Assessing the impact of explosive eruptions of Fogo volcano (São Miguel, Azores) on the tourism economy

. The Azores are an active volcanic region that offers exceptional conditions for nature-based tourism, one of the main axes of economic growth in the archipelago. A future volcanic eruption may have long-term consequences to this economic sector. Therefore, it is fundamental to assess its vulnerability to volcanic hazards in order to try to mitigate the associated risk. This study proposes a new approach to assess the economic impact of explosive eruptions on the tourism sector. We considered two eruptive scenarios for Fogo volcano (São Miguel Island), the most probable (VEI 4 sub-Plinian 15 eruption) and the worst-case (VEI 5 Plinian eruption), both producing tephra fallout and pyroclastic density currents. The results of numerical simulations were overlaid with tourism-related buildings and infrastructure of Vila Franca do Campo municipality to identify the elements at risk. The Loss Present Value method was used to estimate the benefits generated by the accommodation units over 30 years for different economic scenarios. The assessment of the economic impact using 2018 indicators reveals that in a near total destruction scenario the economic loss is approximately 145 million euros (considering 2 20 % discount rate). This approach can also be applied to other volcanic regions, geologic hazards and economic sectors.

Despite the numerous hazards related to explosive volcanism, in several places volcanoes are regarded as attractions and not as a potential sources of problems. For some communities, nature-based tourism plays a significant role in the development of the local economy. However, the potential of some volcanoes to produce hazardous explosive eruptions is often overlooked or underestimated. 35 economy. As the Azorean tourism suffered a significant boost since 2015 with the alteration of the accessibility and mobility model, the importance of this sector in the regional economy has grown over the years. Therefore, it is fundamental to assess its vulnerability to volcanic hazards in order to implement appropriate mitigation strategies. São Miguel Island is the largest (744 km 2 ) and most populated (> 137 000 inhabitants) of the Azores archipelago, where most of the tourism industry is concentrated. The volcanic history of São Miguel shows that it is the most active island in the Azores, 70 with the highest eruptive frequency of explosive events. At least 34 explosive eruptions (sub-Plinian and Plinian) are known to have occurred in the last 5000 years. Fogo volcano, located in the central part of São Miguel, was the only volcano that produced a Plinian eruption in this time frame, although it has the lowest eruptive frequency of the three active central volcanoes of the island (Gaspar et al., 2015a). Nonetheless, its potential to produce hazardous explosive eruptions must not be underestimated and the economic impact of future eruptions should be properly assessed. 75 In this framework, we aim to evaluate the impact of future explosive eruptions of Fogo volcano on the tourism of São Miguel, taking Vila Franca do Campo municipality as study area. To assess which areas and exposed elements are susceptible to be affected by tephra fallout and pyroclastic density currents (PDCs), two eruptive scenarios were considered: the most probable scenario, corresponding to a sub-Plinian eruption with Volcanic Explosivity Index (VEI, Newhall and Self, 1982) of 4; and the worst-case scenario, corresponding to a Plinian eruption with VEI 5. The assessment was carried out on buildings related 80 to tourism and allowed to estimate which exposed elements could be affected, with physical damage and loss of functionality.
The methodology proposed consists in evaluating the benefits generated by the tourism industry, restricted to the quantification of revenues generated by accommodation units, to determine the current loss of that revenue over 30 years. Such analysis is especially important as tourism is a growing and promising activity in the Azores in general (Vieira et al., 2019), in São Miguel Island (Vieira and Antunes, 2017) and certainly in Vila Franca do Campo, which due to the existence of bathing areas, a 85 moderate climate, and maritime and recreational infrastructures attracts many visitors. This study represents the first attempt to quantify the economic loss related to future explosive eruptions of Fogo volcano on the island of São Miguel. However, the present approach may be adopted to other active volcanic regions and economic activities.

