Human Health Impacts of High Altitude Emissions Sebastian D. Eastham
Human Health Impacts of High Altitude Emissions
by
Sebastian D. Eastham
B.A., M.Eng., University of Cambridge 2011
Submitted to the Department of Aeronautics and Astronautics in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in Aeronautics and Astronautics at the
Massachusetts Institute of Technology
June 2015
© Massachusetts Institute of Technology 2015. All rights reserved.
Signature of author:
Department of Aeronautics and Astronautics
April 13, 2015
Certified by:
Professor Steven Barrett
Associate Professor of Aeronautics and Astronautics
Thesis Supervisor and Committee Chair
Certified by:
Professor Ronald Prinn
TEPCO Professor of Atmospheric Science
Thesis Committee Member
Certified by:
Professor Susan Solomon
Ellen Swallow Professor of Atmospheric Chemistry and Climate Science
Thesis Committee Member
Certified by:
Professor David Keith
Professor of Applied Physics, Harvard University
Thesis Committee Member
Accepted by:
Paulo C. Lozano
Associate Professor of Aeronautics and Astronautics
Chair, Graduate Program Committee
Page 1 of 159
Human Health Impacts of High Altitude Emissions
by
Sebastian D. Eastham
Submitted to the Department of Aeronautics and Astronautics on April 13, 2015 in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Aeronautics and Astronautics
Abstract
Millions of deaths worldwide are attributed annually to exposure degraded surface air quality and UVinduced skin cancer. However, the focus has been on surface emissions, and the contribution of high altitude emissions to these issues is rarely examined. In this thesis, potential links are investigated between high altitude emissions and damages or benefits to human health via photochemical effects.
Changes in population exposure to fine particulate matter, ozone and UV-B radiation resulting from current and future high altitude emissions are calculated, applying epidemiologically-derived impact functions to estimate resultant mortality and morbidity.
A stratospheric extension is developed for the widely-used tropospheric model GEOS-Chem, which has been shown to accurately model tropospheric conditions and used in simulations of remote and urban pollution. This extended model, the GEOS-Chem UCX, can propagate a stratospheric perturbation through to a tropospheric impact, including shortwave UV fluxes, long-lived species, stratospheric water chemistry and high altitude aerosols.
This model is employed to estimate the impacts of reversing 1 K of global warming using stratospheric sulfate aerosol injection. In total, it is projected that 85,000 additional premature mortalities would occur in 2040 due to particulate matter exposure, but that reduced ozone loading would prevent 64,000 mortalities worldwide. Aerosol injection also results in a 5.7% reduction in the global ozone column and a
3.0% increase in surface UV-B, which could cause 3,700 additional melanoma mortalities per year. By comparison, surface air quality and UV-B impacts due to aviation emissions are found to have resulted in
16,000 premature mortalities globally in 2006, of which 450 occurred in North America. Ozone exposure contributes 43% of this total. The increase in tropospheric ozone due to aviation emissions is found to have prevented 390 skin cancer mortalities in 2006.
This thesis quantifies the photochemical mechanisms connecting future and proposed high altitude emissions schemes to human health impacts and provides an estimate of mortality and morbidity attributable to aviation and sulfate aerosol injection.
Thesis Supervisor: Steven Barrett
Title: Associate Professor of Aeronautics and Astronautics
Page 2 of 159
Acknowledgements
This work would not have been possible without the support and guidance of my advisor, Professor
Steven Barrett. I would also like to thank my thesis committee members Professors David Keith, Susan
Solomon and Ronald Prinn, for their invaluable guidance and advice in the course of completing this thesis.
Page 3 of 159
Contents
Page 5 of 159
Page 6 of 159
Page 7 of 159
1 Background
Anthropogenic emissions have long been recognized as potentially harmful to human health. Concerns date as far back as 1273, when the English King Edward I passed a law prohibiting the burning of coal due to its suspected health effects (1). As technology has advanced, so has the scale at which humanity can impact the environment, and the potential for unchecked emissions to damage not only the environment but also human health. Early examples include London smog and photochemical smog due to industrial emissions (1, 2). More recent examples include acid rain caused by sulfur emissions, ocean acidification due to carbon dioxide and, more generally, climate change due to greenhouse gas emissions
(3). As the cause and the effect become more spatially and temporally separated, however, uncertainty regarding the strength or even the existence of a causal link grows, exemplified by the ongoing debate surrounding anthropogenic climate change.
In the last century, the domain into which we emit species directly has expanded beyond the surface of the earth. These new emissions take place in a chemical environment fundamentally different to the nearsurface atmosphere, with impact pathways that may have been of negligible importance to surface-level emissions. While investigation of the atmospheric impacts resulting from aircraft emissions have been a topic of research (4, 5), there are few quantitative estimates of the human health response to high altitude emissions. Of those that exist, all but two have been concerned with mortalities due to surface PM
2.5
impacts (6–11). In this research, the pathways by which high altitude emissions can impact human health are quantified, including both mortality and non-mortality impacts via ozone, PM
2.5
and UV-B exposure.
1.1
Human health and the atmosphere
A 2012 study identified 67 risk factors to which global mortality in 2010 could be attributed using globalscale modeling (12). Three fell into the category of “Air Pollution”: ambient particulate matter pollution; household air pollution from solid fuels; and ambient ozone pollution. Of these three factors, the first and
Page 8 of 159
last can be directly attributed to anthropogenic emissions and are clustered under the term “surface air quality” (SAQ). Particulate matter at the sub-2.5 µm diameter level (PM
2.5
) and ozone (O
3
) therefore constitute the most significant link between surface air quality changes and human health impacts (13).
Degradation of surface air quality has also been linked to chronic bronchitis and increases in hospital admissions, along with other non-fatal outcomes (14–16).
In addition to PM
2.5
and ozone exposure, high altitude emissions pose an additional health risk. The longer lifetime and greater interaction with the stratosphere exhibited by high altitude emissions allows them to have significant impacts on the surface shortwave radiation budget. The potential for ozone depletion represented by NO x
emissions from stratospheric supersonic aircraft was identified in the early
1970s (4, 17), and stratospheric ozone depletion is linked to skin cancer incidence through increased tropospheric UV-B radiation (18). Future high altitude anthropogenic emissions have the potential to also affect this balance.
1.2
Impact mechanisms
The mechanisms linking anthropogenic emissions to surface air quality and UV-B impacts can be broadly separated into direct perturbations and climate feedbacks. Photochemical mechanisms can be explicitly simulated by a CTM, and involve small changes around an established climatological state. Climate feedbacks, however, can affect these direct perturbations by modifying the climatological state. This study focuses on the former, with some direct parameterization of the latter. By evaluating the former in a
CTM without feedbacks, the partial derivatives of impacts with respect to emissions are estimated independent of how those emissions might affect the climate.
Page 9 of 159
1.2.1
Photochemical mechanisms
Figure 1. Representation of primary versus secondary impact pathways for surface air quality impacts.
The simplest impact mechanism is direct emission of a harmful species, where exposure to the emissions themselves has negative health consequences. In the case of particulate matter, these emissions are known as primary PM
2.5
. Although these are significant contributors to the overall burden of disease from anthropogenic emissions (19), PM
2.5
emitted at high altitude is removed through precipitation processes and therefore has only a limited capacity to directly affect
Figure 2. Rates of vertical transport (shading) and contours of precipitation (black lines) from GEOS-5 data.
surface chemistry (20). Figure 2 shows GEOS-5
estimated vertical transport velocities, with red for upwelling and blue for downwelling. Contours of rate of precipitation are shown as black lines. Aerosols are rapidly removed within the contours, resulting in low survival rates for transport of primary PM
2.5
emitted at high altitudes.
Page 10 of 159
However, as long as they are present these aerosols can also have local chemical and radiative effects, discussed below.
Where primary emissions can often include some fraction of pre-formed PM
2.5
, surface ozone is typically a secondary pollutant. Ozone is naturally produced throughout the troposphere by equilibrium cycling of
NO and NO
2
(NO x
), along with descent of air masses containing stratospheric ozone. Similarly, cycling between OH and HO
2
(HO x
) radicals has a similar effect, and the two sets of cycles can interact via species such as HONO or HNO
3
. Anthropogenic ozone is produced when primary emissions, typically of
NO x
or unburnt hydrocarbons (HC), chemically alter the background steady state, resulting in an increased or decreased ozone concentration. The exact effect of a NO x or HC emissions is a non-linear function of their local mixing ratios (21). These null cycles produce no net ozone in the troposphere but establish a steady-state background level dependent on factors including the local spectrum and intensity of UV radiation, air pressure, temperature and humidity. Since NO x
, HC and ozone lifetimes typically increase with altitude, emissions in the upper troposphere or lower stratosphere have the capacity to produce significant effects over a large domain when compared to emissions near to the surface. An increase in NO x
concentrations throughout the mid-troposphere will result in increased ozone concentrations as well, which will mix vertically to the surface at varying rates depending on the season.
Similarly, changes in HO x
concentrations can also affect ozone concentrations. Existing aircraft emissions have been estimated to reduce the lifetime of tropospheric methane through enhanced OH production
(22), a process which itself involves a photochemical feedback as oxidation of methane provides additional water for production of OH.
The effect of anthropogenic emissions on ozone changes at higher altitudes as ozone-destroying catalytic cycles begin to dominate the null cycles. Stratospheric ozone is controlled by a combination of cycles involving nitrogen, chlorine, bromine and hydrogen among others (23). Whereas tropospheric NO x
Page 11 of 159
emissions typically increase local ozone mixing ratios, stratospheric NO x
emissions have the opposite effect due to enhanced catalytic conversion by increased local NO x
(4). Emissions of aerosol into the stratosphere, meanwhile, will enhance chlorine activation rates, resulting in faster halogen cycling (24).
On the other hand the addition of aerosol could result in enhanced sequestration of NO x
into reservoir species, reducing or even reversing ozone depletion due to elevated chlorine concentrations (25). The stratospheric ozone layer absorbs incoming UV radiation, thereby controlling both surface UV exposure and the photochemical balance of the atmosphere. Changes in stratospheric ozone will also affect the overall source of tropospheric ozone from stratosphere-troposphere exchange (26). Furthermore, stratospheric ozone is crucial to the temperature structure of the atmosphere, and loss of ozone is linked to lower stratospheric temperatures as heating by ozone photolysis is reduced (27).
The above mechanisms can all change the UV budget for the troposphere and surface, resulting in adjustments to photochemical balance. The contribution of photochemical processes to smog formation has been a subject of research for over 60 years (1). The exact response of surface air quality and DNA damage are dependent on the specific spectrum of the impacts. Increases at longer wavelengths could, for example, favor faster cycling of NO x
, potentially increasing ozone concentrations due to the welldocumented photochemical relationship (28). However, the complex dependence of surface concentrations on photochemistry means that a change in UV influx could have a range of impacts.
As surface and free tropospheric photochemistry adjust to the changing conditions in the upper troposphere and stratosphere, so too will formation of secondary PM
2.5
. This “secondary” pathway is
illustrated in Figure 1, which shows one mechanism by which surface sulfates can be enhanced by high
altitude NO x
emissions. Local concentrations of PM
2.5
are dictated by a mix of transport, aerosol thermodynamics and chemistry, wet scavenging, dry scavenging and oxidation rates, amongst other factors. Although direct and precursor emissions are key to understanding concentrations of any pollutant,
Page 12 of 159
secondary inorganic PM
2.5
is highly sensitive to the balance between its precursors: nitrogen oxides
(NO y
), sulfur oxides (SO x
) and ammonia (NH
3
) (28, 29). The resultant ions - nitrate (NO
3
), sulfate (SO
4
2) and ammonium (NH
4
+ ) – exist in an approximate state of thermodynamic equilibrium dependent on temperature, pressure, humidity and the presence of other aerosols such as mineral dust and sea-salt.
Furthermore, changes in the atmosphere’s oxidative capacity and in the availability of nitrogen compounds can affect formation of secondary organic aerosols (SOA). For example, a study of PM
2.5
composition during high pollution events in China in 2013 found that VOC emissions accounted for an average of 25-30% of local PM
2.5
via formation of SOA.
1.2.2
Climate feedbacks
Anthropogenic emissions result in a climate forcing. The best known climate response to anthropogenic emissions is increasing temperature through the modified radiative balance of the atmosphere and heat uptake by the ocean. High altitude emissions have the capacity to change the temperature distribution throughout the atmosphere, as is the case with large volcanic eruptions (30). Since almost all chemical reactions are temperature dependent, a change in the local air temperature has the capacity to accelerate or slow down the local rate of reaction. Changes in temperature can also affect partitioning of species between gas and aerosol phase, with important consequences for particulate matter concentrations in particular (13). Higher temperatures are also linked to greater ozone production. These changes in temperature are also projected to modify both the pattern and the rate of precipitation worldwide (31).
Precipitation is a removal mechanism for both aerosols and soluble gases throughout the troposphere, and is the dominant removal process for several species including atmospheric sulfates (28, 32). Changes in precipitation could therefore have wide-ranging consequences on atmospheric chemistry.
As the earth warms, increased evaporation is expected be countered by increased precipitation and relative humidity will remain unchanged (33). This implies that tropospheric specific humidity is
Page 13 of 159
expected to increase, as the saturation water vapor of air rises with temperature. The increased atmospheric loading of H
2
O will increase the free tropospheric photochemical sink of ozone, resulting in greater OH but potentially reduced ozone mixing ratios depending on the local loading of volatile organic compounds (VOCs) (13). However, the stratospheric implications of water vapor are different.
Notwithstanding coupling between stratospheric water vapor and climate change (34), changes in stratospheric water vapor affect formation of stratospheric aerosols and HO x
-catalyzed ozone depletion.
These changes are also expected to affect cloud formation. Clouds reflect both solar and longwave radiation, and alter the radiative balance of the planet (35). They thereby influence the availability of shortwave radiation for photochemistry in the troposphere (36). Chemical processing by clouds, for example the activation of bromine from reservoir species (37), provides another path for photochemical feedbacks, and for which estimation of cloud location and thickness is important.
An additional significant consequence is modification of the patterns of atmospheric transport. The rate of exchange of air mass between the stratosphere and troposphere (stratosphere-troposphere exchange, STE) is expected to increase as a result of climate change (38), thereby increasing the degree of coupling between stratospheric and tropospheric responses. The general circulation of the atmosphere in a broader sense is also expected to respond to climate change, which could influence factors such as stagnation and ventilation of populated areas (13). A key factor could be the response of the Hadley cell to climate change, which could expand poleward but weaken in the process, altering patterns of mass transport globally (39).
1.2.3
PM
2.5
exposure
If long-term PM
2.5
concentrations rise in a populated area, the increased population exposure can have chronic health impacts. A large number of epidemiological studies have addressed the potential link between exposure to particulate matter and loss of life, but an early, significant cohort study was the 1993
Page 14 of 159
Harvard Six Cities study (40). This study found that exposure to PM
2.5
was positively correlated with both lung cancer and cardiopulmonary mortality. Analysis of a second, larger study by the American Cancer
Society, which ran from 1982 to 1998 in 51 cities, supported these findings (41), confirming that mortality was more strongly correlated with PM
2.5
than with larger-diameter particulate matter such as
PM
10
. Earlier cross-sectional studies (33) and later cohort studies supported this finding, while confounding variables such as sex and body mass index were investigated and discounted. A metaanalysis performed in 2013 which included these results along with those from analyses and reanalyses of
18 other studies results from found that long-term exposure to PM
2.5
increased all-cause mortality at a rate of 6% per 10 µg/m 3 , but that the association was stronger for cardiovascular mortality (11% per 10
µg/m 3 ) than for nonmalignant respiratory disease (3% per 10 µg/m 3 ) (42). This meta-analysis included studies from a wide geographic area, supporting its use in global-scale assessments.
Studies of the relation between PM
2.5
composition and its overall toxicity have thus far been inconclusive, although black carbon is often found to be more toxic than other compounds (43–45). A review by the
EPA found that there was insufficient evidence to support the application of differential toxicity when calculating health impacts (46). In light of this, only the total mass of PM
2.5
is typically considered for the purpose of calculating health impacts.
Long term exposure to a number of other species has been linked with premature mortality due to ambient exposure, including NO
2
, SO
2
and elemental carbon (42). However, since these are both markers of combustion and either precursors to or constituents of PM
2.5
and ozone, they are not considered in this thesis.
The European “Externalities of Energy” (ExternE) project included an analysis of non-fatal health impacts associated with surface air quality degradation (14, 15), which has since been used in assessments of health and economic damages related to air quality for both Europe and China (47, 48). Outcomes of
Page 15 of 159
exposure to PM
2.5
range from minor restricted activity days (MRADs) through increased bronchodilator usage up to hospital admissions and congestive heart failure.
1.2.4
Ozone exposure
Although several studies have been conducted regarding short-term ozone exposure and mortality (16, 49,
50), only one major study has addressed the link with chronic exposure (51). They find that ozone exposure exacerbates pre-existing respiratory conditions, resulting in a 4% increase in respiratory diseaserelated mortality per 10 ppbv of long-term ozone exposure. The results are based only on the 51-city
American Cancer Society (ACS) study and may not be applicable globally. However the Jerrett et al parameters have been used in ozone-attributable mortality estimates in several global studies (12, 52–54).
Although short-term and long-term mortality due to ozone exposure have been calculated together in at least one study (55), the calculations are usually separated due to the possibility of double-counting.
As for PM
2.5
, non-fatal health impacts associated with ozone were also collated for the ExternE project, and have been used in a global assessment of ozone impacts due to climate change and emissions in 2050
(55). These impacts again range from usage of a bronchodilator up to hospital admissions, although new cases of chronic bronchitis due to ozone were not calculated due to lack of confidence in the published estimates.
1.2.5
UV exposure
Anthropogenic emissions, particularly those of CFCs, have been conclusively linked to depletion of the stratospheric ozone layer which shields the earth’s surface from damaging UV radiation (23). However, there is less confidence in the link between UV exposure and skin cancer incidence. Although exposure to excessive UV-B radiation is known to cause skin and DNA damage (56), low UV-B exposure has also been linked to skin cancer due to lack of vitamin D which is formed through photosynthesis in human skin (57–59). A model was developed in 1996 which estimates skin cancer incidence and mortality due to
Page 16 of 159
changes in global UV (60). The model projects incidence of squamous cell carcinoma (SCC), basal cell carcinoma (BCC) and cutaneous malignant melanoma (CMM).
SCC and BCC are collectively referred to as carcinomas or non-melanoma skin cancers (NMSC), whereas CMM is referred to as melanoma. Based on a review of German medical data, SCC makes up approximately 80% of all NMSC, with BCC making up 20% and other NMSC less than 1% (61). Under the Slaper model, SCC is estimated based on accumulated life-long UV exposure, whereas BCC and
CMM are calculated based on recent, specific burns. NMSC is the most common of all cancers amongst white-skinned populations, estimated to occur at age-standardize rates of 100/100,000 (62). However, reliable incidence data is rare, possibly due to the low mortality rate when compared to CMM. CMM is much less common, but is difficult to treat and often fatal. It is therefore responsible for the majority of all skin cancer mortalities (63). A WHO report in 2006 estimated that CMM is responsible for 1.6% of global skin cancer incidence but 80% of global skin cancer mortality (64).
1.3
Atmospheric modeling
Global chemistry-transport models (CTMs) typically have a single domain of interest, defined by the range of pressures between which they are considered accurate. Tropospheric models, such as GEOS-
Chem v9 (65), employ a simplified stratosphere as a boundary condition for the troposphere, exploiting the slow rate of stratosphere-troposphere exchange. In the case of GEOS-Chem v9, stratospheric ozone is estimated using the linearized ozone (Linoz) method (66), providing an estimate of monthly-mean concentrations. Stratospheric chemistry is estimated based on archived monthly mean production rates and loss frequencies from NASA’s Global Modeling Initiative (GMI) code (67, 68). UV fluxes into the troposphere are then estimated using an archived climatology. This configuration has been extensively used for investigations of the atmospheric response to emissions and natural phenomena (69–72).
However, the use of simplified parameterizations of the stratosphere limits the scope of investigations,
Page 17 of 159
since coupling between tropospheric and stratospheric processes cannot be accurately modeled. Other models using this approach include TOMCAT (73), MOZART-4 (74) and CAM-chem (75).
Stratospheric models, meanwhile, take the opposite approach, often using a simplified troposphere to provide boundary conditions for the stratosphere. The AER model, for example, prescribes tropospheric washout processes using fixed removal rates and exploits the relative zonal uniformity of the stratosphere to collapse the stratosphere into two dimensions, allowing rapid analysis at the expense of some accuracy
(32, 76). Other full 3-D models of the stratosphere such as SLIMCAT (73) and CMAM (77) use similar approaches to ensure a reasonable lower boundary condition, but cannot directly model tropospheric chemistry or feedbacks.
Some models, such as MOZART-3 (78), TM-5 (79) and Oslo CTM-3 (80) model both domains. These models are typically used sparingly, since modeling both the troposphere and stratosphere is computationally intensive. MOZART-3 was superseded by MOZART-4, which has more accurate tropospheric chemistry but does not include stratospheric chemistry. Oslo CTM-3 has separate tropospheric and stratospheric mechanisms, neither of which are as comprehensive as the tropospheric mechanism in GEOS-Chem. Other unified models include ULAQ-CTM, WACCM and EMAC (81).
Page 18 of 159
Table 1. Comparison of existing models to GEOS-Chem UCX, which was developed as part of this thesis. Models with distinct chemical mechanisms for the troposphere and stratosphere are shown here with two counts (tropospheric/stratospheric). Although
GEOS-Chem can be run with secondary organic aerosols (SOA), these are not modeled in the default setup and the SOA mechanism is in the process of being rebuilt.
GC UCX 132 344 89 2 2 7 Y Y/N 2 2 T, S
GEOS-Chem (65) 104 240 55 2 2 7 Y Y/N 2 0 T
CAM-chem (75) 117 263 69 2 2 4 Y Y 4 4 T
MOZART-3 (78) 108 236 71 2 2 0 Y Y 4 4 T, S, M
MOZART-4 (74) 85 161 39 2 2 4 Y Y 4 0
TM-5 (79) 42 64 16 0 0 0 Y N 0 0
T
T, S
GMI (82, 83) 46 116 38 1 0 0 N N 0 4 T, S
OSLO CTM3 (80) 97 88/112 17/47 4 4 2 Y Y 8 3 T, S
TOMCAT (84)
SLIMCAT (73)
CMAM (77)
41 93 25 0 0 0 N N 0 0 T
43 109 29 0 0 0 N N 0 3
44 93 34 0 0 0 N N 0 1
S
S, M
1.4
Existing estimates of global mortality attributable to surface emissions
1.4.1
Surface air quality impacts
Lim et al (2012) found that, in 2010, 3.2 million deaths could be attributed to ambient particulate matter,
3.5 million to household air pollution, and 150,000 to ambient ozone pollution, making no distinctions regarding anthropogenic or natural emissions. Their finding for PM
2.5
has been supported by several studies, but their assessment of ozone mortalities is likely an under-estimate. A multi-model ensemble analysis of the impacts of anthropogenic outdoor air pollution found that anthropogenic ozone exposure alone is responsible for a further 470,000 mortalities per year (53). They also estimated that 2.1 million of the death attributable to PM
2.5
were attributable to anthropogenic emissions. A study of global, regional
Page 19 of 159
and megacity premature mortality, meanwhile, estimated that 2.2 million deaths globally were attributable to anthropogenic PM
2.5
and 770,000 attributable to anthropogenic ozone (54). These studies, along with
other relevant estimates, are summarized in Table 2.
Table 2. Existing estimates of premature mortality attributable to anthropogenic emissions. All studies shown used the same
PM
2.5
and O
3
CRFs (51, 85) except for Lim et al, which used an integrated risk function for PM
2.5
which has since been published
(86). *Lim et al estimates are for all PM
2.5
and O
3
, not just anthropogenic.
Study
Lim et al.
2012 (12)
Anenberg et al. 2010 (52)
Silva et al.
2013 (53)
Lelieveld et al. 2013 (54)
Fang et al.
2013 (87)
Method
CTM (TM-5) modeling constrained by satellite observations. Estimates include natural emissions.
CTM (MOZART-2) modeling of preindustrial and present-day (2000) concentrations.
Multi-model ensemble analysis subtracting preindustrial from modern-day estimates.
GCM-CTM (EMAC) modeling of preindustrial and present-day (2005) concentrations.
GCM (AM3) modeling of preindustrial and present-day concentrations including climate change and CH
4
.
PM
2.5
(thousands)
3,200*
3,700
2,100
2,100
1,600
O
3
(thousands)
152*
700
470
770
380
Surface air quality is impacted by these emissions through both direct and photochemical processes.
Direct transport of, for example, soot emitted from coal-fired power plants, can be breathed in and thereby result in public health damages. When enhanced by background fog formation, this is the principle behind the formation of “London-type” industrial smog, such as was reported in London in the
1950s (2). On the other hand, formation of photochemical smog (“Los Angeles-type”) is the result of NO x and SO x
interacting with solar UV radiation, resulting in ozone episodes which can be dangerous to human health (1). These phenomena interact with local topography, climate and other factors such as volatile organic carbon (VOC) loading. Although local emissions can have long-range impacts, such as by formation of the intermediate-lifetime NO x
reservoir PAN, the turbulent nature of the lower atmosphere results in a general tendency towards rapid washout of pollutants.
Page 20 of 159
1.4.2
UV impacts
Slaper et al estimated in 1996 that, due to implementation of the chlorofluorocarbon (CFC) emissions regulations negotiated under the Montreal Protocol and Copenhagen Amendments (collectively the
Vienna Convention), 132 excess cases of skin cancer per million population per year in the USA will be avoided in 2030, rising to 6,500 in 2100. A later study improved on this model, using gridded skin reflectance data (88), multi-decadal GCM forecasts and a global skin cancer mortality database, compensating for under-reporting, to estimate the total global impact of the Montreal Protocol and amendments (89). They estimate that 2 million skin cancer cases per year globally will be avoided in
2030, corresponding to a 16% increase in incidence and a 10-20% increase in UV exposure. However, the authors are now questioning the relative importance of ambient UV-B compared to behavior (Arjan van
Dijk, personal communication).
1.5
Existing estimates of surface impacts due to high altitude emissions
Most research into high altitude emissions and health impacts – not including research into nuclear weapons – has focused on aviation. A review by Brasseur et al in the late 1990s identified many of the mechanisms linking aircraft emissions to atmospheric impacts (5), and this work was later extended by
Lee et al to include more recent developments (90). However, a prevailing assumption has been that surface air quality impacts due to aviation would be dominated by local, rather than high altitude, emissions impacts (91). This assumption was challenged by an investigation which found that, of 12,600 premature mortalities per year attributable to aviation emissions, ~8,000 were due to cruise-level operation (6), and a recent study using nested modeling at the global, regional and urban scales supported this finding, while also estimating that ozone impacts due to aviation incur an additional 2,100 mortalities per year (11). However, there is disagreement over the total attributable impact, with other studies finding that PM
2.5
due to modern-day aviation emissions has only a small or statistically insignificant impact on surface air quality (10, 92).
Page 21 of 159
There has been little work to quantify the UV impact of high altitude emissions since research into supersonic travel slowed down in the 80s and 90s. A 2009 SM thesis (93) used scaling arguments based on EPA recommendations to estimate the change in non-melanoma skin cancer (NMSC) mortality in
1992 and 2002 due to subsonic aviation, along with the change in BCC and SCC incidence. The study found that approximately 20 NMSC mortalities were prevented throughout the US in 2002. However, this study used a reduced order model to estimate aviation’s impacts, and did not include aerosol effects on ozone or scattering. The study was also limited to the USA and did not consider global impacts.
1.6
Thesis objectives and contributions
This thesis has three objectives:
1.
Development of a state-of-the-art chemical transport model linking high altitude emissions to surface impacts;
2.
Estimation of the sensitivity of human health to current and proposed high altitude emissions; and
3.
