Benefit-Cost Assessment of Aviation Environmental Policies by Christopher K. Gilmore B.S.E., Mechanical Engineering, Duke University, 2010 Submitted to the Department of Aeronautics and Astronautics in partial fulfilln ent of the requirements for the degree of ARCHNES Master of Science in Aeronautics and Astronautics MA SSACHUSETTS INsTTrfE O- TECH NLOGY at the JLj 10 2i12 MASSCHUSETTS INSTITUTE OF TECHNOLOGY uBARE June 2012 @ Massachusetts Institute of Technology 2012. All rights reserved. I t ....................... Signature of Author: ............................................... Department of Aeronautics and Astronautics May 24, 2012 0/7 Certified by: ............................. A / ....................... ( Steven R.H. Barrett Charles Stark Draper Assistant Professor of Aeronautics and Astronautics Thesis Supervisor Accepted by: .................................................... I 7 A E a. 7 ..... Modiano Eytan H. Professor of Aeronautics and Astronautics Chair, Committee on Graduate Students Benefit-Cost Assessment of Aviation Environmental Policies by Christopher K. Gilmore Submitted to the Department of Aeronautics and Astronautics on May 24, 2012 in Partial Fulfillment of the Requirements for the Degree of Master of Science in Aeronautics and Astronautics ABSTRACT This thesis aids in the development of a framework in which to conduct global benefit-cost assessments of aviation policies. Current policy analysis tools, such as the aviation environmental portfolio management tool (APMT), only consider climate and air quality impacts derived from aircraft emissions within the US. In addition, only landing and takeoff (LTO) emissions are considered. Barrett et al., however, has shown that aircraft cruise emissions have a significant impact on ground-level air quality. Given the time-scale and atmospheric lifetimes of species derived from aircraft emissions at these higher altitudes, a global framework for assessment is required. This thesis specifically investigates the global as well as regional implementation of an ultra-low sulfur jet fuel (ULSJ). The expected result from this policy is a reduction in aircraft SOx emissions, which in turn would reduce the atmospheric burden of primary and secondary sulfate aerosols. Sulfate aerosols have both climate and air quality impacts as they reflect incoming solar radiation (and thus provide atmospheric cooling) and are a type of ground-level pollutant that have generally been correlated to premature mortalities resulting from cardiopulmonary disease and lung cancer. Benefit-cost techniques are applied in this analysis. The framework developed within this thesis includes the ability to calculate expected avoided premature mortalities outside of the US. In addition, a monetization approach is used in which different values of statistical lives (VSLs) are applied depending on the country in which a premature mortality occurs. Also, the economic impact of increased fuel processing to reduce the FSC is estimated. This analysis is performed using Monte Carlo techniques to capture uncertainty, and a global sensitivity analysis (GSA) is utilized to determine the primary sources of uncertainty. The benefit-cost analysis results show that for US and global implementation, there is -80% chance of ULSJ implementation having a not cost beneficial outcome when climate, air quality, and economic impacts are included. On average, however, the air quality benefits do exceed the climate disbenefits. In addition, the GSA reveals that the largest contributor to the uncertainty in this analysis is the assumed US VSL distribution, where approximately 60% of the variance in the final output distribution can be attributed to this uncertainty. In addition, a fast policy tool approach is investigated using sensitivity values calculated from an adjoint model built-in to the global chemical transport model (GCTM) used for the atmospheric modeling within this analysis. From this fast policy tool, first order estimates of the impact of ULSJ on premature mortality are calculated. Thesis Supervisor: Steven R.H. Barrett Title: Charles Stark Draper Assistant Professor of Aeronautics and Astronautics i Acknowledgements First I would like to thank my advisor, Professor Steven Barrett, for his guidance throughout my first two years at MIT. He has helped me develop into a capable researcher, a better writer, and a more critical thinker, all of which I greatly appreciate. I would also like to thank Dr. Steve Yim for his patience with me and tolerating all of my questions when I first got to PARTNER. His help day in and day out has been invaluable. I also want to thank Dr. Jim Hileman for all of his help. His door was always open, and he was always there to provide his insights. In addition, I want to thank Dr. Christopher Wollersheim and Dr. Robert Malina for their wisdom with regards to all things related to economics. Their help and guidance with the ULSJ project made my job easier. I also would like to thank Dr. Jon Levy for his help with the public health analysis and everyone at the FAA, including Chris Sequeira and Daniel Jacob, for their advice and help with regards to the ULSJ project. I also want to thank all the students in PARTNER for helping to make my experience at MIT so far a great one. In particular, I want to thank Akshay and Jamin for always being willing to help me with whatever problem I found too daunting to take on by myself, thanks to Kristy, Naki, (Lt.) Nick, Tony, and Gideon for putting up with my sense of humor, Phil for answering all of my climate code related questions, and finally thanks to Fabio, Seb, and Sergio for always lightening the mood. I also would like to thank Dr. Jon Protz, my advisor from Duke University for helping me get this far and convincing me that grad school was the right choice for me. Finally, and most importantly, I would like to thank my parents, Mike and Ai-Chi, for always being there and providing encouragement and advice whenever I needed it. I would not be here without them. ii Contents Chapter 1: Introduction............................................................................................................... 1 1.1. Aircraft and the Environment ........................................................................................ 1 1.2. Policy Objectives ....................................................................................................... 4 1 .3 . Mo tiva tio n ........................................................................................................................ 5 1.4. Thesis Structure .......................................................................................................... 6 Chapter 2: Background .......................................................................................................... 8 2.1. Atmospheric Modeling ................................................................................................. 8 2.1.1. Global and Nested GEOS-Chem Model Descriptions............................................ 2.2. Impact of Aerosols on Climate ..................................................................................... 8 9 2.3. Air Quality and Mortality............................................................................................. 11 2.4. Elements of EBCA...................................................................................................... 13 2.4.1. Benefit/Cost Analysis .......................................................................................... 14 2.4.2. Discounting .............................................................................................................. 15 2.4.3. Premature Mortality............................................................................................... 16 2.4.4. Valuing Lives........................................................................................................ 16 2.4.5. Cost Analysis ........................................................................................................ 18 2.5. Ultra-Low Sulfur Fuels ................................................................................................. 18 2.5.1. Ultra-Low Sulfur Diesel Case Study ..................................................................... 18 2.5.2. QinetiQ Report on Jet Fuel Sulfur Limit Reduction .............................................. 20 2.5.3. Energy Information Administration (EIA) Report on Market Effects Due to ULSD.....21 iii 2.5.4. Other Transportation Sectors .............................................................................. 21 2.6. Role of Sensitivity Analysis ........................................................................................ 21 Chapter 3: EBCA Methodology Development ....................................................................... 26 3.1. Em issions Scenarios ................................................................................................. 26 3.2. Determ ining Climate Impacts...................................................................................... 27 3.2.1. Sulfate RF Calculation...........................................................................................28 3.2.2. RF Uncertainty ...................................................................................................... 30 3.2.3. Sulfate Lum ping ................................................................................................... 31 3.2.4. W TW GHG Emissions.............................................................................................. 32 3.2.5. APMT-Impacts Climate Module............................................................................ 33 3.3. Country Dependent VSLs .......................................................................................... 34 3.4. Concentration Response Functions ............................................................................ 36 3.5. Econom ic Analysis ..................................................................................................... 39 3.5.1. Price History Analysis........................................................................................... 40 3.5.2. Cost Buildup Approach........................................................................................ 42 3.5.3. Cost Distribution....................................................................................................46 3.6. Sensitivity Analysis ................................................................................................... 46 3.6.1. Monte Carlo Analysis Framework........................................................................ 46 3.6.2. Nom inal Range Sensitivity Analysis ..................................................................... 47 3.6.3. Global Sensitivity Analysis ................................................................................... 47 3.7. Additional Operational Concerns ................................................................................. 48 iv 3.7.1. Change in Fuel Properties.................................................................................... 48 3.7.2. Fuel Lubricity.........................................................................................................49 Chapter 4: EBCA Results...................................................................................................... 52 4.1. Aerosol RF Results...................................................................................................... 52 4.2. Mortality Results by Country...................................................................................... 53 4.3. VSL Results by Country............................................................................................. 55 4.4. Global and US ULSJ O utcom es ................................................................................ 56 4.4.1. Assum ptions for Global and US Im plem entation Analysis .................................... 56 4.4.2. Assum ed Uncertainty Distributions...................................................................... 58 4.4.3. Global Im plem entation Analysis Results .............................................................. 61 4.4.4. US Implem entation Analysis................................................................................. 62 4.4.5. US-Only Im plem entation Analysis ........................................................................ 64 4.4.6. Constant VSL Analysis ........................................................................................ 65 4.4.7. Cost Effectiveness Analysis ................................................................................. 66 4.5. Policy Im plications ...................................................................................................... 66 4.6. Nom inal Range Sensitivity Results ............................................................................ 68 4.6.1. Discount Rate ..................................................................................................... 70 4.7. Global Sensitivity Analysis (GSA) Results ................................................................... 71 Chapter 5: Fast Policy Analysis............................................................................................. 75 5.1. Adjoint Model and Policy Tool.................................................................................... 75 5.2. USLJ Analysis Com parison ........................................................................................ 76 V Chapter 6: Conclusions and Future W ork............................................................................. 82 6.1. Global ULSJ Implementation..........................................................................................82 6.2. Fast Policy Analysis................................................................................................... 83 6.3. Limitations ...................................................................................................................... 83 6.4. Future W ork....................................................................................................................84 Appendix A: Tables...................................................................................................................87 W o rks Cited .............................................................................................................................. vi 97 List of Figures Figure 1: Radiative forcing estimates of aircraft emissions..................................................... 2 Figure 2: Multi-branch hysteresis behavior of an ammonium sulfate aerosol particle. Plot from 0 W ang et a l.................................................................................................................................1 Figure 3: The product supplied of ULS, LS, and HS diesel fuel plotted simultaneously with the price differential for ULS-HS and LS-HS for Jan 2001 to February 2011 .............................. 41 Figure 4: NG Price history required for ULSJ hydroprocessing where prices are presented based on 2.37 scf/gal (0.018 scm/L) of NG per gallon of ULSJ. ............................................ 44 Figure 5: Capital costs for hydrotreating and SMR units as a function of HDS capacity depreciated over 30 years................................................................................................... 45 Figure 6: Benefit-cost distribution for global implementation analysis for three different discount ra tes (DRs ). .............................................................................................................................. 61 Figure 7: Benefit-cost distribution for US implementation analysis. ....................................... 63 Figure 8: Benefit-cost distribution for US-only implementation analysis................................ 64 Figure 9: Benefit-cost distribution for a constant US VSL analysis. ....................................... 65 Figure 10: NRSA results for global implementation of ULSJ................................................ 68 Figure 11: NRSA results for US implementation of ULSJ. ................................................... 69 Figure 12: Net benefit-cost plotted against discount rate of the deterministic model used in the NR SA ........................................................................................................................................ 70 Figure 13: Global Implementation GSA main effect sensitivity index results......................... 72 Figure 14: Global Implementation GSA total effect sensitivity index results........................... 72 Figure 15: US Implementation GSA main effect sensitivity index results ............................... 73 Figure 16: US Implementation GSA total effect sensitivity index results ................................ 73 vii Figure 17: Using sensitivities to compute the total impact from an aircraft emissions policy s ce n a rio .'...................................................................................................................................7 Figure 18: Forward model premature mortality results from the ULSJ EBCA. ....................... 6 77 Figure 19: Adjoint policy tool results and adjusted results for global ULSJ implementation. ...... 78 Figure 20: Adjoint policy tool results for monetized avoided premature mortalities with an assumed income elasticity of 1. ............................................................................................ 80 Figure 21: Proposed structure of multi-year study adjoint policy tool.................................... 85 viii List of Tables Table 1: Assumed properties of "sulfates" optical bin in GEOS-Chem.................................. 11 Table 2: On-road implementation timeline of ULSD in the US ............................................ 19 Table 3: Off-road implementation timeline of ULSD in the US.' ............................................ 20 Table 4: Source-receptor mortality impact from full-flight aircraft emissions. ......................... 24 Table 5: Emission indices methodology for air quality simulations, where bolded variables are from AE DT . ............................................................................................................................... 27 Table 6: Coefficients in Eq. (11) and associated values and uncertainties. ........................... 30 Table 7: Percentage increase in avoided mortalities given a 1 pg/m 3 increase in ground-level PM2 5 concentration. Values from Pope et al. and Laden et al. .............................................. 38 Table 8: Process energy shares required for jet fuel production........................................... 44 Table 9: Aviation sulfate RF by component and region. ........................................................ 52 Table 10: Avoided mortalities by country due to global ULSJ implementation. ..................... 54 Table 11: Regional simulations avoided mortalities results for the US from global im plem entation of ULSJ ..................................................................................................... . . 54 Table 12: VSL and valuation of avoided premature mortalities (when cruise emissions are included) due to ULSJ implementation by country in US 2006 $. .......................................... 56 Table 13: Brief description of each input parameter. ............................................................. 59 Table 14: Monte Carlo Input Values and Distributions (Triangular: [Low, High, Nominal]).........60 Table 15: Global implementation EBCA results, given in 2006 US $ Billion.......................... 61 Table 16: Global implementation results from EBCA where no implementation cost has been included, given in 2006 US $ Billion. .................................................................................... ix 62 Table 17: Global implementation results from EBCA where no climate cost has been included, given in 2006 US $ Billion. ................................................................................................... Table 18: US EBCA results for global implementation, given in 2006 US $Billion. ................ . 62 63 Table 19: Valuation of US health impacts due to global implementation from CMAQ and GEOSC hem Nested sim ulations. .................................................................................................... 63 Table 20: US-only implementation EBCA results, given in 2006 US $ Billion. ...................... 64 Table 21: Constant US VSL EBCA results, given in 2006 US $ Billion. ................................. 65 Table 22: Cost effectiveness analysis results, given in 2006 US $ Million. ........................... 66 Table 23: Aircraft SO x Em issions.......................................................................................... 77 x THIS PAGE INTENTIONALLY LEFT BLANK. xi Chapter 1: Introduction While aircraft emissions represent only -1% of total fossil fuel use in the world,' aircraft operations are expected to increase by -5% annually.2 If this growth is realized, current aviation operations will double by 2025 and represent the fastest growing mode of transportation. As with any transportation sector, the environmental impact of aircraft emissions is of growing concern. The prioritization of improving air quality is reflected historically in legislation passed within the United States (US), such as the Clean Air Acts of 1970 and 1977,5 which motivated the restrictions placed on allowable pollutant concentrations. With regards to the transportation sector, the US has also required the use of desulfurized diesel fuel for on and off highway purposes to reduce ground-level pollutant concentrations. 6 Similar regulations also exist in many European nations. Specifically regarding aviation, the International Civil Aviation Organization's (ICAO) has routinely recommended increased aviation NOx stringencies to reduce negative health impacts derived from aircraft operations through an improvement in air quality.7 Aircraft emissions have both climate impacts, through their impact on the atmospheric radiative balance, and human health impacts, due to increased ground-level pollution concentrations by way of the chemical reaction of emissions with the ambient atmosphere and vertical transport. Policymakers have placed increased emphasis on mitigating the environmental impacts of aviation where the Federal Aviation Administration (FAA) has defined a set of prospective policy 8 goals in their Destination 2025 plan. This thesis aims to provide a framework in which to evaluate potential aviation policies within the context of a global environmental benefit-cost analysis (EBCA) with an emphasis on quantifying and monetizing air quality and climate impacts in order to achieve these policy goals. 1.1. Aircraft and the Environment Aircraft emissions are comprised mostly of carbon dioxide (CO2) and water vapor (H20), each making up 71 % and 28% of emissions, respectively. A small, but significant, portion of these emissions are comprised of nitrogen oxides (NOx), sulfur oxides (SOx), hydrocarbons (HCs), carbon monoxide (CO), and primary PM2.5. PM 2 .5 refers to particulate matter with a diameter less than or equal to 2.5 pm, and primary PM 2 .5 is particulate matter that is directly emitted from jet engines and most notably consists of black carbon (BC), or soot.9 With regards to climate, 1 aircraft emissions on the whole have a net warming effect on the atmosphere as estimated based on the most current understanding of how aircraft perturb the state of the atmosphere. Figure 1 shows the estimated radiative forcing (RF) contribution of these emissions as well as species (e.g. aerosols) and phenomena (e.g. contrail formation) derived from aircraft emissions with uncertainties. I 2 Spatial ) m) (W m (W sale RF Terms Global Carbon dioxide NOx 00 Total NOx Water vapour I iI fl Linar ontail I -0.0048 Local to Low 0.0034 (0.002S) Local to global (0.010) Local to continental 1.. -0.04 . OW LO Local to Very hemispheric Low 1 I global (IPCC AR4 values) 1 Total aviation (Excl. induced cirrus) -0.08 Low L (-0.0035) I0.0118 Induced cirrus cloudiness Total aviation (Incl. Induced cirrus) *lBest estimate jEstimate S Global Hemispheric to global I(0.0020) -90% confidence Soot aerosol Linear contrails High 0.0263 Continental (0.219) hemiheric -0.0125 Global (-0.0104) Ozone production Methane reduction Sulphate aerosol LO SU .J........... . 0.055 (0.0478) Global Low 0.078 Global Low ,1 0.08 0.04 0 Radiative Forcing (W M-2) 0.12 Figure 1: Radiative forcing estimates of aircraft emissions. 10 RF is a metric used to quantify the net perturbation to the atmospheric radiative balance from a particular atmospheric species. It is typically taken at its top-of-the-atmosphere (TOA) value, i.e. the net incoming minus the net outgoing radiative flux at the border between the atmosphere and space, although measures at other levels of the atmosphere are also possible. RF is not the only way by which to measure the climate impact of atmospheric species, where global warming potential and integrated temperature change are also common metrics." For the purposes of this analysis, however, RF provides the most convenient impact measure. 