Geographical and geological setting 90
The Azores archipelago is located in the North Atlantic Ocean and comprises nine volcanic islands. From a geodynamic point of view, this region corresponds to the triple junction where the North American, Eurasian and Nubian lithospheric plates meet ( Fig. 1a,b). Due to this particular geodynamic framework, the archipelago is subject to frequent seismic and volcanic activity (e.g. Gaspar et al., 2015b). São Miguel Island is formed by three active central volcanoes, Sete Cidades, Fogo and Furnas, linked by Picos and Congro 95 fissure volcanic systems. The eastern part of the island comprises the older and inactive Povoação volcano and Nordeste volcanic system (Fig. 1c). All three active central volcanoes are truncated by summit calderas related to paroxysmal explosive eruptions. In the last 5000 years at least 33 sub-Plinian eruptions and one Plinian eruption are known to have occurred in São Miguel (Pacheco et al., 2013;Gaspar et al., 2015a).
Located in the central part of São Miguel, Fogo volcano (also known as Água de Pau volcano) is the largest of the three active 100 central volcanoes of the island, reaching a maximum altitude of 947 m above sea level. The 3.2 km-wide summit caldera was formed by several collapse events and is presently occupied by a lake. Several volcanic structures are present inside the caldera (lava domes, pumice and tuff cones), as well as on the northern and southern flanks of the volcano (lava domes, scoria and pumice cones) (Wallenstein, 1999;Wallenstein et al., 2015).
The stratigraphy of Fogo volcano is divided into two major lithostratigraphic units (Wallenstein, 1999;Wallenstein et al., 105 2015): the Lower Group, which comprises all volcanic products older than 40 ka; and the Upper Group, which includes all products emitted in the last 40 ka, including from the historical eruptions. The recent eruptive period was marked by at least two paroxysmal explosive eruptions of Plinian dimensions: the Ribeira Chã eruption (8000-12 000 years ago) and the Fogo A eruption (~4600 years ago). In particular, the last 5000 years were characterized by five sub-Plinian eruptions: Pisão, Fogo B, C, D, and 1563 (historic). 110 Fogo A was one of the largest eruptions recorded on São Miguel. Its deposit encompasses a complex and widespread succession of trachytic pyroclastic products emitted from the summit caldera (e.g. Walker and Croasdale, 1971;Booth et al., 1978;Bursik et al., 1992;Wallenstein, 1999;Pensa et al., 2015a,b). The eruption started with a short-lived hydromagmatic phase, followed by a Plinian eruptive column that produced a major pumice fall deposit. The radial and almost symmetrical distribution of the fall deposit indicates that weak wind was blowing from the west during the eruption. The eruptive column 115 experienced partial collapses that generated small volume PDCs and, in the final stage, the total collapse of the column led to the emplacement of a voluminous ignimbrite, reaching > 20 m thick in Ribeira Grande graben (Wallenstein, 1999;Pensa et al., 2015a,b;Wallenstein et al., 2015).
The last sub-Plinian eruption of Fogo volcano occurred in CE 1563. The deposit corresponds to a stratified succession of trachytic pumice lapilli and ash fall layers (e.g. Walker and Croasdale, 1971;Booth et al., 1978;Wallenstein, 1999;Aguiar, 120 2018). The eruption started on June 28 th in the centre of the caldera, on a previously existing cone (know as Pico da Lagoinha or Pico das Berlengas) (Frutuoso [1522(Frutuoso [ -1591, 1981). The first phase was hydromagmatic and was followed by a sub-Plinian eruptive column with repeated hydromagmatic pulses. Tephra was mostly dispersed to the eastern part of the island due to the strong west-southwest-blowing wind (Walker and Croasdale, 1971;Wallenstein, 1999;Aguiar, 2018). The eruptive activity lasted five days and ceased on July 3 rd . Four days after the onset of the sub-Plinian eruption, a basaltic flank eruption occurred 125 on Pico Queimado dome (then called Pico do Sapateiro), on the north flank of Fogo volcano. A subsequent phreatic explosion was also reported inside the caldera on February 1564 (Frutuoso [1522(Frutuoso [ -1591(Frutuoso [ †], 1981Wallenstein et al., 2015;Aguiar, 2018).
The present study area, Vila Franca do Campo municipality, is located on the southern flank of Fogo volcano and it borders with Lagoa municipality to the west, Ribeira Grande to the north and Povoação to the east. To the south it is bounded by the ocean in a costal extension of 15.5 km. Vila Franca do Campo has an area of 77.9 km 2 and is divided in six parishes: Água 130 d'Alto, São Pedro, São Miguel, Ribeira Seca, Ribeira das Tainhas and Ponta Garça (Fig. 2).

Definition of the eruptive scenarios
The first step in the assessment of the impact of future explosive eruptions of Fogo volcano on São Miguel Island was to define 135 the eruptive scenarios. Based on the recent geological record of Fogo volcano (last 5000 years) and the frequency and magnitude of past trachytic explosive eruptions, two scenarios were defined: the most probable scenario (i.e. the most likely eruption) and the worst-case scenario (i.e. the largest magnitude eruption), in agreement with previous studies (e.g. Gaspar et al., 2015a). The most probable scenario is a sub-Plinian eruption with VEI 4, similar to the Fogo 1563 eruption. Such an eruption would produce widespread tephra fallout and could also generate PDCs in proximal areas, although this was not the 140 case of the 1563 eruption. The worst-case scenario is a Plinian eruption with VEI 5, similar to the Fogo A eruption. Such an eruption would produce thick widespread tephra fallout and generate voluminous PDCs along the flanks of the volcano.
Taking into account that in the Azores region there are differences in the wind patterns of summer (May to September) and winter periods (October to April) (see wind statistical analysis in Pimentel et al., 2006;Cole et al., 2008;Gaspar et al., 2015a) different tephra fallout scenarios should also be considered for the two periods. 145

Numerical simulations
To identify which areas of São Miguel are susceptible to be affected by trachytic explosive eruptions of Fogo volcano, the dispersion of tephra fallout and PDCs was simulated using VORIS (Volcanic Risk Information System) version 2.0.1 (Felpeto et al., 2007) implemented in a geographic information system (GIS) (ArcGIS 9.1 ESRI®). VORIS 2.0.1 is a tool used in the assessment of volcanic hazards that provides users with the necessary instruments for the production of scenarios and hazard 150 maps.

