Content uploaded by Mike Hulme
Author content
All content in this area was uploaded by Mike Hulme on Nov 10, 2015
Content may be subject to copyright.
http://ppg.sagepub.com/
Progress in Physical Geography
http://ppg.sagepub.com/content/38/4/499
The online version of this article can be found at:
DOI: 10.1177/0309133314538644
2014 38: 499 originally published online 5 June 2014Progress in Physical Geography Mike Hulme
Attributing weather extremes to 'climate change': A review
Published by:
http://www.sagepublications.com
can be found at:Progress in Physical GeographyAdditional services and information for
http://ppg.sagepub.com/cgi/alertsEmail Alerts:
http://ppg.sagepub.com/subscriptionsSubscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
http://ppg.sagepub.com/content/38/4/499.refs.htmlCitations:
What is This?
- Jun 5, 2014OnlineFirst Version of Record
- Aug 7, 2014Version of Record >>
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
Progress report
Attributing weather extremes
to ‘climate change’: A review
Mike Hulme
King’s College London, UK
Abstract
Over the last 30 years, scientific research has increasingly implicated human activities in contemporary
regional- to global-scale climatic change. Over the last decade, this research has extended to the detection of
the fingerprint of human activities on individual extreme weather events. Is it possible to say that this or that
weather extreme was ‘caused by’ human activities? Pursuing answers to this question raises many difficult
philosophical, epistemological and political issues. In this progress report, I survey the nascent science of
extreme weather event attribution by examining the field in four stages: motivations for extreme weather
attribution, methods of attribution, some example case studies and the politics of weather event attribution.
There remain outstanding political dangers and obstacles for extreme weather attribution if it is to be used, as
some claim it can and should be, for guiding climate adaptation investments, for servicing the putative loss and
damage agenda of the UN Framework Convention on Climate Change or for underpinning legal claims for
liability for damages caused by extreme weather.
Keywords
attribution, climate adaptation, climate change, extreme weather, loss and damage
I Introduction
Humans and the societies in which they live
have always sought to establish causes for their
daily weather. Many cultural beliefs and prac-
tices throughout human history have been
inspired by this desire. When weather ‘misbe-
haves’, or delivers meteorological devastation
through windstorm, torrent, blizzard, drought
or intense heat, the psychological need to attach
blame to such events becomes overwhelming.
Behringer’s (2010) cultural history of climate
offers one account of early-modern European
blame narratives of deviant weather, which
included the trickery of witches and the moral
deviancy of human behaviour. From the perspec-
tive of Pacific cultures, Donner (2007, 2011)
argues that the performance of weather has tradi-
tionally been understood as ‘the domain of the
gods’. The search for culpability for weather,
especially adverse weather (e.g. Grattan and
Brayshay, 1995), seems to be an enduring quest
across all human cultures. This is especially true
in the new and confusing world in which climate,
in a discursive sense at least, seems to be chang-
ing daily (Rudiak-Gould, 2014).
Over the last 30 years, scientific research has
increasingly implicated human activities in con-
temporary regional- to global-scale climatic
change. It is hardly surprising then that interest
has risen in the possibility of detecting the fin-
gerprint of human activities, not just on such
Corresponding author:
Department of Geography, King’s College London,
London WC2R 2LS, UK.
Email: mike.hulme@kcl.ac.uk
Progress in Physical Geography
2014, Vol. 38(4) 499–511
ªThe Author(s) 2014
Reprints and permission:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0309133314538644
ppg.sagepub.com
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
broad-scale changes in climate, but on individ-
ual extreme weather (and short-term climate)
events. Yet this possibility raises many difficult
philosophical, epistemological and political
questions. What does it mean for something to
be caused by something else – especially in
complex systems? Is the sought-after cause of
extreme weather deterministic (‘this caused
that’) or stochastic (‘this made that more
likely’)? In what ways are answers to the ques-
tion of causation shaped by political or moral
considerations rather than by scientific inquiry?
The difficulties involved in extreme weather
event attribution were very well illustrated in the
early weeksof 2014 in theUK, following a series
of tempestuous and rain-laden winter storms.
Extensive flooding ensued across southern Eng-
land and claim and counter-claim, from scientists
and politicians alike, filled broadcast and online
media (Shukman, 2014). As stated by the Inter-
governmental Panel on Climate Change (IPCC)
in their 2012 report into extreme weather events
and climate adaptation, ‘it is very difficult to
attribute an individual [weather] event to exter-
nal forcing’ (IPCC, 2012: 128).
My two previous progress reports on climate
change published in this journal were concerned
with the nature and status of the IPCC (Hulme
and Mahony, 2010) and with the possibility of
global climate engineering through strato-
spheric aerosol injection (Hulme, 2012). In this
third and final progress report of this series,
I survey the nascent science of extreme weather
event attribution. The article proceeds by exam-
ining the field in four stages: motivations for
extreme weather attribution, methods of attribu-
tion, some example case studies and the politics
of weather event attribution.
II Motivations for attribution
As many climate scientists can attest, one of the
most frequent questions asked by media jour-
nalists and other interested parties following the
latest meteorological extreme is: ‘Was this
weather event caused by climate change?’ (Stott
and Walton, 2013). I refer to this asthe ‘extreme-
weather blame’ question. The question, though
ubiquitous, is also ambiguous because it is not
clear what exactly is meant by the causal agent
‘climate change’. As explained by Fleming and
Jankovic (2011), in recent decades the meaning
of climate change in popular western discourse
has changed from being a descriptive index of
a change in climate (as in ‘evidence that a cli-
matic change has occurred’) to becoming an
independent causative agent (as in ‘climate
change caused this event to happen’). Rather than
being a descriptive outcome of a chain of causal
events affecting how weather is generated, cli-
mate change has been granted power to change
worlds: political and social worlds as much as
physical and ecological ones.
