chapters, this part of the research is built upon the same hydrogen
pathway assumptions and methodological framework used for estimating primary
pollution, resulting from physical transport only and without considering chemical
transformation. See Chapter 2 for more details on the hydrogen pathway scenarios and air
quality model applications.
We assume that in Sacramento, California, the current light duty gasoline fleet is held
constant and still on the road in the same numbers, whereas additional hydrogen fuel cell
vehicles have been introduced and are operating at market penetrations of 9% and 20%,
respectively. Thus, the total vehicle population is the sum of the current light duty fleet
1 We used the 2004 ambient VOC and NOx concentrations as the " baseline" from which changes were
calculated on a daily basis. The rationale is as follows: first, there is no ambient air quality standard for
VOC, and second the NOx air quality in 2004 ( see Figure 27) met the air quality standard. Thus, while one
can try to project improved VOC and/ or NOx air quality in a future year, say 2025, based upon a state
implementation plan ( SIP), the degree of improvement is difficult to estimate since regional growth will
likely reduce any gains due to emission control strategies. Also, we used daily air quality data to predict
ozone for a given day and there is no good way to predict the VOC and NOx concentrations on a daily basis
in a distant future year.
To estimate the typical changes in VOC and NOx concentrations, the year 2004 meteorology was not used;
instead, a typical meteorological year ( TMY2) was used, which was extracted for the region statistically
from 1961- 1990 and is the most representative meteorology ( TMY2, 2006). The TMY2 is a complete
annual cycle of hourly meteorological data extracted from the 30- year period to represent a typical, rather
than a worst- case, long- term meteorological condition in a specific region.
The choice of using 2004 initial VOC and NOx concentrations in conjunction with the TMY2 meteorology
as inputs to the regression model could be considered " inconsistent," i. e., in the sense that a high initial
VOC and NOx concentration level could have been used on a day with good ventilation ( meteorology),
however running the analysis for an entire season averages out the impact of such occurrences.
61
and added hydrogen vehicles. Table 10 shows the two scenarios of estimated hydrogen
demand for regional vehicle use. See Section 2.2.2 ( in Chapter 2) for more details and
discussions.
Table 10. Regional hydrogen demand for vehicle use ( part of Table 2)
Scenario 1 Scenario 2
Hydrogen FCV market penetration 9% 20%
Number of hydrogen FCVs 111,400 278,600
Hydrogen fuel demand 78,000 kg/ day 195,000 kg/ day
Number of hydrogen stations 27 66
We assume that in a steady state, hydrogen demand meets hydrogen supply on a daily
basis. To meet the hydrogen demand, three hypothetical hydrogen supply pathways were
considered in this study: onsite hydrogen production, centralized hydrogen production
with pipeline delivery, and centralized hydrogen production with liquid hydrogen truck
delivery.
The lifecycle emissions of ozone precursors associated with each hydrogen pathway are
used to determine the impact on ozone production. Based on the lifecycle emission
inventories and location information, a Gaussian dispersion model ISCST3 was run,
together with Typical Meteorological Year ( TMY2) data for the region as the
meteorological inputs to the model ( ISCST3, 2006; U. S. EPA, 1995). The TMY2 dataset
is a complete annual cycle of hourly meteorological data extracted from the 30- year
period spanning 1961- 1990 to represent a typical, rather than a worst- case, long- term
62
meteorological condition in a specific region ( TMY2, 2006). Therefore, the typical
incremental concentrations of ozone precursors at each receptor ( i. e., the air quality
monitoring stations, shown in Figures 7 and 8) due to atmospheric transport of emissions
associated with hydrogen pathways can be determined. The incremental VOC and NOx
concentrations were then added to the baseline VOC and NOx concentrations and used to
estimate subsequent ozone formation, though not necessarily in proximity to the origin of
the emissions.
3.3.3. Data and the ozone regression model
The air quality data used in this research are selected from the Air Quality System ( AQS),
which is maintained by U. S. EPA and also contains profiles of many air quality
monitoring stations throughout the country ( AQS, 2006). However, the data on VOC and
ozone are not complete for the year 2004. Even though more than 10 air quality
monitoring stations were operating in Sacramento County in 2004, only one of them ( see
station 9 shown in Figures 7 and 8) has relatively good- quality VOC data available.
