Scientists’Views about Attribution of Global Warming
Rob van Dorland,
and Leo Meyer
PBL Netherlands Environmental Assessment Agency, PO Box 303, 3720 AH Bilthoven, The Netherlands
Energy Research Centre of The Netherlands ECN, PO Box 1, 1755 ZG Petten, The Netherlands
University of Queensland, 4072 Brisbane QLD 4072, Australia
University of Western Australia, Crawley Washington 6009, Australia
Royal Netherlands Meteorological Institute (KNMI), PO Box 201, 3730 AE De Bilt, The Netherlands
ABSTRACT: Results are presented from a survey held among
1868 scientists studying various aspects of climate change,
including physical climate, climate impacts, and mitigation. The
survey was unique in its size, broadness and level of detail.
Consistent with other research, we found that, as the level of
expertise in climate science grew, so too did the level of
agreement on anthropogenic causation. 90% of respondents
with more than 10 climate-related peer-reviewed publications
(about half of all respondents), explicitly agreed with
anthropogenic greenhouse gases (GHGs) being the dominant
driver of recent global warming. The respondents’quantitative
estimate of the GHG contribution appeared to strongly depend
on their judgment or knowledge of the cooling eﬀect of
aerosols. The phrasing of the IPCC attribution statement in its fourth assessment report (AR4)providing a lower limit for the
isolated GHG contributionmay have led to an underestimation of the GHG inﬂuence on recent warming. The phrasing was
improved in AR5. We also report on the respondents’views on other factors contributing to global warming; of these Land Use
and Land Cover Change (LULCC) was considered the most important. Respondents who characterized human inﬂuence on
climate as insigniﬁcant, reported having had the most frequent media coverage regarding their views on climate change.
The general public is strongly divided over the question of
human causation of climate change.
Many believe that climate
scientists are equally divided with respect to the same question,
in contrast to what several studies
have found. Perceptions
about the level of agreement or disagreement among scientists
inﬂuence people’s acceptance of scientiﬁc conclusions and their
support for related policies.
Public perception of climate
change and of the scientiﬁc consensus on the subject, in turn, is
inﬂuenced by ethical, social, and political values and attitudes.
Scientists are considered a trusted source of climate
hence, public commenters frequently use the
purported existence of either strong agreement or strong
disagreement as an argument pro or contra the validity of
assessments by the Intergovernmental Panel on Climate
Change (IPCC). Leviston and Walker
showed that the
general public has a tendency to overestimate the prevalence of
contrarian opinions in climate science and to underestimate the
level of agreement.
Science is an evidence-based process, but the evidence has to
be interpreted (research) and weighed (assessment). These
interpretations and assessments are inﬂuenced by personal
knowledge of the evidence, but also by weighing the competing
credibility of diﬀerent experts and of diﬀerent explanations.
Thus, the “networks of trust”of survey respondents and their
views on the “consilience of evidence”will impact any survey
results. As the available evidence converges over time, scientists’
aggregate opinion can be expected to reﬂect this convergence,
resulting in a broadlythough not necessarily unanimously
We performed a detailed survey under a large group of
scientists studying various aspects of global warming and
climate change (including impacts and mitigation) and who
have published in peer-reviewed or, in a few cases, gray
literature. We explored the distribution of scientiﬁc opinion on
the causes of recent global warming, using the latest two IPCC
assessment reports, AR4
as a benchmark. An
attempt has been made to elucidate, in precise terms, the points
of both agreement and disagreement regarding the inﬂuence of
Received: April 25, 2014
Revised: July 14, 2014
Accepted: July 22, 2014
© XXXX American Chemical Society Adx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXX
anthropogenic GHGs. We investigated how the interplay
between climate warming by GHGs and cooling by aerosols
complicates the issue of attribution to GHGs only, as phrased
by the IPCC in 2007 in AR4. The relation between the
respondents’views on attribution and their self-reported
frequency of media coverage is brieﬂy explored, as this is
relevant to the public perception of consensus.
Several studies have investigated levels of consensus among
scientists in the discourse on climate change, using diﬀerent
questions, approaches, and sample sizes. However, scientiﬁc
consensus usually is characterized in imprecise ways. This is
one of the reasons, in addition to the pivotal role it plays in
public perception and policy support, that the debate about the
existence of consensus among scientists continues. Our study
distinguishes itself from earlier work, in the large size and broad
makeup of its survey sample and in the level of detail with
which we explored the distribution of scientiﬁc opinion; thus,
providing a more detailed description of what exactly is agreed
upon. Our survey, conducted in 2012, covered a wide range of
the physical science issues that are at the center of the public
debate on climate change.
In this article, we focus on the level of agreement or
disagreement regarding attribution of global warming to various
anthropogenic and natural causes. One of our survey questions
(Q1) was designed to be directly comparable with the well-
known statement of AR4: “Most of the observed increase in
global average temperatures since the mid-20th century is very
likely due to the observed increase in anthropogenic GHG
concentrations”. The comparable AR5 statement reads as
follows: “It is extremely likely that human activities caused
more than half of the observed increase in global average
surface temperature from 1951 to 2010.”The equivalent AR5
statement diﬀers from AR4 in two important aspects: the
likelihood level is increased in the AR5 statement, and it is
written in terms of “human activities”, whereas the AR4
statement speciﬁes “anthropogenic GHG concentrations”. This
last distinction is very relevant to our analysis and we argue that
the AR5 statement is clearer and less open to misinterpretation
than its AR4 equiv.