To be more precise, then, what people mean
when they ask the ‘extreme-weather blame’
question is: ‘Was this particular weather event
caused by greenhouse gases emitted from
human activities and/or by other human pertur-
bations to the environment?’ In other words, can
this meteorological event be attributed to
human agency as opposed to some other form
of agency? – for example, purely natural pro-
cesses, such as volcanic eruptions or ocean
variability, or (for some) the actions of God,
gods or spirits (Donner, 2007).
From a human psychological perspective, the
‘extreme-weather blame’ question seems a rea-
sonable one to ask. Knowing the cause of some
adversity or personal affliction is often the first
step towards living with it or overcoming it; it
helps reconcile an undesirable lived reality with
an alternative reality we might desire (Jankovic,
2006). But why would climate scientists then be
interested in developing an answer to the ques-
tion? Why have climate scientists over the last
10 years embarked upon research to provide
an answer beyond the stock phrase ‘no individ-
ual weather event can directly be attributed to
greenhouse gas emissions’? There seem to be
four possible motives.
500 Progress in Physical Geography 38(4)
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
The first is because the question piques the
scientific mind; it acts as a spur to develop new
rational understanding of physical processes
and new analytic methods for studying them.
As stated by the World Climate Research Pro-
gramme in 2013, attributing extreme weather
events is one of the six grand challenges in
climate science (cited in Peterson et al., 2013).
Most climate system modelling has been driven
by the desire to understand past and future
large-scale changes in climate parameters and
the relative influences of human and natural
influences on such changes (Edwards, 2010).
Extreme weather attribution asks a different set
of questions, however, and requires climate sys-
tem models to be reconfigured and deployed in
different ways (Hegerl and Zwiers, 2011; Stone
and Allen, 2005). The ‘extreme-weather blame’
question therefore tests scientific understanding
and the veracity of climate simulation model-
ling in new ways. As models have gained in
complexity, and especially as their animating
computers have increased in computational
power, so it has become possible to use them
to answer reductive questions about the causes
of specific extreme weather events; and, from
a (climate) science credibility perspective,
being able to offer even tentative answers to
such public-interest questions is better than
offering no answer at all.
A second argument, put forward by some, is
that it is important to know whether or not spe-
cific instances of extreme weather are human-
caused in order to improve the justification,
planning and execution of climate adaptation.
For example, Stott et al. (2013) in their over-
view paper of weather attribution science for the
World Climate Research Programme argued
that failure to accurately attribute extreme
weather to human causes could lead to poor
adaptation decisions (see also Allen, 2011; Stott
and Walton, 2013). Conversely, Pall et al.
(2011: 385) contend that the ability to quantify
the contribution of human-based emissions to
the risk of damaging weather events ‘could
prove a useful tool for evidence-based climate
change adaptation policy’. As another apologist
for extreme weather attribution expressed it,
such knowledge is simply part of ‘evidence-
based planning and informed decision-making’
and hence self-evidently necessary (Washington,
2014). If climate adaptation is understood
through the paradigm of optimization, then such
claims seem logical. However, as others have
argued, there are alternative ways of thinking
about adaptation practices (Barnett and O’Neill,
2010; Dessai and Hulme, 2004; Dessai et al.,
2009; Weaver et al., 2013).
A third argument for pursuing an answer to the
‘extreme-weather blame’ question is inspired by
the possibility of pursuing legal liability for dam-
ages caused. An early example of this case was
laid out by Allen (2003) in the context of flood
damage caused by a meteorological extreme:
‘If insurance premiums rise as insurers factor in
the increased risk of flooding due to climate
change, and house prices consequently fall, some
of this loss can straightforwardly be blamed on
past greenhouse-gas emissions’ (p. 891). This
argument then drives the desire to calculate what
fractional increase in probability of an extreme
weather event can be attributed to a specific
cause, in this case to greenhouse gas emissions
(Allen et al., 2007). This motivation connects
with wider arguments about litigation for climate
change damages (Grossman, 2003; Masson,
2010). If specific loss and damage from extreme
weather can be attributed to greenhouse gas
emissions – even if expressed in terms of
increasedrisk rather than deterministically – then
lawyers might get interested (Adam, 2011).
A counter-argument to this conjecture has been
expressed by others; the diversity of attribution
methods and the questionable credibility of cli-
mate models used for extreme weather attribu-
tion suggests to some that such evidence
would unlikely hold sway in court cases (Stott and
Walton, 2013: 278; see also Anonymous, 2012).
The liability motivation for research into
weather event attribution also bisects the new
Hulme 501
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
international political agenda of ‘loss and dam-
age’ which has emerged in the last two years.
Although the notion of loss and damage due to
climate change has been on the political agenda
for some years (Warner and Zakieldeen, 2012),
it was first given a formal place in the ongoing
UN Framework Convention on Climate Change
(FCCC) negotiations at the 18th Conference of
theParties(COP18)heldinDohain2012,and
then carried forward at COP19 in Warsaw in
2013. The basic idea is to legitimate loss and
damage caused by climate change as grounds for
less-developed countries to gain access to new
international climate adaptation funds. Although
the details of this arrangement are a long way
from being agreed, including resolution of the
definitional questions about what constitutes ‘cli-
mate change’ and what constitutes ‘loss and
damage’ (Warner et al., 2012), there is an inter-
section here with the putative claim from climate
scientists to be able to attribute specific extreme
weather events to greenhouse gas emissions.