2 Greenhouse gases (GHGs), such as carbon dioxide (CO2), ozone (03), methane (CH4), and water vapor, warm the atmosphere. As can be seen from Figure 1, direct emissions of C02 and water vapor both lead to warming as aircraft directly increase the concentration levels of these species, although the net RF impact (where red/positive denotes warming) of C02 is an order of magnitude larger than water vapor. The formation of ozone also leads to a positive RF impact by way of the NOx cycle: NO +034 N0 2 +0 2 NO2 + hv + NO + 0 O + 0 2 + M+ 0 3 + M NOx emissions, which are primarily NO (except at low thrust), destroy ozone, but can also produce ozone given the production of NO2. In addition, NO and NO2 are cycled between one another and can result in ozone production and loss through the following reactions: HO 2 + NO H0 2 +0 3 -- OH + NO 2 -+OH+0 2 +0 2 OH + 03 4 HO 2 + 02 Thus, ozone formation is a function of not only NOx emissions, but also ambient concentrations of OH and HO2 , two very important atmospheric oxidizing agents. Ozone formation is also a strong function of altitude where ozone production efficiency (i.e. proportion of NOx that is ultimately converted to ozone) is higher at cruise altitudes than at ground-level. NOx molecules also interact with the methane cycle, where aircraft emissions lead to an atmospheric decrease in methane (thus an associated cooling), but an associated decadal loss of tropospheric ozone leads to further cooling.9 NOx emissions from aircraft that are closer to the ground can either create or destroy 03, and this behavior is a function of the ambient hydrocarbon (HC) concentration in a particular region. Aircraft emissions also lead to the formation of primary and secondary aerosols. As mentioned previously, primary BC is a direct emission from combustion and leads to a warming effect. Secondary sulfate aerosols (as well as ammonium nitrate aerosols) are formed from either direct SO 4 emissions or oxidized SO 2 emissions, where inorganic aerosols of this type refract 3 solar radiation back into space, thus providing a net cooling effect. Linear contrail formation in the wake of the aircraft provides an additional net warming effect. The most uncertain climatic impacts of aircraft emissions, however, are the induced cirrus cloudiness and soot-cirrus. Cirrus cloud formation is caused by the spreading and shearing of linear contrails as a result of increased particulate matter concentrations in the atmosphere, which provide nuclei from which clouds can grow. 9 The effect of this indirect climate impact can be either warming or cooling. Figure 1 shows it as a net warming effect, but with an associated large uncertainty. Primary and secondary particulate matter is also harmful to human health through ground-level population exposure. Exposure to PM, such as sulfate aerosols and BC, has been shown to lead to increased risk in cardiopulmonary disease (CP) as well as lung cancer (LC), and longterm PM exposure have been associated with overall increases in human premature mortality, where premature mortality comprised approximately 85% of all monetized health impacts from Although population exposure to ozone also has human health impacts, it is not considered in this thesis. In addition, morbidity impacts (hospital admissions, missed work PM 2.5 exposure. days, etc.) are also not considered. 1.2. Policy Objectives In their Destination 2025 document, the FAA has stated 6 distinct goals to improve sustainability 8 and reduce the overall impact of aircraft transportation operations by 2018. These goals are the following: - Reducing those exposed to aircraft noise to less than 300,000 people - Implementing an alternative fuel for current leaded general aviation fuel - Improve fuel efficiency by 2% annually - Reduce the health impacts of aircraft emissions by 50% relative to a 2005 baseline - Set aviation on a trajectory for carbon neutral growth relative to a 2005 baseline - Use at least one billion gallons of renewable jet fuel This thesis focuses primarily on the impact of aviation emissions on the environment and human health, thus topics concerning other aspects of sustainability or aircraft noise reduction will not 4 be discussed. The major focus of this thesis is, however, on global as well as regional implementation of an ultra-low sulfur jet fuel (ULSJ), where the emphasis is placed on reducing ground-level PM 2.5 concentrations with the ultimate goal of reducing the human health impact of aviation operations. ULSJ would serve as a drop-in and immediate replacement fuel where rulemaking would mandate a fuel sulfur content of less than 15 ppm by mass. 1.3. Motivation In order to determine the viability of a particular piece of environmental legislation or rulemaking, standard benefit/cost analysis (BCA) techniques are employed in which potential economic, climate, and health benefits, disbenefits, or societal costs are monetized and aggregated in order to determine a net benefit/cost outcome. The general BCA framework will be discussed in greater detail in Section 2.4. Analysis currently conducted within the Partnership for AiR Transportation Noise and Emissions Reduction (PARTNER), such as that performed for the NO) stringency analysis, is regionally focused on the US, and BCA has only historically considered landing/take-off (LTO) emissions. A suite of tools known as the aviation environmental portfolio management tool (APMT) has been developed by PARTNER that determine noise, air quality (LTO emissions, only), and climate impacts and monetization of damages of aviation policy scenarios within the US. APMT can assess policies from a US perspective, but does not currently have the ability to account for a global-scale analysis. Barrett et al.,14 however, has shown that cruise emissions have a significant impact on groundlevel air quality, where -8000 premature mortalities per year are attributable to aircraft cruise emissions, which constitutes 80% of the total mortality impact derived from all aircraft emissions. Given the time scale of removal and transport of PM 2.5 and its associated precursors, aircraft emissions then have an inherently global impact. As such, it becomes necessary to adjust the scope of policy analysis to incorporate the global atmospheric impact of a particular aviation policy. This requires global atmospheric modeling through the use of global chemical transport models (GCTMs), rather than only more regional or localized air quality models. In addition to other difficulties, it is also necessary to determine the ground-level health impact in countries outside of the US where epidemiological data may not be available. The primary goal of this thesis is to expand the analytical capabilities of PARTNER and the APMT tool suite to include global scale BCA studies of aircraft policy emissions scenarios. This is not only important for the specific case of ULSJ, but also relevant for any future alternative jet 5 fuel study where a comprehensive BCA may be necessary to validate or invalidate a transition away from today's standard jet fuel. This requires the development of a comprehensive global environmental benefit/cost analysis framework in which to streamline current and future aviation policy analyses in order to achieve the afore stated FAA policy goals. 1.4. Thesis Structure This thesis has the following structure: Chapter 1 generally addresses how aircraft impact the environment and more specifically discusses some of these prospective policy goals and potential aviation policy options. Chapter 2 provides background on general EBCA methodology, including current benefit-cost analysis techniques employed by the Environmental Protection Agency (EPA). Chapter 3 discusses the EBCA methodology developed for the purposes of global policy analysis, including determining aviation emissions impacts and monetizing these impacts. This is developed within a framework to address the environmental and economic impacts of an ultra-low sulfur jet fuel (USLJ) scenario that would require all commercial jet fuel to have a fuel sulfur content (FSC) of less than 15 ppm. Chapter 4 presents the results of an EBCA as applied to a case study of global and US implementation of ULSJ. Chapter 5 extends this policy analysis framework to a discussion of a fast policy tool derived from an adjoint sensitivity analysis. A comparison between the previous EBCA mortality outcomes versus those calculated from the adjoint fast policy tool is also provided. Chapter 6 presents results from an adjoint sensitivity analysis as applied to temporal variations in aviation emissions impacts and a discussion of the potential policy implications of these variations. Finally, chapter 7 concludes this thesis and provides a discussion of possible future work. 6 THIS PAGE INTENTIONALLY LEFT BLANK. 7 Chapter 2: Background This chapter provides background information for the EBCA methodology that will be developed in Chapter 3. It includes a discussion on the types of atmospheric models used as well as some basic information concerning how aerosols (i.e. PM2.5 ) impact the radiative balance of the atmosphere. In addition, literature is reviewed related to the link between PM2.5 concentrations and premature mortality risk and current EBCA practices. Information is also presented from an ultra-low sulfur diesel (ULSD) case study that was performed given the similarities between diesel and jet fuel. Finally, a brief description of adjoint sensitivity analysis and how it relates to policy analysis is provided. 2.1. Atmospheric Modeling EBCA is directly dependent on the ability to chemically and physically model the perturbation to the atmosphere resulting from anthropogenic emissions. Once emissions scenarios have been accurately defined, it is then necessary to incorporate these scenarios into the appropriate physical models. Two different models are generally employed within PARTNER to conduct air quality studies. On a global scale, GEOS-Chem, a GCTM, is typically used, although the nested version can also be applied to smaller, higher resolution domains. For more local studies (i.e. US, only), the Community Multiscale Air Quality (CMAQ) model is used. CMAQ has been used by the EPA in their air quality policy analyses, but because CMAQ was used as a sensitivity study compared to GEOS-Chem and nested GEOS-Chem simulations for the ULSJ analysis, no detailed description will be provided in this thesis. 2.1.1. Global and Nested GEOS-Chem Model Descriptions GEOS-Chem is a tropospheric model (with an approximated stratosphere) that simulates global gas and aerosol phase chemistry (including aerosols that are relevant to aircraft emissions), accounts for wet and dry deposition, and models transport on an intercontinental scale. 15 The model takes in emissions and meteorology as inputs and at the end of each time step computes new tracer concentrations. A typical chemistry time step is 60 minutes and a year simulation with a 3 month spin-up period, which is required to eliminate the impact of the initial conditions, takes approximately 12 hours to perform. 8 GEOS-Chem generates three dimensional, gridded data. For the ULSJ analysis, a 40 x 50 horizontal grid is typically used, while nested simulations use a 0.5* x 0.6660 horizontal grid where boundary conditions are provided by the coarser resolution simulation. These simulations also use the GEOS-5 reduced vertical grid that is defined up to 0.01 hPa, where the atmosphere is split into 47 layers. GEOS-Chem also has a built-in aerosol optical property module in which it calculates optical depths of the primary aerosol species, but it currently does not have an integrated radiative transfer model (RTM) to allow for climate studies. An aircrafts emissions module was previously implemented into GEOS-Chem where species concentrations in the appropriate grid cells were perturbed given aircraft emissions at that location. The model currently allows the user to switch on or off particular aircraft emissions types, to choose to include full aviation emissions or only LTO emissions, and to set the FSC. 2.2. Impact of Aerosols on Climate Inorganic aerosols, such as ammonium nitrate and ammonium sulfate, scatter incoming solar radiation. In atmospheric models, sulfate aerosols are considered to be purely scattering, i.e. extinction of incoming solar radiation is achieved by only refraction and not through absorption. 16 With sulfate aerosols, the amount of solar radiation that is refracted back into space is determined by a given particle's backscattering coefficient, 0. 1is a function of several parameters, including solar incidence angle and radiation wavelength, but most notably it is a function of particle size. In general, aerosols can exist as a solid or as an aqueous particle. Transition between solid and aqueous states for sulfate aerosols is governed by a hysteresis cycle. In this cycle, aerosol particle radius is a function of relative humidity given the hygroscopic growth (i.e. water vapor condensation) that occurs on the particle, which in turn impacts the optical properties of the particle.17 Thus, an increase in relative humidity can cause an increase in particle radius through water condensation in order to maintain thermodynamic equilibrium, but this growth behavior is specific to the relative humidity history of the particle. Figure 2 gives an illustration of this hysteresis cycle. 9 s-. 2.5 0- DRo DRH 0 0z~ U CR1-I 0- 1.0 0 20 40 60 80 100 Relative Humidity (%) Figure 2: Multi-branch hysteresis behavior of an ammonium sulfate aerosol particle. Plot from Wang et al." From Figure 2, there are two distinct pathways that the ammonium sulfate aerosol particle can follow, which are in general called the upper and lower hysteresis branches. In the lower branch (solid line), no hygroscopic growth is experienced until a threshold RH value of 80% (DRHo) is reached if the particle is initially solid. Similarly, an aqueous particle will not begin to crystallize if it exists on the upper branch (dashed line) until a threshold RH value of 35% (CRHo) is reached. Thus, relative humidity history dictates on which branch a given aerosol particle exists given some assumption concerning the initial phase of the particle of interest. In addition, different aerosol types have different hysteresis pathways, which add to modeling complexity when aerosol mixing ratios need to be determined or assumed. From this particle growth data as well as total atmospheric burdens, the aerosol optical depth can be calculated, which is a dimensionless measure of the atmospheric burden of the aerosol as it relates to extinction of solar radiation and is mathematically defined as (x1, x 2 ; b(x,2) dx 2=f Eq. (1).16 As can be seen in Eq. (1), the optical depth, -r, is a function of the two integration bounds as well as the wavelength of the incoming solar radiation. b is known as the extinction coefficient and is proportional to the burden of the species within the atmosphere as well as the optical properties of that particular species. Optical depth is directly related to the RF contribution of that species. 10 The methodology implemented for this thesis for RF due to sulfate aerosol scattering is discussed further in Section 3.2.1. GEOS-Chem takes a simplified approach to the treatment of inorganic aerosols and its calculation of optical depth. GEOS-Chem currently tracks atmospheric burdens of nitrates, sulfates, and ammonium individually, but they are lumped together within the "sulfates" bin when performing optical calculations, where the primary underlying assumption is that the three different species are internally mixed (i.e. mixed aerosols are homogenous). GEOS-Chem, however, does not consider more complex internally mixed aerosols, such as liquid H2 SO4 deposited on BC particles. The particle size distributions for dry "sulfates" (i.e. 0% RH) are assumed to be as shown in Table 1. Table 1: Assumed properties of "sulfates" optical bin in GEOS-Chem. Property Density Geometric Mean Radius Geometric Standard Deviation Effective Radius Refractive Index Value 1.7g/cm 0.07 pm 1.6 0.12 pm 1.43-10-8i Additionally, the hygroscopic growth behavior of sulfate aerosols within GEOS-Chem is outlined in Martin et al. The assumed refractive index is that of sulfuric acid aerosol and no complex hysteresis behavior is assumed. Different mixing fractions, however, and the impact on the assumed refractive index are not considered and is the subject of ongoing development within the model. 2.3. Air Quality and Mortality This section contains significant contributions from Dr. Jon Levy from Boston University School of Public Health. Many studies have linked PM 2.5 exposure to adverse health end-points in the US, most notably the American Cancer Society (ACS) 19,20 and Harvard Six Cities2 cohort studies in which populations were monitored over time and the associations between several health incidences and long-term exposure to ground level PM2.5 concentrations were monitored. For this analysis, only premature mortality is considered given that approximately 85% of all monetized health impacts are due to premature mortality.12 Changes in concentrations of PM 2.5 can be related to 11 avoided mortalities through a concentration-response function (CRF). For this analysis, a linear CRF is used based on follow-up analyses of the aforementioned cohort studies and an EPA expert elicitation study,23 which comprehensively describes current interpretation of epidemiological evidence based on expert opinion. Studies, as outlined in Pope et al. 24 provide evidence of a link between premature mortality and long-term exposure to PM2.5 . The EPA expert elicitation study2 showed that several leading experts believe that the main contributors to mortality from short-term exposure are acute cardiac or respiratory events, possibly from pre-existing health conditions. Mortality due to longterm exposure is most likely a result of cardiovascular disease, chronic respiratory disease, and lung cancer. The biological drivers of these premature mortalities are still uncertain, although it is strongly believed that PM exposure can lead to health issues such as chronic obstructive pulmonary disease (COPD)25,26 and atherosclerosis.27 An impact on lung cancer risk due to PM exposure is thought to exist, but the relationship remains uncertain as compared to cardiopulmonary illnesses,28,29,30 although relative risks have been calculated for lung cancer. Given these established links to human health, deaths resulting from cardiopulmonary (CP) diseases and lung cancer (LC) are assumed to be the dominating health impact pathways given a change in ground-level PM2.5 concentrations. Uncertainties for each disease are captured in a range of relative risk values obtained through literature review. A linear CRF is assumed given the results from multiple analyses of long-term studies. In addition, no substantial evidence has been found for a concentration level at which health impacts exhibit a significantly different concentration response. 24 Specifically, Schwartz et al. 31 conducted an analysis that included threshold and non-linear models, and also concluded that a linear fit does indeed best fit the data. This interpretation concerning CRF shape is also reiterated within the EPA expert elicitation study. Nonlinearities are observed, specifically in the Pope ACS cohort analysis. Goodness-of-fit tests, however, have shown that the probability of a non-linear fit being statistically different from that of a linear fit is not significant.24 These cohort studies, however, did not consider very high pollutant concentration levels that may be seen in portions of the developing world (-50 pg/m 3 versus -10 pg/m 3 in the US). In these cases, the assumption of a linear CRF may be inappropriate as the marginal impact of pollution at very high concentration levels may be decreasing and the number of premature mortalities predicted may be overestimated. To account for this, the CRF applied in Ostro86 as part of a World Health Organization (WHO) study assumes a log-linear relationship with changes in concentrations, 12 which results in a lower premature mortality response at higher background PM2.5 concentrations. Differential toxicity among PM 2.5 species is a subject of ongoing research and is relevant given that the change in ground-level PM 2.5 due to ULSJ implementation is seen primarily in S04 species. The relative impact of S04 species on air quality becomes important in determining the 32 magnitude of avoided mortalities seen due to a reduction in FSC. Hedley et al. showed that a reduction in SOx emissions in Hong Kong due to a sulfur content restriction of 5000 ppm for fuel used in power plants and road vehicles led to 80% and 41% reductions in S02 pollution and SO4 particulate concentrations, respectively. This improvement in overall air quality was accompanied by a 2.1% reduction in all-cause mortality, a 3.9% reduction in respiratory disease related mortalities, and a 2.0% reduction in cardiovascular related mortalities in annually averaged trends after the fuel regulation was enforced. In addition, some studies have shown a link between sulfate aerosols and CP related health incidences,3 3 34 3, 5 and committees have concluded that there is currently not enough evidence to discredit the assumption of equal toxicity. 53,36 More recent results presented by Levy et al.,37 where a probabilistic analysis was applied to determine the relative toxicity of constituent species relative to PM 2.5 as a whole, however, show that sulfate has only a 42.7% of being more toxic than PM2.5 with regards to cardiopulmonary illnesses. This is in stark contrast to elemental carbon and nitrate, which have a 100% and 97.7% chance of being more toxic than PM2.5. Probabilities are lower in general for respiratory illnesses. Levy et al.,37 however, does acknowledge the large uncertainty present in this analysis and state that the results only suggest the possibility of differential toxicity. In this thesis, differential toxicity is not directly considered and all PM2.5 species, including those derived from SOx emissions, are assumed to have equal health impacts. This is, however, a significant uncertainty. 2.4. Elements of EBCA For the purposes of this thesis, many of the guidelines set by the EPA have been incorporated into the techniques employed within this particular analysis. This section will serve as a brief overview of these practices and any assumptions that were required when these guidelines could not be strictly met. 13 2.4.1. Benefit/Cost Analysis BCA (or also commonly, CBA) is a general framework in which policy analysis is commonly performed. It requires the identification of both benefits and costs, each of which can be positive or negative for society (negative benefits are also referred to as "disbenefits"). Distinguishing between benefits and costs is, in some cases, not always straightforward. For the case of environmental policy analysis, a general rule of thumb is that positive or negative outcomes derived from the actual implementation of a policy fall under the benefits category (e.g. decrease in premature mortality or increase in climate damages) while costs are assigned towards the actual implementation of the policy, be it the need for additional infrastructure, land, etc. Furthermore, BCA requires a definition of both a baseline and policy scenario. Within aviation policy, the baseline scenario has been defined using aviation in the year 2006 where the most complete set of aircraft emissions inventory data available. The policy scenario is then defined relative to this baseline aviation scenario. As an example, the ULSJ policy scenario would have aircraft SOx emissions reduced by a factor of 15/600, where this factor is the ratio of the proposed FSC of 15 ppm to the current average FSC of 600 ppm. As the atmosphere is a highly nonlinear system, the definition of the background emissions, i.e. emissions from all other sources not including aviation, is also important for both the baseline and policy scenario outcomes. As the focus of this thesis is on aviation policy, the topic of background emissions will not be addressed in detail, but a brief description of the source for these background emissions is provided in Section 3. Sensitivity analysis and uncertainty quantification also play a vital role in BCA. From the atmospheric modeling to metric monetization, policy analysis is a highly uncertain process. This is due in large part to the limitations of the physical models used and, particularly for environmental analysis, the lack of real markets in which to obtain monetary values that can be easily assigned to climate damages or premature mortalities. As such, it becomes necessary to utilize Monte Carlo techniques in order to provide both a policy outcome distribution as well as a comprehensive sensitivity analysis to determine the primary sources of uncertainty given the assumptions within the analysis. 14 2.4.2. Discounting A key component of BCA is the discounting of all future benefits/costs. In an economic sense, money now is worth more to a person than money in the future and the degree to which a person experiences that opportunity cost is contained within the discount rate, which is directly tied to the concept of a net present value (NPV). NPV is defined in the following equation: NPVCt t1(1 +r)t Eq.(2) where Ct is the cash flow for that time period, t is the time period, and r is the discount rate. Within the base year of the analysis, the time period is 0 (i.e. costs in the initial year of a policy are not discounted). The cash flow can be defined as the relevant benefit or cost quantity within that time period. The US government has identified a range of appropriate discount rates (27%)38 to be applied in BCA, where this range is also applied for aviation environmental policy analysis. The higher the discount rate, the more current costs are valued against future costs, and vice versa. Thus, different choices in discount rate will cause different policy assessment outcomes, especially when the time scales of benefits and/or costs are significantly different. Discounting plays an important role in valuing both climate damages as well as human health impacts. Aviation emissions in a given year can have impacts out to hundreds of years, thus it becomes necessary to take the NPV of all costs over a time horizon that captures these impacts to fully quantify the climate benefit/disbenefit associated with a given policy scenario. In the case of climate damages, Eq. (2) can be directly applied where the cash flow for a specific year is the climate damage in that year, all of which is computed within APMT. A more detailed description of how APMT assigns climate damages will be discussed in Section 3.2.5. As suggested by the EPA, it is also necessary to discount health impacts based on a prescribed lag schedule. It assumes that 30% of avoided mortalities are seen in the year of implementation, 50% in years 2-5, and the remaining 20% spread out over years 6-20.39 The assumed discount rate can then be applied to each year as required, and then premature mortalities can be summed over the full 20 years to obtain a "discounted" premature mortality outcome. This thesis applies the same discount rate for all discounting purposes as the EPA does not provide any insight on application-specific discount rates. 