United States. Congress. House. Committee on Scien.

Road from Kyoto : hearing before the Committee on Science, U.S. House of Representatives, One Hundred Fifth Congress, second session (Volume pt. 2) online

. (page 107 of 137)
Online LibraryUnited States. Congress. House. Committee on ScienRoad from Kyoto : hearing before the Committee on Science, U.S. House of Representatives, One Hundred Fifth Congress, second session (Volume pt. 2) → online text (page 107 of 137)
Font size
QR-code for this ebook


39*


61U



Al=<Jevelooe<l (annei 1) countnes: NAl=deveiODing (non-anne» 11 countries; BAU=Dusiness-as-usual scenano: CP=climatfrDOlicv scenario
Table 1: Fossll-hiel energy use (10**J) under buslness-a»4isual and climate-policy scenarios



markedly differ from other annex 1 countries, in that
sharp declines in energ>' consumption are not assumed to
reach 1990 levels again before 2020. As a consequence,
future increases in emissions in annex 1 countries under
the BAU scenano are masked by these reduced
consumption patterns in Russia and the countries of the
former Soviet Union.

The ex-pected growth in energ>-related CO. emissions
of non-annex 1 countries under BAU by 2020 exceeds
the total level of emissions of annex 1 countries m 1990.
Non-annex 1 countries' carbon emissions resulting from
consumption of fossil fuels are expected to rise from 1 -9
BMTC in 1990 to 41 BMTC in 2010, a 116% increase
over 30 years. In 1990, annex 1 countries produced 68%
and non-annex 1 countries produced only 32% of total
global carbon emissions. Under the BAU scenario, by
the years 2010 and 2020, non-annex 1 region's projected
contributions «ill be 41 and 6 BMTC, respectively. On
the other hand, if the climate-polic>- scenario is
implemented, by the years 2010 and 2020, non-annex 1
region's projected carbon emissions could be reduced to
3 7 and 5-4 BMTC, respectively. Figure 2 shows the
relative contributions of developed and developing
countries to fossil-fuel use for the years 1990, 2010, and
2020 under both scenarios. The global fossil-fuel
consumption is expected to be less imder the climate
polic>' scenario, which reflects increased fuel-combustion
efficiencies, conservanon measures, and improved
technologies. We assume that there is no cross-border
trading of greenhouse gases. Under the climate-polic>'
scenario, CO, emissions decrease more than fossil-fiiel
consumption as countries switch away from coal to less
polluting fuels, such as natural gas.

Pollutant emissions and concentrations
The followmg discussion outlines the modelling protocol
followed in this study along with a brief description of key
assumptions and sensitivity analyses.

A PM source-receptor matrix'' was used to derive
concentration panems for PM,„ and PM,, per unit of
energy use for each of the four sectors and three fossil-^
fuel types (coal, oil, and natural gas). The source-receptor
coefficients were calculated with large-scale air-dispersion
models that were based on emission inventories calibrated
with and matched to extensive monitoring data in USA
for 1990-1994." In addition to calculation of
concentrations of primar>' particles (those emined directly
by sources), the model also incorporates the secondary
conversion of gaseous precursors (SO, and NO,) to fine
particulates as they are transported through the lower
atmosphere. The source-receptor-matrix model yields



smoothed, area-wide averages of PM,„ and PMj,
concentrations for each IPCC region and does not
esumate local airborne-pollutant levels at specific receptor
sites; the laner air-poUutant estimates are generally
calculated with Gaussian air-dispersion models that
incorporate only local emission inventories and
meteorological data. To do this analysis, we assume that
airborne fine particles travel hundreds of miles through
the atmosphere and their estimated regional
concencrauons are not ver>' sensitive to the details of the
modelling matrix. Tliis contrasts with tlie well-knowTi
difficulty of modelling for shon-term ozone episodes over
a small geographic area. Evidence supporting the long-
range transpon of fine particles can also be found — eg, m



a



Natural gas

Oil

Coal




350-1

300-

250-

200

150-

100-

5 50-

o

y

J 350-
g 300-

250-

200-

150-

100-

50-

0-'

1990 Business Climate Business Climate
-as-usual -policy -as-usual -policy
scenario scenario scenario scenario
2010 2010 2020 2020

Year
Figure 2: Fossil-fuel use under buslness-as-usual and
climateimlicy scenarios In developed (annex 1; A) and
developing (non-annex 1; B) countries




Vol 350 • November 8, 1997



1343



846



THE LANCET



the repon of the second World Meteorological
Organization meeting of experts to assess the response to
and atmospheric effects of the Kuwait oil fires."