Identification and quantification of the relative contributions of different chemical and meteorological processes to these impacts
The scope of this thesis is limited to photochemical effects which link emissions to chronic health impacts for a given climatological state. It does not address meteorological, dynamical or climatological feedbacks. The first contribution of this thesis is the development of a unified tropospheric-stratospheric chemical transport model, enabling simulation of the chemical response of the atmosphere to perturbations from the surface to the stratopause. The second contribution is to provide an estimate of total morbidity and mortality impacts attributable to aviation emissions via both surface air quality and
UV exposure, as well as those which might be expected to result from deployment of sulfate aerosol engineering. The modeling work and analytical methods also provides the tools for future investigations into other high altitude emissions and first order estimates of their potential impacts.
Page 22 of 159
2 Approach
Health impact estimates were produced in three steps. First, an atmospheric model was extended so that it would be capable of modeling emissions and their impacts in a consistent fashion from the stratosphere through the troposphere and onwards to the surface. Simulation scenarios were then developed to model high altitude emissions and converted into a set of input conditions for the model. Finally a postprocessing model was developed to estimate human health impacts based on the model output.
2.1
Atmospheric modeling
GEOS-Chem is a chemical transport model in use by over 70 groups worldwide to assess chemical feedbacks throughout the troposphere for a broad variety of scenarios and situations (65, 69–71).
However, it uses a linearized model of stratospheric chemistry, and changes to the stratospheric chemical environment cannot evolve or be propagated to the troposphere. Tropospheric perturbations are also rapidly removed upon reaching the stratosphere through relaxation to an archived chemical state. This
limitation is addressed in chapter 3 of this thesis through development of the GEOS-Chem Unified
Chemistry eXtension (UCX). The UCX adds a stratospheric chemistry model based on the existing GMI stratospheric chemistry mechanism (68) which allows chemical coupling between the stratosphere and troposphere to be captured. Additionally, investigation of chemical feedbacks between stratospheric ozone and aerosols and tropospheric photochemistry is made possible through extension of photolysis to the stratopause, calculation of J-values for shorter wavelengths and improved modeling of high-altitude
aerosols. Development of this model is outlined in chapter 3.
When being driven by GEOS-5 meteorology, GEOS-Chem typically uses a reduced vertical resolution in the stratosphere with a total of 47 atmospheric layers from the surface to the lower mesosphere. Since this work is concerned with stratospheric chemistry, the full vertical resolution of 72 layers is used in all
Page 23 of 159
simulations. For the sake of computational efficiency, calculations are performed at a global horizontal resolution of 4°×5°.
2.2
Investigations
Two scenarios are investigated which span future and existing high altitude emissions. Both are simulated using the GEOS-Chem UCX developed as part of this thesis, and the resultant atmospheric impacts are analyzed in the context of establishing an atmospheric mechanism linking each health risk factor to the original emission. These atmospheric impacts are linked to health impacts using functions derived from
epidemiological data which are described in section 2.3.
2.2.1
Stratospheric aerosol injection
Recent studies estimate that Earth’s surface will warm by 2-4°C by 2100 relative to 2010 (94), and expected impacts include sea level rise, changing precipitation patterns and migration of tropical diseases
(95). Existing environmental health risks will also respond, as chemical and dynamical atmospheric systems adjust to the changing conditions. Temperature, precipitation, humidity and other meteorological variables control the rate at which harmful species are produced from anthropogenic emissions, along with rate at which they are removed.
Climate change therefore has the potential to either improve or worsen surface air quality even in the absence of emissions regulation (13).
In isolation, increasing temperatures are expected to worsen ozone pollution but the sign of the PM
2.5
response is not definitely known (13). Enhanced sulfate formation due to faster oxidation is countered by preferential partitioning of nitrate aerosols into the gas phase as evaporation increases (29). Meanwhile increasing precipitation is expected to enhance aerosol washout and thereby improve surface air quality, notwithstanding changes in precipitation patterns and distributions. Other variables such as venting, mixing layer depth and cloud cover, although they are expected to affect surface air quality, are more uncertain.
Page 24 of 159
Global-scale climate engineering could slow or prevent some aspects of climate change (96). The most widely investigated approach is stratospheric aerosol injection (SAI), the deliberate thickening of the existing stratospheric aerosol layer. When Mt. Pinatubo erupted in 1991, the increased aerosol optical thickness resulted in global average surface temperatures for 1992 being reduced by 0.5°C (30, 97). By emitting sulfate into the stratosphere on a continuous basis, SAI would ideally reduce or reverse impacts of climate change. However, it will also introduce impacts of its own. The surface cooling achieved by the Mt. Pinatubo, for example, was accompanied by 6% global ozone column depletion (97) as the sulfate
these effects schematically. Enhancement of the stratospheric sulfate layer, shown in the figure as effect
[1], would increase planetary albedo, decrease the longwave radiation penetrating to the troposphere [2], potentially inducing a surface temperature reduction (98). However, it would also affect stratospheric chemistry by providing a greater reaction surface area [3] (99), impacting both the flux of UV to the troposphere [4] and the amount of ozone available for transport to the surface [5]. Sulfate would also be transported to the surface directly [6], and the reduced temperatures and precipitation anticipated to result from geoengineering (100) would impact the rate of removal from the atmosphere [7] which could in turn increase surface sulfate loading.
Page 25 of 159
Figure 3. Schematic of a global sulfate aerosol engineering strategy, along with some potential intended and unintended impacts.
SAI represents an unprecedented high-altitude intervention. Previous high altitude impacts have been side effects of some other industry, such as supersonic aviation in the case of NO x
or domestic refrigeration in the case of CFCs. With SAI the intent is to modify the stratosphere, bringing about an order of magnitude change in its sulfur loading and sustaining in the long term a state which has only been achieved for a year or two at a time by natural processes such as volcanic eruptions.
Previous studies have evaluated the sensitivity of climatological (100–102) and chemical (24, 25, 103) metrics to SAI. In this work, the GEOS-Chem UCX is applied to quantify the sensitivity of air-quality and UV-B exposure related health outcomes to SAI in the context of ongoing anthropogenic climate change.
Page 26 of 159
2.2.2
Subsonic aircraft emissions
Mortalities attributable to civil aircraft cruise emissions via surface PM
2.5
and ozone have been quantified in other studies, including the original Barrett et al study in 2010 (6) (PM
2.5
), a study on ultra-low-sulfur jet fuel (7) (PM
2.5
), and a recent study using multi-scale modeling (11) (PM
2.5
and ozone). Two other studies have investigated this problem using coupled global chemistry-climate models (9, 10). Skin cancer impacts, meanwhile, were estimated for the US in a 2009 S.M. thesis (93). However, there is as yet no global estimate of the combined PM
2.5
, ozone and UV impacts of aviation emissions.
Several studies have been published investigating spatial and temporal attribution (8, 104) in addition to possible mechanisms for these impacts (7, 20, 92, 105, 106). Whitt et al showed that vertical mixing timescales from cruise altitude to the surface are significantly longer than average tropospheric washout timescales outside of the tropics, suggesting that primary particulate matter from aircraft engines is not responsible for increases in surface PM
2.5
. Lee et al found that nighttime NO x
chemistry in hemispheric winter may be a key pathway for cruise aviation impacts to affect surface air quality, but did not calculate associated health impacts.
Using the GEOS-Chem UCX with a Federal Aviation Authority (FAA) emissions inventory, the year
and seasonal responses to identify the mechanism linking aviation emissions to surface-level impacts.
This chapter therefore provides two contributions; a single combined estimate of global mortality and non-mortality aviation impacts via changes in PM
2.5
, ozone and UV-B, and the mechanisms by which these impacts are incurred.
Page 27 of 159
2.3
Calculation of health impacts
All health impacts are calculated in three steps:
1.
Simulate local conditions under each scenario;
2.
Multiply calculated concentrations or incident radiation by the total affected population to determine exposure; and
3.
Convert to an incidence rate using an exposure/concentration response function (ERF/CRF).
Three health risk factors are considered: particulate matter (PM
2.5
), ozone (O
3
) and biologically-active ultra-violet radiation (UV-B). The population in a given grid cell is estimated based on the LandScan gridded population data product for 2012 (107), which is defined at a resolution of 30″×30″. Each cell is assigned to a country based on the Global Rural Urban Mapping Project national identifier grid (108).
The local population is scaled to the target year based on country-specific projections from the UN
Population Prospects Division (109). Since some ERFs are calculated for specific age fraction only, projected age distributions are retrieved from the same UN database.
The derivation and meaning of ERFs is outlined below, followed by the health outcomes assessed in the thesis. The terms ERF and CRF are used interchangeably throughout the thesis.
2.3.1
Exposure response functions
Exposure or concentration response functions are applied to estimate health outcomes resulting from a change in population exposure to a given risk factor (PM
2.5
, ozone or UV-B). The specific health outcomes are derived separately for each risk factor, based on a review of recent epidemiological studies, meta-analyses, and other studies of global and regional health impacts. Both morbidity (non-fatal, e.g. annual rate of hospital admission) and mortality outcomes are calculated. Mortality impacts are calculated in units of “premature mortality”, where 1 premature mortality means that 1 additional person per year
Page 28 of 159
will die as a result of the modeled change in exposure. This approach has been used to calculate global impacts attributable to both specific policies and background air quality, and is the WHO methodology for calculation of air quality health impacts (7, 12, 52–54). A review of the existing literature shows that three functional forms are used for ERFs relating PM
2.5
and ozone exposure to mortality: log-linear (9,
52–54), log-log (6, 110) and, more recently, an integrated risk function (IRF) (10, 12) developed by
Burnett et al (86). The log-linear form has also been approximated with a linear function for computational convenience (87). The log-linear form requires only an estimate of the gradient of mortality rate with respect to exposure, whereas the IRF and log-log functions require more extensive calibration. The parameters for the log-linear function can therefore be taken directly from epidemiological studies, providing additional flexibility. This form is used to calculate the central estimates of PM
2.5
and ozone mortality in this study, with parameters outlined for each risk factor below.
Sensitivity to possible exposure thresholds and alternative CRFs is explored in the context of mortality
mortality to ozone exposure have been developed (16, 111, 112), there currently exists only one estimate of chronic mortality impacts of ozone exposure. As such, sensitivity to the choice of CRF for ozone is not quantified.
Some studies also report the number of years of life lost (YLL) by multiplying the number of mortalities in each age bracket by the remaining time in the reference-standard life expectancy for that age bracket
(113). Prior to the 2010 Global Burden of Disease assessment, 3% time-discounting rates were also recommended to increase the weighting applied to deaths earlier in life. Globally, each mortality due to respiratory disease or cardiovascular disease results in an average of 20 YLL, compared to 24 YLL for skin cancer (114). These estimates are approximately 2 times greater than those calculated based on the old method, as used by Fang et al and Lelieveld et al when calculating global mortality impacts attributable to anthropogenic air pollution (54, 87). As premature mortalities are more frequently reported
Page 29 of 159
than YLL and have a consistent standard, only premature mortalities are reported in this study for mortality impact assessments.
Where necessary, baseline incidence rates in this thesis are taken from the WHO mortality and burden of disease estimates for 2012 (114). To estimate the mortality attributable to a specific pathway, some studies employ a counterfactual lower threshold exposure below which it is assumed that there is no negative health effect (12, 110). Others calculate the difference in estimated mortality between two scenarios (52, 53). Inclusion of a threshold value is estimated by Anenberg (52) to reduce estimated mortality attributable to each factor by 30%. Use of a lower threshold value reflects a lack of observational data regarding health impacts at low concentrations. As there are is limited observational data for PM
2.5
exposure below approximately 5 µgm -3 , application of a linear model beyond this point extrapolates beyond the bounds of the epidemiological studies on which the ERF is based. However, reviews have consistently found no evidence to support the use of a lower threshold when calculating health impacts (15, 115). A recent study by Pope et al which compared exposure due to ambient PM
2.5
to that resulting from cigarette smoke found that the ERF is instead likely to be steeper at low concentrations
(116).
2.3.2
PM
2.5
exposure
Annual average PM
2.5
exposure is estimated based on the total mass of inorganic aerosol (sulfate, ammonium and nitrate), carbonaceous aerosol (organic and black/elemental carbon) and fine mineral dust. Secondary organic aerosol is included in organic carbon based on the standard GEOS-Chem mechanism (117). Cardiovascular mortalities are calculated using the results of the Hoek meta-analysis
(42), which found an 11% increase in mortality (a reference relative risk of 1.11) due to cardiovascular disease per additional 10 µg/m 3 . Although the ACS cohort reanalysis estimates of cardiopulmonary disease and lung cancer mortality (85) have been extensively employed in the literature (see section
Page 30 of 159
1.4.1), the Hoek meta-analysis includes both the ACS results and those from studies outside the US
including Asia. The central impact estimate is calculated using a log-linear ERF with the Hoek relative risk estimate. It is therefore a more appropriate estimator of global-scale damages than other, localized
ERFs (40, 85), while being more recent than the WHO method (110) and more readily applicable than the
cardiovascular disease only, rather than cardiopulmonary disease as a whole. I am also not accounting for lung cancer or respiratory mortality. The relative risk of cardiovascular disease due to PM
2.5
exposure as calculated by Hoek et al is smaller than the risk of cardiopulmonary disease calculated based on the ACS study alone; Lelieveld et al estimate the relative risk from the ACS study for cardiopulmonary mortality to be 12.9% per 10 µg/m 3 PM
2.5
.
The WHO 2012 Global Burden of Disease database is used to estimate baseline mortality incidence rates at the national level (114). Cardiovascular disease is classified as stroke, ischemic heart disease, hypertensive heart disease, cardiomyopathy, myocarditis and endocarditis. This set was chosen based on the diseases considered for inclusion by other studies (12, 52–54), limited by the scope of the Hoek et al
ERF which is applicable only to cardiovascular rather than cardiopulmonary disease. The sensitivity of calculated PM
2.5
impacts to different choices of ERF, health outcomes and exposure threshold
Non-fatal impacts of PM
2.5
exposure are calculated according to exposure response functions (ERFs) collated by the ExternE project (14, 15), which have been used in regional and global analyses of health and economic damages related to air pollution (47, 48).The most recently published estimates for each of these ERFs, from the ExternE 2005 Methodology Update, is applied to the annual mean change in PM
2.5
to determine non-fatal health outcomes of each emissions class. ERFs which were considered to be appropriate only for “sensitivity analyses” by the authors are not included here. The ERFs are listed in
Page 31 of 159
a 1998 study of childhood asthma (118) (about 20%), and for established adult asthmatics (assumed to be
4.5% of the population above 20) (47). Age brackets for affected populations are matched to within 5 years to those given in ExternE. Since many of the ERFs are defined based on PM
10
exposure, it was assumed that 60% of PM
10
is PM
2.5
, which is considered a typical ratio by the ExternE authors. Restricted activity days due to PM exposure are calculated either by the combination of “work loss days” (valued in the ExternE project at €82 per day) and “minor restricted activity days” (€38 per day) or by direct calculation of restricted activity days (€130 per day). This analysis uses the estimate resulting from the latter calculation.
Consistent with these studies, all non-mortality ERFs are assumed to be linear with no threshold, whereas mortality impacts are assumed to be log-linear. Although unit cost data is available for Europe and China
(48), economic impacts are not evaluated. Impacts are grouped by severity as follows, based on the order of magnitude of the cost of a single case in the European Union: bronchodilator usage, minor impacts
(symptom days and restricted activity days), hospital admissions and chronic bronchitis.
Table 3. Non-fatal health outcomes of PM
2.5
exposure. Relevant populations are given in parentheses; asterisks denote outcomes relevant to asthmatics only. 95% confidence intervals are given in parentheses. Impacts are grouped by severity based on the estimated unit cost in Europe.
Outcome
Bronchodilator usage (5-14)*
Bronchodilator usage (20+)*
Lower respiratory symptoms (5-14)
Lower respiratory symptoms (15+)
Restricted activity days (15-64)
Cardiac hospital admissions (all ages)
Respiratory hospital admissions (all ages)
Chronic bronchitis (25+)
Unit cost
(€
2000
)
Exposure-response function
(cases yr -1 µg -1 m 3 PM
2.5
)
1 3.00×10 -2
1 1.52×10 -1
38 3.10×10 -1
38 2.17×10 -1
130 9.02×10 -2
2,000 7.23×10 -6
2,000 1.17×10 -5
190,000 4.42×10 -5
(-1.15×10
(-1.52×10
(1.53×10 -1
(2.50×10 -2
(7.92×10
(3.62×10
(6.38×10
(-3.17×10
-1
-1
-2
-6
-6
-6
, 1.77×10
, 4.62×10
, 4.62×10 -1
, 4.05×10 -1
, 1.01×10
, 1.09×10
, 1.72×10
, 9.02×10
-1
-1
-1
-5
-5
-5
)
)
)
)
)
)
)
)
Page 32 of 159
2.3.3
Ozone exposure
Ozone exposure is calculated based on the six-month ozone season average population exposure in each grid cell. Ozone season is identified as the six continuous months which show the highest average exposure in the year. This reflects the approach used elsewhere and by which the original relative risk estimate was calculated (51, 52, 54). As a consequence, changes in ozone concentrations outside of the local ozone season will not affect calculated exposure, based on the assumption that health impacts are correlated with peak exposure rather than accumulated exposure. Based on the Jerrett et al study, a reference relative risk of 4% ( RR
R
= 1.04) increase in mortality due to respiratory disease per 10 ppbv increase in mixing ratio is applied with a log-linear ERF. Respiratory disease is assumed to consist of asthma and chronic obstructive pulmonary disease (COPD). This is a broader classification than that used by Lim et al (12) but narrower than that used by Lelieveld et al (54) and Silva et al (53). The aim of this thesis is to evaluate the health impacts resulting from a change in the atmospheric steady-state. Therefore, although short-term mortality exposure response functions have been calculated for day-to-day variations in ozone exposure (16, 50, 112), they are not applied in this study. However, for comparison with
impacts. To avoid double-counting mortality impacts, the acute mortality outcomes of ozone are not included. The listed ERFs were used in a study of ozone impacts due to climate change (55) and are derived from work on impacts of ozone in Europe and China. Model ozone mixing ratios were converted to mass concentrations using the ExternE ozone conversion factor (1 ppbv = 1.997 µ gm -3 ) for consistency.
As for non-fatal PM
2.5
exposure impacts, outcomes are grouped by case severity.
Table 4. Non-fatal health outcomes of ozone exposure. Relevant populations are given in parentheses; asterisks denote outcomes relevant to asthmatics only. 95% confidence intervals are given in parentheses.
Page 33 of 159
Outcome
Bronchodilator usage (15-64)*
Lower respiratory symptoms (cough) (5-14)
Lower respiratory symptoms (other) (5-14)
Minor restricted activity days (15-64)
Respiratory hospital admissions (65+)
2.3.4
UV-B exposure
Unit cost
(€
2000
)
Exposure-response function
(cases yr -1 µg -1 m 3 ozone)
1 7.30×10 -2 (-2.55×10 -2 , 1.57×10 -1 )
38 9.30×10 -2
38 1.60×10 -2
(-3.3×10 -4 , 2.22×10 -1 )
(-4.3×10 -2 , 8.10×10 -2 )
38 1.15×10 -2
2,000 1.25×10 -5
(4.4×10 -3 , 1.86×10 -2 )
(-5.00×10 -6 , 3.00×10 -5 )
A scaling approach is adopted to estimate changes in skin cancer mortality and incidence rates due to high altitude emissions. The existing dose-response function derived by Slaper et al (60) to calculate scenario incidence rate y s
of melanoma in a cohort of age a is 𝑎 𝑦 𝑠
(𝑎) ∝ ∑ 𝐷 𝑠
(𝑥) × Φ(𝑥) 𝑐−1 × (𝑎 − 𝑥) 𝑑−𝑐 𝑥=0 where Φ(𝑥) is the accumulated UV dose up to age x , D
S
is the annual average dose of UV-B and c and d are disease-specific constants. D s
is calculated based on weighting of UV exposure by the SCUP-h action spectrum (56). If D s
is held constant, as in the case of a generational change in annual average UV, this relation can be reduced to 𝑦
𝑆
∝ 𝐷 𝑐
𝑆
In spite of being responsible for less than 2% of all skin cancers, around 80% of skin cancer mortalities are the result of cutaneous malignant melanoma (59, 63). The WHO mortality database reports only aggregate skin cancer mortality. In this study, skin cancer mortality impacts are calculated by scaling this value under the assumption that all skin cancer mortalities are due to melanoma, for which the coefficient c is 0.6±0.4. For a baseline dose rate D
0
, multiplying the baseline skin cancer mortality rate by
(
𝐷
𝑆
𝐷
0
) 𝑐
− 1 therefore yields an estimate of skin cancer mortalities attributable to a scenario. While SCC and BCC are responsible for 20% of total global skin cancer mortality, the factor c is greater for both
Page 34 of 159
diseases. Using only the CMM value of the parameter c will therefore result in a lower-bound estimate of mortalities due to a change in the average UV exposure.
Although a function has been developed for squamous and basal cell carcinomas (otherwise known as non-melanoma skin cancer, or NMSC), the WHO mortality database does not carry incidence data for these diseases. Worldwide incidence of NMSC is not well catalogued, as discussed by Lomas et al (62), with widespread under-reporting due to, for example, different reporting standards, lack of awareness in less developed regions, and the possibility of treatment of small tumors before registration even occurs.
Absolute estimates of NMSC incidence and mortality are therefore not included in the calculation of skin cancer mortality impacts. However, the percentage change in the skin cancer incidence rate is estimated for each grid cell using the Slaper et al relation. The power factor c is taken to be 1.4±0.4 for BCC and
2.5±0.7 for SCC, as stated in the original CRF. An approximate estimate of the global average increase in the incidence rate for each cancer is then produced by population-weighting the percentage change in incidence.
Of the different disease ERFs evaluated in the ExternE project, the unit cost of chronic bronchitis is relatively rare but expensive, costing 95 times more than a single hospital admission. Meanwhile the total treatment cost of skin cancer for the US alone has been estimated at $8.1 billion per year as of 2015, with
4.9 million adults per year being treated (119). Although I do not attempt to calculate the cost of high altitude emissions in this thesis, the methods shown here provide a starting point for future research into the associated economic impacts.
Page 35 of 159
3 Development of a unified tropospheric-stratospheric atmospheric chemical transport model
This chapter is structured as follows: section 3.1 outlines the changes and extensions which have been applied to GEOS-Chem with the aim of producing an accurate, unified model of the troposphere and
these extensions (with supporting validation and intercomparison) is to enable an existing and extensively validated community tropospheric chemical transport model (CTM) to be used to investigate the link between stratospheric changes, previously beyond the scope of GEOS-Chem’s abilities, and the tropospheric chemistry already accurately modeled by GEOS-Chem. By coupling tropospheric and stratospheric chemistry, the model can then be applied to calculate the health impacts of high altitude emissions.
3.1
New chemistry and physics
The UCX extends the chemistry mechanism to include reactions relevant to the stratosphere. As shown in
with 34 photolytic decompositions. These were identified by comparing reactions and processes already present to those included in the GMI stratospheric chemistry mechanism (68). Reaction constants were updated in line with JPL 10-06 (120).
The UCX includes expanded gas-phase chemistry in both the troposphere and stratosphere. Unlike the base model, the UCX explicitly models atomic oxygen in the ground state form O( 3 P) as well as the most common excited form O( 1 D). The base model neglects them as intermediates due to their short lifetimes in the troposphere. Although these species still exhibit short lifetimes in the stratosphere, their importance in correctly modeling stratospheric chemistry result in the need for these reactions to be explicitly considered. The UCX also explicitly models atomic hydrogen (H) and nitrogen (N) as active species.
Page 36 of 159
The UCX combines GEOS-Chem’s bromine mechanism (37) with an adapted version of GMI’s chlorine mechanism. Chlorine is transported to the stratosphere in long-lived organic species such as CFCs and
HCFCs, followed by the release of active chlorine and subsequent sequestration in reservoir species
(ClONO
2
and HCl). Halon species H-1211, H-1301 and H-2402 are added as organic sources of bromine.
Heterogeneous halogen chemistry is also modeled. Dry deposition and wet scavenging of HCl is modeled by assuming similar behavior to HBr, using a large Henry’s law coefficient to yield near-100% uptake.
The base version of GEOS-Chem used a modified version of the Fast-JX v6.2 photolysis rate solver
(121), which efficiently estimates tropospheric photolysis. It uses the wavelength bands from the older
Fast-J tropospheric photolysis scheme and does not consider wavelengths shorter than 289 nm, instead assuming they are attenuated above the tropopause. However, these high-energy photons are responsible for the release of ozone-depleting agents in the stratosphere as well as key ozone-forming reactions such as photodecomposition of molecular oxygen. DNA damage is also caused by UV radiation with these wavelengths. The standard Fast-JX model (122) addresses this limitation by expanding the spectrum analyzed to span 18 wavelength bins covering 177-850 nm, extending the upper altitude limit to approximately 60 km. Fast-JX v7.0a is therefore incorporated into GEOS-Chem UCX. Fast-JX includes cross-section data for many species relevant to the troposphere and stratosphere. However, accurately representing sulfur requires calculation of gaseous H
2
SO
4
photolysis, a reaction which is not present in
Fast-JX but which acts as a source of sulfur dioxide in the upper stratosphere. The mean cross-section between 412.5 and 850 nm is estimated at 2.542×10 -25 cm 2 (123). Photolysis of ClOO and ClNO
2
is also included in the UCX, given their importance in catalytic ozone destruction, using data from JPL 10-06
(120). Fast-JX v7.0a also includes a correction to calculated acetone cross sections. Accordingly, where hydroxyacetone cross-sections were previously estimated based on one branch of the acetone decomposition, a distinct set of cross sections from JPL 10-06 are used.
Page 37 of 159
The base version of GEOS-Chem uses satellite observations of total ozone columns when determining ozone-related scattering and extinction. The UCX allows either this approach, as was used for the production of the results shown, or can employ calculated ozone mixing ratios instead, allowing photolysis rates to respond to changes in the stratospheric ozone layer. For the scenario analyses in
chapters 4 and 5, online ozone mixing ratios were used.
To capture seasonal ozone depletion (23), eight new heterogeneous reactions are modeled with uptake coefficients and reaction probabilities from GMI (68, 124) and JPL 10-06 (125). Heterogeneous polar stratospheric cloud (PSC) reactions are considered to be pseudo-first-order, limited to a minimum reactant lifetime of 1 ms. Application of a minimum reactant lifetime is recommended by Kirner et al (124) as it prevents overshoot while permitting rapid halogen activation upon formation of a suitable reaction surface.
3.1.1
Stratospheric H
2
O
Given the complexity of the tropospheric hydrological cycle, H
2
O was left as a passive tracer throughout the troposphere, with mixing ratios derived from GEOS-5 specific humidity fields. However, within the stratosphere H
2
O is now treated as a chemically-active advected tracer. The UCX can therefore respond to events such as dehydration through sedimentation of ice PSCs. The model can be initialized based on the existing meteorological data if distributions from a previous simulation are not available, to ensure reasonable initial conditions. Alternatively, this feature can be deactivated entirely in the model options.
3.1.2
Aerosols
Two new aerosols are implemented, split into stratospheric particulate (solid) and liquid aerosol (SPA and
SLA). The former consists of type Ib and type II PSCs, made up of nitric acid trihydrate (NAT) and ice.
The latter covers all stratospheric sulfate aerosols, ranging from H
2
SO
4
liquid binary solutions (LBS) to
Page 38 of 159
supercooled ternary solution (STS) with parameterized uptake of HNO
3
, ClNO
3
, HOCl, HCl, BrNO
3
,
HOBr and HBr (126).
Stratospheric aerosol formation is modeled on a recent implementation in ECHAM-MESSy (124).
Available H
2
SO
4
is calculated by assuming that all stratospheric sulfates take the form of sulfuric acid, using a phase partitioning parameterization (127). This results in gas phase H
2
SO
4
above approximately
35 km and condensed liquid aerosol below. Interpolation is applied over 0.01 μPa to prevent instantaneous conversion between aerosol and gas phases resulting from small changes in temperature, and only gas-phase H
2
SO
4
is considered to be available for photolysis. Uptake of HNO
3
and trace halogens into the liquid aerosol is calculated according to Carslaw et al (126) based on gridbox mean aerosol properties (128). As the temperature falls, HNO
3
uptake is promoted, resulting in formation of
STS. If the temperature falls below the HNO
3
vapor saturation temperature over NAT, calculated using the Hanson and Mauersberger relation (129), HNO
3
freezes to form NAT. This process is modeled by calculating NAT formation prior to calculating HNO
3
uptake into STS, removing HNO
3
from the liquid into the solid phase. NAT forms either heterogeneously on existing solid PSCs or homogeneously with a supercooling criterion of 3 K below the vapor saturation temperature. Aqueous-phase reaction rates are calculated based on JPL 10-06 (120, 125, 130).
