Communicating Risk in Glacier Tourism: A Case Study of the Svínafellsheiði Fracture in Iceland

Every day in early summer 2018, an estimated 1000 tourists went on guided tours of Svínafellsjökull, an outlet glacier in southeast Iceland. However, this changed on 22 June 2018, when a warning was issued against glacial travel due to the risk of a large landslide caused by a fracture in the surrounding mountainside. Tourists often entrust tourism employees with responsibility for their safety; however, there is a dearth of research into the ways in which tourism employees receive and respond to risk communication. These dynamics were explored in this ethnographic study, which drew on 50 semistructured interviews and extensive participant observation. The results indicate that despite demographic shifts, Icelandic inhabitants remain the basic unit on which risk management processes are centered, with repercussions for the ways in which exposure is calculated and risk is communicated. Tourists and tourism employees have a limited understanding of the risk and emergency protocols compared with local inhabitants. We argue that, for their own safety and that of customers, risk communication needs to be tailored to the needs of tourism employees, including guides and hospitality workers. The recommendations that emerge from this research can guide risk communication strategies in other mountainous regions of the world where tourism is an important source of livelihoods.


Introduction
Given the natural beauty and accessibility of Svínafellsj€ okull (64.01868N; 16.82158W), an outlet glacier of Vatnaj€ okull in southeast Iceland, it is unsurprising that it emerged as one of the most popular destinations in the country for glacier hiking (see Figure 1). During the 2017 summer peak season, an estimated 1000 tourists went on guided tours of the glacier each day. However, this changed on 22 June 2018, when local police together with the Department of Civil Protection and Emergency Management (DCPEM) issued a warning against travel on Svínafellsj€ okull due to the risk of a large landslide caused by a fracture in the surrounding Svínafellsheiði mountainside. Guided glacier tours were discouraged, and tourists were ''advised to stop only for a short while at viewpoints by the glacier tongue'' (IMO 2018). In the days that followed, all guiding operations shifted to nearby glaciers.
The emergence of the Svínafellsheiði fracture is part of a global trend whereby climate change has contributed to the decline of glaciers, snow, and permafrost in recent years (Hock et al 2019: 133). This has led to changes in the frequency, magnitude, and location of natural hazards, as well as the emergence of landslide risk where there was no record of previous events (Hock et al 2019: 133-134). People are also increasingly exposed to these hazards due to growing populations and tourism (Hock et al 2019: 133).

Background Risk communication
Risk communication is widely accepted as a key strategy for mitigating vulnerability and reducing the effects of disasters (Paton 2006). The term ''risk communication'' refers to interactive flows of information to notify people of the probability of a hazard occurring, likely consequences, and mitigation actions (Plough and Krimsky 1987: 6). Since the 1980s, the theory and practice of risk communication have shifted from a top-down flow of information from experts to the public to a broader approach that takes into account community participation, cultural factors, social relationships, and trust (Khan et al 2017;Lin et al 2020).
Risk communication strategies are more effective when tailored to intended audiences (Paton 2006: 9;Seeger 2006). Factors that affect how people engage with, and act upon, risk communication include, but are not limited to, cultural background (Lindell and Perry 2004), trust in risk management authorities (Haynes et al 2008), confidence in emergency procedures (Barberi et al 2008), duration of exposure to risk (Stancu et al 2020), and level of social capital (Cadag et al 2017;Anderson-Berry et al 2018). Social capital refers to the networks of relationships among people who live and work in a particular society. As communities become more diverse, the social context in which information is received is characterized by increasingly varied experiences, beliefs, needs, expectations, and interpretations of risk (Paton 2006;Lin et al 2020).
Participatory risk communication involves community members engaging in two-way dialogue with risk managers; it also provides an opportunity for both groups to raise and understand concerns (Pelling 2007;Cadag and Gaillard 2012;Demeritt and Nobert 2014: 319;Cadag et al 2017). When local actors are not familiar or do not recognize response and evacuation plans in advance, these protocols are less likely to be followed during an emergency (Voight 1990;Carey et al 2012). A sizable body of literature indicates that community participation improves the implementation and quality of risk management (eg Pelling 2007; Cadag and Gaillard 2012; Cadag et al 2017; Kerstholt et al 2017). One criticism of participatory risk communication is that it reinforces existing power dynamics within a community, for example, through the exclusion of marginalized groups or disadvantaged individuals (Cadag et al 2017: 484;Nguyen et al 2017).

Risk communication in Iceland
Situated in the North Atlantic Ocean, Iceland is prone to a range of hazards, including avalanches, volcanoes, earthquakes, landslides, floods, and extreme weather. DCPEM is responsible for preparing for and managing risks due to natural hazards (Parliament of Iceland 2008). At the local level, police are responsible for developing and implementing preparedness and response strategies together with Regional Civil Protection Committees (Parliament of Iceland 2008).