Applying such a large-scale air-dispersion model carries
with it several assumptions that could result in a
misestimation of actual PM concentrations. In the
absence of data on typical chimney heights in other parts
of the world, for example, we have assumed that the
distribution of chimney heights in the USA is
representative of that found elsewhere. Should average
heights of chimneys in a given region be lower than in the
USA, the likely consequence would be higher local
concentrations of particles and their precursors, and
lower concentrations at greater distances downwind.
Conversely, higher chimneys could yield lower
concentrations near a given source but higher
concentrations further downwind. A second key
assumption is that the spaual association between
emission sources and population densities in the USA is
representative of the rest of the world. In some
developing regions, such as the mega-cities of Southeast
Asia, whose population patterns are more geographically
dense, this approach is likely to underestimate acttial
human exposures to airborne pollutants. Also, the source-
receptor-matrix model makes no estimates of indoor air-
pollutant exposures to PM from coal or biomass used for
cooking and heating. Such exposures are known to be
very high in several developmg regions, such as China,
Africa, and India."

Our modelling indicates that concentrations of PM
from a fixed quantity of emissions in hot and dry regions
are about one-third of what would be expected from
those same emissions under most other climatic
conditions. The likely reasons for this are: higher
ambient temperatures comcide with greater mixing
heights, hence a greater volume of air for dispersion; and
the conversion of gaseous PM precursors such as SO.
mto mtermediate products (sulphuric acid) requires the
presence of water vapour. Where less water vapour is
available, conversion rates are slower and conversion
takes place over longer distances. Therefore, regional
levels of PM are expected to be lower in dryer climates.

To capture these influences of climate on PM
concentrations, the USA was divided into four roughly
equal geographic zones based on long-term climatic data:
hot/dry, houwet, cold/dry, and cold/wet. Unique source
strengths for each sector/fuel were derived for each zone.
Similar climatic data for each of the remammg eight
global regions were used along with data on population
densities to obtain the appropnate, representative
weighting among USA climate zones — eg, 90% of the
population of the Southeast Asia group was found to
reside m a climate which most closely resembles the
hot'wet designanon, while 10% resides in a hot/dry
grouping; none was in the cold/dry or cold/wet zones. To
derive an overall Southeast Asia estimate, the source-
receptor association for the hot/wet USA zone was
weighted 90%, and the hot/dry zone 10%. Had we not
adjusted the source-receptor matrix to account for
regional differences for climate, and instead used the
USA climate as a proxy for the rest of tile world, then the
modelled concentrations of PM would have been from
16% (climate policy Asia) to 130% (climate policy
Europe) lower.

This analysis projected future global patterns of energy
consumption and adjusted the overall USA source-



receptor association to account for climate differences in
other global regions. To project likely changes in
concentrations of PM, it is also necessary to estimate
future efficiencies of pollution-control equipment and
policies unrelated to climate-control mitigation. In our
analysis, two alternative pollution-control strategies were
assumed to occur, in developed and developing regions.
The USA Organisation for Economic Co-operation and
Development (OECD)-West and OECD-Asian countries
were assumed to use USA controls, including SO; and
NO, controls outlined m the USA, 1990, Clean Air Act.
These do not reflect controls that individual states of the
USA might adopt as a result of recent revisions to ozone
and PM National Ambient Air Quality Standards, which
are not expected to go into effect until after the adoption
of climate-change minganon strategies. All other regions
were assumed to start with no controls in the year 2000.
By the year 2020, these other regions were assumed to
attain 1990 USA control levels for transportation-related
emissions and 1970 USA control levels for point-source
emissions.

Our results are sensitive to the estimation of pollution-
control efficiencies for fuel combustion because emission
rates change in relation to these values. This is especially
true for situations where high levels of control are
estimated, because the difference between 90% and 99%
control is a ten-fold difference in emissions. For countries
that may have more stringent control programmes than
the USA (eg, certain sectors in Japan), use of USA
standards as a proxy may overestimate future PM
concentranons. It is also possible that our assumptions
about future pollution-control efficiencies are too
optimistic for developmg countries; if so, then we will
have underestimated emissions and associated health
effects of the control scenarios. The sensitivity of the
analysis to assumed control levels was evaluated by
calculating PM concentrations in non-annex 1 regions
and Eastern Europe and former Soviet Umon, which
would result in the absence of controls. Analysis for 2020
showed that removing controls would yield PM,„ and
PM;, concentrations of 65% and 32% higher,
respectively, than the values calculated in the BAU
analysis.