Tephra fallout
Numerical simulations of tephra fallout were computed using an advection-diffusion model that assumes that above the vent the mass is distributed along a vertical line following the Suzuki approach (Suzuki, 1983). Far from the vent, the transport of particles is controlled by the advective effect of the wind, the diffusion due to atmosphere turbulence and the terminal settling 155 velocity of the particles (see details in Folch and Felpeto, 2005;Felpeto et al., 2007).
The eruptive source parameters used in the simulations were obtained from the literature related to Fogo A and Fogo 1563 eruptions, and when unavailable, from published data of similar explosive eruptions. For the most probable scenario (a VEI 4 sub-Plinian eruption), we considered a total bulk volume of 1 km 3 (Booth et al., 1978) and a column height of 18 500 m (Carey and Sparks, 1986). For the worst-case scenario (a VEI 5 Plinian eruption), we used a total bulk volume of 3.2 km 3 (Booth et 160 al., 1978) and a column height of 27 000 m (Bursik et al., 1992). The simulations were conducted assuming a vent located in the centre of Fogo caldera. Eruptive input parameters are shown in Table 1.
Wind parameters were compiled by Pimentel et al. (2006) Terceira, between 1947 and2003. Statistical analysis of 56 years of radiosonde data revealed significant differences between summer (May to September) and winter periods (October to April). In the troposphere (and lower levels of the stratosphere, i.e. up to 17 000 m altitude), the most frequent directions in the summer period are west to northwest and north to northwest blowing winds, whereas in the winter period the dominant trend is west to southwest blowing winds. At higher altitudes (above 17 000 m), a strong eastern direction prevails during the summer period, while in 170 the winter period a western direction dominates (Pimentel et al., 2006;Gaspar et al., 2015a). The most probable combinations of wind direction and intensity were chosen for different vertical heights, according to the simulated column height. Wind input parameters are summarized in Table 1.

Pyroclastic Density Currents
Simulations of PDCs were performed with the energy cone model (Malin and Sheridan, 1982), which provides a fast and conservative approach to assess the maximum potential extent of these volcanic products (e.g. Alberico et al., 2002Alberico et al., , 2008Felpeto et al., 2007;Toyos et al., 2007). The maximum potential extent of a PDC is directly related to the VEI of the eruption 180 and the topography around the vent. A higher VEI implies that the PDC can reach larger distances (Alberico et al., 2008).
Input parameters used in the simulations were collapse equivalent heights of 300 and 500 m (for VEI 4 sub-Plinian and VEI 5 Plinian scenarios, respectively) and a constant collapse equivalent angle of 6°, in agreement with previous studies (e.g. Alberico et al., 2002Alberico et al., , 2008Alberico et al., , 2011Cole et al., 2008). Both simulations were preformed assuming a source area equal to the floor of the caldera, where each 50 m cell had an equal probability of generating PDCs. 185

Exposed elements
The vulnerability of buildings and infrastructure to different volcanic products depends on the type of construction materials, the quality of the workmanship, the age and maintenance level, their shape and orientation (Pomonis et al., 1999), but also on the loss of benefits from the activity for which they are intended to. Thus, in the particular case of explosive eruptions, we considered the vulnerability of buildings to tephra fallout and PDCs. In this study, we do not intend to assess the damage to 190 the buildings but instead the loss of functionality.
To identify the exposed elements, a detailed inventory of all buildings and infrastructure related to the tourism sector in Vila Franca do Campo municipality was carried out during the summer of 2017. This followed a similar approach to the studies carried out by Pomonis et al. (1999) in Furnas parish and Gomes et al. (2006) in the parishes located on the flanks of Sete Cidades volcano (São Miguel Island), and by Cabral (2015) in Santa Catarina municipality (Fogo Island, Cape Verde). 195 The inventory included accommodation (hotels, guest houses, rural tourism, local accommodations), restaurants (ice cream shops, pastry shops, cafes, pubs), tourism animation/activities (travel agencies, rent-a-cars, nautical and terrestrial activities, souvenirs shops) and culture (churches and places of cult, museums, libraries, theatres, marketplaces, cultural centres).
Although this study does not intend to assess building damage but rather its loss of functionality, the inventory was based on a classification method developed by the Centre for Volcanology and Geological Hazards Assessment (CVARG) of the 200 University of the Azores, now Research Institute for Volcanology and Risk Assessment (IVAR), to study building vulnerability to different geological hazards in the Azores. This method classifies buildings according to use, number of floors, type of materials used in the construction, roof inclination, type of windows, etc. Details on the classification method can be found in Gaspar et al. (2004) and Gomes et al. (2006). The exposed elements (i.e. buildings and infrastructure associated with tourism in Vila Franca do Campo) were mapped in detail in a GIS. 205 The inventoried elements were then combined with the maps resulting from the different numerical simulations of tephra fallout and PDCs. This allowed us to identify which buildings and infrastructure would be affected by a certain thickness of tephra, as well as those located within the maximum potential extent of PDCs.

