Recent extreme weather events and their impacts on societies have highlighted the need for timely adaptation to the changing odds of their occurrence. Such measures require appropriate information about likely changes in event frequency and magnitude on relevant spatiotemporal scales. However, to support robust climate information for decision-making, an effective communication between scientists and stakeholders is crucial. In this context, weather event attribution studies are increasingly raising attention beyond academic circles, although the understanding of how to take it beyond academia is still evolving. This paper presents the results of a study that involved in-depth interviews with stakeholders from a range of sectors about potential applications and the general usefulness of event attribution studies. A case study of the hot and dry summer 2012 in southeast Europe is used as a concrete example, with a focus on the applicability of attribution results across sectors. An analysis of the interviews reveals an abundant interest among the interviewed stakeholders and highlights the need for information on the causes and odds of extreme events, in particular on regional scales. From this data key aspects of stakeholder engagement are emerging, which could productively feed back into how probabilistic event attribution studies are designed and communicated to ensure practical relevance and usefulness for the stakeholder community.
Changing extreme weather events are likely to be among the most notable consequences of anthropogenic climate change with potentially highly significant impacts on societies (Seneviratne et al. 2012; Barriopedro et al. 2011; Rahmstorf and Coumou 2011; Huntingford et al. 2014). A number of recent extreme events with considerable adverse impacts have raised questions as to whether and to what extent those events might be linked to climate change (Rahmstorf and Coumou 2011; Coumou and Rahmstorf 2012). These have included droughts in the Horn of Africa (Funk et al. 2013; Lott et al. 2013); flooding in Pakistan (Christidis et al. 2013), Bosnia and Serbia (van Oldenborgh 2014) and Thailand (van Oldenborgh et al. 2012); and exceptionally hot and dry summers in Russia, southeast Europe (Christidis et al. 2013; Otto et al. 2012; Sippel and Otto 2014), and California (Swain et al. 2014; Funk et al. 2014), among others.
Although the frequency and magnitude of some extreme events such as heat waves have increased in recent years (Donat et al. 2013; Seneviratne et al. 2012), the question of causality (i.e., whether the odds of occurrence of any particular extreme event have changed due to anthropogenic climate change) remains largely unanswered (Stott et al. 2013). Over the last decade, however, scientists have started to use and explore the prospects of probabilistic event attribution (PEA) as a way to make probabilistic statements about the links between climate change and particular extreme events (Allen et al. 2007). After the method had been initially described (Allen 2003; Stone and Allen 2005), scientists attributed large-scale heat events, such as the European heat wave 2003, and found that human influence on the climate “has at least doubled the risk of a heat wave exceeding this threshold magnitude” in a European summer (Stott et al. 2004, p. 610). Subsequently, PEA was extended to smaller spatial scales and more directly impact-relevant variables, such as floods (Pall et al. 2011; van Oldenborgh et al. 2012) or dryness (Sippel and Otto 2014). Since 2012, an annual report has been published by the American Meteorological Society that aims to investigate extreme weather events within a wider climatic context (Peterson et al. 2012, 2013; Herring et al. 2014), which has received large interest both within the scientific community and beyond. Currently, a near-real-time quasi-operational attribution service is being discussed (Stott et al. 2013; Nature 2012) and the prospects are explored within the recently launched European Union (EU)-funded project European Climate and Weather Events: Interpretation and Attribution (EUCLEIA 2014; http://eucleia.eu). It is anticipated that this would allow a preliminary answer to the question whether or not anthropogenic climate change played a role with respect to an extreme event while the event is happening or shortly after, but crucially “while the world is listening.” The real-time opportunities that attribution scientists are striving to achieve could provide a wide range of applications (Stott and Walton 2013). However, even with the current time lag in the attribution process a number of roles for the science have already been established, for example, by offering a long-term perspective on extreme weather events and, in doing so, addressing the public’s demand for rigorous climate information (Schiermeier 2011; Stott et al. 2013). This would potentially contribute to closing communication gaps in media and public discourse in how daily weather is influenced by long-term climate. Furthermore, it has been suggested that on regional levels, attribution information could assist government authorities, nongovernmental organizations (NGOs), and businesses in adaptation planning (Pall et al. 2011; Schiermeier 2011; Hoegh-Guldberg et al. 2011; Stott et al. 2013), emphasizing changes in extremes that are not only projected for the future but are already under way. Further potential applications of event attribution lie in legal claims (Grossman 2003; Lord et al. 2012) or in the loss and damage agenda under the United Nations Framework on Climate Change (UNFCCC; Huggel et al. 2013; James et al. 2014), where under the adaptation program the Warsaw International Mechanism for Loss and Damage (UNFCCC 2013) has been established to address impacts of climate change we cannot adapt against (Thompson and Otto 2015).