■MATERIALS AND METHODS
Survey Sample. Participation in our survey was sought
from scientists having authored or coauthored peer-reviewed
articles or assessment reports related to climate change.
Approximately 6000 names were assembled from articles with
the keywords “global warming”and/or “global climate change”,
covering the 1991−2011 period via the Web of Science.
Around 2000 names were collected from a public database
assembled by Jim Prall, based on scientiﬁc literature up to
supplemented by an additional ∼500 authors of recent
(2009−2011) climate science peer-reviewed literature. Prall’s
database also includes signatories of public statements
disapproving of mainstream climate science. They were
included in our survey to ascertain that the main criticisms of
climate science would be captured. This last group amounts to
less than 5% of the total number of respondents, about half of
whom had only published in the gray literature on climate
There was some overlap between these sources, with the
unique total number of names amounting to ∼8000. Based on
email address availability, 7555 of them were contacted. Of
these emails, ∼1000 were returned undelivered or unread,
leaving a total of 6550 people that were successfully
approached. 1868 questionaires were returned, although not
all of these were fully completed. This amounts to a response
rate of 29%. Each respondent could only respond to the survey
once. Survey results were analyzed anonymously.
Sample Representation. It is important to consider the
extent to which the group of people contacted was
representative of “climate scientists”. Having applied the
above-mentioned sources and criteria, we are conﬁdent that
most of the main players in climate science were invited. The
key word searches “global warming”and “global climate
change”ensured that we also sampled the wider scientiﬁc
ﬁeld, including those studying impacts and mitigation of global
warming. By also soliciting responses from signatories of public
statements who are not necessarily publishing scientists, it is
likely that viewpoints that run counter to the prevailing
consensus are somewhat magniﬁed in our results. This is
further exacerbated by this group exhibiting a relatively higher
response rate (see below). With the exception of this group, the
criteria used for selecting our survey sample are similar to those
used in other surveys studying the distribution of scientiﬁc
opinion on climate change, as discussed, for example, by Bray.
Survey invitees were tagged with certain characteristics,
which allowed us to check the level of representation of the
response group. These characteristics included information
regarding expertise in the form of one or more keywords, see
Supporting Information (SI). These were subsequently
grouped into Working Groups (WG) 1, 2, or 3, or according
to certain ﬁeld of expertise that refer to IPCC nomenclature:
WG1 (the physical science basis), WG2 (impacts, adaptation,
and vulnerability) and WG3 (mitigation of climate change).
Some people were tagged with multiple ﬁelds of expertise;
therefore, the total of these ﬁelds exceeds 100%. 619 invitees
were tagged as having been IPCC AR4 WG1 coordinating, lead
or contributing authors, and 218 were tagged as being
“unconvinced”of the evidence, based on their published
articles or signed public declarations critical of mainstream
climate science as embodied by the IPCC. The latter
information was extracted from Jim Prall’s public database.
Invitees were also tagged according to their country of
employment, based on their email addresses.
The relative prevalence of respondents with certain tags was
compared to their prevalence in the total group of invitees (see
Figure S1 in SI). The absence of a strong systematic bias led us
to conclude that the group of respondents overall could be
considered representative of the total group invited, with some
minor diﬀerences. WG1 and “other”ﬁelds of expertise were
slightly overrepresented among the respondents, as were
invitees tagged as “unconvinced”(3% of invitees against 5%
of respondents) and IPCC AR4 WG1 authors (8% and 9%,
respectively). Around 80% of both invitees and respondents
were from either North America or Europe, with the remainder
being predominantly based in Asia or Oceania.
Survey Questions. The survey focused on important topics
in the public debate on climate science, while also covering a
wide range of scientiﬁc topics related to the scientiﬁc basis of
climate change, not all of which are discussed in this article.
Answer options reﬂected a variety of viewpoints, all of which
were phrased as speciﬁc and neutral as possible. Questions and
answers were previewed by physical and social scientists and
climate change public commentators with a wide range of
opinions, to minimize the chance of bias. The main questions
investigated in this article are listed below, with a brief
Environmental Science & Technology Policy Analysis
dx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXXB
description of associated answer options. The complete set of
survey questions and answer options is freely available.
Q1. What fraction of global warming since the mid-20th century
can be attributed to human-induced increases in atmospheric GHG
concentrations? Quantitative answer options in percentage
ranges of GHG contribution. Answer options included
>100% (i.e., GHG warming has been partly oﬀset by aerosol
cooling) and <0% (i.e., GHG caused cooling).
Q1b. What confidence level would you ascribe to the
anthropogenic GHG contribution being more/less than 50%?
Answer options according to the IPCC likelihood scale.
Q3. How would you characterize the contribution of the
following factors to the reported global warming of ∼0.8 °C since
preindustrial times: GHGs, aerosols, land use, sun, internal
variability, spurious warming? Qualitative answer options ranged
from “strong cooling”to “strong warming”.“Spurious warming”
refers to global mean surface temperature change being
overestimated due to artifacts in the data, such as Urban
Heat Island (UHI) eﬀects.
Q3b. How would you describe the level of scientif ic under-
standing for each of these factors? Answer options ranged from
very low to high.
Q3c. An explanatory question was asked regarding all factors
other than GHGs (which were assigned a warming inf luence), each
with specific multiple choice answers.
Q4. What is your estimate of equilibrium (Charney) climate
sensitivity, i.e. the temperature response (degrees C) to a doubling of
CO2?Open, numeric answer.