Economic value of tourism
The methodology used to assess the impact of future explosive eruptions of Fogo volcano on the tourism economy consists in 210 evaluating the benefits generated by the accommodation units of Vila Franca do Campo municipality. With that aim, we determined the current loss of revenue over a period of 30 years for different eruptive scenarios and discount rates, and considered the evolution of the occupancy rate of the accommodation capacity. This study follows the approach proposed by Vianna et al. (2012) that quantifies the value of a tourism industry based on shark diving in the Republic of Palau (Pacific Ocean), one of the main activities that contributes to the country's economy. 215 In the present case and following the building inventory, we calculated the annual income of 46 accommodation units by estimating the average price per night of each unit. To calculate this value, we considered the price of one night in March 2018 (low season) and the cost of one night in August 2018 (high season). Some of the costs were obtained directly, provided by a representative of the accommodation unit, while others were obtained indirectly through online booking and shopping platforms, as well as through own websites. In specific cases, when was not possible to obtain this information, an average 220 cost was calculated considering the type of accommodation.
The annual revenue was obtained by multiplying the average cost per night by 365 days for the cases where the rent corresponds to the entire accommodation building. For the cases where the rent corresponds to only one room (e.g. hotels, apartments or chalets), the average cost of one room was multiplied by 365 days and by the number of rooms of the accommodation unit.
The total annual revenue of the tourism sector in Vila Franca do Campo at 2018 values was obtained by adding up the annual 225 revenues of each of the 46 accommodation units. However, this total revenue presupposes an occupancy rate of 100% of the existing accommodation capacity. To consider a more realistic situation, the total annual revenue was multiplied by a factor of 0.5 and 0.65 to assume an occupancy rate of 50 % and 65 %, respectively. This originates two alternative scenarios for the existing accommodation capacity (explained later on).
To calculate the economic loss in Vila Franca do Campo municipality due to future explosive eruptions, we considered three 230 economic scenarios resulting from the numerical simulations: 1) Destruction of accommodation buildings affected by ≥ 20 cm of tephra fallout from a VEI 4 sub-Plinian eruption during the summer period; 2) Destruction of accommodation buildings located within the maximum potential extent of PDCs generated by a VEI 4 sub-Plinian eruption; and 3) Destruction of accommodation buildings located within the maximum potential extent of PDCs generated by a VEI 5 Plinian eruption.
Economic scenarios 1 and 2 (tephra fallout and PDCs, respectively) from a VEI 4 sub-Plinian eruption are more conservative 235 and realistic in case of a future explosive eruption of Fogo volcano. Regarding tephra fallout, the summer period was chosen as it is the high season of touristic activity, but also because it corresponds to the wind conditions that would affect a larger portion of the study area. Economic scenario 3 (PDCs from a VEI 5 Plinian eruption) is not the most likely but represents the worst case possible, assuming near total destruction of the study area. The annual revenue of the accommodation units was calculated for these three scenarios. 240 The next step was to calculate the Loss Present Value (LPV) for each of the economic scenarios. The LPV corresponds to the value lost (in Euros) after a certain period of time, discounted to the initial period (2018), which corresponds to year zero t0, following the expression: Where is the year (0 to i), t is the revenue of the accommodation units in year , taking into account the occupancy rate, 245 and is the discount rate.
The occupancy rate ( ) is the percentage of occupied units or rooms in a certain area during a specific period. The discount rate ( ) is an intertemporal preference rate that allows to convert future values into present values, and accounts for the notion that a given monetary amount does not have the same value in the present and in the future. A discount rate equal to zero is used when it is intended to give the same weight to future and present values in a long-term analysis. This rate is considered a 250 critical element in cost-benefit analysis when costs and benefits accrue over a number of years. The social discount rate is used when examining costs of benefits accruing to the society and its value has been calculated by researchers, or sometimes externally proposed by government agencies, for purposes of project or even environmental losses evaluation (Courard-Hauri et al., 2020). Evans and Sezer (2004) propose discount rates for six major countries ranging between 3.5 % (France) and 5 % (Japan). Evans and Sezer (2005) argue that social discount rates in European Union member countries mostly lie in the range 255 of 3 to 5.5 %, while Florio and Sirtori (2013)  In this study, we considered a period of 30 years for the analysis, following the example of the eruption of Soufrière Hills 260 volcano (Montserrat, Lesser Antilles) that started in 1995. This eruption also occurred on an island and produced tephra fallout, PDCs and lahars, with devastating consequences that last until today, even after more than two decades. Currently Montserrat's tourism industry is still in a recovery phase (Pacheco and Lewis-Cameron, 2010). The LPV was calculated for each of the chosen economic scenarios, considering two occupancy rates of the existing accommodation capacity: one more conservative, with an occupancy rate of 50 % and discount rates of 0, 2 and 4 %; and another less conservative, considering that the Azores 265 tourism shows a growing tendency, with an occupancy rate of 65 % and the same discount rates (0, 2 and 4 %). Discount rates between 2 and 4 % lie in the range of values estimated for Portugal and mentioned above. Moreover, the choice of two discount rates is also justified by the factors related to uncertainty. Moreover, the use of a higher (lower) discount rate can be seen to some extent as a way of giving less (more) weight to future and therefore uncertain (or risky) monetary flows associated with the touristic activity. Finally, the impact of discounting can be assessed through comparison with a zero-discount baseline. 270