As reported by Kirner et al (124), observational studies have found that NAT particles do not form until a supercooling of 2-3 K below the saturation vapor temperature is achieved. This is likely because homogeneous NAT formation takes place via condensation of nitric acid dehydrate (NAD), which occurs
2-3 K below the NAT saturation temperature, and subsequent metastable conversion to NAT (131).
However, laboratory studies have shown that homogeneous NAT formation requires temperatures to remain continuously below 190 K for at least a day (132), although if temperatures fall below the frost point then heterogeneous nucleation on ice will occur instead. Since the model does not track air masses and only bulk, equilibrium aerosol properties are calculated, it is not possible to tell whether this criterion
Page 39 of 159
is satisfied. Based on these arguments, Buchholz concluded that homogeneous formation is not a significant pathway for NAT formation in the polar stratosphere (133). Homogeneous NAT nucleation was therefore disabled in the validation simulations.
Stratospheric aerosols are modeled in thermodynamic equilibrium. Physical properties are recalculated based on local temperature, pressure and mixing ratios. If no particles already exist in a given grid cell, a
20% supersaturation ratio is required before solid PSC formation is permitted, based on laboratory and modeling studies (134–136). This approximates a model of PSC formation whereby ice particles form homogeneously, followed by NAT condensing onto the surface. When calculating surface reaction probabilities (γ), the surface is assumed to be NAT if there is any NAT present in the grid cell. A kinetic aerosol growth parameterization, which will allow aerosol evolution and persistence, is to be implemented in future work. However, the current parameterization is fast and allows calculation of approximate surface area densities for heterogeneous chemistry while also permitting PSC denitrification and dehydration.
Effective aerosol radii are estimated for optical depth calculations, allowing stratospheric aerosols to affect both stratospheric and tropospheric photochemistry. Scattering and absorption parameters for SLA are based on stratospheric background sulfate, with irregular ice cloud properties used for SPA. Relevant optical data are already present in GEOS-Chem.
The new stratospheric aerosol formation mechanism replaces ISORROPIA II above the tropopause, such that ammonium is ignored and allowed to advect freely if it reaches the stratosphere. High altitude emissions from aviation and volcanoes, which were simplified or removed above the tropopause in the base model, now take place with no restrictions.
Gravitational settling of stratospheric soot and sulfates is calculated using Stokes’ law with a slip correction factor. Trace species taken up into aerosols are also sedimented. SPA are sedimented using a
Page 40 of 159
trapezoidal scheme which accounts for non-uniform vertical particle number density profiles within individual gridboxes by estimating the location of aerosol number density maxima and minima (133).
3.1.3
Surface emissions and boundary conditions
Emissions of biogenic bromine species were implemented in GEOS-Chem prior to the UCX (37). A fixed global surface mixing ratio boundary condition is applied for N
2
O, CFCs, HCFCs, halons, OCS and longlived organic chlorine species, using 500 pptv for OCS (32) and monthly WMO values (137) for the others.
The base model contains a methane emissions inventory for use in offline methane-only simulations with prescribed OH (138–142). This inventory is now used in place of the fixed zonal average methane fields previously employed by GEOS-Chem. The estimated annual methane emissions fluctuate depending on soil wetness, but the total emissions are approximately 470 Tg/yr before soil uptake is taken into account.
Alternatively, setting a switch in the UCX input menu will replace methane emissions with a surface boundary condition, ensuring that the atmospheric methane mixing ratios stay close to projected values at the expense of model flexibility.
3.1.4
Mesospheric treatment
The chemical mechanism is limited to 0.1 hPa, preventing explicit calculation of mesospheric chemistry.
Species with super-stratospheric sinks may therefore form non-physical reservoirs, which could disrupt chemistry elsewhere. To prevent this, the UCX includes a simple high-altitude NO y
mechanism. Based on zonal mean O( 3 P), O( 1 D) and J-rates taken from a 2D model (32), NO x
and N
2
O are destroyed, producing molecular nitrogen and removing active nitrogen above the stratosphere. Similarly, H
2
SO
4
photolysis to convert mesospheric sulfates to sulfur dioxide is simulated by propagating the H
2
SO
4
J-rate upwards from the stratopause in a separate calculation. For all other species, a relaxation to climatology is applied throughout the mesosphere, as was previously employed in both the mesosphere and stratosphere.
Page 41 of 159
Lyman-Alpha line photochemistry and mesospheric chemical sources such as solar proton events and photodissociation of molecular nitrogen are not considered. This may result in an underestimation of upper stratospheric NO x
(143), but modeling of such effects is left as a future research aim.
3.2
Evaluation
To verify the ability of the UCX to accurately model Earth’s atmosphere chemistry, results from four simulation years are presented for both models. A comparison is first performed between base GEOS-
Chem and the UCX to demonstrate that implementation of stratospheric chemistry has not impaired tropospheric modeling (which has already been extensively validated). GEOS-Chem UCX’s ability to reproduce the atmosphere is then assessed when compared to other models and historical observations.
Finally, source gas stratospheric lifetimes and their correlations with active species are assessed and again compared to observation data and other stratospheric models.
3.2.1
Model setup
Both the base and extended model were initialized with identical atmospheric loading of common species.
For new species, initial distributions were taken from the AER 2-D chemistry transport model (32). The only exceptions were H
2
O, which was initialized according to meteorological data, and stratospheric NO y and sulfates, which were initialized from 2-D data in the UCX. Methane was left as a prescribed field for
the base model, while the UCX employed the methane boundary condition described in section 3.1.3.
Both models were driven using GEOS-5 meteorological inputs on the same underlying model grids with identical emission inventories for common species. Simulations ran for 2004-2007 at a horizontal resolution of 4°×5°. This resolution was chosen to minimize run time, in light of the eight-fold simulation time increase resulting from each doubling of spatial resolution and the additional doubling in simulation time required due to CFL constraints. Although an exhaustive timing analysis was not performed, the models were run on physically identical hardware and the UCX simulation completed in approximately
Page 42 of 159
160% the time of the base simulation without code optimization. This level of performance suggests that maintaining a single unified version of GEOS-Chem is practical (rather than long-term use of branched unified and tropospheric versions). In light of this work, the UCX has now been adopted into the main model and released to the community in an update to v10-01 of GEOS-Chem.
3.2.2
Comparison to GEOS-Chem v9-01-03
Historically, global CTMs have been compared based on the estimated mean tropospheric hydroxyl radical (OH) concentration. An approach from the literature is used (144), weighting grid box OH concentrations by air mass to calculate mean OH for each of 12 regions. These results are compared to
climatology (145) and a multi-model analysis (146). The results are shown in Figure 4.
Figure 4. Comparison of air mass-weighted model OH to results from a multi-model mean (146) and a reference climatology
(145). The vertical limits on each box range from 0 to 2.5×10 6 molec/cm 3 on a linear scale. Data for both versions of GEOS-
Chem are annual means for 2007.
Page 43 of 159
The base and extended model OH are equal to within 4% in all 12 regions. Global mean OH and estimated CH
4
lifetimes against destruction by OH were calculated using air-mass weightings (144), averaging daily means from the target year. The relevant reaction rate is calculated based on JPL 10-06.
UCX increased global mean OH from 11.9×10 5 to 12.0×10 5 molec/cm 3 . The primary OH generation mechanism is more detailed, with OH production via O( 1 D) + H
2
O calculated explicitly rather than assuming photochemical steady-state. The change in atmospheric OH results in increased atmospheric oxidizing power, reducing mean CH
4
lifetime from 7.27 to 7.20 years compared to Spivakovsky’s 8.23 years. Naik et al. (146) found a multi-model mean global OH of 11.12±1.6×10 5 molec/cm 3 . Although the
UCX results are within this range, already high OH concentrations have been slightly increased by the implementation of online stratospheric chemistry. Moving to the regional distribution, the UCX has increased mean OH in all 12 regions by 1-4%. This change is greatest in the extreme southern latitudes.
An analysis of trends in CH
3
CCl
3
(147) estimated that southern hemispheric (SH) OH concentrations are
14±35% greater than those in the northern hemisphere (NH). The base model finds the difference to be -
25.2%, compared to -23.7% in the UCX and -28% from the multi-model analysis. The UCX therefore does not affect the pre-existing problem of models overestimating relative NH to SH OH production. It should be noted that these results are sensitive to the chosen meteorological data and sampling period.
A central goal of this modeling work has been to ensure that the ability of GEOS-Chem to model the troposphere is not compromised by the addition of stratospheric chemistry. To test this, the mean mixing ratio from the simulation period was calculated for each surface grid cell in GEOS-Chem and the UCX, isolating the largest change. Of the 53 tracers present in the base model, the absolute mean difference is within 5% for most tracers. The exceptions are nitrogen species and bromine species. Peroxypropionyl nitrates (PPN) and nitrate aerosols (specifically the NO
−
3
ion) show maximum variations between 5 and
10%, likely the result of changes in the photolysis code resulting in adjusted NO x
partitioning. However,
Page 44 of 159
the balance of inorganic aerosols is maintained, with variations of both NH
+
4
and SO
2−
4
below 5%. Of the bromine species, Br and HBr variations are between 5 and 10%, while BrNO
2
and BrONO
2
, which are sensitive to both nitrogen and bromine partitioning, show maximum variations of 10.6 and 22.4% respectively. There are two likely causes of this. Firstly, the changes in OH and photolysis rates affect the partitioning of both bromine and nitrogen, which coupling species will be particularly sensitive to.
Secondly, implementation of the UCX included a fix for the CH
3
Br boundary condition which increases the available bromine at the surface by approximately 7-8 pptv.
3.2.3
Global stratospheric chemistry
The ozone layer is a key feature of the stratosphere dependent on a large number of stratospheric processes including halogen cycles, aerosol formation and short-wavelength photochemistry. I therefore
use it to demonstrate the improved stratospheric modeling of the UCX. Figure 5 compares column ozone
estimates from the base model’s monthly-mean relaxation scheme, the UCX online calculations and
observations by the Total Ozone Mapping Spectrometer (TOMS), respectively. Figure 6 shows the zonal
mean and range for each in 2007, highlighting the extent of the correlation between base and UCX results at midlatitudes and suggesting that the new online chemistry reliably replicates results previously achieved through relaxation to a known climatology. However, during Southern hemispheric winter the
UCX exhibits formation of an Antarctic ozone hole which is not well replicated by the base model. Ozone depletion over the Antarctic influences the rest of the southern hemisphere after the breakdown of the polar vortex each spring. The chemistry local to the ozone hole is explored more thoroughly in section
3.2.4. The overall column ozone discrepancy has been reduced globally from 9.9% to 3.6% through the
Page 45 of 159
Figure 5. Zonal-mean column ozone for 2004-2007 from the base model (top), UCX (middle) and TOMS data (bottom). Model results are shown in terms of odd oxygen to allow fair intercomparison. The dashed line corresponds to Neumayer station, from
which ozonesonde measurements were used in section 3.2.4.
Following Emmons et al (74), ozonesonde data was obtained from the World Ozone and Ultraviolet radiation Data Center (WOUDC, retrieved 5 th September 2013 from http://www.woudc.org
) and compared to model results. Results were binned by latitude based on launch location, using the same boundaries as for the OH analysis. Vertically, samples were averaged in four regions defined by pressure boundaries. Three were tropospheric, using the same boundaries as in the MOZART-4 evaluation at ±100 hPa around 900 hPa, 650 hPa and 400 hPa. The final bin was chosen to demonstrate stratospheric ozone behavior, bounded at 30 and 5 hPa. The samples were weighted uniformly and averaged by month and
Page 46 of 159
year. Comparison data were averaged from the nearest grid box, weighted uniformly to account for bias resulting from the nonuniform frequency and spacing of the measurements. The full comparison is shown in Figure 4 with root mean square coefficient of variation (CV) and mean bias (∆μ) given for each bin.
Figure 6. Mean column ozone in the base model, UCX and TOMS measurements. Solid lines are 2007 means, with the range of measurements from 2007 shown as dotted lines for the model and as a shaded region for TOMS.
The mean bias is within ±10% for 13 of the 16 zones and ±17% globally. The stratospheric band not used by Emmons et al shows the greatest variation between the UCX and base model. The model spinup process is clearly shown as the UCX exhibits low ozone mixing ratios throughout the first 3 years, during which the stratospheric air is going through its first turnover. The base model, relaxing to climatological
Page 47 of 159
means rather than calculating ozone mixing ratios online, does not require such a spinup. However, by the final simulation year, the UCX simulation shows mixing ratios that are closer to the TOMS observations at all latitudes except the southern midlatitudes.
Figure 7. Comparison of tropospheric ozonesonde data to GEOS-Chem UCX and the base model.
Moving to the tropospheric results, at high Northern latitudes the model underestimates ozone recovery during polar springtime in the upper troposphere, resulting in errors between -15% and -25% and a mean
Page 48 of 159
bias of -5.37%. Arctic ozone depletion shows finer spatial and temporal structure than in the Antarctic, with greater local variability associated with atmospheric waves (23). Since GEOS-Chem averages over
‘polar caps’ beyond ±84° to compensate for artificial polar singularities resulting from a global longitudelatitude grid, such fine structure is lost, resulting in the reduced variability. This will be addressed in an upcoming grid-independent version of GEOS-Chem. Stratospheric ozone depletion is also highly sensitive to aerosol loadings. The thermodynamic equilibrium model for aerosol and PSC formation underestimates ozone variability in regions with rapid temperature fluctuations.
The negative bias observed elsewhere is positive at high southern latitudes, where calculated ozone has a
due to the fact that well-mixed Eulerian frameworks cannot effectively isolate the air within the polar vortex.
Under-prediction of upper tropospheric Northern ozone remains consistent down to 750 hPa, with an increase in both mean bias, to -8.2%, and variance. This trend is not visible at the surface as high
Northern hemispheric surface emissions dominate the impact of the coarsely resolved polar stratosphere.
Above 750 hPa in the Northern tropics, the results show low mean biases of the order of +1-3%, and CVs below 1%. The model captures some interannual variation such as the peak in August 2004, but does not register other events such as a large spike in Northern tropical ozone at the end of 2005. This discrepancy propagates to all altitudes, indicative of a specific event. This could be caused by a tropopause folding event bringing stratospheric ozone into the troposphere.
3.2.4
Polar ozone depletion
The mechanisms underlying stratospheric ozone depletion by chlorine have been extensively investigated and documented since the discovery of the Antarctic ozone hole (23). These mechanisms involve rapid
Page 49 of 159
catalytic cycling, with strong dependence on stratospheric aerosol surface area densities. These in turn are highly sensitive to meteorological conditions. Capturing such depletion requires mechanisms not typically present in tropospheric models, including polar stratospheric cloud (PSC) microphysics and heterogeneous halogen chemistry.
As a test of these features within the UCX, Figure 8 shows the modeled ozone depletion behavior over
72°S, 10°W for 2006. This year was chosen due to the extent of the Antarctic ozone hole. The grid box for this location contains the Neumayer observation station (70.65°S, 8.26°W), which supplied ozonesonde data for comparison. A sampling pressure altitude of 73 hPa was chosen on the basis of a recent study investigating rapid ozone depletion which informed the chemical mechanism choices in the
UCX (148). Meteorological conditions including temperature and the extent of local and polar nighttime are also shown. The light shading corresponds to polar nighttime at 89°S, while darker shading corresponds to local nighttime at Neumayer.
Figure 8 compares modeled ozone depletion to ozonesonde data, showing odd oxygen (O
x
) from the base model and ozone from UCX. The base model’s chemical mechanism is limited to the troposphere, instead using the Linoz method to solve for ozone and storing the result in the O x
tracer. Taking 2 ppmv as an approximate springtime mean, and recognizing that the ozonesondes show 100% depletion, the extended model achieves an approximate maximum depletion of 90% compared to 40% in the base model.
Although the modeled depletion is 50% greater than the base model, UCX overestimates the rate of ozone recovery beginning around mid-November. By comparison, observations at Neumayer show ongoing depletion until mid-December. This is mirrored by underestimated ozone destruction rates. Observed ozone is near-zero by late September, while the modeled minimum occurs in mid-October. There are two likely causes. Firstly, well-mixed grid boxes in an Eulerian framework compromise the isolation of the polar vortex, which is made worse by the combination of GEOS-Chem’s large (84°S and above) averaged
Page 50 of 159
polar cap and the low resolution used in these simulations. This degraded isolation results in undepleted extravortical air replenishing the modeled ozone concentrations.
Figure 8. Clockwise, from top left: ozone burden, inorganic chlorine partitioning, PSC surface area density and temperature for the year 2006 at 73 hPa over Neumayer. The shaded regions correspond to local and polar night, and the dashed line is the typical chlorine activation threshold temperature (148).
Secondly, the rate of chlorine-based ozone depletion is a function of available ClO x
, which depends on the relative chlorine activation and deactivation rates. Activation occurs through heterogeneous chemistry
and solid aerosol surface areas for this period. The thermodynamic aerosol parameterization, coupled with the coarse horizontal resolution that precludes localized PSC formation, will underestimate the overall
Page 51 of 159
surface area density (SAD). In reality, aerosols can form in small low-temperature regions and then persist even once local temperatures have exceeded the frost point, but this is not captured in a pure thermodynamic equilibrium model at coarse grid resolutions. This manifests as a cessation in ozone depletion when aerosol loadings return to pre-vortex levels, which in turn occurs immediately after the temperature rises above the approximate PSC formation or “chlorine activation” temperature. PSC measurements (149) suggest typical SADs of 2-10 µm 2 /cm 3 , whereas modeled densities are 0.5-2
µm 2 /cm 3 .
This analysis is informed by the total inorganic chlorine (Cl y
) partitioning between molecular chlorine
(Cl
2
), active chlorine (Cl + ClO = ClO x
), the ClO dimer Cl
2
O
2
, reservoir chlorine in the forms of HCl and
ClONO
2
, and other modeled inorganic chlorine species (ClOO + OClO + ClNO
2
+ HOCl = Other Cl).
Ozone recovery begins immediately after complete deactivation into reservoir chlorine; based on analysis by Grooß et al (148), the modeled unrealistic recovery of ozone is likely due to insufficient reactivation of chlorine which would otherwise balance ongoing chlorine deactivation.
For comparison, Figure 9 shows column ozone over the Arctic for 2007. Data from the UCX, the base
model (‘Trop’) and reanalyzed TOMS data are shown. The UCX consistently estimates a smaller total ozone column than the base model, although both models capture the overall annual trend with a spring maximum and winter minimum and the reanalyzed observations agree with each of the two models at different times in the year. Modeling Arctic ozone poses additional challenges compared to Antarctic ozone, due to more variable temperatures, the influence of Northern Hemispheric landmasses on zonal flow and the resultant reduced stability of the polar vortex (23, 150).
The large grid resolution used in the current implementation of the UCX means that local extremes in temperatures are averaged out, preventing formation of either liquid or crystal polar stratospheric clouds. In the more isolated Antarctic vortex, cooling is large-scale and relatively uniform, such that PSC formation is likely even when using
Page 52 of 159
simple thermodynamic equilibrium models and large, averaged grid cells. In the warmer Arctic stratosphere, solid PSCs may not form at all (151), and ozone concentrations are likely to be much more sensitive to small variations in temperature. Investigation of the relative sensitivity of modeled Arctic ozone to temperature uncertainty and grid resolution would be a useful future research objective.
Figure 9. Arctic ozone at 74°N for 2007 as modeled by the UCX, the base model (‘Trop’) and from TOMS reanalysis
3.2.5
Source gas stratospheric lifetimes
Stratospheric lifetimes of some tracers are calculated for comparison with recent estimates by Brown et al
(152) based on the ACE-FTS experiment, summarized in Table 5. As recommended in the Brown et al
paper, their results are linearly scaled to reflect a lifetime for CFC-11 of 60 years rather than WMO estimates of 45 years, corresponding to estimates by Douglass et al. (153). The “previous estimates” are referenced from the 2006 and 2010 Scientific Assessments of Ozone Depletion (137, 154) and a previous study of stratospheric lifetimes (155). The UCX lifetimes are calculated from global burdens and stratospheric removal rates, while the Brown et al lifetimes are derived from tracer-tracer correlations with CFC-11.
Page 53 of 159
Table 5. Stratospheric lifetime of selected species. Range shown for ACE-FTS corresponds to uncertainty, range for UCX corresponds to absolute range in daily-mean values for 2007.
CFC-11
CFC-12
CH
N
2
3
CCl
Cl
4
O
CH
4
Brown et al (152)
60.0
1
[56.0-64.0]
150
[127-184]
92.4
[62.4-179]
46.2
[37.2-61.2]
164.4
[127-235]
260
[203.4-359.4]
GEOS-Chem UCX
58.9
[48.9-66.5]
123
[108-136]
43.8
[37.9-46.1]
50.6
[42.1-57.1]
148
[132-162]
164
[147-175]
Previous Estimates
45
-
2
[26-50]
114 3
93 4
[75-111]
All of the estimated lifetimes lie between the previous estimated lower bound and Brown et al’s upper bound, apart from CH
3
Cl. Both CH
3
Cl and CH
4
exhibit stratospheric lifetimes of the order of 50% lower than that estimated by Brown et al. Of the species shown, these are the only two with significant OH reaction branches. As such, their relatively low lifetimes are likely due to the aforementioned high OH.
3.2.6
Correlation studies
Popp et al. (156) showed a robust correlation between stratospheric ozone and nitric acid (HNO
3
), based on in-situ and satellite measurements. Their analysis (specifically Popp et al. Figure 5) is replicated using
the UCX in Figure 10. The smooth lines correspond to polynomial fits based on ACE-FTS interferometry
and MLS satellite data. Daily mean data are plotted corresponding to measurements on the same day from the mean location of all measurements for each instrument.
1 CFC-11 lifetime from Douglass et al (153) used to scale other lifetimes
2 Montzka et al (154)
3 Daniel et al (137)
4 Volk et al (155), based on estimated lifetime for CFC-11 of 45 years
Page 54 of 159
Figure 10. HNO
3
-O
3
correlations in GEOS-Chem UCX (point data) compared to polynomial fits based on measurements from
ACE-FTS (top) and MLS (bottom) (156). Different markers correspond to different measurement dates.
The form of the correlation between the trend in both cases follows the polynomial fit, with HNO
3 peaking at the same ozone mixing ratio as was observed for each of the three observation sites. The MLS data also observed narrow peaks shown in the model, which are not reflected in the polynomial fits. The largest discrepancy is in the MLS equatorial data, although this fit is to a small number of measurements
Page 55 of 159
and does not appear to follow the observed trends elsewhere. The explanation by Popp et al is that air parcels traveling with the Brewer-Dobson circulation evolve from an initial high ozone condition, moving through regions of high HNO
3
production to achieve larger HNO
3
-O
3
ratios with increasing latitude. The
UCX results support this hypothesis, exhibiting evolution of an ozone-rich sample towards a high HNO
3 state with increasing latitude even where the sparsely-populated MLS data is inconclusive. The UCX results depend on the initial conditions selected by the user. In the case of the simulation used to generate these results, the initial stratospheric NO y
distribution increased absolute HNO
3
mixing ratios, resulting in the observed upward bias. Another study by Sankey and Shepherd (157) compared correlations between stratospheric species in the Canadian Middle Atmosphere Model (CMAM) to data from the ATMOS
experiment (158). Sankey and Shepherd Figure 14 is replicated in Figure 11, below.
Page 56 of 159
Figure 11. NO y
/O
3
relations in January (top panel) and July (bottom panel) 2007 from the UCX.
The UCX results, unlike those from Sankey and Shepherd, are disperse below 50 hPa. This is especially pronounced in the tropics, suggesting strong tropospheric influence at these altitudes not present in
CMAM. The subtropical mixing barrier observed by Michelsen is therefore not well reproduced below 85 hPa. However, at 50 hPa and above, the aforementioned barrier is reflected in a sharp increase of 50-
100% in the NO y
/O
3
ratio between the equator and ±30°, corresponding to the “tropical pipe” transport barrier which separates tropical from extratropical air. Future research may benefit from a dedicated investigation into overrepresentation of tropospheric influence in the GEOS-5 lower stratosphere, since the primary references for this work focus on data above 70 hPa (158–160). At high southern latitudes in
July, the model demonstrates another transport barrier due to the polar vortex.
Page 57 of 159
Figure 12. NO y
versus N
2
O throughout the year in the Southern and Northern Hemispheres (top and bottom respectively).
NO y
is also compared to N
2
O, as in Figure 11 of Sankey and Shepherd (157), in Figure 12. Excluding
very high altitude sources, N
2
O is the only major stratospheric source of NO y
, producing NO through reaction with O( 1
D). These species are plotted in Figure 12. The UCX successfully reproduces the trend
in NO y
relative to N
2
O, showing the same consistent maximum of 15-20 ppbv NO y
at 50-70 ppbv N
2
O.
The same vertical variation is also achieved. As altitude increases, N
2
O is converted into NO, resulting in the aforementioned peak. The relation’s compactness results from both tracers being accurately represented as long-lived compared to the horizontal mixing timescales. The spread in the low altitude
Southern Hemispheric data likely results from interference by the polar vortex, with denitrification
Page 58 of 159
reducing NO y
mixing ratios at the poles without affecting N
2
O. This spread is not visible in CMAM, which does not model denitrification.
3.3
Summary
An online unified chemistry extension has been successfully applied to the GEOS-Chem chemistrytransport model, expanding modeling capabilities and improving the agreement between model and observations without compromising tropospheric chemistry. In particular, the UCX shows improved correlation between modeled column ozone and satellite data, capturing medium- and long-term ozone responses to the Antarctic polar vortex. This was achieved without reducing model stability and only a minimal increase in data requirements in the form of 2-D initial conditions and boundary conditions for stratospheric source gases. Using a thermodynamic aerosol parameterization, ozone depletion trends in southern hemispheric winter are reproduced. By expanding the wavelength range considered for photolysis calculations, stratospheric photochemistry modeling is enabled; by combining online radiative transfer calculation in the stratosphere with the aforementioned aerosol parameterization, it has also been made possible to model the potential effects of stratospheric aerosol formation on tropospheric photochemistry through changes in optical depth.
The UCX does not significantly affect the calculated global tropospheric oxidative capacity, resulting in an estimated lifetime of CH
4
against destruction by OH of 7.20 years compared to climatological estimates of around 8.3 years. In particular, zonal seasonal OH means are calculated to be within the range of 80-150% of estimated climatology, although this lies within the estimated zonal seasonal variability of air mass-weighted OH concentrations. This result is also likely to be sensitive to the choice of emissions inventory and meteorological data, which is not investigated here.
Page 59 of 159
4 Human health impacts of sulfate aerosol climate engineeering
Following development and validation of the GEOS-Chem UCX, the goal of assessing human health impacts of high altitude emissions can be addressed. In this chapter the potential impacts of sulfate aerosol injection (SAI) are estimated, broken down by photochemical and climate feedback pathways.
Although previous studies have evaluated the sensitivity of climatological (100–102) and chemical (24,
25, 103) metrics to SAI, temperature and precipitation impacts calculated using global circulation model are applied to existing meteorological fields to estimate the photochemical response to climate change reversal using stratospheric sulfates.
4.1
Experimental design
Simulations are run using the GEOS-Chem UCX to estimate the sensitivity of surface impacts, in the form of population exposure, to the following geoengineering metrics: rate of aerosol injection, total stratospheric aerosol burden, or the achieved surface cooling. Literature concentration and exposure response functions (CRF/ERFs) are then applied to estimate human health impacts, as described in
and we account only for the impacts of a ‘steady-state’ geoengineering policy. Transient effects and changes in short-term mortality are not accounted for due to their dependence on accurately modeling intermediate climatic states. An injection rate of 1 TgS/yr was chosen based on work by Rasch et al (101), who found that injection of an idealized aerosol at this rate could result in 0.9 to 1.7 K of surface cooling.