However, a more general use of event attribution science will present a number of challenges. Hulme et al. (2011) and Hulme (2014) suggest that the attribution of extreme weather events could be counterproductive by emphasizing the uncertainty and possibly subjective outcomes of these studies, and also that there can be a discrepancy between the (meteorological) hazard (which is being attributed) and the (real) risk of the event, including vulnerability and exposure. The latter point also involves other complex and interacting drivers, for example in assessments of biological impacts (Parmesan et al. 2011).
While it is recognized that there is a compelling case to better understand the impacts of a changing climate and the potential change in risk from extreme weather events (Moss et al. 2013), these impacts have not yet been universally adopted by stakeholders (Schiermeier 2011; Stott and Walton 2013). Nonetheless, it has been shown that robust decision-relevant climate science is more likely to emerge from extended scientist–stakeholder dialogues, including active learning, support and feedback loops, rather than from “predict-then-act” types of scientist–stakeholder communication (Weaver et al. 2013). The latter places the expert climate change scientist at the top of an information chain feeding knowledge down to the nonexpert, passive stakeholder (Burns et al. 2003; Young and Matthews 2007). This process does not provide a mechanism for the decision-maker to communicate back to the scientist any concerns, misunderstanding, relevance, or timeliness of the issues. In contrast, the contextual model underlying recent stakeholder engagement efforts defines communication as a process of dialogue between an expert and nonexpert where information is socially contextualized (Weaver et al. 2013). That is, the information is not communicated in a social vacuum, but rather all parties will have their own cultural values, norms, heritage, political perspectives, and so on that will influence how the information is received, processed, and applied (Kahan et al. 2012). Dialogue between the parties, therefore, not only allows the science to be framed within one context, but also helps to situate the science in a range of contexts, providing parties with an equal stake in the knowledge construction (Sarewitz and Pielke 2007; Weaver et al. 2013). Using this approach the stakeholder becomes a central part of the research process (O’Riordan 1994; Harrison et al. 2004; Sarewitz and Pielke 2007) by a mutual exchange of information that can support decision-makers’ understanding and develop trust in the process (Wagner 2007; Krauss et al. 2012). The development of trust is a key aspect of the communication and engagement process as highlighted by Cash et al. (2003) and Kahan et al. (2012) as a way to develop credibility, saliency, and legitimacy of the research and researcher. Here, “credibility” refers to the adequacy of the scientific methodologies deployed, which ideally generates knowledge relevant to the stakeholder (salience) without compromising potentially opposing values and beliefs of the stakeholder or other relevant groups in its production (legitimacy; see Cash et al. 2003).
Within the context of climate change, extensive stakeholder dialogues have been conducted for certain climate information products, such as the UK Climate Projections 2009 (UKCP09; Street et al. 2009), to assess the relevance, usefulness, and uptake across the stakeholder community. Notably, Tang and Dessai (2012) found that while these forecasts where mostly considered as legitimate and credible, the practical uptake for specific decisions was limited due to, but not limited to, the complexity and information richness provided by probabilistic forecasts that are generally not incorporated into adaptation decisions.
Although it is not as developed as other climate services, there is expected to be a high degree of stakeholder engagement within event attribution, which is currently being tested in recent projects. However, at present extensive dialogues or platforms for dialogue have not been developed or institutionalized more broadly (Stott and Walton 2013). Moreover, the literature on practical exercises that assess how this could be applied within attribution science is still sparse (Stott and Walton 2013). A possible reason for such limited stakeholder engagement with attribution science could center on the issue that, even compared to other climate science problems, it is both a complex issue and a relatively new subject, sometimes even compounded by a range of apparently conflicting attribution statements (Otto et al. 2012; James et al. 2014).
