The 2010 eruption of Eyjafjallajökull volcano, and the social consequences across the world, demonstrated some key issues in volcanological science and its application. Scientists in several nations were called upon to advise governments, to justify models and to give guidance about likely future activity. This is symptomatic of many other fields: scientists increasingly have a role in governance, and their work may be driven by questions that arise as a result. This article considers the role of scientists in different national contexts and the challenges faced in formulating scientific advice for policymakers. It concludes by assessing future challenges, and the key role that social scientific research can play. While this is a research paper and presents new data, it takes a commentary approach to elucidate some of the challenges involved in governing volcanic hazards.
 The timescales of volcanic eruptions do not correlate well with the timescales of politics—volcanoes may erupt very rarely [e.g., Mason et al., 2004], while political terms tend to be around 4 years in length. In addition, governments may struggle to deal with so-called “black swan” events [Taleb, 2007]—and volcanic eruptions may fall into this category. During volcanic crises, the dependence on scientific advice tends to be very high [e.g., Aspinall et al., 2002; Bostok, 1978]. This puts the onus on scientists to mobilize appropriate advice in a crisis, requiring a degree of reflexivity: a term used in the social sciences for research that is context sensitive [e.g., Alvesson and Skoldberg, 2000]. Recent debates in climate science, for example, have called into question the use of different scientific methods and models [e.g., Hulme and Mahony, 2010]—and this is a social question about the nature and authority of knowledge and its acquisition, as well as a scientific one. The combination of social and physical scientific ideas is not a trivial process, partly because it requires the acceptance of qualitative methods—not readily acknowledged by many in the scientific community, but highly regarded in the social sciences. In recent years, however, there have been many studies of the social context of volcanic eruptions, particularly in the field of risk perception and communication [e.g., Johnston et al., 1999; Gregg et al., 2004; Perry and Lindell, 2008; Paton et al., 2008; Dibben, 2008; Gaillard, 2008; Solana et al., 2008; Haynes et al., 2007, 2008a, 2008b; Bird et al. 2009, 2010, 2011]. These studies have revealed the cultural complexities of social perceptions of risk and subsequent human behavior, and can aid both scientists and decision makers in local contexts. They also have aspects that can be more generally applied, such as the idea that scientists may be more trusted than the media [e.g., Haynes et al., 2008a; see also Eiser et al., 2009]. However, the 2010 eruption of Eyjafjallajökull produced challenges that were apparently new, in that multiple governments were involved in nations that do not themselves have active volcanoes on their mainland. This creates some challenges for the scientific community as well as for eruption management more generally, because it crosses cultural boundaries. These may be exhibited in governments as well as in populations.
 The history of volcanic crises has provided a range of case studies for the involvement of scientists in the decision-making process. A key example is the noneruption of La Soufrière de Guadeloupe in 1976 [Bostok, 1978]. In this case, disagreement between scientists proved detrimental to management. However, Hincks  suggested that the evacuation of residents was justified in this case because the scientific uncertainty was so high, based on retrospective Bayesian analysis. There have been many studies of the process of decision-making during volcanic eruptions, and tools have been developed for this purpose [e.g., Newhall and Hoblitt, 2002; Marzocchi et al., 2004, 2008; Marzocchi and Woo, 2009; Aspinall et al., 2003; Aspinall, 2006; Neri et al., 2008; Baxter et al., 2008]. Many of these tools have been used effectively in simulated and real events, such as the eruption on Montserrat [Aspinall et al., 2002]. However, they can only be used where volcano advisory structures exist within the relevant government. In the case of Eyjafjallajökull, the UK government did not have such structures in place. This paper uses examples from the eruption and its management in the UK and Iceland to draw conclusions about the role of scientists in volcanic crises, and about eruption management. It applies both quantitative and qualitative data obtained before, during and after the eruption. Initially, results from a survey prior to the eruption will be discussed to set the broader context. This will be followed by analysis of results from the 2010 event.
 The 2010 eruption of Eyjafjallajökull volcano in South Iceland began on 20 March, following several months of unrest. Initially, lava was extruded on the eastern flank of the volcano through a small fissure that cut across the Fimmvorðuhálsi region between Mýrdalsjökull and Eyjafjallajökull. The magma was a primitive, crystal-rich basalt inferred to have risen from depth relatively rapidly [Sigmundsson et al., 2010]. The fissure eruption ceased on 12 April and appeared to be over, although little deflation of the magma source had been observed. Late on 13 April, accelerating seismicity was detected under the glacier, and at midnight on had remobilized the trachydacitic magma chamber from the 1823 eruption, generating a trachyandesitic magma [Sigmarsson et al., 2011]. Explosivity was considerably amplified by the presence of the glacier. This phase of the eruption resulted in airspace closures across Europe, widespread media coverage, and government anxiety. In spite of warnings from scientists, European governments were not prepared to deal with this scenario. This paper will focus on the UK and Icelandic management of the eruption, looking at them comparatively and drawing conclusions about the role of scientists as advisors to governments in this and other eruptions.