Figure 26 illustrates the Sacramento windrose for 2004 ( WRCC, 2008), and this windrose
pattern is very typical of this region. Note that the prevailing wind in the region is
commonly in the direction from southwest to northeast in the summer months ( see Figure
26), therefore station 5 is very often downwind of station 9, and it makes sense to assume
that the early morning ( say, 6: 00 to 9: 00am) pollution level at station 9 provides the
initial VOC and NOx concentrations that form ozone that reaches a maximum at the
downwind station 5 ( see Figures 7 and 8). This study uses data for 93 days in the summer
63
( most of the period July 3, 2004 through October 26, 2004), the season during which
ozone pollution mainly occurs.
Figure 26. Sacramento windrose for 2004 ( including wind speeds and directions)
An intrinsically linear regression model was developed to explore the relationship of
ozone formation specific to the region. Consider initially that the peak ozone
concentration is related to a number of factors, shown as
64
3 ( max) ( , , ( max), ( ), , , .) x O f VOC = NO Temp RH avg Solar radiation Wind speed etc ,
( Equation 2)
where
f: represents a certain functional relationship;
O3( max): the peak ozone concentration ultimately reached ( i. e., 1- hour maximum ozone
concentration of the day) at the receptor station 5 ( representing a location downwind of
the urban area and often reflecting the maximum ozone pollution level observed in
Sacramento), in units of parts per billion ( ppb) by volume;
VOC ( or NOx): the initial ambient VOC ( or NOx) concentration at air quality monitoring
station 9 ( representing a typical central urban Sacramento pollution level). They are the
3- hour average of the ambient VOC ( or NOx) concentrations during 6: 00 to 9: 00am, in
units of ppbC ( or ppb) by volume;
Temp( max): the 1- hour maximum temperature of the day, degrees Celsius; and
RH( avg): the daily average relative humidity, %.
Using regression analysis, only four factors, namely the initial VOC concentration, initial
NOx concentration, maximum hourly temperature of the day, and daily average relative
65
humidity, were observed to be statistically significant and theoretically meaningful, and
were therefore selected. The initial ambient VOC, the initial ambient NOx, and the
observed ambient peak ozone during the period of regression are shown in Figure 27, and
the meteorological conditions used in the regression model are shown in Figure 28.
0
50
100
150
200
250
20040703
20040708
20040712
20040716
20040727
20040731
20040804
20040808
20040812
20040816
20040902
20040906
20040910
20040914
20040918
20040922
20040926
20041001
20041005
20041009
20041014
20041018
20041022
20041026
Ambient conc. ( ppb)
VOC ( ppb C)
NOx ( ppb)
Obseved O3 ( ppb)
Figure 27. Actual initial ambient concentrations of pollutants used in the regression ( 6: 00- 9: 00am)
66
0
5
10
15
20
25
30
35
40
45
20040703
20040708
20040712
20040716
20040727
20040731
20040804
20040808
20040812
20040816
20040902
20040906
20040910
20040914
20040918
20040922
20040926
20041001
20041005
20041009
20041014
20041018
20041022
20041026
Temp. ( max, ℃)
0
10
20
30
40
50
60
70
80
90
100
RH ( avg, %)
Temp.( max,℃)
RH ( avg,%)
Figure 28. Actual meteorological conditions used in the regression
The following linear regression model ( or equation) was estimated with Ordinary Least
Squares ( OLS) and ascertained to best correspond to all the groups of air quality data ( see
Equation 3).
2
3 ( max) 54.268 3.069 ( max) 0.406 ( ) 0.474 0.521
(- 2.644) ( 7.628) ( 2.478) ( 2.376) (- 2.017)
O = − + Temp + RH avg + NOx− NOx VOC
( Equation 3)
The coefficient of determination is R2= 0.65 for the regression, and the sample size is
N= 93. The numbers in parentheses under the equation are t- statistics for the
corresponding regression coefficients. The t- distribution table shows that at a significance
level of 0.05, the critical value of t- statistic is 1.9873, with 88 degrees of freedom for the
67
above regression. That means if the magnitude of the t- statistic for a regression
coefficient exceeds 1.9873, we can say with at least 95% confidence that the regression
coefficient is significantly different from zero. The Durbin- Watson test gives DW= 1.580,
which provides no evidence of the existence of autocorrelation in the model specification
using the time series data.
The regression is limited to the training dataset and hence to the corresponding region
and time period. As we are only interested in the relationship between ozone and its
precursors VOC and NOx, the explanatory variables Temp( max) and RH( avg) are the
control variables for our analysis.