While there are gaps between the state of the science and the possibility of its application, the potential for its use is slowly being realized. Although the growing attribution scientific community can in principle assess the attributable risk of a wide range of extreme weather events and their impacts (Stott et al. 2013), the needs of potential users of such data are largely unexplored. Failure to adequately communicate and contextualize the science of event attribution would potentially present a missed opportunity (Sarewitz and Pielke 2007) in terms of improving or optimizing climate-related decisions on various temporal and spatial scales.
This explorative study explores the gap between stakeholder needs and the scientific potential of probabilistic event attribution and considers the design principles necessary for meaningful engagement and application.
Our study involved a series of in-depth interviews with stakeholders to identify their perspectives on attribution science in relation to their capacity as decision-makers to determine what kind of attribution information might be relevant and useful. The interview data were gathered to help identify potential improvements and shortcomings of the PEA methodology, in particular whether the current attribution research is relevant to stakeholders. For this study the term “probabilistic event attribution” is defined as the detection of changes in the probabilities of extreme events, and the subsequent establishment of causes, following Christidis et al. (2013) and Bindoff et al. (2013).
The sample population for the study was drawn from sectors already engaged in climate policy and/or adaptation. This population was selected to represent the stakeholder community who would potentially make use of the data presented to them, in particular with respect to extreme weather events of drought and heat.
As the study was designed as an explorative enquiry involving focused participant selection criteria, a small sample size was recruited (N = 14). Hence, with the qualitative analysis of the data it is not possible to extrapolate results to a wider population. However, the study has been able to identify key principles that can be tested as part of a future larger research study. Table A1 in the appendix shows the breakdown of the sectors and participants, including further details of each (anonymized) interview.
The semistructured interviews lasted approximately 1 h and were based on the following protocol: First, stakeholders were shown an analysis of an extreme seasonal heat and drought event that had occurred over southeast Europe in summer 2012. Preliminary results of the attribution study (Sippel and Otto 2014) were shown, explained, and discussed in detail (see below for further information on that study). Second, the interviews continued with a series of key questions (Table 1). Those questions were chosen to elicit sector-specific information relevant to event attribution (e.g., current level of use of climate services, anticipated applications, etc.), to identify potential methodological shortcomings and to discuss possible improvements. During the questioning, the interviewer and participant were able to explore other related issues dependent on the responses. Last, the interviews finished with a freely evolving dialogue based on further issues or comments brought up by the participants that had not been addressed during the predetermined structure. The interviews were recorded on electronic media based upon approval by the participants, and were subsequently transcribed.
A free-coding analysis (Flick 2009) of the interviews identified a range of emerging themes that could be considered as important issues for users of event attribution information. Consideration needs to be given to the relatively small sample size, which could lead to potential biases, along with the semistructured nature of the interviews and coding process. However, the qualitative methodology was not intended to provide a comprehensive, unbiased reflection of (potential) users of, or perspectives on, event attribution science, but to offer the opportunity to elucidate broad cross-sectoral insights into stakeholder perspectives. A selection of relevant statements from the interviews is detailed in the appendix tables.
The interviews were primed with an exemplary illustration of an event attribution study of an exceptional season of heat and drought in southeast Europe, which two of the authors had previously conducted (Sippel and Otto 2014). This was used to demonstrate what attribution information could look like and provide a focus for discussion. The event was chosen because it had significant impacts on agriculture and human health (G. Simic 2013, personal communication) and occurred in a region where no event attribution study had previously been conducted. Therefore, the choice of event, regional focus, and its extent and magnitude might have influenced the interview process and the interest of participants in the issues.