Tephra fallout from VEI 4 sub-Plinian eruption
The VEI 4 sub-Plinian simulation for the summer period (Fig. 3a) shows that Vila Franca do Campo, located southeast of Fogo caldera, is the most affected municipality of São Miguel, with > 3 m of tephra fall deposited in the caldera and immediately to east-southeast of the vent. Almost all the municipality is affected by tephra deposition up to 2 m thick. The predominant winds 275 from west-northwest also lead to the deposition of tephra on the western half of Povoação municipality (between 50 cm and 1 mm, from west to east) and on the southern part of Ribeira Grande municipality (up to 3 m close to the caldera).
For the winter period (Fig. 3b), the prevailing winds from the west and the higher wind intensities promote the deposition of tephra on the eastern part of the island. Although Vila Franca do Campo is located on the southeast flank of the volcano, the strong westerly winds cause most of the tephra to be deposited on the northern sector of the municipality, with > 3 m 280 immediately to the east of the caldera. However, in this scenario the entire Povoação municipality is affected by tephra fall that can reach maximum thicknesses of 1 to 2 m in much of its area. Ribeira Grande and Nordeste municipalities are also affected on their southern sectors, although with smaller thicknesses (1 m to 1 mm, from south to north).

Tephra fallout from VEI 5 Plinian eruption
The VEI 5 Plinian simulations show the same dispersal patterns of the VEI 4 sub-Plinian simulations described above, but 285 with greater extents. For the summer period (Fig. 4a), Vila Franca do Campo municipality is the most affected, with > 3 m of tephra deposited on its northern part, east-southeast of the caldera. Much of the municipality is affected by tephra deposition between 1 and 3 m thick. The neighbouring Povoação municipality is affected by tephra fall up to 1 m thick, particularly on the western side. The southern part of Ribeira Grande municipality is also affected by thicknesses of > 3 m of tephra close to the caldera. 290 For the winter period (Fig. 4b), the predominant stronger winds blowing from west lead to the deposition of tephra on the eastern part of São Miguel. Vila Franca do Campo is affected by thick tephra deposition (> 3 m) on the northern sector of the municipality. Given the higher intensities of the wind, tephra thicknesses of 1 to 3 m are deposited on much of Povoação municipality, reaching > 3 m on the western part. Ribeira Grande and Nordeste municipalities are also affected on their southern sectors, with thicknesses of 2 m to 1 mm, from south to north. 295

Pyroclastic Density Currents
The simulations of PDCs for the VEI 4 sub-Plinian and VEI 5 Plinian eruptions (Fig. 5) show the maximum potential extent of these volcanic products generated from within Fogo caldera. In both cases, PDCs are not contained inside the caldera and flow down the flanks of the volcano. The entire central part of São Miguel may be affected by PDCs, which reach the sea on the north and south coasts. As expected, the VEI 4 and VEI 5 simulations differ in the maximum potential extent that the PDCs 300 can achieve. The VEI 4 simulation (blue dotted line in Fig. 5) shows that the western half of Vila Franca do Campo, most of Lagoa and the central part of Ribeira Grande municipalities are affected by PDCs, which can reach maximum distances of 9 km from the caldera. In the case of the VEI 5 simulation (red dashed line in Fig. 5), more than two-thirds of Vila Franca do Campo, all of Lagoa and the central part of Ribeira Grande municipalities are affected by PDCs that can reach maximum distances of 11 km from the source. 305