The explanatory variable, Temp( max), is extremely important, and its coefficient is
significantly different from zero ( t- statistic is 7.628). The temperature effect is correlated
with sunlight and also other meteorological conditions associated with the build- up of
pollutants, e. g., low wind speed, so it is possible that several meteorological factors are
included in Temp( max). However, the effect of RH appears to be largely separate from
that of Temp( max), as model hypothesis testing and auxiliary regressions ( equivalently,
variance inflation factors, or VIFs) have not identified strong collinearity between
Temp( max) and RH( avg) for this training dataset even though RH is often inversely
related to temperature.
Based on the standard ozone isopleth diagram produced by EKMA ( see Figure 25), there
can be divergent predictions of ozone formation ( increment or decrement) depending, to
68
some extent, on the ambient ratio of initial VOC to NOx as the concentration of NOx
increases. This is reflected by the regressor 2 / , which is equal to the NO x NO VOC x
concentration divided by the ratio of VOC to NOx. The ratios for the training dataset
happen to be within 0 ( see Figure 29). There is no case in which the
ratio is greater than 20 in the dataset of the research, so the regression
equation should perhaps not be applied to situations where . Intuitively,
the functional relationship where is very likely to be, or be close to, a
straight line parallel to the VOC axis based on a standard EKMA ozone isopleth diagram
when plotted in terms of initial VOC and NO
/ 20 x < VOC NO ≤
VOC/ NOx
/ 20 x VOC NO >
/ 20 x VOC NO >
x ( see Figure 25).
Ambient ratio of VOC to NOx
0
2
4
6
8
10
12
14
16
18
20
20040703
20040708
20040712
20040716
20040727
20040731
20040804
20040808
20040812
20040816
20040902
20040906
20040910
20040914
20040918
20040922
20040926
20041001
20041005
20041009
20041014
20041018
20041022
20041026
Ambient VOC/ NOx
Figure 29. Ambient ratio of VOC to NOx at station 9 ( 6: 00- 9: 00am). ( Note that not all calendar days are
present in the available data, due to lack of good- quality measurements of VOC and ozone. This study uses
data for 93 days in the summer, i. e., most of the period July 3 - October 26, 2004, the season during which
ozone pollution mainly occurs.)
69
The comparison of the observed and predicted peak ozone concentrations at receptor
station 5 are shown in Figure 30. Even though the prevailing wind commonly blows from
southwest to northeast in the summer, due to the wind speed/ direction variability the
geographical location where peak ozone concentrations occur may differ from receptor
station 5 ( as can be observed from photochemical grid model simulations), and thus the
predicted concentration at station 5 is simply an approximation of the observed or
theoretical concentration derived from smog chamber experiments or chemical reaction
mechanisms. That in part explains why R2 is not higher in the regression; that is, about
65% of the variation in peak ozone concentrations can be accounted for by the regression
variables selected, but the regression cannot account for the day- specific variation of
spatial transport of the predicted emissions to receptor 9 by the ISCST3 model or of
subsequent ozone formed by those emissions to receptor 5.
0
20
40
60
80
100
120
140
20040703
20040708
20040712
20040716
20040727
20040731
20040804
20040808
20040812
20040816
20040902
20040906
20040910
20040914
20040918
20040922
20040926
20041001
20041005
20041009
20041014
20041018
20041022
20041026
Ambient conc. ( ppb)
Observed O3 ( ppb)
Predicted O3 ( ppb)
Figure 30. Comparison of the observed and predicted peak ozone concentrations
70
3.3.4. Applying the regression model
Three hydrogen supply pathways are considered, and for each of them there are two sets
of market penetrations: 9% and 20%. Therefore, there are a total of six scenarios. Based
on a previous study ( see Chapter 2), the changes in ambient concentrations of primary
pollutants, including VOC and NOx, at monitoring stations have been determined. Below
are the steps to estimate the changes in ozone air quality due to lifecycle emissions of
each hydrogen pathway. We should sequentially:
( 1) Estimate the incremental VOC and NOx concentrations ( i. e., the 3- hour average of
6: 00- 9: 00am or the daily average, at station 9), caused by atmospheric physical transport
and associated with each of six hydrogen supply scenarios. This step is accomplished
using atmospheric dispersion models along with typical meteorological year data. No
atmospheric chemical transformation is considered and only directly emitted primary
pollutants are investigated for this step; therefore, secondary ozone pollution has not been
included so far in the analysis. The percentage change in VOC or NOx is expressed as
Equations 4 and 5.