The study region seen in Fig. 1 encompasses mostly states on the Balkan Peninsula and the Pannonian Plain and is often considered as a region particularly vulnerable to the impacts of climate change; thus, appropriate adaptation and planning measures are of high interest in a range of sectors of the regional economy (Whitlock 2012). In addition, southeast Europe was considered a good case study to present to participants as it has seen a number of extreme events in recent years, most notably a severe drought and heat wave in the summer in 2012 (Hydrometeorological Service of Serbia 2012), with mean summer temperatures up to 5°C above the average and lower than normal precipitation (Fig. 1). Based on extensive climate model ensemble simulations, Sippel and Otto (2014) found pronounced changes in the probability of 5-day summer heat waves compared to an earlier decade (1960s; Fig. 1). For example, the frequency of a 100-yr monthly heat extreme in July or August in the 1960s now occurs around once in a decade and hence the frequency has increased by more than an order of magnitude, which might have profound implications for climate adaptation planning in the region.
3. Results and discussion
The coding analysis of the interview data identified a range of themes that emerged from the stakeholders’ perceptions of how PEA could be applied in practical decision-making processes, including in support of climate adaptation decisions, as tools for communication, and for the allocation of funds and liability within international negotiations. In addition, we construct and discuss some guidelines that could be used by scientists to develop meaningful stakeholder engagement.
a. Stakeholder perspectives on event attribution
In initially establishing whether event attribution research has a role to play in decision-making it was stated by all interviewed stakeholders that the presented methodology is potentially of high practical relevance to their work (Table A1). Although the perceived usefulness of PEA varied between “it is certainly interesting” and “this is absolutely a step forward,” we interpret these statements as indicating that a close engagement with the stakeholder community could yield very productive results (Table A2). A central aspect of those claims was that “event attribution” answers the question of how changes in the climate impact current weather, which is not answered by projections (Table A2). However, various sector-specific reasons might have additionally motivated these statements, and are discussed below.
Initially, the PEA framework had been motivated by its potential applicability in the insurance and risk sectors (Allen 2003). This view was corroborated by an insurance expert, who emphasized that the availability of attribution insights would help to determine the additional risk of extreme events imposed by climate change—the “delta” (participant N; Table A2). The interviewee further suggested applications in which smallholder farmers in vulnerable regions could potentially be reimbursed for the risk premium due to climate change, justifying public or donor subsidies (keeping in mind that meteorological risk is only one contributing factor to disaster risk; see, e.g., IPCC 2012). Moreover, there was agreement that investment decisions are often decisively influenced by risk assessments—an area where any better description of how the different risk components are changing would constitute an improvement (Table A2). However, this is not to say that changes in climate necessarily exert major influence on the integrated risk: “I get the impression that in developed countries, too few emphasis [sic] is put on the increasing value of assets and their vulnerability, because it’s easy to blame the climate. […] But this does not mean that one does not need to understand how climate has changed” (participant A).
Regardless of those general statements, however, stakeholders that are actively engaged in southeast Europe recognized the importance of attribution information on regional scales, particularly where detailed climate information products are scarcely available (Table A2). A public health expert explained that in this region the development of early warning and adaptation tools (e.g., the development of national “Heat-Health Action Plans”; WHO 2011) has made use of probabilistic climate predictions and aims to reduce heat impacts among the most vulnerable population. Here, participant H emphasized that an additional layer of information that targets the current state of the climate would enhance those activities, as it would unambiguously highlight the relevance for addressing and reducing public health related risks now, not only in a somewhat distant future. Furthermore, PEA information could assist to illustrate the complexity and interconnectedness of the impacts of extreme events. In the case of southeast Europe these have resulted by a combination of heat and drought affecting both animal and human health through various pathways (participant K).
Stakeholders who engage with communication activities stated that attribution results could help raise awareness of climate change impacts, which previously has been achieved only through using anecdotal experiences. However, the latter had led to accusations of being selective in their choice of events, which would possibly undermine the message. This highlights the need of a comprehensive approach to PEA aspired by projects like EUCLEIA. Participant J noted that “One has to see the wood [climate change] for the trees [single weather events]. I think the wood becomes much clearer to see. However, it is still important to better recognize single trees in order to accurately determine the problem.” The interviewee further explained that studies with clear statements about changes in the odds of specific extreme events would enter and influence political debates and potentially raise awareness among politicians and policy-makers by increasing the “signal-to-noise ratio” and would therefore help to shape common norms.