Geoengineering emissions are assumed to be in the form of sulfate aerosol on the basis of microphysical simulations by Pierce et al (161) which showed that aerosol lifetimes are greatest when emitting directly as H
2
SO
4
rather than emitting gaseous SO
2
and allowing aerosol to form in situ by oxidation.
Calculated exposure sensitivities are shown and discussed in section 4.2. These are used to estimate
Page 60 of 159
changes are grouped under “climate change reversal”, whereas the remaining changes are grouped under
“side effects” as they are unique to SAI. Second-order effects are discussed in section 4.2.4, while
sensitivity to the choice of CRF is explored in section 4.3.1.
4.1.1
Sensitivity decomposition procedure
For the Geoengineering Model Intercomparison Project (GeoMIP), the Environment Canada modeling group used CanESM2 to simulate the RCP 4.5 emissions scenario over the period 2000-2100 with and without an enhanced stratospheric aerosol layer (13, 19). This was achieved by assuming a horizontally uniform increase of 0.0472 in AOD, distributed between 0 and 10 km immediately above the climatological tropopause, finding an overall annual average surface temperature change of -0.6 K. This aerosol loading represents a 5 TgS/yr stratospheric injection scenario, using large, volcanic aerosols. The monthly mean temperature and precipitation differences are extracted from their model results.
Temperature differences are maintained as three-dimensional fields, whereas only the global percentage change in precipitation for each month is retained. An alternative approach using local precipitation
adjustment is discussed in section 4.2.2.
The precipitation and temperature responses are divided by 5 to give small deltas approximately appropriate to a 1 TgS/yr injection scenario and then applied directly to the meteorological input fields for
GEOS-Chem. The resulting change in mixing ratios and UV irradiance is then attributed to the modeled change in temperature and precipitation. This yielded the sensitivity of each quantity with respect to temperature and precipitation. In the case of temperature, dividing the resulting change (for some quantity
X, e.g. ozone in a grid cell) by the change in global annual average surface temperature gives
𝜕𝑋
. The
𝜕𝑇 same applies for precipitation, giving
𝜕𝑋
𝜕𝑃
. For PM
2.5
, the corresponding units are
gm -3 K -1 and µgm -3 ‰ -1 , respectively.
Page 61 of 159
Sensitivity of surface air quality and UV to aerosol loading and subsidence of injected mass are calculated in a similar fashion. GEOS-Chem is run without any temperature or precipitation perturbation but injecting 1 Tg per year of some inert, sulfate-like aerosol with the same molar mass. The aerosol has no chemical sources or sinks and does not interact chemically with other species in any way but does descend to the surface via sedimentation and advection. It is also washed out in the troposphere in the same fashion as online sulfate aerosols. The change in each quantity as a result of descent of aerosol mass is then the “Subsidence” impact, and is normalized by the injection rate (because it relies on the fact that what goes up must come down and is expected to scale with injection rate). This is the quantity Δ𝑆
𝜕𝑋
.
𝜕𝑆
Only PM
2.5
is affected as chemical interactions are not simulated. All microphysical processes relevant to sulfate in the full chemical model, such as gravitational settling and aerosol evaporation and condensation, are included for this inert aerosol calculation.
This simulation is then repeated but this time allowing photochemical interactions. Ignoring the meteorological aspects, the assumption is that the results of this can be decomposed into: a) The descending aerosol mass. Since there are no sulfate loss mechanisms in the stratosphere except for descent into the troposphere, impacts due to descending aerosol mass are normalized by the injection rate [
𝜕𝑋
𝜕𝑆
in
gm -3 (TgS/yr) -1 for PM
2.5
] b) Everything else, which is the result of an aerosol layer in the stratosphere enabling faster heterogeneous chemistry and influencing surface-bound radiation. This is therefore normalized by the stratospheric layer burden (
𝜕𝑋
𝜕𝐵
in
gm -3 TgS -1 for PM
2.5
)
Subtracting
𝜕𝑋
𝜕𝑆
from the result of the second simulation therefore gives Δ𝐵
𝜕𝑋
𝜕𝐵
. Since ΔS is prescribed as 1
TgS/yr and ΔB is calculated as 2.4 TgS from the forward GEOS-Chem UCX run, the following have therefore been calculated:
Page 62 of 159
2.
3.
1.
4.
𝜕𝑋
– Sensitivity of X to temperature (e.g. µgm
𝜕𝑇
-3 K -1 )
𝜕𝑋
– Sensitivity of X to precipitation (e.g. µgm
𝜕𝑇
-3 ‰ -1 )
𝜕𝑋
– Sensitivity of X to rate of injection S due to subsidence of aerosol (e.g.
gm -3 (TgS/yr) -1 )
𝜕𝑆
𝜕𝑋
𝜕𝐵
– Sensitivity of X to the presence of a static aerosol layer (e.g.
gm -3 TgS -1 )
The overall impact is estimated as
Δ𝑋 = Δ𝑆 (
𝜕𝑋
𝜕𝑆
+
𝜕𝐵
𝜕𝑆
𝜕𝑋
𝜕𝐵
) + Δ𝜏
𝜕𝑇
𝜕𝜏
(
𝜕𝑋
𝜕𝑇
+
𝜕𝑃
𝜕𝑇
𝜕𝑋
)
𝜕𝑃 with ΔS = 1 TgS/yr, and
𝜕𝐵
𝜕𝑆
= 2.4 yr (i.e. the aerosol lifetime calculated from simulation). The above equation is a simplification, however. The only fully independent variable is ΔS, the rate of injection.
Although the meteorological sensitivities are multiplied by the change in optical depth Δ τ , this parameter is itself also a function of the injection rate via the calculated aerosol lifetime
𝜕𝐵
𝜕𝑆
and the aerosol optical
𝜕𝜏 properties which control
𝜕𝐵
. The calculation therefore expands to
Δ𝑋 = Δ𝑆 ((
𝜕𝑋
𝜕𝑆
+
𝜕𝐵
𝜕𝑆
𝜕𝑋
𝜕𝐵
) +
𝜕𝐵
𝜕𝑆
𝜕𝜏 𝜕𝑇
𝜕𝐵 𝜕𝜏
(
𝜕𝑋
𝜕𝑇
+
𝜕𝑃
𝜕𝑇
𝜕𝑋
𝜕𝑃
))
The sensitivities
𝜕𝑇
𝜕𝜏
= −12.4
K per unit OD and
𝜕𝑃
𝜕𝑇
= 2.43
% K -1 are estimated directly from the
CanESM2 output data. This approach neglects second order terms, such as
𝜕 2 𝑋
𝜕𝑇 2
𝜕 2 𝑋
and
𝜕𝑇𝜕𝑃
. The
consequences of this assumption are assessed in section 4.2.4
4.1.2
Model setup
All model runs are performed at a 4°×5° resolution with 72 model layers from the surface to 0.01 hPa.
Meteorological fields for 2004-2010 inclusive are cycled as a surrogate for 2040, and the final 5 years of
Page 63 of 159
simulation data are averaged to yield a representative annual mean. Geoengineering temperature and precipitation fields are calculated using the output data from the GeoMIP G4 and CMIP5 RCP45 experiments as run in CanESM2 (102). Precipitation is perturbed according to the global net monthly mean percentage change due to geoengineering. Temperature is perturbed according to the interpolated three-dimensional monthly mean absolute change. Output data are scaled by aerosol optical depth;
CanESM2 results correspond to an AOD of 0.0472. Transport and convection are not adjusted to reflect the effects of geoengineering or of climate change.
Stratospheric sulfate aerosols are assumed to follow a log-normal distribution, implemented specifically for this experiment over the original background parameters. Aerosol parameters for each simulation are estimated from a 2D model with full aerosol microphysics (32, 76). Both the baseline and 1 TgS/yr scenarios were simulated, and parameters for a log-normal aerosol distribution were estimated based on the best fit to the 2D model results. Simulations with no aerosol injection use a modal radius of 0.06 µm and standard deviation σ = 1.8. For an injection rate of 1 TgS/yr, a modal radius of 0.16 µm and standard deviation σ = 1.6 is used. Mie scattering parameters are calculated using a NASA FORTRAN code
(available at http://www.giss.nasa.gov/staff/mmishchenko/ftpcode/spher.f
[accessed 2014-11-02]) (162).
Gravitational settling for a log-normally-distributed aerosol is calculated according to Chen et al (163).
Heterogeneous chemistry and interactions with radiation are described in Eastham et al (164).
RCP 4.5 anthropogenic emissions for 2040 are used to produce a surrogate future atmosphere (165–167).
Uniform surface boundary conditions, also based on RCP 4.5 projections, are prescribed for long-lived species such as CFCs and N
2
O. Methane and CH
3
Br surface mixing ratios are fixed for 4 separate zonal bands to allow latitudinal variation and scaled to RCP 4.5 projections. When estimating the relative effect of the vertical distribution of emissions on aerosol transport, all sulfur emissions using standard modernday GEOS-Chem emissions inventories are summed for each month to find a surface emissions
Page 64 of 159
distribution. 1 Tg of inert aerosol per year is then emitted with this temporal and spatial distribution to find the primary aerosol exposure resulting from 1 Tg/yr of aerosol emissions at the surface.
4.2
Exposure sensitivities
Sensitivity to each of the impact drivers is presented here. Meteorological drivers are presented first, including a sensitivity analysis of the impact of regional changes in precipitation compared to global net reductions. Finally, sensitivity of the exposure metrics to the subsidence of aerosol mass and the presence of a static aerosol layer is shown.
4.2.1
Sensitivity to temperature
The global Earth-system model CanESM2, which incorporates fully coupled three-dimensional atmosphere and ocean general circulation models (19), was run for the GeoMIP project (13) to quantify the climatological impact of sulfate geoengineering. CanESM2 estimated that an increase in the global stratospheric sulfate aerosol optical depth of 0.0472 would change global surface temperatures by an average -0.6°C and decrease net global precipitation by 1.4%. The calculated distribution of temperature is used to compute the global sensitivity of population-weighted PM
2.5
(µ gm -3 K -1 ), ozone (ppbv K -1 ) and erythemal UV (%K -1 ) with respect to temperature.
Warmer air promotes evaporation of ammonium nitrate, resulting in decreased nitrate aerosol concentrations. However, the rate of sulfate formation increases with temperature, due to faster oxidation by OH. This results in an overall negative sensitivity of PM
2.5
to temperature when nitrate is the dominant
terms of both the total population exposure and the population-weighted mean increase in loading from each impact pathway for PM
2.5
.
Page 65 of 159
Table 6. Sensitivity of population PM
2.5
exposure to sulfate aerosol injection
Global
N. America
S. America
Europe
Asia
Africa
W. Pacific
Global
N. America
S. America
Europe
Asia
Africa
W. Pacific
Temperature
(K -1 )
-1,500
-27
11
-77
-1,300
-41
-0.0087
-0.29
-0.065
0.039
-0.16
-0.44
-0.054
-0.0003
Precipitation
(‰ -1 )
-13
-0.44
-0.18
-1.1
-9.7
-1.5
-0.018
-0.0026
-0.0011
-0.00060
-0.0022
-0.0032
-0.0020
-0.00061
Aerosol subsidence
(TgS/yr) -1
91
5.3
3.3
3.6
63
15
0.30
0.018
0.013
0.011
0.0074
0.021
0.020
0.010
Strat. aerosol burden
(TgS -1 )
-6.4
0.34
0.76
-4.4
-9.3
6.2
0.060
-0.0013
0.00082
0.0026
-0.00891
-0.0031
0.0082
0.0021
Figure 13 shows the horizontal distribution of PM
2.5
sensitivities to temperature. Evaporation of nitrate results in negative sensitivities throughout Europe, Asia and the US East Coast, and nitrate has a global population-weighted mean sensitivity of -0.28 µgm -3 K -1 . Precursor emissions (NO x
) in India are projected to increase by over 100% between 2000 and 2040, resulting in a peak sensitivity of –1.3 µgm -3 K -1 in northern India. Changes in dust are not shown or included in the temperature impact estimate, as the net change in dust exposure is -0.02 µgm -3 K -1 . Organic carbon loading is positively correlated with temperature due to increased biogenic emissions, increasing at a rate of up to 0.17 µgm -3 K -1 in equatorial
Page 66 of 159
Africa and 0.31 µgm -3 K -1 in South America. These changes are focused in remote regions and carbonaceous aerosol exposure increases by 0.021 µgm -3 K -1 globally.
Figure 13. Sensitivity of PM
2.5
loading to temperature, broken down by component. Clockwise, from top left: total non-mineral
PM
2.5
, nitrate (NO
3
), sulfate (SO
4
2) and carbonaceous aerosol. All plots are in µ gm -3 K -1 .
The total global mean sensitivity of PM
2.5
to temperature is shown in Figure 14, weighted by area, land-
area and population. An annual-average land-weighted global response of -0.084 µgm -3 K -1 is calculated for nitrate aerosol. Sulfate concentrations over land change at 4.2×10 -3 µgm -3 K -1 due to the temperature dependency of SO
2
oxidation, at the lower end of the 1.6-34×10 -3 µgm -3 K -1 range reported by Dawson et al for the US (23). The global land-area-weighted average change is -0.10 µgm -3 K -1 inorganic PM
2.5
,
Page 67 of 159
increasing in magnitude to -0.31 µgm -3 K -1 when population weighting is applied. The net global response is dominated by evaporation of nitrate aerosols in densely populated areas, resulting in a negative sensitivity of combined PM
2.5
of -0.29 µgm -3 K -1 . Reversal of climate warming by sulfate aerosol engineering therefore increases PM
2.5
exposure as nitrate aerosol formation is enhanced, partially mitigated by a small decrease in sulfate loading due to decreased SO
2
oxidation.
Figure 14. Sensitivity of PM
2.5
loading to temperature, weighted by area, land-area and 30+ population. Reference estimates are the US land-weighted values reported by Dawson et al for January and July in 2001-2002 (23).
Table 7 shows the surface ozone impacts of SAI in the same format as Table 6. Note that it is not
meaningful to compare the magnitudes of any two columns directly as they are not related to a common injection rate. Temperature change has mixed impacts on ozone (10). The lifetime of PAN, a NO x reservoir, decreases at a rate of 10% K -1 , meaning that NO x
concentrations are increased in polluted
Page 68 of 159
regions but transport to remote regions is reduced. The population-weighted global average sensitivity of ozone to temperature is estimated to be 2.1 ppbv K -1 , based on the annual 6-month maximum. The sensitivity is positive at all locations, peaking at up to 4 ppbv K -1 in areas with high background NO x
. The approach presented here estimates the partial sensitivity with respect to temperature without including humidity feedbacks. The current model configuration also uses a fixed specific humidity condition, such that an increase in temperature will result in a corresponding fall in relative humidity. This is likely the cause of increased ozone in remote and marine areas as O( 1 D) removal is slowed by reduced H
2
O availability. However, ozone is only weakly dependent on humidity in polluted regions (10), so this discrepancy should not significantly affect population exposure.
Table 7. Sensitivity of population ozone exposure to sulfate aerosol injection
Global
N. America
S. America
Europe
Asia
Africa
W. Pacific
Global
N. America
S. America
Europe
Asia
Africa
W. Pacific
Temperature
(K -1 )
10,000
840
510
830
6,700
1,400
29
2.1
2.1
1.7
1.7
2.2
1.8
1.0
Precipitation
(‰ -1 )
5.9
0.39
0.093
0.59
4.3
0.53
0.030
0.0012
0.00095
0.00031
0.0012
0.0014
0.00071
0.0011
Aerosl
Subsidence
(TgS/yr) -1
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Strat. aerosol burden
(TgS -1 )
-2,000
-190
-130
-200
-1,100
-290
-19
-0.39
-0.46
-0.43
-0.42
-0.37
-0.38
-0.66
Net incident erythemal UV exposure at the surface is decreased by increasing temperature due to the associated increases in the tropospheric ozone column. The overall sensitivity of SCUP-h weighted UV exposure to temperature is -0.85% K -1 .
Page 69 of 159
4.2.2
Sensitivity to precipitation
A uniform percentage reduction in net global precipitation results in increased particulate matter loading.
The global population-weighted sensitivity is -2.6×10 -3 µgm -3 ‰ -1 . This calculation includes fine mineral dust, which is transported further when precipitation is reduced and contributes -0.8×10 -3 µgm - 3 ‰ -1 of the total impact. This result is less sensitive to the choice of emissions pathway than the temperature impact as there is no change in the sign of the sensitivity between species. Ozone exposure is found to be relatively insensitive to precipitation, responding at a rate of 1.2 pptv ‰ -1 , c.f. the result of Dawson et al
(24) who found a sensitivity of 2 pptv ‰ -1 for the US. The net global erythemal UV exposure response is
-0.0018 % ‰ -1 .
As a sensitivity test, a simulation was run in which the change in precipitation was adjusted as a function of location instead of using a single global factor. An investigation of regional climate change in 2000 split global land area into 21 regions, representing “different climatic regimes and physiographic settings”
impacts of geoengineering used these regions, modified to split Australia into North and South Australia, to estimate regional impacts and rates of change with respect to different quantities of solar radiation management (169). The relative change in precipitation in each region per unit decrease in solar
insolation is shown in Figure 15 (bottom), where -1 corresponds to the relative change in global net
precipitation.
Page 70 of 159
Figure 15. Top: Giorgi regions (168) mapped to the 4°×5° grid. Bottom: Mean relative change in precipitation per unit geoengineering for each region.
Reduction in net global precipitation has negligible effects on ozone and UV-B exposure (1% of total impacts), but is responsible for 16% of the change in global PM
2.5
exposure. However, this does not
capture the potential impact of regional changes in precipitation. Figure 16 shows impacts on PM
2.5
as
change in precipitation.
Page 71 of 159
Figure 16. Impacts of precipitation change due to geoengineering on global PM
2.5
.Top: Impacts due to reduction in global net precipitation only. Bottom: Impacts modelled using region-specific precipitation impact factors.
Application of regional scaling factors increases the sensitivity of global PM
2.5
exposure to precipitation effects by 15%. This is the result of increases in exposure in the most negatively affected regions and decreases in areas with already low exposure. North American impacts fall by 35%, and South American impacts fall by 60%. Although most of North-Eastern Europe shows increased exposure due to regional scaling factors, the Mediterranean experiences a net decrease in PM
2.5
exposure due to increased precipitation, and overall European PM
2.5
exposure attributable to changes in precipitation fall by 64%.
Meanwhile Asian impacts are increased by 40%, as precipitation in South Asia, North Asia and Tibet is decreased yet further. These results are not used to calculate health impacts, but should be investigated as part of future work.
Page 72 of 159
4.2.3
Direct impacts of SAI
Using a single log-normal aerosol size distribution with 𝑟̃ = 0.16
µm and 𝜎̃ = 1.6
, an injection rate of S =
1 TgS/yr between 20 and 25 km and ±30° latitude results in the average stratospheric sulfur burden increasing from 0.1 to 2.5 TgS ( 𝛥𝐵 = 2.4
TgS). This corresponds to a 2.4 year stratospheric aerosol lifetime due to low sedimentation velocities associated with a small, fixed modal radius. Average column peak surface area density increases to 15 µm 2 /cm 3 , with an associated increase in global mean optical depth 𝛥𝜏 of 0.079.
The effect of SAI on stratospheric ozone, a subject of ongoing research, is discussed first (24, 25, 99,
103). Increasing aerosol surface area density (SAD) results in sequestration of NO x
as HNO
3
through
N
2
O
5
hydrolysis, slowing destruction of O
3
through NO x
cycling. This is reflected in a 10% reduction in the ratio of NO x
to NO y
compounds at 26 km. Meanwhile the additional surface area increases activation of chlorine and bromine, with ClO x
/Cl y
ratios 20% greater at 21 km altitude as a result of aerosol injection. The newly-available chlorine not only depletes ozone directly, it also oxidizes methane and thereby increases the water available for HO x
formation. Figure 17 shows the annual average rate of
methane oxidation by O( 1 D), OH and Cl with and without SAI, with the sulfur injection region highlighted in grey. SAI increases both HO x
-catalyzed ozone destruction and oxidation of CH
3
Cl and
CH
3
Br, further increasing halogen availability. These results are therefore sensitive to the chosen projections of halocarbon emissions.
Page 73 of 159
Figure 17. Methane loss rates under baseline (solid lines) and engineered (dashed lines) conditions.
ozone. The change of loss rate is expressed as a percentage of the overall loss rate calculated for the baseline scenario. ClO x
cycles, including the HO x
-ClO x
coupled cycle and ClOOCl dimer cycle, are the largest contributor to increased stratospheric ozone loss, increasing overall ozone destruction by 28% at around 18 km pressure altitude. HO x
, BrO x
and other coupled halogen cycles each contribute between 12 and 15% additional ozone destruction between 10 and 20 km. However, sequestration of NO x
into HNO
3 due to enhanced N
2
O
5
hydrolysis acts to reduce the overall loss rate, reducing ozone depletion at 17 km by 16%. The net change is an increase in ozone destruction rates in the lower stratosphere, peaking at
+50% at 17 km pressure altitude.
Overall ozone loss peaks at 19 km. Between 20 and 33 km, the suppressed NO x
cycling results in a reduction in overall ozone destruction rates, as halogen cycles become less dominant. This effect is greatest at 25 km, where ozone destruction rates with SAI are 20% lower than without. This results in
Page 74 of 159
increased ozone concentrations above 25 km, although the net effect is a decrease in the global ozone column.
For an increase in stratospheric sulfur burden 𝛥𝐵 = 2.4
TgS, a 5.7% reduction in the global average ozone column is calculated. The greatest percentage loss is observed at the poles (10% and 9.3% for the
Antarctic and Arctic respectively). These numbers are unchanged by inclusion of temperature and precipitation impacts, as they have no effect on the stratospheric surface area density. The sensitivity of stratospheric ozone to injection, expressed with respect to the change in stratospheric burden, is therefore
-2.4 %TgS -1 .
For comparison, Heckendorn et al (24) calculated a 2.3% depletion for an approximately 1.4 TgS burden
(-1.6%TgS -1 ), using a microphysical model and simulating injection with SO
2
. The relatively high sensitivity of stratospheric ozone to burden in the GEOS-Chem simulations is due to the idealized aerosol size distribution, which uses a fixed model radius and variable number density, preventing the formation of large aerosols with lower area-to-volume ratios. Tilmes et al (99) calculated polar ozone depletion using a prescribed surface area density distribution (101), produced using a fixed modal radius bulk aerosol. They estimated that winter Arctic polar ozone loss would lie in the range of 50-130 DU for a 1.5
TgS/yr injection, corresponding to a burden of approximately 4 TgS. For comparison, the 2.4 TgS burden increase modeled in this study results in an average Arctic winter ozone loss of 41 DU. The loss estimates from this work are therefore within the range of published values.
Page 75 of 159
Figure 18. Global mean changes to ozone-destroying cycles and the overall ozone column, based on noontime photolysis rates.
Rate changes are for the limiting step in each cycle and are given as a percentage of the overall baseline rate, as in Heckendorn et al (24).
Loss of stratospheric O
3
increases shortwave UV flux, while enhanced aerosol optical depth decreases flux at all frequencies. Total incident UV power increases by 11% TgS -1 at wavelengths below 300 nm, but decreases by 0.45% TgS -1 above 300 nm. The shortwave O( 1 D) branch of ozone photolysis is on average 2.1% TgS -1 faster worldwide, increasing from 1% TgS -1 at the equator to 8% TgS -1 at the poles.
The surface ozone response is dominated by reduced ozone transport from the stratosphere, as stratospheric ozone depletion reduces the amount of ozone subsiding to the surface. The resultant
ratios fall by between -0.16 and -1.2 ppbv TgS -1 . The global average sensitivity of ozone exposure to
Page 76 of 159
stratospheric aerosol burden is -0.39 ppbv TgS -1 . Peak reductions occur in subsidence regions such as
Central Asia, while regions with strong upwelling such as the equator observe changes an order of magnitude smaller.
Figure 19. Sensitivity of surface ozone mixing ratios to increased stratospheric aerosol burden, expressed as ppbvTgS -1 .
When changes in temperature and precipitation are neglected, the direct transport of aerosol from the stratosphere contributes 50% of the total population-weighted change in sulfate loading following aerosol injection. Subsiding aerosol is subject to wet scavenging, which removes 85% of injected mass above 500 m pressure-altitude. The sensitivity of population-weighted PM
2.5
to SAI via aerosol subsidence is 0.018
µgm -3 (TgS/yr) -1 . For comparison, injection of an inert sulfate-like aerosol at the same rate with a modern distribution of anthropogenic sulfur emissions results in a 0.45 µgm -3 (TgS/yr) -1 increase in populationweighted PM
2.5
. By direct transport alone, a unit of inert aerosol injected into the stratosphere therefore results in human PM
2.5
exposure 25 times smaller than the same unit of additional emission from current sources.
Page 77 of 159
21 shows the sensitivity of surface PM
2.5
, not including directly transported sulfates. OH formation impacts are dictated by a tradeoff between lower ozone mixing ratios and greater O( 1 D) branch ozone photolysis. Shortwave photodissocation of VOCs is also faster, supplying a small additional source of OH in polluted regions. This results in an overall increase in OH over land in the Northern Hemisphere, where anthropogenic pollution is responsible for the majority of surface ozone (26), but a net zero or negative change elsewhere. The northern hemispheric increase in OH results in surface sulfate exposure increasing by 0.0077 µgm -3 TgS -1 as SO
2
oxidation is accelerated. However, overall PM formation is reduced in populated areas as lower local ozone concentrations result in slower production of NO
3
radicals, reducing formation of nitric acid and therefore nitrate aerosol. Nitrate aerosol exposure changes by -0.011 µgm -3
TgS -1 worldwide. This reduction is focused in densely populated regions within Western Europe and
South-East Asia, whereas positive sensitivities are smaller and more diffuse worldwide. Since ammonia reacts preferentially with sulfates over nitrates, ammonium loading increases by 0.0024 µgm -3 TgS -1 .
Overall PM
2.5
exposure is reduced, with a global net sensitivity of -0.0013 µgm -3 TgS -1 . Due to the decreased stratospheric ozone coverage, population-weighted UV-B exposure increases at a rate of 1.27%
TgS -1 .
Page 78 of 159
Figure 20. Percentage change in surface OH loading per TgS increase in the stratospheric aerosol burden using an idealized aerosol.
Figure 21. Sensitivity of surface PM
2.5
to the stratospheric sulfur burden after subtracting the contribution of direct transport.
4.2.4
Response linearity
The method outlined above does not take into account second order effects, specifically linearity and cross-terms. The error associated with assuming linearity is addressed first. This refers to the fact that
Page 79 of 159
higher-order terms may result in the response to a 1.0 K change being different from 10 times the response to a 0.1 K change.