Exposed elements at risk
The inventory and characterization of buildings related to tourism in Vila Franca do Campo municipality (140 in total) revealed that 46 correspond to accommodation units (33 %), 51 to restaurants (36 %), 29 are destined to culture or cultural activities (21 %) and 14 buildings are related to tourism animation/activities (10 %) (Fig. 6). Infrastructure such as the marina, the fishing port and the water park were also considered in the analysis. 310 The identification of the elements at risk was attained by overlapping the mapped exposed elements (buildings and infrastructure) with the results of the simulations (Figs. 7 to 9). Regarding tephra fallout scenarios, it is important to distinguish the type of destruction that the buildings can suffer depending on the tephra thicknesses. When a building (represented by a polygon) was overlapped by more than one thickness class we considered the most conservative option, by choosing the class with the higher thickness. The impact on the buildings was assessed according to the expected type of destruction following 315 critical thickness thresholds.
Buildings affected by 1 mm to 20 cm of tephra and subject to constant and careful cleaning of the roof are usually not damaged or may suffer only minor damage. For the summer period (Figs. 7a and 8a), 62.1 % and 61.4% (VEI 4 sub-Plinian and VEI 5 Plinian scenarios, respectively) of all buildings related to tourism are likely to suffer thicknesses of up to 20 cm. On the other hand, for the winter period (Figs. 7b and 8b) buildings in Vila Franca do Campo municipality will only be covered by a 320 maximum of 5 cm of tephra. The affected buildings are located in Ponta Garça, the easternmost parish, and correspond to 5 % and 15 % (VEI 4 sub-Plinian and VEI 5 Plinian scenarios, respectively) of all touristic buildings (see Table 2).
Buildings covered by ≥ 20 cm of tephra can suffer significant damage, such as roof collapse (Pomonis et al., 1999). For both summer scenarios (Figs. 7a and 8a), approximately 31 % of the buildings are in these conditions. In more extreme cases, buildings affected by ≥ 1 m of tephra are likely to suffer total collapse (Blong, 1984;Spence et al., 2005). For this case, the 325 percentage of affected buildings is higher in the VEI 5 Plinian scenario with 19 %, whereas in the VEI 4 sub-Plinian scenario is 9.3 % ( Table 2).
On what concerns PDCs, it should be noted that the distinction between dense and dilute currents was not taken into account in this analysis and therefore only total destruction was considered. The percentage of tourism-related buildings located within the maximum potential extent of PDCs (Fig. 9) from a VEI 4 sub-Plinian eruption is 87.1 %, whereas from a VEI 5 Plinian 330 eruption is 95 %.
In addition to buildings, infrastructure related to the tourism sector in Vila Franca do Campo municipality will suffer similar damage to the buildings when affected by tephra fallout (summer period) and PDCs. Like the other exposed elements, infrastructure will not be affected by tephra fallout in the winter period.

Assessment of the economic impact on tourism 335
The economic impact of future explosive eruptions of Fogo volcano on the tourism sector of Vila Franca do Campo municipality was calculated considering three economic scenarios. The number of accommodation buildings affected in each economic scenario is shown in Table 3.
The sum of the annual revenue of each accommodation unit resulted in an estimated total annual revenue of approximately 9.5 million euros, assuming that the accommodation capacity was fully occupied. Combining the revenue data with the chosen 340 occupancy and discount rates allowed to estimate the LPV for a period of 30 years for each of the economic scenarios (Fig.   10). Table 4 summarizes the occupancy rates ( ), discount rates ( ) and LPV for the three economic scenarios considered.
The graphs of figure 10 show that in all scenarios the LPV is greater for higher occupancy rates and lower discount rates. For example, in economic scenario 3 (PDCs from a VEI 5 Plinian eruption) the LPV after 30 years is almost 145 million euros, when the applied occupancy rate is 65 % and discount rate is 2 %. On the other hand, if an occupancy rate of 65 % and a 345 discount rate of 4 % are considered, the loss is approximately 113 million euros. The lowest value of the updated loss is 87 million euros for an occupancy rate of 50 % and a discount rate of 4 %. In terms of LPV per year, there is a decreasing trend over time in all scenarios. For economic scenario 3 and considering both discount rates, the loss value in year zero does not reach 5 million euros, for occupancy rates of 50 %, and does not exceed 6 million euros, for a higher occupancy rate of 65 % (Fig. 10c). However, over the years the LPV downward trend is more evident when a discount rate is applied. For an occupancy 350 rate of 65 %, the loss value in year 30 is 1.9 and 3.4 million euros for discount rates of 4 % and 2 %, respectively, while for an occupancy rate of 50 %, this value is approximately 1.4 and 2.6 million euros, considering discount rates of 4 % and 2 %, respectively.
The use of a discount rate equal to 0 % means that the society would attribute to a given monetary amount the same value in the future and in the present. Although this scenario is considered less realistic among economists and other agents, its use 355 might be important to assess the impact of discounting. As expected, this exercise yields the highest LPV in all economic scenarios (Table 4). For instance, and regarding economic scenario 3, LPV would amount to 192 million euros, at zerodiscounting and when a 65 % occupancy is applied. This means that, in such a case, discounting at 2 % and 4 % rates reduces LPV by nearly 47 and 79 million euros, respectively. Other similar calculations can be performed, namely for economic scenarios 1 and 2, based on the values included in Table 4. 360