However, many stakeholders were keen to stress that any consideration of event attribution information would be in addition to the climate information they are currently using (often on a very detailed level; Table A2). It is therefore important to recognize that any attribution statement would not be presented in isolation, but rather would need to be presented in such a way as to complement the existing information, making it as simple as possible for the stakeholder to assimilate this new information (Table A2). Importantly, this would preclude neither low-regret adaptation strategies (Wilby and Keenan 2012) nor a general build-up of capacity to deal with disaster risk (Hulme et al. 2011).
While the limited sample size and the explorative nature of this study does not allow us to derive definitive conclusions, we still can infer from the interviews that stakeholders across different sectors would welcome the availability of information regarding the attribution of extreme events. The statements also reveal different sector-specific information needs, which corroborates previous findings in the context of probabilistic climate projections (Street et al. 2009; Gawith et al. 2009). Hence, a future development of sector-specific dialogues seems to be promising in this context. Besides, these dialogues could tackle questions that remain currently unanswered, namely whether some of the anticipated applications prove mutually exclusive or whether there is a risk that certain groups within sectors could be marginalized by PEA-related activities.
As previously mentioned, the question of whether probabilistic event attribution could have a role to play in allocating funds or liability in international negotiations is currently under debate (Hulme 2014; James et al. 2014), which was reflected in the participants’ responses during the interviews.
Some stakeholders argued that there might be a role for scientific evidence taken from event attribution in the allocation of adaptation funds or in the emerging loss and damage agenda [e.g., a “compensation mechanism with a scientific element to it” (participant L)], although a more integrated risk assessment framework would be needed (Huggel et al. 2013). It was also pointed out that an attribution of cost estimates (e.g., by using empirical damage databases; see Bouwer et al. 2007) would be an important prerequisite for the allocation of funds or liability, which could be a relevant extension to attribution studies but would necessarily have much higher uncertainties attached: “I think the holy grail is to pin the costs of particular extreme weather events to hold some parties liable for those costs. […] As I understand it the holy grail of being able to attribute the costs of a particular event to a particular party is difficult to envisage, but I think that there could be some particular use in determining uses of global funds for climate adaptation, for example” (participant H).
Yet, other participants expressed the opinion that the allocation of climate funds will remain fully within a political process (both proponents and opponents of this idea were actively engaged in international climate negotiations). This implies that the issue of compensation would not be linked to any scientific risk analysis, as summarized by participant J: “Particular studies that have come up with a framework [to allocate funds] have not been taken up by the [UNFCCC] process, because in the end it is a political negotiation process, which involves normative decisions. […] I believe that in the future the process will be that attribution informs decision-makers, which will then determine allocation decisions in a political process.” Thus, scientific (case) studies of event attribution could contribute to clarify the overall picture, but would not directly influence normative decisions. A similar view is also conveyed by Hulme (2014), who argues that in a political context “it seems unlikely that there will be a role for the sort of forensic science being offered by extreme weather attribution science” (p. 9).
Nonetheless, we found agreement among stakeholders that the issue of compensation, closely connected to individual parties’ liabilities, is still widely tabooed within climate negotiations (which had been stated before; see, e.g., Müller 2004), even if a scientifically unambiguous probabilistic attribution of a damaging weather event were possible. While this seems true in the emerging negotiations around loss and damage where parties so far are not yet discussing a definition of the term (James et al. 2014), an indirect influence of scientific evidence gathered in event attribution studies seems plausible as the UNFCCC secretariat actively requests background information on the topic.
Hence, the mixed views of our stakeholders seem to mirror the academic debate on event attribution (Table A3). While clearly indicating a potential demand for event attribution assessments we emphasize that further qualitative studies, and in particular a broader dialogue with relevant stakeholders, are needed to explore in more detail whether and how probabilistic event attribution studies could be designed. An example would be addressing how PEA could help to base instruments like the Warsaw mechanism on scientific evidence, beyond providing a clearer overall picture by exemplarily assessing the risk of extreme weather events attributable to greenhouse gas emissions as is currently done (e.g., Peterson et al. 2013; Herring et al. 2014).
b. Designing attribution studies that are relevant beyond academia
The interviewed stakeholders saw probabilistic event attribution as potentially having a positive impact on decision-making (Table A2). However, during the interviews it became evident that it is not just the information that is important but also how the information is constructed and communicated. Here, we highlight design criteria for event attribution studies that were considered important by the participants. The following presentation and discussion is structured along the three central themes that frequently lie between knowledge and action: credibility, legitimacy, and salience (Cash et al. 2003).