15 2.4.3. Premature Mortality Based on the information presented above, the EPA recommends the use of a linear CRF with a nominal estimate of 1.06% for the CRF coefficient. A Weibull distribution was assumed for the CRF coefficient within The Benefits and Costs of the Clean Air Act from 1990 to 2020,39 but no distribution parameters were provided. For this thesis, upper and lower bounds were determined from the EPA expert elicitation study concerning CRFs. 2.4.4. Valuing Lives Valuing premature mortalities is a very complex issue, both from an economic as well as social justice perspective. The US recommends that the value of a statistical life (VSL), largely calculated from wage-risk studies, be applied in matters of policy analysis when mortality monetization is required. As this tends to be a sensitive issue within policy analysis, a brief overview of VSL use standardization within the US is provided below. Standardizing VSL use within US cost-benefit analyses was first addressed in OMB Circular A4.40 A literature review concluded that the "substantial majority of the resulting estimates of VSL vary from roughly $1 to $10 million per statistical life." Based on a review by the EPA's science advisory board (SAB), two important conclusions were drawn: it is appropriate to adjust VSL relative to income and a lag structure for health effects should be applied. OMB advised a standard value of $5 million, but acknowledged the value was lower than other estimates in federal agencies, specifically the EPA. In DOT's 2009 VSL guidance memorandum, 4 1 the standard estimate applied within DOT was updated from a previous value provided in a 2008 memorandum. In this update, five different US VSL studies were taken and adjusted for real income growth (with an assumed income elasticity of 0.55 based on a literature review) as well as inflation using the consumer price index (CPI). $5.8 million was given as the mean value adjusted to 2007 prices. While the original OMB range of $1 to $10 million was not ruled out, DOT suggested a more specific uncertainty range of $3.2 to $8.4 million. A normal distribution with a standard deviation of 2.6 million was recommended, although due to the unrealistic negative values that this encompassed, other distributions, such as Weibull or lognormal, were also recommended, but no distribution parameters are provided. In terms of policy applications, DOT stated "the same standard is to be applied to all individuals at risk, regardless of age, location, income, or mode 16 of travel," thus setting requirements that all individuals within the US be monetized by the same VSL. In EPA's Guidelines for Preparing Economic Analyses,38 a central estimate of VSL is given as $7.4 million. This value is based on 26 VSL studies. VSLs are adjusted year to year by a GDP deflator, although it appears that no adjustment is made for real income growth or decline. A Weibull distribution was then fit to these 26 studies, resulting in a scale parameter of 7.75 and a shape parameter of 1.51. The EPA lists several limitations to the provided estimate, including its lack of specificity to environmental hazards, as is true of all current VSL estimates. It concludes that for now, the abovementioned central estimate is the most appropriate estimate of US VSL and should be applied uniformly without the consideration of differences of income within the US. In order to be consistent with previous analyses conducted on the impact of changes in air quality on human health, the EPA approach was applied in both calculating and monetizing avoided premature mortalities. Due to the global scope of this thesis, differentiating between premature mortalities in different regions in the world by applying different VSLs could be interpreted to suggest that a person's life in a higher income country is "worth" more than a life in a lower income country. Heinzerling42 argues that VSL is in fact an inappropriate valuation to apply to premature mortalities as it does not capture the value of lost (or saved) life but rather just the value of the increased risk associated with a more "unsafe" environment. In this thesis, country specific VSLs are applied, and the justification for this approach is provided in Section 3.3, although a constant VSL approach is also applied as a comparison. One of the disadvantages of using VSL as a mortality monetization metric is that it does not capture the temporal aspect of when the premature mortality occurs. For instance, complications from CP illnesses may place the elderly at much greater risk than younger members of the population, but a premature mortality from either age group is valued equally. Thus, an alternative metric, as applied in the ExternE43 study, is the value of lost life years (VOLLY). This involves calculating the expected age of mortality relative to a reference life expectancy, and then monetizing each of the years "lost" due to premature mortality. Although this helps resolve the temporal issue related to the VSL, this approach is not applied in this thesis given the lack of data related to value of a life year lost on a disease specific basis as well as a lack of precedent in US environmental policy analysis. 17 2.4.5. Cost Analysis Here it is important to distinguish between a financial BCA, in which "real" prices are analyzed, versus a societal BCA, in which the "true" costs seen by society are analyzed.4 As environmental BCA certainly falls under the latter category, it is necessary to obtain the "shadow" prices (i.e. prices that are not distorted by market effects) when determining the cost of policy implementation. An example of a market distortion might be the oligarchical pricing effect due to the small number of firms within the oil refining industry. As such, the cost analysis portion of this thesis attempts to determine the "true" costs seen by society as a result of the implementation of ULSJ through both a top-down and bottom-up approach. 2.5. Ultra-Low Sulfur Fuels 2.5.1. Ultra-Low Sulfur Diesel Case Study Given the similarities between diesel and jet fuel, 45 ultra-low sulfur diesel (ULSD) implementation within the United States was used as a comparative case study for ULSJ, especially for the cost analysis. ULSD, through EPA rulemaking, was phased-in to production for on and off-road uses. Table 2 shows the timeline for on-road implementation of ULSD, while Table 3 shows the timeline for off-road implementation of ULSD. US implementation of ULSD exhibited a pattern of a gradual increase in fuel sulfur content (FSC) stringency as the fuel passed from refineries to retail outlets. This implementation was seen over a time period of 10 years for both on and off-road uses. 18 Table 2: On-road implementation timeline of ULSD in the US.46,47 Requirement Announcement of Diesel Fuel Sulfur Content Regulation for On-Road Vehicles Proposed Heavy-Duty Engine and Vehicle Standards and Highway Diesel Fuel Sulfur Control Requirement Heavy-Duty Engine and Vehicle Standards and Highway Diesel Fuel Sulfur Control Requirements Final Rule Refiners and Importers: 80% of Diesel Fuel Imported/Produced must be ULSD Fuel Terminals: Fuel listed as ULSD must meet 15 ppm specification Retail Outlets: Fuel listed as ULSD must meet 15 ppm specification Refiners and Importers: 100% of Diesel Fuel Imported/Produced must be ULSD Fuel Terminals: All highway diesel must be ULSD Retail Outlets: All highway diesel must be ULSD Date May 1997 May 2000 January 2001 Description Reducing sulfur content of diesel fuel for heavyduty diesel engines is identified as a potential pathway to improve air quality. Proposed requirement to reduce sulfur content of diesel fuel for highway vehicles to no greater than 15 parts per million (ppm) with a start date of June 1, 2006. Final rule requires that refiners begin producing 15 ppm sulfur content diesel fuel beginning June 1, 2006. June 2006 N/A September 2006 October 2006 June 2010 N/A October 2010 December 2010 N/A 19 N/A N/A Based on a ULSD pump survey, 85% of pumps were dispensing ULSD in the 4th quarter of 2006. 100% of highway diesel fuel pumps are now dispensing ULSD as of the 3 rd quarter of 2010. Table 3: Off-road implementation timeline of ULSD in the US.46,47 Requirement Proposed Clean Air NonRoad Diesel Rule Date April 2003 Final Clean Air Non-Road Diesel Rule May 2004 Description Reduce diesel fuel sulfur content to a maximum of 500 ppm starting in 2007 for non-road applications (including locomotive and marine applications). Reduce diesel fuel sulfur content to a maximum of 15 ppm by 2010. Non-road diesel fuel sulfur content must be reduced from current levels (about 3000 ppm) to 15 ppm by 2010. Fuel must meet 500 ppm standard. Refiners and Importers: Non- June 2007 Road, Locomotive, and Marine Fuel Fuel Terminals: Non-Road, August 2007 Locomotive, and Marine Fuel Retail Outlets: Non-Road, October 2007 Locomotive, and Marine Fuel Refiners and Importers: Non- June 2010 Road Fuel Refiners and Importers: June 2010 Locomotive and Marine Fuel Fuel Terminals: Non-Road August 2010 Fuel Fuel Terminals: Locomotive August 2012 and Marine Fuel Retail Outlets: Non-Road October 2012 Fuel Retail Outlets: Locomotive October 2012 and Marine Fuel Fuel must meet 500 ppm standard. Fuel must meet 500 ppm standard. Fuel must meet 15 ppm standard. Fuel must meet 15 ppm standard. Fuel must meet 15 ppm standard. Fuel must meet 15 ppm standard. Fuel must meet 15 ppm standard. Fuel must meet 15 ppm standard. 2.5.2. QinetiQ Report on Jet Fuel Sulfur Limit Reduction A report by QinetiQ48 addressed ULSJ implementation in Europe. It estimated that due to the additional hydroprocessing required, there will be a 0.01 to 0.015 EUR/liter additional required cost in ULSJ production, which is approximately 4 to 7 cents (2006 US$) per gallon. The report also outlined many of the potential impacts the additional hydroprocessing would have on fuel properties as well as operational effects due to the reduction in fuel sulfur content. Potential climate impacts were described, but were not quantified. SO2 emissions as a function of FSC were estimated for a representative local airport by scaling against a previous dispersion model study at Heathrow Airport49 based on emissions derived from the First Order Approximation methodology.50 The report concluded there is unlikely to be any measurable health effect due to FSC reduction when only LTO emissions are considered. Full-flight emissions impacts were not addressed. 20 2.5.3. Energy Information Administration (EIA) Report on Market Effects Due to ULSD An EIA report from 200151 analyzed the possible effects of ultra low sulfur diesel (ULSD) implementation on the diesel fuel market within the US. Based on a cost curve analysis, a 6.5 to 8.2 (2006 US$) cent/gal marginal cost increase for ULSD production was estimated to cover additional capital and hydrotreating costs for an assumed future supply and demand. In comparison, the EPA52 predicted a full compliance US average cost of 5.2 cents/gal (2006 US$). 2.5.4. Other Transportation Sectors Marine fuels have also received significant attention in terms of their global air quality impact. A policy analysis performed by Corbett and Winebrake 53 estimated a 70 to 85% reduction in marine SO,, emissions due to marine gas oil (MGO) and marine diesel oil (MDO) implementation (lower sulfur alternatives) over the standard marine residual oil (RO). Additional C02 emissions were estimated to be less than 1%. In a human health policy analysis, Winebrake et al.54 estimated the total health impacts due to a global marine fuel sulfur content limit. Findings showed a 41,200 reduction in premature mortalities for a global fuel sulfur content limit of 5000 ppm as compared to 87,000 premature mortalities with the assumed baseline emissions scenario. Marine fuel use is defined as an off-road diesel fuel and all marine fuel must meet the 15 ppm standard by the end of 2012, as outlined in Table 3, for all US marine applications except for RO used by ocean-going ships. 2.6. Role of Sensitivity Analysis In general, sensitivity analysis is a useful tool to supplement policy analysis. In the case of performing large scale atmospheric chemistry simulations, sensitivity analysis allows for the determination of perturbations to atmospheric concentrations given a small perturbation in emissions without the need to completely rerun entire GCTM simulations. For GCTMs in particular, however, sensitivity calculation to emissions is not always a simple task, either due to model complexity or excessive computation time. Sensitivity analysis begins by the definition of a cost function, J, which can be any function of the inputs to a given model. For instance, one can define the cost function of a GCTM simulation to be the total atmospheric concentration of a particular trace gas (e.g. ozone) or aerosol (e.g. sulfate) at the final time step, N, of the simulation. Thus, through sensitivity analysis techniques, one can determine the impact of 21 emissions in each of the time steps (i.e. 1 to N) on the atmospheric concentration of a particular species in the final time step. There are two primary methods by which sensitivities can be calculated. In the forward calculation, each input is perturbed individually, and for each perturbation, a simulation is performed to determine the impact on the defined cost function, J. While simple, this approach very quickly becomes computationally expensive given the number of forward model simulations required. For GCTMs such as GEOS-Chem, a forward calculation of sensitivities for just one particular emission would require the number of forward simulations equal to the number of grid boxes times the number of time steps. In a standard GEOS-Chem simulation performed for this thesis, there are a total of 72 x 46 x 47 grid boxes (155,664) and within a year simulation, there are 8,760 time steps, where the standard time step for emissions is 60 minutes. Given that a single year long simulation takes about 12 hours to complete, this problem then becomes far too large given the relatively limited computational resources available. An alternative procedure for sensitivity calculation is the adjoint method, which has already been implemented within GEOS-Chem. Adjoint methods take a "backward integration" approach to sensitivity calculation. First, the governing equations of the GCTM are linearized so that the following approximation holds true: Ax" = xn+ Eq. (3), where A represents the approximate linear operator, xn is the input vector (i.e. grid concentration values) for the current time step, and x"*< is the output vector (i.e. grid concentration values to serve as inputs in the next time step). As mentioned previously, the cost function, J, can be defined in terms of the output, x''. In that case, the perturbation to J, or 5J, can be computed by the following relationship: SJ = axn+l J xn+1 Eq. (4), where the transpose notation is a result of the matrix multiplication required given that x is a vector input. From Eq. (3), given that it is a linear approximation, small perturbations in the output can be calculated from the following equation: 8xn+1 = ASx" 22 Eq. (5). Substituting Eq. (5) into Eq. (4) and rearranging some of the notation, the following relationship is obtained: Sx" ] SJ =AT Eq. (6). Equivalently, Eq. (4) can be written in terms of sensitivity to inputs: 6J = axn 6x' Eq. (7). Comparing Eq. (6) to Eq. (7), the first order derivative becomes a = AT Eq. (8), where it can be observed that A =axni Eq. (9), which, based on the previous definitions, is the Jacobian of the system. From Eq. (8), it is then clear that the sensitivity of the cost function to concentrations in previous time steps can be calculated iteratively by multiplying the sensitivity at the current time step by the Jacobian evaluated at that time step. Determining the Jacobian can be a difficult task if the GCTM is inherently complex or poorly organized, where substantial understanding of the underlying code structure is required to implement the adjoint method. The primary advantage to this "backwards" approach, however, is the wealth of sensitivity data obtained from a single adjoint simulation. It is possible to obtain full 3-D sensitivity data, as well as temporal sensitivity data, for, in the case of GEOS-Chem, an approximate 2.5 times increase in runtime over a standard forward model simulation. These equations are relevant for discrete adjoint method implementation. The topic of continuous adjoint methods will not be discussed in this thesis. Given that the adjoint method has been implemented within GEOS-Chem, there is considerable freedom in how J is defined. Currently, GEOS-Chem is only compatible with cost functions that are first order differentiable to the outputs. For instance, the cost function can be the total atmospheric ozone burden in the final time step or, alternatively, in all time steps. Thus, J is a sum of the ozone mass in every grid cell over the appropriate time horizon, which is first order differentiable given it is a linear cost function. J can also be defined as a weighted sum of concentration, where the cost function has the following general form: 23 I = 2%=1 L=1X% k4=1 Wi,jkXi,j,k,n Eq. (10), where w is the grid (defined by i, j, k) specific weight value (assumed to be constant in time), x is the grid and time specific concentration of the species relevant to the cost function, n denotes the time step index, i denotes the longitudinal index, j denotes the latitudinal index, and k denotes the altitudinal index, and N, /, J, and K are the limits on these indices, respectively. The weighting term is defined based on the policy metric of interest. If, for instance, the total atmospheric burden is the desired output, then the weighting in each grid cell will be 1 (assuming the concentration is outputted as a mass). Alternatively, if a population weighted metric is required, then the weighting will be the population specific to that grid. In addition, the definition of the cost function can be restricted to a certain domain, where grid cells within the domain (e.g. the contiguous US) have a non-zero weighting and grid cells outside of the domain have a zero weighting. The exact definition of the weighting functions used in this thesis will be addressed in Section 5.1. The adjoint model was developed and validated within GEOS-Chem by Henze et al.55 Previous research performed by Koo5 focused on the application of the GEOS-Chem adjoint model on the air quality impact of aircraft emission, where both spatial and temporal studies were conducted and validated. For these aviation air quality studies, the cost functions were defined as the (weighted or non-weighted) PM 2.5 burden at ground-level. One of the primary results from Koo showed the impacts of aviation emissions from one region (source) on the expected premature mortality in another region (receptor) based on adjoint sensitivity simulations. For instance, aviation emissions in the US have the largest premature mortality response in the Asian domain given the intercontinental nature of the transport of PM 2.5 and its precursors. This data is presented in Table 4, where the regions considered are the US, North America (NA), Europe (EU), Asia, and the entire world. Table 4: Source-receptor mortality impact from full-flight aircraft emissions. 56 US NA EU Asia World To US 180 210 20 30 320 NA 220 310 50 60 490 24 EU 490 750 2010 380 3600 Asia 1620 2470 1630 2830 8760 World 2240 3400 3010 3300 12150 THIS PAGE INTENTIONALLY LEFT BLANK. 25 Chapter 3: EBCA Methodology Development This chapter provides a detailed description of the EBCA methodology that was developed or adapted for the specific case of ULSJ implementation globally as well as just within the US. It covers the entire policy analysis pathway, from the definition of the emissions scenario to the final monetization and comparison of policy outcomes. While the methodology in this section was developed for the ULSJ case, it is also helps to establish a general framework in which to conduct EBCA on global aviation policies, particularly those that are air quality and human health motivated. 3.1. Emissions Scenarios Emissions were derived from output from the FAA's Aviation Environmental Design Tool (AEDT). AEDT calculates aircraft fuel burn and emissions on a flight-by-flight basis, covering the majority of civil aviation. A procedure similar to that applied by Barrett et al.57 was used to modify AEDT output for use in this thesis, which is outlined in Table 5 where AEDT outputs are bolded. For the baseline case, a FSC of 600 ppm was assumed, while for the ULSJ (i.e. policy) scenario, a FSC of 15 ppm was applied. The relationship between FSC and SO 2 and SO 4 emissions is also provided in Table 5. The background emissions inventory used within GEOS-Chem will not be discussed in-depth within this thesis, but a description of these background emissions is provided in Donkelaar, et al. 58 There are also several expected changes in fuel properties that will be discussed in greater detail in Section 3.7.1, but these expected changes were not considered in the definition of the policy emissions scenario. 26 Table 5: Emission indices methodology for air quality simulations, where bolded variables are from AEDT. Species CO2 Baseline Emissions (g) 3159 x FUEL H20 NOx as NO2 1231 NO, CO HC as CH 4 TOG BC < 3000ft CO HC 1.16 x HC PMNV BC > 3000ft OC < 3000ft 0.03 x FUEL PMFO OC > 3000ft SO 2 0.03 x FUEL (FSC/1 000) x [(100 - E)/100] x FUEL x (64/32) (FSC/1000) x (E/100) x FUEL x (96/32) Svi as S04 x Description/Notes C02 aircraft emissions, constant is adjusted to 3150 FUEL for ULSJ, FUEL represents the fuel burn value obtained from AEDT Aircraft H20 emissions AEDT default value, NO/NO 2 mole fraction partitioning changes between LTO/non-LTO AEDT default value CH 4 equivalent, AEDT default value, speciated Total organic gases aircraft emissions, speciated Black carbon emissions below 3000 ft, AEDT default value for Black carbon with small amounts of metals (PMNV) Black carbon emissions above 3000 ft Organic carbon emissions below 3000 ft, AEDT default value for organic PM from fuel (PMFO) Organic Carbon emissions above 3000 ft S02 emissions, based on fuel sulfur content (FSC) in ppm (by wt) and wt-% of fuel sulfur emitted at Sv (E) Assumes Sv' emitted as SO 4 3.2. Determining Climate Impacts Two changes in emissions due to ULSJ implementation are considered in this analysis: a decrease in aircraft SOx emissions directly due to fuel desulfurization and an increase in well-towake (WTW) greenhouse gas (GHG) emissions due to increased fuel processing requirements. Because atmospheric sulfate aerosol concentrations are reduced given the decrease in the sulfur content in the jet fuel, the impact on RF, a metric used to quantify the net effect of a particular species on the global radiation energy balance, is determined. Lee et al.59 provides values for aviation RF impacts, estimating the sulfate aerosol impact as -4.8 (90% Cl: -0.79 to 29.3) mW/m 2 for 2005 aviation, where the negative RF value implies cooling. Sulfate aerosols have a cooling effect on the atmosphere, thus a decrease in the sulfur content of fuel is expected to cause a net warming effect when comparing ULSJ to standard aviation jet fuel when only direct effects are considered. The increase in WTW GHG emissions was previously analyzed in Stratton et al.61 While direct climate impacts are the primary focus, indirect effects are also possible. In general, the formation of particulate matter within the atmosphere provides nuclei from which clouds may 27 propagate, also known as cloud condensation nuclei (CCN). Thus, there is a possibility that a decrease in sulfate aerosols would remove possible CCN, thereby decreasing overall cloud cover as CCN formation is a function of specie particle number, not specie mass. Whether or not this leads to a net warming or cooling, however, is difficult to determine as it depends largely on the change in size distribution of the aerosol particles. In general, smaller particles are not as easily removed and generate longer-living clouds.16 Thus, a decrease in sulfate aerosol concentration may or may not have a significant impact on cloud formation and/or lifetimes. Additionally, when H2SO 4 condenses on BC, a phenomenon known as optical focusing occurs, which results in a net warming. Thus, desulfurization of jet fuel can have a cooling effect to the extent that optical focusing is important. This phenomenon, however, is not considered as it is not modeled within GEOS-Chem. As mentioned in Section 1.1, indirect climate effects are highly uncertain, and given this uncertainty, these indirect effects are not addressed in any detail within this thesis. An analysis of this type may be possible in the future given sufficient development of a climate-coupled GCTM and a better understanding of aerosol mass impact on particle number and size distributions. 3.2.1. Sulfate RF Calculation While higher fidelity calculation of RF can be accomplished through the use of a radiative transfer models (RTM), GEOS-Chem does not currently have a built-in RTM to allow for such calculations, thus a simplified approach is taken here. Eq. (11) is used to determine the globally averaged direct radiative forcing due to sulfate aerosols, or sulfate direct radiative forcing 1 (DRF), from aerosol optical depth (-) quantities calculated in GEOS-Chem 7,62: DRF = where 4 FT -FTT2(1 - Ac)2(1 -RS) 2 (saTsd + /aqTaq) Eq. (11) is the global mean top-of-the-atmosphere radiative flux, T is the fraction of incident light transmitted by the atmospheric layer above the aerosol layer, Ac is the fractional amount of cloud cover, f, is the area averaged albedo of the underlying surface, #,d is the backscattering coefficient of a solid particle of interest, Paq is the backscattering coefficient of an aqueous particle of interest, -rd is the optical depth of a solid particle of interest, and Taq is the optical depth of an aqueous particle of interest. This equation assumes the aerosol is a purely scattering particle (i.e. no absorption of solar radiation) and is optically thin (i.e. - << 1), 17,62 28 which are appropriate assumptions for the sulfate aerosol species present in the atmosphere.' 6 A derivation of Eq. (11) can be found in Seinfeld and Pandis.16 Eq. (11) is a simplified one box model representation of the atmosphere. In this one box representation, a single aerosol layer is assumed through which the net flux is determined by using globally and temporally averaged parameter values present in Eq. (11). Because a more rigorous calculation of DRF which incorporates speciated hysteresis behavior, as was done in Wang et al.,17 is outside the scope of this thesis, it is assumed that this one box model approach is first order accurate. As described in Section 2.2, transition between solid and aqueous states for sulfate aerosols is governed by a hysteresis cycle,' 7 where the relative humidity (RH) history of a particle is related to the hygroscopic growth that occurs. No hysteresis loop behavior is assumed in the GEOS-Chem simulations and sulfate aerosol particles are assumed to always 63,64 Several studies have attempted to quantify the impact of sulfate hysteresis behavior on sulfate aerosol RF.17,64,65 be on the upper hysteresis branch. can be estimated based on a particle's asymmetry factor, g, where g is an intensity-weighted average of the cosine of the scattering angle 16,66and is also a function of RH.65,67 Wiscombe and Grams66 estimate the average value of f to be the following: -- g Eq. (12) where, g is a function of the size of the particle, thus fl (overbar denotes time average) is a function of RH given the hygroscopic growth that occurs due to water condensation. Aerosol optical depth (AOD) values are obtained from GEOS-Chem simulations for the background (not including aviation), baseline aviation (background + aviation with standard jet fuel), and ULSJ aviation (background + aviation with ULSJ fuel) cases. These AOD values, however, are presented for a 400 nm wavelength of incoming solar radiation. Wang et al. 17 evaluates AODs at 550 nm, "a wavelength that is representative of the mean across the solar spectrum." In general, RFs are calculated by taking a weighted average over the entire solar radiation spectrum as aerosol optical properties are wavelength dependent.6 7 A simplified weighted RF calculation is described in Nemesure et al.67 For this analysis, however, AODs at just 550 nm are computed to avoid significant computation times and are used to determine sulfate DRF. The version of GEOS-Chem used in this analysis cannot compute AODs at a specified wavelength other than at the default wavelength of 400 nm, although a recently 29 developed post-processing module, FlexAOD, 68 does have this functionality. FlexAOD can also compute asymmetry factors, which can be used to determine the backscattering coefficient based on Eq. (12). Using FlexAOD, sulfate aerosol AODs are recomputed at 550 nm and used in Eq. (11). As mentioned in Section 2.2, the sulfate species bin within GEOS-Chem and FlexAOD also includes nitrates and ammonium where no distinction is made in optical properties between the different species. Limited research has been performed on the direct RF impacts of nitrates and ammonium alone, but nitrate contributions to overall aerosol mass is small relative to sulfate when background concentrations are considered and impact on direct RF is uncertain. 