Further refinements to transportation-related emissions
in all non-USA regions were made to account for the
much higher reliance on diesel fuel in most of the world
than in the USA."-' The fraction of the energy consumed
in the transponanon sector which is diesel fiiel is different
m each country (eg, 60% in Latin America, 20% in the
USA) and the fraction is assumed to remain constant
throughout the study period. These refinements also
assume a steady reduction in use of lead in petrol starting
from present levels in the year 2000 to zero in 2010.
Increased use of catalytic converters to control vehicle-
e.xhaust pollutants will accompany phase-out of lead.

In the USA, data indicate that ambient PM,„
concentrations are on average 65% of total suspended
particulate concentrations; PM. , concentrations are on an
average about 50% of PM,„ concentrations; and fossil-
fuel combustion accounts for about 45% of measured
PM., concentranons.-'-' To validate modelled PM
concentrations, ambient monitoring data were gathered
for about 40 countnes. Virtually all of the data, however,
were for total suspended paniculate and no data were
found which established the fracnonal contribuuon from
fossil-fuel combusnon. If one were to assume that the



1344



Vol 350 • November 8, 1997



847



THE LANCET



percentage of total suspended particulate that is
contributed by fuel-combustion sources in the USA holds
throughout the world, then estimated PM concentrations
should be two to ten times higher than those developed in
this study. In our judgment, it is highly unlikely that other
regions experience long-term average PM concentrations
from fossil-fuel combustion that approach such levels.
Total suspended particulate concentrations in most of the
world are at least twice as high as those in the USA; this
is likely due to factors such as much higher contribution
to the coarser fraction (>2-5 (im) from such sources as
wind-blown dust (eg. Middle East) as well as a higher
fine fraction (<2-5 )im) contribution from sources such as
biomass burning from heating, cooking, and forest
clearing (eg, Southeast Asia).

Health-impact estimates

Extensive public-health literature in several countries
shows that both monality and morbidity are significantly
associated with exposure to PM.* Besides mortality,
studies in several countries have associated PM with
cardiovascular and respiraton' morbidity (eg, increased
hospital admissions and emergenc\'-room visits for acute
and chronic pulmonary effects, mcluding bronchitis,
asthma, cough and wheeze) ."'-'' LittJe is known about the
mechanism of action of PM Some toxicological studies
have produced bronchitic responses," arrhythmias," and
exacerbation of pneumonia in rats" from exposure to fine
particles, which are similar to those observed in
epidemiological studies. It has been found that tiny
respirable particles or airborne aerosols can interfere
directly with pulmonary function and plasma viscosity in
animals and human beings.""'' Mortality effects have
consistently been associated with short-term and long-
term exposure to PM in more than 40 studies done in
over ten countries.*"^"

The short-term exposure studies use daily time-series
analysis to assess changes in mortality in the general
population that are associated with changes in daily
concentrations of PM. Long-term exposure studies of
PM have used a prospective cohon design, which allows
the adjustment of potential confounder? in the
assessment of the association between PM and mortality.

Health benefits from reductions in PM have been
derived previously from epidemiological studies done in
the USA and in Europe.""" Long-term exposure studies
were used as the basis for the quantitative analysis
because the resvilts from these analyses have followed
single cohorts and adjusted for important confounders.
These analyses suggested mortality effects from both
long-term and short-term exposure to PM.""
Epidemiological studies were chosen for this quantitative
analysis only if they reported the relative risk (RR) as a
monotonic function of PM concentrations. The two age
groups for which RR has been estimated in long-term
studies are adults aged over 30 years and infants aged
between 1 month and 1 year. To evaluate the impact of
PM concentrations on adult mortality, the paper by Pope
and colleagues was selected as the basis for quantitative
analysis." For infant mortality, we used a study by
Woodruff and colleagues" that reported RR as a
continuous monotonic function of PM concentrations.
Pope and colleagues found that a 24-5 jig/m' increase in
PM,, is associated with a 11 7 (95% CI 1 09-1 -26)
increase in RR of total mortality for adults over 30 years.
Woodruff and colleagues, found that for a 10 ng/m'



increase in PM,„, an odds ratio of 104 (95% CI
1 02-1 07) for increases in total infant mortality between
1 month and 1 year associated v.ixh increases in PM,,,
concentrations

We assume that the association is the same in the rest
of the world as it is in the USA. Epidemiological studies
have found a relatively consistent association between
shon-term exposures to PM and mortalitj' in many
countries, including Chile, Brazil, and Western Europe." ''
Given the relatively consistent association found between
short-term exposure to PM and mortality across
countries, it is reasonable to assume that the health
effects from long-term exposures to PM in other regions
of the world parallel those of the USA."

