M. Granger Morgan, Peter J. Adams, and David W.
Keith conducted a survey, Elicitation of Expert Judgments of Aerosol
Forcing, compiled and analyzed the results and produced the following
article which has been accepted for publication by Climatic Change.
ELICITATION
OF EXPERT JUDGMENTS OF AEROSOL FORCING
M. GRANGER MORGAN1, PETER J. ADAMS2, AND DAVID W. KEITH3
1Department of Engineering and Public Policy, Carnegie Mellon University,
Pittsburgh, PA 15213 USA E-mail: granger.morgan@andrew.cmu.edu
2Department of Engineering and Public Policy and Department of Civil
and Environmental Engineering, Carnegie Mellon University, Pittsburgh,
PA 15213 USA E-mail: petera@andrew.cmu.edu
3Department of Chemical and Petroleum Engineering and Department of
Economics, University of Calgary, Calgary, Alberta, Canada T2N 1N4
E-mail: keith@ucalgary.ca
Abstract: A group of twenty-four leading atmospheric and climate scientists
provided subjective probability distributions that represent their
current judgment about the value of planetary average direct and indirect
radiative forcing from anthropogenic aerosols at the top of the atmosphere.
Separate estimates were obtained for the direct aerosol effect, the
semi-direct aerosol effect, cloud brightness (first aerosol indirect
effect), and cloud lifetime/distribution (second aerosol indirect
effect). Estimates were also obtained for total planetary average
forcing at the top of the atmosphere and for surface forcing. Consensus
was strongest among the experts in their assessments of the direct
aerosol effect and the cloud brightness indirect effect. Forcing from
the semi-direct effect was thought to be small (absolute values of
all but one of the experts' best estimates were < 0.5 W/m2). There
was not agreement about the sign of the best estimate of the semi-direct
effect, and the uncertainty ranges some experts gave for this effect
did not overlap those given by others. All best estimates of total
aerosol forcing were negative, with values ranging between - 0.25
W/m2 and - 2.1 W/m2. The range of uncertainty that a number of experts
associated with their estimates, especially those for total aerosol
forcing and for surface forcing, was often much larger than that suggested
in 2001 by the IPCC Working Group 1 summary figure (IPCC, 2001).
1.
Introduction
Working
Group 1 of the third assessment of the IPCC (2001) reported that their
largest uncertainty about the magnitude of anthropogenically induced
radiative forcing was associated with the direct and indirect effects
of aerosols (Figure 1). Evidence on this forcing is available from
a variety of sources including: laboratory studies; local measurements; satellite-based global
measurements; mesoscale or cloud resolving models; and modeling studies
based on general circulation models (GCMs). Estimates can be based
on "forward" calculations in which observed or modeled
aerosols are used to calculate radiative forcings or on "reverse"
calculations in which the aerosol forcing is computed as the "missing"
forcing required for a GCM simulation of past climate to reproduce
the observed temperature record (Anderson et al., 2003). The IPCC noted that the "quantification of aerosol radiative
forcing is more complex than the quantification of radiative forcing
by greenhouse gases because aerosol mass and particle number concentrations
are highly variable in space and time…The quantification of indirect
radiative forcing by aerosols is especially difficult…[because] in
addition to the variability in aerosol concentrations [it depends]
on some quite complicated aerosol influences on cloud processes…"
(IPCC, 2001).
Taken
alone, none of the available sources of
evidence can provide definitive answers about the magnitude of aerosol-related
radiative forcing, nor is there any unambiguous objective method of
combining results in the literature to produce a probabilistic estimate.
Accordingly, we asked a group of leading experts to consider and carefully
synthesize the full range of available evidence and then provide their
judgments in the form of subjective probability distributions for
a number of standard measures of aerosol forcing. Such estimates
implicitly depend on each expert's weighing of current scientific
evidence.