Let 𝑓(𝑇) correspond to the response of some quantity (e.g. population-weighted PM
2.5
concentration) to input T (e.g. temperature). The Taylor expansion of the response to a perturbed input 𝑇 + 𝛿𝑇 is then
𝜕𝑓 𝑓(𝑇 + δ𝑇) = 𝑓(𝑇) + δT
𝜕𝑇
+
δ𝑇
2
𝜕
2 𝑓
2 𝜕𝑇 2
+ 𝑂(𝛿𝑇
3 )
The quantity of interest is the response per unit perturbation. Discarding all but the first two terms yields
𝜕𝑓
𝜕𝑇
= 𝑓(𝑇 + δ𝑇) − 𝑓(𝑇)
+ 𝑂(𝛿𝑇 2 ) 𝛿𝑇
In this case, 𝑓(𝑇 + 𝛿𝑇) and 𝑓(𝑇) are calculated directly by simulation. For some larger perturbation Δ𝑇 , the same expansion gives 𝑓(𝑇 + Δ𝑇) = 𝑓(𝑇) + Δ𝑇
𝜕𝑓
𝜕𝑇
+
Δ𝑇
2
2
𝜕
2 𝑓
𝜕𝑇 2
+ +𝑂(Δ𝑇 3 )
The quantity of interest is now the overall response to the perturbation, i.e. 𝑓(𝑇 + Δ𝑇) − 𝑓(𝑇) = Δ𝑇
𝜕𝑓
𝜕𝑇
+
Δ𝑇 2
2
𝜕 2 𝑓
𝜕𝑇 2
+ 𝑂(Δ𝑇
3
)
An estimate of this quantity, 𝑓̂(𝑇 + Δ𝑇) , is produced using the results from the smaller perturbation. This discards terms beyond the first order, approximating the true result as 𝑓̂(𝑇 + Δ𝑇) = 𝑓(𝑇) + Δ𝑇 ( 𝑓(𝑇 + 𝛿𝑇) − 𝑓(𝑇) 𝛿𝑇
)
Page 80 of 159
The linearity error is therefore 𝜖
𝐿
(𝑇) = 𝑓̂(𝑇 + Δ𝑇) − 𝑓(𝑇 + Δ𝑇)
=
Δ𝑇 𝛿𝑇
(𝛿𝑇
𝜕𝑓
𝜕𝑇
+
δ𝑇 2
2
𝜕 2
𝜕𝑇 𝑓
2
) − (Δ𝑇
𝜕𝑓
𝜕𝑇
+
Δ𝑇 2
2
𝜕 2 𝑓
𝜕𝑇 2
) + 𝑂(Δ𝑇
3 )
= (Δ𝑇
𝜕𝑓
𝜕𝑇
+
δTΔ𝑇
2
𝜕 2
𝜕𝑇 𝑓
2
) − (Δ𝑇
𝜕𝑓
𝜕𝑇
+
Δ𝑇 2
2
𝜕 2
𝜕𝑇 𝑓
2
) + 𝑂(Δ𝑇
3
) 𝜖
𝐿
(𝑇) =
Δ𝑇(δ𝑇 − Δ𝑇)
2
𝜕 2 𝑓
𝜕𝑇 2
+ 𝑂(Δ𝑇
3
)
The same error applies for precipitation P and aerosol injection rate S. A second error results from neglecting cross terms. Say that perturbations to both temperature T and precipitation P are calculated directly as 𝑓(𝑇 + Δ𝑇, 𝑃) − 𝑓(𝑇, 𝑃) = Δ𝑇
𝜕𝑓
𝜕𝑇
+
Δ𝑇 2
2
𝜕 2 𝑓
𝜕𝑇 2
+ 𝑂(Δ𝑇
3 ) and
𝜕𝑓 𝑓(𝑇, 𝑃 + Δ𝑃) − 𝑓(𝑇, 𝑃) = Δ𝑃
𝜕𝑃
+
Δ𝑃 2
2
𝜕 2 𝑓
𝜕𝑃 2
+ 𝑂(Δ𝑃
3 )
The goal is then to calculate the effect of perturbing both variables. We can approximate this as 𝑓̂(𝑇 + Δ𝑇, 𝑃 + Δ𝑃) = (𝑓(𝑇 + Δ𝑇, 𝑃) − 𝑓(𝑇, 𝑃)) + (𝑓(𝑇, 𝑃 + Δ𝑃) − 𝑓(𝑇, 𝑃))
= Δ𝑃
𝜕𝑓
𝜕𝑃
+ Δ𝑇
𝜕𝑓
𝜕𝑇
+
1
2
(Δ𝑃
2
𝜕 2 𝑓
𝜕𝑃 2
+ Δ𝑇
2
𝜕 2 𝑓
𝜕𝑃 2
)
Page 81 of 159
However, the Taylor expansion is 𝑓(𝑇 + Δ𝑇, 𝑃 + Δ𝑃) − 𝑓(𝑇, 𝑃) = 𝛥𝑇
𝜕𝑓
𝜕𝑇
+ Δ𝑃
𝜕𝑓
𝜕𝑃
+
1
2
(𝛥𝑇
2
𝜕
2 𝑓
𝜕𝑇 2
+ 𝛥𝑃 2
𝜕
2 𝑓
𝜕𝑃 2
+ 2𝛥𝑇𝛥𝑃
𝜕
2 𝑓
𝜕𝑇𝜕𝑃
) + 𝑂(Δ
3 )
Therefore the cross term error is 𝜖
𝐶
(𝑇 × 𝑃) = 𝑓̂(𝑇 + Δ𝑇, 𝑃 + Δ𝑃) − 𝑓(𝑇 + Δ𝑇, 𝑃 + Δ𝑃)
𝜕 2 𝑓 𝜖 𝑐
(𝑇 × 𝑃) = −𝛥𝑇𝛥𝑃
𝜕𝑇𝜕𝑃
+ 𝑂(Δ
3 )
To determine the effect of this error on the overall outcome, additional simulations were run which
Page 82 of 159
Table 8. Experimental setup to determine linearity and cross-term errors. Values in orange are considered small perturbations.
Values in green correspond to a full-scale geoengineering scenario with 1.0 TgS/yr injection. Experiments highlighted in bold were used to construct the central estimate.
ID
Base
T
P
S
T
P
S
T
P
T
S
P
S
All
T (K)
0.12
1.0
1.0
1.0
1.0
P (‰)
2.9
24
24
24
24
S (TgS/yr)
0.5
1.0
1.0
1.0
1.0
Term estimated 𝑓(𝑇, 𝑃, 𝐵)
𝜕𝑃 2
𝜕
2 𝑓
2 𝜕𝑆
𝜕 2 𝑓
𝜕𝑇𝜕𝑃
𝜕
2 𝑓
𝜕𝑇𝜕𝑆
𝜕 2 𝑓
𝜕𝑃𝜕𝑆
𝜕𝑓
𝜕𝑇
𝜕𝑓
𝜕𝑃
𝜕𝑓
𝜕
𝜕𝑆
2 𝑓
2 𝜕𝑇
𝜕
2 𝑓
𝑂(Δ
3 )
Associated error term 𝜀
𝐿
(𝑇) 𝜀
𝐿
(𝑃) 𝜀
𝐿
(𝑆) 𝜀
𝐶
(𝑇 × 𝑃) 𝜀
𝐶
(𝑇 × 𝑆) 𝜀
𝐶
(𝑃 × 𝑆)
𝑂(Δ
3 )
Global population-weighted ozone, PM
2.5
and UV exposure are used as metrics. Figure 22 shows linearity
and cross-term errors for a 1.0 K surface cooling induced by 1 TgS/yr injection. Errors are normalized by the overall increase in exposure calculated by scaling temperature and precipitation impacts to match the
1 TgS/yr injection scenario. Since the full injection rate S = 1 TgS/yr is used to calculate the central estimate of mortalities, the second order error term
𝜕 2
𝜕𝑆 𝑓
2
is relevant only when scaling to other scenarios.
Page 83 of 159
Figure 22. Error resulting from first-order approximation of impacts of geoengineering
The largest relative error is the second-order aerosol response term
𝜕
𝜕𝑆
2 𝑓
2
in UV-B exposure. This suggests that directly simulating injection at a rate of 1 TgS/yr results in 18% lower UV-B exposure than if the results of a 0.5 TgS/yr injection scenario are doubled. However, this is not relevant to the central scenario, as the full ΔS response is used when estimating mortality sensitivities. All other error terms affect the total by less than 5%. The overall change in calculated mortalities is shown by the right-most bars, which represent the relative change in calculated mortalities when directly simulating co-perturbation of temperature, precipitation and aerosol burden injection in comparison to linear addition of the scaled response to each stimulus. Direct simulation results in calculated ozone, PM
2.5
and UV-B exposure changing by -2.7%, +1.7% and -3.4% respectively. Second order terms are therefore not considered to significantly affect the final result. However, these assessments are limited by the lack of climatological
Page 84 of 159
feedbacks in GEOS-Chem. Increasing temperatures could affect precipitation in a non-linear fashion, as could additional upper-tropospheric aerosol loading through (for example) cloud formation. Assessment of these feedbacks using a coupled chemistry-climate model is a future objective.
4.2.5
Seasonal trends
shows the monthly mean change in ozone and PM
2.5
attributable to each of the four drivers after scaling to a 1 TgS/yr injection rate. The implied precipitation sensitivity is 2.4% K -1 , and the temperature change is calculated based on the CanESM2 temperature sensitivity to stratospheric sulfate optical depth
(-12 K per unit AOD). This gives a temperature sensitivity of -0.98 K (TgS/yr) -1 . Multiplying the calculated sensitivities by a representative perturbation allows the impacts of a geoengineering scenario to be estimated. Beyond the specific results, the four-factor decomposition allows the identification of the mechanisms which are likely to dominate impacts in specific regions. This will enable further investigation that could reduce the most important uncertainties.
Figure 23. Seasonality of ozone (left) and PM
2.5
(right) exposure resulting from geoengineering. Temperature impacts are shown in blue, precipitation in red, sulfur injection in yellow and sulfur burden in purple.
Page 85 of 159
As already observed in the annual average signal, the PM
2.5
response is dominated by the rise in Asian concentrations due to decreased temperatures. This increase is greatest in Northern Hemispheric winter, during which the change in temperature will have a greater relative effect. This is also correlated with the smallest temperature-related ozone impacts. During Northern Hemispheric summer, PM
2.5
impacts in
Asia are reduced to near-zero, whereas the temperature induced reduction in ozone is maximized.
Ozone impacts are more heterogeneous. These are calculated based on exposure during the ozone season, which is taken as the six consecutive months over which average ozone exposure is maximized (53). In
North America, Europe and Asia, this is usually the period April through September inclusive.
Temperature-related ozone impacts are maximized for these regions within this period, reaching up to -2 ppbv, whereas outside ozone season the change in exposure due to cooling is 50 to 100% smaller. There is an overall change of 2 ppbv due to a 1 K cooling in Europe and North America. However, in the
Southern Hemisphere and Africa, ozone impacts due to temperature changes are constant throughout the year, with an overall amplitude of approximately 0.5 ppbv. In all locations, the response of ozone exposure to cooling is proportional to the background ozone concentration.
Impacts due to stratospheric burden are out of phase with this pattern. North American and European ozone due to aerosol injection is level at around -1 ppbv for most of the year, but rises to -1.5 ppbv in late winter. This pattern is consistent with the theory that the fall in surface ozone is due to reduced ozone in descending stratospheric air, which has been estimated to contribute the greatest fraction of surface ozone during hemispheric winter (26). However, this effect is not included in the health impact as it occurs outside ozone season.
In the Southern Hemisphere, ozone exposure again varies by a smaller degree than in the Northern
Hemisphere with the exception of the West Pacific Region (WPR). Here ozone exposure varies by 1 ppbv throughout the year, again reaching a maximum absolute change of -2 ppbv during hemispheric winter.
Page 86 of 159
This suggests that the ozone impact estimates due to geoengineering, and in particular due to stratospheric ozone depletion, are sensitive to the choice of time period over which exposures are calculated.
4.3
Health impact sensitivities
Applying mortality concentration response functions (CRFs) described in section 2.3, projected
due to the affected diseases. Skin cancer mortalities are estimated based on surface UV flux but are not weighted to account for under-reporting or skin reflectance. The environmental factors are broadly classified into “Reversal” and “Side Effect”. Global cooling and precipitation are counted as reversals of the existing effects of climate change, whereas sulfate injection and aerosol burden impacts are counted as side-effects of SAI specifically. Quantities in bold are central estimates, with the CRF 95% confidence interval given in brackets. 95% confidence intervals for totals are estimated by assuming a triangular distribution for each CRF input variable and applying Latin Hypercube sampling with 100,000 samples.
shows the breakdown of global central estimates by disease and by environmental driver.
Table 9. Mortality impacts attributable to a 1 TgS/yr geoengineering scenario. A surface cooling of 0.98 K is assumed.
Mortalities were calculated based on scaled model output. Central estimates are in bold, 95% confidence intervals in square brackets.
O
3
PM
2.5
UV
Total
Cooling
Reversal
Drying Subsidence
Side effect
Strat. burden Total
-45,000 -670
[-11,000 : -75,000] [-170 : -1,100]
69,000 14,000 3,900 -1,100
[32,000 : 98,000] [6,400 : 19,000] [1,800 : 5,600] [-540 : -1,100]
440
[150 : 720]
-22
[-7 : -37]
-
-
-18,000 -64,000
[-4,600 : -30,000] [-16,000 : -110,000]
3,200
[1,100 : 5,500]
85,000
[40,000 : 120,000]
3,700
[1,200 : 6,200]
24,000 13,000 3,900 -16,000 25,000
[-24,000 : 69,000] [5,700 : 19,000] [1,800 : 5,600] [-2,200 : -28,000] [-39,000 : 86,000]
Page 87 of 159
The global impact is dominated by PM
2.5
resulting from surface cooling, mitigated by reductions in surface ozone. In North and South America, PM
2.5
cooling and drying impacts are either counteracted by negative ozone impacts or are negative themselves, resulting in a surface air quality impact of 1 prevented mortality per million people in North America and 2 prevented mortalities per million in South America.
Total impacts in North and South America are also small per capita compared to those in Afro-Eurasia, where large PM
2.5
increases are only partially balanced by decreased ozone concentrations. Mortalities in
Afro-Eurasia due to cooling and drying are increased by up to 5 per million and 3 per million, respectively. Globally the PM
2.5
impacts of reversing climate change exceed the ozone benefits, resulting in 37,000 additional premature mortalities per year.
Figure 24. Premature mortalities incurred globally per TgS/yr injected. Error bars show CRF 95% confidence intervals based on
Monte Carlo simulation (Latin Hypercube sampling with n = 100,000)
Page 88 of 159
By contrast, the surface air quality side effects of SAI result in a reduction in mortalities in Asia at a rate of 2 prevented mortalities per million, but are responsible for an increase in mortality rates in Europe and
Africa. An aggregate change of +15,000 prevented mortalities is found due to surface air quality side effects of sulfate aerosol engineering. Combining the surface air quality consequences of reversal of climate change and unique SAI side effects, sulfate engineering has net surface air quality benefits for the
Americas but damages throughout Afro-Eurasia.
The mortality impacts of increased surface UV are well correlated with latitude and greater in the southern hemisphere, following the pattern of stratospheric ozone depletion. This is the only impact mechanism for which the side effects of sulfate aerosol injection are greater than the direct consequences of reversing climate change. A 3,700 increase in global melanoma mortality corresponds to an additional
0.4 melanoma mortalities per million population globally, but 2 per million in Europe. This does not account for under-reporting of existing cases, behavioral change or protection due to skin reflectance. It also does not take into account non-melanoma skin cancers, which are typically less fatal but occur at greater rates (63).
Although the overall change in mortality from respiratory disease due to ozone exposure is 25% smaller than the increase from cardiovascular disease due to PM
2.5
, the background rate of respiratory disease mortality is 78% smaller than the rate of cardiovascular disease mortality. Proportionally, the impact on respiratory disease mortality is therefore greater than on overall cardiovascular disease mortality.
Page 89 of 159
Table 10. Baseline and geoengineering-attributable mortality rate per million people worldwide
Respiratory disease
Cardiovascular disease
Melanoma
All cause
Baseline
490
2,200
13
7,900
SAI
-7.2
[-1.8 : -12]
+9.6
[4.5 : 14]
+0.41
[0.13 : 0.70]
+2.8
[-4.4 : 9.6]
The reported uncertainty is a 95% bound and arises from uncertainty in each of the underlying CRFs.
Some other sources of uncertainty, such as the choice of CRF and non-linearity, are discussed and quantified above and below. Uncertainties due to differences between the 2040 and 2006 climate, as well as due to climatological changes attributable to geoengineering which were not included here, are not quantified. Since these will affect, for example, baseline stratospheric ozone concentrations and the rate of exchange between the stratosphere and troposphere, quantification of these uncertainties is an important future research goal. However, it is outside of the scope of this work.
Since ozone and PM
2.5
impacts mostly cancel each other out in terms of net premature global mortalities, the sign of the result varies within the 95% bounds, with an estimated 24% probability of an overall negative result. The overall confidence interval is 5 times as great as the estimated total mortality impact.
Geoengineering approaches are often evaluated in the context of reversing the temperature increase associated with a doubling of CO
2
, estimated at a 2.1 K change (170). Based on the calculated sensitivities, such a scenario would result in 53,000 premature mortalities per year in 2040; 79,000 additional mortalities due to the temperature and precipitation reduction, moderated by 26,000 fewer mortalities due to the direct effects of SAI. This can be contrasted against a ‘peak-shaving’ scenario, in which a smaller cooling is produced with the aim of limiting the rate of climate change (171). For example a 0.5 K cooling scenario would cause 13,000 mortalities per year.
Page 90 of 159
4.3.1
Sensitivity to choice of ERF
Mortality attributable to PM
2.5
has been investigated in a large number of epidemiological cohort studies and meta-analyses over the past several decades, resulting in a variety of different rates and functional forms for the response of chronic mortality rates to an increase in long-term PM
2.5
loading (41, 42, 85,
86). Two factors in particular are uncertain: which mortality outcomes are affected, and the shape of the response function.
The sensitivity of calculated PM
2.5
impacts to this result is calculated using a variety of different response
functions. The functions considered are shown in
Page 91 of 159
incidence rates for each relevant disease. Note that the Hoek and ACS all-cause CRFs are identical.
Although the Hoek cause-specific CRF considers only cardiovascular disease, it was determined based on meta-analysis of other CRFs, such as that derived from the ACS re-analysis, which are relevant to cardiopulmonary mortality. It is therefore a lower-bound estimate of mortalities attributable to PM
2.5
. The
ACS study also produced several different estimates of the relative risk (RR) of cardiopulmonary mortality per 10 µgm -3 PM
2.5
. This sensitivity analysis is conducted using the RR for 1999-2000 from the random-effects Cox model analysis, consistent with Anenberg et al (52), yielding a central excess risk of
13% per 10 µgm -3 PM
2.5
. However, Hoek et al use the 1979-1983 estimate to build their CRF, which has the same central excess risk of cardiovascular disease of 6% per 10 µgm -3 PM
2.5
.
Page 92 of 159
Table 11. Literature CRFs for PM
2.5
. The Hoek and EPA reviews include a large number of additional CRFs not shown here.
ID Mortality endpoint
ACS (85) All-cause
Form Study area Method
Log-linear USA Cohort
ACS (85) Cardiopulmonary disease, lung cancer Log-linear USA
EPA (172) All-cause Log-linear USA
Cohort
Expert elicitation
Hoek (42) All-cause
Hoek (42) Cardiovascular disease
Log-linear Global
Log-linear Global
IRF-U (10) Cardiopulmonary disease, lung cancer IRF (86)
Ostro (110) Cardiopulmonary disease, lung cancer Log-log
Global
Global
Meta-analysis
Meta-analysis
Meta-analysis
Meta-analysis
health endpoint resulting from a change in local concentration, and has been used in other global studies of mortality attributable to changes in surface air quality (52, 53, 87). However, the log-linear form does not capture the saturation at very high concentrations observed in cohort studies of frequent and persistent cigarette smokers (86). This has motivated the development of alternative risk functions. Some studies employ high concentration thresholds beyond which no increase in risk is assumed. In these studies a lower and/or higher concentration threshold (LCT/HCT) is assumed on the basis that health risks beyond these points are negligible or that the epidemiological data on which the function was calculated cannot be extrapolated beyond these points (173). The minimum and maximum concentrations observed in the underlying study are often used for this purpose. Alternatively, more complex forms are employed. These include a log-log form developed for the WHO in 2004 (110) which saturates at high exposures, and which was built upon to create an integrated risk function (IRF) in 2014 (86). Although recommended coefficients for the former are supplied by the WHO based on data available at the time of publication, the latter function is designed to be flexible and therefore requires an iterative procedure to be applied to find appropriate concentration response function. For the purposes of this sensitivity analysis, coefficients are taken from calculations by Morita et al (10), who used the IRF to estimate global mortalities
Page 93 of 159
baseline concentration of PM
2.5
, the steeper the initial rate of increase of mortality rate once the lower threshold is exceeded; however, at high concentrations the CRF saturates, and the gradient decreases.
Figure 25. Graphical representation of the different PM
2.5
CRFs. For the IRF-U and Ostro CRFs, three baseline conditions are shown: 5 µgm -3 (dashed), 15 µgm -3 (solid) and 25 µgm -3 (dash-dot)
endpoint. Cardiovascular diseases are shown in blue, while respiratory diseases are shown in red. Lung cancer is shown separately in yellow. The Hoek cause-specific CRF was used to produce the central estimate listed throughout this study, as it is the most recent globally-applicable CRF which does not rely on an iterative procedure to produce its response function. Where risks of cardiopulmonary disease are reported, as is the case with the cause-specific ACS (85) and Ostro (110) CRFs, this is assumed to include both cardiovascular diseases and respiratory diseases.
Page 94 of 159
Figure 26. Mortalities calculated using various alternative CRFs. HD here stands for Heart Disease, and COPD for Chronic
Obstructive Pulmonary Disorder
The all-cause CRFs have the advantage that they require only the total mortality rate for a given population fraction. However, these do not account for areas with high mortality rates not linked to air quality concerns. The 2009 American Cancer Society (ACS) CRF is the oldest of the three shown; the
2011 EPA CRF is based on an expert elicitation, and was chosen to lie between the ACS rate (85) and the higher reported rate from the 2002 Pope et al Harvard Six Cities analysis (41). Subsequent studies have tended towards the lower ACS rate, and have focused on cause-specific CRFs. The global results from the all-cause CRFs are, with the exception of the EPA result, within the bounds of the cause-specific CRFs.
The Hoek CRF yields the lowest overall mortality rate, but is equal to within 1% to the result calculated using the Burnett integrated risk function (IRF-U). All CRFs find cardiovascular mortality to be the dominant factor, contributing between 70 and 100% of the overall increase in premature mortality.
Application of the ACS cause-specific CRF results in a more than 100% increase in premature mortalities
Page 95 of 159
due to PM
2.5
. Globally, the reduction in ozone mortalities corresponds to 75% of the increase in premature mortalities due to PM
2.5
calculated using the Hoek CRF. After accounting for the reduction in mortality due to lower ozone exposure, application of the ACS CRF instead of the Hoek CRF would therefore increase the total estimated mortalities by a factor of approximately 5.
There is also uncertainty regarding which health impacts should be included when applying the ozone and
PM CRFs. This arises from uncertainty in the epidemiological studies from which the CRFs are derived, from changing classification standards, and from the challenges of collecting global mortality data, as different countries report mortality rates with differing degrees of accuracy and reliability. For example,
Lim et al (12) apply the Jerrett ozone CRF only to mortality rates attributable to chronic obstructive pulmonary disorder (COPD, ICD-10 codes J40-J44), whereas Lelieveld et al (54) apply it to all respiratory diseases, including respiratory infections and asthma (ICD-10 codes J00-J99). Yet other studies specify only “respiratory disease”, such as Fang et al (87). This study applies the Jerrett CRF only to COPD and asthma (ICD-10 codes J40-J46). Non-inclusion of asthma would reduce total calculated ozone-related mortalities attributable to geoengineering by 11%.
The same issues apply for PM
2.5
, where studies have differed over whether to include respiratory disease, and which specific diseases to include under the umbrella terms of cardiovascular or cardiopulmonary disease. As we apply the Hoek cardiovascular CRF, we include only heart disease and stroke
(cerebrovascular disease). However, other studies have included respiratory disease when applying the
Krewski et al CRF (85). Using the same definition for respiratory disease as was applied for ozone impacts, inclusion of respiratory disease when calculating PM
2.5
impacts using the Hoek CRF would increase calculated mortalities by 47%. Using the ACS lung cancer CRF would increase mortalities by
9%. The effect of excluding or including different diseases under cardiovascular disease is shown
graphically in Figure 26. The most important contributor is cerebrovascular disease (stroke), which
Page 96 of 159
contributes 42% of the total calculated mortalities. Although some studies also include upper and lower respiratory diseases when calculating respiratory disease related mortalities, we find this increases estimated mortalities by less than 0.001%.
Figure 27 shows the breakdown of PM
2.5
and ozone impacts based on hypothetical exposure thresholds, beyond which it is assumed that there are no further health impacts. Pale bars correspond to the fraction of premature mortalities avoided or incurred in regions where the simulated exposure is below a low concentration threshold (LCT), while the darker bars show the fraction in regions exceeding a high concentration threshold (HCT). Estimates are calculated based on the log-linear, no-threshold Hoek CRF.
LCTs represent the possibility that, at very low exposures, small changes in exposure may not affect human physiology. HCTs represent the fact that linear extrapolation of relative risk with concentration results in unrealistic values at very high concentrations, based on our understanding of the risks of extreme PM
2.5
exposure (e.g. in the case of active smokers) (86). Note that the Ostro CRF saturates at high concentrations for this reason, and the Burnett IRF (IRF-U) is designed to allow for both an LCT and
HCT.
Page 97 of 159
Figure 27. Effect of applying low and high concentration thresholds (LCTs and HCTs) when calculating mortalities with the
Hoek PM
2.5
and Jerrett O
3
chronic mortality CRFs
For ozone, an LCT of 33.3 ppbv was chosen; this was the lowest concentration measured in the ACS study which forms the basis of the Jerrett CRF. Application of the LCT reduces ozone mortality estimates by 1%, as almost all regions worldwide are projected to exceed this threshold value. The exceptions are
South America and the West Pacific. Due to the relatively clean air in these regions, application of a lower concentration threshold reduces ozone exposure by 12% and 70% respectively.
The application of a 5.8 µgm -3 LCT for PM
2.5
exposure reduces global PM
2.5
mortality by 8%, 55% of which comes from lightly-polluted areas in Asia. Premature Asian mortalities attributable to PM
2.5
resulting from geoengineering are reduced by 5%, the smallest relative decrease of all regions, compared to a 62% reduction in North America and a 100% reduction in the West Pacific region. This reflects the fact that the majority of air quality related premature mortalities occur in already-polluted regions of Asia.
This is further highlighted by the application of a 30 µgm -3 HCT, which reduces premature mortality due to geoengineering-attributable PM
2.5
by 26%. The HCT has no effect outside of Africa and Asia, as only
Page 98 of 159
these regions locally exceed the HCT on an annual-average basis. However, applying this HCT reduces attributable African PM
2.5
by 8%, and Asian by 29%. No HCT is applied for ozone, as peak ozone season
1-hour maximum ozone values do not exceed the 100 ppbv maximum concentration observed in the ACS study in any simulation.
Overall, application of concentration thresholds has a negligible impact on ozone mortalities but could significantly affect PM
2.5
mortality. An LCT of 5.8 µgm -3 reduces global mortality by 8%, predominantly affecting Western and developed regions. An HCT of 30 µgm -3 , on the other hand, reduces global mortality by 26%, with 99% of this decrease occurring in Asia.
4.3.2
Non-fatal health impacts
both bronchodilator usage and symptom day frequencies are decreased overall by geoengineering, due to the greater influence of reduced ozone levels compared to increased PM
2.5
. Net hospital admissions also decrease, with 54,000 fewer respiratory hospital admissions per year but an additional 22,000 cardiac hospital admissions. The relative impact of PM
2.5
and ozone is shown in Figure 28. As in the case of
mortality impacts, the conflict between increased PM
2.5
and decreased ozone exposure results in a small net change with high uncertainty relative to the central estimate. However, the confidence intervals on
PM
2.5
exposure impacts with respect to both hospital admissions and symptom/restricted activity days are significantly smaller than those for ozone relative to the magnitude of the central estimate.
Increased PM
2.5
exposure also results in 90,000 additional cases of chronic bronchitis per year. In an analysis of Chinese health impacts, Matus et al estimate each case of chronic bronchitis to incur economic damages 12 times greater than a mortality due to acute exposure, and 28 times greater than a hospital admission. Cost ratios are greater still in Europe, with each case of chronic bronchitis estimated to cost the same as 95 hospital admissions (15, 48). Although the uncertainty interval on this result is over 200%
Page 99 of 159
the magnitude of the central estimate, the magnitude and severity of this outcome makes further investigation of this impact pathway a priority.
An important next step will be to monetize these impacts. Based on the ExternE costs, which are calibrated for the European Union, the greatest economic effect of the listed non-mortality outcomes would be due to the reduction in lower respiratory symptom days in children and minor restricted activity days in adults. Although these are individually less severe than an additional hospital admission or case of chronic bronchitis, the cumulative impact of 390 million fewer minor restricted activity days in the working population could be significant.