The association between change in PM,, and PM,, and
change in mortality risk is modelled as an exponential
function. The exponenual functional form for the RR
comes from the specification of the logistic regression
model used in the study by Woodruff and colleagues, and
the Cox's proportional hazard model used m the Pope
study for evaluation of the association between PM and
mortality. In this model, the RR associated with a PM
change of A PM is:
RR (PM) = exp (P A PM)

The slope coefficient, (3, is denved from the chronic-
exposure epidemiological studies selected for use in this
analysis, and A PM is the incremental change in the
aimual mean (or median) PM concentration, had a
pohcy-based scenario not been adopted. The risk
reduction associated with adoption of the policy is:
[exp (P APM)-1].

The estimated regional incidence of avoided premature
morality is:

A mortahty=(RR [A PM]-1) (mortality rate) (population),
where the mortality rate and population are the adult or
post-neonatal regional estimates, derived from projections
of population by age and infant monality rates by
coimtry. These were obtained from the Bureau of Census
(USA) international database and aggregated to a
regional level. TTie projected baseline age-adjusted
regional adult (>30 years) monality rates were calculated
from individual country population projections (by 5-year
age group), adjusted for immigration and emigration with
World-Bank projected net-migration rates.

The source-receptor-matrix estimates annual mean PM
concentrations. The RRs in the Pope study are based on
the annual median of daily PM concentrations, which is
usually lower than the mean for air-pollution
concentrations.'- PM data, from the USA, suggests that a
■y distribution best describes the observed skewed
distribution of daily PM levels." Assuming a y
distribution, an estimated median can be derived from
the mean and an estimated maximum daily level.
Observed population-weighted peak-to-mean ratios from
^four USA geographic zones are applied to the source-
receptor association. A maximum likelihood routine was
used for each region in the climate-policy scenario to
estimate the 7 distribution with the estimated mean and
maximum daily level.

Results

The scenarios that form the basis for this exercise rest on
complex models of energy-use trends, estimated
emissions of carbon, projected levels of PM, and
associated impacts on mortality. Each of these models



Vol 350 • November 8, 1997



1345



848



THE LANCET



Year



2000
2005
2010
2015
2020

Total cumulative* avoided mortalrty 2000-2020



56000131000-83 000)
89 000149000-1310001
106 000 1 58 000-156 0001
138 000 ( 76 000-203 000)

16700001900 000-2 500 000)



137 000 1 75 000-204 OOOI
286 000 1 157 000-425 0001
508000 I27800O-750 000)
5630001309000-831000)

6 340 000 1 3 500 000-9 400 OOOl



193 000 1 106 000-287 0001
375 000(206 000-557 000)
614000(337000-906000)
701 000 ( 385 OOO-l 034 000)

8010000(4 400000-11900000)



"Cumulatjve is sum o' annual deatris eacn year between 2000 and 2020 Al=anne« 1 (developed) countries. NAl=non-annex 1 (developing) countnes.

Table 2: Annual avoided mortality with 90% CIs (combined adult and neonatal infants) from climate-policy scenarios (15% below

1990 levels by 2010 for Al countries and 10% below BAU levels by 2010 for NAl countries)



necessarily has several assumptions and includes many
uncertainties. Of the many limitations of this exercise,
two stand out: we have only calculated mortality impacts
of PM for adults over age 30 and infants of age 1 month
to 1 year; and we do not provide estimates of morbidities
related to air pollutants from fossil-fuel combustion. The
quantitative estimates of public-health impacts generated
through this assessment should be understood as a first
approximation of the likely magnitude of the effect, and
not as representing precise predictions of avoided
mortalities.

With these caveats in mind, if the climate-policy
scenario were to be implemented, substantial
improvements in fuel efficiency would be realised by
2020, and major reductions m PM pollution would be
achieved. By the end of the second decade of the next
century, the policy we have outlined here would avoid
about 700000 deaths annually (90% CI
385000-1 034 000), of which 563000 (309 000-831 000)
and 138 000 (76 000-203 000) would occur in non-annex
1 (developing) and annex 1 (developed) countries,
respectively (table 2). Note that the CI only reflects
uncertainties in the epidemiological studies. If BAU
climate policies continue, the greatest impacts will occur
in developing coimtnes.