% VOC newVOC baselineVOC 100% incremental VOC 100%
baselineVOC baselineVOC
−
Δ = × = × ,
( Equation 4)
71
% x x 100% x 100%
x
x x
NO new NO baseline NO incremental NO
baseline NO baseline NO
−
Δ = × = × ,
( Equation 5)
( 2) Add the incremental VOC and NOx to the baseline VOC and NOx concentrations
( i. e., the current ambient background VOC and NOx in 2004), respectively, and use the
sum as inputs to the ozone- and- precursors regression model developed in this study ( see
Equation 3);
( 3) Calculate the new peak ozone concentrations day by day using the same
meteorological data as in the regression model, as we treat meteorological factors as
control variables. The calculated results are for station 5 ( see Figures 7 and 8); and
( 4) Compute the difference between the new ozone and the previously predicted ozone
that is estimated using ambient background VOC and NOx as inputs to the regression
model. Now, the changes in ozone air quality, denoted by ( max) 3 ΔO , and the percentage
changes in peak ozone levels, denoted by % ( max) 3 ΔO , associated with a hydrogen
pathway can be determined. Below are the formulas ( see Equations 6 and 7).
3 3 3 ΔO ( max)= newO ( max)− baseline O ( max) , ( Equation 6)
3 3
3
3
% ( max) ( max) ( max) 100%
( max)
O new O baseline O
baseline O
−
Δ = × , ( Equation 7)
72
Note that the changes in initial VOC and NOx are small relative to the baseline pollution
level, so it makes sense to apply the regression model to these “ new” input data since
they remain within the range of observations in the region.
3.4. Results and discussion
3.4.1. Incremental 3- hour average pollution of ozone precursors
The incremental 3- hour average concentrations of VOC and NOx at receptor station 9 are
shown in Tables 11 and 12. Those numbers represent additional pollution, caused by
lifecycle emissions from six hypothetical hydrogen pathways respectively and occurring
at station 9, which subsequently results in ozone pollution at station 5 commonly
downwind of station 9. Relative to the actual ambient pollution level at station 9, Figures
31- 34 compare the incremental 3- hour average pollution of physically transported VOC
and NOx associated with each hydrogen pathway.
At the 9% market penetration, the onsite pathway causes additional VOC concentrations
of 0% to 0.007%, the pipeline pathway causes 0% to 0.014% ( there is one outlier of
0.041%), and both are much smaller than the truck pathway that causes additional VOC
concentrations of 0% to 0.027%. At the 20% market penetration, the onsite pathway
causes additional VOC concentrations of 0% to 0.058%, the pipeline pathway causes 0%
to 0.034% ( again, there is an outlier of 0.102%), and the truck pathway causes additional
VOC concentrations of 0% to 0.067%.
73
At the 9% market penetration, the onsite pathway causes additional NOx concentrations
of 0% to 0.140%, the pipeline pathway causes 0% to 0.381% ( one outlier is 0.795%), and
the truck pathway results in additional NOx concentrations of 0% to 0.750%. At the 20%
market penetration, the onsite pathway causes additional NOx concentrations of 0% to
0.443%, the pipeline pathway causes 0% to 0.952% ( the outlier is 1.987%), and the truck
pathway results in additional NOx concentrations of 0% to 1.861%.
In conclusion, compared to the background initial VOC and NOx ( the 3- hour averages,
6: 00- 9: 00am), the truck pathways have the greatest impact on both VOC and NOx
pollution, the onsite pathways have the smallest impact, and the pipeline pathways are
between them ( even though the pipeline pathways and the onsite pathways are almost
comparable in terms of the resulting additional VOC or NOx pollution). In particular, the
real- world NOx pollution at station 9 was often relatively low in 2004 ( see Figure 27),
and diesel hydrogen- delivery trucks emit substantial amounts of NOx, which explains
why the truck pathway at the 20% market penetration can lead to up to a 2% increase of
the current NOx pollution levels.
74
Table 11. Descriptive statistics for incremental 3- hour average VOC concentrations, ppbC
Scenario N Range Minimum Maximum Mean Std. deviation
onsite, 9% 93 0.00398 0.00000 0.00398 0.00066 0.00074
pipeline, 9% 93 0.00817 0.00000 0.00817 0.00080 0.00164
truck, 9% 93 0.01479 0.00000 0.01479 0.00206 0.00301
onsite, 20% 93 0.03617 0.00000 0.03617 0.00282 0.00487
pipeline, 20% 93 0.02043 0.00000 0.02043 0.00200 0.00410
truck, 20% 93 0.03689 0.00000 0.03689 0.00520 0.00748
Table 12. Descriptive statistics for incremental 3- hour average NOx concentrations, ppb
Scenario N Range Minimum Maximum Mean Std. deviation
onsite, 9% 93 0.02015 0.00000 0.02015 0.00299 0.00330
pipeline, 9% 93 0.06097 0.00000 0.06097 0.00516 0.01089
truck, 9% 93 0.07298 0.00000 0.07298 0.01063 0.01588
onsite, 20% 93 0.18763 0.00000 0.18763 0.01369 0.02490
pipeline, 20% 93 0.15242 0.00000 0.15242 0.01290 0.02721
truck, 20% 93 0.18194 0.00000 0.18194 0.02686 0.03932
75
9% market penetration
0.000%
0.005%
0.010%
0.015%
0.020%
0.025%
0.030%
0.035%
0.040%
0.045%
Period ( Jul 3 - Oct 26, 2004)
Percentage change in VOC conc.