Table 12. Non-fatal health impacts per TgS/yr injected. *LRS: Lower Respiratory Symptom
Cause
O
3
Outcome Incidence
Adult bronchodilator usage -53,000
(thousands) [18,000 : -110,000]
O
3
O
3
Child LRS* days (cough) -720
(millions) [150 : -1,700]
Child LRS* days (other)
(millions)
-120
[330 : -620]
O
3
Minor restricted activity days
(millions)
-390
[-150 : -630]
O
3
Respiratory hospital admissions -90,000
[40,000 : -220,000]
PM
2.5
PM
2.5
Child bronchodilator usage 220
(thousands) [-860 : 1,300]
Adult bronchodilator usage 7,400
(thousands) [-7,400 : 22,300]
PM
2.5
PM
2.5
PM
2.5
PM
2.5
Child LRS* days
(millions)
Adult LRS* days
(millions)
Restricted activity days
(millions)
Cardiac hospital admissions
130
[66 : 200]
160
[19 : 300]
190
[160 : 210]
22,000
[11,000 : 34,000]
PM
2.5
Respiratory hospital admissions 36,000
[20,000 : 53,000]
PM
2.5
Chronic bronchitis 90,000
[-6,400 : 180,000]
Page 100 of 159
Figure 28. Non-mortality impacts related to surface air quality impacts due to geoengineering. Note that only minor restricted activity days are calculated for ozone
Global incidence of non-melanoma skin cancer (NMSC) is projected to increase as a result of increased
above, this does not take into account different likelihoods of incidence by different population groups.
Although data is sparse regarding absolute NMSC incidence rates, annual average treatment cost of skin cancer in the US between 2007 and 2011 was $8.1 billion per year (119). An increase in incidence of the magnitudes shown here is therefore likely to have a significant human and economic impact.
Table 13. The change in skin cancer case rates resulting from geoengineering
Skin cancer sub-type Δ 95% interval
Squamous-cell carcinoma +11% (+7.6%, +14%)
Basal-cell carcinoma +5.9% (+4.2%, +7.6%)
Cutaneous malignant melanoma +2.5% (+0.84%, +4.2%)
Page 101 of 159
Figure 29. Population-weighted mean percentage change in skin cancer incidence resulting from geoengineering
4.4
Summary
Surface air quality impacts associated with sulfate aerosol engineering are shown to be dominated by the direct consequences of reducing global temperatures. In the absence of temperature and precipitation effects, an established 1 TgS/yr sulfate aerosol injection scheme would result in 18,000 fewer mortalities per year in 2040 due to reductions in surface ozone but incur an additional 1,100 premature mortalities due to increased exposure to particulate matter. By comparison, condensation of nitrate due to lower temperatures coupled with high projected aerosol precursor emissions in Afro-Eurasia results in 69,000 premature mortalities per year, while slowing of the hydrological cycle results in a further 14,000 premature mortalities as rainout and washout are reduced. Lower temperatures also suppress ozone formation, resulting in a compensatory effect of 45,000 fewer mortalities per year due to respiratory disease. A net decrease in respiratory hospital admissions of 54,000 per year is expected due to the
Page 102 of 159
combined change in ozone and PM
2.5
, but cardiac hospital admissions are projected to rise by 22,000 per year. An additional 90,000 cases of chronic bronchitis per year are also expected due to PM
2.5
exposure, with no mitigating effect from reduced ozone.
Due to their relative isolation and low precursor emission projections, North and South America are found to benefit from climate engineering, with mortality rates falling by 1.7 and 2.6 per million population respectively (12,000 and 13,000 prevented mortalities). Although Asia benefits from sulfate aerosol engineering’s side effects, the disbenefits of climate change reversal and large projected population result in an overall increase of 4.2 mortalities per million population (21,000 premature mortalities). Europe and
Africa are the only regions negatively affected by both climate change reversal and sulfate engineering side effects, resulting in mortality rate increases of 8.0 and 0.29 per million (5,800 and 570 premature mortalities) respectively.
However, the increased stratospheric aerosol surface area density resulting from sulfate aerosol engineering is found to result in a 5.7% reduction in the global ozone column. This ozone layer depletion would increase surface UV-B exposure by 3.0% and incur 3,700 premature mortalities due to skin cancer every year, concentrated in Europe and Asia. Meanwhile incidence of non-melanoma skin cancer would increase by between 4 and 14% globally. Tropospheric cooling would also reverse the ozone enhancement resulting from global warming and therefore increase UV-B exposure, resulting in an additional 440 skin cancer mortalities per year.
This study does not address uncertainty regarding stratospheric ozone impacts of geoengineering, which are the dominant factor for the calculated health impacts. An investigation of interactions between geoengineering aerosols and future very short lived halogen species emissions by Tilmes at al (174) showed that different inventories of these species could result in changes of up to 12% in local erythemal
UV. This study also does not address the known sensitivity of geoengineering effectiveness to the
Page 103 of 159
stratospheric aerosol distribution achieved (101, 161, 175). Inclusion of an aerosol microphysics module for stratospheric aerosols would help to address this issue.
Due to the overall dominance of temperature and precipitation over net health impacts, a more comprehensive assessment of climatological interactions with surface air quality and UV-B exposure is needed. Accurate calculation of local precipitation impacts in particular is a critical future research objective due to the transboundary impacts of regional precipitation change. A natural next step would be to run a unified assessment of geoengineering in a climate model and compare the overall impact with the linear sum of sensitivities presented here. This would also capture changes in impacts due to the change in background climate which will occur between 2006 and 2040, which is not within the scope of this work.
Page 104 of 159
5 Human health impacts of aircraft emissions
In this chapter, the contribution of aviation emissions to the global burden of disease is calculated. This provides the first unified global estimate of aviation’s impacts on surface air quality and UV-B, including mortality and non-mortality impacts. The mechanisms responsible for these impacts are also identified.
5.1
Experimental design
All model runs are performed using the GEOS-Chem UCX model, with 5 years of model spinup.
Analysis is performed using the final year of data, relevant to 2006. The aviation emissions dataset is described in detail below. Non-aviation emissions, both natural and anthropogenic, are simulated using the standard GEOS-Chem global emissions data set. This includes the EDGAR v3 global anthropogenic emissions inventory for each year (176), overwritten with regional inventories where available. Biogenic emissions are calculated using the MEGAN inventory (177). Population exposure and health impact
a long-term change in exposure.
Aircraft emissions are taken from a global inventory produced for 2006 from the FAA Aviation
Environmental Design Tool (AEDT). This provides flight tracks and estimated fuelburn along with NO x
,
CO and hydrocarbon emission rates, with a total fuelburn of 188 Tg in 2006. Black carbon (BC) and organic carbon (OC) emissions indices, along with the average speciation of hydrocarbon emissions, are
Page 105 of 159
Figure 30. Annual average fuel burn distribution (kgm -2 , log scale) for 2006
5.2
Exposure estimates
shows the sensitivity of population exposure to global aviation emissions. These numbers are calculated by subtracting a zero-aviation-emissions baseline simulation from one including aviation emissions. Globally, full-flight aviation emissions increase population-weighted PM
2.5
exposure for 2006 by 0.053 µgm -3 . 83% of the aviation-attributable increase in population-weighted PM
2.5
exposure is nitrate aerosol by mass. The greatest single-cell increase in annual-average PM
2.5
is a change of +0.29 µ gm -3 in
Eastern Europe, 5.4 times the global mean increase in exposure. As discussed in Barrett et al and Lee et al
(6, 92), PM
2.5
formation in Europe is nitric acid limited, and additional nitrate PM
2.5
forms as long as
ammonia nitrate formation globally. With the exception of South Asia, the regions with the greatest background ammonia concentrations also experience the greatest increases in nitrate aerosol concentrations as a result of aviation emissions. European PM
2.5
exposure is most sensitive to aviation, increasing by 0.10 µgm -3 .
Page 106 of 159
Table 14. Sensitivity of population PM
2.5
exposure to aviation emissions
Region Response
Global 160
N. America
S. America
3.8
0.18
Europe
Asia
Africa
W. Pacific
48
106
1.8
0.014
Global
N. America
S. America
Europe
Asia
Africa
W. Pacific
0.053
0.014
0.0011
0.10
0.058
0.0064
0.00081
Figure 31. Background ammonia concentrations for 2006.
Annual average population ozone exposure resulting from aviation is shown in Table 15. Aviation
emissions are responsible for between 20 and 40% of cruise-altitude NO y
, although the specific fraction is sensitive to the time of year, the choice of emissions inventories and estimates of lightning NO x
(5). The
Page 107 of 159
ozone which is subsequently produced by aviation NO x
emissions mixes throughout the free troposphere in the Northern Hemisphere, resulting in diffuse exposure compared to locally-produced PM
2.5
shows the response of column total and surface average tropospheric ozone to full-flight aviation
ppbv at 11-12 km, from 30-90°N. Decreases of up to -4 ppbv are observed in the stratosphere. The pattern matches that of Köhler et al (22), who found that aviation NO x
increases cruise-altitude ozone by 4-5 ppbv.
Table 15. Sensitivity of population ozone exposure to aviation emissions
Region Response
Global
N. America
S. America
Europe
Asia
Africa
W. Pacific
1600
180
40
230
1000
140
3.6
Global
N. America
S. America
Europe
Asia
Africa
W. Pacific
0.54
0.66
0.24
0.49
0.57
0.50
0.20
Surface 1-hour maximum ozone exposure during local ozone season is increased in all locations by an average of 0.54 ppbv, with greater concentrations in the Northern Hemisphere. The change in surface ozone is the result of descending free tropospheric air, which contains increased ozone as a result of enhanced free-tropospheric NO x
concentrations. Zonal average surface NO x
between 30 and 60°N is actually decreased during Northern hemispheric winter (December, January and February, DJF) as a result of aviation emissions, coinciding with the peak increase in surface ozone mixing ratios. This is explored further below. The greatest increase in surface ozone occurs over the Tibetan plateau due to the
Page 108 of 159
combination of a high surface elevation and local large-scale subsidence of air from the free troposphere.
The smallest ozone changes occur in regions with high background NO x
concentrations, such as continental Europe, the North-Eastern states of the USA and East China.
Figure 32. Annual average column and surface ozone perturbation due to full-flight aviation emissions. Change in column ozone is relative (%), whereas surface ozone changes are shown in ppbv.
Page 109 of 159
Figure 33. Annual average zonal ozone perturbation due to full-flight aviation emissions
Figure 34 shows the seasonal trend of ozone and PM
2.5
exposure resulting from global aviation emissions.
Ozone impacts are greatest during hemispheric winter, with peak impacts of 1 ppbv in North America,
Europe and Asia occurring in December and January. The increase in surface ozone attributable to aviation during Northern Hemispheric winter has been observed in other studies (92, 179). However, since ozone season in the Northern Hemisphere is April through September, this peak is not included in the estimate of ozone health impacts. The same pattern holds for the Southern Hemisphere, although total impacts are of the order of 0.1 ppbv.
The seasonal trend of PM
2.5
exposure matches that of ozone. In Europe and Asia, which are the only two regions for which regional average changes greater than 0.03 µgm -3 are calculated, the peak increase in
PM
2.5
exposure occurs during hemispheric winter, coinciding with the peak increase in ozone exposure.
The pattern closely matches that of the regional ozone perturbation. This highlights the connection between aviation-induced oxidation of particulate matter rather than direct transport of primary emissions to the surface, which has been estimated to be a negligible contributor to surface PM
2.5
(20). Aviation
Page 110 of 159
emissions result in zonal concentrations of NO x
(NO + NO
2
) falling during the DJF season, indicating that the increase in nitrate is not due to the local presence of additional aviation-attributable NO x
.
Figure 34. Seasonality of ozone (left) and PM
2.5
(right) exposure resulting from global aviation
To determine the underlying mechanism, chemical mixing ratios are isolated for a region with high
sensitivity to aviation emissions. Figure 35 shows the seasonal cycle of ozone and NO
x
from 0 to 3 km pressure altitude over two simulation years over Western Europe (2.5°W to 22.5°E, 40 to 56°N). The absolute contribution of aviation-attributable ozone is maximized when background ozone is at a minimum, coinciding with a net decrease in surface NO x
as it is converted to nitrate aerosol. The increase in surface nitrates mirrors the decrease in NO x
, as shown in Figure 36. The reduction in NO
x
occurs below
2 km altitude in all months, with no visible signal in Northern Hemispheric summer. Three events therefore occur simultaneously: background surface NO x
levels increase, background ozone levels decrease, and aviation-attributable ozone increases without any increase in aviation-attributable NO x
.
Ozone, nitrate radical and N
2
O
5
levels rise simultaneously starting in November 2005, each increasing by
4-5% in December relative to baseline levels. This is accompanied by a 4% increase in HNO
3
and 3% increase in nitrate aerosol concentrations, while NO x
concentrations fall to 2% below baseline values. The combination of a large relative increase in ozone and high background NO x
therefore results in increased night-time nitrate production via the NO
2
+ O
3
⟶ ∙ NO
3
+ O
2
reaction. This then increases local nitrate
Page 111 of 159
aerosol via N
2
O
5
and/or HNO
3
. Nitrate aerosol formation is fastest during hemispheric winter due to the cold temperatures, which prevent evaporation of volatile nitrate aerosol (29). The ratio of free ammonia
([NH
4
+ ] + [NH
3
] - 2×[SO
4
2]) divided by total nitrate ([HNO
3
] + [NO
3
]), known as the gas ratio (180), averaged 1.6 throughout the period for this region, falling rapidly once temperatures are low enough to permit nitrate formation. Ammoniated aerosol formation is therefore nitric-acid limited in this region.
This is the same mechanism as postulated by Lee et al (92), although they calculate a smaller surface response of 0.1 ppbv.
The identified impact mechanism for exposure to both ozone and PM
2.5
is vertical mixing of aviationattributable ozone to the surface. The calculated exposures are therefore sensitive to the rate of vertical mixing in the free troposphere in the GEOS-5 meteorological dataset. An unquantified source of uncertainty is the error in baseline vertical mixing rates calculated by the underlying meteorological model, and the change in mixing and downward transport induced by aviation emissions, which another study has found to be significant (9).
The increase in tropospheric ozone also affects surface UV-B exposure. Globally, aviation reduces population-weighted UV-B exposure by 0.63%. UV-B exposure in the USA specifically is reduced by
0.80%. This is greater than the global average due to the smaller relative increase in ozone column over
Asia.
Page 112 of 159
Ozone
NO x
Figure 35. Baseline (bottom) and aviation-attributable (top) ozone and NO x
from the surface to 3 km pressure altitude over
Western Europe. Years 2005 and 2006 are shown
Page 113 of 159
Figure 36. Time series of baseline (bottom) and aviation-induced (top) surface ozone, NO x
and inorganic aerosol over Western
Europe for 2005 and 2006
5.3
Health impact sensitivities
are presented to allow comparison to other studies. PM
2.5
exposure attributable to global aviation emissions result in an additional 9,500 premature mortalities per year, similar to existing studies (see
section 0). These impacts are concentrated in Europe and Asia due to high background ammonia
concentrations, and these two regions account for 28% and 70% respectively of global aviationattributable premature mortalities due to PM
2.5
. Meanwhile premature mortalities due to ozone are concentrated in Asia, in which 83% of the additional 7,000 mortalities are incurred. Total skin cancer mortalities are an order of magnitude smaller, with 390 prevented mortalities per year due to reduced UV-
B exposure. Since this decrease in exposure is the result of increased free tropospheric ozone, these benefits are linked spatially to aviation emissions. 47% of skin cancer benefits are incurred in Europe and
24% in North America. Asia accounts for a further 23% of prevented skin cancer mortalities. Total
Page 114 of 159
mortality information is also expressed as the increase in total mortality rate per million people globally in
in 2040 , with a smaller absolute number of mortalities due to a larger projected population.
Table 16. Mortality impacts attributable to aviation emissions. Central estimates are in bold, 95% confidence intervals in square brackets. Total impact confidence intervals are calculated using Monte Carlo methods (Latin Hypercube sampling, n = 10 5 )
O
3
PM
2.5
UV
Total
Aviation
(Global)
7,000
[1,800 : 11,000]
9,500
[4,200 : 14,000]
-390
[-130 : -660]
16,000
[8,400 : 23,000]
Aviation
(N. America)
450
[120 : 750]
130
[59 : 180]
-95
[-31 : -160]
490
[130 : 790]
Table 17. Baseline and aviation-attributable mortality rate per million people worldwide
Respiratory disease
Cardiovascular disease
Melanoma
All cause
Baseline
490
2,200
13
7,900
Aviation
+1.1
[0.27 : 1.8]
+1.5
[0.65 : 2.2]
-0.06
[-0.02 : -0.10]
+2.5
[1.3 : 3.5]
Total SAQ-related non-fatal health impacts of aviation emissions are shown in Table 18. Globally,
aviation incurs an additional 12,000 hospital admissions per year via ozone and PM
2.5
exposure. Since both species are increased by aviation, the relative uncertainty of these impacts after addition is smaller
by ozone, most significantly 530 additional respiratory hospital admissions per year.
Page 115 of 159
Figure 37. Non-mortality impacts due to surface air quality effects of aviation emissions
Additional tropospheric ozone due to aviation emissions is estimated to decrease non-melanoma skin cancer incidence worldwide by 1.8/0.97% (SCC/BCC), and by 2.2/1.2% in the US.
Page 116 of 159
Table 18. Non-fatal health impacts due to aviation. *LRS: Lower Respiratory Symptom
Cause
O
3
O
3
O
3
O
3
PM
2.5
PM
2.5
PM
2.5
Outcome
Adult bronchodilator usage
Aviation
(Global)
3,100
(thousands) [-1,100 : 6,700]
Child LRS* days (cough)
(millions)
Child LRS* days (other)
(millions)
Minor restricted activity days
(millions)
Child LRS* days
(millions)
Adult LRS* days
(millions)
Restricted activity days
(millions)
130
[-27 : 310]
22
[-60 : 110]
54
[21 : 87]
Aviation
(USA)
390
[-140 : 840]
4.3
[-0.88 : 10]
0.74
[-2.0 : 3.8]
2.7
[1.0 : 4.3]
O
3
Respiratory hospital admissions 6,300 530
[ -2,500 : 15,000] [-210 : 1,300]
PM
2.5
Child bronchodilator usage
(thousands)
15
[-58 : 90]
0.49
[-1.9: 2.9]
PM
2.5
Adult bronchodilator usage 500
(thousands) [-500 : 1,500]
16
[-16 : 49]
14
[6.9 : 21]
15
[1.8 : 28]
19
[16 : 21]
0.22
[0.11 : 0.33]
0.27
[0.031 : 0.51]
0.32
[0.28 : 0.36]
PM
2.5
Cardiac hospital admissions 2,200
[1,100 : 3,300]
PM
2.5
Respiratory hospital admissions 3,500
[1,900 : 5,200]
PM
2.5
Chronic bronchitis 8,000
[-580 : 16,000]
38
[19 : 57]
62
[34 : 90]
151
[-11 : 310]
5.4
Comparison to other studies
shows central mortality impacts, for global aviation emissions only, against existing literature estimates. Only Jacobson et al and Yim et al calculate ozone mortalities, and only Yim et al report chronic mortality using the Jerrett et al CRF. Jacobson et al use a short-term mortality CRF which employs a 35 ppbv lower concentration threshold and a sensitivity of +0.4% all-cause mortalities per 10 ppbv increase in 1-h maximum ozone. This is based on a 2006 study of acute mortality in California due to changes in daily ozone exposure (16). Yim et al report 2,100 additional mortalities world-wide resulting from a 0.6 ppbv increase in global average ground-level ozone, compared to 7,000 in this study resulting from a 0.5
Page 117 of 159
ppbv increase in population-weighted ozone. The factor of 3 difference between this result and the Yim et al result is likely due to differences in the underlying CRF. Based on the figures quoted above, they find an approximate sensitivity of 3,500 mortalities per ppbv. Other studies using the Jerrett ozone CRF report sensitivities of 18,000 to 19,000 mortalities per ppbv. The sensitivity in this study is approximately
12,900 mortalities per ppbv. This smaller figure is the result of classifying only COPD and asthma as respiratory diseases exacerbated by ozone exposure, which have a global mortality rate in the 30-plus age range of 0.1%/year compared to the figure of 0.134% used by Anenberg et al (52). Scaling the results from this study to match the Anenberg et al sensitivity yields a revised value of 17,300 mortalities per ppbv. The result is also dependent on the distribution of impacts.
Table 19. Central mortality impacts in this and other studies of aviation emissions. *Jacobson et al report short-term, rather than chronic, ozone-related mortality.
O
3
PM
2.5
UV
Total
Barrett et al
(6)
-
13,000
-
13,000
Morita et al
(10)
-
400
-
400
Jacobson et al (9)
310*
310
-
620
Yim et al
(11)
2,100
14,000
-
16,000
Koo et al (8)
-
26,000
-
26,000
This study
7,000
9,500
-390
16,000
PM
2.5
mortalities calculated in this work are of the same order of magnitude as three of the studies but are two orders of magnitude greater than those calculated by Jacobson et al (9) and by Morita et al (10). Both
Jacobson and Morita use global chemistry-climate models, as opposed to the chemical-transport models used in this study and those of Koo, Yim and Barrett et al. Morita et al use the global chemistry-climate model NASA GISS ModelE2 to calculate the contribution of aviation emissions to surface PM
2.5
in 2006, finding a 0.002 µ gm -3 increase in global population-weighted exposure, compared to 0.053 µ gm -3 in this study. The cause of this difference is not known, although climate models have typically not been able to consistently resolve aviation’s impacts, as reported in a recent model intercomparison (181).
Page 118 of 159
The Yim et al study used high-resolution (between 36 and 50 km horizontal resolution) simulations in the regional model CMAQ to refine global results from GEOS-Chem and found a global average PM
2.5
perturbation of 0.0062 µ gm -3 compared to the value of 0.0053 µ gm -3 in this study. The smaller impact calculated by Morita et al is therefore not a function of grid resolution or of GEOS-Chem being a uniquely sensitive model. Yim et al also find that high-resolution modeling of nested grid domains yields a 12% increase in ozone impacts but a 29% decrease in PM
2.5
impacts. Applying these factors to the results of this study, PM
2.5
impacts are reduced to 6,800 premature mortalities per year, while ozone impacts are increased to 7,800 premature mortalities per year.
Although Barrett et al calculate an overall nitrate exposure impact smaller than in this study, their estimate of mortalities attributable to PM
2.5
is 40% greater.
This cannot be explained by differences in
CRF. Barrett et al employ the more sensitive Ostro log-log CRF but do not include cerebrovascular
mortality estimates to the Hoek CRF, including stroke, for the same increase in exposure. The observed
compares the baseline cardiopulmonary rates reported by the Barrett et al study for 14 global sub-regions to the average rate for the 7 regions analyzed in this study, based on the country-specific data given in the
WHO database. All but one of the sub-regions shows an average baseline cardiopulmonary mortality rate greater than the regional average from the WHO 2012 data. The Barrett et al baseline mortality data is therefore expected to be significantly greater than the country-specific data in this study, resulting in a greater sensitivity to PM
2.5
.
Page 119 of 159
Table 20. Cardiopulmonary mortality incidence per 1,000 people as reported by Barrett et al and in the latest WHO estimates
Sub-region
Africa (AFR/D)
Africa (AFR/E)
N. America (AMR/A)
S. America (AMR/B)
S. America (AMR/D)
Middle East (EMR/B)
Middle East (EMR/D)
Europe (EUR/A)
Europe (EUR/B)
Europe (EUR/C)
S. E. Asia (SEAR/B)
S. E. Asia (SEAR/D)
W. Pacific (WPR/A)
W. Pacific (WPR/B)
Barrett et al.
4.62
4.52
4.76
3.58
3.37
4.27
5.64
4.34
6.13
10.38
4.47
6.34
2.98
3.80
WHO 2012
3.7
3.4
3.2
4.1
5.5
4.1
2.7
Difference
+25%
+22%
+41%
+10%
+3.7%
+4.5%
+38%
-22%
+11%
+87%
+9%
+55%
+11%
+41%
The only existing estimate of skin cancer mortality due to aviation is a 2009 Master’s thesis (93). They used column ozone data from Köhler et al (22), along with linear scaling relations and latitude-dependent incidence data, to calculate the number of US NMSC cases and mortalities resulting from aviation emissions in 2002. They found that 20 (95% CI: 13 to 28) NMSC mortalities were prevented by increased column ozone due to aviation emissions, or a rate of approximately 40 prevented NMSC mortalities per
1% column ozone increase. By comparison, analysis of the results from the GEOS-Chem UCX show that
95 (95% CI: 31 to 160) fewer melanoma skin cancer mortalities occur in the US as a result of a 0.52% reduction in column ozone, or a rate of 180 prevented melanoma mortalities per 1% increase in ozone.
Brunelle-Yeung also found that SCC and BCC incidence rates in the US were reduced by 1.8-2.0% and
1.0-1.1% respectively. The results from this study (2.2% and 1.2%, discussed above) agree qualitatively, noting also that their results are based on 2002 aviation with total NO x
emissions of 0.68 TgN/yr compared to 0.81 TgN/yr in the 2006 AEDT inventory (22, 178).
Page 120 of 159
5.5
Summary
Global aviation emissions are found to contribute 16,000 premature mortalities in 2006 to the global burden of disease through surface air quality impacts alone, not including local and transcontinental impacts due to contrails. Of these impacts, 7,000 are due to increased ozone exposure, over three times greater than calculated in a recently published global study. Ozone mixing ratios are found to increase the most during winter, driving nighttime nitrate aerosol formation and thereby increasing annual average population PM
2.5
exposure by 0.053 µ gm -3 globally. However, the increase in tropospheric and lowerstratospheric ozone reduces surface UV exposure sufficiently to prevent 390 mortalities per year, 95 of which are in the USA. This is the first estimate of global melanoma mortality due to aviation, and the first estimate of skin cancer mortality impacts due to aviation using a global chemical transport model.
Uncertainty in the concentration response functions results in a 95% bound of 8,400 to 23,000 mortalities per year due to aviation. Uncertainty due to the climatological response to aviation is not quantified but could be significant due to the importance of vertical transport of ozone from the troposphere to the surface.
Future research objectives are split into two groups. The first is to increase the scope of the research.
Work on contrail modeling, for example on the CERM contrail prediction model (182), has found that contrails could produce up to 50 mW/m 2 of radiative forcing locally (183). The increase in cloud cover could have both local and global consequences through reduced UV flux in the tropospheric column underneath the contrails. In addition, natural cirrus clouds constitute an increase in the available surface area for heterogeneous chemistry and could therefore increase local ozone catalysis rates (184). This could therefore result in stratospheric ozone depletion. Reduced ozone would likely result in increased
UV at lower altitudes and offset some of the UV reductions due to contrail optical depth. The
Page 121 of 159
investigation of local and transcontinental impacts of aviation emissions given here could therefore be expanded to include contrail effects.
The second group of research objectives are related to modeling. The simulations for this study were run at a coarse global resolution, but Yim et al find that simulation at higher resolution increases ozone impacts by 12% while decreasing PM
2.5
impacts by 29%. This study also used a bulk aerosol model to capture aerosol impacts, and includes only a simple model of secondary organic aerosol formation. The bulk model assumes that inorganic aerosols exist only on the deliquescence hysteresis branch, which may lead to significant underestimates of summertime nitrate concentrations. Finally, a fixed surface boundary condition was applied for methane, which prevents capture of the methane feedback mechanism from aviation ozone. Future studies should quantify the impact of these assumptions.
Page 122 of 159
6 Conclusions
The first major contribution of this thesis is the production of a unified tropospheric-stratospheric extension, now available to the global modeling community. The GEOS-Chem UCX extends a community chemical-transport model, used by over 70 groups globally, into the stratosphere, allowing interactions between the troposphere and stratosphere to be explicitly modeled. Although other models have previously attempted to model both domains simultaneously, the GEOS-Chem UCX allows stratospheric perturbations to be propagated into one of the most widely used and chemically complete global chemical transport models currently available. As it is now incorporated into the main branch of the GEOS-Chem code, the UCX provides an estimate of stratospheric impacts resulting from any tropospheric or stratospheric chemical perturbation. Stratospheric gas-phase chemistry based on the
NASA GMI stratospheric CTM is implemented along with a state of the art photolysis model, stratospheric aerosol and heterogeneous chemistry to provide a comprehensive model of stratospheric chemistry. The UCX is shown to capture 90% of stratospheric Antarctic ozone depletion without compromising the previously demonstrated ability of GEOS-Chem to model the troposphere.