69 Within GEOS-Chem, all three of these species are treated identically (i.e. purely scattering), thus Eq. (11) is still applicable. Using this bin to compute RFs also captures the nitrate bounce-back effect and its potential impact on direct climate forcing due to a reduction in atmospheric sulfate concentrations. This lumping, however, assumes that sulfate, nitrate, and ammonium aerosols are identical in their optical properties, which is certainly not the case. This issue is addressed in greater detail in Section 3.2.3. RF values for standard aviation minus the background and ULSJ aviation minus the background are calculated for four regions: global, northern hemisphere, Europe, and Asia. Values for all anthropogenic and biogenic sources of sulfate are also calculated. These RF values are area weighted to reflect the differences in grid box size given that GEOS-Chem uses a uniform polar grid (40 x 50). 3.2.2. RF Uncertainty The IPCC Third Assessment Report70 provides uncertainty values and ranges (based on Penner et al. 71) for all of the coefficients in Eq. (11). The minimum and maximum values provided in the paper are used as bounds for a triangular distribution. No uncertainty estimate was provided for FT. Table 6: Coefficients in Eq. (11) and associated values and uncertainties. Coefficient 7 1 -Ac (1 - Rs)2 Nominal Value 0.58 0.39 0.72 30 Uncertainty Range 0.4 - 0.83 0.35-0.44 0.65 - 0.8 Wiscombe and Grams66 provide a high, low, and mean value for the backscattering coefficient, f. The mean value is previously shown in Eq. (12). The upper and lower bounds can be estimated as the following: #high - g Eq. (13) Aw - - g Eq. (14) = Uncertainty in the optical depth values from GEOS-Chem is not easily obtained due to the inherent complexity that exists in the model. Rather than attempt to determine how uncertainty propagates through the model based on initial uncertainties in the input data, the uncertainty for optical depth was determined by survey. IPCC Fourth Assessment Report72 provides sulfate aerosol optical depth values across nine different models that used identical input emissions. These nine values are used to define an uncertainty factor for the optical depth values. This approach also captures uncertainty related to atmospheric processing and removal of SOx emissions. 3.2.3. Sulfate Lumping Again, aerosols are assumed to be internally mixed, thus different mixing fractions of sulfate, nitrate, and ammonium are not considered. The potential effect of different mixing fractions, however, is approximated in this section. In this case, differences in hysteresis behavior are not considered as this is rather complex. Instead, the focus is placed on the magnitude of the backscattering coefficient for a given particle. Wang et al. provides estimates of the backscattering coefficient given differences in the refractive index and particle size. It can be seen that if a composition of purely ammonium sulfate is assumed as compared to sulfuric acid (which is the assumed composition of sulfates in GEOS-Chem), the change in the backscattering coefficient would increase results by about 13%. Similarly, Jarzembski et al.73 shows that for short-wave radiation, the calculated backscattering coefficients for ammonium sulfate versus ammonium nitrate are nearly equal, assuming ammonium nitrate is representative of the nitrate aerosol present in the atmosphere. Given that variations in the calculated RFs are -10%, the lumping was assumed to be reasonable given the large aforementioned uncertainty in premature mortality calculation and monetization as well as climate impact monetization. 31 3.2.4. WTW GHG Emissions 61 As a part of their analysis of alternative jet fuels, Stratton et al. performed a life cycle green house gas (GHG) emissions analysis of standard jet fuel and ULSJ using the Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) framework developed by Argonne National Laboratories. Their approach used a weighted average of GHG emissions from all potential crude oil sources (12 different countries/regions) feeding into US refineries, i.e. just the extraction and raw material transportation aspects of the cycle. The baseline case was further defined by the assumptions made concerning the refining efficiency. It was assumed that the refining energy efficiency of conventional jet fuel is 93.5% (i.e. MJ of fuel for a unit quantity of jet fuel/MJ input to refinery). A life cycle analysis of standard jet fuel using 2005 production data yielded a total WTW GHG emissions value of 87.5 gCO 2e/MJ. This value includes the extraction and transportation emissions as well as refining and combustion emissions. For the corresponding baseline case, ULSJ differs from conventional jet fuel in terms of its life cycle analysis in two important ways. First, from a 2001 General Motors study, 74 a 2% energy penalty is assumed for reducing the sulfur content in diesel fuel from 350 to 5 ppm, i.e. 2% more energy is required during the processing and refining stage of the life cycle given the additional HDS of the diesel fuel that is required. Given the similarities between diesel and jet fuel, this same 2% penalty assumption was extended to the case of ULSJ. This penalty was seen in the refining energy efficiency, which was reduced to 91.5% for ULSJ. Second, there is an expected 0.4% decrease in combustion C02 emitted per unit of fuel energy due to a change in the hydrogen-carbon ratio as a result of the additional hydroprocessing. Both of these factors were accounted for in the life cycle analysis for ULSJ and yield a WTW GHG emissions total of 89.1 gCO 2e/MJ. Thus, this 2% assumed reduction in refining efficiency caused a 2% increase in WTW GHG emissions was seen between the baseline cases for conventional jet fuel and ULSJ. A high and low value of 90.7 and 87.5 gCO 2e/MJ are assumed as the uncertainty range in this analysis. Given that regional jet fuel data for the US shows an average FSC of 600-700 ppm and the ULSD study that is the basis of this work used a FSC of 350 ppm, it can then be expected that the assumed energy penalty could be as great as 4%. As determined in the baseline emissions scenario, a 2% energy penalty yielded an 89.1 gCO 2e/MJ WTW GHG lifecycle emissions value. By extrapolating this expected change, it then follows that a 4% energy penalty yields an additional 1.6 gCO 2e/MJ, or 90.7 gCO 2e/MJ. The low value is equal to the 32 WTW GHG emissions value of conventional jet fuel where the FSC of the inputted crude oil is assumed to be less than 15 ppm, thus no additional processing would be required. A higher baseline energy penalty could be assumed given the higher FSC of jet fuel on average compared to diesel fuel. There is, however, considerable uncertainty in what energy penalty will be seen in ULSJ production because the chemical make-up of jet fuel (as outlined in Hileman et al. 75) in general has simpler hydrocarbon structures than diesel fuel. Thus the energy input required to desulfurize jet fuel from 350 to 5 ppm is potentially less than the energy input required to desulfurize diesel fuel the same amount. 3.2.5. APMT-Impacts Climate Module As part of the aviation environmental portfolio management tool (APMT) project that focuses on quantifying and valuing the environmental effects of aviation activity, a framework in which climate impacts are assessed in a computationally inexpensive manner was initially developed 76 77 by Marais et al. and Mahashabde et al. This section will briefly describe the overall structure of the APMT-Impacts Climate module and explain the relevant portions of code important to this analysis. The APMT-Impacts Climate Module (from here on referenced as "APMT-Climate") is used to value climate impacts on a global scale in this analysis. APMT-Climate takes as inputs fuel burn, C02, and NOx emissions. Climate impacts due to a variety of species and effects, including sulfate aerosol cooling, are derived or scaled from these inputs. Sulfate aerosols are designated "short-lived" effects, i.e. effects that are scaled directly from fuel burn and whose RF effects are assumed only to last the year in which it is emitted. For this climate analysis, only two climate impacts are considered: reduction in sulfate cooling due to FSC reduction and an increase in CO 2 RF due to the increase in WTW GHG emissions. The RF due to S04 can be calculated using the following relationship: (RF RFso4(t) = RFGEOS-Chem 10, Ashort~J C = emission year t > emission year Eq. (15), where RFGEOS-Chem is the RF value for sulfate aerosols calculated from the GEOS-Chem simulations as outlined in Section 3.2.1, A is the climate efficacy value for the short-lived and C02 effect, and t is a time variable in integer years. The climate efficacy relates the proportional change between the RF of the given species and the resulting temperature response of the 33 system. Given the uncertainty in these estimates, this analysis assumes all efficacy values to be one, i.e. each effect produces the same proportional response in temperature given a unit RF input to the system. The increase in C02 emissions is reflected in the emissions inputs required by the code, as mentioned previously. Although these are not direct aviation emissions, the method by which the C02 RF value is determined is based on total atmospheric C02 concentrations, and thus the source of the C02 is inconsequential. APMT-Climate deals with C02 effects by determining the overall change in concentration given the input emission index based on an impulse-response function and then integrating the product of these values over time. A logarithmic relationship is then used to determine the RF of C02 in the atmosphere for the study year relative to preindustrial C02 levels. Further details are provided in Marais et al. 76 7 and Mahashabde et al.77 Once RFs have been computed, surface temperature changes are calculated by using a simplified heat transfer model.78 After the induced temperature changes are calculated, these values are then passed to a damage function that computes the impact on global GDP, as is the case with the DICE-2007 model. 79 Although RF effects for short-lived terms are only considered for the year of emission as atmospheric lifetimes of these species are assumed to be short, the temperature effects can last multiple years resulting from the transient heat model, thus the total damage is provided as a net present value (NPV) of damages taken in relation to the base year of 2006. This analysis is limited to a one-year pulse emission scenario for the policy (ULSJ) and baseline (standard jet fuel) cases. As mentioned previously, RF effects for sulfate aerosols last only in the year they are emitted, but C02 RF impacts are tracked into the future due to the lifetime of the species. Temperature effects are also tracked into the future for all species given the analytical heat transfer model. A time horizon of 800 years is assumed. 3.3. Country Dependent VSLs The EPA suggests the use of the value of a statistical life (VSL) as a means to value avoided premature mortalities when conducting benefit-cost analyses (BCA). 80 Many studies have explored VSL values in the United States and other relatively high income countries, 81,82,83 but uncertainty remains in determining how to apply VSLs from higher income countries to lower 34 income countries in order to provide an appropriate estimate in these countries where no VSL estimates have been made. VSL is constructed from a person's willingness to pay (WTP) for an arbitrarily small but finite reduction in risk. Wage-risk studies estimate WTP by comparing an individual's perceived risk within a certain type of employment versus the amount of compensation the individual receives, i.e. wage. Beyond higher income countries, few credible VSL estimates exist. Miller 8 ' provided estimates for 49 countries, including several low income countries. For this analysis, it is necessary to extrapolate a VSL from one country and apply it to another country in which no estimate has been made. From Hammitt and Robinson,83 who conducted a study on applying US VSL specifically to sub-Saharan Africa, the following relationship can be used to extend a VSL from one country to another. VSLB = VSLA - ()IE Eq. (16), where A and B denote the base country and country of interest, respectively, VSL are the VSLs for the respective countries, / is a measure of income for each country, and IE is the income elasticity associated with VSL. This method requires the selection of a base VSL. For the purposes of this thesis, the US is used as the base country with a VSL of $7.4 million in 2006 dollars as suggested by the EPA.80 This value is derived from 26 US VSL studies where values have been adjusted for inflation. A Weibull distribution was then fitted to the data with a scale parameter of 7.75 and a shape parameter of 1.51.80 Hammitt and Robinsons3 recommend using gross national income (GNI) per capita as an income measure for each country. The major source of uncertainty lies in the value of IE used. IE is a reflection of the proportion of an individual's income that is used towards risk reduction. Within higher income countries, it has been shown that lEs less than one are appropriate,83 meaning that an increase in income level does not cause a proportional increase in VSL. When performing cross-country comparisons where there is a large discrepancy in income level, however, lEs greater than one are plausible given that as the average income of a person is reduced, a reduction in the proportion of income used towards risk reduction follows. What this value for IE should be, however, remains highly uncertain. Hammitt and Robinson83 suggest applying a range of lEs from 1 to 2, where a uniform distribution is assumed in this analysis. 35 Hammitt and Robinson8 3 also suggest comparing calculated VSL values to the expected future earnings and consumption so as not to undervalue VSLs in low income countries since VSLs should be at least equal to the net present value of future earnings lost due to premature mortality. Hammitt and Robinson 83 make an estimate of expected future earnings by taking the NPV of unadjusted GNI per capita for half the expected lifetime for a person in the country of interest assuming a 3% discount rate, where NPV was previously defined in Eq. (2). For this analysis, the cash flow for the period is defined as the GNI per capita for the base year of 2006 for each country. Where appropriate (i.e. when VSL falls below the NPV of future earnings), this value is substituted for the VSL. Valuing lives across countries, however, has ethical implications in that it may be interpreted as a life being more highly valued in higher income countries than in lower income countries. VSLs are rather a reflection of an individual's willingness to reduce his or her own risk in premature mortality subject to the economic constraints present in that country. 83 Policymakers may object on moral grounds to a variable VSL approach. As such, the Department of Transportation (DOT) sets guidelines so that all individuals in the US are valued equally, as previously discussed in Section 2.4.4. This idea can also be extended across countries, i.e. assume a constant VSL valuation for all avoided mortalities as a result of ULSJ implementation where this may be viewed as a policy choice rather than a concept strongly supported by economic theory. Because no "global" VSL exists, a valuation using a constant US VSL is provided. The benefit of ULSJ implementation is valued by multiplying the number of avoided mortalities for a given country by that country's corresponding VSL and summing while also taking into account discounted health benefits in the future. The standard mortality lag structure as defined in Section 2.4.3 is used in this analysis. Non-discounted costs are also presented in Error! Reference source not found. as a comparison. 3.4. Concentration Response Functions For this analysis, the EPA defined CRF is extrapolated to other countries in order to evaluate global health impacts of ULSJ implementation, but several assumptions are required. First, a linear CRF is applied across all pollutant concentration levels. Second, it is assumed that all avoided mortalities that result from ULSJ implementation will be seen in a reduction in cardiopulmonary disease and lung cancer related deaths. Third, consideration is limited to 36 members of the population with an age greater than 30 as the original cohort studies focused their analysis to this age bracket. The CRF then has the following form: A(Premature Mortalities) = Zj[fl fk,30+ PijAxijBcP + fp/Cfk, 3 0+PijAxijBkc] Eq. (17), where k denotes the country of interest, which is a function of the grid cell location, ij (i.e. k = k(ij)), fk,30+ is the fraction of the population above 30 years of age in a specific country, B is the disease specific baseline per capita mortality rate in a specific country, CP denotes values in terms of cardiopulmonary disease, LC denotes values in terms of lung cancer, # is the fractional increase in mortality given one ptg/m 3 increase in annual average PM2.5 (i.e. risk coefficient) and is a function of the disease of interest, Pij is the total population in grid cell, ij, and AXij is the change in PM 2.5 concentration in a grid cell, ij with units of Ig/m3 . In Eq. (17), the summation symbol and indices, ij, show that in order to obtain the total avoided mortalities, it is necessary to sum across all grid points defined by the gridded population data, GRUMP, 84 which has a finer grid resolution (2.5' x 2.5') than global GEOS-Chem (40 x 5*). Each population grid cell is assigned to the closest corresponding GEOS-Chem grid cell (by grid center point) to determine population/concentration products as required by Eq. (17). As stated in Section 2.3, only CP and LC related premature mortalities are considered. The change in the number of premature mortalities is split between both disease groups with two different # values, which are assumed not to change across countries and are based on epidemiological data for the US. Because it is not appropriate to apply the US all-cause risk coefficient globally given that countries in different parts of the world can have very different mortality profiles, it is necessary to scale the disease-specific risk coefficients appropriately. Under the assumption that CP and LC premature mortalities dominate and comprise all premature mortalities seen by a change in ground-level PM2.5 concentration, the diseasespecific risk coefficients within the US are related to the all-cause (AC) values by # B= BC #3lBfc + #!%B&c Eq. (18), where flcP and fls are unknown. Baseline incidence rates for the US are computed using data from the WHO Global Burden of Disease (GBD)85 database and flC is assumed to be the distribution of values obtained through the EPA expert elicitation study,23 which is centered on a 1.06% increase in premature mortality given a 1 pg/m 3 increase in ground-level PM2.5 concentration. This central estimate is based on the EPA recommended Weibull distribution, a 37 distribution based on the ACS and Six Cities studies.20, This CRF approach will be here-on referred to as the "EPA CRF." The number of CP related deaths is much greater than LC related deaths in the US, and the CP risk coefficient is known with more certainty. As a result, Eq. (18) can be defined in terms of flI and a ratio between the uncertain disease specific relative risks, y. Eq. (18) then becomes flAC = fl[BP + yBL] = (RRLC-1) RR LC Eq. (19), where (RRCP-1) Eq. (20), RRCP and RR is the ratio between the number of health incidences in the baseline pollution case to the number of health incidences when only background pollution is considered. (RR-1)/RR is then the percentage change in mortality given a change in ground-level PM2.5 concentration. Rearranging Eq. (19) produces the following equation for the risk coefficient. #s =L[C us[BUS+yBbU] Eq. (21). To solve for flcP, an appropriate value of y must be determined. Table 7 shows the central estimates for these RR, adjusted to percent increase per pg/m 3 increase in PM 2.5 . Table 7: Percentage increase in avoided mortalities given a 1 pg/m 3 increase in ground-level PM 2.5 concentration. Values from Pope et al.20 and Laden et al. Pope Laden All Cause 0.6 1.4 Cardiopulmonary Lung Cancer 0.8 1.2 2.2 2.1 Based on the values in Table 7, if each RR value is allowed to vary within the specified range above, y can range from 0.5 (1.2/2.2) to 2.6 (2.1/0.8) where a uniform distribution is assumed for each RR. Eqs. (9) and (12) yield the following result: A(Premature Mortalities) = Z'ijfk,30+ PijAXi j f CP(Bf + yBLc)] Eq. (22). As a comparison to the EPA CRF approach, the WHO methodology described by Ostro86 was also used and it was applied by Barrett et al.87 to determine the number of mortalities that result from full-flight operations by aircraft. This function has the following form: 38 RRk =kXB+1) Premature Mortalities = Ek Eq. (23) RRk BkPk Eq. (24), where XA and XB are the concentrations for the policy and baseline cases, which for this study would be ULSJ aviation and standard aviation, respectively, fl is a disease specific power coefficient, Bk is the baseline incidence rate for a disease category, Pk is the exposed population, and k corresponds to a given country and total mortalities are determined by summing across all countries. Thus, this methodology uses background concentrations compared across policy cases in order to determine a relative risk per country (RRk), which is then related to a percent increase in premature mortality given some change in PM 2 .5 concentration. This method results in a lower marginal risk at higher concentrations. Baseline incidence rates are a function of grid cell location (i.e. the country that coincides with that grid cell) and are determined using the WHO GBD database, which provides cause specific mortality information bracketed by age group for each country. Given that the ACS and Harvard Six Cities cohort studies focused on populations 30 and 25 years and older, respectively, country specific mortality data is required specifically for the 30 years and older (30+) age bracket. WHO GBD data, however, are provided only for the 15+ bracket. As a first approximation, it is assumed that no CP and LC deaths occur in the 15-30 age bracket, and that the mortality data provided for the 15+ bracket can be exactly applied to the 30+ bracket. Relative uncertainties for all cause mortality rates are also provided for each country. Applying these uncertainties to cause-specific deaths underestimates the uncertainty for CP and LC related deaths, but no cause-specific relative uncertainties are provided. 30+ population data for each country are obtained using the US Census IDB. 88 Given that similar data is used to determine the WHO GBD database values as are used to perform the population projections in the US Census IDB, the relative uncertainties from the GBD database are also applied to the population projections. 3.5. Economic Analysis Two separate cost analyses are performed. The first uses historical price data from ULS diesel to estimate ULSJ costs while the second relies on cost-curve estimation based on petroleum refining data. 39 3.5.1. Price History Analysis This section includes significant contributions from Dr. Robert Malina and Dr. Christoph Wollersheim. The only cost aspect considered in the price history analysis is the expected increase in price due to the additional processing required to desulfurize jet fuel. For this analysis, it is assumed that any price increase for ULS production is a function of increased capital and refining costs and not a function of any other market factors that may be relevant. This price analysis is based on price history data of ULS diesel as it has recently been implemented for on-road use in 2006 and is currently being phased into off-road use (ships, locomotives, etc.) as detailed previously in Section 2.5. The US Energy Information Administration (EIA)89 provides price history data for three diesel fuel types: high sulfur (HS) (500+ ppm), low sulfur (LS) (15-500 ppm), and ultra low sulfur (ULS) (<15 ppm). The price differential between ULS/LS and LS/HS are plotted against the amount of ULS/LS and HS diesel fuel supplied, and is shown in Figure 3. Note that throughout this section, references to HS, LS, and ULS refer to diesel fuel. The black and red lines show the price differentials and correspond to the right-hand axis, while the purple, green, blue, and orange lines show the product supplied and correspond to the lefthand axis. There appears to be a spike in the price differential beginning in 2005 and ending in early 2008. This spike coincides with a change in supply of ULS and LS fuel, which are assumed to be direct substitutes given that a decrease in LS supply is accompanied by an increase in ULS supply, while the total amount of ULS and LS supplied remains approximately constant over the entire time period. This price spike also coincides with the initial phase-in of ULS diesel in 2006 shown in the timeline of standards for on-road implementation provided in Table 2. The cause of this price spike is unclear. It could be a result of market fluctuations given the shift in supply away from LS to ULS. This shift in supply, however, may be exaggerated in Figure 3 given that ULS diesel in Jan. 2005 may already have been produced in non-negligible quantities. No ULS diesel was reported due to the fact that it was not required to be labeled as ULS diesel by regulations at this point. 