 In the UK, volcanic eruptions were not included in the National Risk Assessment. There were no plans for such an event, and the government had not heeded scientists' warnings [e.g., Oppenheimer, 2010], in spite of extensive research on the impact of Icelandic eruptions across Europe [e.g., Thordarson and Self, 2003; Thordarson and Larsen, 2007]. The response of the government was therefore reactive. A Scientific Advisory Group in Emergencies (SAGE) was formed, under the auspices of the Office of the Chief Scientific Advisor (CSA). The SAGE comprised scientists already known to the government, either through previous work (such as involvement with the BGS and the Meteorological Office) or through recommendation. The SAGE met four times between 21 April and 24 June 2010. The membership evolved slightly over time, reflecting the involvement of experts in particular fields. The later two meetings also involved representatives from the Icelandic Meteorological Office (IMO) and the Institute of Earth Sciences (IES) at the University of Iceland, by telephone.
 In March 2011, the UK House of Commons Science and Technology Committee published a report entitled “Scientific Advice and Evidence in Emergencies” (SAEE), which included an assessment of the SAGE mechanism as displayed in the “ash crisis.” A number of comments, suggestions and criticisms emerged, and will be discussed later in this paper. Also of relevance here is a slightly earlier document published by the Council for Science and Technology in 2008 entitled “How government and academia can work together.” This report identified some weaknesses in current structures for the integration of science and policy – particularly a lack of clear channels for advice. This was picked up in the SAEE report, which recommended that the Office of the CSA identify in advance experts in particular fields so that they can be consulted in a timely manner.
 The scientific advisory procedures in Iceland are considerably less complex than those in the UK. They also reflect the fact that Iceland is much smaller, and that Icelanders are very much aware of their geology and its potential impacts. The challenges in Iceland were also more extensive in some ways—local communities were significantly impacted by ashfall—but also less disruptive to large numbers of people. When the ash cloud was moving south toward Europe, it was sometimes possible to open the airspace over Reykjavik to flights from the U.S. and Canada, for example. Iceland was also relatively well equipped to monitor the cloud using the radar installation at Keflavik. Responsibility for monitoring volcanoes in Iceland lies with the IMO in Reykjavik, which has a seismic network around the island, continuous GPS stations and the capacity to extend the networks in places that are showing signs of activity. The IMO works closely with the IES, which has additional seismic and GPS stations, as well as analytical facilities and expertise. Scientists from the IMO and the IES advise the Icelandic government on geological risks through a committee that meets every few months. In the event of an emergency, operational decisions are made by the Risk Office of the Icelandic Police, which has expertise in disaster management. Procedures are then executed by the police chiefs at the local level.
 Three sets of data were used for this analysis: interviews with scientists and local officials; reports and documents from the UK government; and a survey carried out in 2008–2009. Twelve interviews (semistructured) with UK and Icelandic scientists and local officials were carried out between 2008 and 2010. Interviews were carried out both before and after the eruption and ranged from 30 min to 2 h in length—five interviews were carried out before and during it, but all contain data relevant to the argument of this paper. In addition, documentary analysis was undertaken using the minutes of the SAGE meetings and the reports in Table 1. This included interview data from the hearings preceding the reports. It should be noted that it was not possible to interview SAGE scientists about the meetings because of legal constraints from the government. Thus, discussion of the SAGE material relies on documentary analysis of the minutes, reports and supplementary documents provided on the government websites. Documentary and archival analysis is a standard social scientific method, particularly in the political sciences where interviews may be limited in scope.
Table 1. Main Documentary Sources Used in this Papera
Reference URL and Notes
Note that the numbers are used to reference the sources later in the paper.
1. Scientific Advice in Emergencies Report (SAEE)
http://www.publications.parliament.uk/pa/cm201011/cmselect/cmsctech/498/498.pdf This includes evidence from stakeholders and scientists
2. Additional written data for SAEE
http://www.publications.parliament.uk/pa/cm201011/cmselect/cmsctech/498/498vw.pdf - this includes statements from a range of stakeholders and experts from volcanology and aviation
 The paper also draws on data from an online survey of 186 volcanologists in 2008 and 2009, distributed using the Volcano Listserv. The survey preceded the Icelandic eruption, but the results are highly relevant to the issues raised by the eruption and its management. The survey was concerned with the role of volcanologists as scientific advisors, and the interface between science and the public. It also examined the relationship between volcano monitoring and research. The survey results were analyzed using a range of statistical tests. Initially, results were assessed for normality. As several of the statements had provoked strong opinions, it was appropriate to analyze them using nonparametric tests. In particular, the Kruskal-Wallis test (denoted by “U”) and the Jonckeheere-Terpstra test (denoted by “J”) were used to compare the medians of different groups. The effect size, r, was also calculated. Predictor variables used in the analysis were: highest qualification, experience in decision-making, experience working in a volcano observatory and experience in academic institutions. A further predictor was “resources,” which is a combined variable. It was developed to measure ability to cope with volcanic eruptions, and has three levels: “0” represents countries which do not have volcanoes, “1” represents developed countries with volcanoes, and “2” developing countries with volcanoes. Significant results were those where the probability (p) of a false result was less than 5%. It should be noted that many of the respondents were from Anglophone countries—particularly the UK, U.S. and New Zealand—and mainland European countries.
4. Survey Results
 This section presents results from the survey carried out prior to the 2010 eruption. These results are shown in Table 2. They will be used to contextualise the focused results from Iceland. A detailed demographic is provided in the work of Donovan et al. [2012a]. Respondents were dominantly from Anglophone countries (57%), and most had postgraduate degrees (52.9% had PhDs). Observatory experience was claimed by 62%, though 12% of that number were “affiliated scientists.”