In Iceland, information about risks and emergency protocols are shared through different media, including broadcasters, public meetings, email, telephone calls, brochures, face to face, government websites, social media, the Safe Travel website, and the 112 (national emergency number) mobile application (Bird and Gísladóttir 2020). Public meetings conducted by risk managers with presentations by scientists have been espoused as a particularly effective way to communicate information to local residents, listen to their concerns, build trust, and learn from local knowledge (Guðmundsdóttir 2016: 37). Emergency response and evacuation drills have also been trialed for some volcanic hazards (Bird et al 2010). Authorities typically rely on risk communication rather than access restrictions to promote personal safety in Icelandic glacial, volcanic, and geothermal landscapes (Bird and Gísladóttir 2020).

Risk communication in tourism
Both tourism and risk management literatures indicate that the tourism sector globally is poorly prepared for disasters related to natural hazards (eg Prideaux et al 2003;Hystad and Keller 2008;Bird et al 2010). Reasons include unclear communication by emergency management authorities, a lack of formal consideration of tourism in risk management, poor adoption of risk management protocols by tourism businesses, and a passive approach by tourism businesses toward risk management (eg Glaesser 2003;Cioccio and Michael 2007;Hystad and Keller 2008;Becken and Hughey 2013).
Tourists visiting hazardous environments have different perceptions of and aversions to risk. Visitors to volcanic sites often: adopt their own precautionary measures; hope nothing dangerous will happen when they are in the area; rely on having enough time to get to a safe location; and expect hazards will be managed by tourism operators (Erfurt-Cooper 2010: 6;Purdie et al 2015: 198). However, studies of volcanic tourism indicate tourists are not always provided with sufficient information to make informed decisions about their personal safety (Knafou 2019;Bird and Gísladóttir 2020). This was the case for the Whakaari/White Island disaster, when 21 tourists and guides died in a volcanic eruption in December 2019, despite an increased eruption alert (March et al 2020). Tourism operators may fear that communicating risk with tourists will lead to a reduction in demand and associated profits (Bird and Gísladóttir 2020); however, a reputation for unsafe practices can have the same effect (Purdie et al 2015: 197). There have been calls to improve risk management within the tourism sector, including tourism disaster spokespeople, dedicated web pages, disaster drills, mainstreaming risk management into training, and tourism-oriented risk communication strategies (eg Hystad and Keller 2008;Bird et al 2010;Becken and Hughey 2013;Mair et al 2016;Aliperti and Cruz 2019).
Iceland has experienced a dramatic increase in tourism in the last decade, from 448,000 foreign visitors in 2010 to over 2 million in 2019 (Icelandic Tourist Board 2018). By 2019, tourism was the most important source of foreign currency earnings and the largest provider of jobs in the country (M€ uller et al 2020: 51). The number of foreign tourism employees rose from 2427 in 2008 to 10,551 in 2019 (Halldórsdóttir and J ulíusdóttir 2020: 13). An estimated 30% of staff in the sector-and 75% of staff in hotels and guesthouses-are not from Iceland (Statistics Iceland cited in Wendt 2019: 27). Foreign workers are often employed in low-paid, low-skilled jobs, on short-term or temporary contracts (Skaptadóttir and Wojtynska 2019). Foreign tourism employees typically have strong social ties with their coworkers but few connections to Icelandic society beyond their employer (Wendt 2019;Halldórsdóttir and J ulíusdóttir 2020). English is often the main language of communication (Mirra 2019), with almost half of all foreign tourism employees reporting inadequate or nonexistent Icelandic language skills (Hauksson 2019).
The few studies available suggest that tourism employees and foreigners are commonly overlooked in risk communication. In Japan, researchers found that disaster management structures were not well adapted to the needs of foreigners, for whom language barriers and inexperience with emergency protocols can lead to panic and confusion during disasters (Sakurai and Adu-Gyamfi 2020). In Iceland, Bird et al (2010) found that tourism companies operating in an area with significant volcanic risk (63.41868N; 19.00608W) had not provided emergency training for their employees.
Glacier guides constantly observe and respond to hazards, including steep and slippery terrain, crevasses, rockfalls, extreme weather, and river crossings (Purdie et al 2015). In addition, climate change has made glacier tourism in some regions more dangerous and challenging for both clients and guides (Purdie et al 2015;Salim et al 2021), adding new and large-scale hazards to the existing risks.

Research design Methodology
This article is based on ethnographic fieldwork conducted in theÖraefi district of Iceland between August 2018 and April 2021 (see Table S1, Supplemental material, https://doi.org/10. 1659/MRD-JOURNAL-D-21-00051.1.S1). Fifty semistructured interviews were conducted with 52 people (25 female; 27 male) identified through purposive sampling. The participants included 14 Icelandic local inhabitants, 9 foreign glacier guides, 9 tourists, 7 foreign inhabitants working in the tourism sector, 3 risk management experts, 3 scientists, 2 municipal government officials, 2 tourism experts, 2 search and rescue coordinators, and 1 park ranger. While some tourism employees in the area were Icelandic, most were foreigners. This research focused on the latter, with comparisons made between tourists and local inhabitants. While most interviews were conducted with individuals, 5 were conducted with 2 people at the same time, and 1 interview was conducted with 4 people together at their request. Some participants were interviewed twice. All interviews were conducted in English, except 1, which was held in Icelandic, with a translator. The interviews typically took between 60 and 90 minutes and were conducted face to face (47 interviews) or online, due to regulations related to the coronavirus-19 (Covid-19) pandemic (3 interviews).