Such
formal elicitation of expert judgment has been widely used in applied
Bayesian decision analysis (DeGroot, 1970; Spetzler and Staël von
Holstein, 1975; Watson and Buede, 1987; von Winterfeldt and Edwards,
1986; Morgan and Henrion, 1990), often in business applications, and
in climate and other areas of environmental policy through the process
of "expert elicitation" (Morgan et al., 1978a; Morgan et
al., 1978b; National Defense University, 1978; Morgan et al., 1984;
Morgan et al., 1985; Wallsten and Whitfield, 1986; Stewart et al.,
1992; Nordhaus, 1994; Morgan and Keith, 1995; Budnitz et al., 1995
and 1998; Morgan, Pitelka and Shevliakova, 2001). The results of
such studies provide a clear indication of the nature and extent of
agreement within a scientific community and also allow conclusions
to be drawn about how important the range of expert opinions is to
the overall policy debate. Sometimes apparent deep disagreements
make little difference to the policy conclusions; sometimes they are
of critical importance (Morgan and Henrion, 1990).
The
method used here is based on a structured elicitation of each expert's
judgment. It differs from group-based methods such as Delphi (Dalkey,
1969; Linstone and Turoff, 1975) of the more recent expert group method
developed by Budnitz et al. (1995) in that we do not seek consensus
between experts, nor do we provide a mechanism for iterative communication
between experts. An advantage of the method used here is that it
can effectively test the range of expert judgments unhampered by social
interactions, which may constrain discussion of extreme views in group-based
settings.
Expert
judgment is not a substitute for definitive scientific research.
Nor is it a substitute for careful deliberative expert review of the
literature of the sort that is undertaken by the IPCC. It can, however,
provide a more systematic representation of the diversity of expert
judgment than is typically provided in consensus reports, and thus
valuable input to experts performing such reviews. Indeed, it was
a request from two of the authors of the current IPCC assessment that
led us to conduct this study. It can also provide insights for policy
makers and research planners while research to produce more definitive
results is ongoing. It was for these reasons that Moss and Schneider
have argued such elicitations should become a standard input to the
IPCC assessment process (Moss and Schneider, 2000).
2.
The Survey Instrument
We
developed a survey instrument that asked experts to consider the factors
that contribute to uncertainty in four standard global measures of
average anthropogenic aerosol forcing at the top of the atmosphere:
· the direct aerosol effect,
which we defined as the "change in radiative flux by scattering
and absorption of unactivated aerosol particles in the absence of
any other climate changes or feedbacks";
· the semi-direct aerosol
effect, which we defined as the "change in radiative flux
resulting from a change in cloud distribution because of local heating
by absorptive (e.g., black carbon) aerosols";
· the first aerosol indirect
effect, which we defined as the "change in cloud reflectivity
resulting from a change in concentration of cloud condensation nuclei
holding other cloud properties constant (e.g., total liquid water
and cloud cover)"; and
· the second aerosol indirect
effect, which we defined as the "change in cloud cover/lifetime
resulting from a change in cloud condensation nuclei."
We provided these definitions at the beginning of the survey and then,
to minimize confusion about definitions, repeated each one before
asking questions about each type of forcing.
We
also asked for a judgment of total aerosol forcing. Finally,
since absorbing aerosols cause large surface forcings that affect
the hydrological cycle, but are not captured by total aerosol forcing
at the top of the atmosphere (Ramanathan et al., 2001), we also asked
for a judgment of the surface forcing for aerosols. In order
to help experts keep the taxonomy of direct and indirect forcings
straight, we included the simple diagram shown in Figure 2.
Given
the substantial uncertainty in estimates of aerosol forcing, it is
important to know how soon uncertainties might be reduced by future
research. We probed experts' judgment about the rate at which uncertainty
might change in the future. After the experts provided their estimates
of top of the atmosphere forcings, we posed the following question:
Suppose
we were to come back to you in 20 years and ask this question again.
Consider the full range of your uncertainty from lower to upper bound.