Demographic results for this survey are provided in the work of Donovan et al. [2012a]. The IAVCEI Crisis Protocols document [Newhall et al., 1999] is a set of guidelines for volcanologists produced to manage some of the issues that arise during volcanic eruptions.
1. Monitoring volcanoes is more important in the short term than research.
2. Monitoring is vital for volcanic risk management.
3. Research is the key to effective monitoring.
4. During eruptions, priority must be given to monitoring volcanoes rather than carrying out research.
5. The role of volcanologists in periods of volcanic unrest is purely to provide scientific information.
6. Volcanologists should adhere to the IAVCEI Crisis Protocols 1999 document.
7. The role of volcanologists inevitably includes recommending evacuations.
8. Volcanologists at academic institutions should actively seek to share resources with observatories.
9. Public safety is threatened when scientists do not trust one another.
10. Scientific distrust is most often related to data usage.
11. Volcano observatories should be run by scientists.
12. Volcano observatories should be run by local officials (non-scientists).
13. Volcano observatories should be government-funded.
14. Observatories must actively engage with the public to educate them about volcanic risk.
15. Risk education is the responsibility of local government.
16. Pressure to publish can impede monitoring activities at observatories.
5. Monitoring and Research (Statements 1–4)
 Volcanologists with PhDs were slightly more likely to believe that research is the key to monitoring than those with masters degrees (U = 1753.5, r = −0.22; p < 0.025). Of the entire group, 85.2% agree or strongly agree that monitoring is vital for risk management, with 71.4% agreeing or strongly agreeing that research is the key to effective monitoring. The response to statement 1, that monitoring is more important in the short term, was most positive among nonvolcanologists, with chief scientists and those with PhD disagreeing. Interestingly, these latter groups were also less positive about monitoring being more important during eruptions (statement 4), perhaps reflecting the huge volumes of research generated by eruptions such as that of Mount St Helens or the Soufrière Hills. Those with decision experience were more likely to rate monitoring as vital—perhaps also suggesting the value of monitoring preeruption. The response to statement 4 was also significantly affected by the resources variable, which would be consistent with developing countries having fewer resources to put into research during an eruption (see below).
6. Volcanologists' Behavior in Crises (Statements 5–10)
 A number of respondents, including some with extensive observatory experience, admitted that they had never read the IAVCEI Crisis Protocols document (statement 6). Responses to this statement were significant with high effect for the “resources” predictor. Categories 1 and 2 show a medium to large effect (U = 22.00, r = −0.38, p < 0.025) on the responses—hence the K-W test picked up the effect, but the J-T did not. The majority of responses for this statement were unsure or no response (58.2%). Of those who did respond, 76.9% agreed or strongly agreed.
 Statement 7 is particularly relevant to those working on active volcanoes in advisory roles. The decision to call for an evacuation may be largely dependent on volcanologists' advice, which puts them in a difficult position, particularly where local authorities tend to “blame” the scientists, thus deflecting public anger away from themselves. Responses to this question were bimodal, as were responses to statement 5 (that scientists should give purely scientific information). Interestingly, chief scientists and observatory scientists both disagreed with statement 7, and disagreement increased with level of highest degree for the whole group. The difference between technicians and chief scientists in particular had a huge effect size (r = −0.74), and difference between chief scientists and affiliated scientists (not employed by an observatory) had r = 0.35. These data suggest that responses to this question were linked to observatory experience (as opposed to academic experience), to level of observatory employment and to level of highest degree. Those ranked highest in all of these categories tended to disagree with the statement, suggesting that making evacuation decisions in best practice is the role not of scientists but of local officials. These trends were also identified as highly significant in the Jonckheere-Terpstra test, which analyses the development of the group medians with each category of predictor in the sequence.
 Statement 8 produced a highly skewed distribution over the whole group (zskew = −1.167): 93% in the affirmative. However, there are variations within the demographic, perhaps the most interesting being the highly significant impact of observatory experience (J = 4691.0, z = −2.348, r = −0.19, p < 0.025): those with experience as scientists or chief scientists at observatories felt less strongly that academic volcanologists should seek to share resources with observatories. Statements 9 and 10 address some of the communication issues faced by scientists at observatories. Statement 9 concerns the importance of trust, underlining that scientific distrust is linked to public safety in volcanic crises (65% agree or strongly agree). Statement 10 relates scientific distrust to data usage, and suggests that many volcanologists are not convinced that this is the case (only 31% agree or strongly agree), but around 44% were unsure or neutral. This reflects the highly personal nature of the question: supplementary research suggests that scientific distrust is related not only to data usage but also to individual suspicions and misinterpretations and complex interpersonal dynamics. Again, this depends on communication between scientists.