The core issues covered in the interviews with people living or working inÖraefi included their role in the community, understanding of the hazard, perception of risk information, and involvement in risk management. The questions were open-ended to allow important issues, perceptions, and ideas to be raised and discussed. The interviews were recorded, transcribed, and analyzed using QSR NVivo 12t. The results were analyzed through a bottom-up, inductive approach to allow themes, commonly held views, and connections to emerge from the data (Hammersley and Atkinson 2007: 153). Data coding was initially open to facilitate the identification of themes and categories, but it became increasingly focused over time (Esterberg 2002: 160). In some cases, further interviews were conducted to shed light on a specific topic or to saturate identified categories (Hammersley and Atkinson 2007: 158). To ensure anonymity, interviewees were broadly categorized and coded as local inhabitants (LI), foreign inhabitants excluding glacier guides (FI), glacier guides (GG), municipal government authorities (MG), scientists (S), risk managers (RM), tourism experts (TE), and tourists (T).
The interview findings were triangulated with data gathered through participant observation conducted during 2 scientific monitoring missions (October 2018 and August 2019), 3 formal public risk briefings (October 2018), and 12 additional study trips to the field site. The first author conducted active research while working as a glacier guide in the community between April and October 2019.

Study site and overview of risk management
Svínafellsj€ okull is an outlet glacier ofÖraefaj€ okull, an icecovered volcano that extends south from the massive Vatnaj€ okull ice cap, in theÖraefi district of southeast Iceland (Evans 2016: vi; see Figure 2). Between 1890 and 2010, Svínafellsj€ okull retreated approximately 800 m and decreased in volume by 30% (Hannesdóttir et al 2015: 567). This is part of a global trend of recent decline in glaciers due FIGURE 2 Location ofÖraefi district showing Svínafellsj€ okull and the approximate location of the fracture in orange. (Adapted from Roberts and Gudmundsson 2015: 18) to climate change (Hock et al 2019: 8-42). As glaciers retreat, they provide less buttressing support for oversteepened valley flanks, leaving them susceptible to failure (Seneviratne et al 2012: 189;Hock et al 2019: 159).
The fracture in the Svínafellsheiði mountainside is understood to be 1.7 km long (see orange dots in Figure 2 for approximate location) and widening at a rate of up to 1.3 cm per year (Saemundsson et al 2019). The potential landslide volume is estimated to be from 60 to 100 million cubic meters (Saemundsson et al 2019). A large landslide resulting from the fracture could fall up to 500 m onto the surface of the glacier, with glacial ice incorporated into the body of material moving downhill. The Icelandic Meteorological Office (IMO) has warned that such a slide may sweep water from the proglacial lake ''creating a fast-flowing slurry of rock, ice, water and even air'' (IMO 2018). The risk to downhill settlements and infrastructure is predicted to increase over coming decades as the proglacial lake grows (IMO 2020).
Recent decades have seen dramatic changes as the community shifted from a dependence on agriculture to large-scale tourism (Welling and Abegg 2019). The population ofÖraefi was reported to be 151 people in 2018 (Statistics Iceland 2019); however, this does not include most non-Icelanders living and working in the area. Until June 2018, Svínafellsj€ okull was one of the most important sites for glacier tourism activities in the country (Welling et al 2020: 4). Some 37% of tourists who visited the area joined a guided glacier tour, while 76% viewed glaciers at a short distance (Welling et al 2020: 8).
The neighboring hamlet of Freysnes (63.99078N; 16.89698W) and the westernmost sections of Svínafell (63.97928N; 16.89138W) are potentially at risk from the hazard. Freysnes lies approximately 800 m southwest of the proglacial lake, while Svínafell lies 600 m southeast (see Figure 2). Freysnes consists of 17 buildings, including a hotel, petrol station, cafeteria, search and rescue coordination center, farm, and several houses; it is traversed by the country's main highway, referred to locally as the Ring Road. In 2018, approximately 80 people lived in Freysnes, including 25 Icelanders and 55 foreigners working in tourism-oriented hospitality jobs (LI.1). Most foreign tourism employees were from central Europe, including the Czech Republic, Hungary, Poland, and Slovakia. People working in these positions were provided with accommodation and lived full-time inÖraefi during the peak season.
In June 2018, before operations shifted to other glaciers, 4 main companies conducted daily commercial glacier walks on Svínafellsj€ okull. Typically, in peak summer season, an estimated 1000 tourists went on the glacier every day (GG.2). Some 80 to 100 glacier guides and support personnel lived in temporary accommodation in Skaftafell (64.07048N; 16.97528W), 6 km northwest of Freysnes. Most glacier guides were from countries with strong mountaineering traditions. While some glacier guides were highly skilled and experienced, most had less than 2 years of glacier guiding experience. Some guides were based in the district permanently, while others worked on shifts of up to 2 weeks. There was a high staff turnover rate among glacier guides, with many only working 1 or 2 seasons.