40 Product Supplied and Price Differential 140000 0.3 120000 - 0.25 - 015 80000 - -- * - L Supplied LS Supplied 2 160000 -- 0.05 a. 0 40000 - 60000 ULS + LS Supplied 0 2-0.05 -- HS Supplied -ULS-HS 0 r- r-I o o 4 N M 0000 < Lfn '0 r_ r- < , 0 0 , o 0 0 00 0 0) 0 < 0 -'- 0 -0.1 - Price Duff. LS-HS Price Diff. 0 Date Figure 3: The product supplied of ULS, LS, and HS diesel fuel plotted simultaneously with the price differential for ULS-HS and LS-HS for Jan 2001 to February 2011. All price history data is condensed into a single representative price differential. For this analysis, the observed price differential in ULS and HS diesel fuel (after adjusting the nominal prices by inflation to the real prices for a base year of 2006) is weighted against the amount of ULS and HS diesel supplied for a given month in order to capture the interaction between price and quantity within the fuel market. HS diesel fuel is defined by a FSC of greater than 500 ppm, and because jet fuel FSC is between 600 and 700 ppm, the ULS/HS price differentials are used rather than the ULS/LS price differentials. One limitation of this method is weighting against negligible amounts of HS fuel (which would drive price differentials downward) due to a phasing out of HS diesel fuel for non-road applications beginning in January 2007. Figure 3 shows a decline in HS diesel production starting around this time, but its production is currently nonnegligible and thus not an issue in this analysis. Three different weighted estimates are determined by considering three distinct time periods. The first time period considered is the "steady state" period in which the price spike stabilizes relative to the noise already present in the data. This period is set from January 2008 to February 2011 (where the end date was based on the most up-to-date data available given the time the price analysis was conducted). The second time period includes all ULS price data, which is only reported from January 2007 and onward although ULS production numbers are 41 provided starting in January 2004. Not all relevant price data is available (i.e. price data for ULS since introduction of ULS production), for the entire price spike period. For this reason, a third period is analyzed from January 2005 to the present where the LS/HS price differential is used as a substitute for the ULS/HS price differential from 2005-2007 when ULS price data was unavailable. Regardless, all price differentials for this third price scenario are weighted against ULS and HS production quantities. For the three scenarios described above, the following weighted averages have been calculated: 3.7 cents for the steady state period (low price differential), 5.6 cents for all ULS price data, and 6.5 cents for all price data (high price differential). 3.5.2. Cost Buildup Approach This section includes an analysis originally performed by Dr. Jim Hileman, Russ Stratton, and Matt Pearlson and then supplemented by the thesis author. An alternative approach to estimating production costs is used to corroborate the price differentials determined above. The analysis includes estimating the capital costs for the hydrotreater unit, the steam methane reformer (SMR) unit, and the natural gas (NG) feedstock costs for a representative refinery. In hydrodesulfurization (HDS), or more generally hydroprocessing, hydrogen gas (H2 ) is used to desulfurize jet fuel. Steam methane reformation is used to create H2 at the expense of NG consumption. The additional NG requirement for HDS is determined from the GREET model90 as follows. GREET is used to calculate the differences in NG consumption (in units of energy of NG consumed per energy of ULSJ produced) based on the refining efficiencies and process shares defined for ULSJ and conventional jet fuel. Refining efficiency for a specific petroleum product is defined as 7 EF EF EF+E1 Eq. (25), where EF is the specific energy of the fuel and Er is energy input to the refinery per unit mass of jet fuel produced. The total change in input specific energy between standard jet fuel and ULSJ is required for this analysis, which is defined as A = E1,u - E1,1 42 Eq. (26). U indicates values for ULSJ, and J indicates values for standard jet fuel. Inverting Eq. (25) and taking the difference between the standard jet fuel and ULSJ cases produces the following relationship: 1 1 EFU+EIU EF,J+EJ f7u 77j EF,U EF,J Eq. (27). If the specific energy of standard jet fuel and ULSJ are assumed to be approximately equal (within 0.3% as outlined in Section 3.2.4), then Eq. (27) reduces to 1 lU E,-Eg 7j Eq. (28). - EF EF To determine the change in energy associated with additional NG consumption per mass of jet fuel, it is necessary to multiply Eq. (28) by the appropriate process energy share. ANG = (- - EF 'fNG Eq. (29), where fNG is the total process energy share associated with NG use within the refinery. To determine the additional amount of NG required, Eq. (29) must be multiplied by the density of jet fuel, where energy density is equal to the product of specific energy and density of the jet fuel as reflected in Eq. (30). This expression is then divided by the energy density of NG to acquire the additional volume of NG required (at standard conditions) per unit volume of jet fuel. ANGvolume = where DF -1 - -)-DF -f DN Eq. (30), is the energy density of jet fuel, and DNG is the energy density of natural gas. 61 The process energy shares assumed in Stratton et al. (see Table 8 below) are used for this analysis. The process fuels shown in Table 8 are used to operate the SMR to produce H2 for HDS and also power the hydroprocessing itself. Use of natural gas and refinery gas in refining processes accounts for 80.9% of the total energy consumption, thus the costs due to electricity, coke, and residual oil consumption are not considered. The OECD defines refinery gas as a "non-condensable gas obtained during distillation of crude oil or treatment of oil products (e.g. cracking) in refineries. It consists mainly of hydrogen, methane, ethane and olefins. It includes gases which are returned from the petrochemical industry." 43 Table 8: Process energy shares required for jet fuel production. 6 1 Type of process fuel Process energy share (%) 3.5 41.3 39.6 14.3 1.3 100 Electricity Natural Gas Refinery Gas Coke Residual Oil Total The energy density of NG is 983 BTU/ft3, or 3.66x104 kJ/m 3 (lower heating value (LHV), GREET v1.8A). The energy density of ULSJ is 34.3 MJ/L. 8 From Stratton et al., 6' the assumed jet fuel refinery efficiency is 93.5% and ULSJ refinery efficiency is assumed to be 91.5% due to the energy penalty described with the WTW GHG emissions calculation. Using Eq. (30) and the above stated values, the amount of refinery and natural gas required for the additional hydroprocessing required to desulfurize jet fuel is 2.37 standard cubic feet (scf) per gallon, or 0.018 standard cubic meters (scm) per liter. The NG price history is plotted in Figure 4 where the price shown is the cost for 2.37 standard cubic feet (scf), the amount of NG computed previously that is required for HDS for each gallon of jet fuel. Because refinery gas is created and used within the refineries, there is no market for refinery gas from which to obtain a per volume price. Thus, refinery gas is assumed to have the same cost as natural gas given they are used for the same purpose within the refinery. 0.03 I -. 12 0.025 810 0.02 a ii: - 0 00 C 1000 0.01 ULSJ. 0 (Nf C ~ 0 C C C C 0 C 4 C C C - C Figure 4: NG Price history8 required for ULSJ hydroprocessing where prices are presented based on 2.37 scf/gal (0.018 scm/L) of NG per gallon of ULSJ. Capital expansion of hydrotreating and steam methane reformation capacity is required to handle the increase in ULSJ production. Here it is assumed that the 2.37 scf of natural gas 44 calculated previously is used only for the H2 production required for hydroprocessing. Cost curves presented in Gary and Handwerk 200791 are used to derive the additional capital costs for increasing the hydrotreater and SMR unit capacities for ULSJ production. Capacity is varied from 5,000 to 50,000 barrels per day (bpd), or 800,000 to 8,000,000 liters per day for the hydrotreater and a corresponding SMR capacity is determined from the per gallon requirement of NG. Steam methane reformation is assumed to yield 3.8591 moles of H2 per mole of NG where a water gas shift reaction is assumed; therefore the product of hydrotreating capacity and the required NG/gallon is multiplied by 3.85 to obtain the total additional amount of H2 required to treat a gallon of ULSJ fuel. The cost curves of the hydrotreater and the corresponding SMR are aggregated and plotted in Figure 5 (blue line). For example, the total capital cost associated with steam methane reformation and hydrotreating for a capacity of 22,000 bpd of ULSJ is $58 million. Straight-line depreciation is then applied to these capital costs. The US mandates a 10-year depreciation time horizon for tax purposes,92 but capital costs in this analysis are depreciated over 30 years'l to reflect the useful lifetime of an average refinery to yield a per gallon capital cost (red line). For 22,000 bpd of capacity, the additional per-gallon cost for increasing the hydrotreating capacity to accommodate full scale ULSJ production is $0.017 per gallon when a 10 year depreciation time horizon is considered, but $0.006 per gallon when depreciated over 30 years. 0.030 120 , 0 0.025 .100 80 0.020 L860 2 0.015 40 0.010 20 0.005 ' 0.000 0 -. 0 10,000 20,000 30,000 40,000 50,000 Hydrotreating Capacity (bpd) Figure 5: Capital costs for hydrotreating and SMR units as a function of HDS capacity depreciated over 30 years. 45 The average capacity of ULS fuel at US refineries, according to the EIA, is approximately 22,000 bpd, where this value was obtained by dividing the total amount of ULS diesel produced in 2009 (3.1 million bpd) by the total number of US refineries producing ULSD (141). For this representative capacity, the per gallon additional capital cost ranges from 0.004 and 0.006 $/gal. A more rigorous cost buildup approach would involve estimating the additional per gallon cost of ULSJ using cash flows that includes capital cost factors such as loan payments, depreciation over 10 years, escalation from 2005 costs (the year in which this analysis is based), a change in location factor seeing as capital costs would rise outside of the US Gulf Coast, a compounding of increased capital costs due to loan payments and depreciation/capital recovery, inclusion of fixed operating expenses (additional staff, insurance, maintenance, etc.), additional supporting utilities, and discounted cash flows for the "true value of money" for an economic investment. This approach will result in a higher cent/gal cost than estimated above. 3.5.3. Cost Distribution From the price history analysis, minimum and maximum price differentials of 3.7 and 6.6 cents/gallon are calculated. From the cost-buildup approach, minimum and maximum price differentials of 1.6 and 3.6 cents/gallon are calculated. While the cost-buildup approach is useful in validating the price history analysis, it provides a minimum price differential as it only considers the additional methane required for H2 production for hydroprocessing and additional refinery processes that may use the methane as a fuel. Other costs that the refineries see as a result of ULSJ implementation that may be hidden within the price history analysis are not captured. Thus, 1.6 cents/gallon is taken as the minimum price differential, while the 6.6 cents/gallon is taken as the maximum price differential. 3.7 cents/gallon is chosen as the nominal price differential given it is the expected price differential when production reaches a steady state as defined in Section 3.5.1. A triangular distribution is assumed, which is described in Section 4.4.2. 3.6. Sensitivity Analysis 3.6.1. Monte Carlo Analysis Framework Monte Carlo techniques are used to quantify the uncertainty present within the EBCA. From Allaire and Willcox, 9 3 within a general model, f(x), with an arbitrary number of input parameters, the expected outcome can be determined from a Monte Carlo simulation with the form 46 1E =1f(xm) -+ IE[f(x)] as N -> oo Eq. (31), where x denotes a vector of input parameters. Eq. (31) states that as the number of Monte Carlo simulations, N, goes to infinity, then the mean value of all simulations will approach the expected outcome of the model system. For each simulation, the input parameters are randomly selected based on distributions assigned to each variable. The outputs presented in each of the results section have all been produced using Eq. (31) unless otherwise noted. Given computational time constraints, N is chosen to be 2000. Due to the complexity of GEOS-Chem, it is not practical to perform 2000 air quality simulations. Rather, a 60% uncertainty is assumed for ground-level concentrations and these results are scaled for each Monte Carlo (MC) simulation based on a triangular uncertainty distribution. 3.6.2. Nominal Range Sensitivity Analysis A nominal range sensitivity analysis (NRSA) as detailed in Jun 94 is used. A NRSA is a local, first order sensitivity analysis that is used for a deterministic model and shows changes in the final output given these deterministic inputs. Each input is varied from a nominally low to high value as it is inputted into the deterministic model while all other parameters not being tested are held at their modal values. This type of analysis does not capture any interaction sensitivities and is most effective for linear systems. The high, low, and nominal values are defined in Table 14, except for the US VSL, which for the purposes of this analysis, is assumed to have a high, low, and nominal value of $12 million, $1million, and $7.4 million, respectively. Also, nominal values for the uniform distributions are assumed to be the value midway between the defined endpoints. 3.6.3. Global Sensitivity Analysis While the results of an NRSA may be useful to understand first order and absolute effects of input parameter values on the output value, it provides no information concerning how much uncertainty each parameter contributes to the total output uncertainty. A global sensitivity analysis (GSA) serves to quantify the contribution to variance. GSA is detailed in Allaire and Willcox, 9 3 but the method used for this analysis is implemented as specified by Salteli. 95 A description outlining this approach is provided here. Three matrices are defined with random variables: A, B, and C. A is an N x k matrix that contains a set of randomly generated variables, where N is the number of MC simulations and k 47 is the number of input parameters. B is an N x k matrix that contains a different set of randomly generated variables from those in A. C is an N x k matrix that is formed by all columns of B except for the fh column, which is replaced by the /h column of A. Based on these constructed matrices, Monte Carlo simulations are performed where each column of each matrix is one simulation defined by randomly defined input parameters and each unique matrix defines one complete Monte Carlo run. Thus, because there are k + 2 unique matrices formed, a total of Nx(k + 2) runs are required. From these simulations, the variance can be computed as Var(Y) = Var[Y(A); Y(B)] Eq. (32), meaning the expected variance of the output is defined by both set of randomly generated variables and Y is the vector of expected outcomes generated by the simulations. The main effect index can be computed as Si = 1Z[Y(A) - Y(Ci) - Y(A) - Y(B)]/Var(Y) Eq. (33), where the multiplication shown above is scalar component-wise multiplication which gives the fraction of output variance explained by that input parameter alone, but does not include interaction effects between parameters.93 The total effect sensitivity index is defined as STi = 1 - N= 1 [Y(B) - Y(Ci) - Y(A) - Y(B)]/Var(Y) Eq. (34), which does account for input interactions.93 The results from a GSA can indicate which input parameters require further research and understanding to reduce overall uncertainty in a model or associated analysis. To improve convergence times and reduce the value of N, Salteli9 5 suggests the use of Sobol quasi-random numbers instead of completely random variables. Sobol sets are used in this analysis. 3.7. Additional Operational Concerns 3.7.1. Change in Fuel Properties One of the operational concerns of ULSJ is the impact on fuel energy density and specific energy given the additional processing required for desulfurizing jet fuel. From Hileman et al.,75 a 1% reduction in energy density and a 0.3% increase in specific energy of the fuel are expected post-processing of a fuel with an average FSC of 600-700 ppm. These changes in fuel properties are most likely due to the breakdown of aromatic rings that constitute approximately 48 20% of jet fuel. Thus, more fuel will be burned by volume, but less by mass. This has two potential consequences. First, if more fuel is required to be burned by volume, then it is possible that airlines will have to purchase more fuel by the gallon at a given price. The impact of this effect is unclear as market adjustments could take place given that the consumers know the fuel has reduced energy density, and thus airlines may not incur any cost penalty as a result. Second, if less fuel is burned by mass (assuming that the total fleet energy requirement remains the same and there are no airline operational impacts given the increase in fuel volume carried), there is a potential reduction in climate and health impacts due to a reduction in overall emissions, although this is uncertain and not considered in this analysis. The reduction in fuel energy density may cause a higher percentage of fuel to fall below standard jet fuel specifications, which in turn has an effect on profits seen by the refineries given that less fuel can be sold as jet fuel. Based on ASTM turbine fuel specifications for Jet A or A1,96 jet fuel must have a specific energy content of at least 42.8 MJ/kg and a density between 0.775 and 0.840 kg/L, which implies that the lowest possible energy density that is within specification is 33.2 MJ/L. Fuel data for JP-8 was obtained through the Petroleum Quality Information System (PQIS) database.97 Given the similarities between Jet A and JP-8, it is assumed that the fuel specifications mentioned above could also be applied to the PQIS data. The values in the data set are shifted by the expected reduction in energy density. A 1 % energy density reduction corresponds to an energy density of 34.4 MJ/L, thus a 0.4 MJ/L shift is applied to all values in the data set and compared against the minimum fuel specification value of 33.2 MJ/L. No energy density values fall below the fuel specification. As a limiting case, a 2% energy density reduction shift is also applied to all the data points. Of the available data where energy density could be computed, three points fell below the fuel specification, which corresponds to 0.07% of the total fuel volume. It is then assumed that no significant additional costs would be seen by refineries or consumers with regards to meeting fuel specification standards. 3.7.2. Fuel Lubricity One other operational issue is a potential decrease in fuel lubricity. Decrease in fuel lubricity can lead to engine fuel pump failure as these components tend to be partially fuel lubricated. Decreased fuel lubricity can cause more rapid fatigue of the mechanical components due to an increase in wear scar diameter (WSD). An example of this was the full or partial failure of at least eight pumps on New Zealand airlines flights due to poor lubricity fuel delivered from a local 49 New Zealand refinery. Tests indicated excessive wear in the spline-drives, which connect the fuel-pumps to the fuel-control units, from at least three different manufacturers. These splines were expected to have a service life of 3000-5000 hours, but were wearing out in 150 hours. 98 This issue was addressed in several ways. First, the refinery added 5% (although 30% has also been suggested99 ) straight-run (non-hydroprocessed) kerosene to production when possible to hydroprocessed streams which resulted in a reduction of hydrotreater severity. As a result, WSD decreased from 0.78 mm to 0.65 mm from June 1994 to December 1996. Of the airlines affected, one was supplied with DCI-4A (corrosion inhibitor for use in jet fuels) doped fuel, one added a different, unspecified corrosion inhibitor, and one did not use any additive. In addition, hardware modifications were made by the engine manufacturers by offering improved pump splines. There have not been any further reported lubricity issues with this fuel since these changes were implemented. 100 If it were needed, a fuel additive could be used with ULSJ to improve its lubricity. The US military currently uses a Corrosion Inhibitor/Lubricity Improver (CI/LI) additive in all of its JP8 fuel. This additive is obtained through a contract price of $19.706/gallon when purchased in a 55 gallon drum. CI/LI is typically added at 20 mg/L (25 ppm m/m) of JP8. 101 It then follows, on a per gallon volumetric basis, the additional price as a result of the CI/LI additive is 0.05 cents/gallon. Given that this additional price is two orders of magnitude less than the cost of HDS, it is neglected in the benefit-cost analysis. There is also a possible air quality impact due to the additive since it is burned with the fuel during engine operations, but this is unknown. 50 THIS PAGE INTENTIONALLY LEFT BLANK. 51 Chapter 4: EBCA Results This chapter presents the results from the EBCA conducted for global and US implementation of ULSJ. First, results pertaining to the climate and air quality impacts of the policy implementations are presented in Sections 4.1, 4.2, and 4.3. Section 4.4 presents the BCA outcomes when all impacts are monetized and aggregated. Finally, results from the sensitivity analysis are shown to determine the main contributors to the uncertainty present within this EBCA. All assumptions made within this analysis are also provided in this chapter. 4.1. Aerosol RF Results Table 9 shows the calculated RF values for the baseline minus the ULSJ scenario, which corresponds to sulfate aerosol formation from direct aviation SOx emissions. These values correspond to a global implementation of ULSJ. Total background sulfate RF values are also provided in order to further compare against values from the literature. Table 9: Aviation sulfate RF by component and region. RF in m W/m2 Region Global Northern Hemisphere Europe Asia Background (W/m 2) 2.5% Percentile -6.0 -11.2 -15.9 -8.2 -1.44 Average Median -3.4 -6.3 -9.0 -4.7 -0.82 -3.3 -6.1 -8.8 -4.5 -0.80 97.5% Percentile -1.4 -2.6 -3.7 -1.9 -0.35 The total direct radiative forcing for sulfate, nitrate, and ammonium aerosols estimated from this analysis is -0.82 W/m 2 . Results from Kiehl et al. 65 and Wang et al.1 7 are not directly comparable as they only consider direct RF due to sulfate aerosols alone, where the latter only considers anthropogenic aerosols, although direct RF estimations of -0.56 W/m 2 and -0.389 W/m 2, respectively, show that the estimates made in this analysis are on the same of magnitude. Directly comparable results are found in Martin et al.,10 2 which also used GEOS-Chem and reported a direct RF of -0.605 W/m 2 for sulfate, nitrate, and ammonium aerosol.species when biogenic and anthropogenic emissions are both considered under the assumption that aerosols were on upper hysteresis branch. The results from this analysis show a 36% bias when compared to Martin et al. Liao et al.1 03 reports a sulfate aerosol RF of -0.49 W/m 2, but the annually and globally weighted average AOD at 550 nm is reported to be 0.024, while the simulations for this analysis produce an AOD of 0.034. Also, the use of RTMs (which were 52 implemented in all of the aforementioned studies) more accurately models the flux between aerosol layers as well as accounting for the attenuation in solar radiation intensity as light penetrates the atmosphere and interacts with other molecules. The SOx pathway median RF of -3.3 mW/m 2 is about 45% lower than the Lee et al.59 RF of -4.8 mW/m 2 , although this value is captured within the 95 percentile range. The Lee et al. 59 value is based off of a scaling from fuel burn indices relative to a reference value determined in Sausen et al. 104 and differences could be a result of different assumptions in the underlying chemistry or emissions indices. In addition, the above forcing calculation results also take into account the nitrate bounce back effect where a decrease in sulfate aerosol concentration actually causes an increase in nitrate aerosol concentration due to the availability of free ammonia in the atmosphere, which is in general a limiting factor for aerosol formation. Thus, the warming experienced due to a reduction in FSC may be partially mitigated by this bounce back effect. It can also be seen that the change in RF is very regionally dependent. As expected, the change in RF is larger in the northern hemisphere (NH) given that the majority of aircraft emissions occur in this region. Also, the largest RF change is in the European domain, most likely due to the fact that nitrate aerosols are more prevalent in this region and thus a not as strong bounce back effect is observed. 4.2. Mortality Results by Country Avoided mortalities for select countries due to ULSJ implementation are presented in Table 10 for both EPA and WHO CRFs. The WHO values are deterministic. Results are generated from GEOS-Chem and CMAQ simulations performed by Dr. Steve Yim. 53 Table 10: Avoided mortalities by country due to global ULSJ implementation. EPA-derived CRF, Full-Flight Emissions EPA-derived CRF, LTO Emissions WHO CRF Full- LTO FliahtI Country Australia Canada China Egypt Germany India Japan Kenya Saudi Arabia United Kingdom United States of America Total 2.5% Percentile 0.4 2.3 85 15 32 340 5.7 0.5 4 9.7 Mean 0.9 6 220 39 83 870 15 1.5 11 25 97.5% Percentile 1.6 11 390 70 150 1600 27 2.9 22 45 2.5% Percentile 0.1 0.3 -270 1.1 1.8 -93 1.3 0 0.4 2.8 Mean 0.2 0.7 -150 3 4.7 -55 3.4 0 1 7.2 97.5% Percentile 0.3 1.2 -59 5.2 8.4 -20 6.3 0.1 1.9 13 1.4 5.9 72 46 25 390 14 4.2 13 14 0.3 0.8 -13 3.4 1.4 -15 3.6 0.1 1.1 4.2 46 890 120 2300 210 4200 12 -390 31 -130 56 100 140 1500 34 58 Note: positive values indicate avoided mortalities (i.e. saved lives) while negative numbers indicate an increased mortalities (i.e. lives lost) after ULSJ implementation. Avoided mortality numbers from the US nested GEOS-Chem and CMAQ are provided in Table 11. Note that based on the EPA CRF formulation, for the US specifically, the avoided mortalities are independent of y, i.e. the ratio between the disease specific relative risks. Table 11: Regional simulations avoided mortalities results for the US from global implementation of ULSJ. Full-Flight LTO Emissions 2.5% Nested GEOSChem CMAQ Emissions 97.5% 2.5% 97.5% Percentile Mean Percentile Percentile Mean Percentile 56 93 140 230 260 430 18 33 44 83 81 150 When compared to the global GEOS-Chem simulation avoided mortality numbers, the nested GEOS-Chem values are 17% higher than the global GEOS-Chem results, while the CMAQ results are 92% higher on average when full-flight emissions are considered. 