onsite, 9%
pipeline, 9%
truck, 9%
Figure 31. Comparison of percentage changes in 3- hour average VOC concentrations ( 9% market
penetration)
20% market penetration
0.00%
0.02%
0.04%
0.06%
0.08%
0.10%
0.12%
Period ( Jul 3 - Oct 26, 2004)
Percentage change in VOC conc.
onsite, 20%
pipeline, 20%
truck, 20%
Figure 32. Comparison of percentage changes in 3- hour average VOC concentrations ( 20% market
penetration)
76
9% market penetration
0.0%
0.1%
0.2%
0.3%
0.4%
0.5%
0.6%
0.7%
0.8%
0.9%
Period ( Jul 3 - Oct 26, 2004)
Percentage change in NOx conc.
onsite, 9%
pipeline, 9%
truck, 9%
Figure 33. Comparison of percentage changes in 3- hour average NOx concentrations ( 9% market
penetration)
20% market penetration
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
Period ( Jul 3 - Oct 26, 2004)
Percentage change in NOx conc.
onsite, 20%
pipeline, 20%
truck, 20%
Figure 34. Comparison of percentage changes in 3- hour average NOx concentrations ( 20% market
penetration)
77
3.4.2. Changes in peak ozone concentrations, ΔO3 ( max)
Ozone formation in the atmosphere is a complicated issue. Table 13 summarizes the
estimated changes in 1- hour peak ozone concentrations due to lifecycle emissions of each
hydrogen pathway. The range of ( max) 3 ΔO results increases as the market penetrations
increase from 9% ( corresponding to 27 refueling stations) to 20% ( corresponding to 66
refueling stations). Given the same market penetration, truck pathways correspond to the
widest range of results, which means that truck pathways tend to result in
much more variation in degradation or improvement of ozone air quality.
( max) 3 ΔO
As shown in Table 13, all the minimum ( max) 3 ΔO results are negative and all the
maximum results are positive, which means that increases in initial VOC and
NO
( max) 3 ΔO
x concentrations do not necessarily increase the peak O3 concentration, and may even
result in a decrease. This is consistent with a standard EKMA ozone isopleth ( see Figure
25). Accounting for the meteorological conditions, this phenomenon depends on the
ambient ratio of initial VOC and NOx, i. e., relative abundance of initial VOC and NOx.
NOx is relatively abundant in the NOx- rich zone ( or so- called " VOC- limited" zone) on a
typical ozone isopleth diagram, which corresponds to the situations where / x VOC NO is
generally less than 7- 10 ( Chang et al., 1989; NRC, 1991). However, VOC is relatively
abundant in the VOC- rich zone ( or so- called " NOx- limited" zone) on an ozone isopleth
diagram, which corresponds to the situations where / x VOC NO is generally greater than
7- 10 ( Chang et al., 1989; NRC, 1991).
78
Table 13. Descriptive statistics for changes in peak ozone concentrations, ( max) , ppb 3 ΔO
Scenario N Range Minimum Maximum Mean Median Std. deviation
onsite, 9% 93 0.00512 - 0.00237 0.00275 0.00031 0.00019 0.00083
pipeline, 9% 93 0.01761 - 0.00574 0.01187 0.00081 0.00000 0.00213
truck, 9% 93 0.02684 - 0.00648 0.02036 0.00155 0.00018 0.00371
onsite, 20% 93 0.03351 - 0.02523 0.00828 0.00016 0.00084 0.00472
pipeline, 20% 93 0.04401 - 0.01437 0.02964 0.00201 0.00000 0.00530
truck, 20% 93 0.06702 - 0.01641 0.05061 0.00377 0.00046 0.00926
3.4.3. Percentage changes in peak ozone concentrations, % ( max) 3 ΔO
Figure 35 presents the comparison of % ( max) 3 ΔO results associated with these three
hydrogen supply pathways. For the 9% market penetration scenario, the onsite pathway
causes within the range of - 0.007% to 0.008%, and the pipeline pathway
causes within the range of - 0.008% to 0.021%. The truck pathway causes
within the range of - 0.009% to 0.039%, which is a wider range than for the
onsite and pipeline pathways.