It was emphasized that in order to use an event attribution result for adaptation policies or to include in science-based NGO campaigns, a high level of credibility is crucial (see also Wagner 2007; Cash et al. 2003). It was stated that scientific credibility in a policy context includes the reputation of individual researchers and their institutions, a wide acceptance of the applied methodologies by the scientific community, and the publication of studies in peer-reviewed scientific journals (participant L; Table A4). However, credibility goes beyond any single researcher, institution, or journal. For instance, event attribution studies that are being carried out remotely would gain credibility if local or regional researchers were involved, for example through a collaborative study design or evaluation of results as shown earlier (O’Riordan 1994; Harrison et al. 2004).
Further, most stakeholders emphasized that studies legitimately addressing policy-makers or the public must show “a great deal of confidence” and “have to be careful to communicate fully the uncertainty that comes with that” (Table A4). Those statements might raise concerns about whether there is a tension between the need for relevant/robust information and methodologically inherent uncertainties. However, participants further stressed that uncertainty in a modeling study as such does not impede its usefulness, as long as it is clearly illustrated, because “politicians and decision-makers all over the world make very important and far-reaching decisions on much less certainty. […] The fact that we are dealing with decision-making under uncertainty is a fact, and it doesn’t mean we should only do it in economics and not in climate change” (participant B).
The need to clearly define terminology and methodological limitations is of particular importance when it comes to attribution studies, since the risk that is being attributed (which is meteorological hazard in many, but not all cases) might differ from risk perceptions in the stakeholder community (Hulme et al. 2011).
However, as a result of our study we found that the participants were well aware that “hazard is a central component of risk, but not the only one,” which was reinforced by several similar statements from stakeholders across sectors (Table A4). It was emphasized that there could not be one single dataset or methodology that provides for the full range of risk or that determines political decisions (see also Stirling 2010; Huggel et al. 2013), but a methodology dedicated to identify changes in the hazard component was nevertheless regarded as highly useful (Table A4). For example, an insurance expert stated that “[Poor discrimination between hazard and risk] is not my problem, because I wish that science could answer the question how hazard probabilities have changed. Of course vulnerability changes as well and I include this into my calculations […]. But the question I wish to be answered is about causality and changes in the hazard component” (participant N). However, we note that the interviewed stakeholders were (high level) experts in their respective fields, and hence we argue for caution in extrapolating those statements and awareness illustrated here into the wider stakeholder community where understanding of the role of risk may not be so developed.
Nonetheless, while assessments of uncertainty and a clear definition of a study’s limitations and terminology are a crucial part of scientifically credible methodologies, an open discussion and illustration of an attribution study’s limitations might simultaneously enhance its legitimacy among the stakeholder community.