What is the probability that after 20 years of additional research
at current levels of support the outer tails of your box plot for the current global average magnitude of
<here we inserted the name of the specific forcing we were asking
about>:
Please
enter a separate number for each of the four contingencies.
______
will have gotten longer (i.e., taller)
______
will have gotten shorter by 0 to 50%
______
will have gotten shorter by 50% to 80%
______
will have gotten shorter by more than 80%
total
probability = 1.0
Part
1 of the survey explained the motivation of the study. Part 2 defined
the forcings we would be asking about. Part 3 provided background
information on problems of bias and overconfidence in expert judgment.
Parts 4-9 contained the elicitation itself. Finally, in Part 10 we
asked the experts to assess their own expertise on a scale from 0
(not familiar with this literature) to 7 (among the handful of top
experts in the world) and to tell us about the kinds of information
that played the most important role in shaping their judgments.
The
survey was designed to mimic protocols that we have employed in face-to-face
interviews. These protocols aim to minimize the effect of various
common cognitive biases found in lay and expert judgment. For each
of the four standard measures of aerosol forcing, we first asked the
experts to consider the factors that contribute to uncertainty before
asking for their judgments about the strength of the forcing. In the
elicitation itself, we first asked for extreme values so as to reduce
the impact of anchoring, then asked the expert to consider counterfactual
conditions that might widen their distributions so as to minimize
over confidence, and then asked for interior points in the distribution
before finally asking for a best estimate (Morgan and Henrion, 1990,
Dawes, 1988, Kahneman et al., 1982).
After
doing this systematically for the case of the direct aerosol effect,
we introduced the idea of "box plots" (Tukey, 1977) and
asked the experts to transcribe their first set of responses into
a box plot to represent their judgment about the forcing resulting
from the direct aerosol effect. In all subsequent questions, we asked
experts to respond directly by constructing box plots. In each case,
the instructions read:
To minimize the risk of overconfidence, please start by drawing
short horizontal lines to denote the lower and upper extreme values.
Ask yourself if you could explain smaller and larger values if they
were found in the future, and if so, revise your bounds accordingly.
Then fill in the other elements of the box plot (X=5%; =25%; = best estimate; =75%; X = 95%).
In
the box plots displayed in this paper, a slightly different graphical
convention was employed, as explained in the caption of Figure 3.
The
survey instrument was iteratively tested and refined with two advanced
Ph.D. students and one post-doctoral scholar studying aerosol forcings
and was also reviewed by a senior colleague in the field. A copy
of the survey instrument is available at http://cdmc.epp.cmu.edu/docs/pub/survey.doc.
The
survey was administered to experts by mail. On January 18, 2005 we
sent e-mail inviting 60 experts asking them to participate. Resource
constraints precluded inviting the entire expert community to participate.
Instead, in developing the list of experts we attempted to generate
a representative sample of modelers and observationalists and to include
experts in general climate science, cloud physics and aerosols. We
built the initial list from our own knowledge of the field, a review
of recent publications, and list of participants in recent meetings
and workshops. After asking a senior colleague to review the draft
survey, and proposed list of participants, we concluded that the draft
and list were too model-centric. We revised the survey to make the
treatment more balanced and made a special effort to add names in
the other categories so as to achieve balance across different areas
of expertise.
In
our initial e-mail, we indicated that completing the survey would
require approximately two hours and that we could offer a modest honorarium
to those who could accept it. Our e-mail messages were accompanied
by a PDF copy of a "Dear Colleague" letter from Prof. Ron
Prinn of MIT which explained that he had encouraged us to conduct
the study. He asked the experts to participate, noting that the IPCC
Third Assessment had been "reluctant to provide even subjective
estimates of the mean and standard deviation for indirect radiative
forcing by aerosols." He argued that "it is very important
for policy-making to know if this situation persists." He concluded:
"I believe that the results from this effort, presuming its acceptance
in a peer-reviewed journal, will prove valuable to the scientific
community in the months ahead as they work to summarize the state
of knowledge in this field in the Fourth Assessment."