We estimate that implementation of the cUmate-polic>'
scenario could prevent up to 8 million deaths (90% CI
4-4-ir9 million) worldwide during the first 20 years of
the next century, corresponding to 1-67 million (0-9-2-5
million) and 6-34 million (3-5-9-4 million) avoided
deaths in developed and developing countries,
respectively. There are higher expected deaths in
developing countnes compared with developed countries
because of the higher proportional expected changes in
PM concentrations under the climate-policy scenario.

To assess the sensitivity of the numbers to certain key
assumptions in the estimation of PM concentrations, we
considered two analyses. TTie first assesses the effect of
assuming that the USA climate is representative of that in
each of the other global regions examined. The second
assesses the effects of assuming that there is no change in
emission-control levels in the BAU baseline for the non-
annex 1 countries and in Eastern Europe and the former
Soviet Union. Table 3 presents the results for the





NAl couimin


AlCOWItltU


ToUJ


Primary nealth analysis


6 300000


1700 000


8000000


Holding worldwide climate estimate


3 800000


1200000


5 000000


same as m trie USA








Holding emission control levels


7600000


1700000


9 300000


constant in trie BAU in NA 1 and








Eastern Europe









NA l=develooing (nonanne* 1 countries: Al=ileveloDed (anne* 11 countries.
BAU=Dusiness-as-usual scenario

Table 3 : Sensitivity analysis of avoidable deaths from climate
policy: 2000 to 2020



cumulative number of deaths avoided under each of these
scenarios. If we assume that the climate in the USA is the
same in the rest of the world, this results in fewer
estimated avoidable deaths than our primary health
analysis between 2000 and 2020 in both non-atmex 1
countries (3-8 million vs 6-3 million in our primary
analysis) and annex 1 countries (1-2 million vs 1-7
million). The alternative assumption, about control levels
in the BAU baseline, results in more estimated avoidable
deaths in non-annex 1 countnes (7-6 million vs 6-3
million).

Discussion

One of the most cntical set of assumptions in this
scenario assessment is that present levels of PM are
causing adverse effects on mortality of the magnitude
reponed by Pope and colleagues. The assumption that
fme particles are the direct cause of these effects, and not
merely conelated with them is also critical to our analysis.
It is these fine particles that can be transported globally
because they are not trapped by conventional paruculate-
control technologies, are buoyant, and can travel
thousands of miles through the air before they are
deposited. Fine PM can produce effects on health in
distant countries.

The climate-policy scenano that we use makes several
assumptions that affect the estimates of the public-health
impact of improved policies for climate control. We
assume an ambitious phase-in of poUuuon-control
equipment in developing countries from policies
unrelated to the climate-change-mingation scenarios. If
countnes fail to use such pollution-control equipment,
air-pollutant emissions will increase, and associated
deaths and illnesses will be higher than those estimated
under the BAU scenano, as seen in the sensitivity
analysis. The effectiveness of policies to reduce
greenhouse gas in the climate-policy scenano would be
even greater. Second, the au - quality modelling protocol
we use assumes that the USA distribuuon of chimney
heights is representanve of that throughout the world.
Many developing countnes have lower chimneys for
power plants and industnal boilers and greater population
densines surrounding mduscnal facilities, so this
assumption probably causes an underesnmation of
avoided deaths from greenhouse-gas reduction in the
non-annex 1 countnes. However, if populations are
concentrated at greater distance downwind, then lower
chimney heights would probably yield lower PM
exposures and result in an overestimation of avoided
deaths. Third, the analysis does not include reductions in
morbidity due to lower PM, or reductions in morbidity
from other pollutants.

There are several uncenainues associated with the use
of the prospective cohon studies for estimation of
monality effects from PM exposure worldwide. While the



1346



1 350 • November 8. 1997



849



THE LANCET



prospective cohort studies have anempted to account for
all important confounders, some potential confounders
may not have been considered. However, sensitivit>-



Online LibraryUnited States. Congress. House. Committee on ScienRoad from Kyoto : hearing before the Committee on Science, U.S. House of Representatives, One Hundred Fifth Congress, second session (Volume pt. 2) → online text (page 107 of 137)