7. Observatory Management (Statements 11–16)
 These statements deal with how observatories are run, and how they engage with the population. There is a very strong feeling among respondents that observatories should be run by scientists. This is particularly strongly felt in developing countries, while those who felt that local officials should run them tended to be based in nonvolcanic countries. In practice this is closely linked to local knowledge and circumstances: in observatories that are required to make recommendations to local decision makers via their director, it is important that the director be a scientist who is able to weigh up results from different disciplines. This requires a particularly versatile scientist, who has interactional expertise (the ability to engage with multiple disciplines at a high level; Collins and Evans ). There is a roughly inverse correlation between this and level of observatory experience, and also highest degree: nonvolcanologists were more likely to support the running of observatories by local officials. However, of all respondents, only 2% disagreed with statement 11, and agreed with 12. This is in part a reflection on the group that was sampled but also a result of experience of operating at observatories.
 The respondents also felt that observatories should be government funded, which is unsurprising given the demographic and the fact that observatories are important for civil protection. The questions about outreach, however, show some variations. While respondents felt very strongly that observatories should engage with the public about risk (83% agree or strongly agree), there was also a strong feeling that risk education is the responsibility of local governments (53% agree or strongly agree, with only 12.2% disagree or strongly disagree). This statement (15) also had significant correlations with a number of other statements. Given that the dominant feeling was that observatories should be government funded, it makes sense that they then take on responsibility for risk education. The USGS, for example, is a government employer and bears educational responsibilities.
8. Qualitative Results: Implications From Interviews and Documentary Analysis
 This section discusses the findings from interviews and document analysis before, during and after the Eyjafjallajökull eruption. We review the recommendations made by the SAEE Report and the relationship between government and science. Several key areas were identified and will be discussed below with particular reference to the Eyjafjallajökull eruption.
9. Government Structures and Complexity
 As mentioned in the introduction, the government structures that are in place prior to an eruption play a significant role in determining crisis management. The UK government seeks to apply five principles in its management and communication of risk—openness and transparency, involvement, proportionality, evidence and responsibility. However, during the “ash crisis,” it was difficult to adhere to these principles. The following quotation from an Icelandic scientist demonstrates that the simplicity of systems in Iceland contributes to their effectiveness.
 “How short the communication links are here in Iceland—I guess that is what separates our community from others in similar situations. It means that there is a quite close personal contact between everybody involved so if you need to get a message to somebody you don't have to go through complicated channels of approval here and there, and then official contact between different agencies… it's usually just one phone call, and I think that's quite essential for efficient communication between scientists who are doing the monitoring, and people who have to react to it.”
 In the UK, the Volcanic Ash Advisory Centre at the Meteorological Office is responsible for issuing warnings to the aviation industry during eruptions. However, channels of advice to the UK government during the eruption were less clear (according to interviewees and the SAEE Report). Since volcanic ash was not a part of the National Risk Assessment (NRA) (document 8), advisory procedures had not been developed either by the Government Office for Science (GO-Science) or by the Civil Contingencies unit (in the Cabinet Office). The Civil Contingencies Act in 2004 specified generic procedures for dealing with emergencies that require central government response (Figure 1). Identification of the “Lead Government Department” by the Cabinet Office was an area identified as problematic by the SAEE Report. This is a key part of crisis management in the UK—an LGD has to be identified to coordinate the government's response. In the case of the ash crisis, multiple departments were affected, including the Department for the Environment, Food and Rural Affairs (DEFRA), GO-Science and the Department of Transport. That the effects on each department were very different—ranging from transport chaos to the possibility of air pollution and health impacts—complicated the crisis response.
 Responses to emergencies in the UK depend upon the magnitude of the emergency as defined in Figure 1, and different levels set in motion particular procedures. The UK government is highly complex in its structure, and its responses are tightly controlled by the civil service. While there is extensive cross-government discussion about risk assessment, the SAEE Report noted that the committee were “surprised and concerned” (p. 22) at the lack of input from GO-Science into the NRA. In part, this is because the Civil Continencies and NRA planning come under the Cabinet Office, not GO-Science. The Report recommended that GO-Science be located within the Cabinet Office. This was in fact rejected by the government, which resisted some of the conclusions of the SAEE Report, arguing that the communication links across government were not as bad as implied by the Report. This is an interesting result, because many external scientists and indeed industrial parties felt that the coordination across the government was not adequate.
 From this brief (and simplified!) summary, it is very clear that the size of the UK government, its complexity and its bureaucracy are in stark contrast with the Icelandic case. In part, this is due to the much higher population in the UK and its legal system. However, it also shows the limitations of complexity on rapid response. The fact that governments cannot prepare for every eventuality requires that a generic procedure be defined for events that require such a response. A further point made by the SAEE Report was that the government clarify means by which scientists can engage with the Cabinet Office, particularly since the Scientific Advisory Panel on Emergency Response (SAPER) was abolished. SAPER had had a centralizing role in emergency planning, but ceased operations in favor of subject-specific advisory groups (including the SAGE mechanism).