54 Table 10 shows that when using the EPA CRF formulation, considering only LTO emissions results in a global net increase in mortalities (net health disbenefit), while using the WHO formulation results in a global net decrease in mortalities (net health benefit) when compared to the baseline scenario. Based on the previously defined CRFs, any increase in ground-level PM2.5 concentrations will result in an increase in mortalities (i.e. negative values). The differences in global avoided mortalities between the two CRFs can be explained by the nonlinearity of the WHO function. For instance, China's (the US's) change in concentration due to ULSJ implementation for LTO emissions is +0.002 pg/m 3 (-0.001 pg/m 3) while the background concentration is 45.5 pg/m 3 (6.2 pg/m 3), where the values are population weighted. A gradient can be defined for each CRF, where the WHO gradient is defined as (RR-1)/RR, and the EPA gradient is defined as the product of the risk coefficient and the change in ground-level PM2.5 concentration. Applying the data that was used to derive the population weighted values, the WHO CRF gradient is 12 times lower than the gradient assumed by the EPA CRF in China and 8 times lower than the gradient assumed by the WHO CRF in the US. As a result, the WHO CRF predicts -13 avoided mortalities compared to the -150 avoided mortalities predicted by the EPA CRF (whereas in the US, it is 34 versus 31 avoided mortalities, WHO and EPA, respectively). Thus, countries such as China, including India where background PM 2.5 concentrations are also high, cause an overall increase in mortalities when health impacts are scaled linearly to ground-level concentration changes for LTO impacts (-130 on average, not discounted). 4.3. VSL Results by Country Table 12 presents the VSLs determined for select countries as described previously as well as total valuation from health impacts when no lag structure is considered (non-discounted health impacts). All values are in 2006 US dollars, and the GNI data, which is purchasing power parity adjusted, was obtained from the World Bank database.10 5 55 Table 12: VSL and valuation of avoided premature mortalities (when cruise emissions are included) due to ULSJ implementation by country in US 2006 $. ULSJ Country Australia Canada China Egypt Germany India Japan Kenya Saudi Arabia United Kingdom United States of America Total GNI/capita 2006pUS$00 3,0 2.5% d Percentile 33,010 930,000 36,410 500,000 4,790 51,000 4,700 30,000 34,410 430,000 2,540 5,100 32,770 400,000 1,430 1,700 22,590 220,000 35,110 450,000 45,640 690,000 77 VL 2006 US$ Mean 4,400,000 5,100,000 290,000 280,000 4,600,000 130,000 4,300,000 61,000 2,500,000 4,800,000 7,100,000 Valuation 2006 US$ 97.5% Percentile 11,000,000 13,000,000 1,100,000 1,100,000 12,000,000 560,000 11,000,000 300,000 6,800,000 12,000,000 18,000,000 Mean 4,000,000 30,000,000 63,000,000 11,000,000 390,000,000 110,000,0001 64,000,000 89,000 28,000,000 120,000,000 830,000,000 2,500,0000 From the selected countries, the estimated VSL and avoided premature mortality valuations are both largest for the US. The predicted VSLs vary by several orders of magnitude and although China has almost twice as many avoided premature mortalities as the US, the valuation of these mortalities is -10 times lower than the US. 4.4. Global and US ULSJ Outcomes 4.4.1. Assumptions for Global and US Implementation Analysis For each of the Monte Carlo (MC) simulations, the following assumptions are made for the global implementation of ULSJ analysis: " Changes in energy density and specific energy are not considered within this analysis. " APMT-Climate input parameters are used as described in Table 13 and the distributions and associated values are provided in Table 14. " Three discount rates are applied to climate costs deterministically: 2, 3, and 7%. * The EPA CRF methodology is used to calculate the number of avoided mortalities due to ULSJ implementation in this analysis where a mortality lag structure is implemented assuming the same discount rates as applied to climate costs. 56 0 Full-flight health impacts are considered. " Gross National Income (GNI) per capita adjusted for purchasing power parity (PPP) is used as the income measure to determine VSLs across countries. * Price differentials are assumed to be applicable on a global level although they are based on US price history data. The primary assumptions for the US implementation analysis are very similar to those used in the global implementation analysis. The only additional or differing points are the following: e The calculations provided in the US implementation analysis are the costs seen by the US due to a global implementation of ULSJ. * Climate costs are scaled by a regional GDP factor to obtain climate costs for just the US (7-23% of total damages"'). * Avoided mortality benefits are those seen by the US due to global implementation. e Implementation costs are a result of US fuel burn, only (6.74x 1010 kg for aviation year 2006), i.e. the amount of fuel burn seen in the US region as defined by the nested GEOS-Chem grid. A US-only implementation analysis is also performed. The overall structure of the analysis is the same as the global implementation analysis, except for the following distinctions: " Climate costs are calculated based on US fuel burn, only. Given that the DICE-2007 damage function calculates impact on global GDP, the 7-23% fraction for US damages is again applied. * Avoided mortality benefits are those seen by the US due to US-only implementation, where US-only implementation is approximated by a nested GEOS-Chem simulation with baseline boundary conditions and ULSJ for all flights within the domain * Implementation costs are a result of US fuel burn. 57 4.4.2. Assumed Uncertainty Distributions Table 13 provides a brief description of each input parameter used in the MC analysis. As described in the cost build-up section, the additional price associated with increased hydroprocessing and hydrogen gas capacity determines the amount of additional lifecycle C02 emissions. The amount of additional hydroprocessing is also directly related to the expected change in fuel energy density and specific energy, as described in the operations section below. Table 14 shows the values and assumed distributions for each of the described input parameters in Table 13. 58 4.4.3. Global Implementation Analysis Results Table 15 presents the primary results from the EBCA for global implementation of ULSJ. Note that values in parentheses are non-cost beneficial values. Table 15: Global implementation EBCA results, given in 2006 US $ Billion. Component -Climate 2% 3% 7% Air Qualit _2% 3% 7% Implementation Total 2% 3% 17% Mean Median 95% Interval 2.35 1.64 (0.82) 2.07 1.46 (0.73) 0.13 0.10 -34.26 0.06 - 20 2.34 2.27 2.05 (2.52) 1.83 1.77 1.60 (2.49) 0.21 -7.55 0.20 -7.32 0.18 -6.59 (1.31) - (3.80) (2.53) (1.89) (1.29) (2.63) (2.11) (1.62) (7.70) - 3.37 (5.98)- 3.59 (4.15) - 3.48 % Cost Beneficial 16.32 15 17 20 Figure 6 shows the benefit-cost distribution produced by the MC simulations for each of the three scenarios described in Table 15. Positive values represent net cost beneficial scenarios while negative values represent net non-cost beneficial scenarios. From Table 11, there are an estimated -2300 avoided premature mortalities resulting from global ULSJ implementation. 2% DR -10 -8 6 -4 -2 0 2 4 6 8 10 2 4 6 8 10 2 4 6 8 10 3% DR -10 -8 -6 -4 -2 0 7% DR -10 -8 -6 -4 -2 0 Net Benefit-Cost (US 2006 $ billion) Figure 6: Benefit-cost distribution for global implementation analysis for three different discount rates (DRs). 61 Table 16 shows statistics when climate impacts are weighed against air quality impacts only. Note that climate and air quality statistics are the same as in Table 15. Table 16: Global implementation results from EBCA where no implementation cost has been included, given in 2006 US $ Billion. Component Climate 2% 3% 7% Air Qualit 2% 3% 7% Total 2% 3% 7% Mean Median 95% Interval (2.35) (1.64) (0.82) 2.07) (1.46) 0.73 (0.13) - (6.32 (0.10) -. (4.26) (0.06) - 2.081 2.34 2.27 2.05 1.83 1.77 1.60 0.00 0.63 1.23 (0.18) 0.31 0.85 %Cost Beneficial 10.21 -7.55 10.20 -7.32 10.18 -6.59 4.68)- 5.65 3.06 - 5.84 1.16 - 5.78 46 57 77 Alternatively, implementation costs are weighted against air quality impacts only due to the uncertainty in global climate impacts. This is shown in Table 17. Table 17: Global implementation results from EBCA where no climate cost has been included, given in 2006 US $ Billion. Com onent Mean Air Quali 2% 2.34 3% 2.27 7% 2.05 Implementation 12.521 Total 1 2% 0.18 3% 0.25 7% 0.47 Median 95% Interval 1.83 1.77 1.60 2.9) 0.21 -7.55 0.20 -7.32 0.18 -6.59 (1.31) -13.80 0.61 0.66 0.84 2.94 - 5.20 2.95 - 4.98 3.00 - 4.26 % Cost Beneficial 37 35 31 4.4.4. US Implementation Analysis Table 18 presents the primary results from the EBCA for US implementation of ULSJ. Note that values in parentheses are not cost beneficial values. Table 19 provides health impacts and valuations for the US from global implementation for the other two models i.e. nested GEOSChem and CMAQ. Valuations are discounted for the lag in health impacts and a nominal VSL of 2006 US $7.4 million is assumed. All values in Table 19 are nominal values. 62 Table 18: US EBCA results for global implementation, given in 2006 US $Billion. Component Climate 2% 3% 7% Air Quality 2% 3% 108 I % Cost "Ran~ Mean Median 0.75 0.60 89) 0.06 -2.37 (0.56) (1.7) - (1.326 2 (0.50) (0.45) (1.46- 1.32 (1.26) - 1.20 23 2 i 0) 2%pl(0.48)io (0.40) (.35) 3% 7% Table 19: Valuation of US health impacts due to global implementation from CMAQ and GEOSChem Nested simulations. Valuation 2006 US $Million 940 300 1,500 560 Avoided Mortalities 140 44 230 83 Nested GEOS-Chem Nested GEOS-Chem, LTO CMAQ CMAQ, LTO Figure 7 shows the benefit-cost distribution produced by the MC analysis for each of the three scenarios described in Table 18. Positive values represent cost beneficial scenarios while negative values represent not cost beneficial scenarios (i.e. plotted as benefit minus cost). 2%DR -4 -3 -2 -1 0 1 2 3 4 1 2 3 4 1 2 3 4 3%DR -4 -3 -2 -1 0 7%DR -4 -3 -2 -1 0 Net Benefit-Cost (US 2006 $ billion) Figure 7: Benefit-cost distribution for US implementation analysis. 63 4.4.5. US-Only Implementation Analysis Table 20 presents the primary results from the EBCA for US-only implementation of ULSJ. Note that values in parentheses are non-cost beneficial values. Table 20: US-only implementation EBCA results, given in 2006 US $ Billion. n IMa~nn Component udn s n I I 4501. Inta nual Madian Cost I RianafiniaI Climate 2% 3% 7% Air Qualit 2% 3% 7% Implementation 0.09 0.04 0.07 0.00) - (0.24 (.40.00)1-(0.12) 0.50 0.48 0.43 (0.90) 0.38 0.38 0.35 (0.89) 0.04-1.57 0.04-1.52 0.04 -1.37 (0.47) - (1.36) 2% (0.53) (0.58) (1.33) - 0.65 1 3% (0.51) (0.56) (1.27) - 0.64 1 7% (0.51) (0.56) (1.20) - 0.52 1 Figure 8 shows the benefit-cost distribution produced by the MC simulations for each of the three scenarios described in Table 20. Positive values represent net cost beneficial scenarios while negative values represent net non-cost beneficial scenarios. 2%DR -2 -1.5 -1 -0.5 0 0.5 1 0 0.5 1 0.5 1 3% DR -2 -1.5 -1 -0.5 7%DR -2 -1.5 -1 -0.5 0 Ne t Benefit-Cost (US 2006 $ billion) Figure 8: Benefit-cost distribution for US-only implementation analysis. 64 4.4.6. Constant VSL Analysis As mentioned previously, the US VSL can be applied to all avoided mortalities to reflect a policy choice that values all premature mortalities equally. The results from this analysis are presented in Table 21 and Figure 9. Table 21: Constant US VSL EBCA results, given in 2006 US $ Billion. 7.41 (4.94) - 43.02 (3.73) - 42.55 (2.64) - 38.94 7.81 7.33 2% DR -10 0 10 20 30 40 50 60 30 40 50 60 50 60 3%DR -10 0 10 20 7% DR -10 0 10 20 30 40 Net Benefit-Cost (US 2006 $ billion) Figure 9: Benefit-cost distribution for a constant US VSL analysis. 65 4.4.7. Cost Effectiveness Analysis As an alternative to a benefit-cost analysis, implementation costs and climate disbenefits are presented on a per premature mortality basis within a cost effectiveness framework. Results are presented in Table 22 and are expressed in 2006 US $ million and are presented for the global implementation, US implementation, and US-only implementation scenarios. Table 22: Cost effectiveness analysis results, given in 2006 US $ Million. Discount Rate Global Implementation 2% 3% 7% Median 95% Interval 2.57 2.26 2.02 2.18 1.97 1.78 0.72 - 6.75 0.73 - 5.66 0.73 - 4.72 13.0 12.3 12.2 13.4 10.8 10.8 4.29 - 32.2 4.40 - 30.0 4.43 - 28.9 Mean 1_7117_71 Global Implementation-US 2% 3% 7% US-Only Implementation b 2% 16.7 14.8 5.86 - 40.3 3% 16.6 17.6 14.6 15.5 5.95 - 38.9 6.43-4.12 7% 4.5. Policy Implications The results from the global implementation analysis, shown in Table 15, indicate that the majority of predicted outcomes are not cost beneficial, although the 95% confidence interval does include the cost neutral outcome (i.e. costs and disbenefits are equal to benefits). All three monetized components of the analysis (climate, air quality, and implementation) are on a similar order of magnitude, which indicates no single component, be it climate damages or health benefits, dominates the outcomes. While higher discount rates tend to narrow the distribution shown in Figure 6, for this analysis, higher discount rates only slightly increased the percentage of cost beneficial outcomes. Increased cost beneficial outcomes is generally expected given that it will reduce the overall impact of C02 climate damages given their time scale, but increasing discount rate also reduces the avoided premature mortality benefit due to the implemented health benefit lag structure. Because implementation cost reflects the economic impact of ULSJ policy implementation, Table 16 shows the results from an analysis where only the environmental components are considered. When comparing climate damages to air quality benefits on a global scale, the average outcome of the analysis shows either cost-neutral or cost beneficial outcomes, 66 indicating that on average, the benefits from air quality will outweigh the damages seen by the expected increase in atmospheric warming. Alternatively, if the air quality impacts are measured against only the implementation cost, Table 17 shows that on average, the predicted cost required to produce ULSJ will exceed the air quality benefit from reduced ground-level PM 2.5 concentrations. From a US implementation outlook, where, again, the US implementation scenario is defined by global implementation of ULSJ but costs and benefits are considered only within the US domain, similar results as seen in the global implementation are found. All three discount rates show that the majority of outcomes (77-78%) are not cost beneficial, although the proportion of costs and benefits seen by the US are different than the relative magnitude of the costs and benefits seen in the global implementation analysis. In the case of US implementation, about 1/3 of the implementation cost and air quality benefit is taken on by the US, but only 1/7 of the climate cost is attributed to US emissions, thus resulting in a slightly larger percentage of cost beneficial outcomes. This proportionally small amount of climate damages seen by the US is a result of the assumed climate damage fraction stated in Section 4.4.1. When US-only implementation is considered, which is defined by ULSJ implementation only within the US domain, the air quality benefits dominate climate damages, but the implementation cost is a factor of 2 larger than the benefit achieved from a reduction in premature mortalities, thus only 11-13% cost beneficial scenarios are observed. Given the wide range in VSLs obtained from the methodology developed in Section 3.3 and the ethical considerations of valuing all avoided premature mortalities equally, US VSLs are also applied to all premature mortalities in the global study. As expected, ULSJ implementation on average is significantly cost beneficial (-US $10 Billion net outcome) where the majority of outcomes are cost beneficial. A constant VSL approach, however, should be considered carefully as it does not reflect current economic practice. It is difficult to say, however, whether or not this policy "should" or "should not" be implemented. Besides the constant VSL approach, where air quality benefits are likely being overvalued, all implementation scenarios show an expected cost detrimental outcome. The uncertainty ranges, however, do always capture the $0 outcome, indicating that there is a certain likelihood of an overall cost beneficial outcome, albeit this likelihood is at or under -20% for all cases. A decision on the actual policy implementation will then largely be a result of the priorities of the policymakers. If the goal is to reduce the environmental impact of aviation, then the climate 67 versus air quality approach indicates that ULSJ implementation may be a viable option. If the goal is to simply reduce the air quality impact of aviation, then ULSJ, given the expected air quality impacts, is a very attractive option given that it can be used as a drop-in fuel and can be implemented quickly given the trends and rulemaking for on and off-road diesel fuel. If, however, implementation cost is an issue given that refineries would incur the capital costs to update their refineries to meet ULSJ demand, then the economic impact of ULSJ implementation may need to be investigated further given the uncertainties present in this aspect of the analysis. 4.6. Nominal Range Sensitivity Results This section presents the results of the NRSA for both global and US implementation, where the primary results are shown in Figure 10 and Figure 11, respectively. Again, US implementation here refers to the US net benefit-cost due to a global implementation of ULSJ. Blue and green bars represent the change in net benefit-cost attributed to a low or high parameter value, respectively. Only the change in net benefit-cost relative to a base deterministic model output is shown. Tornado Plot of Net Cost/Benefit Sensitivities Price Differential US, VSL I $0.066 I $1 Million US, All Cause, Beta 0.0139 GEOS-Chem PM Conc. -60 E U $0.016 $12 Million 0.0145 +60% Global Income Elasticity 2 Discount Rate, Climate Costs 2% 1 Low 4.5 Climate Sensitivity Param. 0.0015 +75% Aerosol Optical Depth -2.5 -2 -1.5 -1 -0.5 -70% 0 NPV Change Figure 10: NRSA results for global implementation of ULSJ. 68 High 2 0.0041 Damage Function Coeff. -3 7% 0.5 1 1.5 2 2.5 x10 Tornado Plot of Net Cost/Benefit Sensitivities, US US, VSL - $1 Million Price Differential - $0.066 US, All Cause, Beta a. $12 Million $0.016 0.0139 GEOS-Chem PM Conc. 0.0145 -60% +60% US GDP Scale Factor - 23% 7% High Discount Rate, Climate Costs 2% Climate Sensitivity Param. - 4.5 Damage Function Coeff. - 0.0041 -10 -8 -6 -4 -2 0 NPV Change % 2 0.0015 2 4 6 8 x 10 Figure 11: NRSA results for US implementation of ULSJ. In the global and US analysis, the total net benefit-cost output is most significantly impacted by the US VSL, the price differential, the assumed percent change in premature mortality given a 1 pg/m 3 change in PM 2.5 concentration, and the uncertainty assumed for the ground level PM concentration change found in GEOS-Chem. Other important input parameters are the assumed global income elasticity and components specific to the climate impacts such as the climate sensitivity parameter, damage function coefficient, and the various components of the sulfate RF calculation method. The other APMT-Climate inputs do not appear as significant parameters. The global income elasticity, CP percent increase in premature mortality, and LC percent increase in premature mortality values have no effect on the US analysis because no values applied in the EBCA are derived from those parameters as they are in the full global analysis. Likewise, the GDP fraction associated with US climate costs has no impact on the global analysis. Uncertainty analysis of APMT-Climate has been performed previously and can be found in Jun. 108 This sensitivity analysis is useful in that it provides a method in which to gauge the response of the system for a perturbation in an individual parameter. The values shown in the tornado plots above can be used to estimate the benefit-cost response to an increase in US VSL or change in 69 ground PM 2 .5 concentration relative to the nominal value. For instance, if the US VSL is actually $12 million rather than $7.4 million, then one would expect a $1.5 billion increase in net benefit for the global implementation case, which is shown as a positive $1.5 billion shift in NPV of the net benefit-cost value. This type of analysis, however, is potentially misleading. This analysis provides relatively little insight into each components contribution to uncertainty, i.e. which parameters have the largest impact on the distribution seen from the MC analysis. It is misleading in the sense that the US VSL and the price differential are shown to have the largest impact on the value of the output metric, but the range in values applied in the analysis is significant compared to the other inputs. While this NRSA approach is useful as it provides some insight into what the most influential factors in this CBA within a deterministic framework are, to further understand the major sources of uncertainty in this analysis, a GSA is performed. The results of the GSA are discussed in the next section. 4.6.1. Discount Rate Within the US NRSA, both endpoints for discount rate produce a decrease in net benefit-cost. This is possible due to the interaction of discounting health benefits and climate disbenefits. Net benefit-cost is plotted against discount rate for deterministic outcomes to better understand this relationship. Discount Rate vs. Net Benefit-Cost 10 9 S-0.5 Co4 0 C ~ 1.5 Global z - - - us 1 7 2 8 9 10 Discount Rate (%) Figure 12: Net benefit-cost plotted against discount rate of the deterministic model used in the NRSA. 70 Figure 12 shows that in the global and US analyses, the net benefit-cost increases through the nominal rate of 3% (as defined, net benefit-cost at 3% will be 0) and then plateaus as the discount rate increases to 10%. The plateau can be explained by the decrease in value in future costs/benefits, thus the net benefit-cost of the system approaches the benefit-cost in the year of implementation. Also, the US analysis (dashed line) never produces a net benefit, as shown in Figure 11. The shape of the response can be explained by the decreasing climate disbenefits coupled with decreasing health benefits as the discount rate increases. At lower discount rates, climate costs more rapidly decrease compared to the decrease in health impacts. After a local minimum is reached at approximately 6% in the global analysis and 3% in the US analysis, the decrease in climate costs no longer outweighs the decrease in health benefits and a slight downturn is observed. The net result of each analysis then approaches a steady state value as the discount rate continues to increase. 4.7. Global Sensitivity Analysis (GSA) Results A GSA was performed in order to determine the contribution of each input parameter to the total output variance. Main and total effect indices are reported. Main effect indices report the specific input parameters direct impact on the output variance while the total effect indices also account for input parameter interaction. The results for the most significant factors for the global implementation are shown in Figure 13 and Figure 14. The results for the most significant factors for the US implementation are shown in Figure 15 and Figure 16. Only input parameters with main effect indices of greater than 2% are plotted. Both the US and Global implementation results yield similar results, although the climate factors were less significant in the US analysis than in the global analysis. It is clear that the US VSL input parameter has the largest impact on output variance with a main effect sensitivity index of approximately 55% and 60% for the global and US analysis, respectively, while all other significant effects are approximately 10% or below. This is not surprising given that the US VSL forms the basis for all potential benefits derived from ULSJ implementation as well as being a highly uncertain value in itself due to the assumed Weibull distribution as defined by the EPA. This analysis also shows that the same parameters shown to be significant in the NRSA are also shown to be significant in the GSA, but relative impacts on the output variance are much different than in the differences seen on the absolute value of the output. Total effect sensitivity indices are not significantly higher than the main effect sensitivity indices, indicating that second order interaction effects between the input parameters are present but not significant. 71 US, VSL Price Differential US, All Cause, Beta GEOS-Chem PM Conc I I Damage Function Coeff Global Income Elasticity 0 0.1 0.5 0.3 0.4 0.2 Main Effect Sensitivity Index 0.6 0.7 Figure 13: Global Implementation GSA main effect sensitivity index results. US, VSL US, All Cause, Beta Price Differential GEOS-Chem PM Conc Global Income Elasticity Damage Function Coeff I I 0 0 0.1 0.1 0.2 0.3 0.4 0.5 0.4 0.5 0.2 0.3 Total Effect Sensitivity Index 0. 0..6 Figure 14: Global Implementatio n GSA total effect sensitivity index results. 72 0.7 US, VSL Price Differentia US, All Cause, Beta GEOS-Chem PM Cor I 0 0.1 0.2 0.3 0.4 0.5 Main Effect Sensitivity Index 0.6 0.7 Figure 15: US Implementation GSA main effect sensitivity index results. US, VSL US, All Cause, Beta GEOS-Chem PM Conc. Price Differential 0 0.1 0.2 0.3 0.4 0.5 Total Effect Sensitivity Index Figure 16: US Implementation GSA total effect sensitivity index results. 73 0.6 0.7 THIS PAGE INTENTIONALLY LEFT BLANK. 