Q6. Please indicate your field(s) of expertise in climate science.
Multiple choice answer options.
Q7. Please indicate the approximate number of climate-related
articles you have published in peer-reviewed scientif ic journals,
including as coauthor. Open, numeric answer.
Q11. How frequently have you featured in the media regarding
your views on climate change? Answer options ranging from “very
Aggregation of Results. Given the large sample size
(1868) and the diversity in scientiﬁc backgrounds of our
respondents, results were segregated according to ﬁelds of
expertise and publication metrics, as indicated by the
respondents in their respective answers to questions 6 and 7.
The self-declared ﬁelds of expertise were categorized as WG1,
2, 3, or other ﬁelds of expertise, analogous to the tagged
expertise ﬁelds (see SI). Around 65% of those with self-declared
WG1 ﬁelds of expertise also were tagged with WG1 ﬁelds of
expertise. Respondents who were labeled as “unconvinced”
indicated more often than other respondents that they had
expertise in one or more of the WG1 ﬁelds and they indicated
more expertise ﬁelds in general. For a subgroup of invitees,
Google Scholar metrics regarding number of publications were
also available as tagged information. For ﬁelds of expertise as
well as publication metrics, aggregated results did not strongly
depend on tagged or self-declared numbers. More details can
be found in the SI.
Attribution. The responses to Q3, on the qualitative GHG
contribution to global warming since preindustrial times, are
shown in Figure 1. Responses were segregated according to the
self-declared number of climate-related peer-reviewed publica-
tions, in four ranges of approximately equal size. About half the
respondents stated that they had authored or coauthored more
than 10 peer-reviewed climate-related publications. Responses
indicating a cooling inﬂuence of GHGs (11 responses or less
than 1% of the total) were grouped under the category
“insigniﬁcant”, for graphing purposes. The majority of
respondents selected the highest score (“strong warming”)
for the GHG contribution. This majority was even stronger for
respondents with the highest number of self-declared
publications. A similar, though less pronounced trend was
found for respondents with increasingly relevant ﬁelds of
expertise (see SI). Furthermore, 82% of AR4 WG1 authors
selected the “strong warming”option for this question (not
Q1 also concerned the contribution of GHGs, but then as a
percentage of observed warming since the mid-20th century.
This enabled a direct comparison with the well-known AR4
statement on attribution, which states that this contribution is
very likely (probability >90%) to be more than 50%. Less well-
known is the fact that IPCC in AR4 also states that GHG
forcing alone was likely (probability >66%) to have resulted in
greater than observed warming if there had not been an
oﬀsetting, cooling eﬀect from aerosol and other forcings. In
AR5 this was further clariﬁed. The net cooling eﬀect of aerosols
means that the sum of all warming contributions exceeds
This is the reason for including the answer option
“>100%”, which, even if counterintuitive, would be consistent
with both AR4 and AR5 and with recent research.
Their awareness of or judgment about the oﬀsetting eﬀect of
aerosols appears important in how respondents answered Q1,
as is discussed in more detail below. The proportion of
respondents who chose GHG > 100% was higher among
respondents with expertise in “attribution”or “aerosols and
clouds”(see Figure 2).
AR4 WG1 authors (not shown) responded similarly to those
with (self-declared) expertise in attribution or aerosols, also
preferentially selecting “>100%”. As the self-declared number of
publications increased, so did the proportion of respondents
selecting “>100%”, although still below the answer option of
Four respondents tagged as AR4 WG1 authors chose the
“26−50%”option and, as such, disagreed with AR4’s attribution
statement. Those who were tagged as “unconvinced”(N= 88,
not shown) consisted of two main subgroups: one claiming
only a minor eﬀect of anthropogenic GHGs (GHG < 25%),
Figure 1. Qualitative contribution of anthropogenic GHGs to global
warming since preindustrial times (Q3). Responses are shown as a
percentage of the number of respondents (N) in each subgroup,
segregated according to self-declared (SD) number of peer-reviewed
Environmental Science & Technology Policy Analysis
dx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXXC
and the other claiming the answer was “unknown due to lack of
knowledge”. Six of the “unconvinced”respondents selected the
option GHG > 50%, thus agreeing with AR4’s attribution
Consensus. Responses to Q1 and Q3 were both condensed
into three categories: (1) agreement; (2) disagreement; and (3)
undetermined (“unknown”,“I do not know”, and “other”).
Those who selected any of the options of GHG > 50% in
answer to Q1 were included in the “agreement”category. The
answer “no warming”was included in the “disagreement”
category. For Q3, responses were interpreted as “agreement”if
GHGs were accredited with strong warming or with moderate
warming if none of the other natural or anthropogenic factors
were deemed to have caused strong warming. So, according to
these respondents, GHGs were either the strongest or tied for
the strongest contributor to global warming.
In Figure 3 the distribution of respondents over the
categories “agreement”,“undetermined”, and “disagreement”
is shown for all respondents and for ﬁve diﬀerent subgroups:
the group of AR4 WG1 authors (N= 174) and four quartiles of
approximately equal size (N=∼400), based on their self-
reported number of publications. Results are shown separately
for the questions of qualitative (Q3) and quantitative (Q1)
Undetermined responses (unknown, I do not know, other)
were much more prevalent for Q1 (22%) than for Q3 (4%);
presumably because the quantitative question (Q1) was
considered more diﬃcult to answer. This explanation was
conﬁrmed by the open comments under Q1 given by those
with an undetermined answer: 100 out of 129 comments
(78%) mentioned that this was a diﬃcult question.