 The importance of personal contact and the long-term building up of relationships within the science-policy interface is something that was identified in the government's report into the use of academic advice (document 5). The biggest challenge in the case of the eruption at Eyjafjallajökull was the lack of government planning for such an eventuality. This meant that experts had to be identified reactively by the civil service using existing networks. It also involved activation of the SAGE system, a relatively new form of scientific advisory practice in the UK and one that received several criticisms in the SAEE Report. In the UK, resources do not allow for such a committee unless a disaster is ongoing. The SAGE was disbanded after the eruption was over, and government work on the problem was wound down as other issues arose. This also relates to the NRA, because risks in the UK are compared so that the most important in terms of both likelihood and impact are those that receive the most attention (Figure 2). The government is clearly aware that different risks are perceived in different ways by different people, and has commissioned work on this topic [e.g., Eiser, 2004]. The SAEE Report emphasized the importance of social scientific analysis and its absence from government scientific activities. This was evidenced on the SAGE, which did not include any social scientists in spite of the obvious social relevance of the work. Other factors identified in the SAEE Report included the lack of openness and transparency surrounding the work of the SAGE, which was problematic both for its members, who were unsure how much they could discuss with colleagues, and for the government in that openness has been identified as important for public confidence. The SAGE meeting minutes, published online following the SAEE Report, demonstrate that significant scientific debate and discussion took place. This would have been of interest to other scientists, and indeed to the public and aviation industry. It is also, importantly, central to the way that science works, involving a range of approaches, views and alternatives.
 Icelandic scientists noted the importance of continuity in expert advice to civil defense personnel in Iceland:
 “We have meetings regularly and often with the civil defence people to inform them of the status, not necessarily because there is something going on—just to keep them informed… also to keep this personal contact…that you actually know the persons you are talking to … and you know that they are not overreacting… if everybody is unprepared, if there is no personal contact, if there is no prior knowledge or understanding of this, people can overreact…and think that somebody is crying “wolf,” whereas he isn't, he's just making an observation.” Icelandic scientist.
 Scientific governance structures in Iceland are considerably simpler than the UK (Figure 3). This is also demonstrative of the different nature of the volcanic hazard in Iceland: there is a need for a local response as well as a central one. In the UK, the aviation hazard was sufficiently general that it required a central response. The importance of local police chiefs in Iceland was emphasized by interviewees, who commented that the interest of these individuals in understanding volcanic hazards had a significant impact on their participation in preparation exercises and drills. Again, local and personal contact was highlighted as crucial. This is in fact also the case in the UK, in that individuals play an important part in how efficiently a crisis is managed, as do local and regional networks. However, the complexity of the UK government and its departments can make the channels for advice very unclear to those outside of the civil service.
 The importance of facilitating links between government and academia has been the subject of previous reports, but the “ash crisis” in particular posed a further question: how can international expertise be integrated into assessments and international policy questions be addressed? While the SAEE Report touches on this, and subsequent material from the UK government's work on the Fukushima nuclear disaster also raises these issues, they require a response beyond any single government. In the evidence collected by the UK government for the SAEE Report, both written and oral, the question of “international coordination” was raised. Many respondents felt that this was a key problem at the policy level within Europe. This is an important question that is worthy of further discussion [e.g., Sparks, 2007].
10. Data Usage and Distribution: Scientific Networks
 While the scientific community has limited ability to effect such changes, the 2010 eruption of Eyjafjallajökull did demonstrate the importance of openness and debate within academia—a number of workshops and conferences took place in order to investigate ash hazard (e.g., see www.wovo.org). On the subject of transparency, one scientist commented:
 “There is a very remarkable step … and this is the webpage of the weather office… they put everything on there, everybody in the whole world can go there and see what's happening, and make interpretations if they like, and this has become very popular… I think this is a very positive thing. Besides being extremely popular, it also increases the trust – people know that no information is being held back.” Icelandic scientist, referring to the Icelandic Meteorological Office website.
 This is a controversial statement, in that many scientists are naturally protective of their data, either because the public might misinterpret it, or because its collection required time and patience, and they are concerned that someone might take it and publish it without acknowledgment. These concerns are part of the experience of doing science, and take on new dimensions when science engages with the public. This can involve cultural considerations as well as practical and moral ones. However, it is also worthy of discussion in the volcanological and hazards fields (see Bird et al. , for a full discussion of the IMO website).
 Data availability is also important in the assessment of risk, particularly if subjective probabilistic methods are used [e.g., Hincks, 2005; Marzocchi et al., 2004, 2008]. These methods seek to use multiple data sets to assess the risk from active volcanoes, and involve a level of expert judgment (e.g., in selecting prior probabilities, or providing a set of quantitative answers to questions). The SAGE, for example, carried out some analyses using expert elicitation. This was aided by the availability of data from the Icelandic scientists, and the transfer of data between Icelandic and UK Meteorological Offices. Similarly, scientists involved in media work were also helped by the availability of data on the Internet.
11. Public Preparation and Awareness
 Much important work on public risk perception and communication in Iceland has been carried out by Bird et al. [2010, 2011], particularly concerning the risk to tourists. Here, we are focusing on scientific management of the eruption and the uncertainties involved. The communication of the uncertainty inherent in volcanological science is a particularly important aspect of risk management.
 “Even if scientists disagree, we don't necessarily keep that away from people – there are things that we disagree about, and it's something we reveal most of the time so people know what the margins are.” Icelandic scientist.
 Uncertainty and risk differ—risk is generally regarded as quantifiable, at least in principle, whereas uncertainty may not be [e.g., Stirling, 2007; Donovan et al., 2012b]. UK interviewees expressed concern that the public regard science as a source of certainty and do not appreciate the uncertainty and error that are necessarily involved in assessing the natural system. This is clearly amplified during a “crisis” that was unexpected by the majority. Explaining the reaction to the 2000 earthquakes in South Iceland, one scientist commented:
 “We started talking about this in the 1970s already – that we had to prepare for a sequence of earthquakes as had happened in the past. So there was sort of this building up of the prediction, gradually, to prepare the population for this. And whereas everybody started out being rather angry about this – we were disturbing their peace, making them afraid and so on – they gradually got used to this and the kids got used to it, they learned about it in school, and gradually people were actually quite well prepared in 2000, so their reaction … was not so much panic, it was more like, well, this is just an earthquake.” Icelandic scientist.