% ( max) 3 ΔO
% ( max) 3 ΔO
% ( max) 3 ΔO
For the 20% market penetration scenario, the onsite pathway causes within
the range of - 0.075% to 0.022%, and the pipeline pathway causes within the
range of - 0.020% to 0.052%, shown in Figure 36. Moreover, Figure 36 also demonstrates
that the truck pathway causes
% ( max) 3 ΔO
% ( max) 3 ΔO
% ( max) 3 ΔO within the range of - 0.023% to 0.100%, which
is also wider than the range for the onsite and pipeline pathways.
79
During the modeling period ( July 3, 2004 through October 26, 2004), two obviously
different ozone pollution trends appear. One occurs in the days preceding September 15,
2004, when the incremental VOC and NOx from hydrogen pathways typically lead to an
increase in the ozone level, and the resulting ozone pollution goes up in the following
order: onsite pathway < pipeline pathway < truck pathway. That is, during the worst
summer months for ozone pollution, July, August, and September, the changes in
are almost all positive, therefore a worsening of ozone air quality would
occur. The other trend occurs after September 15 ( mostly in October), and all pathways
often lead to no ozone pollution or even a decrease in ozone formation; especially onsite
and truck pathways lead to greater decreases in ozone pollution.
% ( max) 3 ΔO
At the 20% market penetration, the greatest increase can be up to 0.1% of the current
ozone level ( corresponding to the truck pathway), and the greatest decrease can be around
- 0.1% of background ozone pollution ( corresponding to the onsite pathway). In summary,
all the pathways result in very small changes in ozone air quality. However, Figures 35
and 36 both show that the truck pathway causes % ( max) 3 ΔO to fluctuate much more than
do the other two hydrogen pathways ( especially in July, August, and September).
Therefore, just in terms of ozone pollution, all the three hydrogen pathways in some cases
would result in a better ozone air quality, corresponding to a negative , and
in some cases will result in a worse ozone air quality, corresponding to a positive
, but there is little doubt that the truck pathway tends to lead to a much wider
fluctuation in degradation or improvement of ozone air quality.
% ( max) 3 ΔO
% ( max) 3 ΔO
80
9% market penetration
- 0.010%
- 0.005%
0.000%
0.005%
0.010%
0.015%
0.020%
0.025%
0.030%
0.035%
0.040%
20040703
20040714
20040731
20040810
20040902
20040912
20040922
20041003
20041014
20041024
Period ( Jul 3 - Oct 26, 2004)
Percentage change in O3 conc.
onsite, 9%
pipeline, 9%
truck, 9%
Figure 35. Comparison of percentage changes in peak ozone concentrations ( 9% market penetration)
20% market penetration
- 0.08%
- 0.06%
- 0.04%
- 0.02%
0.00%
0.02%
0.04%
0.06%
0.08%
0.10%
20040703
20040714
20040731
20040810
20040902
20040912
20040922
20041003
20041014
20041024
Period ( Jul 3 - Oct 26, 2004) Percentage change in O3 conc.
onsite, 20%
pipeline, 20%
truck, 20%
Figure 36. Comparison of percentage changes in peak ozone concentrations ( 20% market penetration)
81
3.4.4. Further discussion on ozone pollution
The ozone pollution caused by the truck pathways fluctuate most widely; percentage
changes in peak ozone concentrations are approximately - 0.01% to 0.04% for the 9%
market penetration scenario, and approximately - 0.03% to 0.1% for the 20% market
penetration scenario. Note that the 20% onsite pathway can cause a decrease around -
0.1% of background ozone pollution. The federal ozone standard is 80 ppb within the 8-
hour averaging time ( NAAQS, 2006), and the California ozone standard is 70 ppb within
the 8- hour averaging time and 90 ppb within the 1- hour averaging time ( CARB, 2006).
Figures 27 and 30 show that the California 1- hour ozone standard is violated on some
days during the modeling period, but the ambient peak ozone is often in the vicinity of
the standard. Therefore, the truck pathways ( and even more so the onsite and pipeline
pathways) are unlikely to lead to a serious ozone problem in Sacramento.