The UCX is applied to quantify sensitivities of human health to both modern and future high altitude emissions scenarios. This work contributes the first estimate of human health impacts resulting from hypothetical stratospheric sulfate aerosol injection schemes, including both premature mortalities and non-fatal health impacts. Novel findings include an overall negative mortality impact of stratospheric ozone depletion on geoengineering aerosols due to the conflicting effects of ozone-exacerbated respiratory disease and UV-B induced melanoma. Depletion of stratospheric ozone reduces surface ozone exposure as the ozone content of stratospheric intrusions and subsiding upper tropospheric air is decreased. The calculated reduction in mortality exceeds the increase in melanoma mortalities due to UV exposure, although non-fatal non-melanoma skin cancer incidence is also found to increase by 4-14%
Page 123 of 159
globally. The cancellation of large positive PM
2.5
and negative ozone exposure terms result in an overall estimate of 25,000 premature mortalities per year, with a 95% confidence interval of -39,000 to +85,000.
Reduction of the uncertainty in these results, which is a function of the underlying concentration response functions only, is an important future research objective. Although second-order effects due to crossterms and non-linearity are found to affect the outcome by less than ±5%, the influence of decreased temperature on other climatological quantities and the subsequent effect on surface PM
2.5
concentrations is highly uncertain (13, 185). Modeling changes in regional, rather than global, precipitation rates is also shown to significantly affect the distribution of impacts, increasing PM
2.5
exposure in Asia while decreasing exposure in Western Europe and North America. In light of the number of premature mortalities depending on this calculation, the contribution of climate feedbacks is an open research question which will require coupled chemical-climate modeling.
This work also provides the first estimate of global skin cancer mortality due to aviation emissions included as part of a unified estimate of aviation’s impacts on human health. Aviation emissions are found to reduce global skin cancer incidence by 390 mortalities per year, but increased surface ozone and
PM
2.5
result in a net additional 16,000 premature mortalities per year, with a 95% confidence interval of
8,400 to 23,000 mortalities. A mechanism for these impacts is identified which does not rely on transport of primary emissions to the surface. Cruise aviation emissions significantly increase free tropospheric ozone content, resulting in increased surface mixing ratios in winter months through mass subsidence.
Simultaneously, background NO x
levels are increased and the colder temperatures allow persistence of volatile nitrate aerosols. Together these conditions result in enhanced conversion of background NO x
to nitrate aerosol without the presence of primary aviation emissions. Global non-fatal skin cancer incidence is decreased, with a 2% reduction in squamous cell carcinoma incidence and a 1% reduction in basal cell carcinoma incidence. Combined with 8,000 additional cases of chronic bronchitis and 12,000 additional
Page 124 of 159
respiratory hospital admissions per year, this demonstrates a significant non-mortality health impact associated with aviation. However, this does not include impacts due to decreased oxidation in the free troposphere resulting from the formation of contrails. Quantification of these impacts constitutes a new research direction. A second question for future research is the potential effect of contrails on local chemistry through heterogeneous chemistry on the surface of the contrail ice crystals.
More general future research directions include quantification of meteorological feedbacks and a comprehensive assessment of UV-B exposure response functions. The Slaper et al CRF used for this thesis provides only an approximate estimate of skin cancer impacts due to an increase in local UV-B.
Given the finding of the tradeoff between ozone exposure and skin cancer due to stratospheric ozone depletion identified in chapter 4, refinement of this CRF will be needed to improve estimates of the human health risks of sulfate aerosol injection.
Finally, there are several improvements which could be made to the modeling approach. These include higher resolution simulations, implementation of size-resolved aerosols, and implementation of a full secondary organic aerosol formation mechanism. The changes in atmospheric oxidation rates responsible for the results stated above will also affect SOA production, likely increasing the magnitude of the change in PM
2.5
.
Page 125 of 159
Appendix A. Auxiliary information
Figure 38 shows global population density regridded to 4°×5° after scaling to national population
estimates for 2006 and 2040. Although impact calculations were performed at the native map resolution
(30″×30″), this lower resolution map is shown to provide a direct comparison to the various atmospheric responses shown throughout the text.
Figure 38. Global population density estimates for 2006 (top) and 2040 (bottom), downscaled to match the simulation grid
Page 126 of 159
Figure 39 shows the region designations used for the 6 continents when isolating regional impacts.
Figure 39. Region designations
Figure 40 shows the average global wind pattern between 1 and 11 km for 2006 through 2010. This is
provided to give context to local and transcontinental impact calculations.
Figure 40. Average horizontal wind speed between 1 and 11 km pressure altitude from 2006 to 2010
Page 127 of 159
Appendix B. Exposure response function formulae
The simplest ERF is linear, linking a concentration χ to a disease incidence rate I as
𝐼 = 𝛼 + 𝛽𝜒 where 𝛼 is some hypothetical zero-exposure incidence rate and 𝛽 is the observed rate of change of incidence with respect to concentration calculated in an epidemiological study. For the linear ERF this is equal to the excess risk ER, which is in turn the relative risk (RR) minus 1. The change in incidence between a baseline scenario B and policy scenario P is therefore
𝐼
𝑃
− 𝐼
𝐵
= Δ𝐼 = (𝛼 + 𝛽𝜒
𝑃
) − (𝛼 + 𝛽𝜒
𝐵
)
= 𝛽(𝜒
𝑃
− 𝜒
𝐵
) = 𝛽Δ𝜒
This is then multiplied by the local affected population to yield the change in total incidence. The parameter 𝛽 can be estimated based on epidemiological data. If an increase Δχ
𝑅
of the pollutant in question results in a relative risk 𝑅𝑅
𝑅
, this corresponds to an excess risk 𝐸𝑅
𝑅
= (𝑅𝑅
𝑅
− 1) . The slope 𝛽 is then simply 𝑑𝐼 𝛽 = 𝑑𝜒
≃
𝐸𝑅
𝑅
Δ𝜒
𝑅
For example, data from a 1987 study of restricted activity days (RADs) due to exposure to particulate matter (186) suggests that there are an additional 902 restricted activity days per 10 µgm -3 of ambient
902
PM
2.5
per 1,000 adults aged 15-64 (15). This corresponds to a relative risk 𝑅𝑅 = 1 + (
1000
) = 1.902
for
Page 128 of 159
Δχ
𝑅
= 10 µgm -3 and therefore 𝛽 = 0.0902
. Say that a small town with 15,000 adults aged 15-64 were to experience an annual increase of 0.1 µgm -3 PM
2.5
; the total increase in RADs would then be
15000 × 0.0902 × 0.1 = 135 additional RADs per year
All non-mortality health outcomes are calculated in this fashion, consistent with the ExternE 2005 methodology. Sometimes the relative risk is expressed as a change in the existing incidence rate. In this case, the above equation must be re-stated as
Δ𝐼 = 𝐼
𝐵 𝛽Δ𝜒 where 𝛽 is now the relative increase in incidence rate per unit increase in exposure. An advantage of the linear ERF is that only the absolute delta in concentration Δ𝜒 between two scenarios is required. For example, say that the change in incidence between two scenarios was to be calculated. This change would then be
Δ𝐼 = 𝐼
𝐵 𝛽((𝜒
2
− 𝜒
𝐵
) − (𝜒
1
− 𝜒
𝐵
))
= 𝐼
𝐵 𝛽(𝜒
2
− 𝜒
1
)
Knowledge of the baseline concentration 𝜒
𝐵
is therefore not necessary. This property is not true for the next example.
For mortality associated with chronic ozone and PM
2.5
exposure, the log-linear ERF is used instead. This has been used in several other studies estimating mortality impacts of air quality (6, 52, 53, 87). It takes the form ln 𝐼 ≈ ln 𝛼 + 𝛽𝜒
Page 129 of 159
∴ 𝐼 = 𝛼 exp(𝛽𝜒)
Relative risks for mortality attributable to chronic ozone and PM
2.5
exposure are typically expressed as relative increases in underlying mortality rates (42, 51). The change in incidence rate can therefore be expressed as
Δ𝐼 = 𝐼
𝑃
− 𝐼
𝐵
= 𝐼
𝐵
(
𝐼
𝑃
𝐼
𝐵
− 1)
= 𝐼
𝐵
( 𝛼 exp(𝛽𝜒 𝛼 exp(𝛽χ
𝑃
)
𝐵
)
− 1) = 𝐼
𝐵
(exp(𝛽(𝜒
𝑃
− 𝜒
𝐵
))
∴ Δ𝐼 = 𝐼
𝐵
(exp(𝛽Δ𝜒) − 1)
Once again the parameter 𝛽 can be estimated from a reference relative risk value linked to a reference increase in concentration. However, in this case the derivation is less straight forward. A recent metaanalysis of mortality attributable to PM
2.5
exposure (42) estimates an 11% increase in cardiovascular mortality per 10 µg/m 3 of PM
2.5
. This corresponds to a reference relative risk 𝑅𝑅
𝑅
= 1.11
for a reference concentration change Δ𝜒
𝑅
= 10 . Since
𝑅𝑅
𝑅
=
𝐼
𝐵
𝐼
𝑅 where I
R
is a reference incidence rate, this gives – for a log-linear CRF – the result
𝑅𝑅
𝑅
= exp(𝛽𝜒
𝐵
) exp(𝛽𝜒
𝑅
)
= exp(𝛽(𝜒
𝐵
− 𝜒
𝑅
)) = exp(𝛽Δ𝜒
𝑅
)
∴ 𝛽 = ln(𝑅𝑅
𝑅
)
Δ𝜒
𝑅
Page 130 of 159
This allows the overall calculation of the change in incidence rate to be condensed to
Δ𝐼 = 𝐼
𝐵
(𝑅𝑅
𝑅
(
Δ𝜒
𝛥𝜒
𝑅
)
− 1)
Since this ERF is non-linear, the change in incidence between scenarios now still requires knowledge of the baseline concentration 𝜒
𝐵
. However, if RR
R
is close to 1 and Δ𝜒 is small, the log linear CRF can be approximated as a linear CRF. Again defining the excess risk as the relative risk minus 1, i.e.
𝐸𝑅
𝑅
= 𝑅𝑅
𝑅
− 1
∴ Δ𝐼 = 𝐼
𝐵
Δ𝜒
((1 + 𝐸𝑅
𝑅
) Δ𝜒
𝑅
− 1)
By the binomial theorem
(1 + 𝑥) 𝛼
= 1 + 𝛼𝑥 + O(𝑥
2 ) ≃ 1 + 𝛼𝑥
Δ𝜒
∴ Δ𝐼 ≃ 𝐼
𝐵
(1 + 𝐸𝑅
𝑅
Δ𝜒
𝑅
− 1)
= 𝐼
𝐵
𝐸𝑅
𝑅
Δ𝜒
Δ𝜒
𝑅
For a relative risk of 1.05 (5%) per µ gm -3 , this approximation is accurate to within 3% for ±0.15 µ gm -3 , the peak change in PM
2.5
calculated by the Barrett et al study of aviation cruise emissions (6).
Page 131 of 159
Bibliography
1.
Isaac PCG, Linsky B, Larson GP (1953) Air Pollution and Man’s Health.
Public Health Rep
68:868–874. Available at: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2024105/ [Accessed
November 3, 2014].
2. Wilkins ET (1954) Air pollution aspects of the London fog of December 1952. Q J R Meteorol
Soc 80:267–271. Available at: http://doi.wiley.com/10.1002/qj.49708034420.
3. IPCC (2014) in Climate Change 2014: Impacts, Adaptation, and Vulernability/ Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change , eds Field CB et al. (Cambridge University Press,
Cambridge), pp 1–32.
4. Cunnold DM, Alyea FN, Prinn RG (1977) Relative effects on atmospheric ozone of latitude and altitude of supersonic flight. AIAA J 15:337–345. Available at: http://arc.aiaa.org/doi/abs/10.2514/3.7327.
5. Brasseur GP et al. (1998) European scientific assessment of the atmospheric effects of aircraft emissions. Atmos Environ 32:2329–2418. Available at: http://linkinghub.elsevier.com/retrieve/pii/S135223109700486X.
6. Barrett SRH, Britter RE, Waitz I a (2010) Global mortality attributable to aircraft cruise emissions.
Environ Sci Technol 44:7736–42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20809615.
7. Barrett SRH et al. (2011) Public Health, Climate and Economic Impacts of Desulfurizing Jet Fuel.
1–22.
Page 132 of 159
8. Koo J, Wang Q, Henze DK, Waitz I a., Barrett SRH (2013) Spatial sensitivities of human health risk to intercontinental and high-altitude pollution. Atmos Environ 71:140–147. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1352231013000502 [Accessed October 29, 2014].
9. Jacobson MZ, Wilkerson JT, Naiman AD, Lele SK (2013) The effects of aircraft on climate and pollution. Part II: 20-year impacts of exhaust from all commercial aircraft worldwide treated individually at the subgrid scale. Faraday Discuss 165:369–382.
10. Morita H, Yang S, Unger N, Kinney PL (2014) Global Health Impacts of Future Aviation
Emissions Under Alternative Control Scenarios. Environ Sci Technol 48:14659–14667.
11. Yim SHL et al. (2015) Global, regional and local health impacts of civil aviation emissions.
Environ Res Lett .
12. Lim SS et al. (2012) A comparative risk assessment of burden of disease and injury attributable to
67 risk factors and risk factor clusters in 21 regions, 1990-2010: a systematic analysis for the
Global Burden of Disease Study 2010. Lancet 380:2224–60. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23245609 [Accessed July 10, 2014].
13. Jacob DJ, Winner D a. (2009) Effect of climate change on air quality. Atmos Environ 43:51–63.
Available at: http://linkinghub.elsevier.com/retrieve/pii/S1352231008008571 [Accessed July 16,
2014].
14. European Commission (1998) in ExternE - Externalities of Energy , ed Holland MR
(Luxembourg), p 501.
15. European Commission (2005) ExternE - Externalities of Energy: Methodology 2005 Update eds
Bickel P, Friedrich R (Luxembourg).
Page 133 of 159
16. Ostro BD, Tran H, Levy JI (2006) The health benefits of reduced tropospheric ozone in California.
J Air Waste Manag Assoc 56:1007–1021.
17. Johnson H (1972) The Effect of Supersonic Transport Planes on the Stratospheric Ozone Shield.
Bost Coll Environ Aff Law Rev 1. Available at: http://lawdigitalcommons.bc.edu/cgi/viewcontent.cgi?article=2059&context=ealr [Accessed
November 4, 2014].
18. Madronich S, Gruijl F (1994) Stratospheric Ozone Depletion Between 1979 and 1992:
Implications for Biologically Active Ultraviolet-B Radiation and Non-Melanoma Skin Cancer
Incidence. Photochem Photobiol 59:541–546. Available at: http://onlinelibrary.wiley.com/doi/10.1111/j.1751-1097.1994.tb02980.x/abstract [Accessed
November 4, 2014].
19. Caiazzo F, Ashok A, Waitz I a., Yim SHL, Barrett SRH (2013) Air pollution and early deaths in the United States. Part I: Quantifying the impact of major sectors in 2005. Atmos Environ 79:198–
208. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1352231013004548 [Accessed July
11, 2014].
20. Whitt DB, Jacobson MZ, Wilkerson JT, Naiman a. D, Lele SK (2011) Vertical mixing of commercial aviation emissions from cruise altitude to the surface. J Geophys Res 116:1–13.
Available at: http://www.agu.org/pubs/crossref/2011/2010JD015532.shtml [Accessed July 22,
2011].
21. Lin X, Trainer M, Liu SC (1988) On the Nonlinearity of the Tropospheric Ozone Production. J
Geophys Res 93:879–888.
Page 134 of 159
22. Köhler MO et al. (2008) Impact of perturbations to nitrogen oxide emissions from global aviation.
J Geophys Res 113:1–15. Available at: http://www.agu.org/pubs/crossref/2008/2007JD009140.shtml [Accessed November 23, 2011].
23. Solomon S (1999) Stratospheric ozone depletion: A review of concepts and history. Rev
Geophys :275–316. Available at: http://www.agu.org/pubs/crossref/1999.../1999RG900008.shtml
[Accessed May 31, 2013].
24. Heckendorn P et al. (2009) The impact of geoengineering aerosols on stratospheric temperature and ozone. Environ Res Lett 4:045108. Available at: http://stacks.iop.org/1748-
9326/4/i=4/a=045108?key=crossref.dade9338b483538df3be60e23373263c [Accessed February 3,
2014].
25. Tilmes S, Garcia RR, Kinnison DE, Gettelman A, Rasch PJ (2009) Impact of geoengineered aerosols on the troposphere and stratosphere. J Geophys Res 114:D12305. Available at: http://doi.wiley.com/10.1029/2008JD011420 [Accessed October 28, 2014].
26. Emmons LK, Hess PG, Lamarque J-F, Pfister GG (2012) Tagged ozone mechanism for
MOZART-4, CAM-chem and other chemical transport models. Geosci Model Dev 5:1531–1542.
Available at: http://www.geosci-model-dev.net/5/1531/2012/ [Accessed September 19, 2014].
27. Lin P, Fu Q, Solomon S, Wallace JM (2009) Temperature Trend Patterns in Southern Hemisphere
High Latitudes: Novel Indicators of Stratospheric Change. J Clim 22:6325–6341. Available at: http://journals.ametsoc.org/doi/abs/10.1175/2009JCLI2971.1 [Accessed September 4, 2014].
28. Seinfeld JH, Pandis SN (2006) Atmospheric Chemistry and Physics (Wiley). 2nd Ed.
Page 135 of 159
29. Pye HOT et al. (2009) Effect of changes in climate and emissions on future sulfate-nitrateammonium aerosol levels in the United States. J Geophys Res 114:D01205. Available at: http://doi.wiley.com/10.1029/2008JD010701 [Accessed September 26, 2014].
30. Hansen J, Ruedy R, Sato M, Reynolds R (1996) Global surface air temperature in 1995: Return to pre-Pinatubo level. Geophys Res Lett 23:1665–1668. Available at: http://onlinelibrary.wiley.com/doi/10.1029/96GL01040/full [Accessed October 29, 2014].
31. Kirtman B et al. (2014) in Climate Change 2013: The Physical Science Basis. Contribution of
Working Grou I to the Fifth Assessment Report of the Intergovernmental Panel on Climate
Change , eds Stocker TF et al. (Cambridge University Press, Cambridge), pp 953–1028. 1st Ed.
Available at: http://www.climatechange2013.org/images/report/WG1AR5_Chapter11_FINAL.pdf
[Accessed September 21, 2014].
32. Weisenstein DK et al. (1997) A two-dimensional model of sulfur species and aerosols. J Geophys
Res 102. Available at: http://www.atmos.washington.edu/~dipierro/classes/Aerosol_chemistry/Weisenstein_model_of_S
AL_97JD00901.pdf [Accessed September 21, 2012].
33. Held I, Soden B (2000) Water vapor feedback and global warming. Annu Rev Energy Environ
25:441–475. Available at: http://www.annualreviews.org/doi/abs/10.1146/annurev.energy.25.1.441 [Accessed November 4,
2014].
34. Solomon S et al. (2010) Contributions of stratospheric water vapor to decadal changes in the rate of global warming. Science 327:1219–23. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20110466 [Accessed August 28, 2014].
Page 136 of 159
35.
Kiehl JT, Trenberth KE (1997) Earth’s Annual Global Mean Energy Budget.
Bull Am Meteorol
Soc 78:197–208. Available at: http://journals.ametsoc.org/doi/abs/10.1175/1520-
0477(1997)078<0197:EAGMEB>2.0.CO;2.
36. Liu H et al. (2006) Radiative effect of clouds on tropospheric chemistry in a global threedimensional chemical transport model. J Geophys Res 111:D20303. Available at: http://doi.wiley.com/10.1029/2005JD006403 [Accessed November 5, 2014].
37. Parrella JP et al. (2012) Tropospheric bromine chemistry: implications for present and preindustrial ozone and mercury. Atmos Chem Phys 12:6723–6740. Available at: http://www.atmoschem-phys.net/12/6723/2012/ [Accessed September 9, 2013].
38. Zeng G, Morgenstern O, Braesicke P, Pyle J a. (2010) Impact of stratospheric ozone recovery on tropospheric ozone and its budget. Geophys Res Lett 37:n/a–n/a. Available at: http://doi.wiley.com/10.1029/2010GL042812 [Accessed November 4, 2014].
39. Lu J, Vecchi G a., Reichler T (2007) Expansion of the Hadley cell under global warming. Geophys
Res Lett 34:L06805. Available at: http://doi.wiley.com/10.1029/2006GL028443 [Accessed
November 5, 2014].
40. Dockery DW et al. (1993) An Association between Air Pollution and Mortality in Six U.S. Cities.
N Engl J Med 329:1753–1759. Available at: http://dx.doi.org/10.1056/NEJM199312093292401.
41. Pope III CA et al. (2002) Lung Cancer, Cardiopulmonary Mortality, and Long-term Exposure to
Fine Particulate Air Pollution. J Am Med Assoc 287:1132–1141. Available at: http://jama.jamanetwork.com/article.aspx?articleid=194704 [Accessed October 28, 2014].
Page 137 of 159
42. Hoek G et al. (2013) Long-term air pollution exposure and cardio- respiratory mortality: a review.
Environ Health 12:43. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3679821&tool=pmcentrez&rendertype
=abstract [Accessed August 27, 2014].
43. Lepeule J, Laden F, Dockery D, Schwartz J (2012) Chronic exposure to fine particles and mortality: an extended follow-up of the Harvard Six Cities study from 1974 to 2009. Environ
Health Perspect 120:965–70. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3404667&tool=pmcentrez&rendertype
=abstract [Accessed October 21, 2014].
44. Levy JI, Diez D, Dou Y, Barr CD, Dominici F (2012) Systematic Reviews and Meta- and Pooled
Analyses A Meta-Analysis and Multisite Time-Series Analysis of the Differential Toxicity of
Major Fine Particulate Matter Constituents. Am J Epidemiol .
45. Laden F, Neas LM, Dockery DW, Schwartz J (2000) Association of Fine Particulate Matter from
Different Sources with Daily Mortality in Six U.S. Cities. Environ Health Perspect 108:941–947.
Available at: http://www.ehponline.org/ambra-doi-resolver/10.1289/ehp.00108941.
46. United States Environmental Protection Agency (2010) Quantitative Health Risk Assessment for
Particulate Matter .
47. Measuring welfare loss caused by air pollution in Europe: A CGE analysis Available at: http://ac.els-cdn.com/S0301421510003083/1-s2.0-S0301421510003083-main.pdf?_tid=d783bcc4-
9b3d-11e4-bfe2-00000aacb362&acdnat=1421165244_ca5243256c34b4d36e350546c1f3c859
[Accessed January 13, 2015].
Page 138 of 159
48. Matus K et al. (2012) Health damages from air pollution in China. Glob Environ Chang 22:55–66.
Available at: http://linkinghub.elsevier.com/retrieve/pii/S0959378011001233 [Accessed
November 8, 2014].
49. Ji M, Cohan DS, Bell ML (2011) Meta-analysis of the Association between Short-Term Exposure to Ambient Ozone and Respiratory Hospital Admissions. Environ Res Lett 6:11. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3138529&tool=pmcentrez&rendertype
=abstract [Accessed November 4, 2014].
50. Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F (2004) Ozone and short-term mortality in 95 US urban communities, 1987-2000. J Am Med Assoc 292:2372–2378. Available at: http://archpsyc.jamanetwork.com/article.aspx?articleid=199808 [Accessed December 30, 2014].
51. Jerrett M et al. (2009) Long-term ozone exposure and mortality. N Engl J Med 360:1085–95.
Available at: http://www.ncbi.nlm.nih.gov/pubmed/19279340.
52. Anenberg SC, Horowitz LW, Tong DQ, West JJ (2010) An estimate of the global burden of anthropogenic ozone and fine particulate matter on premature human mortality using atmospheric modeling. Environ Health Perspect 118:1189–95. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2944076&tool=pmcentrez&rendertype
=abstract [Accessed October 10, 2014].
53. Silva R a et al. (2013) Global premature mortality due to anthropogenic outdoor air pollution and the contribution of past climate change. Environ Res Lett 8:034005. Available at: http://stacks.iop.org/1748-
9326/8/i=3/a=034005?key=crossref.4619a48f65a5bc77307c00cf38f9299f [Accessed September 9,
2014].
Page 139 of 159
54. Lelieveld J, Barlas C, Giannadaki D, Pozzer a. (2013) Model calculated global, regional and megacity premature mortality due to air pollution. Atmos Chem Phys 13:7023–7037. Available at: http://www.atmos-chem-phys.net/13/7023/2013/ [Accessed September 27, 2014].
55. Selin NE et al. (2009) Global health and economic impacts of future ozone pollution. Environ Res
Lett 4:044014. Available at: http://stacks.iop.org/1748-
9326/4/i=4/a=044014?key=crossref.0d2025b77173b9122ac34105d3b8bcb0 [Accessed January 7,
2015].
56. De Gruijl FR, Van der Leun JC (1994) Estimate of the wavelength dependency of ultraviolet carcinogenesis in humans and its relevance to the risk assessment of a stratospheric ozone depletion. Health Phys 67:319–325.
57. Webb AR, Kline L, Holick MF (1988) Influence of Season and Latitude on the Cutaneous
Synthesis of Vitamin D3: Exposure to Winter Sunlight in Boston and Edmonton Will Not Promote
Vitamin D3 Synthesis in Human Skin. J Clin Endocrinol Metab 67:373–378.
58. Lindqvist PG et al. (2014) Avoidance of sun exposure is a risk factor for all-cause mortality: results from the Melanoma in Southern Sweden cohort. J Intern Med 276:77–86. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24697969 [Accessed July 12, 2014].
59. Lucas R, McMichael T, Smith W, Armstrong B (2006) in Environmental Burden of Disease , eds
Prüss-üstün A, Zeeb H, Mathers C, Repacholi M (World Health Organization, Geneva).
60. Slaper H, Velders G, Daniel J (1996) Estimates of ozone depletion and skin cancer incidence to examine the Vienna Convention achievements. Nature . Available at:
Page 140 of 159
http://www.nature.com/nature/journal/v384/n6606/abs/384256a0.html [Accessed October 28,
2013].
61. Eisemann N et al. (2014) Non-melanoma skin cancer incidence and impact of skin cancer screening on incidence. J Invest Dermatol 134:43–50. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23877569.
62. Lomas a., Leonardi-Bee J, Bath-Hextall F (2012) A systematic review of worldwide incidence of nonmelanoma skin cancer. Br J Dermatol 166:1069–1080.
63. De Vries E, van de Poll-Franse L V., Louwman WJ, de Gruijl FR, Coeburgh JWW (2005)
Predictions of skin cancer incidence in the Netherlands up to 2015. Br J Dermatol 152:481–488.
Available at: http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2133.2005.06386.x/full [Accessed
November 2, 2014].
64. Lucas R, McMichael T, Smith W, Armstrong B (2006) in Environmental Burden of Disease , eds
Prüss-üstün A, Zeeb H, Mathers C, Repacholi M (World Health Organization, Geneva).
65. Bey I et al. (2001) Global modeling of tropospheric chemistry with assimilated meteorology:
Model description and evaluation. J Geophys Res 106:23073. Available at: http://doi.wiley.com/10.1029/2001JD000807.
66. McLinden CA et al. (2000) Stratospheric ozone in 3-D models: A simple chemistry and the crosstropopause flux. J Geophys Res - Atmos 105:14653–14665. Available at: http://www.ess.uci.edu/~prather/publications/2000JGR_McLinden-Linoz.pdf [Accessed June 4,
2012].
Page 141 of 159
67. Murray LT, Jacob DJ, Logan J a., Hudman RC, Koshak WJ (2012) Optimized regional and interannual variability of lightning in a global chemical transport model constrained by LIS/OTD satellite data. J Geophys Res 117. Available at: http://doi.wiley.com/10.1029/2012JD017934
[Accessed December 19, 2013].
68. Rotman DA et al. (2001) Global Modeling Initiative assessment model : Model description , integration , and testing of the transport shell. J Geophys Res 106:1669–1691.