 All interviewees asserted that Iceland is culturally prepared for geological hazards in a way that the UK has not had to be in recent years—and public expectations of convenience in travel and mobility has increased dramatically [e.g., Adey et al., 2011]. The role of cultural cognition in risk judgments requires further research [e.g., Kahan et al., 2009]—expectations, emotions and lifestyle factors all affect worldviews, and these may be threatened by risk.
12.1. Role of Scientists in Decision-Making
 The survey carried out on the global volcanology email list in 2008–9 demonstrates that volcanologists are divided concerning their responsibilities in advisory contexts (Table 2). A closer look at the data suggests that those with experience in decision-making during a volcanic crisis are less likely to think that recommending evacuations is inevitable; this is consistent with normative scientific advisory guidelines, which would argue that scientists are there to advise while policymakers and elected politicians make the decisions, often in accordance with the precautionary principle. While the precise position of scientific advisors in governments is usually derived from the structure of the government itself, the nature of volcanic events can result in the breakdown of official structures because the dependence on the scientists is so high and the impact of the events can destabilize infrastructure. In the case of the UK, the fact that the 2010 eruption was unexpected rendered its management complex, as new structures had to be put in place. In developing such structures, the specificities of volcanic eruptions (and noneruptions) have to be taken into account: eruptions may be long lived, and rarely exhibit deterministic behavior [Sparks, 2003]. This means that the nature of the advice may change during the eruption, as scientific confidence rises and falls, and this in turn may affect the expectations of stakeholders – including government officials. In Iceland, however, a committee of scientists meets with government officials semiregularly, and relationships have been built up over a number of years. This means that people know that they can trust one another, and know what kinds of information they can expect.
 To give a further example, countries that have to manage chronic eruptions, such as on Etna or Kilauea, have of necessity developed advisory structures. In Italy, for example, the Istituto Nazionale per la Geofisica e Vulcanologia (INGV) provides scientific advice to the Protezione Civile, which then takes the necessary steps in emergency management and decision-making. However, these two structures remain under review, and have undergone several stages of evolution over the past few years, with the combination of volcanology and seismology occurring in 2000, and then being redistributed thematically and spatially. The advisory structures, and in particular the demarcation of the advisors and the managers via INGV and PC, respectively, is currently a source of some debate.
 In the UK, there is no permanent arena for planning for these events, even though globalization and economic networks are increasingly rendering natural hazards transnational—for example, through aid packages, commerce and NGOs. The threat of the “unknown unknowns” can be reduced in part by importing knowledge from other cultures— including expert groups. This requires channels for knowledge economies, well recognized in other disciplines—such as the life sciences—but as yet poorly developed for geophysical hazards. Scientific governance [Irwin, 2008], and risk governance [Renn, 2008] are much discussed in science and technology studies, acknowledging that the demarcation of science, risk and politics is increasingly problematic [e.g., Irwin and Wynne, 1996; Rayner, 2003]. Recent work [e.g., Stirling, 2007] has demonstrated that even the separation of risk assessment and risk management may be an oversimplification, and this requires analyses of the use of the precautionary principle in the former as well as the latter [see also Wynne et al., 2007]. Recent techniques for risk assessment on volcanoes would seem to support this: expert elicitation techniques, for example, cannot guarantee that the precautionary principle is not implicit in the values provided by at least some of the experts, and scientific reports can be worded to urge caution in governance. The line between science and policy making is unclear even in very specific situations, let alone in the complexities of international expert advice.
 The UK government is currently carrying out a study of “black swan” events, in the knowledge that these events are difficult to govern. The ash crisis, as far as many people were concerned, was such an event. Analysis of the SAEE Report, and discussions with interviewees, demonstrated that although the SAGE was mobilized rapidly and incorporated a high level of expertise, the political mechanisms for its formation were unclear even to those involved. Oppenheimer  noted that scientists had repeatedly sought to inform the UK government about the ash threat from Iceland. In this case, there seems to have been a breakdown of communication between academia and government – in spite of efforts on both sides (e.g., document 5). This raises an important question for scientists: how can scientific research have impact at the policy level prior to a crisis? Networks are important in this: personal relationships developed between scientists and decision makers over long periods of time, even if they are not working together frequently. Scientists can also register interest with government science offices, some of which keep records of experts who can be contacted if their expertise is required. It is also apparent, both from the survey results and interviews, that involvement in the policy process is time consuming. It may also not be rewarding – particularly in terms of scientific publications, by which scientific careers are judged. There is perhaps a challenge here to national and international government agencies to provide a form of recognition for scientists who are involved in this way. The implementation of policy fellowships is one example of this. In light of the L'Aquila court case – seismologists being tried for manslaughter for failing to predict the 2009 earthquake [e.g., Bojanowski, 2011; see also Geller, 1997]—it may be that legal protection for scientists should also be provided by governments. These events clearly demonstrate the challenge for scientists of balancing a social role with the uncertainties of geophysical hazards.