On
February 09 we sent follow-up email to 54 experts. We received six
declinations, in most cases on the basis that they did not have time.
We received no response to either message from 19 experts.
Printed
copies of the survey instrument were mailed to all 29 experts who
agreed to participate. Upon receiving the questionnaire, two responded
that they did not believe they had the necessary expertise. Of the
remaining 27, 24 completed and returned the survey. As explained
below, one additional expert completed the survey well after this
paper was submitted. Table 1 lists the experts whose responses are
included in this paper along with their affiliations. In reporting
results, numbers assigned to experts were randomized so that all results
are anonymous. We do this to assure that experts feel free to offer
their frank opinions, uninfluenced by possible peer expectations or
similar pressures.
Because
all of the authors read each other's papers, some participated together
in the preparation of the last IPCC consensus review, or are participating
in the current IPCC review, and some have written papers together,
we make no claim that the responses we have received are "independent"
in the sense that they have not been influenced by each other's views.
However, because our objective is to sample the range of current expert
opinion, and it is the nature of expert communities to engage in such
consultation, we do not view this as a problem. Readers are reminded
we are not sampling from a distribution which describes the true values.
The judgment of one of the outliners may be correct, and those who
share a consensus view may be wrong. It is for this reason that we
have cautioned against combining individual responses in the past,
and do so as well in this case (Keith, 1996).
In
two cases, experts returned a box plot that appeared highly anomalous.
We asked the two experts to confirm that they meant what they had
drawn. In both cases, they responded that they had not correctly
understood the question, and revised their answers. After this paper
was submitted for publication, we received one additional response
which contained three very anomalous responses. We asked this expert
for clarification but after waiting several weeks and receiving no
response, chose to leave those responses out when we submitted the
final version of the paper.
We
received several responses to the question about how the uncertainty
might change in 20 years, which did not sum to 1 (or to 100%). When
the error was small, we simply renormalized the four numbers. In
a few cases, where the error was > 20%, we checked with
the experts, who, in each case, indicated they had not correctly understood
the response mode, and provided us with a corrected response. Once
a draft of this paper was complete, we provided each expert with a
copy and with their own code number, and gave each an opportunity
to correct any major mistakes we might have made in interpreting their
responses.
While
we asked for anthropogenic aerosol forcing estimates only, Expert
9 chose to provide estimates of present day total (natural plus anthropogenic)
values for sulfate, nitrate, organic carbon, and black carbon but
did not include sea-salt or mineral dust. Therefore, his estimates
are only comparable to the others to the extent that the former are
predominantly anthropogenic and the latter are predominantly natural.
Uncertainties regarding the fraction of mineral dust and organic carbon
that are natural versus anthropogenic are not included in his range.
All
of the survey responses we received were completed with a level of
detail that clearly indicated that the experts had taken the task
seriously. Judging from the written comments and other indications
of effort, all respondents probably devoted the two hours we had estimated,
if not more, to completing the survey. However, we did not ask the
experts to report on how long they took to complete the survey.
3.
Results
Figures
3-8 report experts' judgments of the six measures of forcing that
we elicited. The table at the bottom of Figures 3-7 report how the
experts anticipate their uncertainty judgment for absolute upper and
lower bounds might change if we were to return to pose the same question
"after 20 years of additional research at current levels."
All
but one of the experts placed their best estimate of the value of
the forcing from the direct aerosol effect between 0 and -1 W/m2.
Six of those assigned a small probability (< 0.05) to the
forcing from the direct aerosol effect turning out to be as negative
as -2 W/m2 or more. Half the experts assigned at least
some probability to the forcing from the direct aerosol effect being
positive, but all but two placed the upper bound on possible positive
forcing at or below 1 W/m2. Note that six of the nine
experts whose self-assessment of expertise was 6 or 7 assigned no
probability to the forcing being positive, and the remaining three
assigned probabilities of between 0.05 and 0.25. While the range
of uncertainties reported by several of the experts is somewhat broader
than that reported in 2001 by the IPCC (see Figure 1), the results
are generally consistent.