On an average day in summer 2018, an estimated 1500 people spent time in the area exposed to the Svínafellsheiði hazard, including 1000 on glacier tours or staying at the hotel, 100 foreign tourism employees, 25 local inhabitants, and 1200 people passing on the road, many of whom also participated in glacier activities (Matti andÖgmundardóttir 2021). The vast majority of people living in the at-risk area were foreign tourism employees. When glacier tours shifted from Svínafellsj€ okull, they still passed areas exposed to the risk for approximately 15 minutes each way on the road. As a result, the overall number of people exposed to the hazard on an average day remained similar, but the level of risk decreased significantly.
Planning and risk management related to the Svínafellsheiði fracture were coordinated by local police, DCPEM, and Regional Civil Protection Committees, with hazard monitoring conducted by the IMO and University of Iceland. Risk communication activities took the form of emails and risk meetings with local residents, as well as announcements in the media, on government websites, and through the Safe Travel website (www.safetravel.is). Large warning signs were erected on access routes ( Figure  3), and information boards were set up at viewpoints. During the first year of this research, businesses pushed back on the idea of displaying information about the fracture at accommodation venues, fearing it would discourage tourists. Boards were later developed and displayed by tourism businesses in collaboration with risk managers.
Risk meetings inÖraefi were conducted with differentoften overlapping-audiences. These included closed meetings with members of the community living in the immediate vicinity, meetings of the Regional Civil Protection Committee, and townhall briefings open to the public. In addition, in December 2020, a meeting was held for managers of tourism businesses operating in the area. Table  S2 (Supplemental material, https://doi.org/10.1659/MRD-JOURNAL-D-21-00051.1.S1) lists the public meetings conducted in the area or online (due to Covid-19 restrictions) during the study period. All meetings were conducted in Icelandic (correspondence with risk manager, 8 June 2021).
A large landslide from Svínafellsheiði could fall with little warning (S.2; RM.2). This has complicated emergency protocols, including evacuations. A phone alert system allows DCPEM to send emergency alerts to all mobile phones in a given area (RM.2). However, a risk manager warned that the landslide may have ''already collapsed, and when the message arrives, it's too late'' (RM.2). At the time of writing, these messages were only available in Icelandic and English, but a system is being developed to allow translations in up to 10 languages (RM.1). In 2018, there were discussions about setting up a siren system in high-risk areas; however, as of September 2021, no such system had been established. Experience from New Zealand suggests that using sirens as a warning system can ''confuse or frighten tourists without providing clear instructions for what to do'' (Becken and Hughey 2013: 81). As of early 2021, the emergency protocol was to exit the area either to the east or west by vehicle or to take shelter in Freysnes. With the risk set to escalate in coming decades due to the expansion of the lake, taking shelter in Freysnes will no longer be an option. There was no written response plan available.

Focus on Icelandic inhabitants
Foreign tourism employees and tourists represented the vast majority of people exposed to the risk. However, risk managers and scientists initially underestimated the number of people exposed from both groups (GG.5; LI.3; S.5). For example, a guide recalled how a scientist appeared ''surprised'' during a public briefing in early 2018, when he was informed about the daily number of tourists visiting Svínafellsj€ okull (GG.5). The guide explained the scientist had estimated there to be ''20, 30, or 40 people on the glacier every day,'' but the guide clarified that ''in the summer there can be hundreds and hundreds of people each day'' (GG.5).
In an interview, one scientist referred to there being 100 people on the glacier each day, which still drastically underestimated the scale of glacial tourism (S.1). Another interviewee explained that ''foreign workers are a group of people that definitely get left out in decisions and information about these kinds of things.'' She believed that it was because ''people outside the area don't realize how many people live here'' (LI.3).
In later public meetings, more accurate figures were cited; however, the focus on Icelandic inhabitants persisted. In late 2018, a risk manager described theÖraefi district as a ''sparsely populated area'' (RM.3); while another estimated the people living close by the hotel at ''about 10'' (RM.2). In both cases, these estimations appeared to only account for Icelandic inhabitants. Foreign tourism employees, who constituted the largest demographic residing in the area, were not included. A scientist reflected that ''maybe risk managers should have looked at the demographics of people'' in the area before making risk management decisions (S.5). Several interviewees reasoned that the risk management system in Iceland has not adapted to the expansion of the tourism sector over the preceding decade (GG.2; GG.3; LI.1; S.5). One scientist explained that ''it's a relatively recent phenomenon that foreign workers come here to work in tourism; it started on a larger scale 5 or 6 years ago. I think they are not taken into consideration, not because risk managers don't want to include them, but because the bureaucracy is lagging behind'' (S.5).
The focus on Icelandic local inhabitants was also evident in risk management policies and funding structures. In Iceland, acceptable risk for floods and avalanches is calculated based on time spent in different residential and commercial buildings at risk (IMO 2020; Parliament of Iceland 2000). The applicability of this system to tourists is undermined by the different behavior patterns of tourists. The government had not established a policy on acceptable risk limit for tourists; it was also unclear how risk was evaluated for temporary foreign tourism employees (IMO 2020). Local-level policing remained staffed and funded based on the Iceland population, despite the impact of tourism on the number and diversity of people in the area, leaving police severely understaffed and struggling with an immense workload (RM.2; RM.4). Risk managers acknowledged that procedures should change in light of the influx of tourism, including the review of emergency plans across the country (RM.2).