There are two ways of expressing the level of consensus,
based on these data: as a fraction of the total number of
respondents (including undetermined responses), or as a
fraction of the number of respondents who gave a quantitative
or qualitative judgment (excluding undetermined answers).
The former estimate cannot exceed 78% based on Q1, since
22% of respondents gave an undetermined answer. A ratio
expressed this way gives the appearance of a lower level of
agreement. However, this is a consequence of the question
being diﬃcult to answer, due to the level of precision in the
answer options, rather than it being a sign of less agreement.
As a fraction of the total, the level of agreement based on Q1
and Q3 was 66% and 83%, respectively, for all respondents, and
77% and 89%, respectively, for the quartile with the highest
number of self-declared publications. As a fraction of those who
expressed an opinion (i.e., excluding the undetermined
answers), the level of agreement based on Q1 and Q3 was
84% and 86%, respectively, for all respondents, and 91% and
92%, respectively, for the quartile with the highest number of
The similarity between the fractions as derived from Q1 and
Q3 (excluding the undetermined responses) suggests that it is
reasonable to interpret the answer option “moderate warming”
(provided no other factor was deemed to have caused “strong
warming”) as agreeing with the IPCC. The fraction of
respondents that disagreed with a dominant human inﬂuence
on climate was 12% and 14%, based on the answers to Q1 and
Q3, respectively. This group becomes smaller, 8% in both cases,
for the quartile with the highest number of publications. A table
with consensus estimates for the diﬀerent subgroups and
expressed in the above-mentioned two diﬀerent ways can be
found in the SI (Table S3). Excluding undetermined answers,
90% of respondents, with more than 10 self-declared climate-
related peer-reviewed publications, agreed with dominant
anthropogenic causation for recent global warming. This
amounts to just under half of all respondents.
Diﬀerent surveys are not directly comparable, due to
diﬀerent groups of people being asked diﬀerent questions.
However, since climate science surveys typically drew from the
same overall pool of climate-related scientists, Bray
that these can be meaningfully compared, to study the net
change in aggregate opinions. He concluded that the level of
consensus has grown over time. This is consistent with the
analysis of the peer-reviewed literature that shows a similar
increase in consensus.
Our results for the level of consensus are similar to those
found in other surveys.
Doran and Kendall-Zimmer-
reported an 82% consensus among 3146 earth scientists,
Figure 2. Percentages for the contribution of anthropogenic GHG to
global warming since the mid-20th century (Q1). Responses are
shown as a percentage of respondents (N) in each subgroup,
segregated according to self-declared (SD) ﬁelds of expertise “WG1”
(categorized as Working Group 1) and “attr or aer”(expertise in
attribution or aerosols and clouds).
Figure 3. Responses shown as percentages of agreement and
disagreement about the dominant inﬂuence of GHGs on global
warming, based on responses to Q3 (qualitative GHG contribution)
and Q1 (quantitative GHG contribution). Also shown are the
percentages of responses for the answer options “unknown”,“Ido
not know”, and “other”, combined and labeled as “undetermined”.
These answer options were much more prevalent for the quantitative
question (Q1). The level of agreement increases for respondents with
increased self-declared number of peer-reviewed climate-related
publications and is highest for AR4 WG1 authors.
Environmental Science & Technology Policy Analysis
dx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXXD
which rose to 88% for those who identiﬁed themselves as
climatologists, which is very similar to our ﬁndings. However,
Anderegg et al.,
and Cook et al.
reported a 97%
agreement about human-induced warming, from the peer-
reviewed literature and their sample of actively publishing
climate scientists, as did Doran and Kendall-Zimmermann
the most published climatologists. Literature surveys, generally,
ﬁnd a stronger consensus than opinion surveys. This is related
to the stronger consensus among often-published−and
arguably the most expert−climate scientists. The strength of
literature surveys lies in the fact that they sample the primary
fora where the evidence is laid out, whereas the strength of
opinion surveys such as ours relates to the fact that much more
detail can be achieved about the exact opinions of scientists. As
such, these two methods for describing scientiﬁc consensus are
complementary. Diﬀerent surveys typically use slightly diﬀerent
criteria to determine their survey sample and to deﬁne the
consensus position, hampering a direct comparison. It is
possible that our deﬁnition of “agreement”sets a higher
standard than, for example, Anderegg’sdeﬁnition (e.g., AR4
WG1 author or having signed a public declaration) and Doran
and Kendall-Zimmermann’s survey question about whether
human activity is “a signiﬁcant contributing factor”.
As indicated, contrarian viewpoints are likely overrepresented
in our sample (amounting to ∼5% of respondents), about half
of whom have published peer-reviewed articles in the area of
climate. However, this does not fully explain the diﬀerence with
the abovementioned studies. Excluding those tagged as
“unconvinced”more closely approximates the methodologies
of earlier studies and increases the level of agreement, for
example, from 84% to 87% based on Q1, excluding
undetermined responses. Moreover, we solicited responses
from a wide group of scientists. A larger proportion of those
not specializing in climate science research may be unconvinced
by or unaware of the scientiﬁc evidence for anthropogenic
causation, as was also found by Doran and Kendall-
Our results agree with Anderegg’s and Doran
and Kendall-Zimmermann’sﬁndings that the level of consensus
is strongest for actively publishing climate scientists. For
example, the level of agreement−excluding undetermined
responses−among AR4 WG1 authors, usually highly published
domain experts, for Q1 and Q3, was 97% and 96%, respectively.