74 Chapter 5: Fast Policy Analysis 5.1. Adjoint Model and Policy Tool As stated in Section 2.1, a single GEOS-Chem forward model simulation takes approximately 12 hours to complete. Thus, if any reanalysis needs to be performed, a complete rerun of a forward simulation is not a trivial task and using GCTMs such as GEOS-Chem as the direct policy analysis tool can be both cumbersome and inefficient. If variations in the emissions scenario are small, such as is often the case with aviation, it may be more effective to use calculated sensitivities of outputs to input emissions. The task of calculating these sensitivities was described in detail in Section 2.6 through the use of the GEOS-Chem adjoint model. A policy tool is currently being developed based on these computed adjoint sensitivities. The general approach is based off of the work done previously by Koo.56 His work focused on determining a set of "source-receptor" matrices, the source referring to the geographical source of aircraft emissions, and receptor referring to the geographical receptor of health impacts based on the defined cost function. The primary functions of the policy tool are to compute the number of premature mortalities resulting from a particular policy scenario as well as to monetize these premature mortalities. For premature mortalities, the tool uses the following defined cost function: = = 1= _1 j=lf 3 0+,i,j,kPi,J,kBi,j,kXi,j,k,n Eq. (35), where f3o. is the fraction of the population greater than 30 years of age, P is the population, B is the baseline incidence rate for either CP or LC related mortalities, and X is the concentration of PM 2 5 , where all of these values are a function of the grid cell, ijk,where X is also a function of time step, n. As can be seen, Eq. (35) is essentially identical to Eq. (22), except that it does not include non-grid specific values (such as the relative risk ratio), which are multiplied later on within the tool. For monetization, the tool uses the following defined cost function: = k=1(f30+,i,j,kPi,I,kBi,j,kXi,j,k,n) =1 = 1X 1 _1 E= I) ijk Eq. (36), where Eq. (36) is Eq. (35) multiplied by the income ratio used to determine the extrapolated VSL. Again, these cost functions can be limited to the desired receptor region by limiting the summation limits in the above equations. 75 Once the appropriate source-receptor matrix has been calculated from an adjoint simulation, the total impact can be calculated as described in Figure 17. 3D Gridded Emissions Data 3D Gridded Sensitivity Data Inner Product Defines Policy Scenario = 0 Total Atmospheric Impact of Aircraft Emissions Figure 17: Using sensitivities to compute the total impact from an aircraft emissions policy scenario.109'110 Figure 17 shows that using sensitivities, the total impact can be computed very easily as it is just a matrix inner product between the 3D sensitivity data and the defined emissions scenario. Currently, the tool also has many of the uncertainty aspects as described in Chapter 3 built-in to the code so that MC simulations can be simultaneously performed. The primary assumption that allows the use of first-order sensitivities to compute the approximated perturbation to the atmosphere is that GEOS-Chem is indeed linear to small perturbations in input emissions. This was shown to be an appropriate assumption in Koo. 56 5.2. USLJ Analysis Comparison This section provides a policy comparison between the forward model calculated premature mortalities and those calculated by the newly implemented policy tool. Here, premature mortalities from global ULSJ implementation are determined. The distribution in Figure 18 shows the MC results from the forward model data where the mortality lag-structure has not been taken into account. 76 Avoided Mortality Distribution Based on Sensitivity Analysis 200 180 -160 140 120a) 100_ CD - 80 6040 - 20 0 0 1000 2000 3000 4000 Avoided Mortalities 6000 6000 Figure 18: Forward model premature mortality results from the ULSJ EBCA. Again, these results were based on the methodology described in Section 3.4 where the data show a modal value of approximately 1800 premature mortalities derived from global ULSJ implementation. Next, the adjoint sensitivity policy tool is applied. First, an appropriate aircraft emissions scenario needs to be applied. Again, the empirical relationship between total aircraft SOx emissions and FSC is shown in Table 5, which is shown again in Table 23. Table 23: Aircraft SOx Emissions S02 1(FSC/1000) x [(100 - E)/100] x FUEL x (64/32) S04 (FSC/1000) x (E/100) x FUEL x (96/32) Given that Table 23 shows a linear relationship between SOx emissions and FSC, a very simple emissions scaling factor can be applied to the baseline emissions scenario. Again, the ULSJ scenario assumes a decrease in FSC of approximately 600 to 15 ppm, meaning the SOx scaling factor would be 15/600. Applying this scaling factor and using the tool based on the methodology presented in Figure 17, the following distribution is calculated and plotted against the distribution shown in Figure 18: 77 Comparison of Distributions 200 | 180 Adjoint Forward 160140 120- 100 60 40 200 1000 2000 4000 3000 Avoided Mortalities 6000 5000 Comparison of Distributions, Adjusted 200 Adjoint 180 - Forward - 160 140 120 S100- u- 80 - 60 - 40 20 0 1000 2 300 4000 Avoided Mortalities 6000 6000 Figure 19: Adjoint policy tool results and adjusted results for global ULSJ implementation. As can be seen in top half of Figure 19, the shapes of the distributions are very similar, but the modal value produced from the adjoint policy tool is only 1500 avoided premature mortalities. If the distribution produced from the adjoint policy tool is multiplied by the ratio of the average predicted avoided premature mortalities, the forward model distribution is accurately reproduced, as shown in the bottom half of Figure 19. This outcome indicates that the policy tool is systematically underestimating the avoided premature mortalities for this particular policy scenario, suggesting that further development of the policy tool will require some sensitivity tuning. This systematic underestimation relative to the forward model results may be a 78 consequence of several factors. First, while the linear approximation relative to changes in aircraft emissions has been shown to be appropriate, there are still non-negligible second-order errors (possibly up to 10%). In addition, this analysis did not incorporate any sort of temporal variations in sensitivities nor aircraft emissions given the increase in data storage that would be required for the tool. Thus, taking the inner product between yearly averaged sensitivities and aircraft emissions could bias the results given in or out of phase temporal patterns in both the sensitivities and emissions. Although this bias does exist, the results from the adjoint policy tool do show a first-order accurate estimate of the resulting decrease in premature mortalities given a global implementation of ULSJ. Future implementation of temporal sensitivities (on the order of a month or week rather than a year) and its impact on this bias is currently being investigated. The premature mortality valuation portion of the code is still being developed and is based on the cost function defined by Eq. (36). The primary difficulty in the development of this aspect of the code is the inability to post-process the adjoint sensitivities. Eq. (36) can only be defined for a set income elasticity value, and this elasticity cannot be altered after the simulation has been performed given that the cost function is a global summation. Thus, it will be necessary to perform several adjoint simulations with different assumed income elasticities and then interpolate to approximate the mortality valuation. A comparison, however, is provided between the forward model results when a constant income elasticity of 1 is assumed and the adjoint policy tool. A distribution comparison is shown in Figure 20. 79 Comparison of Valuation Distributions 450 Adjoint Forward 400 350 300 u 250 I 200 U- 150 100 50 0'- 0 0.5 1 1.5 Avoided Mortality Valuatoin 2 2.5 10 10 Figure 20: Adjoint policy tool results for monetized avoided premature mortalities with an assumed income elasticity of 1. Figure 20 again shows a systematic underestimation of the valuation of avoided premature mortalities for global ULSJ implementation. Even when adjusting for the bias, the distribution produced by the adjoint policy tool does not exactly match the distribution based on the forward model results. Differences in this case can be attributed to the fact that for this sample analysis, some of the uncertainty factors pertaining to disease specific premature mortalities (such as a variable relative risk ratio) were excluded based on the definition of the cost function. 80 THIS PAGE INTENTIONALLY LEFT BLANK. 81 Chapter 6: Conclusions and Future Work As the aviation sector of transportation continues its growth, the effect of aircraft operations on the environment will have increased importance with regards to its impacts on climate and air quality. As such, policies proposed to address these environmental issues need to be evaluated within a consistent framework. This thesis introduced the concepts of environmental benefitcosts analysis as it pertains to aviation policy. It also helped to develop a framework in which to conduct global policy analysis given that tools currently available can perform regional analysis, only. 6.1. Global ULSJ Implementation This thesis focused largely on the global implementation of an ultra-low sulfur jet fuel (ULSJ). Government rulemaking would require the fuel sulfur content of jet fuel to be reduced to at most 15 ppm, a measure already in place for on and off-road diesel fuel. There are two primary expected consequences of ULSJ implementation: a reduction in ground-level PM2.5 as a result of a decrease in atmospheric sulfate aerosol burden and an increase in global temperature given the loss in cooling as sulfate aerosols reflect incoming solar radiation and an overall increase in C02 emissions due to additional fuel processing requirements. After determining the magnitude of these potential impacts, the valuation of climate damages as well as air quality benefits as it pertained to decreases in premature mortality were calculated and the economic impact of increased fuel processing to bring ULSJ to specification was estimated. It was calculated that ULSJ would have a global benefit of approximately 2300 avoided premature mortalities, of which 120 are a result of decreased PM2. concentrations within the US. Using the variable VSL approach and without discounting the premature mortalities, these 2300 mortalities were valued at approximately 2006 US $2.5 billion. In terms of climate damages, a 3.3 mW/m2 increase in global RF was calculated due to the sulfate aerosol reduction. This RF increase, as well as the one associated with the 2% increase in WTW C02 emissions due to the fuel processing requirements, resulted in expected climate damages of 2006 US $0.8 - 2.4 billion, where variations are due to the assumed discount rate. The economic impact of global ULSJ implementation was valued at approximately 2006 US $2.5 billion for the year of policy implementation. 82 While an air quality benefit due to ULSJ implementation is expected, the total aggregated outcome from this analysis indicates that there is between an 80-85% likelihood that global ULSJ implementation will result in a not cost beneficial outcome based on the MC uncertainty analysis. If, however, the health benefits are weighted against only the climate disbenefits, then there is between a 46-77% chance that there will be a cost beneficial outcome. Thus, the decision of whether or not to implement ULSJ will be a largely motivated by the priorities of the policymaker. The FAA has already stated their goal of a 50% reduction in air quality impacts of aircraft emissions by 2018, a goal that ULSJ implementation would help achieve. If a stronger priority is placed on achieving this goal rather than about the underlying cost-effectiveness, then the policymakers may indeed opt to implement ULSJ. 6.2. Fast Policy Analysis Based on the adjoint model currently implemented within the GEOS-Chem GCTM, a fast policy tool was developed in order to provide rapid policy analysis. The ULSJ study was burdened by the computational intensity of GEOS-Chem. The calculation of a particular cost function's sensitivity to aircraft emissions allows for a first order approximation of the effect of a particular policy scenario. While the tool is still being developed, initial estimates of the amount of expected premature mortalities due to a sample ULSJ case are promising, where the tool estimates a modal value of 1500 avoided premature mortalities compared to 1800 calculated based on the forward model analysis. The underestimation is thought to occur due to the annually averaged approach taken by the adjoint policy tool. Temporal resolution is lost due to the desire to reduce overall data storage and to reduce simulation times, and given that temporal variations in sensitivities as well as aircraft emissions are no longer captured, an underestimation occurs. This will be investigated further in future work. 6.3. Limitations An obvious limitation for any policy analysis is the amount and quantification of uncertainty within a given study. Based on the global sensitivity analysis performed for the global ULSJ implementation case, nearly 60% of the uncertainty present within this analysis was a result of the distribution assumed for the base (US) VSL value. This indicates that reductions in uncertainty within the total analysis can be achieved given more refined estimates of US VSL values. 83 The time-scale of the ULSJ study was also a major limitation in this analysis. Impacts due to ULSJ implementation were only considered for the year of implementation. Aviation policies, however, have much longer time scales (on the order of decades), thus a more informed policy decision may be obtained by considering ULSJ implementation out to 20 or 30 years in order to incorporate changes in the atmospheric state that could increase or decrease the air quality and climate impacts relative to the baseline. Lower economic costs could also occur in the future as production efficiency of ULSJ improves. The primary limiting factors for why a multi-year study was not performed within this thesis were the lack of global background (as well as aircraft) emissions data for future years and the significant computation times that would be associated with such simulations. A potential option for addressing these issues, however, is presented in Section 6.4. In addition, higher fidelity options must be available in the future when considering the climate impacts of different scenarios. For this particular case, the APMT-Climate tool was sufficient as it was not difficult to obtain a first order estimate of the RF impact due to a reduction in sulfate aerosols as well as the expected increase in CO 2 emissions. For other scenarios where there may be significant changes in other trace gases (such as ozone), a full radiative transfer model will be required to provide more accurate RF calculations. Also, coupling a radiative transfer model with a GCTM such as GEOS-Chem will allow for multi-year climate studies, making policy studies on the order of decades easier to conduct. 6.4. Future Work As mentioned above, integration of an RTM is of great importance, especially for more refined calculations in RF change across policy scenarios as well as for climate impact due to aviation studies. Future work will focus on the development of a simplified environment in which results from GEOS-Chem simulations can be passed to an RTM and higher fidelity RF calculations can be obtained. This work, however, will not focus on direct implementation of an RTM coupled within GEOS-Chem as this is currently in development within the GEOS-Chem community. With regards to the adjoint policy tool, the premature mortality monetization component needs to be fully implemented. In addition, given that policies have on the order of 20 year lifetimes, development of the policy tool to predict policy impacts out to several decades would greatly improve the ULSJ policy analysis performed here as well as improve overall decision making by policymakers. Figure 21 shows the proposed structure of a multi-year adjoint policy tool. 84 Define Emissions Scenarios I Scale Emissions Pre-Defined Bg Scenario 1 Pre-Defined Bg Scenario n Pre-Defined Bg Scenario N Determine First Order Impact Compute Air Quality Impact Year 1 --- Interpolate Compute Air --Quality ImpactQuality Year n Interpolate Compute Air Impact Total Impact Year N Figure 21: Proposed structure of multi-year study adjoint policy tool. A defined policy emissions scenario would be passed to several pre-defined background scenarios in which the total atmospheric impact for each of the years corresponding to the background scenarios would be calculated based on sensitivities computed from corresponding adjoint model simulations. Each background scenario would be defined by the predicted increase or decrease in background anthropogenic emissions over time. Growth and decline in anthropogenic emissions and its impact on air quality for the US have been analyzed by Ashok."' Impacts in years where these is no pre-defined background scenario can be interpolated from the years built-in to the policy tool, and the impacts can be monetized and discounted as appropriate to obtain the total air quality impact of the policy. The primary difficulty is determining the interpolation scheme that will be required. Sensitivities are a function of the defined background scenario as the adjoint tool produces a linear projection relative to the total amount of emissions present at a given time step. Thus, to minimize the overall error based on these first order estimates, it will be necessary to study how nonlinearly sensitivities are related to each type of emissions given defined variations in the background scenarios. 85 THIS PAGE INTENTIONALLY LEFT BLANK. 86 Appendix A: Tables Full table of estimated premature mortalities due to global ULSJ implementation. Table Al: Premature mortalities by country. EPA-derived CRF from Eq. (13), Full-Flight Emissions EPA-derived CRF from Eq. (13), LTO Emissions ' Country Afghanistan Albania Algeria Angola Antigua and Barbuda Argentina Armenia Australia Austria Azerbaijan Bahrain Bangladesh Belarus Belgium Belize Benin Bhutan Bolivia BosniaHerzegovina Botswana Brazil Brunei Darussalam Bulgaria Burkina Faso Burundi Cambodia Cameroon Canada Cape Verde Central African Republic Chad Chile China Colombia ' WHO CRF from Eq. (7) FullFlight LTO 16 1 7.3 2.7 97.5% Percentile 31 1.9 13 5.5 2.5% Percentile 0.0 0.1 0.4 -0.20 Mean 0.0 0.3 1.2 -0.10 97.5% Percentile 0.0 0.4 1.9 0.0 14 0.6 12 1.20 -0.1 0.1 1.6 0.0 0.0 0.7 1.6 0.4 2.3 1.8 0.1 32 5.2 4.1 0.0 0.4 0.4 0.1 0.0 1.7 4.1 0.9 5.8 4.5 0.2 81 13 11 0.0 1 1.3 0.3 0.0 3.1 7.9 1.6 11 8.3 0.4 150 24 20 0.1 2 2.6 0.5 0.0 0.1 0.0 0.1 0.1 0.0 0.0 0.7 -0.5 0.1 0.0 0.1 0.0 0.0 0.0 0.3 0.0 0.2 0.4 0.1 0.0 1.8 -0.3 0.3 0.0 0.3 0.0 0.0 0.0 0.5 0.0 0.3 0.6 0.2 0.0 3.2 -0.1 0.6 0.0 0.5 0.1 0.0 0.0 2.8 2.6 1.4 1.8 3.7 0.2 39 4.8 3.1 0.1 1 0.6 0.4 0.0 0.4 0.0 0.3 0.2 0.1 0.0 0.8 -0.1 0.1 0.0 0.2 0.0 0.0 0.6 0.0 3.2 1.5 0.1 8.2 2.7 0.1 15 0.1 0.0 0.6 0.3 0.0 1.6 0.6 0.0 2.9 0.7 0.1 15 0.2 0.0 2.4 0.0 1.4 1.8 0.1 0.4 2 2.3 0.0 0.0 3.5 5 0.4 1.2 5.4 6 0.0 0.0 6.3 9.6 0.8 2.4 11 11 0.1 0.0 0.1 0.2 0.0 0.0 0.4 0.3 0.0 0.0 0.3 0.4 0.0 0.1 0.9 0.7 0.0 0.0 0.5 0.8 0.0 0.2 2 1.2 0.0 0.0 1.6 7.9 0.5 1.7 5.1 5.9 0.2 0.0 0.1 0.4 0.0 0.1 0.9 0.8 0.0 -2.6 2.2 0.1 85 0.6 -1.3 5.8 0.2 220 1.5 -0.5 12 0.4 390 2.7 0.6 0.8 0.0 -270 0.0 1.5 2 0.1 -150 0.0 3.2 4.5 0.1 -59 0.0 -0.3 6.9 0.3 72 4.1 0.6 1.7 0.1 -13 0.1 2.5% Percentile 5.2 0.4 2.9 1.0 Mean 87 EPA-derived CRF from Eq. (13), Full-Flight Emissions ' EPA-derived CRF from Eq. (13), LTO Emissions WHO CRF from Eq. (7) Full- LTO linh* Country Commonwealth of Dominica Comoros Congo Congo Democratic Republic Costa Rica Croatia Cyprus Czech Republic Denmark Djibouti Dominican Republic East Timor Ecuador Egypt El Salvador Equatorial Guinea Eritrea Estonia Ethiopia Federated State of Micronesia Fiji Finland France FYROM/Macedonia Gabon Gambia Georgia Germany Ghana Greece Grenada Guatemala Guinea Guinea-Bissau Guyana Honduras Hungary Iceland India Indonesia Iran Iraq Ireland 2.5% Percentile Mean 97.5% Percentile 2.5% Percentile . Mean 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.0 1.2 0.1 1 0.1 4.3 1.2 0.1 3.3 0.2 2.5 0.3 11 3.1 0.2 6.3 0.3 4.6 0.5 20 5.8 0.4 -0.7 0.0 0.3 0.0 -1.3 0.1 0.0 -0.4 0.0 0.7 0.0 -0.7 0.3 0.0 -0.1 0.0 1.3 0.1 -0.3 0.5 0.0 2.6 0.7 1.2 0.3 2.9 1.5 0.5 -0.2 0.0 0.3 0.0 -0.2 0.1 0.0 0.8 0.0 0.0 15 0.2 0.0 0.3 0.3 6.5 1.9 0.0 0.1 39 0.5 0.0 0.9 0.8 18 3.4 0.0 0.2 70 0.9 0.1 1.7 1.5 36 0.0 0.0 0.0 1.1 0.0 0.0 0.0 -0.2 0.3 0.1 0.0 0.0 3 0.0 0.0 0.0 -0.1 0.8 0.1 0.0 0.0 5.2 0.0 0.0 0.1 0.0 1.8 5.7 0.0 0.4 46 1.2 0.1 1.3 0.5 30 0.2 0.0 0.1 3.4 0.0 0.0 0.0 -0.1 1 0.0 0.0 0.4 8.3 0.3 0.0 0.1 1.1 32 0.7 1.2 0.0 0.3 0.6 0.1 0.0 0.2 2.7 0.0 340 4.8 12 5.8 0.2 0.0 0.0 1.1 22 0.9 0.1 0.4 2.8 83 2 3.1 0.0 0.8 1.7 0.3 0.0 0.6 7 0.0 870 12 29 15 0.6 0.0 0.0 2 39 1.6 0.1 0.8 5.1 150 4 5.7 0.0 1.5 3.2 0.7 0.1 1.1 13 0.0 1600 22 54 28 1.1 0.0 0.0 0.1 1.1 0.1 0.0 0.0 0.0 1.8 0.2 0.2 0.0 0.0 0.2 0.0 0.0 0.0 0.2 0.0 -93 0.5 0.1 0.2 0.1 0.0 0.0 0.2 2.9 0.2 0.0 0.2 0.0 4.7 0.5 0.5 0.0 0.0 0.5 0.1 0.0 0.0 0.6 0.0 -55 1.3 0.3 0.6 0.2 0.0 0.0 0.3 5.1 0.3 0.0 0.2 0.0 8.4 1 0.9 0.0 0.0 1.1 0.1 0.0 0.0 1.1 0.0 -20 2.3 0.5 1.1 0.3 0.0 0.0 1 11 0.5 0.1 0.6 1.9 25 2.3 2.8 0.0 2 1.6 0.5 0.1 1.6 2.2 0.0 390 20 28 15 0.7 0.0 0.0 0.1 1.8 0.1 0.0 0.2 0.0 1.4 0.4 0.5 0.0 0.1 0.3 0.1 0.0 0.0 0.2 0.0 -15 2.6 0.4 0.6 0.2 88 97.5% Percentile EPA-derived CRF from Eq. (13), Full-Flight Emissions Country Israel Italy Ivory Coast Jamaica Japan Jordan Kazakhstan Kenya Kiribati Korea Kuwait Kyrgyz Republic Lao People's Democratic Republic Latvia Lebanon Lesotho Liberia Libyan Arab Jamahiriya Lithuania Luxembourg Madagascar Malawi Malaysia Maldives Mali Malta Mauritania Mauritius Mexico Mongolia Morocco (includes Western Sahara) Mozambique Namibia Nepal Netherlands New Zealand Nicaragua Niger Nigeria Norway Oman Pakistan Panama EPA-derived CRF from Eq. (13), LTO Emissions WHO CRF from Eq. (7) FullFlight LTO :2I 2.5% Percentile Mean 0.8 2 9.8 25 0.7 1.8 0.3 0.7 5.7 15 0.6 1.5 2.9 7.5 0.5 1.5 0.0 0.0 3.3 8.5 0.2 0.5 2.3 0.9 97.5% Percentile 3.7 46 3.6 1.3 27 2.8 14 2.9 0.0 16 1 4.2 2.5% Percentile 0.1 3.2 0.0 0.0 1.3 0.0 0.3 0.0 0.0 0.5 0.0 0.1 Mean 0.2 8.4 0.1 0.0 3.4 0.1 0.7 0.0 0.0 1.3 0.0 0.2 97.5% Percentile 0.3 15 0.2 0.0 6.3 0.2 1.2 0.1 0.0 2.3 0.0 0.5 2.5 15 2.6 2.1 14 1.9 6.9 4.2 0.0 2.1 0.5 2.3 0.2 4.9 0.1 0.1 3.6 0.1 0.6 0.1 0.0 0.3 0.0 0.2 0.3 0.6 0.7 0.0 0.1 0.9 1.5 2 0.1 0.2 1.7 2.8 3.6 0.1 0.5 -0.1 -0.3 0.1 0.0 0.0 -0.1 -0.2 0.2 0.0 0.0 0.0 -0.1 0.3 0.0 0.0 0.6 0.9 2.1 0.1 0.6 0.0 -0.1 0.2 0.0 0.0 0.5 1.5 0.1 0.1 0.1 0.4 0.0 2.1 0.1 0.4 0.0 6.5 0.1 1.4 3.9 0.3 0.2 0.3 1.1 0.0 5.7 0.2 1 0.0 17 0.3 2.6 7.1 0.6 0.4 0.6 2.1 0.1 11 0.3 2.2 0.0 30 0.6 0.1 0.0 0.0 0.0 0.0 0.0 0.0 -0.3 0.0 0.0 0.0 0.5 0.0 0.2 0.0 0.0 0.0 0.0 0.1 0.0 -0.2 0.0 0.0 0.0 1.3 0.0 0.3 0.1 0.0 0.0 0.0 0.2 0.0 -0.1 0.1 0.0 0.0 2.3 0.0 2.1 1.4 0.1 0.7 0.6 1.5 0.1 9.9 0.2 2.5 0.1 26 0.5 0.3 0.0 0.0 0.0 0.0 0.1 0.0 -0.2 0.0 0.0 0.0 2.1 0.0 3.9 0.1 0.0 7.3 3 0.0 0.2 2.7 11 0.3 0.3 37 0.0 10 0.4 0.0 18 7.7 0.0 0.4 8 30 0.7 0.8 95 0.1 18 0.7 0.1 34 14 0.0 0.8 15 60 1.3 1.6 170 0.2 0.3 0.0 0.0 -1.9 0.3 0.0 0.0 0.2 0.9 0.0 0.0 -18 0.0 0.7 0.0 0.0 -1.1 0.8 0.0 0.0 0.5 2.2 0.1 0.0 -9.7 0.0 1.3 0.0 0.0 -0.4 1.5 0.0 0.0 1.1 4.9 0.2 0.1 -3.8 0.0 19 0.8 0.1 8.3 2.8 0.1 1.5 15 37 0.9 1 51 0.5 1.4 0.0 0.0 -0.3 0.4 0.0 0.0 1.1 2.9 0.2 0.1 -3.1 0.0 89 EPA-derived CRF from Eq. (13), Full-Flight Emissions Country Papua New Guinea Paraguay Peru Philippines Poland Portugal Republic of Moldova Romania Russia Rwanda Saint Kitts and Nevis Saint Lucia Saint Vincent EPA-derived CRF from Eq. (13), LTO Emissions WHO CRF from Eq. (7) FullFliaht LTO 2.5% Percentile 0.0 0.1 0.2 1.7 11 0.7 Mean 0.0 0.2 0.5 4.5 27 1.7 97.5% Percentile 0.1 0.4 0.9 8.3 50 3.1 2.5% Percentile 0.0 0.0 0.0 0.1 -1.6 0.2 Mean 0.0 0.0 0.0 0.2 -0.9 0.5 Percentile 0.0 0.0 0.1 0.3 -0.3 0.9 U.2 U.U 0.3 1.1 13 7.5 2.6 0.0 0.1 0.5 -0.5 0.7 0.9 3.7 28 0.1 2.3 9.3 73 0.4 4.2 17 140 0.7 -0.5 -4.8 -18 0.0 -0.3 -2.8 -11 0.0 -0.1 -1 -3.6 0.0 1 3 37 0.4 -0.1 -0.8 -2.