Understanding risk and emergency protocols
No tourist interviewed was aware of the risk of a large landslide falling onto Svínafellsj€ okull. Of the 8 tourists interviewed at the Svínafellsj€ okull viewpoint, only 1 had read the warning signs on the access road, but they had misinterpreted the warning to be about small-scale rockfall (T.1; Figure 3). Another tourist mentioned safety was a top priority during their trip. However, when asked about the warning signs, she had understood them to include ''some tourist information, some history information about this place, and probably some safety warning,'' but explained that ''we didn't read them'' (T.8). Another tourist intentionally ignored the signs: ''Sometimes it better not to know what's happening around you. . . sometimes you want to cut yourself off'' (T.5). A third tourist mentioned that natural hazards were among the experiences that attracted him to the country, and that if there was a risk that he ''wouldn't like to hear about it'' (T.6). In identifying risks in the area, tourists focused on the most visible and immediate risks such as slipping on ice, falling into cold water, small-scale rockfall, or weather conditions (T.1; T.7; T.8). During these discussions, a wide range of risk aversity among tourists was evident.
Communicating risk information with tourists in Iceland is complicated by the large number of tourists on short visits to Iceland who have little previous experience with the terrain and hazards of the country. Risk managers drew attention to the low risk knowledge of most tourists visiting Svínafellsj€ okull and Iceland (RM.1; RM.3). Tourists lacked an understanding of emergency protocols beyond normal instincts to avoid debris and move to higher ground (T.1-T.8), which is not an appropriate response given the massive scale of the hazard and the degree of exposure at the viewpoint. This was unsurprising given the almost total lack of awareness that tourists showed about the potential risk of a large landslide onto the glacier.
A common sentiment expressed by foreign tourism employees and local inhabitants was feeling responsible for the safety of others, including clients, tourists, staff members, and family members. This was particularly pronounced among glacier guides, hotel receptionists, and tourism managers (GG.4; GG.6; GG.8; LI.1; LI.3; LI.6; FI.6). A local inhabitant reflected that ''if you have a company, then you're always responsible for the people who are working for you, and you're also a bit responsible for your guests'' (LI.6). A glacier guide also explained that ''I feel like I have responsibility for other people, even if they are not on my tour'' (GG.6).
All local inhabitants interviewed had a comprehensive understanding of the fracture and emergency protocols (LI.1-LI.16). Local inhabitants frequently referred to the findings presented in risk briefings and were aware of recent developments (LI.1-LI.16). However, there was some initial disagreement between scientific and local knowledge about what areas would likely be affected (Matti and Ogmundardóttir 2021). One local inhabitant reasoned that floodwaters from the glacier have ''always come down these two rivers. . . why would this be any different?'' (informal discussion, 24 October 2018). Based on their risk knowledge, some local inhabitants had developed personal response plans (LI.1; LI.12). All local inhabitants were aware that a warning of imminent collapse would come through the DCPEM phone alert system (LI.1-LI.16). Some expressed concerns about poor mobile reception (LI.1; LI.3; LI.13), whether they would wake up to a message sent at night (LI.3), and about how tourists would react to such a message (LI.1).
By comparison, foreign tourism employees varied greatly in their understanding of the risk and how to respond. All those interviewed who had lived in the area for less than a year had heard of the fracture but were aware that they lacked a comprehensive understanding of the risk: ''I don't know very much about this'' (FI.2); ''I know about the fracture but I don't know exactly what caused it or what is actually happening'' (FI.4). Glacier guides and foreign tourism employees who had lived in the area for longerespecially those present when operations shifted from Svínafellsj€ okull-tended to have a stronger understanding. However, the knowledge of glacier guides and foreign tourism employees was typically less up-to-date, less evidence based, and less technical than that of local inhabitants (FI.3; FI.5; LI.3). Foreign tourism employees were also unsure of emergency protocols, including communications (Box 1) and response strategies (Box 2). Furthermore, a foreign tourism employee living in an area at risk mentioned that in the case of an emergency ''most of us here don't have a car, so it's hard for us to go some place'' (FI.4).
Glacier guides were aware and actively mitigated risks in the course of their work (GG.1-GG.8). Focus was placed on the type of risks experienced regularly, including crevasses, minor rockfalls, crossing rivers, poor crampon technique, poor weather, and unstable glacial features. Guides regularly engaged in discussion and training drills-especially crevasse rescues-with more experienced guides and management on how to manage these risks. A similar culture has been documented among guides in other adventure tourism activities, including ski touring (Grímsdóttir 2004: 23) and white-water rafting (Morgan and Fluker 2006: 157). However, there was virtually no discussion among guides of the Svínafellsheiði fracture and risk of a large landslide. One glacier guide also noted that his company did not have a record of the names of all tourists going on glacier walks, so in case of a landslide ''we could have easily had 200 people buried, we actually wouldn't even know their names'' (GG.3).