Impact of tephra fallout and PDCs from Fogo volcano
Fogo volcano is considered one of the most hazardous volcanoes in the Azores archipelago. Despite the low eruptive frequency of sub-Plinian and Plinian eruptions, with a recurrence interval of 833 years (Gaspar et al., 2015a), Fogo volcano produced the 365 largest eruption in the Azores in the last 5000 years (Fogo A) and one of the most recent explosive eruptions of the archipelago (Fogo 1563) (Walker and Croasdale, 1971;Booth et al., 1978;Wallenstein, 1999). Even in non-eruptive periods, Fogo volcano is a hazardous landform with significant seismicity (Silva et al., 2015; and geothermal activity, manifested in fumarolic fields, CO2-cold and thermal springs, and diffuse degassing areas (Viveiros et al., 2015). Several unrest episodes have occurred in the last 20 years, with intense earthquake swarms lasting for several months and accompanied by ground deformation 370 (inflation episodes), such as in May-September 2005 (Wallenstein et al., 2007;Silva et al., 2012). Due to its location, in the central part of São Miguel, a future explosive eruption of Fogo volcano will severely impact the parishes located on the flanks of the volcano, but also others in more distal areas depending on wind conditions. Such an eruption will have significant longterm economic consequences for the island and even for the entire Azores archipelago. Next, we discuss the impact of tephra fallout and PDCs on São Miguel, following the eruptive scenarios defined in this study. 375 The simulation of tephra fallout from a VEI 4 sub-Plinian eruption (most probable scenario) during the summer period, shows that the prevailing winds blowing from west-northwest promote the deposition of tephra within the caldera and to the eastsoutheast of the vent, making Vila Franca do Campo municipality the most affected, with tephra deposition up to 2 m thick.
For an eruption occurring during the winter period, the predominant winds blowing from the west and the higher wind intensities, lead to the deposition of tephra on the eastern part of the island. Validation of the simulation results can be made 380 by comparison with the geological record, namely with the deposit of the Fogo 1563 eruption, which has a well-defined dispersal axis towards the east (see the isopach map of the Fogo 1563 deposit in figure 20 of Walker and Croasdale, 1971).
Regarding tephra fallout from a VEI 5 Plinian eruption (worst-case scenario), the simulation results show the same dispersion patterns as the VEI 4 sub-Plinian eruption scenario, for summer and winter conditions, though with thicker tephra deposition, due to the larger volume of ejected material. 385 The impact of tephra fallout on buildings and infrastructure will depend on the thickness of accumulated tephra, which is translated as static load. In localities affected by the accumulation of 20 cm or more of tephra, such as those on the central and eastern parts of Vila Franca do Campo, buildings would suffer significant damage. While in localities where the 1 m threshold is exceeded, such as Ponta Garça in the summer period, buildings would likely suffer total collapse, including constructions reinforced with concrete (Blong, 1984;Spence et al., 2005). However, if tephra is wet these critical thickness thresholds are 390 substantially reduced (Spence et al., 2005). This possibility should not be overlooked given the rainy Azorean climate (Hernández et al., 2016).
Other elements such as the ground transportation network would also be affected by tephra fallout. As the main roads of Vila Franca do Campo are located on the southern part of the municipality, they are mostly vulnerable to the accumulation of tephra fall deposits during the summer period. Marine transportation can also be affected by tephra fallout as ports and marinas 395 become inoperable. Particularly during the summer period, marine operations along the south coast of São Miguel would be severely affected and important fishing ports, such as Vila Franca do Campo, would be brought to a halt.
The simulations of PDCs, for VEI 4 sub-Plinian and VEI 5 Plinian eruptions, show that these currents are not contained inside the caldera but are able to overcome the caldera walls and flow down the flanks of the volcano. The eruption of such volcanic products would affect the entire central part of São Miguel, reaching the sea on the north and south coasts of the island. The 400 maximum potential extent of the PDCs is in the order of 9 and 11 km from the caldera for VEI 4 and VEI 5 scenario, respectively. It should be noted that in case of a real event, the progression of PDCs is strongly controlled by topography and channelled through valleys and depressions along the flanks of the volcano. Validation of these simulations can be done by comparison with the geological record of PDC deposits. Fogo A ignimbrites were emplaced quasi radially outward from the caldera, within narrow paleo-valleys on the southern flank and along Ribeira Grande graben on the northern flank, reaching 405 the sea in both cases (see distribution maps of the Fogo A ignimbrites in Pensa et al., 2015a,b).
The impact of PDCs on buildings and infrastructure is mostly related to their dynamic pressure and temperature. However, in this case we do not distinguish between dense and dilute currents, and assumed a binary impact approach which considered the absence of damage or total destruction of buildings and infrastructure by PDCs. Ground and marine transportation networks are also highly vulnerable to PDCs as roads, ports and marinas would become buried or suffer irreparable heat damage. 410