Likelihood of a Dominant Human Inﬂuence. Responses
of the conﬁdence level of the anthropogenic GHG contribution
being larger or smaller than 50% are shown in Figure 4.
Respondents who estimated this contribution to be more than
50% (GHG > 50%) did so in combination with a higher level of
likelihood, than respondents who estimated this contribution to
be smaller than 50% (GHG < 50%). Of the group GHG > 50%,
89% assigned at least the same likelihood as the AR4 (“very
likely”) to GHGs contributing more than 50% to recent
warming; 65% chose a likelihood at least as high as that in AR5
for net anthropogenic activities (“extremely likely”). In fact,
“virtually certain”was selected most often by these respondents
(Figure 4). Those with more relevant self-declared ﬁelds of
expertise assigned a higher likelihood to their particular choice
than those with less relevant expertise. Only 39% of the group
GHG < 50% assigned a likelihood of “very likely”or stronger to
Contribution of Other Factors to Warming. Besides
asking for the qualitative contribution of GHGs to the warming
of ∼0.8 °C since preindustrial times, Q3 also asked about the
inﬂuence of other factors (see Figure 5). To allow averages to
be computed, the qualitative answer scale was transcribed
numerically, assuming the scale to be equidistant, meaning that
the distance between the diﬀerent answer options is assumed to
be identical. Average sample sizes are given in brackets; they are
not constant as they vary slightly according to the proportion of
undetermined responses (“unknown”and “I do not know”).
Consistent with AR4 and AR5, anthropogenic GHGs were
estimated to have had by far the strongest contribution to
There are however some diﬀerences with the IPCC
assessments in how some of the other factors were estimated.
Land Use and Land Cover Change (LULCC) is estimated by
IPCC to have exerted a small negative forcing (cooling) of
−0.15 (−0.25 to −0.05) W/m2due to an increase in surface
albedo. LULCC can also cause surface warming due to reduced
evaporation during the summer and in the tropics; this leads to
a vertical redistribution of heat and is thus not captured in the
reported radiative forcing in AR5 due to LULCC. However, in
our survey, on average, LULCC was deemed to have caused
Figure 4. Likelihood of anthropogenic GHG contribution being larger
(GHG > 50%) or smaller (GHG < 50%) than 50% (Q1b). Responses
are shown as a percentage of the respondents (N) in each subgroup.
The sample size is given in the legend. For respondents who selected
the GHG > 50% option, the assigned level of likelihood is segregated
according to self-declared (SD) ﬁelds of expertise “WG1”(categorized
as Working Group 1) and “attr or aer”(expertise in attribution or
aerosols and clouds).
Figure 5. Contribution of diﬀerent factors to the reported ∼0.8 °C
warming since preindustrial times, for diﬀerent groups of respondents
(Q3). Qualitative responses for each group were averaged under the
assumption of being equidistant. Average sample sizes (N) are shown
in brackets, for each group of respondents.
Environmental Science & Technology Policy Analysis
dx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXXE
slight warming. This estimate was substantially lower
(insigniﬁcant, on average) for both self-declared attribution
experts and AR4 WG1 authors. Other potential sources of
warming (from the sun, from natural variability, or warming
being partly spurious), on average, were estimated from
insigniﬁcant (4) to slight warming (5), without being strongly
dependent on the ﬁelds of expertise or AR4 WG1 authorship.
Aerosols were estimated by most respondents to have had a
cooling inﬂuence on climate, in line with the IPCC assess-
ments. However, 16% of all respondents selected (slight,
moderate or strong) warming for the net eﬀect of aerosols (see
also Figure 6).
The spread in results (see SI Table S4) is smaller for the
contribution of GHGs (standard deviation of 0.83 for all
respondents together and 0.78 for those with self-declared
WG1 ﬁelds of expertise) than for the contribution of most
other factors. This is also reﬂected in the higher level of
scientiﬁc understanding (Q3b) that was reported for GHGs
compared to the other factors (see SI Figure S9). The
contribution of aerosols shows a wide spread: highest standard
deviation of 1.44 for all respondents together and 1.26 for those
with self-declared WG1 ﬁelds of expertise.
When a factor, other than GHGs, was estimated to have had
a slight, moderate or strong warming eﬀect, a clarifying
question was asked in Q3c about the reasons for, or indications
of this inﬂuence. These were phrased as multiple choice
questions, for which more than one option per question could
be selected. Figure 6 shows the prevalence of responses,
segregated according to their choice of >50% or <50% GHG
This reveals the similarities and diﬀerences between these
two groups in how certain issues are perceived. Overall, the
UHI eﬀect, changes in solar irradiance, and both options within
the category “LULCC”were chosen as the most prevalent
contributors to global warming. As mentioned above, according
to the IPCC assessments LULCC has probably led to an
increase rather than a decrease in surface albedo.
The group GHG < 50% indicated twice as often alternative
factors to have contributed to the observed warming than the
group GHG > 50%, as expressed by the longer red bars in
Figure 6. For the categories “strong”and “moderate”warming
inﬂuence, the diﬀerence becomes even more pronounced: a
factor of 3.
Warming due to “spurious warming”and “natural variability”
was judged most diﬀerently between the two groups. In the
category “spurious warming”, referring to a warming bias in the
temperature record, the UHI eﬀect was the option chosen most
often by both groups. For the group GHG > 50%, the relative
popularity of this option, compared to other options, was more
pronounced. For the category “Natural Variability”, no large
diﬀerences were found between the indicated causes, except
that ‘spontaneous changes in cloud cover’was selected the least.