Since the same meteorological conditions for each day are used when deriving the
regression model and applying the model to the analysis of hydrogen supply scenarios,
the changes in peak ozone concentrations shown in Figures 35 and 36 are due only to the
variation in VOC and NOx inputs to the model. Most of the largest positive peaks in
% ( max) in the summer ( before September 15, 2004) correspond to the lowest 3 ΔO
/ x VOC NO ratios on those days, about 3.2- 5.2 ( see Figures 29, 35 and 36). Because the
estimate of the ratios of the incremental VOC and the incremental NOx due to a hydrogen
supply pathway is about 0.1- 0.3, the new / x VOC NO ratios inputted to the model would
82
decrease. That is, the / x VOC NO ratio goes down and the peak ozone concentration goes
up accordingly. In other words, VOC, NOx, and peak O3 concentrations are all positive
increases. Therefore, ozone formation is generally in the NOx limited regime ( limited by
NOx) in the summer in that part of Sacramento. On the contrary, most of the largest
negative peaks in in the fall ( after September 15, 2004) correspond to the
lowest
% ( max) 3 ΔO
/ x VOC NO ratios, approximately 1.1- 1.5 ( see Figures 29, 35 and 36). Similarly,
the new / x VOC NO ratios input to the model would decrease. That is, the / x VOC NO
ratio goes down and the resulting peak ozone concentration also goes down. Put another
way, both VOC and NOx increase, but peak O3 concentrations decrease. Therefore, ozone
formation mostly is limited by VOC in the fall. In summary, the ozone production ridge
line on the isopleth corresponds to a / x VOC NO ratio between 1.5 and 5.2 in Sacramento,
which is slightly lower than a typical EKMA value. That could be in part because the
Sacramento region is not a closed smog chamber, station 5 is not always where peak
ozone formation occurs due to the variability of wind speed and direction, and the
statistical model parameters are only estimates.
3.5. Conclusions
In this chapter, we assumed two sets of hydrogen vehicle market penetrations of 9% and
20%, respectively, and considered the following three hypothetical natural gas- to-hydrogen
pathways: onsite hydrogen production, central hydrogen production with
gaseous hydrogen pipeline delivery, and central hydrogen production with liquid
hydrogen truck delivery. Prior to estimating changes in ozone air quality due to each
83
hydrogen pathway, lifecycle emission inventories and optimized spatial layouts of
hydrogen infrastructure were determined.
Atmospheric ozone formation is complicated. In this research, a region- specific linear
regression model was developed to link the peak ozone concentrations to ambient
meteorological conditions and the early morning ambient VOC and NOx as the precursors
to ozone formation. The regression model and data were limited to the Sacramento region
and the time period from July 3, 2004 to October 26, 2004. The model shows that
increases in precursor concentrations do not necessarily increase the peak ozone
concentration, and may even cause it to decrease. The results indicate that, in Sacramento,
ozone formation is generally limited by NOx in the summer and is mostly limited by
VOC in the fall. The ozone production ridge line on the isopleth corresponds to a
/ x VOC NO ratio between 1.5 and 5.2 in Sacramento, which is slightly lower than a
typical value observed in a closed smog chamber or empirical kinetic modeling approach
( EKMA) diagrams. The ozone monitoring station used in the regression analysis is also
not always an accurate indicator of peak ozone formation due to the variability of wind
speed and direction.
Compared to the background initial VOC and NOx ( the 3- hour averages, 6: 00- 9: 00am),
truck pathways have the greatest impact on both VOC and NOx pollution, the onsite
pathways have the smallest impact, and the pipeline pathways are between them. Since
the current light duty fleet is held constant and additional hydrogen cars are added to the
fleet, the incremental VOC and NOx pollution resulting from lifecycle emissions of all
84
hydrogen pathways is a positive quantity. At the 9% market penetration, the truck
pathway caused additional VOC ( or NOx) up to around 0.05% ( or 1%) of current
pollution level in 2004. At the 20% market penetration, the truck pathway caused
additional VOC ( or NOx) up to around 0.1% ( or 2%) of the current pollution level.