Expert
estimates for the semi-direct aerosol effect show greater variation.
The levels of self-reported expertise are lower, and there are no
obvious trends with reported level of expertise. There is not agreement
about the likely sign of the forcing from the semi-direct aerosol
effect. Fourteen of the 21 experts who responded to this question
gave best estimates that are positive, one gave a best estimate of
zero, and five gave best estimates that are slightly negative (all
with absolute magnitude < 0.2 W/m2). For the
direct aerosol effect there was at least some overlap in the uncertainty
range of all the experts, while for the semi-direct aerosol effect
this was not true. For example, two experts assigned no probability
to this forcing being positive, and three assigned no probability
to this forcing being negative. Part of this disagreement over the
sign of the forcing produced by the semi-direct aerosol effect may
result from ambiguity in how the community defines this effect. The
phrase was first used to describe reduction in cloud cover
by atmospheric heating (Ackerman et al., 2000), which implies a positive
forcing. Indeed, Expert 13 commented that "By definition, the
semi-direct effect is positive." However, Johnson et al. (2004)
found that absorbing aerosols above clouds can increase their liquid
water path and reflectivity, which they termed "a negative semi-direct
forcing". Expert 10 explicitly stated that he did not include
this negative effect in his assessment. Expert 5 also commented on
the definitional ambiguity. This expert indicating that it was unclear
where to include the possibility that black carbon "might brighten
clouds in some circumstance more than it might diminish them by heating",
and asked rhetorically, "does this still count as the "semi-direct
effect?"
The
semi-direct effect has been postulated at least since the work of
Hansen et al. (1997), but relatively little attention was devoted
to it until the INDOEX field campaign highlighted its importance (Ackerman
et al., 2000). Therefore, at the time of the Third Assessment Report,
there was little basis for an assessment of its global magnitude.
It is mentioned in the final text of the report but it was not possible
at that time to provide an uncertainty estimate. Therefore, the uncertainties
reported in Figure 4 represent an additional source of uncertainty
not quantified in the IPCC summary figure.
Expert
judgments regarding the indirect effects on cloud brightness (first
indirect effect) and cloud lifetime/distribution (second indirect
effect) are recorded in Figures 5 and 6. Again there are no obvious
trends with self-reported levels of expertise. When comparing their
responses to the IPCC summary figure (our Figure 1), it is important
to bear in mind that the 0 to –2 W/m2 uncertainty range
given there refers only to the cloud brightness or first indirect
effect. The text and figure labels and captions make this clear in
the main body of the report and in the technical summary. However,
the figure in the Summary for Policymakers, on which our Figure 1
is based, mentions this only in the caption. This omission has caused
some confusion.
Section
6.8.5 of IPCC Working Group I report states that "Available GCM
studies suggest that the radiative flux perturbations associated with
changes in cloud lifetime/distribution (second indirect effect) could
be of similar magnitude to that of the first effect". However,
the authors "refrain from giving any estimate or range of estimates
for the second aerosol indirect effect" because of a lack of
observational support for the GCM studies and because of debate as
to whether this effect can be interpreted as a radiative forcing in
the normal sense.
Most
experts who responded to our survey provided uncertainty ranges for
both aerosol indirect effects that are much wider than those provided
for the direct effect. Of the 23 experts who provided judgments regarding
the cloud brightness (first) indirect effect (Figure 5), five said
this could be as strong as –3 W/m2 or even more negative.
Fifteen gave uncertainty ranges from slightly positive to –2.5 W/m2.
These estimates are broadly consistent with the IPCC summary figure.
The remaining three respondents did not provide absolute upper and
lower bounds but provided 5th and 95th percentile
confidence bounds that are similarly consistent with the IPCC summary
figure.
Despite
the greater uncertainty, there was strong agreement that the forcing
associated with the cloud brightness (first aerosol indirect effect)
is negative. While one respondent did not provide an estimate, the
remaining 23 gave best estimates that ranged from - 0.4 W/m2 to -1.25W/m2.