A final reason for scientists to be investing in
this area of climate science – a reason stated
explicitly less often than the ones above and yet
one which underlies much of the public interest in
the ‘extreme-weather blame’ question – is frus-
tration with, and argument about, the invisibility
of climate change. Rudiak-Gould (2013) explains
the problems involved in making climate change
visible – and by implication ‘real’ – to assorted
groups in society. He quotes the American Psy-
chological Association’s Taskforce on the Inter-
face between Psychology and Climate Change:
Because climate change is so hard to detect from per-
sonal experience, it makes sense to leave this task to cli-
mate scientists. This makes climate change a
phenomenon where people have to rely on scientific
models and expert judgment, and/or on reports in the
mass media, and where their own personal experience
does not provide a trustworthy way to confirm the
reports. (Swim et al., 2009: 22)
If this is believed to be true – that only scientists
can make climate change visible and real – then
there is extra onus on scientists to answer the
‘extreme-weather blame’ question as part of
an effort to convince citizens of the reality
of human-caused climate change. One danger
of this motivation is that it may skew research
towards those weather extremes that may seem
a priori to be more attributable to human
agency. The issues involved here, including the
complex human psychology of beliefs and per-
ceptions, are nicely summarized by Doyle
(2011), Kerr (2013) and Rudiak-Gould (2013).
So there are mixed and multiple motives at
work in the community of scientists who have
engaged in and promoted the nascent science
of weather attribution. In the next section, I sur-
vey some of the reasoning and analytical meth-
ods underpinning weather extreme attribution.
III Attribution methods
There is a long history of detection and attribu-
tion studies in the science of climate change
and this field is well reviewed by Allen et al.
(2007), Bindoff and Stott (2013), Hegerl and
Zwiers (2011) and Stott et al. (2010). Typi-
cally, these studies seek, first, to detect a
change in the spatio-temporal mean monthly
or seasonal statistics of some climatic variable
on global, continental or regional scales and,
then, to attribute such a change to a specific
causal factor (e.g. volcanic eruption, elevated
greenhouse gas concentration) with the aid of
one or more climate simulation models. Such
studies underpin the successive statements on
climate change attribution that have been made
by the IPCC in their 1st to 5th Assessment
Reports between 1990 and 2013.
Attributing extreme weather events to human
influences requires different approaches, how-
ever, of which four broad categories can be
identified. The first and most general approach
to attributing extreme weather phenomena to
rising greenhouse gas concentrations is to use
simple physical reasoning. For example, ther-
modynamic arguments would suggest that more
502 Progress in Physical Geography 38(4)
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
intense precipitation events are to be expected
in an atmosphere which holds more water
vapour (O’Gorman and Schneider, 2009); an
upward shift in mean monthly temperature
would lead to a disproportionate increase in the
frequency of extreme hot daily temperatures
(Trenberth, 2011). The danger of this approach
is that general physical reasoning is either too
vague to be of value or may turn out to be
wrong. For example, it used to be argued that
under conditions of ocean warming tropical
cyclones would become more frequent as the
area of ocean above the trigger temperature
expanded. However, better understanding of the
dynamics of hurricane formation now suggests
that frequencies seem likely to decrease even
though average intensity might increase (Knut-
son et al., 2013).
General physical reasoning can only lead to
broad qualitative statements such as ‘this
extreme weather is consistent with’ what is
known about the human-enhanced greenhouse
effect. Such statements offer neither determinis-
tic nor stochastic answers and clearly underde-
termine the ‘extreme-weather blame’ question.
It has given rise to a number of analogies to try
to communicate the non-deterministic nature of
extreme event attribution. The three most
widely used ones concern a loaded die (the
chance of rolling a ‘6’ has increased, but no sin-
gle ‘6’ can be attributed to the biased die), the
baseball player on steroids (the number of home
runs hit increases, but no single home run can be
attributed to the steroids) and the speeding car-
driver (the chance of an accident increases in
dangerous conditions, but no specific accident
can be attributed to the fast driving) (Peterson
et al., 2013: S64).
A second approach is to use classical statisti-
cal analysis of meteorological time series data
to determine whether a particular weather (or
climatic) extreme falls outside the range of what
a ‘normal’ unperturbed climate might have
delivered. Such analysis estimates the likeli-
hood of a specific observed meteorological
extreme occurring given no external human for-
cing. This was the approach used in the study of
the 2003 European heatwave by Luterbacher
et al. (2004: 1503) in which they concluded
that ‘the late 20th- and early 21st-century
[European] warmth very likely exceeds that
of any time during at least the past 500 years’ and
that 2003 was by far the hottest summer in this
period. The difficulties of such an approach are
revealed in a later study of the 2003 heatwave
by Charpentier (2011) and the subsequent com-
mentary by Stott et al. (2011). All such extreme
event analyses of meteorological time series
are at best able to detect outliers, but can never
be decisive about possible cause(s). A different
time series approach therefore combines observa-
tional data with model simulations and seeks to
determine whether trends in extreme weather pre-
dicted by climate models have been observed in
meteorological statistics (e.g. Min et al., 2011, for
precipitation extremes; Zwiers et al., 2011, for
temperature extremes). This approach is able to
attribute statistically a trend in extreme weather
to human influence, but not a specific weather
event. Again, the ‘extreme-weather blame’ ques-
tion remains underdetermined.
To move beyond both qualitative physical
reasoning and the limitations of observed trend
analysis, a third method has been developed in
recent years. Originally proposed by Allen
(2003), and first applied to an extreme climatic
event by Stott et al. (2004), the approach was
outlined theoretically in Stone and Allen
(2005). Taking inspiration from the field of epi-
demiology, this method seeks to establish the
Fractional Attributable Risk (FAR) of an
extreme weather (or short-term climate) event.
It asks the counterfactual question, ‘How might
the risk of a weather event be different in the
presence of a specific causal agent in the climate
system?’
The single observational record available to
us, and which is analysed in the statistical meth-
ods described above, is inadequate for this
task. The solution is to use multiple model-
Hulme 503
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
simulations of the climate system, first of all
without the forcing agent(s) accused of ‘caus-
ing’ the weather event and then again with
that external forcing introduced into the model.