Communication channels
Official risk meetings were the main channel by which local inhabitants received information about the Svínafellsheiði fracture and emergency protocols. Several local inhabitants reevaluated their personal response strategies based on advice and updates given at these forums (LI.1; LI.2); this Hopefully through a government source, but maybe the search and rescue teams or the park rangers or the police. I feel they should be a little bit more organised. I hope that if they were alerting people that something was happening, they would have a plan in place. I wouldn't expect a personal communication but a general announcement, maybe over the radio, or the [national park] rangers talking to shift managers and then giving us instructions on how to evacuate people. (GG.6) I would probably just hear about it from somebody else. I'm not entirely sure about how that would come to us. Either through the news, national park, or from another guide. I would hope that the company that we work for would be like quick on the uptake and send an email and make sure that everyone was out of the area. (GG.7) indicates that information provided was trusted and taken seriously. The main concern about meetings was that they were not conducted according to the pre-arranged time line. For example, in October 2018, people were told that they would receive an update in spring 2019, yet it did not take place until the end of autumn (LI.1). The meetings were typically dialogic and participatory, with information from local inhabitants used to shape emergency protocols and risk management plans (RM.2; RM.4). While the townhall-style briefings were public, email invitations were sent almost exclusively to Icelanders (email communication, local police, 28 November 2019). All local inhabitants interviewed were aware of the meetings, and most had attended. By comparison, most foreign tourism employees were not aware that the briefings occurred; the few who were aware of the meetings understood them to be conducted ''in Icelandic for locals'' (FI.6). Some local inhabitants expressed concern about the exclusion of foreign inhabitants from formal risk management processes and communication channels. One woman declared that ''it's not good enough that some people are left out . . . I think there's a big information gap between locals and people that are living here as foreigners'' (LI.3).
Foreign tourism employees working in hospitality typically received information about the fracture from their employer or other staff members (FI.1; FI.3; FI.5; FI.6). One interviewee accessed information through mainstream media outlets (FI.3). Some employers were proactive in providing information to employees, including one who maintained that the ''more information you can give to the staff is better . . . we try to update them as much as possible'' (LI.1). This included in-house meetings and updates on employee-specific social media pages (LI.1). However, some foreign tourism employees had not received any information about the risk from their employer. For example, one woman explained: ''I wasn't informed about anything like this when I came here. After some time, I hear something but not officially and not how we should proceed if it happens'' (FI.2). Furthermore, when asked if foreign tourism employees would be in direct contact with risk managers, one local business owner responded that they are ''probably going to come to us as they don't know where to get information'' (LI.6). There was no evidence of information about the fracture being provided to glacier guides by their employers. Instead, developing an understanding of the risk was viewed by guides as something they have to personally ''put the time into'' (GG.4). Some experienced glacier guides had attended a scientific briefing at the University of Iceland in early 2018 in their free time, but they had not received any subsequent updates (GG.4; GG.6; GG.7).
Risk managers acknowledged that they ''relied on locals'' to ''tell your employees about the hazard'' (RM.2). However, at least one local hotel was managed by a foreigner who did not attend the briefings (LI.3), and even the most proactive employers did not update their staff after each briefing (FI.6). The information passed to employees was left to the discretion of employers (R.2; FI.6). There was no evidence of risk managers providing tourism employers with material or support on how to inform their staff or of tourism businesses requesting this information.
Foreign tourism workers typically had very limited interaction with local inhabitants. One interviewee mentioned that ''I don't get a chance to talk much with local people, and we have this language barrier'' (FI.4). An Icelandic local inhabitant explained that it was a ''very split community between the people who have always been here, who speak Icelandic, and then all the young people, who are glacier guides or working at the hotels and don't speak Icelandic'' (LI.3). Opportunities for interaction were also affected by job position, with those working in housekeeping understood to have less interaction and access to information than people at reception (FI.3). Risk managers acknowledged that as a foreign worker, ''you could come to work in a hotel cleaning the rooms and you will have absolutely no idea about the risk you could be facing'' (RM.2). Experienced glacier guides generally had more social capital and more regular contact with local inhabitants, especially locals who had also worked as guides (GG.2; FI.1; FI.3). In recent years, large glacier guiding companies had reduced wages and working conditions, which reduced the hiring of Icelandic staff and led to the emergence of small local operations, both of which eroded opportunities for regular interaction between Icelandic and foreign glacier guides (LI.3).
Tourism operators were concerned about how risk management decisions were made, including the choice to shift operations from Svínafellsj€ okull. Interviewees felt that risk management and communication would have been conducted differently with a better-established sector of the Icelandic economy. One manager argued that, ''if this was somewhere else and there was a fish factory in the way, it would have been dealt with in a very different way . . . but BOX 2: Understanding of response protocols by foreign tourism employees I don't think anything has been discussed about what we would do in that situation. As far as I know, nothing has been done about it. No protocols, nothing.
(FI.1) I would just try to escape by some means I guess if that is what needed to happen. I admit I haven't given it much thought. But yeah I would literally just try to get away. I would try to tell people on the way out what would be the danger zone, I would tell people to come with me.
(GG.4) I don't know. I want to say that I would yell at everyone to get in whatever bus or car and get out of there as soon as possible. . .I guess if you're driving along you could take a right [south], go through some fences, get as much distance between you and the landslide as possible, head towards the sea. Who knows if that works? That would be the first thing that comes to mind.