12.2. Quantitative Risk Assessment and Uncertainty
 At present, many countries do not carry out quantitative risk assessments on active volcanoes because the uncertainty is too high. Some scientists express an ambivalence toward the statistical methods put forward to deal with the management of uncertainty in volcanology – a phenomenon also witnessed in the seismological literature [e.g., Castaños and Lomnitz, 2002]. Further discussion with interviewees concerning expert elicitation [e.g., Aspinall, 2010] and Bayesian methods [e.g., Marzocchi et al., 2004; Newhall and Hoblitt, 2002] suggested that their empirical weakness was a key factor in generating scepticism, but that the methods are the best available options for providing meaningful, quantitative advice on volcanoes. Importantly, the value of discussion and reflection throughout the expert elicitation procedure was noted. One source of anxiety, however, concerns the nature of probabilities: the definition of being “right” or “wrong” is not straightforward. As De Finetti  observed, “probability does not exist”: it is a concept that has been designed to aid human reasoning around uncertainty [see also Spiegelhalter and Riesch, 2011]. Probabilistic assessment allows the inclusion of aleatory and epistemic uncertainties, giving it an advantage over deterministic methods. However, some scientists remain cautious about putting numbers on risk assessments, and there is some evidence that probabilities are poorly understood, even by scientists [e.g., Gigerenzer et al., 2005]. To some extent, too, this may be culturally determined: some countries are far more rigorous in their use of risk assessments than others.
 The UK National Risk Register (NRR) is at the center of the government's management of emergencies. It categorizes different risks according to their likelihood and impact. Risks are identified across the government and put forward for inclusion. This involves the government assessing the level of each risk. If a risk is very high impact—even if it is unlikely—then contingency plans are made. The impacts may range across departments, as occurred in the “ash crisis.” The importance of quantitative—or at least semiquantitative—risk assessment is thus evident from the political structures in place. Policymakers want to know how much they should invest in preparing for particular events. This requires the use of decision metrics [e.g., Marzocchi and Woo, 2009; Aspinall, 2006, 2010]. A useful comparison might be with seismic hazard analysis, where the vast majority of assessments are probabilistic, because that is more politically useful than a purely deterministic assessment [e.g., Jordan et al., 2011]—even if the latter has some scientific advantages [e.g., Castaños and Lomnitz, 2002; McGuire, 2001]. The implication from considering the interface between scientists and policymakers is that in order to bridge the gap between scientific uncertainty on one hand and political requirements on the other, some estimation of likelihood is necessary. While probabilities are frequently not empirically falsifiable in the short or even medium term, they have a political expedience. Subjective assessments may represent the best possible knowledge of the physical system.
12.3. Public Education and Outreach
 The survey respondents also recognized that responsibility for public education is shared by local authorities and observatories. This was felt very strongly by interviewees from Iceland, who described the importance of outreach activities in helping communities to cope with the earthquakes in South Iceland in 2000. The role of scientists in education is implied by the findings of multiple writers [e.g., Haynes et al., 2008a; Bird et al., 2009] that scientists are generally well trusted by the public. A further comment from interviewees was that some cultures are more effective at dealing with uncertainty than others [e.g., Paton et al., 2008]. While this has yet to be rigorously tested, it is an interesting idea in the present context because it affects how eruptions are managed locally. Donovan and Oppenheimer  note that on Montserrat, 15 years after the onset of the eruption, Montserratians frequently note that “the volcano is very uncertain.” Several identified the communication of this fact as pivotal in their own personal management of the crisis [see also Haynes et al., 2007]: they stopped expecting the scientists to predict the volcanic activity and started trying to live with the uncertainty. The implication of this is that open acknowledgment of the uncertainties inherent in scientific processes is an important factor both in building trust and in preparing people [e.g., Krebs, 2011]. Science in the “risk society” [Beck, 1992] is simultaneously a source of uncertainty and of the potential for its reduction.
 In the unprepared UK, flight disruption caused widespread unease and even resentment, while in South Iceland the ash was several centimeters thick and continues to impact on farmers' livelihoods over much longer time periods. The key differences are that volcanoes are an integral part of Iceland's national culture, Icelanders are prepared for eruptions [Bird et al., 2009, 2010, 2011; Donovan and Oppenheimer, 2011], the definition of the risk was clearer in Iceland (less dependent on models) and the governmental structures for scientific advice in volcanic crises (and, crucially, between eruptions) are well established. In the context of the international impacts of the 2010 eruption of Eyjafjallajökull, this raises a further issue: to what extent do volcanologists in, for example, nonvolcanic northern Europe have roles in public outreach? This is a particular issue where there is no obvious institution to take the lead—as volcano observatories might in a local context. In the UK, for example, steps have been taken by several organizations following the 2010 events, including large collaborative ventures involving both government agencies (such as the BGS and the Met Office) and universities.