For
the cloud lifetime/distribution or second indirect effect, 22 experts
provided responses. Five said this effect could be positive. Two
of these said they thought it was conceivable that this effect is
greater than 2 W/m2 although these estimates fell outside
their 95th confidence intervals. Eight experts said this
effect could be –3 W/m2 or more negative. Expert 24 did
not give absolute bounds to this effect while the remaining 13 experts
said this effect could not be more negative than –2 W/m2.
Collectively, these results indicate that the experts who responded
assigned generally greater uncertainty to the cloud lifetime/distribution
(second) effect than even the cloud brightness (first) indirect effect.
Given that the IPCC Working Group I did not estimate an uncertainty
range for this effect, this represents a significant, but often unappreciated,
uncertainty in aerosol climate forcing.
There
was similar agreement about the sign of the best estimates of the
cloud lifetime/distribution or second aerosol indirect effect, although
in this case, the uncertainty range produced by five experts assigned
probabilities of > 0.05 that the value may actually be positive.
Two experts did not respond, one gave a best estimate of 0, the remaining
21 all gave best estimates that were negative, of which only 5 exceeded
an absolute value of <1W/m2.
Figure
7 reports the experts' estimates of total aerosol forcing. Again
we see no trends with levels of self-reported expertise. All but
one of the experts made a separate estimate of this value. Expert
9 instructed us to treat his four distributions as independent, and
perform a sum.
There
is complete consensus that the best estimate of total aerosol forcing
is negative. Values of the experts' best estimates range from - 0.25
W/m2 to - 2.1 W/m2. Fifteen of the 24 responses
involve best estimate values between -1.0 W/m2 and -1.5
W/m2. A direct comparison with the IPCC results
is difficult because this work includes uncertainties associated with
the semi-direct effect and cloud lifetime/distribution (second) indirect
effect that were not estimated in the Third Assessment Report. While
the range of "best estimates" presented here appears to
be comparable to the estimate of IPCC 2001, the level of associated
uncertainty assessed by almost half the experts appears to be considerably
greater than that suggested by the IPCC in 2001. Most of the additional
uncertainty reported here results from inclusion of uncertainties
associated with the semi-direct effect and cloud lifetime/distribution
(second) indirect effect.
All
of the 17 experts who responded to the question about surface forcing
(Figure 8) gave best estimates that were negative, ranging from -1
W/m2 to - 4 W/m2. Several who reported very
low levels of expertise choose not to respond. All but two of the
responses involved an uncertainty range whose absolute value was wider
than 2.5 W/m2 and five experts provide estimates with an
uncertainty range of > 7 W/m2.
Note
that several of the responses reported in Figures 3-8, the 0.05 and
0.95 points on the box plots are significantly closer to the ends
of the boxes than one would anticipate for a singly peaked probability
density function. Since most of the experts probably did not intend
to report a bimodal distribution this suggests that several have either
reported boxes that are too long, or tails and 0.05/0.95 values that
are too tight.
In
the past, in conducting face-to-face expert interviews using a more
elaborate response mode, it has been possible to avoid obvious inconsistencies
or to rectify them when they arose. However, even in those cases,
we have seen clear signs that experts have produced distributions
that are too narrow (Morgan and Keith, 1995). Given the large literature
on overconfidence (Morgan and Henrion, 1990), we are inclined to believe
that in many of the responses we have received in the current study,
the problem is tails and 0.05/0.95 values that are too tight, not
boxes that are too wide. We pointed out this issue in a follow-up
e-mail to the experts. Three experts revised their distributions.
Six responded that they understood the problem but that since only
the extremes of the distributions were affected, they did not want
to make changes.
In
summary, we caution readers against placing faith in the precise values
of the extremes of some of the distributions we have reported, especially
those in which the 0.05/0.95 tick marks are very close to the ends
of longer boxes. However, we believe that the 0.25-0.75 boxes, and
reported best estimates, do a good job of correctly reporting the
experts' views.