(GG.5)
We have the volcano plans, but I don't think we have landslide plans. If we have a landslide that comes down and it blocks the ring road, I don't think we have any plan of what do to. (GG.7) I don't imagine you would have much time. It would be like ''it's coming, get out of the way.'' I would help with evacuation efforts because as far as I can tell. . .but I'm not sure actually. I don't know whether we would be affected. I don't know whether our office would be affected. I don't know.
(FI.1) since it's just a bunch of guides, and they don't understand tourism, they are just like 'yeah just go somewhere else''' (GG.3). Foreign tourism employees and glacier guides expressed interest in attending public briefings if they were conducted in English. One interviewee explained, ''I live here, so it's also about me, of course I would like to go'' (FI.6). Another reasoned: It's important that everyone in the area at risk is informed and knows what they can do and how they can help because, in those sorts of situations, it needs to be all hands on deck. There should be common knowledge among anyone living in the area temporarily or permanently.
An interviewee noted that the tone of the invitation message was important; she recommended that it be pitched as a responsibility to learn how to respond rather than a general conversation about the fracture (FI.1). Another interviewee further stated that if it were not pitched like this, it would not be considered a priority for tourism managers or employees (GG.6). A local inhabitant recommended that for future hazards, a basic written evacuation plan should be provided at the first briefing (LI.1).

Discussion
This study found most tourists visiting the Svínafellsj€ okull viewing area did not sufficiently understand the risk posed by the fracture or how to respond in the case of a large landslide. Even the one interviewee who had read the sign had misunderstood the warning and vastly underestimated the scale of the hazard. Similar problems of tourists ''walking straight past warning signs, taking no notice of the information displayed'' were reported at the high-risk Reynisfjara beach in Iceland (Iceland Monitor 2016). Communicating information about the risk of Svínafellsheiði to tourists was complicated because there were many significant risks in the area; it was not a tourist attraction, unlike other hazards such as volcanic eruptions and crevasses; it was not visible to visitors; and it was the first time this type of hazard has been managed in Iceland. Tourists generally understand that volcano sites are potentially dangerous, and they choose how much safety information to seek and precautionary measures to take (Bird and Gísladóttir 2020); by comparison, most tourists were not even aware they were potentially exposed to a large landslide.
In the case of an emergency, most tourists stated they would depend on people living or working in the area to keep them safe. In their study of Bolivian mountaineering, Mackenzie and Kerr (2012) found that the expectation that risks will be managed and safety ensured is more pronounced on guided tours. Our results indicate that, while tourism employees and guides feel responsible for the safety of clients, many are poorly informed about the risk and emergency protocols. This trend is more pronounced among newcomers and those with little social capital in the community. Glacier guides tend to focus on managing risks that they encounter more frequently. This may reflect: the massive scale of the Svínafellsheiði hazard; the feeling that they are unable to personally control or mitigate the hazard; the focus of management and fellow guides on other risks; and a lack of involvement in official risk communication structures.
Despite significant demographic changes driven by tourism, Icelandic inhabitants remain the primary unit upon which risk management processes are based. In the case of the Svínafellsheiði fracture, this was evident in initial exposure calculations and communication strategies. Official risk communication was oriented toward local inhabitants and, to a lesser degree, tourists. If foreign tourism employees received any information, it was generally in the form of ad hoc and informal communications from their employer. Language and mode of invitation represented significant barriers for foreign tourism employees to attend official briefings. While some studies have examined language as a barrier to effective risk communication, most have focused on how this excludes tourists rather than tourism employees (Erfurt-Cooper 2010: 6, 20).
Risk communication that depends on informal and personal relationships can exclude demographic groups and exacerbate existing power inequalities and vulnerabilities. Foreign tourism employees typically have lower social capital, especially those who do not interact with the public through their work. This is problematic, as research has found that lower social capital and reduced access to risk information lead to more risk-taking behavior in the event of a disaster (Cadag et al 2017;Anderson-Berry et al 2018). Icelandic employers are typically older and more financially stable, while foreign tourism employees are younger and heavily dependent on their employer for salary, accommodation, and sometimes visas. Having foreign tourism employees depend on Icelandic employers for risk information reinforces the social, financial, and workplace power of Icelandic employers. The arrangement can also increase the burden on employers grappling with the risk and how it affects their livelihoods.
Risk communication channels available to foreign tourism employees were generally unidirectional, such as mainstream media, and informal, such as information conveyed by employers. Foreign tourism employees were hindered from developing a direct relationship and building trust with risk managers and scientists; in addition, there was no clear avenue for them to contribute their knowledge to risk management strategies. This increases the chances that emergency protocols do not take into account their particular circumstances, for example, vehicle ownership. Erfurt-Cooper (2010: 15) similarly found that volcanic risk management that focuses on local inhabitants can result in emergency rescue and evacuation systems that do not take into account tourists and other demographics.