If P
0
is the probability of the specified weather
extreme in the unforced (simulated) climate and
P
1
the probability of the same event in the
forced (simulated) climate, then FAR is simply
1–P
0
/P
1
. FAR may vary between 0 and 1 and
expresses the fraction of a weather risk (e.g. a
rainfall intensity above a given threshold) that
can be attributed to a specified influence. The
credibility of this method of weather attribution
can be no greater than the overall credibility of
the climate model(s) used – and may be less,
depending on the ability of the model in ques-
tion to simulate accurately the precise weather
event under consideration at a given scale (e.g.
a heatwave in continental Europe, a rain event
in northern Thailand) (see Christidis et al.,
2013b).
A fourth, more philosophical, approach to
weather event attribution should also be men-
tioned. This is the argument that since human
influences on the climate system as a whole are
now clearly established – through changing
atmospheric composition, altered land surface
characteristics and so on – there can no longer
be such a thing as a purely natural weather
event. All weather – whether it be a raging tem-
pest or a still summer afternoon – is now attribu-
table to human influence, at least to some
extent. Weather is the local and momentary
expression of a complex system whose func-
tioning as a system is now different to what it
would otherwise have been had humans not
been active. This was an argument made origi-
nally by Bill McKibbin in his book The End of
Nature (McKibbin, 1990), where he lamented
that a child born today would ‘never know a nat-
ural summer ... Summer is becoming extinct,
replaced by something else which will be called
‘‘summer’’’ (p. 55). This argument was also
alluded to by Hulme (2000) in his commentary
on the consequences of flooding in Mozambique
in March of that year and resonates with the
much larger literature about the hybrid nature
of all of today’s ecosystems (e.g. Marris, 2011).
This position is also adopted by the climate
scientist Kevin Trenberth in some of his work.
Trenberth’s argument is that in weather and cli-
mate attribution studies analytical techniques
should be called on to refute the null hypothesis
of ‘humans have influenced this weather
extreme’ rather than one of ‘there is no human
influence on this weather extreme’ (Trenberth,
2011). As he says:
All storms develop in [a] changed environment. How-
ever, most of the time, the resulting weather is within
the realm of previous experience. Yet all storms are dif-
ferent than they would have been ... the question
should not be is there a human component; but what
is [the human component]?’ (Trenberth, 2011: 929)
The counter-argument to this stance from an ana-
lytical position is developed by Allen (2011) and
Curry (2011), although they do not take on the
philosophical dimension of the argument.
IV Some example cases
of extreme weather attribution
Table 1 shows a list of weather attribution stud-
ies drawn largely from the two collections pub-
lished by Peterson et al. (2012, 2013). The first
application of the FAR method for attributing a
specific climatic extreme to human influence
was conducted by Stott et al. (2004) in their
study of the European summer heatwave of
2003. They concluded that ‘human influence
has at least doubled the risk of a heatwave
exceeding this threshold magnitude’ (p. 610).
This was consistent with results from the Luter-
bacher et al. (2004) and Scha¨r et al. (2004) stud-
ies which used statistical methods alone.
However, in the case of other weather extremes,
multiple attribution studies of the same event do
not necessarily lead to convergent answers.
Dole et al. (2011) used a FAR methodology
to investigate the exceptional heat over western
504 Progress in Physical Geography 38(4)
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
Russia during July 2010 and concluded that it
was ‘due mainly’ to natural internal atmo-
spheric variability rather than to human-
caused climate forcings. In contrast, Rahmstorf
and Coumou (2011), using a statistical metho-
dology, concluded that there was a probability
of only 20%that the 2010 July heat record
would have occurred naturally without the
observed regional warming since 1980, which
they largely attributed to anthropogenic green-
house gas forcing. A third attribution study of
this Russian heatwave then sought to under-
stand the reason for these apparently contradic-
tory findings (Otto et al., 2012). Also using a
FAR methodology with large ensemble simula-
tions, these authors concluded that both the
above-mentioned studies were correct, but were
answering different questions: Dole et al. (2011)
whether the magnitude of the extreme heat
could be attributed and Rahmstorf and Coumou
(2011) whether its probability of occurrence
could be attributed. This demonstrates the
importance when designing weather attribution
studies of being clear about which precise ques-
tion is being addressed and also being clear in
any subsequent communication of the findings.
The weather attribution studies listed in
Table 1 adopt a number of different methods,
but it will not always be possible for attribution
studies to be applied to all or to any specific
extreme weather event or short-term climate
anomaly. For statistical methods to be effective,
long homogenous meteorological time series
data are required; for some regions of the world
these will not be available through instrumental
measurements alone nor through combined
series of instrumental and proxy measurements.
For modelling attribution studies – whether
combined modelling and observation studies
or FAR studies based on multi-model ensem-
bles – there are different conditions and con-
straints. Not only do the necessary model
Table 1. Some examples of weather and short-term climate extremes to which different analytic weather
attribution methods have been applied and results published.
Date
of extreme
event Location
Nature of
extreme event Source Method
2000 autumn UK Floods Pall et al. (2011) FAR
2003 summer Western Europe Heatwave Stott et al. (2004) FAR
2010 summer Western Russia Heatwave Dole et al. (2011)
Rahmstorf and Coumou (2011)
Christidis et al. (2013b)
FAR
Statistical
FAR
2010 summer Pakistan Heavy rains Christidis et al. (2013b) FAR
2010/11winter UK Cold Christidis and Stott (2012) FAR
2011December New Zealand 48-hour rainfall Dean et al. (2013) Models and
observations
2011 summer Texas Drought Rupp et al. (2012) Modelling
2011 summer Thailand Heavy rains Van Oldenborgh et al. (2012) Statistical
2011/12 winter Iberia Drought Trigo et al. (2013) Modelling
2012 spring East Africa Drought Funk et al. (2013) FAR
2012 March Eastern Australia Heavy rains Christidis et al. (2013a) FAR
2012 October USA Coastal inundation Sweet et al. (2013) Statistical
2012 July Southwest Japan Heavy rains Imada et al. (2013) FAR
2013 summer Australia Heatwave Lewis and Karoly (2013) FAR
Hulme 505
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
simulation experiments need to be conducted
(constrained by the high computational demands
they entail), but it needs to be demonstrated that
the models used can adequately represent, at the
appropriate spatial scale, the characteristics of
the weather extreme under study.