In
addition to broadening several of his distributions to correct this
problem, Expert 7 also wrote: "My initial estimates were primarily
based on simulations with global models. Now tying it more to observational
data the forcing can be more negative… It is not so much the properties
of aerosol, which introduce these uncertainties but rather the relative
altitude positioning of aerosol with respect to clouds."
Past
work has indicated that "forward" calculations of total
aerosol forcing tend to lead to larger uncertainties than "reverse"
calculations (Anderson et al., 2003). Sixteen experts indicated that
they simply relied on forward calculations to provide their judgments
while only Expert 16 indicated a use of pure reverse calculations.
Of the remaining experts, Expert 3 indicated that he used forward
calculations "tempered by reverse considerations", Experts
8, 10, 11, and 20 indicated consideration of both, and Experts 14
and 23 specified that they used reverse calculations for total aerosol
forcing and forward for the individual effects. Given the small number
of experts who relied on reverse methods for total aerosol forcing,
it is difficult to determine whether there were systematic differences
between experts using forward and reverse methods to assess total
aerosol forcing. While the experts using reverse methods did reject
the possibility that aerosol forcing may be more negative than –2.5
W/m2 (i.e., the anthropogenic greenhouse gas forcing),
they were not the only experts to do so.
Table
2 lists the number of times an expert ranked a given factor as a major
contributor to uncertainty in the direct, semi-direct, or either of
the indirect effects. For the direct effect, factors controlling
the total mass burden of anthropogenic aerosol (anthropogenic mass
emission rates, production of condensable gases) were mentioned by
more than 10 experts in addition to physical and chemical properties
of carbonaceous aerosols (composition of primary emissions and aerosol
mixing state). While composition of primary emissions were also often
mentioned in the context of the brightness (first) indirect effect,
uncertainties with respect to cloud microphysics, dynamics, amount,
and distribution were important for both indirect effects and the
semi-direct effect.
Experts
do not expect that uncertainty about aerosol forcing will be resolved
quickly. Of the 20 experts who provided estimates of how uncertainty
about total aerosol forcing will change in 20 years, all but three
thought that the probability that uncertainty would grow was larger
than, or equal to, the probability that it would shrink by >80%;
and only two thought that there was a greater than even chance that
uncertainty would shrink by more than 50% (Figure 7).
4.
Conclusion
While
best estimates of average anthropogenic aerosol forcings obtained
in this survey are generally consistent with values suggested by the
IPCC in 2001, the range of uncertainty assessed by a number of experts
is significantly larger. Most of the additional uncertainty reported
here results from the inclusion of uncertainties associated
with the semi-direct effect and cloud lifetime/distribution (second)
indirect effect. A minority (five experts) also suggested that the
cloud brightness (first) indirect effect could be as strong as –3
W/m2, somewhat beyond
the range reported by IPCC in 2001. Consensus
was strongest among the experts for the direct aerosol effect and
for the cloud brightness effect (first aerosol indirect effect).
While forcing from the semi-direct effect was thought to be small
(absolute values of all but one of the experts' best estimates were < 0.5 W/m2) there was not agreement on the sign. All best estimates of total aerosol forcing were negative,
with values ranging between -0.25 W/m2 and -2.1 W/m2. Where uncertainties were greatest (i.e., the indirect effects)
or there was strong disagreement (i.e., the sign of the semi-direct
effect), aerosol-cloud interactions were mentioned as major contributors
to uncertainty.
Acknowledgments
We
thank the experts who participated in this study for their cooperation
and assistance. We thank Kaiping Chen, Jeffrey Pierce, Serena Chung
and Spyros Pandis for assistance in the development of the survey.
We thank Barbara Bugosh and Patti Steranchak for administrative and
secretarial assistance. This work was supported by National Science
Foundation award SES-034578.
When
in print, a link will be provided for the article including all tables
and figures.
The Survey