All the hydrogen pathways would result in very small ( either negative or positive)
changes in ozone air quality. In some cases worse ozone air quality ( mostly in July,
August, and September) resulted and ozone increments increased in the following order:
onsite pathway < pipeline pathway < truck pathway. In some cases better ozone air
quality was predicted to result ( mostly in October), and the truck and onsite pathways had
a greater impact than the pipeline pathway. The truck pathway tended to lead to a much
wider fluctuation in degradation or improvement of ozone air quality: percentage changes
in peak ozone concentrations are approximately - 0.01% to 0.04% for the 9% market
penetration scenario, and approximately - 0.03% to 0.1% for the 20% market penetration
scenario. Note that the 20% onsite pathway occasionally resulted in a decrease of around
- 0.1% of background ozone pollution. So the positive and negative limits of changes in
ozone pollution would be around one thousandth of current pollution levels. Compared to
the current ambient pollution level, the truck pathways ( and therefore the onsite and
pipeline pathways) are unlikely to cause a serious ozone problem for market penetration
levels of hydrogen fuel cell vehicles in the 9- 20% range.
The quantified ozone concentrations can be used to estimate agricultural losses and
human health damages. Based on the predicted changes in ozone pollution ( and the other
85
criteria pollutants), dose- response functions, and demographic data in Sacramento, social
costs associated with hydrogen supply pathways can be estimated. This is useful for
urban planners and policy makers.
86
4. Investigating Contributions of Gasoline Pathways to Urban
Air Pollution Using Travel Demand Model Data
4.1. Introduction
The current petroleum- fueled transportation system emits significant amounts of criteria
pollutants and, as a result, accounts for a major fraction of urban air pollution in the U. S.
For example, on- road motor vehicles contribute 30.6- 38.5% of volatile organic
compounds ( VOCs), 34.3- 62.2% of nitrogen oxides ( NOx), and 4.4- 5.7% of particulates
( PM10) to annual ambient concentrations in Sacramento, California for 2005 ( Wang et al.,
2007). Due to close proximity to people and their daily lives, current mobile emission
sources are likely to cause human health damage and, thus, result in a significant social
cost ( ExternE, 1998; McCubbin and Delucchi, 1996; Delucchi and McCubbin, 2004).
For simplicity, in this study we use the term “ gasoline pathway” to refer to the petroleum-based
fuel pathway, including both gasoline and diesel transportation fuels. Compared to
hydrogen pathways which emit all criteria pollutants upstream of vehicle operation,
downstream vehicle operation plays an important role for a gasoline or diesel pathway.
Therefore, our focus is on estimating contributions of gasoline vehicle operations to
urban air pollution, although we also consider the other gasoline pathway steps like
gasoline- delivery truck emissions.
87
As gasoline vehicle technology is evolving, we consider various types of current and
advanced gasoline vehicles. For gasoline pathways examined in the chapter, the 2005
light duty ( LD) fleet is used as the reference, which corresponds to current transportation
technology. To reflect the improvements in vehicle/ fuel technologies and standards over
time, we use the predicted 2025 light duty fleet composition as representative of
advanced or evolved gasoline vehicles in the near future.
In this dissertation, the overall goal is to compare hydrogen to gasoline or diesel in terms
of the resulting impacts on urban air quality ( see next chapter). To do so, in this chapter,
we are addressing the following two prerequisite research questions.
( 1) What would be the impacts of gasoline fleet operations ( and gasoline pathways)
on urban air quality, using traditional 4- step travel demand data and grid- level
emission inventories? and
( 2) How do current and advanced gasoline vehicle pathways compare, in terms of the
resulting impacts on urban air quality, from a lifecycle analysis perspective?
Ambient concentrations of pollutants are correlated with emissions, but the contribution
to ambient air quality of on- road mobile sources is not necessarily equal to their
contribution to regional emissions. This is true for several reasons such as the distribution
of other pollution sources and regional topology, as well as meteorology. The complexity
of spatial and temporal distributions of vehicle emissions/ activities and the mobility of
88
vehicles make it very hard to quantify the proportions of ambient air pollutant
concentrations attributable to on- road mobile sources. To obtain specific results, it is
useful to base the analysis on a particular geographic area, and this study chooses the
Sacramento metropolitan area as the setting. Using the dataset of a travel demand model,
regional vehicle emissions are estimated and disaggregated into hourly, gridded
inventories with a 1×1 km resolution. Transportation- related concentrations of primary
pollutants are then predicted using a Gaussian dispersion model. Finally, concentration
contributions of light duty vehicles to urban air pollution are estimated on a regional scale.
In summary, in this chapter we investigate contributions of various current and advanced
gasoline/ diesel pathways to urban air pollution using travel demand model data. We
examine four gasoline pathway scenarios, and the ground- level concentrations for four
primary pollutants, i. e., carbon monoxide ( CO), NOx, VOC, and PM10, are estimated in
order to compare air quality impacts between gasoline pathway scenar