In their FAR study of three high-impact
events – the 2009/2010 cold winter in the UK,
the heatwave in Moscow in July 2010 and the
floods in Pakistan in July 2010 – Christidis
et al. (2013b) concluded that no reliable attribu-
tion of the intense-rainfall events in Pakistan
could be made since the model’s simulation of
intense rainfall in that region was inadequate.
This was in contrast to the temperature extremes
in the UK (winter cold) and in western Russian
(summer heat) where the model performed bet-
ter. It is likely that attribution of temperature-
related extremes using FAR methods will always
be more attainable than for other meteorological
extremes such as rainfall and wind, which cli-
mate models generally find harder to simulate
faithfully at the spatial scales involved. As dis-
cussed below, this limitation on which weather
events and in which regions attribution studies
can be conducted will place important con-
straints on any operational extreme weather attri-
bution system.
V Political dimensions of weather
attribution
The new science of extreme weather attribution
is developing in the context of politically
charged debates about the causes and conse-
quences of climate change and about appropri-
ate responses. How weather event attribution
science is framed, who funds the research, and
for what reason, have unavoidable political
dimensions. For example, whichever of the four
motivations mentioned above is used to justify
investment in this science – scientific curiou-
sity, adaptation guidance, liability for damages,
‘visualizing’ climate change – influences differ-
ent potential research funders and also how
scientific outcomes might subsequently be
interpreted and used. As observed by Nature’s
editorial in 2012, ‘designers of [such weather
attribution] services must think very clearly
about how others might want to use the knowl-
edge that climate scientists produce’ (Anon-
ymous, 2012: 336).
Consider the case where weather attribution
science is framed as a contribution to ‘guiding
adaptation decisions’. Some argue that there is
an urgent need to develop this science to assist
with decisions about the allocation of new inter-
national adaptation funds. For example, Myles
Allen claims that:
because [adaptation] money is on the table, it is sud-
denly going to be in everyone’s interest to be a victim
of climate change ... so we need urgently to develop
the science base to be able to distinguish genuine
impacts of climate change from unfortunate conse-
quences of bad weather. (Gillis, 2011)
This is a view echoed by others: ‘Quantifying
the impacts of anthropogenic climate change
in this way is ... important in guiding the allo-
cation of resources available for adaptation’
(Hoegh-Guldberg et al., 2011: 72).
Hulme et al. (2011) show why such ambi-
tious claims are unlikely to be realized, how-
ever. Investment in climate adaptation, they
claim, is most needed ‘where vulnerability to
meteorological hazard is high, not where
meteorological hazards are most attributable
to human influence’ (p. 765). Extreme weather
attribution says nothing about how damages
are attributable to meteorological hazard as
opposed to exposure to risk; it says nothing
about the complex political, social and eco-
nomic structures which mediate physical
hazards. Also, separating weather into two
categories – ‘human-caused’ weather and
‘tough-luck’ weather – raises practical and
ethical concerns about any subsequent invest-
ment allocation guidelines which excluded the
victims of ‘tough-luck weather’ from benefit-
ing from adaptation funds.
506 Progress in Physical Geography 38(4)
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
Contrary to the claims of some weather attri-
bution scientists, the loss and damage agenda
of the United Nations Framework Convention
on Climate Change (UNFCCC), as it is cur-
rently emerging, makes no distinction between
‘human-caused’ and ‘tough-luck’ weather.
‘Loss and damage impacts fall along a conti-
nuum, ranging from ‘‘events’’ associated with
variability around current climatic norms (e.g.,
weather-related natural hazards) to [slow-onset]
‘‘processes’’ associated with future anticipated
changes in climatic norms’ (Warner et al.,
2012: 21). Although definitions and protocols
have not yet been formally ratified, it seems
unlikely that there will be a role for the sort of
forensic science being offered by extreme
weather attribution science.
So consider the Thailand floods of 2011
which caused about US$48 billion worth of
damage. The weather attribution study of Van
Oldenborgh et al. (2012) concluded that the
amount of rain falling was not unusual and
could not therefore be attributed to human
agency. One can see why asking – and then
answering – the ‘extreme-weather blame’ ques-
tion with regard to the heavy precursor rains
which triggered the high river flows through
Bangkok is of some interest, but it is unlikely
that van Oldenborgh’s conclusion would alter
the rhetorical or practical use of such events as
justification for funding improvements in flood
defence or enhancing adaptive capacity. Poli-
cies to address loss and damage are far more
likely to be influenced by the prospects for
reducing vulnerability and increasing coping
capacities than by whether or not the meteorolo-
gical component of a disaster can be attributed
to human agency (Klein and Mo¨hner, 2011).
Then there is the obvious question about
which extreme weather events should be inves-
tigated: the cases where human influence on
meteorological extremes is easiest to detect or
the cases where the political, economic or ethi-
cal consequences of extreme weather attribution
are greatest (or perhaps where they are least!)?
The list in Table 1 is indicative of the range of
weather/climate events and their locations that
have currently been studied. Yet, as Peterson
et al. (2013) recognize, there are strong biases
at work in case selection. These include
researcher interest and subjective estimation of
which weather extremes might be tractable:
‘We also see a natural [sic] bias towards scien-
tists addressing local events ... impacting
themselves, their friends and neighbours’
(Peterson et al., 2013: S64). Choices are also
constrained by the (un)availability of long time
series of observed data, either for statistical
attribution methods or for verifying the credibil-
ity of model-based approaches.