69. Barkley MP et al. (2011) Can a “state of the art” chemistry transport model simulate Amazonian tropospheric chemistry? J Geophys Res 116:D16302. Available at: http://doi.wiley.com/10.1029/2011JD015893 [Accessed August 25, 2013].
70. Mao J et al. (2010) Chemistry of hydrogen oxide radicals (HO x
) in the Arctic troposphere in spring. Atmos Chem Phys 10:5823–5838. Available at: http://www.atmos-chemphys.net/10/5823/2010/ [Accessed August 14, 2013].
71. Duncan Fairlie T, Jacob DJ, Park RJ (2007) The impact of transpacific transport of mineral dust in the United States. Atmos Environ 41:1251–1266. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1352231006009915 [Accessed December 19, 2013].
72. Corbett JJ et al. (2007) Mortality from Ship Emissions: A Global Assessment. Environ Sci
Technol 41:8512–8518. Available at: http://pubs.acs.org/doi/abs/10.1021/es071686z.
73. Chipperfield M (2006) The TOMCAT/SLIMCAT Off-Line 3D CTM User’s Manual. Available at: http://homepages.see.leeds.ac.uk/~lecmc/papers/uniman0.8.pdf [Accessed June 4, 2012].
Page 142 of 159
74. Emmons LK et al. (2010) Description and evaluation of the Model for OZone and Related chemical Tracers, version 4 (MOZART-4). Geosci Model Dev 3:43–67. Available at: http://www.geosci-model-dev.net/3/43/2010/ [Accessed June 3, 2012].
75. Lamarque J-F et al. (2012) CAM-chem: description and evaluation of interactive atmospheric chemistry in the Community Earth System Model. Geosci Model Dev 5:369–411. Available at: http://www.geosci-model-dev.net/5/369/2012/ [Accessed August 22, 2013].
76. SPARC Group (2006) SPARC Report No. 4: Assessment of Stratospheric Aerosol Properties .
77. De Grandpré J et al. (2000) Ozone climatology using interactive chemistry: Results from the
Canadian Middle Atmosphere Model. J Geophys … 105:26475–26491. Available at: http://www.agu.org/pubs/crossref/2000/2000JD900427.shtml [Accessed December 20, 2013].
78. Kinnison DE et al. (2007) Sensitivity of chemical tracers to meteorological parameters in the
MOZART-3 chemical transport model. J Geophys Res 112:1–
24file:///home/seastham/Dropbox/Thesis/Reference. Available at: http://www.agu.org/pubs/crossref/2007/2006JD007879.shtml [Accessed March 18, 2012].
79. Huijnen V et al. (2010) The global chemistry transport model TM5: description and evaluation of the tropospheric chemistry version 3.0. Geosci Model Dev 3:445–473. Available at: http://www.geosci-model-dev.net/3/445/2010/ [Accessed March 30, 2012].
80. Søvde O a. et al. (2012) The chemical transport model Oslo CTM3. Geosci Model Dev 5:1441–
1469. Available at: http://www.geosci-model-dev.net/5/1441/2012/ [Accessed August 22, 2013].
Page 143 of 159
81. Morgenstern O et al. (2010) Review of the formulation of present-generation stratospheric chemistry-climate models and associated external forcings. J Geophys Res 115:D00M02.
Available at: http://doi.wiley.com/10.1029/2009JD013728 [Accessed November 4, 2014].
82. Rotman D, Tannahill J (2001) Global Modeling Initiative assessment model- Model description, integration, and testing of the transport shell. J Geophys … 106:1669–1691. Available at: http://www.agu.org/journals/jd/jd0102/2000JD900463/pdf/2000JD900463.pdf [Accessed
December 12, 2012].
83. Considine DB, Douglass a. R, Connell PS, Kinnison DE, Rotman D a. (2000) A polar stratospheric cloud parameterization for the global modeling initiative three-dimensional model and its response to stratospheric aircraft. J Geophys Res 105:3955–3973. Available at: http://www.agu.org/pubs/crossref/2000/1999JD900932.shtml.
84. Arnold SR, Chipperfield MP, Blitz M a. (2005) A three-dimensional model study of the effect of new temperature-dependent quantum yields for acetone photolysis. J Geophys Res 110:D22305.
Available at: http://doi.wiley.com/10.1029/2005JD005998 [Accessed August 22, 2013].
85. Krewski D et al. (2009) Extended follow-up and spatial analysis of the American Cancer Society study linking particulate air pollution and mortality. HEI Research Report 140.
(Health Effects
Institudte, Boston, MA) Available at: http://scientificintegrityinstitute.net/Krewski052108.pdf
[Accessed October 28, 2014].
86. Burnett RT et al. (2014) An integrated risk function for estimating the global burden of disease attributable to ambient fine particulate matter exposure. Environ Health Perspect 122:397–403.
Available at:
Page 144 of 159
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3984213&tool=pmcentrez&rendertype
=abstract.
87. Fang Y, Naik V, Horowitz LW, Mauzerall DL (2013) Air pollution and associated human mortality: the role of air pollutant emissions, climate change and methane concentration increases from the preindustrial period to present. Atmos Chem Phys 13:1377–1394. Available at: http://www.atmos-chem-phys.net/13/1377/2013/ [Accessed October 30, 2014].
88. Chaplin G (2004) Geographic distribution of environmental factors influencing human skin coloration. Am J Phys Anthropol 125:292–302. Available at: http://onlinelibrary.wiley.com/doi/10.1002/ajpa.10263/full [Accessed June 11, 2014].
89. Van Dijk A et al. (2013) Skin cancer risks avoided by the Montreal Protocol--worldwide modeling integrating coupled climate-chemistry models with a risk model for UV. Photochem Photobiol
89:234–46. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22924540 [Accessed June 9,
2014].
90. Lee DS et al. (2009) Aviation and global climate change in the 21st century. Atmos Environ
43:3520–3537. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1352231009003574
[Accessed October 20, 2014].
91. Ratliff G et al. (2009) Aircraft Impacts on Local and Regional Air Quality in the United States
(Cambridge, MA) Available at: http://web.mit.edu/aeroastro/partner/reports/proj15/proj15finalreport.pdf.
Page 145 of 159
92. Lee H, Olsen SC, Wuebbles DJ, Youn D (2013) Impacts of aircraft emissions on the air quality near the ground. Atmos Chem Phys Discuss 13:689–727. Available at: http://www.atmos-chemphys-discuss.net/13/689/2013/ [Accessed January 18, 2013].
93. Brunelle-Yeung E (2009) The Impacts of Aviation Emissions on Human Health through Changes in Air Quality and UV Irradiance. Dissertation (Massachusetts Institute of Technology).
94. Prinn RG, Reilly JM (2014) 2014 Energy and Climate Outlook (Cambridge, MA).
95. Smith KR et al. (2014) in Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A:
Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change , eds Field CB et al. (Cambridge University Press,
Cambridge, United Kingdom), pp 709–754.
96. Keith D (2000) Geoengineering the climate: History and Prospect. Annu Rev Energy Environ
25:245–284. Available at: http://www.annualreviews.org/doi/abs/10.1146/annurev.energy.25.1.245 [Accessed July 21,
2014].
97. McCormick M, Thomason L, Trepte C (1995) Atmospheric effects of the Mt Pinatubo eruption.
Nature 373:399–404. Available at: http://www.nuclear.lu.se/fileadmin/nuclear/Undervisning/Atmosfaerskurs/P04.pdf [Accessed
October 29, 2014].
98. Crutzen PJ (2006) Albedo Enhancement by Stratospheric Sulfur Injections: A Contribution to
Resolve a Policy Dilemma? Clim Change 77:211–220. Available at: http://www.springerlink.com/index/10.1007/s10584-006-9101-y [Accessed July 8, 2011].
Page 146 of 159
99. Tilmes S, Müller R, Salawitch R (2008) The sensitivity of polar ozone depletion to proposed geoengineering schemes. Science (80- ) 320:1201–4. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18436741 [Accessed October 24, 2014].
100. Robock A, Oman L, Stenchikov GL (2008) Regional climate responses to geoengineering with tropical and Arctic SO 2 injections. J Geophys Res 113:1–15. Available at: http://www.agu.org/pubs/crossref/2008/2008JD010050.shtml [Accessed June 14, 2011].
101. Rasch PJ, Crutzen PJ, Coleman DB (2008) Exploring the geoengineering of climate using stratospheric sulfate aerosols: The role of particle size. Geophys Res Lett 35:1–6. Available at: http://www.agu.org/pubs/crossref/2008/2007GL032179.shtml [Accessed August 19, 2011].
102. Kravitz B et al. (2013) An overview of the Geoengineering Model Intercomparison Project
(GeoMIP). J Geophys Res Atmos 118:13,103–113,107. Available at: http://dx.doi.org/10.1002/2013JD020569.
103. Pitari G, Aquila V, Kravitz B (2014) Stratospheric ozone response to sulfate geoengineering:
Results from the Geoengineering Model Intercomparison Project (GeoMIP). J Geophys Res -
Atmos :2629–2653. Available at: http://onlinelibrary.wiley.com/doi/10.1002/2013JD020566/full
[Accessed October 20, 2014].
104. Gilmore CK, Barrett SRH, Koo J, Wang Q (2013) Temporal and spatial variability in the aviation
NOx-related O3 impact. Environ Res Lett 8:034027. Available at: http://stacks.iop.org/1748-
9326/8/i=3/a=034027?key=crossref.58500264c399c496c2acaf3c3a1105bd.
Page 147 of 159
105. Leibensperger EM, Mickley LJ, Jacob DJ, Barrett SRH (2011) Intercontinental influence of NOx and CO emissions on particulate matter air quality. Atmos Environ 45:3318–3324. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1352231011001579 [Accessed November 4, 2014].
106. Vennam LP et al. (2014) Air Pollution Modeling and its Application XXIII. Available at: http://link.springer.com/10.1007/978-3-319-04379-1 [Accessed October 6, 2014].
107. Bright EA, Rose AN, Urban ML (2012) LandScan 2012. Available at: http://www.ornl.gov/landscan/.
108. Center for International Earth Science Information Network - CIESIN - Columbia University,
International Food Policy Research Institute IFPRI, The World Bank, Centro Internacional de
Agricultura Tropical - CIAT (2011) Global Rural-Urban Mapping Project, Version 1
(GRUMPv1): National Identifier Grid. Available at: http://dx.doi.org/10.7927/H40K26HS.
109. United Nations Department of Economic and Social Affairs (Population Division) (2013) World
Population Prospects: The 2012 revision. Available at: http://esa.un.org/unpd/wpp/ASCII-
Data/DISK_NAVIGATION_ASCII.htm.
110. Ostro B, Prüss-üstün A, Campbell-lendrum D, Corvalán C, Woodward A (2004) Outdoor air pollution: Assessing the environmental burden of disease at national and local levels (Geneva,
Switzerland).
111. Ji M, Cohan DS, Bell ML (2011) Meta-analysis of the association between short-term exposure to ambient ozone and respiratory hospital admissions. Environ Res Lett 6. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3138529&tool=pmcentrez&rendertype
=abstract [Accessed December 30, 2014].
Page 148 of 159
112. Levy JI, Chemerynski SM, Sarnat J a (2005) Ozone exposure and mortality: an empiric bayes metaregression analysis. Epidemiology 16:458–468.
113. Murray CJL et al. (2012) Disability-adjusted life years (DALYs) for 291 diseases and injuries in
21 regions, 1990-2010: A systematic analysis for the Global Burden of Disease Study 2010.
Lancet 380:2197–2223.
114. World Health Organization (2014) Disease and injury country mortality estimates for 2012.
Available at: http://www.who.int/healthinfo/global_burden_disease/estimates/en/index1.html.
115. United States Environmental Protection Agency (2006) Expanded Expert Judgment Assessment of the Concentration-Response Relationship Between PM2.5 Exposure and Mortality .
116. Pope III CA et al. (2011) Lung Cancer and Cardiovascular Disease Mortality Associated with
Ambient Air Pollution and Cigarette Smoke : Shape of the Exposure – Response Relationships.
1616:1616–1621.
117. Park RJ (2003) Sources of carbonaceous aerosols over the United States and implications for natural visibility. J Geophys Res 108.
118. Roemer W et al. (1998) Daily variations in air pollution and respiratory health in a multicentre study : the PEACE project. 1354–1361.
119. Guy GP, Machlin SR, Ekwueme DU, Yabroff KR (2015) Prevalence and costs of skin cancer treatment in the u.s., 2002-2006 and 2007-2011. Am J Prev Med 48:183–7. Available at: http://www.sciencedirect.com/science/article/pii/S0749379714005108 [Accessed February 2,
2015].
Page 149 of 159
120. Sander SP et al. (2011) Chemical Kinetics and Photochemical Data for Use in Atmospheric
Studies Evaluation Number 17 NASA Panel for Data Evaluation :
121. Wild O, Zhu X, Prather M (2000) Fast-J: Accurate simulation of in- and below-cloud photolysis in tropospheric chemical models. J Atmos Chem :245–282. Available at: http://link.springer.com/article/10.1023/A:1006415919030 [Accessed December 19, 2013].
122. Prather MJ (2012) Fast-JX v7.0a. Available at: http://www.ess.uci.edu/researchgrp/prather/scholar_software/fast-jx.
123. Mills MJ (2005) Photolysis of sulfuric acid vapor by visible light as a source of the polar stratospheric CN layer. J Geophys Res 110:D08201. Available at: http://www.agu.org/pubs/crossref/2005/2004JD005519.shtml [Accessed March 15, 2013].
124. Kirner O et al. (2011) Simulation of polar stratospheric clouds in the chemistry-climate-model
EMAC via the submodel PSC. Geosci Model Dev 4:169–182. Available at: http://www.geoscimodel-dev.net/4/169/2011/ [Accessed August 31, 2012].
125. Shi Q, Jayne J, Kolb C, Worsnop D, Davidovits P (2001) Kinetic model for reaction of ClONO2 with H2O and HCl and HOCl with HCl in sulfuric acid solutions. J Geophys Res 106:24259–
24274. Available at: http://www.agu.org/pubs/crossref/2001/2000JD000181.shtml [Accessed May
7, 2013].
126. Carslaw K, Peter T, Clegg S (1997) Modeling the composition of liquid stratospheric aerosols. Rev
Geophys 35:125–154. Available at: http://www.agu.org/pubs/crossref/1997/97RG00078.shtml
[Accessed September 12, 2012].
Page 150 of 159
127. Kulmala M, Laaksonen A (1990) Binary nucleation of water–sulfuric acid system: Comparison of classical theories with different H2SO4 saturation vapor pressures. J Chem Phys 93:696. Available at: http://link.aip.org/link/JCPSA6/v93/i1/p696/s1&Agg=doi [Accessed February 14, 2013].
128. Grainger RG, Lambert A, Rodgers CD, Taylor FW (1995) Stratospheric aerosol effective radius, surface area and volume estimated from infrared measurements. J Geophys Res 100:16507–16518.
Available at: http://www.agu.org/pubs/crossref/1995/95JD00988.shtml [Accessed July 2, 2013].
129. Hanson D, Mauersberger K (1988) Laboratory studies of the nitric acid trihydrate: Implications for the south polar stratosphere. Geophys Res Lett 15:855–858. Available at: http://europa.agu.org/?uri=/journals/gl/GL015i008p00855.xml&view=article [Accessed
September 26, 2012].
130. Hanson DR (2003) Reactivity of BrONO 2 and HOBr on sulfuric acid solutions at low temperatures. J Geophys Res 108:4239. Available at: http://doi.wiley.com/10.1029/2002JD002519
[Accessed August 26, 2013].
131. Tabazadeh a., Djikaev YS, Hamill P, Reiss H (2002) Laboratory evidence for surface nucleation of solid polar stratospheric cloud particles. J Phys Chem A 106:10238–10246.
132. Salcedo D, Molina LT, Molina MJ (2001) Homogeneous freezing of concentrated aqueous nitric acid solutions at polar stratospheric temperatures. J Phys Chem A 105:1433–1439. Available at:
<Go to ISI>://000167268600005.
133. Buchholz J (2005) Simulations of physics and chemistry of polar stratospheric clouds with a general circulation model. Dissertation (Gutenberg). Available at: http://d-nb.info/976096129/34
[Accessed September 26, 2012].
Page 151 of 159
134. Tabazadeh A, Toon OB, Jensen EJ (1997) Formation and implications of ice particle nucleation in the stratosphere. Geophys Res Lett 24:2007–2010.
135. Carslaw KS et al. (1998) Particle microphysics and chemistry in remotely observed mountain polar stratospheric clouds. J Geophys Res 103:5785.
136. Daerden F et al. (2006) A 3D-CTM with detailed online PSC-microphysics: analysis of the
Antarctic winter 2003 by comparison with satellite observations. Atmos Chem Phys Discuss
6:8511–8552.
137. Daniel JS et al. (2007) in Scientific Assessment of Ozone Depletion: 2006 (World Meteorological
Organization, Geneva, Switzerland), p 572.
138. European Commission JRC (JRC)/Netherlands EAA (PBL) (2009) Emissions Database for Global
Atmospheric Research (EDGAR), release version 4.0. Available at: http://edgar.jrc.ec.europa.eu/.
139. Yevich R, Logan J a. (2003) An assessment of biofuel use and burning of agricultural waste in the developing world. Global Biogeochem Cycles 17. Available at: http://doi.wiley.com/10.1029/2002GB001952 [Accessed December 18, 2013].
140. Van der Werf GR et al. (2010) Global fire emissions and the contribution of deforestation, savanna, forest, agricultural, and peat fires (1997–2009). Atmos Chem Phys 10:11707–11735.
Available at: http://www.atmos-chem-phys.net/10/11707/2010/ [Accessed December 12, 2013].
141. Pickett-Heaps C a. et al. (2011) Magnitude and seasonality of wetland methane emissions from the
Hudson Bay Lowlands (Canada). Atmos Chem Phys 11:3773–3779. Available at: http://www.atmos-chem-phys.net/11/3773/2011/ [Accessed December 18, 2013].
Page 152 of 159
142. Fung I, John J, Lerner J (1991) Three-dimensional model synthesis of the global methane cycle. J
Geophys Res 96. Available at: http://onlinelibrary.wiley.com/doi/10.1029/91JD01247/full
[Accessed December 18, 2013].
143. Jackman CH, Frederick JE, Stolarski RS (1980) Production of Odd Nitrogen in the Stratosphere and Mesosphere : An Intercomparison of Source Strengths. J Geophys Res 85.
144. Lawrence MG, Jöckel P, Kuhlmann R von (2001) What does the global mean OH concentration tell us? Atmos Chem Phys :37–49. Available at: http://www.atmos-chem-phys.net/1/37/2001/acp-1-
37-2001.pdf [Accessed September 2, 2013].
145. Spivakovsky CM et al. (2000) Three-dimensional climatological distribution of tropospheric OH:
Update and evaluation. J Geophys Res 105:8931. Available at: http://doi.wiley.com/10.1029/1999JD901006.
146. Naik V et al. (2013) Preindustrial to present-day changes in tropospheric hydroxyl radical and methane lifetime from the Atmospheric Chemistry and Climate Model Intercomparison Project
(ACCMIP). Atmos Chem Phys 13:5277–5298. Available at: http://www.atmos-chemphys.net/13/5277/2013/ [Accessed December 24, 2013].
147. Prinn RG et al. (2001) Evidence for substantial variations of atmospheric hydroxyl radicals in the past two decades. Science 292:1882–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11337586 [Accessed January 2, 2014].
148. Grooß J-U, Brautzsch K, Pommrich R, Solomon S, Müller R (2011) Stratospheric ozone chemistry in the Antarctic: what determines the lowest ozone values reached and their recovery? Atmos
Page 153 of 159
Chem Phys 11:12217–12226. Available at: http://www.atmos-chem-phys.net/11/12217/2011/
[Accessed May 28, 2013].
149. Gobbi GP, Donfrancesco G Di, Adriani A (1998) Physical properties of stratospheric clouds during the Antarctic winter of 1995. J Geophys Res 103:10859–10873.
150. Wegner T et al. (2012) Heterogeneous chlorine activation on stratospheric aerosols and clouds in the Arctic polar vortex. Atmos Chem Phys 12:11095–11106. Available at: http://www.atmoschem-phys.net/12/11095/2012/ [Accessed July 8, 2013].
151. Peter T, Grooß J (2012) in Stratospheric Ozone Depletion and Climate Change (Royal Society of
Chemistry), pp 108–144.
152. Brown AT, Volk CM, Schoeberl MR, Boone CD, Bernath PF (2013) Stratospheric lifetimes of
CFC-12, CCl4, CH4, CH3Cl and N2O from measurements made by the Atmospheric Chemistry
Experiment-Fourier Transform Spectrometer (ACE-FTS). Atmos Chem Phys 13:6921–6950.
Available at: http://www.atmos-chem-phys.net/13/6921/2013/ [Accessed September 10, 2013].
153. Douglass a. R et al. (2008) Relationship of loss, mean age of air and the distribution of CFCs to stratospheric circulation and implications for atmospheric lifetimes. J Geophys Res 113:D14309.
Available at: http://doi.wiley.com/10.1029/2007JD009575 [Accessed September 18, 2013].
154. Montzka SA et al. (2011) in Scientific Assessment of Ozone Depletion: 2010 (World
Meteorological Organization, Geneva, Switzerland). Available at: http://oceanrep.geomar.de/10405/2/03-Chapter_1.pdf [Accessed September 23, 2013].
Page 154 of 159
155. Volk C et al. (1997) Evaluation of source gas lifetimes from stratospheric observations. J Geophys
Res 102:25543–25564. Available at: http://onlinelibrary.wiley.com/doi/10.1029/97JD02215/full
[Accessed September 17, 2013].
156. Popp PJ et al. (2009) Stratospheric correlation between nitric acid and ozone. J Geophys Res
114:D03305. Available at: http://doi.wiley.com/10.1029/2008JD010875 [Accessed September 10,
2013].
157. Sankey D, Shepherd TG (2003) Correlations of long-lived chemical species in a middle atmosphere general circulation model. J Geophys Res 108:4494. Available at: http://doi.wiley.com/10.1029/2002JD002799 [Accessed September 10, 2013].
158. Michelsen HA (1998) Correlations of stratospheric abundances of NOy , 03 , NO , and CH 4 derived from ATMOS measurements. 103.
159. Murphy DM et al. (1993) Reactive nitrogen and its correlation with ozone in the lower stratosphere and upper troposphere. J Geophys Res 98:8751. Available at: http://doi.wiley.com/10.1029/92JD00681.
160. Fahey D et al. (1996) In situ observations of NOy, O3, and the NOy/O3 ratio in the lower stratosphere. Geophys Res Lett 23:1653–1656. Available at: http://www.agu.org/pubs/crossref/1996/96GL01476.shtml [Accessed September 15, 2013].
161. Pierce JR, Weisenstein DK, Heckendorn P, Peter T, Keith DW (2010) Efficient formation of stratospheric aerosol for climate engineering by emission of condensible vapor from aircraft.
Geophys Res Lett 37:n/a–n/a. Available at: http://doi.wiley.com/10.1029/2010GL043975
[Accessed February 4, 2014].
Page 155 of 159
162. Mishchenko M, Dlugach J, Yanovitskij E, Zakharova N (1999) Bidirectional reflectance of flat, optically thick particulate layers: an efficient radiative transfer solution and applications to snow and soil surfaces. J Quant Spectrosc Radiat Transf 63:409–432. Available at: http://www.sciencedirect.com/science/article/pii/S002240739900028X [Accessed July 7, 2014].
163. Chen J-P, Tsai I-C, Lin Y-C (2013) A statistical–numerical aerosol parameterization scheme.
Atmos Chem Phys 13:10483–10504. Available at: http://www.atmos-chemphys.net/13/10483/2013/ [Accessed July 28, 2014].
164. Eastham SD, Weisenstein DK, Barrett SRH (2014) Development and evaluation of the unified tropospheric-stratospheric chemistry extension (UCX) for the global chemistry-transport model
GEOS-Chem. Atmos Environ . Available at: http://linkinghub.elsevier.com/retrieve/pii/S1352231014000971 [Accessed February 10, 2014].
165. Clarke L et al. (2007) Scenarios of Greenhouse Gas Emissions and Atmospheric Concentrations.
Sub-report 2.1A of Synthesis and Assessment Product 2.1 by the U.S. Climate Change Science
Program and the Subcommittee on Global Change Research (Washington).
166. Smith SJ, Wigley TML (2006) Multi-Gas Forcing Stabilization with MiniCAM. Energy J :373–
391.
167. Wise M et al. (2009) Implications of limiting CO2 concentrations for land use and energy. Science
324:1183–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19478180 [Accessed September
6, 2014].
168. Giorgi F, Francisco R (2000) Uncertainties in regional climate change prediction: a regional analysis of ensemble simulations with the HADCM2 coupled AOGCM. Clim Dyn 16:169–182.
Page 156 of 159
169. Kravitz B et al. (2014) A multi-model assessment of regional climate disparities caused by solar geoengineering. Environ Res Lett 9. Available at: http://stacks.iop.org/1748-
9326/9/i=7/a=074013?key=crossref.53f35661906a36c1a3eda6f77c188f3d.
170. Rasch PJ et al. (2008) An overview of geoengineering of climate using stratospheric sulphate aerosols. Philos Trans A Math Phys Eng Sci 366:4007–37. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18757276 [Accessed June 14, 2011].
171. MacMartin DG, Caldeira K, Keith DW (2014) Solar geoengineering to limit the rate of temperature change. Philos Trans A Math Phys Eng Sci 372. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25404687.
172. United States Environmental Protection Agency (2011) The Benefits and Costs of the Clean Air
Act from 1990 to 2020 .
173. Cohen AJ et al. (2004) Urban Air Pollution. Comp Quantif Heal Risks 77:1353–1434.
174. Tilmes S et al. (2012) Impact of very short-lived halogens on stratospheric ozone abundance and
UV radiation in a geo-engineered atmosphere. Atmos Chem Phys 12:10945–10955.
175. Niemeier U, Schmidt H, Timmreck C (2011) The dependency of geoengineered sulfate aerosol on the emission strategy. Atmos Sci Lett 12:189–194. Available at: http://doi.wiley.com/10.1002/asl.304 [Accessed August 21, 2011].
176. Olivier JGJ et al. (2002) Applications of EDGAR. Including a description of EDGAR 3.0: reference database with trend data for 1970-1995.
Page 157 of 159
177. Guenther a. B et al. (2012) The model of emissions of gases and aerosols from nature version 2.1
(MEGAN2.1): An extended and updated framework for modeling biogenic emissions. Geosci
Model Dev 5:1471–1492.
178. Barrett SRH et al. (2010) Guidance on the use of AEDT Gridded Aircraft Emissions in
Atmospheric Models (Cambridge, Massachusetts) Available at: http://lae.mit.edu/uploads/LAE_report_series/2010/LAE-2010-008-N.pdf.
179. Tarrason L, Jonson JE, Berntsen TK, Rypdal K (2002) Study on air quality impacts of non-LTO emissions from aviation (Oslo).
180. Ansari AS, Pandis SN (1998) Response of inorganic PM to precursor concentrations. Environ Sci
Technol 32:2706–2714.
181. Olsen SC et al. (2013) Comparison of model estimates of the effects of aviation emissions on atmospheric ozone and methane. Geophys Res Lett 40:6004–6009.
182. Li J et al. (2013) in AIAA Guidance, Navigation, and Control (GNC) Conference , Guidance,
Navigation, and Control and Co-located Conferences. (American Institute of Aeronautics and
Astronautics). Available at: http://dx.doi.org/10.2514/6.2013-5178.
183. Burkhardt U, Kärcher B (2011) Global radiative forcing from contrail cirrus. Nat Clim Chang
1:54–58. Available at: http://www.nature.com/doifinder/10.1038/nclimate1068 [Accessed October
15, 2014].
184. Lelieveld J et al. (1999) Chlorine activation and ozone destruction in the northern lowermost stratosphere. J Geophys Res 104:8201–8213.
Page 158 of 159
185. Fiore AM et al. (2012) Global air quality and climate. Chem Soc Rev 41:6663.
186. Ostro BD (1987) Air pollution and morbidity revisited: A specification test. J Environ Econ
Manage 14:87–98.
Page 159 of 159