12.4. Governing Risks From Volcanoes
 The UK's recent SAEE report acknowledged a number of shortcomings in political structures and reactive governance, but also raised a further question: how can governments prepare for “black swan” events? This is currently the topic of a further review by the office of the Chief Scientific Advisor. Yet between the eruption in 2010 and that in 2011, very little was done to aid the scientific community to prepare for future eruptions, or to clarify procedures by which scientists can alert the government about new techniques, methods and models. Similarly, little was done to aid the public to deal with the uncertainty surrounding volcanic events in 2010/2011. As the SAEE report demonstrated, the fact that the work of the Scientific Advisory Group in Emergencies (SAGE) was kept secret does nothing to aid the process of knowledge transfer. The SAGE met four times between April and June 2010, and its minutes have been published following the SAEE report. The reasons for the original confidentiality of the group are unclear. In giving evidence, one scientist on the SAGE noted:
 “There was a lot of information discussed in SAGE which was not, for any reason, secret. It was about the way volcanoes work, the way meteorology works. All of this information should have been shared as widely as possible, as quickly as possible” [House of Commons, 2011].
 Much of the material handled by the SAGE concerned scientific models of the plume and the weather patterns affecting it. They also discussed risk assessments for Icelandic volcanoes more generally. As noted by the SAEE report, and by scientists involved with the SAGE, some of the research generated by its discussions could have benefited from broader involvement and greater clarity on how much information could be shared with colleagues. One further ramification of the lack of openness was the prevalence of rumor among the media; had the SAGE minutes been available, they would have provided a source for robust scientific information about the state of knowledge and could have reassured travelers that the government was consulting relevant expertise. However, the SAEE report demonstrated that the government's use of the SAGE mechanism requires further refinement to maximize its usefulness:
 “Because of the CAA's groundwork and the relatively late formation of SAGE during the volcanic ash emergency, it appears that SAGE contributed little to scientific understanding of the key issue: the ash tolerances of engines and aircraft. We question how much additional knowledge SAGE added to enable airspace to be reopened” [House of Commons, 2011].
 During 2010–2011, many UK scientists carried out research into the Eyjafjallajökull eruption, using models, geochemical analyses and atmospheric measurements to improve scientific understanding of volcanic ash. International conferences were held in Iceland to facilitate communication between the airline industry, the ICAO and the scientific community, and many of these efforts are continuing (e.g., see www.wovo.org). The fact that a similar event, larger in magnitude and intensity, occurred in 2011 should provoke airlines to rethink their investment in scientific research. However, it should also provoke a review of societal reaction to uncertainty especially natural events such as volcanoes, earthquakes and floods. Following the 2010 and 2011 ash “crises,” the UK government has started to plan for future events, including a worst case scenario, based on the Laki eruptions in 1783–5 (according to interviewees). This includes contingency plans for long-term travel chaos and for potential health impacts.
 A further question that arises from this research, however, is more challenging. Volcanic ash is a transnational hazard: ash clouds do not respect human boundaries. While there are a range of international volcanology networks including the Commissions of IAVCEI, these are not clearly translated at the international level of governance. There are perhaps some steps toward this, and there are emerging paradigms in other fields, of which the Intergovernmental Panel on Climate Change is an example. Others include the International Strategy for Disaster Reduction. However, during the 2010 eruption there was no intergovernmental organization for scientific advice and decision-making at the policy level. This would have included many of the scenarios examined by the SAGE in the UK—such as assessment of the different eruptive scenarios and their likelihood—as well as broader impact assessments across the affected region. It could also have provided a focus for scientific and social scientific outreach.
 Eruptions are “epistemological acts”: they change the directions of knowledge-production in volcanology (G. Bachelard, cited by Foucault ). The 2010 eruption of Eyjafjallajökull volcano demonstrated a number of shortcomings in the UK government's risk assessment and management. It also showed the need for integrated international approaches to risk governance regarding volcanic hazards, which do not adhere to national boundaries. The social context of volcanology influences its practice, and social scientific methods can be applied to elucidate several aspects of this: the philosophy and sociology of scientific enterprise; the social and cultural perceptions of science, risk and uncertainty; and the political implications. The role of scientists as advisors to governments is likely to continue to gain in importance as risks are identified and attempts are made to mitigate them. This puts the impetus on scientists and social scientists to discuss openly both the potential and the limitations of their work. This paper concludes the following: (1) relationships between scientists and decision makers are most effective when formed prior to volcanic crises; (2) different types of uncertainty and complexity that arise in volcanic crises result both from the physical system (the complex dynamics of the volcano), and from the social and political systems—the structure of science and interactions between scientists, and also the structures of governments at local, national and international levels; (3) complex governance structures can hinder both decision-making and public confidence; (4) there is a need for an international body to coordinate intergovernmental responses to events similar to the ash crisis (another example might be the Fukushima events); (5) the provision of incentives (such as fellowships) and legal protection might aid governments seeking scientific advice; (6) bridging tools, such as expert elicitation and Bayesian methods, may have an important role to play in providing politically relevant assessments based on the best science available; (7) reactive advisory structures (like the SAGE) can be effective in a crisis if they are transparent and their procedures are clear.
 A.D. acknowledges a NERC-ESRC PhD studentship. Colleagues at the Institute of Earth Sciences and the Icelandic Met Office are thanked for their assistance during fieldwork. Special thanks to Ágúst Gylfason for his help with Figure 3. Colleagues at DEFRA and in the UK scientific community are also thanked for discussions. Respondents to the survey are thanked for their time. Gill Jolly and Katharine Haynes are thanked for their very thorough reviews, which greatly improved this manuscript.