From our interviews as well as previous research (Stott and Walton 2013; Stott et al. 2013; James et al. 2014) one might conclude that salience of attribution statements strongly depends on the particular sector involved. In addition, however, we find that the eligibility of the results as well as the consideration of spatial and temporal scales (and the variables that are being attributed) seems to determine a large part of the perceived usefulness to many stakeholders. For instance, it was stated that precise, understandable communication of scientific results is crucial if attribution statements are to be included in public education or policy campaigns (Table A4). According to an interviewee, this includes a clear labeling of figures and graphs and the availability of short captions that clearly portray the message; “if there’s also [easily understandable] information with it, for example a blog in addition to a peer-reviewed publication, that is very powerful” (participant F). In addition, it was suggested that numerical data are often hard to contextualize especially among the general public, citing the 2°C global warming target as an example: “It’s warmer, fine. […] But for the rest of the population it means absolutely nothing. But in terms of mortality, flood risks, and all sorts of things, that [attribution] is very important, because it means something to the public” (participant B). Hence, the communication of tangible climatic impacts such as changes in extreme events might be one of the key advantages of event attribution, if adequately contextualized. Further, several participants suggested that attribution scientists should think carefully about the geographical scope of their studies, particularly since PEA studies generally involve the calculation of regional averages that can mask nuanced information within a region. For example, studies relevant for public health would have to consider the local topography in a mountainous environment as the majority of the population live in the valleys and not on the mountains with very different climatic conditions (participant F). Similarly, studies that are meant to be relevant for decision-makers would have to incorporate political and economic realities and state boundaries, particularly in a highly fragmented region such as southeast Europe (participant M). In this regard, the possibility to “zoom in” was suggested in order to distinguish between different regions, which could further increase the usefulness of probabilistic information to regional stakeholders. However, it would require very detailed model simulations and might come at the cost of larger uncertainties. An extension of the geographical scope to regions with highly vulnerable populations and where climate information is often scarce (e.g., sub-Saharan Africa) was considered as a potential hot spot for the usefulness of event attribution assessments (Table A4).
The themes outlined here could serve as a starting point to develop suitable design principles for stakeholder-relevant attribution studies. We note that future in-depth assessments of these and related criteria in relation to the application of event attribution in the wider stakeholder community would constitute a highly topical research activity. In particular, potential trade-offs or tensions between these themes and within specific sectors will have to be explored, as it has been shown that such trade-offs are often present in science advising and assessment, for instance between credibility, salience, and legitimacy (Cash et al. 2003).
Nonetheless, the design ideas for pragmatically relevant event attribution studies presented and discussed in this paper could be readily taken up by attribution scientists as a feedback from the explorative stakeholder dialogue.
The engagement with potential users of attribution information from various sectors has shown a variety of different perspectives on event attribution. The most intriguing result of our scoping study was that, regardless of their backgrounds, all interviewed stakeholders found the presented findings of an event attribution study in southeast Europe interesting, and most argued that results of this kind could potentially be usefully applied in their daily work.
This finding is notable because it suggests that results of attribution studies are relevant beyond pointing the finger about who is to blame for climate change. Our results indicate that this kind of research might become a useful tool in the day-to-day work of adaptation planners or to bolster investment decisions. However, future enquiries are needed to investigate in detail how those applications could be tailored more specifically, for example in the water, agriculture, or health sector. Further, an objective communication of the links between climatic changes and the impacts of extreme events was suggested by participants as another key advantage of the PEA method. This would add an additional layer of information to current discourses about extreme events, where at present some people are quick to blame climate change for the bulk of extreme events, or conversely for none of them.
Nevertheless, stakeholders’ needs with respect to attribution information were not found to be uniform across sectors but they were by no means contradictory, corroborating earlier results (Gawith et al. 2009; Street et al. 2009). The identified themes at the interface between science and its application indicate that detailed and fine-tuned attribution studies, with careful consideration of spatiotemporal scales as well as local meteorological peculiarities, and the (hydro)meteorological variables that are being attributed could add to the range of climate science products that are currently used by stakeholders in various sectors and regions and could provide more objective estimates for meteorological hazard and its changes. Those products should be accompanied with a clear communication of the full range of methodological uncertainties, and would presumably offer a higher credibility if peer-reviewed research papers are produced in collaboration with local scientists.
Moreover, research into how probabilistic estimates of meteorological hazard, as generated by PEA studies, could be made more impact-relevant, for instance through the combination with different (observational) datasets or impact models, seems currently to be highly topical among stakeholders. This being said, we emphasize that these studies should be conducted without claiming to determine political decisions, but to better inform stakeholders, decision-makers, and the public, for the feasibility of which we presented qualitative evidence in this paper.
We thank all stakeholders who took part in this study. Furthermore, we appreciate comments from four anonymous reviewers, which strongly improved the manuscript. S.S. acknowledges funding by the German National Academic Foundation.
Extended Interview Data
Current affiliation: Max Planck Institute for Biogeochemistry, Hans-Knöll-Str. 10, 07745 Jena, Germany.
Publisher’s Note: This article was revised on 30 March 2016 to include the CCBY license that was missing when originally published.