The choice of which potential causal agent(s)
are to be implicated in the attribution study also
has political significance. Although nearly all
cases investigated thus far have sought to isolate
the effect on extreme weather of elevated car-
bon dioxide (or aggregated greenhouse gas)
concentrations, the logic of weather event attri-
bution opens up a wider range of possible causal
agents. For example, if such work is motivated
by the desire to support liability claims for dam-
age caused, then distinguishing between FAR
due to greenhouse gas emissions originating
from fossil fuels versus those originating from
land-use change becomes important. Different
political actors, institutional entities and social
practices lie behind different emissions sources
and so liability would be distributed differently.
There is also the possibility of even finer causal
attribution: distributing the fractional risk of an
extreme and damaging weather event between
fossil fuel, land use, black carbon and sulphate
aerosol emissions is in principle possible using
this methodology. The political significance of
these choices is great and likely to be contested
if such attribution claims reach parliaments or
courts.
A final example of the political dimensions
of this nascent science concerns the possibility
of future ventures to intentionally manipulate
the global climate through solar engineering
Hulme 507
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
(Hulme, 2014; Keith, 2013). Should planetary-
scale injection of sulphate aerosols into the
stratosphere ever take place, then weather attri-
bution capabilities may well be called upon to
adjudicate rival political claims about the cause
of subsequent hazardous meteorological events.
For example, interested parties might ask the
question: ‘Was this devastating typhoon caused
by artificially injected aerosols, or by elevated
concentrations of greenhouse gases, or was it a
naturally occurring tropical storm?’ However,
as argued by Sarewitz (2004) more generally
about the role of science in environmental
controversies, deploying extreme weather attri-
bution science in such circumstances might sim-
ply further aggravate the political and ethical
conflicts that aerosol injection technology
will have unleashed. The historical analogy of
seeking definitive attribution of local rainfall
to precursor cloud-seeding interventions is not
propitious (see Fleming, 2010).
VI What future for extreme
weather attribution?
A series of international workshops organized
by the informal Attribution of Climate-Related
Events (ACE) group (Met Office, 2013), start-
ing in 2009, have already led to two coordinated
‘annual reports’ on weather attribution being
published in the Bulletin of the American
Meteorological Society (Peterson et al., 2012,
2013). There are plans for this to be an ongoing
annual report series complementing the annual
State of the Climate report from the World
Meteorological Organization (WMO). How-
ever, as the science of extreme weather attribu-
tion moves from proof-of-concept to a more
developmental stage, outlines for how an opera-
tional international extreme weather attribution
service might be designed have already been
proposed. In their review for the World Climate
Research Programme, Stott et al. (2013) suggest
how an international operational weather attri-
bution service might be incorporated into the
WMO’s new Global Framework for Climate
Services initiative (Hewitt et al., 2012). In
the UK, a system for configuring the Hadley
Centre climate model to be used in more routine
attribution of extreme weather and climate events
has been demonstrated (Christidis et al., 2013b).
However, questions remain (Schiermeier,
2011). Who would pay for such a real-time attri-
bution service? The computational resource
required for operational FAR analysis of extreme
weather using multiple ensemble modelling
methods is intimidating. Who exactly would be
the stakeholders to champion such a service?
As reported by Nature following the 2012 ACE
workshop, ‘none of the industry and government
experts at the ACE workshop could think of any
concrete example in which an attribution might
inform business or political decision-making’
(Anonymous, 2012: 336).
The idea of extreme weather attribution is of
undoubted scientific interest. It can be used to
drive climate model development and can be
linked to improvement in, for example, seasonal
forecasting capabilities (Stott et al., 2013). It
may also have some limited potential public
value in sharpening pronouncements from
scientists in answer to the ‘extreme-weather
blame’ question, avoiding overly glib or over-
precise answers. However, there remain out-
standing political dangers and obstacles for
extreme weather attribution if it is to be used for
guiding climate adaptation investments, for ser-
vicing the putative loss and damage agenda of
the UNFCCC or for underpinning legal claims
for liability for damages caused by extreme
weather. Climate adaptation is not about opti-
mizing society to withstand attributable physi-
cal hazards, but an exercise in hedging against
a wide variety of unknown and poorly known
future events. Any investments made under the
UNFCCC’s loss and damage agenda are more
likely to be driven by the politics of vulnerabil-
ity than by the sort of forensic science which
weather attribution scientists are offering. And
the courts are a long way from accepting FAR
508 Progress in Physical Geography 38(4)
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
evidence as the basis for awarding damages to
plaintiffs. Meteorological hazards are always
mediated through complex political, social and
economic structures; establishing liability for
meteorological hazard falls a long way short
of establishing liability for subsequent risk to
society.
Acknowledgements
The author acknowledges the remarks of James
Screen and Martin Mahony who commented on an
early draft of this review.
References
Adam D (2011) Climate change in court. Nature Climate
Change 1(3): 127–130.
Allen MR (2003) Liability for climate change. Nature 421:
891–892.
Allen MR (2011) In defense of the traditional null
hypothesis: Remarks on the Trenberth and Curry
WIREs opinion articles. WIREs Climate Change 2(5):
767–766.
Allen MR, Pall P, Stone DA, et al. (2007) Scientific
challenges in the attribution of harm to human influ-
ence on climate. University of Pennsylvania Law
Review 155(6): 1353–1400.
Anonymous (2012) Extreme weather. Nature489: 335–336.
Barnett J and O’Neill SJ (2010) Maladaptation. Global
Environmental Change 20: 211–213.
Behringer W (2010) A Cultural History of Climate
(translated from German). Cambridge: Polity Press.
Bindoff NL and Stott PA (2013) Detection and attribution of
climate change: From global to regional. In: StockerTF,
Qin D, Plattner G-K, Tignor M, Allen SK, Boschung J,
Nauels A, Xia Y, Bex V and Midgley PM (eds) Climate
Change 2013: The Physical Science Basis – Contribu-
tion of Working Group I to the Fifth Assessment Report
of the Intergovernmental Panel on Climate Change.
Cambridge: Cambridge University Press, 867–952.