Even though the question concerned centennial scale warming,
Figure 6. Reasons for, or indications of other factors than anthropogenic GHGs having had a slight (light colored), moderate (medium colored) or
strong (dark colored) warming inﬂuence on global average temperatures since preindustrial times, in response to Q3c. Responses are shown as a
percentage of respondents, who selected either <50% or >50% GHG contribution under question Q1 (in red and blue, respectively).
Environmental Science & Technology Policy Analysis
dx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXXF
the short-term El Niño Southern Oscillation (ENSO) was
considered by both groups as the most important.
In the category “Sun”, most respondents chose total solar
irradiance as the most probable cause. The relative popularity of
this option, in comparison with others, was most pronounced
among the group GHG > 50%. Only a small fraction of this
group believed the inﬂuence of the sun to be magniﬁed
through, for instance, the ultraviolet part of its spectrum or the
cloud-forming potential of cosmic rays. In terms of what was
thought to have had a “strong”warming inﬂuence, among the
group GHG < 50%, “cosmic rays and clouds”were considered
as the second strongest contributing factor to global warming,
after changes in solar irradiance.
In the category “Aerosol”, the option of indirect eﬀects via
clouds was selected somewhat more often than absorption by
black carbon. The former option is a direct contradiction to the
IPCC, which states that although the magnitude of indirect
aerosol eﬀects is highly uncertain, its signa negative radiative
forcing, and thus a cooling inﬂuenceis not. Note that the
majority of respondents indicated a slight to moderate cooling
due to aerosols (Figure 5).
Aerosol Cooling Versus GHG Warming. Recent studies
concluded that it is very likely
(90% probability) or extremely
(95% probability) that GHG-induced warming since
the mid-20th century has been larger than the observed
warming. This is not surprising, considering that the radiative
forcing from GHGs (in 2005 compared to 1750) amounts to
about 140% of the total forcing.
AR4 was more conservative
regarding attribution, stating that GHG forcing alone would
likely (>66%) have resulted in greater than the observed
warming if there had not been an oﬀsetting cooling eﬀect from
aerosols and other forcings.
Most responses to Q1, thus, indicated a smaller GHG
contribution than what could be inferred from AR4 (Figure 2),
although most responses claim a higher level of conﬁdence than
in the AR4 about GHG contribution exceeding 50% (Figure 4).
According to Allen,
the quintessential AR4 attribution
statement, quoted toward the end of the introduction, focused
on GHGs rather than on the net anthropogenic eﬀect, in order
to have a more quantitative conclusion and a more justiﬁable
The relative prevalence in our survey of GHG contribution
estimations of between 50% and 100%, relative to >100%,
suggests that this AR4 statement leads people to underestimate
the GHG contribution. Potential reasons for this are the
following: (1) the AR4 statement only provides a lower limit
(“most”) for the GHG contribution; (2) this lower limit
(>50%) is far removed from an inferred best estimate which
exceeds 100%; (3) the inferred best estimate is counterintuitive
(because how could an isolated contribution exceed 100%?);
and (4) there is less awareness of the cooling eﬀect of aerosols
than of the warming eﬀect of GHGs and, thus, readers may
interpret this statement as referring to the net anthropogenic
The statement taken in isolation, without mentioning
oﬀsetting aerosol cooling, could very well be misinterpreted.
For example, Curry and Webster,
in their critique of this
statement, appeared to interpret “most”as meaning between 51
and 99% implying a nonexistent plateau at 100%. They deemed
the “very likely”designation to be an overstatement of the
probability in light of the highly uncertain amplitude of natural
variability. This in contrast to most respondents of our survey,
who assigned higher levels of likelihood to GHG contribution
exceeding 50%. Hegerl et al.
responded to Curry and Webster
by noting that the IPCC attribution statement assigns a lower
probability to this being correct than is implied in individual
studies, because structural uncertainty is taken into account.
Figure 5 shows that the higher the estimated GHG
contribution, the larger the estimated aerosol cooling. This
eﬀect was relatively strongest between the two highest GHG
categories (“76−100%”and “>100%”; see SI Figure S10).
Q4 asked the respondents’estimate of the equilibrium
climate sensitivity (ECS), that is, the global average temper-
ature increase due to a doubling of the atmospheric CO2
concentration. Figure 7 shows how the response to Q4 relates
to the estimated GHG contribution to recent warming.
Estimates for ECS exceeding 10 °C were excluded under the
assumption that these were made in error. Note that the total
sample size for Figure 7 is 913, as there were fewer responses to
Q4 than to Q1 and Q3. The higher the estimated GHG
contribution, the higher the average estimated ECS, except for
the highest two GHG categories, who both estimated 3.0 °C.
Between all other categories there is a signiﬁcant diﬀerence in
average ECS (ttest, alpha = 0.05). We pose that many
respondents did not distinguish the highest two GHG
categories on the basis of having a diﬀerent opinion about
the GHG inﬂuence, but rather because they had a diﬀerent
opinion or diﬀerent level of awareness about the contribution
This conclusion is also supported by the fact that the option
with the highest GHG contribution (GHG > 100%) actually
becomes relatively less prevalent for sensitivity estimates of
more than 3.5 °C, whereas the second highest GHG category
(GHG 76−100%) becomes relatively more prevalent (see SI
Figure S11). Also note that even for lower sensitivity ranges a
signiﬁcant proportion of the respondents (up to 75% in the
range from 1.5 to 2.5 °C) considers the GHG contribution to
be greater than 50%.