Risk communication in the case of Svínafellsheiði was further inhibited by the lack of written emergency procedures. Perry and Lindell (2003: 340) argued that it is not uncommon for risk managers in small communities to depend on ''informal, personal relationships for risk identification, assessment and reduction'' rather than written protocols. However, the provision of written guidelines in English and other common languages would encourage greater access to information. Research in Japan found that a lack of experience with emergency protocols ahead of time contributes to panic and confusion in the case of a crisis (Sakurai and Adu-Gyamfi 2020). The consequences are compounded as foreign tourism employees are often also responsible for the safety of large numbers of tourists. This research highlights the need for risk communication and training initiatives that specifically target tourism employees and glacier guides. These should be conducted in English, incorporate simulations/drills, and be based on the understanding that employees will likely have some responsibility for the safety of clients. In Japan, a simulationbased exercise was conducted to train tourism and hospitality staff on keeping tourists calm and communicating risk and evacuation procedures in the event of a volcanic eruption (Suzuki 2020). Given the high staff turnover inÖraefi, we recommend conducting interactive training simulations on a regular basis, for example, every 6 months.
Several crises in guided tourism have raised questions about culpability and criminal negligence. On 27 July 1999, 21 tourists and guides died in a flash flood while canyoning near Interlaken, Switzerland. In the subsequent court case, 6 managers and senior guides were convicted of manslaughter due to culpable negligence for putting profits before safety, ignoring warning signs, and not sufficiently training junior guides to assess flood risk (Morgan and Fluker 2006). On 7 January 2020, 39 tourists were stranded in a severe storm on Langj€ okull, Iceland's second largest ice cap, on a guided snowmobile trip. More than 200 search and rescue volunteers rescued the tourists, some of whom spent more than 7 hours in the storm. Similar questions were raisedincluding by the Minister for Tourism-about the decision to run the trip despite a severe weather warning (Ćirić 2020). Both cases underscore the importance of training tourism employees and guides on risk management and safety, communicating risk effectively, and implementing safety protocols.
Our results support calls for an expansion of targeted risk management actions for the tourism sector (Becken and Hughey 2013;Ziegler et al 2021: 13). In their study of the Himalayas, Ziegler et al (2021: 16) suggested that licenses or certificates could be used to demonstrate that guides or tourism workplaces adhere to safety standards. A similar initiative could be explored in Iceland. A further exploration of this goes beyond the scope of this study.
Such initiatives will better position tourism employees and guides to inform tourists of risks faced. Tourists who survived the Whakaari/White Island disaster reported that they received no information about the risk and, as a result, could not make informed decisions about their safety (March et al 2020). We agree with Bird and Gísladóttir's (2020) study, which found that some tourists will continue to ignore warning signs, pursue thrill-seeking behavior, and lack local knowledge, so we must ''find other ways to reach them.'' It is crucial that tourists are equipped with sufficient knowledge to make decisions about their own safety. Reducing risks and improving the safety of tourists and tourism employees are not only ethical pursuits, but they also are important for long-term business sustainability.

Conclusion
In this conclusion, we explore how the results of this research can help to improve risk communication for the Svínafellsheiði fracture and for hazards in other nature tourism hubs around the world. Some features of the specific context should be taken into account. This research relates to a large-scale, potentially fatal, sudden-onset hazard, which has a high degree of uncertainty. The lessons are more likely to be applicable to other such hazards, for example, volcanic eruptions, avalanches, flash floods, and landslides. Tourists inÖraefi are exposed to a multitude of potentially fatal hazards that require customized risk management communication and informed decisions and actions. Even risk-averse tourists are likely to focus on more visible, frequent, and well-known hazards to which they are exposed for longer periods. Tailoring risk communication and preparation to tourism employees is more crucial for largescale but less frequent and less visible hazards. The lessons will be particularly relevant for nature and adventure tourism hubs where large portions of tourism employees face language and cultural barriers when accessing government risk communication (eg the Alps, Japan, and Scandinavia). This case also represents the first time the risk of a large landslide onto a glacier has been managed in Iceland.
The case of the Svínafellsheiði fracture suggests that risk assessment and management processes focused heavily on the scientific basis of risk, while assessments of exposure and vulnerability were built on sweeping and dated assumptions. It should be standard practice in the initial phases of a risk assessment to develop a profile of who is exposed, existing vulnerabilities, and power dynamics within the community, perceived responsibilities in the event of an emergency, and access to evacuation infrastructure. Risk management strategies must expand from the narrow focus on nationals, especially in the context of mass tourism. Semistructured interviews and ethnographic fieldwork represent useful methodological tools for exploring these dynamics.
Authorities in Iceland rely on risk communication rather than access restrictions to ensure tourist safety, as much of the sector is based on experiencing glacial, volcanic, and geothermal landscapes, all of which have inherent risks (Bird and Gísladóttir 2020). Our results call for tailoring risk communication and training to the needs of tourism employees, guides, and tourists, as well as local inhabitants. At a minimum, this should include written emergency protocols, participatory risk briefings, and regular simulation-based training in the main language(s) of operation. This should be part of a broader push for increased dialogue and coordination between tourism operators and risk managers. This requires commitment backed by earmarked resources from the government and tourism operators.
With tourism destinations in glacial and mountainous environments increasingly exposed to hazards due to climate change (Tsai and Chen 2011;Mair et al 2016), ensuring the safety of all people is essential. Important questions for future research include: How can guides and guiding operations adapt efficiently and respond to newly emerging hazards? How can technology be harnessed to tailor risk communication to different audiences? What level of risk understanding is required for informed consent in the context of tourism?