Charpentier A (2011) On the return period of the 2003
heatwave. Climatic Change 109(3/4): 245–260.
Christidis N and Stott PA (2012) Lengthened odds of the
cold UK winter of 2010/11 attributable to human
influence. Bulletin of the American Meteorological
Society 93(7): 1060–1062.
Christidis N, Stott PA, Karoly DJ, et al. (2013a) An attri-
bution study of the heavy rainfall over eastern Australia
in March 2012. Bulletin of the American Meteor-
ological Society 94(9): S58–S61.
Christidis N, Stott PA, Scaife AA, et al. (2013b) A new
HadGEM3-A-based system for attribution of weather-
and climate-related extreme events. Journal of Climate
26: 2756–2783.
Curry J (2011) Nullifying the climate null hypothesis.
WIREs Climate Change 2(6): 919–924.
Dean SM, Rosier S, Carey-Smith T, et al. (2013) The role
of climate change in the two-day extreme rainfall in
Golden Bay, New Zealand, December 2011. Bulletin
of the American Meteorological Society 94(9):
S61–S63.
Dessai S and Hulme M (2004) Does climate adaptation
policy need probabilities? Climate Policy 4: 107–128.
Dessai S, Hulme M, Lempert R, et al. (2009) Do we need
better predictions to adapt to a changing climate? EOS
90(13): 111–112.
Dole R, Hoerling M, Perlwitz J, et al. (2011) Was there a
basis for anticipating the 2010 Russian heat wave?
Geophysical Research Letters 38: L06702.
Donner SD (2007) Domain of the Gods: An editorial essay.
Climatic Change 85: 231–236.
Donner SD (2011) Making the climate a part of the human
world. Bulletin of the American Meteorological Society
92(10): 1297–1302.
Doyle J (2011) Mediating Climate Change. Farnham:
Ashgate.
Edwards PN (2010) A Vast Machine: Computer Models
Climate Data and the Politics of Global Warming.
Cambridge, MA: The MIT Press.
Fleming JR (2010) Fixing the Sky: The Checkered History
of Weather and Climate Control. New York: Columbia
University Press.
Fleming JR and Jankovic V (2011) Revisiting Klima.
Osiris 26(1): 1–6.
Funk C, Husak G, Michaelsen J, et al. (2013) Attribution
of 2012 and 2003–12 rainfall deficits in eastern
Kenya and southern Somalia. Bulletin of the Amer-
ican Meteorological Society 94(9): S45–S48.
Gillis J (2011) Study links rise in rain and snow to human
actions. New York Times 16 February. Available at:
http://www.ndtv.com/article/world/study-links-rise-in-
rain-and-snow-to-human-actions-86118.
Grattan J and Brayshay M (1995) An amazing and por-
tentous summer: Environmental and social responses in
Britain to the 1783 eruption of an Iceland volcano. The
Geographical Journal 161: 125–134.
Hulme 509
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
Grossman DA (2003) Warming up to a not-so-radical idea:
Tort-based climate change litigation. Columbian Jour-
nal of Environmental Law 28: 1–61.
Hegerl G and Zwiers F (2011) Use of models in the
detection and attribution of climate change. WIREs
Climate Change 2: 570–591.
Hewitt C, Mason S and Walland D (2012) The global
framework for climate services. Nature Climate Change
2(12): 831–832.
Hoegh-Guldberg O, Hegeri G, Root T, et al. (2011) Dif-
ficult but not impossible. Nature Climate Change 1: 72.
Hulme M (2000) There is no such thing as a purely natural
weather event. The Guardian 14 March. Reproduced
in: Hulme M (2013), Exploring Climate Change
Through Science and in Society: An Anthology of Mike
Hulme’s Essays, Interviews and Speeches. Abingdon:
Routledge, 45–46.
Hulme M (2012) Climate change: Climate engineering
through stratospheric aerosol injection. Progress in
Physical Geography 36(5): 694–705.
Hulme M (2014) Can Science Fix Climate Change? A Case
Against Climate Engineering. Cambridge: Polity Press.
Hulme M and Mahony M (2010) Climate change: What do
we know about the IPCC? Progress in Physical Geo-
graphy 34(5): 705–718.
Hulme M, O’Neill SJ and Dessai S (2011) Is weather event
attribution necessary for adaptation funding? Science
334: 764–765.
Imada Y, Watanabe M, Mori M, et al. (2013) Contribution
of atmospheric circulation change to the 2012 heavy
rainfall in southwestern Japan. Bulletin of the American
Meteorological Society 94(9): S52–S54.
Intergovernmental Panel on Climate Change (IPCC) (2012)
Managing the Risks of Extreme Events and Disasters
to Advance Climate Change Adaptation.Cambridge:
Cambridge University Press.
Jankovic V (2006) Change in the weather. Bookforum
February/March: 39–40.
Keith D (2013) A Case for Climate Engineering.
Cambridge, MA: The MIT Press.
Kerr R (2013) In the hot seat. Science 342: 688–689.
Klein RJT and Mo¨hner A (2011) The political dimension
of vulnerability: Implications for the Green Climate
Fund. IDS Bulletin 42: 15–22.
Knutson TR, Sirutis JJ, Vecchi GA, et al. (2013) Dyna-
mical downscaling projections of late 21st century
Atlantic hurricane activity CMIP3 and CMIP5 model-
based scenarios. Journal of Climate 26: 6591–6617.
Lewis SC and Karoly DJ (2013) Anthropogenic con-
tributions to Australia’s record summer tempera-
tures of 2013. Geophysical Research Letters 40(14):
3705–3709.
Luterbacher J, Dietrich D, Xoplaki E, et al. (2004) Eur-
opean seasonal and annual temperature variability
trends and extremes since 1500. Science 303:
1499–1503.
McKibbin B (1990) The End of Nature. London: Viking
Books.
Marris E (2011) Rumbunctious Garden: Saving Nature in
a Post-Wild World. London: Bloomsbury.