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.0 4 1.3 0.0 11 3.6 0.0 22 7.2 0.0 0.4 0.3 0.0 1 0.8 0.0 1.9 1.8 0.0 13 5.3 0.0 1.1 1.3 3.1 0.0 0.5 0.1 2.3 0.4 0.0 0.7 5.8 1.5 5.8 0.0 0.0 1.1 1.2 7.9 0.0 1.6 0.3 5.8 1.1 0.0 1.8 15 3.8 16 0.0 0.0 2.8 3 14 0.0 3.3 0.6 11 2 0.0 3.4 27 7.1 31 0.0 0.0 5.1 5.5 0.5 0.0 0.2 0.0 0.1 0.1 0.0 0.1 1.5 0.0 0.2 0.0 0.0 0.0 0.3 1.2 0.0 0.4 0.1 0.3 0.3 0.0 0.2 4.1 0.1 0.7 0.0 0.0 0.0 0.8 2.2 0.0 1.1 0.1 0.5 0.6 0.0 0.3 7.1 0.2 1.2 0.0 0.0 0.0 1.3 3.6 0.0 1.7 0.3 1.3 0.5 0.0 2.2 14 9.2 22 0.1 0.0 2.2 1.1 0.6 0.0 0.3 0.1 0.1 0.1 0.0 0.2 3.7 0.3 0.8 0.0 0.0 0.1 0.2 1.6 0.9 2.6 0.3 0.0 4.1 2.4 6.8 0.7 0.0 7.4 4.6 12 1.4 0.0 0.1 0.1 0.1 0.1 0.0 0.3 0.2 0.2 0.3 0.0 0.5 0.4 0.3 0.5 0.0 3.9 2.2 7.1 0.7 0.0 0.3 0.2 0.4 0.2 0.0 0.0 1.3 8.3 1.4 0.7 0.1 3.4 21 3.5 1.8 0.2 6.1 39 6.4 3.4 0.0 0.3 0.8 0.1 0.0 0.0 0.7 2.3 0.3 0.1 0.0 1.2 3.8 0.5 0.1 0.5 5 16 4.1 2.9 0.0 0.9 1.5 0.3 0.1 97.5% Sao Tome and Principe Saudi Arabia Senegal Serbia and Montenegro Seychelles Sierra Leone Singapore Slovakia Slovenia Solomon Islands South Africa Spain Sri Lanka Sudan Suriname Swaziland Sweden Switzerland Syrian Arab Republic Tajikistan Thailand Togo Tonga Trinidad and Tobago Tunisia Turkey Turkmenistan Uganda 90 EPA-derived CRF from Eq. (13), Full-Flight Emissions Country Ukraine United Kingdom United Rep. of Tanzania United States of America Uruguay Uzbekistan Vanuatu Venezuela Viet Nam Western Samoa Yemen Zambia Total WHO CRF from Eq. (7) EPA-derived CRF from Eq. (13), LTO Emissions FullFlight LTO 2.5% Percentile 23 9.7 Mean 57 25 97.5% Percentile 100 45 2.5% Percentile -22 2.8 Mean -13 7.2 97.5% Percentile -4.9 13 21 14 -4.1 4.2 0.6 1.7 3 0.0 0.0 0.0 4 0.1 46 0.0 3.9 0.0 0.6 4 0.0 3.3 0.2 890 120 0.1 9.9 0.0 1.5 10 0.0 9.6 0.5 2300 210 0.2 18 0.0 2.7 19 0.0 19 1 4200 12 0.0 0.6 0.0 0.0 -0.8 0.0 0.1 0.0 -390 31 0.0 1.5 0.0 0.0 -0.5 0.0 0.2 0.0 -130 56 0.0 2.6 0.0 0.1 -0.2 0.0 0.5 0.0 100 140 0.3 8.9 0.0 4.3 9.4 0.0 14 0.4 1500 34 0.0 1.2 0.0 0.1 0.2 0.0 0.4 0.0 58 Full table of calculated VSLs by country. Table A2: VSLs and non-discounted monetizations by country. GNlIcapita Country Afghanistan Albania Algeria Angola Antigua and Barbuda Argentina Armenia Australia Austria Azerbaijan Bahrain Bangladesh Belarus Belgium Belize Benin Bhutan Bolivia Bosnia-Herzegovina Botswana Brazil 2006 USS VSL 2006 US$ 990 7,010 7,160 3,780 17,060 11,710 4,950 33,010 35,810 5,390 29,410 1,240 9,760 34,450 5,870 1,330 3,740 4,300 7,350 11,740 8,810 Mean 45,000 530,000 540,000 220,000 1,700,000 1,000,000 320,000 4,400,000 5,000,000 350,000 3,700,000 58,000 770,000 4,700,000 390,000 60,000 220,000 260,000 530,000 990,000 670,000 21,000 250,000 240,000 60,000 500,000 340,000 130,000 930,000 950,000 120,000 710,000 26,000 190,000 740,000 110,000 20,000 61,000 68,000 130,000 150,000 140,000 91 ULSJ Valuation 2006 US$ 97.5% Percentile 200,000 1,700,000 1,700,000 870,000 4,800,000 3,100,000 1,200,000 11,000,000 13,000,000 1,300,000 9,700,000 250,000 2,500,000 12,000,000 1,400,000 270,000 860,000 1,000,000 1,800,000 3,100,000 2,200,000 Mean 700,000 530,000 3,900,000 590,000 15,000 1,700,000 1,300,000 4,000,000 28,000,000 1,600,000 900,000 4,700,000 10,000,000 49,000,000 14,000 61,000 270,000 63,000 780,000 50,000 5,400,000 Country Brunei Darussalam Bulgaria Burkina Faso Burundi Cambodia Cameroon Canada Cape Verde Central African Republic Chad Chile China Colombia Commonwealth of Dominica Comoros Congo Congo Democratic Republic Costa Rica Croatia Cyprus Czech Republic Denmark Djibouti Dominican Republic East Timor Ecuador Egypt El Salvador Equatorial Guinea Eritrea Estonia Ethiopia Federated State of Micronesia Fiji Finland France FYROM/Macedonia Gabon Gambia Georgia Germany Ghana Greece Grenada Guatemala Guinea ULSJ Valuation 2006 US$ VSL 2006 US$ GNI/capita 2006 US$ Ou, -Ifu 10,790 1,090 350 1,570 2,010 36,410 2,880 690 1,080 11,380 4,790 7,640 Z.7o Percentile 8uu,uuu 160,000 11,000 3,800 18,000 19,000 500,000 34,000 5,400 7,900 130,000 51,000 79,000 880,000 45,000 12,000 71,000 95,000 5,100,000 150,000 26,000 44,000 950,000 290,000 550,000 21,UUU,UUU 2,800,000 220,000 65,000 330,000 430,000 13,000,000 640,000 140,000 220,000 3,000,000 1,100,000 1,900,000 Mean 5,UUU 3,100,000 220,000 4,800 84,000 520,000 30,000,000 7,500 -34,000 250,000 220,000 63,000,000 810,000 7,490 1,150 2,480 75,000 9,700 17,000 530,000 47,000 120,000 1,800,000 240,000 540,000 4,200 350 16,000 270 9,630 16,310 25,060 21,230 36,700 2,180 6,620 1,990 6,810 4,700 5,920 13,550 630 17,930 700 1,800 86,000 140,000 260,000 200,000 490,000 13,000 48,000 12,000 45,000 30,000 36,000 99,000 3,100 160,000 3,000 8,400 750,000 1,600,000 2,900,000 2,300,000 5,100,000 100,000 450,000 93,000 470,000 280,000 390,000 1,200,000 22,000 1,800,000 25,000 49,000 2,400,000 4,600,000 7,900,000 6,300,000 13,000,000 470,000 1,600,000 430,000 1,600,000 1,100,000 1,400,000 3,700,000 120,000 5,100,000 140,000 27,000 140,000 4,000,000 850,000 25,000,000 16,000,000 20,000 850,000 900 54,000 11,000,000 190,000 33,000 19,000 1,500,000 450,000 3,240 4,310 33,410 31,950 8,520 11,050 1,100 4,130 34,410 1,270 26,410 7,650 4,270 870 16,000 20,000 410,000 390,000 45,000 72,000 3,700 17,000 430,000 3,800 280,000 37,000 14,000 1,900 170,000 250,000 4,400,000 4,200,000 630,000 910,000 44,000 240,000 4,600,000 52,000 3,100,000 550,000 250,000 33,000 730,000 1,000,000 11,000,000 11,000,000 2,100,000 2,900,000 220,000 960,000 12,000,000 260,000 8,400,000 1,900,000 990,000 170,000 470 1,400 4,700,000 89,000,000 560,000 62,000 18,000 660,000 390,000,000 100,000 9,800,000 6,600 200,000 53,000 92 Mean o,2UU,UUU , Percentile ULSJ Valuation 2006 US$ VSL 2006 US$ 97.5% Country Guinea-Bissau Guyana Honduras Hungary Iceland India Indonesia Iran Iraq Ireland Israel Italy Ivory Coast Jamaica Japan Jordan Kazakhstan Kenya Kiribati Korea Kuwait Kyrgyz Republic Lao People's Democratic Republic Latvia Lebanon Lesotho Liberia Libyan Arab Jamahiriya Lithuania Luxembourg Madagascar Malawi Malaysia Maldives Mali Malta Mauritania Mauritius Mexico Mongolia Morocco (includes Western Sahara) Mozambique Namibia Nepal Netherlands New Zealand Nicaragua /,( ( U 3,350 17,300 33,570 2,540 3,040 9,880 2,850 36,670 24,840 30,170 1,520 7,040 32,770 4,850 8,690 1,430 3,630 24,320 51,130 1,790 6,600 8,700 150,000 410,000 5,100 7,200 59,000 6,400 490,000 260,000 350,000 2,000 32,000 400,000 17,000 47,000 1,700 10,000 250,000 820,000 2,700 Mean 38,000 140,000 180,000 1,700,000 4,500,000 130,000 160,000 780,000 150,000 5,100,000 2,900,000 3,800,000 65,000 490,000 4,300,000 300,000 650,000 61,000 200,000 2,800,000 8,400,000 81,000 1,710 14,540 9,870 1,660 250 14,910 15,610 60,210 920 650 12,240 4,650 980 21,470 1,740 10,900 13,520 2,850 2,500 110,000 59,000 2,300 210 120,000 130,000 1,100,000 770 400 83,000 16,000 860 210,000 2,500 70,000 99,000 6,400 3,790 670 5,810 1,010 39,070 25,130 2,400 11,000 420 22,000 910 550,000 260,000 4,600 93 , Percentile , Mean 200,000 620,000 760,000 4,900,000 12,000,000 560,000 680,000 2,500,000 640,000 13,000,000 7,800,000 10,000,000 320,000 1,700,000 11,000,000 1,100,000 2,200,000 300,000 830,000 7,500,000 22,000,000 380,000 13,000 4,700 100,000 12,000,000 34,000 110,000,000 1,900,000 23,000,000 2,200,000 3,000,000 5,800,000 96,000,000 120,000 350,000 64,000,000 450,000 4,900,000 89,000 100 24,000,000 4,400,000 180,000 76,000 1,300,000 780,000 73,000 7,200 1,400,000 1,500,000 11,000,000 35,000 23,000 1,000,000 280,000 38,000 2,300,000 78,000 890,000 1,200,000 150,000 360,000 4,000,000 2,500,000 350,000 45,000 4,100,000 4,300,000 28,000,000 180,000 130,000 3,200,000 1,100,000 200,000 6,400,000 370,000 2,800,000 3,700,000 640,000 66,000 2,000,000 1,500,000 3,800 1,800 1,900,000 5,700,000 3,700,000 7,200 7,400 1,200,000 7,700 210,000 350,000 81,000 14,000 20,000,000 48,000 210,000 24,000 380,000 39,000 5,600,000 2,900,000 120,000 870,000 130,000 1,400,000 200,000 14,000,000 7,900,000 530,000 2,100,000 8,900 17,000 710,000 43,000,000 76,000 51,000 ULSJ Valuation 2006 US$ VSL 2006 US$ Mean Percentile Mean Niger b4u 39jU 22,UUU 1,0uuuu IOU,UUU Nigeria Norway Oman Pakistan Panama Papua New Guinea Paraguay Peru Philippines Poland Portugal Republic of Moldova Romania Russia Rwanda Saint Kitts and Nevis Saint Lucia Saint Vincent Sao Tome and Principe Saudi Arabia Senegal Serbia and Montenegro Seychelles Sierra Leone Singapore Slovakia Slovenia Solomon Islands 1,790 53,330 20,480 2,390 9,380 1,690 4,080 6,360 3,090 14,640 22,180 2,860 10,870 14,560 940 13,270 8,830 7,690 1,560 22,590 1,650 9,935 18,160 670 46,950 17,700 25,140 2,230 2,700 880,000 190,000 4,500 54,000 2,400 13,000 26,000 7,400 110,000 220,000 6,400 70,000 110,000 800 96,000 49,000 37,000 2,100 220,000 2,300 60,000 160,000 420 720,000 160,000 260,000 4,000 81,000 9,000,000 2,200,000 120,000 720,000 75,000 230,000 420,000 160,000 1,300,000 2,400,000 150,000 890,000 1,300,000 36,000 1,200,000 660,000 550,000 68,000 2,500,000 73,000 780,000 1,800,000 24,000 7,400,000 1,800,000 2,900,000 110,000 380,000 23,000,000 6,000,000 520,000 2,400,000 360,000 940,000 1,500,000 700,000 4,000,000 6,700,000 640,000 2,800,000 4,000,000 190,000 3,600,000 2,200,000 1,900,000 330,000 6,800,000 350,000 2,500,000 5,200,000 130,000 19,000,000 5,000,000 7,900,000 480,000 2,400,000 6,300,000 1,800,000 11,000,000 90,000 2,900 46,000 200,000 730,000 37,000,000 4,100,000 330,000 8,200,000 97,000,000 13,000 6,700 6,900 7,000 440 28,000,000 260,000 6,100,000 4,600 38,000 2,400,000 10,000,000 3,100,000 260 South Africa 9,090 51,000 690,000 2,300,000 1,200,000 Spain Sri Lanka Sudan Suriname Swaziland Sweden Switzerland Syrian Arab Republic Tajikistan Thailand Togo Tonga Trinidad and Tobago Tunisia Turkey Turkmenistan Uganda Ukraine United Kingdom 29,810 3,850 1,660 6,360 4,580 36,140 42,510 4,070 1,550 6,970 790 4,310 22,180 6,650 12,250 4,970 970 6,130 35,110 350,000 11,000 2,300 26,000 15,000 470,000 620,000 12,000 2,000 31,000 580 14,000 220,000 29,000 84,000 17,000 850 25,000 450,000 3,800,000 220,000 73,000 420,000 270,000 5,000,000 6,400,000 230,000 67,000 480,000 29,000 250,000 2,400,000 450,000 1,000,000 300,000 37,000 400,000 4,800,000 9,900,000 890,000 350,000 1,500,000 1,100,000 13,000,000 16,000,000 940,000 330,000 1,700,000 160,000 1,000,000 6,700,000 1,600,000 3,200,000 1,200,000 200,000 1,500,000 12,000,000 55,000,000 820,000 1,100,000 6,800 6,700 14,000,000 19,000,000 940,000 160,000 3,200,000 21,000 130 310,000 1,500,000 22,000,000 1,100,000 65,000 23,000,000 120,000,000 94 ULSJ United Rep. of Tanzania United States of America Uruguay Uzbekistan Vanuatu Venezuela Viet Nam Western Samoa Yemen Zambia Total 1,140 45,640 10,170 2,170 3,630 11,010 2,310 3,990 2,120 1,070 # 3t:$:y7jf? 1,100 690,000 62,000 3,800 10,000 71,000 4,200 12,000 3,600 1,000 VSL 2006 US$ Valuation 2006 US$ Mean Percentile 46,000 230,000 7,100,000 18,000,000 810,000 2,600,000 100,000 470,000 200,000 830,000 900,000 2,900,000 110,000 500,000 230,000 920,000 100,000 460,000 42,000 220,000 % %.1% $ Mean 75,000 830,000,000 91,000 1,000,000 230 1,300,000 1,100,000 310 940,000 22,000 2,500,000,000 Not all countries and regions are considered in Tables Al and A2. This is due to either a lack of mortality data within the WHO GBD database or a lack of economic data from the World Bank database. To maintain consistency within the analysis, values from other sources were not used. The following countries or territories have been omitted: American Samoa, Andorra, Anguilla, Aruba, Bahamas, Barbados, Bermuda, British Virgin Islands, Cayman Islands, Cook Islands, Cuba, Faeroe Islands, Falkland Islands, French Guiana, French Polynesia, Gibraltar, Greenland, Guadeloupe, Guam, Guernsey, Haiti, Hong Kong, Isle of Man, Jersey, North Korea, Lichtenstein, Macao, Marshall Islands, Martinique, Mayotte, Monaco, Montserrat, Myanmar, Nauru, Netherland Antilles, New Caledonia, Niue, Norfolk Island, Northern Mariana Island, Occupied Palestinian Territory, Palau, Pitcairn, Puerto Rico, Qatar, Reunion, Saint Helena, Saint Pierre and Miquelon, San Marino, Somalia, Svalbard, Taiwan, Tokelau, Turks and Caicos Islands, Tuvalu, United Arab Emirates, United States Virgin Islands, Wallis and Futuna, and Zimbabwe. Given the low population densities as well as geographical locations of many of these omitted countries and territories, the impact on the overall EBCA is expected to be minimal. 95 THIS PAGE INTENTIONALLY LEFT BLANK. 96 Works Cited 1. International Energy Agency, http://www.iea.org/. 2. ICAO, Environmental Report 2007," Tech. rep., International Civil Aviation Organization, 2007. 3. ICAO, Environment Report, 2010," Tech. rep., International Civil Aviation Organization, 2010. 4. IPCC, "Aviation and the Global Atmosphere," Tech. rep., Intergovernmental Panel on Climate Change, Cambridge, UK, 1999. 5. US Environmental Protection Agency. Clean Air Act. http://www.epa.gov/air/caa/. 6. EPA Heavy-Duty Highway Diesel Program, http://www.epa.gov/otaq/highway- diesel/regs/420f06064.htm. 7. International Civil Aviation Organization, Committee on Environmental Protection (CAEP 8). http://www.icao.int/environmental-protection/Pages/default.aspx. 8. FAA. Destination 2025, 2011. http://www.faa.gov/about/plans-reports/media/Destination2025.pdf. 9. Lee et al. Transport impacts on atmosphere and climate: Aviation. Atmospheric Env. 2010, 44, 4678-4734. 10. Lee, et al, 2009. Aviation and global climate change in the 2 1s' century. Atmos. Environ.. Doi: 10.1016/j.atmosenv.2009.04.024. 11. Dorbian, C.S.; Wolfe, P.J.; Waitz, I.A. Estimating the climate and air quality benefits of aviation fuel and emissions reductions. Environ. Sci. Technol. 2011, 45 (16), pp. 2750-2759. 12. US Environmental Protection Agency. The Benefits and Costs of the Clean Air Act: 1990 to 2020. Washington, DC: Office of Air and Radiation; 2011. 13. Mahashabde, A.; Wolfe, P.; Ashok, A.; Dorbian, C.; He, Q.; Fan, A.; Lukachko, S.; Mozdzanowska, A.; Wollersheim, C.; Barrett, Steven R.H.; Locke, M.; Waitz, Ian A. Assessing the environmental impacts of aircraft noise and emissions. Progressin Aerospace Sciences. 2011, 47 (1), pp. 15-52. 14. Barrett, S. R. H.; Britter, R. E.; Waitz, I. A. Global mortality attributable to aircraft cruise emissions. Environ. Sci. Technol. 2010, 44 (19), pp. 7736-7742. 97 15. Bey, I.; Jacob, D. J.; Yantosca, R. M.; Logan, J. A.; Field, B.; Fiore, A. M., et al. Global modeling of tropospheric chemistry with assimilated meteorology: Model description and evaluation. J Geophys. Res. 2001, 106 (23),073-096. 16. J. H. Seinfeld, S. N. Pandis. Atmospheric Chemistry and Physics; John Wiley & Sons, Inc.: Hoboken, NJ, 2006. 17. Wang, J.; Jacob, D.J.; Martin, S.T. Sensitivity of sulfate direct climate forcing to the hysteresis of particle phase transitions. J. Geophys. Res. 2008, 113, D 11207. 18. Martin, R.V.; Jacob, D.J.; Yantosca, Y.M.; Chin, M.; Ginoux, P. (2003), Global and regional decreases in tropospheric oxidants from photochemical effects of aerosols. J. Geophys. Res., 108, 4097. 19. Pope, C. A., 3rd; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr., Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am JRespir Crit CareMed 1995, 151, (3 Pt 1), 669-674. 20. Pope, C. A.; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D., Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA 2002, 287, (9), 1132-1141. 21. Dockery, D. W.; Pope, C. A.; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G. J.; Speizer, F. E., An association between air pollution and mortality in six U.S. cities. NEngl JMed 1993, 329, (24), 1753-1759. 22. Laden, F.; Schwartz, J.; Speizer, F. E.; Dockery, D. W., Reduction in fine particulate air pollution and mortality: Extended follow-up of the Harvard Six Cities study. Am JRespir Crit Care Med 2006, 173, (6), 667-672. 23. United States Environmental Protection Agency. Expanded Expert Judgment Assessment of the Concentration-ResponseRelationship Between PM 2 5 Exposure and Mortality. EPA, 2006. 24. Pope III, C.A.; Dockery, D.W. Health effects of fine particulate air pollution: Lines that connect. Journal of the Air and Waste Management Association, 2006, 56 (6), 709-742. 25. Brauer, M.; Avila-Casado, C.; Fortoul, T.I.; Vedal, S.; Stevens, B.; Churg, A. Air Pollution and Retained Particles in the Lung. Environ. Health Perspect.2001, 109, 1039-1043. 26. Churg, A.; Brauer, M.; Avila-Casada, C.; Fortoul, T.I.; Wright, J.L. Chronic Exposure to High Levels of Particulate Air Pollution and Small Airway Remodeling. Environ. Health Perspect.2003, 111, 714-718. 98 27. Libby, P.; Ridker, P.M.; Maseri, A. Inflammation and Atherosclerosis. Circulation.2002, 105,1135-1143. 28. Cohen, A.J.; Pope, C.A., III. Lung Cancer and Air Pollution. Environ. Health Perspect. 1995, 103, 219-224. 29. Samet, J.M.; Cohen, A.J. Air Pollution and Lung Cancer. In Air Pollution and Health; Holgate, S.T.; Samet, J.M.; Koren, H.S.; Maynard, R.L., Eds.; Academic Press: London, United Kingdom, 1999; 841-864. 30. Cohen, A.J. Air Pollution and Lung Cancer: What More Do We Need to Know?. Thorax. 2003, 58, 1010-1012. 31. Schwartz, J.; Coull, B.; Laden, F.; Ryan, L., The effect of dose and timing of dose on the association between airborne particles and survival. Environ Health Perspect.2008, 116, (1), 64-69. 32. Hedley, A.J.; Wong, C.-M.; Thach, T.Q.; Ma, S.; Lam, T.-H.; Anderson, H.R. Cardiorespiratory and all-cause mortality after restrictions on sulphur content of fuel in Hong Kong: An intervention study. Lancet. 2002, 360 (9346), 1646-1652. 33. Luttmann-Gibson, H.; Suh, H. H.; Coull, B. A.; Dockery, D. W.; Sarnat, S. E.; Schwartz, J.; Stone, P. H.; Gold, D. R., Short-term effects of air pollution on heart rate variability in senior adudts in Steubenville, Ohio. Journalof Occupationaland EnvironmentalMedicine. 2006, 48, (8), 780-788. 34. Sarnat, S. E.; Suh, H. H.; Coull, B. A.; Schwartz, J.; Stone, P. H.; Gold, D. R., Ambient particulate air pollution and cardiac arrhythmia in a panel of older adults in Steubenville, Ohio. Occupationaland EnvironmentalMedicine. 2006, 63, (10), 700-706. 35. O'Neill, M. S.; Veves, A.; Zanobetti, A.; Sarnat, J. A.; Gold, D. R.; Economides, P. A.; Horton, E. S.; Schwartz, J., Diabetes enhances vulnerability to particulate air pollution Associated impairment in vascular reactivity and endothelial function. Circulation2005, 111, (22), 2913-2920. 36. US Environmental Protection Agency Science Advisory Board Advisory on Plans for Health Effects Analysis in the Analytical Plan for EPA's Second Prospective Analysis - Benefits and Costs of the Clean Air Act, 1990-2020; EPA-SAB-COUNCIL-ADV-04-002; DC, 2004. Washington, 37. Levy, J.I.; Diez, D.; Dou, Y.; Barr, C.D.; Dominici, F. A meta-analysis and multisite timeseries analysis of the differential toxicity of major fine particulate matter constituents. American JournalofEpidemiology. 2012, accepted for publication. 38. US Environmental Protection Agency. GuidelinesforPreparingEconomic Analyses. 2010. 99 39. US Environmental Protection Agency, Office of Air and Radiation. The Benefits and Costs of the Clean Air Act from 1990 to 2020. March 2011. 40. Office of Management and Budget. CircularA-4, September 17, 2003. 41. US Department of Transportation. VSL Memorandum, March 18, 2009. 42. Heinzerling, Lisa. The Rights of Statistical People. Georgetown Law, The Scholarly Commons, 2000. 43. Integrated Environmental Health Impact Assessment System. Methods and measuresfor impact analysis. http://www.integrated-assessment.eu/guidebook/impact-analysis. 44. European Union. Guide to Cost Benefit Analysis of Investment Projects. July 2008. 45. Hileman, J.I.; Ortiz, D.S.; Bartis, J.T.; Wong, H.M.; Donohoo, P.E.; Weiss, M.A.; Waitz, I.A. Near-Term FeasibilityofAlternative Jet Fuels. RAND and Massachusetts Institute of Technology, 2009. 46. US Environmental Protection Agency, Fuels and Fuel Additives; http://www.epa.gov/oms/regs/fuels/diesel/diesel.htm. 47. Clean Diesel Fuel Alliance; http://www.clean-diesel.org/. 48. QinetiQ. Reduction of sulphur limits in aviationfuel standards.EASA, 2010. 49. Department for Transport. Britain's TransportInfrastructure- Adding Capacity at Heathrow: Decisions Following Consultation.DfT, 2009. 50. International Civil Aviation Organization. Airport Air Quality Guidance Manual, 2007. 51. Energy Information Administration. The Transition to Ultra-Low-Sulfur Diesel Fuel: Effects on Prices and Supply. EIA, 2001. 52. US Environmental Protection Agency. Regulatory Impact Analysis: Heavy-Duty Engine and Vehicle Standardsand Highway Diesel Fuel Sulfur Requirements. EPA, 2000. 53. Corbett, J.J.; Winebrake, J.J. Emissions tradeoffs among alternative marine fuels: Total fuel cycle analysis of residual oil, marine gas oil, and marine diesel oil. Journalof the Air and Waste Management Association. 2008, 58 (4), 538-542. 54. Winebrake, J.J.; Corbett, J.J.; Green, E.H.; Lauer, A.; Eyring, V. Mitigating the health impacts of pollution from oceangoing shipping: An assessment of low-sulfur fuel mandates. EnvironmentalScience and Technology. 2009, 43 (13), 4776-4782. 100 55. Henze, D. K.; Hakami, A.; Seinfeld, J. H. Development of the adjoint of GEOS-Chem. Atmospheric Chemistry and Physics, 2007, pp. 2413-2433. 56. Koo, Jamin. Adjoint Sensitivity Analysis of the Intercontinental Impacts of Aviation Emissions on Air Quality and Health. Masters Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2011. 57. Barrett, S.; et al. Guidance on the use of AEDT Gridded Aircraft Emissions in Atmospheric Models Version 2.0, FAA, Washington DC, August 17, 2010. 58. Van Donkelaar, A.; Martin, R.V.; Leaitch, W.R.; MacDonald, A.M.; Walker, T.W.; Streets, D.G.; Zhang, Q.; Dunlea, E.J.; Jimenez, J.L.; Dibb, J.E.; Huey, L.G.; Weber, R.; Andreae, M.O. 2008. Analysis of aircraft and satellite measurements from the Intercontinental Chemical Transport Experiment (1NTEX-B) to quantify long-range transport of East Asian sulfur to Canada. Atmospheric Chemistry and Physics, 8 (11), pp. 2999-3014. 59. Lee, D.S.; Fahey, D.W.; Forster, P.M.; Newton, P.J.; Wit, R.C.N.; Lim, L.L.; Owen, B.; Sausen, R. Aviation and global climate change in the 21st century. Atmospheric Environment. 2009, 43 (22-23), 3520-3537. 60. McCormick , M.P.; Thomason, L.W.; Trepte, C.R. Atmospheric effects of the Mt. Pinatubo eruption. Nature, 1995, 373, pp. 399-404. 61. Stratton, R.W.; Wong, H.M.; and Hileman, J.I. Life Cycle Greenhouse Gas emissionsfrom Alternative Jet Fuels. PARTNER, 2010. 62. Charlson, R.J.; S.E. Schwartz; J.M. Hales; R.D. Cess; J.A. Coakley, Jr.; J.E. Hansen; D.J. Hoffman. Climate forcing by anthropogenic aerosols. Science. 1992, 255, 423-430. 63. Heald, C. GEOS-Chem Optics Description, 2010. 64. Martin, S.T.; Hung H.-M.; Park, R.J.; Jacob D.J.; Spurr, R.J.D.; Chance, K.V. Chin, M. Atmos. Chem. Phys. 2004, 4, 183-214. 65. Kiehl, J.T.; Schneider, T.L.; Rasch, P.J.; Barth, M.C. Radiative forcing due to sulfate aerosols from simulations with the National Center for Atmospheric research Community Climate Model, Version 3. J. Geophys. Res. 2000, 105, 1441-1458. 66. Wiscombe, W.J.; Grams, G.W. The backscattered fraction in two-stream approximations. J. Atmos. Sci. 1976, 33, 2440. 67. Nemesure, S.; Wagener, R.; Schwartz, S.E. Direct shortwave forcing of climate by the anthropogenic sulfate aerosol: Sensitivity to particle size, composition, and relative humidity. J. of Geophysical Research. 1995, 100 (D12), 26105-26116. 101 68. Curci, G. FlexAOD, http://pumpkin.aquila.infn.it/flexaod/. 69. Adams, P.J.; Seinfeld, J.H.; Koch, D.; Mickley, L.; Jacob, D. General circulation model assessment of direct radiative forcing by the sulfate-nitrate-ammonium-water inorganic aerosol system. J. of Geo. Res. 2001, 106 (D1), 1097-1111. 70. IPCC. IPCC Third Assessment Report: Working Group I, The Scientific Basis, 2001. 71. Penner, J.E.; Charlson, R.J.; Hales, J.M.; Laulainen, N.S.; Leifer, R.; Novakov, T.; Ogren, J.; Radke, L.F.; Schwartz, S.E.; Travis, L. Quantifying and Minimizing Uncertainty of Climate Forcing by Anthropogenic Aerosols. Bulletin of the American MeteorologicalSociety. 1994, 75, 375. 72. IPCC. IPCC Fourth Assessment Report: Working Group I Report, The Physical Scientific Basis, 2007. 73. Jarzembski, M.A.; Norman, M.L.; Fuller, K.A.; Srivastava, V.; Cutten, D.R.. Complex refractive index of ammonium nitrate in the 2-20-tm spectral range. Appl. Opt. 2003, 42, 922-930. 74. General Motors Corporation, Well-to-Tank Energy Use and Greenhouse Gas Emissions of Transportation Fuels - North American Analysis, Volume 3, General Motors Corporation, Argonne National Labs, British Petroleum, Exxon Mobile, Shell, 2001; http://www.transportation.anl.gov/pdfs/TA/165.pdf (accessed October 12, 2009). 75. Hileman, J. I.; Stratton, R. W.; Donoho, P. E. Energy Content and Alternative Jet Fuel Viability Journalofpropulsion andpower.2010, 26 (6), 1184-1195. 76. Marais, K.; Lukachko, S.P.; Jun, M.; Mahashabde, A.; Waitz, I.A. Assessing the impact of aviation on climate. Meterologische Zeitschrift. 2008, 17 (2), 157-172. 77. Mahashabde, A. Assessing Environmental Benefits and Economics Costs of Aviation Environmental Policy Measures. Ph.D. Dissertation, Massachusetts Institute of Technology, Cambridge, MA, 2009. 78. Shine, K.P.; Fuglestvedt, J.S.; Hailemariam, K.; Stuber, N. Alternatives to the global warming potential for comparing climate impacts of emissions of greenhouse gases. Climatic Change. 2005, 68(3), 281-302. 79. W. D. Nordhaus. A Question of Balance: Weighing the Options on Global Warming Policies. Yale University Press, New Haven, 2008. 80. US Environmental Protection Agency. GuidelinesforPreparingEconomic Analyses. EPA, 2010. 102 81. Miller, T.R. Variations between countries in values of statistical life. Journal of Transport Economics and Policy. 2000, 34 (2), 169-188. 82. Viscusi, W.K.; Aldy, J.E. The Value of a Statistical Life: A Critical Review of Market Estimates Throughout the World (2003) Journal of Risk and Uncertainty, 27 (1), 5-76. 83. Hammitt, J.K.; Robinson, L.A. The Income Elasticity of the Value per Statistical Life: Transferring Estimates between High and Low Income Populations. JournalofBenefit-Cost Analysis. 2011, 2(1). 84. Center for International Earth Science Information Network (CIESIN), Centro Internacional de Agricultura Tropical (CIAT). Global Rural-UrbanMapping Project (GRUMP). Columbia University, 2004. 85. World Health Organization. GlobalBurden ofDisease (GBD); http://www.who.int/healthinfo/globalburdendisease/en/. 86. B. Ostro. Outdoor air pollution: Assessing the environmental burden of disease at national and local levels. Environmental Burden of Disease Series, No. 5 World Health Organization, 2004. 87. Barrett, S.R.H.; Britter, R.E.; Waitz, I.A. Global mortality attributable to aircraft cruise emissions. Environmental Science and Technology. 2010, 44 (19), 7736-7742. 88. US Census Bureau. InternationalData Base (IDB); http://www.census.gov/population/international/data/idb/informationGateway.php. 89. US Energy Information Administration; http://www.eia.gov/. 90. Argonne National Laboratory. GREET Model v1. 8A; http://greet.es.anl.gov/. 91. Gary, J.H.; Handwerk, G.E.; Kaiser, M.J. Petroleum Refining: Technology and Economics, 5 th, ed; CRC Press: Boca Raton, Fl, 2007. 92. Department Of Treasury Internal Revenue Service. Publication946: How to Depreciate Property.Department of Treasury, Washington, DC, 2009. 93. Allaire, D.; Willcox, K. Surrogate modeling for uncertainty assessment with application to aviation environmental system models. AIAA Journal.2010, 48 (8), 1791-1803. 94. Jun, Mina. Uncertainty Analysis of an Aviation Climate Model and an Aircraft Price Model for Assessment of Environmental Effects. Masters Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2007. 103 95. Saltelli, A.; Ratto, M.; Andres, T.; Campolongo, F.; Cariboni, J.; Gatelli, D.; Saisana, M.; Tarantola, S. Global Sensitivity Analysis: The Primer; Wiley and Sons: England, 2008. 96. ExxonMobil Aviation. World Jet Fuel Specifications: 2005 Edition. Exxon Mobil, U.S., 2005. 97. Defense Energy Support System. Petroleum Quality Information System, 2008. 98. Phelan, P. 'Green' fuel sparks off fears of engine failure. FlightInternational,17-23 August 1994. 99. Coordinating Research Council, Inc. CRC Fuel Lubricity Workshop. CRC, Alexandria, VA, 1996. 100. Coordinating Research Council, Inc. CRC Fuel Lubricity Workshop. CRC, Alexandria, VA, 1996. 101. Baniszewski, D. "Question about cost of CI/LI additive for JP-8." Correspondence with Hileman, J. November 16, 2010. 102. Martin, S.T.; Hung H.-M.; Park, R.J.; Jacob D.J.; Spurr, R.J.D.; Chance, K.V. Chin, M. Atmos. Chem. Phys. 2004, 4,183-214. 103. Liao, H.; Seinfel, J.H.; Adams, P.J.; Mickley, L.J. Global radiative forcing of coupled tropospheric ozone and aerosols in a unified general circulation model. J of Geo. Res. 2004, 109 (D16207), doi: 10. 1029/2003JD004456. 104. Sausen, R.; Isaksen, I.; Grewe, V.; Hauglustaine, D.; Lee, D.S.; Myhre, G.; Kohler, M.O.; Pitari, G.; Schumann, U.; Stordal, F.; Zerefos, C. Aviation radiative forcing in 2000: an update of IPCC (1999). Meteorol. Zeit. 2005, 114, 555-561. 105. The World Bank. GNIper capita, PPP; http://data.worldbank.org/indicator/NY.GNP.PCAP. CD. 106. Interagency Working Group on Social Cost of Carbon United States Government. Appendix 15A. Social Cost of CarbonforRegulatory Impact Analysis Under Executive Order 12866, 2010. 107. Schwartz, S.E. Heat capacity, time constant, and sensitivity of Earth's climate system. J Geosphys. Res. 2007, 112 (24), D24S05. 108. Jun, Mina. Uncertainty Analysis of an Aviation Climate Model and an Aircraft Price Model for Assessment of Environmental Effects. Masters Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2007. 104 109. Atmosphere graphic: http://www.aboutaerosols.com/environment-atmosphere.php. 110. A380 graphic: http://news.cnet.com/i/ne/p/2007/1026airbus 1_440x291.jpg. 111. Ashok, Akshay. The air quality impact of aviation in future-year emissions scenarios. Masters Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2011. 105