In AR5, the principal attribution statement was changed to
include “the anthropogenic increase in GHG concentrations
and other anthropogenic forcings together”, which remedies
some of the issues identiﬁed above with the equivalent AR4
Figure 7. Average estimates of equilibrium climate sensitivity (ECS, in
°C per doubling of the atmospheric CO2concentration), versus
estimates of the quantitative GHG contribution. Sample sizes for each
GHG category are noted on the x-axis (total N= 913). The two
categories with the highest GHG contribution are not distinguished by
the estimates of concomitant ECS, whereas the other categories are.
Environmental Science & Technology Policy Analysis
dx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXXG
Climate Sensitivity. Figure 8 shows the distribution of the
respondents’ECS estimates, shown in ranges for graphing
purposes. The peak for the ECS range (2.5−3.5 °C) is in the
middle of the “likely”range as assessed in AR5 (1.5−4.5 °C).
On the other hand, the skewed distribution shown in Figure 8,
with more responses for lower rather than higher values of
ECS, is diﬀerent from the distribution as inferred from theory
and as assessed by the IPCC, which has a fat tail toward higher
Media Exposure. Figure 9 shows the self-reported
frequency of media coverage (Q11) and how this relates to
the responses to the two above-mentioned questions on GHG
contribution (Q1 and Q3), as well as to their estimates of ECS
(Q4). For ease of presentation, the options “very frequently”
and “frequently”were combined, as were “rarely”and “never”
(see SI Figure S13 for more details). Those who estimated ECS
to be lower than 1.75 °C reported more “frequent”or “very
frequent”media coverage than those who estimated it to be
higher, although not all of these diﬀerences are statistically
signiﬁcant (see SI). Responses to the quantitative attribution
question (Q1) showed that respondents on either side of the
spectrum reported more frequent media coverage than the
group in the middle. The relative highest frequency of media
coverage was reported by those who attributed less than 25% of
global warming to GHGs, and those who attributed >100% to
GHGs, analogous to what can be inferred from the IPCC,
reported only slightly less frequent media coverage. Those who
estimated the qualitative greenhouse contribution (Q3) to be
insigniﬁcant or negative (i.e., cooling) reported signiﬁcantly
more frequent or very frequent media exposure than those who
estimate GHGs to have exerted either slight, moderate or
strong warming. Relative to their total number, 30% of the
group that selected “insigniﬁcant or cooling”reported being
featured frequently or very frequently in the media, as opposed
to 15% of the majority of respondents, who selected a strong
warming contribution of GHGs. When only taking the “very
frequently”responses into account, the diﬀerence between
those who regard GHG to cause warming versus those who do
not is even stronger (12% versus 4%). These diﬀerences are
statistically signiﬁcant (p= 0.01 and p= 0.04, respectively,
using the “Fisher’s exact test”) and indicate that those who
most strongly disagree with a discernible inﬂuence of
anthropogenic GHGs on climate are overrepresented in the
media, relative to the prevalence of these opinions in the
The Supporting Information contains background information
on the following topics: Aggregating ﬁelds of expertise,
comparison between tagged and self-declared ﬁelds of expertise,
attribution, consensus, contribution of other factors to
warming, aerosol cooling versus GHG warming, climate
sensitivity, and media exposure. This material is available free
of charge via the Internet at http://pubs.acs.org.
*Phone: +31 20 525 8271; e-mail: Verheggen.Bart@gmail.com.
Amsterdam University College AUC, PO Box 94160, 1090
GD Amsterdam, The Netherlands
Figure 8. Number of respondents per range of estimated ECS,
segregated according to the respondents’answers regarding the
quantitative GHG contribution (total sample size N= 913).
Figure 9. Self-reported frequency of media coverage, segregated according to responses to the questions on quantitative (Q1) and qualitative (Q3)
GHG contribution, as well as to the question on ECS. Responses are shown as a percentage of the number of people (N) per response category (as
denoted on the x-axis). The most frequent media coverage is reported by respondents who deemed the eﬀect of GHGs to be the smallest and ECS
to be the lowest.
Environmental Science & Technology Policy Analysis
dx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXXH
The authors declare no competing ﬁnancial interest.
We thank the following people for their contributions to this
work: Collection of email addresses: Sanne Boersma, Bä
Winkler, Rob Painting, Rob Honeycutt, Sarah Green, John
Cook, Wendy Cook, Ari Jokimä
ki, Phil Scadden, Glenn
Tamblyn, Anne-Marie Blackburn, John Hartz, Steve Brown,
George W. Morrison, Alexander C. Coulter, and many
unnamed researchers. Survey preview: Marcel Crok, Gavin
Schmidt, Gerbrand Komen, Hans Labohm, Roger Pielke Sr,
Rasmus Benestad, Sybren Drijfhout, James Annan, Mike
Hulme, Ronald Flipphi, Jan Paul van Soest, Gert Spaargaren,
Marjolein de Best-Waldhober, Tom Fuller, Ernst Schrama, Alex
Vermeulen, Iina Hellsten, Arjan Hensen, Remko Kampen, Paul
Baer. Funding: Netherlands Ministry of Infrastructure and the
(1) Center, P. R. Little Change in Opinions about Global Warming;
Pew Research Centre, Washington, D.C., 2010.
(2) Oreskes, N. Beyond the ivory tower. The scientific consensus on
climate change. Science 2004,306, 1686.