Masson J-F (ed.) (2010) Climate Change Litigation and
Law. Hauppauge, NY: Nova Science Publishers.
Met Office (2013) Attribution of climate-related events
(ACE). Available at: http://www.metoffice.gov.uk/
research/climate/climate-monitoring/attribution/ace.
Min S-K, Zhang X, Zwiers FW, et al. (2011) Human
contribution to more intense precipitation extremes.
Nature 470: 378–381.
O’Gorman PA and Schneider T (2009) The physical basis
for increases in precipitation extremes in simulations of
21st century climate change. Proceedings of the
National Academy of Sciences 106: 14773–14777.
Otto FEL, Massey N, van Oldenborgh, et al. (2012)
Reconciling two approaches to attribution of the 2010
Russian heat wave. Geophysical Research Letters 39:
L04702.
Pall P, Aina T, Stone DA, et al. (2011) Anthropogenic
greenhouse gas contribution to flood risk in England
and Wales in autumn 2000. Nature 470: 382–385.
Peterson TC, Hoerling MP, Stott PA, et al. (eds) (2013)
Explaining extreme events of 2012 from a climate
perspective. Bulletin of the American Meteorological
Society 94(9): S1–S74.
Peterson TC, Stott PA and Herring S (eds) (2012)
Explaining extreme events of 2011 from a climate
perspective. Bulletin of the American Meteorological
Society 93(7): 1041–1067.
Rahmstorf S and Coumou D (2011) Increase of extreme
events in a warming world. Proceedings of the National
Academy of Sciences 108: 17905–17909.
Rudiak-Gould P (2013) ‘We have seen it with our own
eyes’: Why we disagree about climate change visi-
bility. Weather Climate and Society 5(2): 120–132.
Rudiak-Gould P (2014) Climate change and accusation:
Global warming and local blame in a small island state.
Current Anthropology, forthcoming.
510 Progress in Physical Geography 38(4)
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
Rupp DE, Mote PW, Massey M, et al. (2012) Did human
influence on climate make the 2011 Texas drought
more probable? Bulletin of the American Meteor-
ological Society 93(7): 1052–1054.
Sarewitz D (2004) How science makes environmental con-
troversies worse. Environmental Science and Policy 7:
385–403.
Scha¨r C, Vidale PL, Lu
¨thi D, et al. (2004) The role of
increasing temperature variability in European summer
heat waves. Nature 427: 332–336.
Schiermeier Q (2011) Climate and weather: Extreme
measures. Nature 477: 148–149.
Shukman D (2014) Barrage over climate change link to
floods. BBC News Online 18 February. Available
at: http://www.bbc.co.uk/news/science-environment-
26242253.
Stone DA and Allen MR (2005) The end-to-end attribution
problem: From emissions to impacts. Climatic Change
71: 303–318.
Stott PA and Walton P (2013) Attribution of climate-
related events: Understanding stakeholder needs.
Weather 68(10): 274–279.
Stott PA, Allen M, Christidis N, et al. (2013) Attribution of
weather and climate-related events. In: Hurrell JW and
Asrar GR (eds) Climate Science for Serving Society:
Research Modeling and Prediction Priorities. Dor-
drecht: Springer, 307–337.
Stott PA, Christidis N and Betts RA (2011) Changing
return periods of weather-related impacts: The attribu-
tion challenge. Climatic Change 109(3/4): 263–268.
Stott PA, Gillett NP, Hegerl GC, et al. (2010) Detection
and attribution of climate change: A regional perspec-
tive. WIREs Climate Change 1: 192–211.
Stott PA, Stone DA and Allen MR (2004) Human contri-
bution to the European heatwave of 2003. Nature 432:
610–614.
Sweet W, Zervas C, Gill S, et al. (2013) Hurricane Sandy
inundation probabilities today and tomorrow. Bulletin of
the American Meteorological Society 94(9): S17–S20.
Swim J, Clayton S, Doherty T, et al. (2009) Psychology
and global climate change: Addressing a multifaceted
phenomenon and set of challenges. Available at: http://
www.apa.org/science/about/publications/climate-change-
booklet.pdf.
Trenberth KE (2011) Attribution of climate variations and
trends to human influences and natural variability.
WIREs Climate Change 2(6): 925–930.
Trigo RM, An
˜el J, Barriopedro D, et al. (2013) The record
winter drought of 2011–12 in the Iberian peninsula.
Bulletin of the American Meteorological Society 94(9):
S41–S45.
Van Oldenborgh G-J, van Urk A and Allen M (2012) The
absence of a role of climate change in the 2011 Thai-
land floods. Bulletin of the American Meteorological
Society 93(7): 1047–1049.
Warner K and Zakieldeen SA (2012) Loss and damage due
to climate change: An overview of the UNFCCC
negotiations. Available at: http://www.oxfordclimate-
policy.org/publications/documents/LossandDamage.pdf.
Warner K, Van der Geest K, Kreft S, et al. (2012) Evidence
from the front lines of climate change: Loss and damage
to communities despite coping and adaptation. United
Nations University Policy Report No 9. Available at:
http://unu.edu/publications/policy-briefs/evidence-from-
the-frontlines-of-climate-change-loss-and-damage-to-
communities-despite-coping-and-adaptation.html.
Washington R (2014) 1852 or 2014? Keble Geography, 10
February. Available at: http://www.keble-oxford-geo-
graphy.info/?p¼1335.
Weaver CP, Lempert RJ, Brown C, et al. (2013) Improving
the contribution of climate model information to
decision-making: The value and demands of robust
decision frameworks. WIREs Climate Change 4(1):
39–60.
Zwiers F, Zhang X and Feng Y (2011) Anthropogenic
influence on long return period daily temperature
extremes at regional scales. Journal of Climate 24:
881–892.
Hulme 511
at Kings College London - ISS on August 7, 2014ppg.sagepub.comDownloaded from