(3) Doran, P. T.; Zimmerman, M. K. Examining the scientific
consensus on climate change. EOS, Trans., Am. Geophys. Union 2009,
(4) Anderegg, W. R.; Prall, J. W.; Harold, J.; Schneider, S. H. Expert
credibility in climate change. Proc. Natl. Acad. Sci. U.S.A. 2010,107,
(5) Cook, J.; Nuccitelli, D.; Green, S. A.; Richardson, M.; Winkler, B.;
Painting, R.; Way, R.; Jacobs, P.; Skuce, A. Quantifying the consensus
on anthropogenic global warming in the scientific literature. Environ.
Res. Lett. 2013,8, 024024.
(6) Ding, D.; Maibach, E. W.; Zhao, X.; Roser-Renouf, C.;
Leiserowitz, A. Support for climate policy and societal action are
linked to perceptions about scientific agreement. Nat. Clim. Change
2011, DOI: 10.1038/nclimate1295.
(7) McCright, A. M.; Dunlap, R. E.; Xiao, C. Perceived scientific
agreement and support for government action on climate change in
the USA. Clim. Change 2013,119, 511−518.
(8) Heath, Y.; Giffort, R. Free-market ideology and environmental
degradation. The case of belief in global climate change. Environ.
(9) Kahan, D. M.; Jenkins-Smith, H.; Braman, D. Cultural cognition
of scientiﬁc consensus SSRN Electronic J. 2010
(10) Leiserowitz, A., Maibach, E., Roser-Renouf, C., Smith, N.
Climate change in the American mind: Americans’global warming
beliefs and attitudes in May 2011. In Yale Project on Climate Change
Communication; Yale University, George Mason University, 2011.
(11) Leviston, Z., and Walker, I. Second Annual Survey of Australian
Attitudes to Climate Change: Interim report CSIRO, 2011.
(12) Collins, H. M.; Evans, R. The third wave of science studies:
Studies of expertise and experience. Soc. Stud. Sci. 2002,32, 235−296.
(13) Petersen, A. Simulating Nature: A Philosophical Study of
Computer-Simulation Uncertainties and Their Role in Climate Science
and Policy Advice, 2nd ed.; CRC Press: Boca Raton, FL, 2006.
(14) IPCC The physical science basis. In Contribution of WG1 to the
Fourth Assessment Report of the IPCC; Cambridge University Press:
Cambridge, UK., 2007.
(15) IPCC Climate Change 2013: The Physical Science Basis.
Contribution of WG1 to the Fifth Assessment Report of the IPCC;
Cambridge University Press, Cambridge, UK., 2013.
(16) Malka, A., Krosnick, J. A., Debell, M., Pasek, J., Schneider, D.
Featuring skeptics in news media stories about global warming reduces
public beliefs in the seriousness of global warming (Woods Institute
for the Environment, Stanford University, Technical Paper), 2009.
(17) Prall, J. W. Most-Cited Authors on Climate Science, 2011.
(18) Bray, D. The scientific consensus of climate change revisited.
Environ. Sci. Policy 2010,13, 340−350.
(19) PBL Netherlands Environmental Assessment Agency. Climate
Science Survey, 2012. http://www.pbl.nl/sites/default/ﬁles/cms/
(20) IPCC. Guidance Note for Lead Authors of the IPCC Fifth
Assessment Report on Consistent Treatment of Uncertainties, Table
(21) Huber, M.; Knutti, R. Anthropogenic and natural warming
inferred from changes in Earth’s energy balance. Nat. Geosci. 2011,5,
(22) Jones, G. S.; Stott, P. A.; Christidis, N. Attribution of observed
historical near-surface temperature variations to anthropogenic and
natural causes using CMIP5 simulations. J. Geophys. Res.: Atmos. 2013,
(23) Wigley, T. M. L.; Santer, B. D. A probabilistic quantification of
the anthropogenic component of twentieth century global warming.
Clim. Dyn. 2012,40, 1087−1102.
(24) Bray, D., and von Storch, H. The Perspectives of Climate
Scientists on Global Climate Change. A Survey of Opinions, Germany,
(25) Lichter, S. R. Climate Scientists Agree on Warming, Disagree on
Dangers, and Don’t Trust Media’s Coverage of Climate Change; Service,
S. A., Ed.; George Mason University, 2008.
(26) Rosenberg, S.; Vedlitz, A.; Cowman, D. F.; Zahran, S. Climate
change: a profile of US climate scientists’perspectives. Clim. Change
(27) Allen, M. In defense of the traditional null hypothesis: Remarks
on the Trenberth and CurryWIREsopinion articles. IN Wiley
Interdisciplinary Reviews: Climate Change, 2011; Vol. 2, pp 931−934.
(28) Curry, J. A.; Webster, P. J. Climate science and the uncertainty
monster. Bull. Am. Meteorol. Soc. 2011,92, 1667−1682.
(29) Hegerl, G.; Stott, P.; Solomon, S.; Zwiers, F. Comment on
“Climate Science and the Uncertainty Monster”J. A. Curry and P. J.
Webster. Bull. Am. Meteorol. Soc. 2011,92, 1683−1685.
(30) Boykoﬀ,M.Who Speaks for Climate? Making Sense of Media
Reporting on Climate Change; Cambridge University Press, 2011.
Environmental Science & Technology Policy Analysis
dx.doi.org/10.1021/es501998e |Environ. Sci. Technol. XXXX, XXX, XXX−XXXI