Impact of explosive eruptions on the economy of tourism of Vila Franca do Campo
The main economic activities in the Azores are public services, retail and wholesale trade, fishing, livestock ranching, and production of dairy products (Vieira et al., 2019). Tourism is a growing and promising activity for job creation and for the development of this archipelago, where the landscape and marine-related activities constitute the main attractions (Calado et al., 2011;Torres et al., 2017;Vieira and Antunes, 2017). From 2001 to 2018, there has been an almost constant increase in 415 jobs related to tourism in the Azores. In 2015 tourism employed 11 847 people and in 2018 it reached 19 614 people. These values give a good note of the importance that this sector has assumed, growing to represent approximately 20 % of all employment in the Azores (Fortuna et al., 2020). The number of visitors sharply expanded after the liberalization of airspace in two corridors between the mainland and the Azores since 2015, which allowed the entrance of low-cost airlines (Vieira et al., 2019). According to satellite tourism accounts, tourism accounted for 6.7 % of the gross value added (GVA). At the same 420 time, the consumption of goods and services by tourists amounted to 14.1 % of the GDP (Azorean Statistical Office, 2018).
The location of Vila Franca do Campo municipality in the southern part of São Miguel Island, approximately 20 km east of Ponta Delgada city (the island's capital), together with its moderate climate and numerous bathing areas along the coast attract many tourists throughout the year. Therefore, tourism is a major source of revenue for this municipality. In 2018, the estimated total annual revenue of the accommodation units on Vila Franca do Campo was approximately 9.5 million euros (assuming 425 that the accommodation capacity would be fully occupied). This figure clearly shows the importance of this sector on the economy of the municipality and of the entire island of São Miguel.
Looking at the calculated LPV for a period of 30 years for each of economic scenario, we can conclude that this value is greater when the occupancy rate is higher and the discount rate is lower. Economic scenario 1 (tephra fallout from a VEI 4 sub-Plinian eruption during the summer period) has the smallest LPVs for the different rates used. The other two economic scenarios 430 (PDCs from a VEI 4 sub-Plinian eruption and from a VEI 5 Plinian eruption) have similar LPVs because they comprise 96 % and 98 % of the accommodation buildings, respectively (Table 3). Still, and as expected, economic scenario 3 represents the higher loss, almost 145 million euros (Table 4). As seen in figure 10, the LPV shows a decreasing trend over the years, as result of the applied methodology, except when the discount rate of 0 % is applied, as it causes LPV to remain constant over time. 435 The method used in this study is a first attempt to quantify the economic loss of the tourism sector resulting from future explosive eruptions of Fogo volcano. It should be noted that both the annual revenue and the LPV obtained for the tourism economy were estimated using only the values from the accommodation units of Vila Franca do Campo municipality. To achieve a more realistic estimation of the total revenue related to the tourism economy, the benefits of other sectors, such as restaurants or tourism animation/activities, should also be considered. The tourism industry also contributes to the economy 440 by generating jobs in hotels, guest houses, local accommodations, restaurants, souvenir shops, etc. and therefore the workers' salaries should also be quantified. Indirect losses should also be accounted for, such as a drop in the consumption of fish, dairy or beef products in the food sector, thus affecting the economy of other sectors such as agriculture and sea. Nevertheless, the present methodology may be adopted to calculate the loss of revenue of other sectors related to tourism and also applied to different volcanic regions vulnerable to explosive volcanism. 445

Conclusions
This study presents a new approach to quantify the impact of explosive volcanic eruptions on the tourism industry. We determined the economic loss related to future explosive eruptions of Fogo volcano (São Miguel Island), by estimating the benefits generated by the accommodation units of Vila Franca do Campo municipality.
Two eruptive scenarios were considered for Fogo volcano, the most probable scenario (a VEI 4 sub-Plinian eruption) and the 450 worst-case scenario (a VEI 5 Plinian eruption). We evaluated the vulnerability of tourism-related buildings and infrastructure in Vila Franca do Campo to tephra fallout and PDCs by analysing their loss of functionality. The Loss Present Value (LPV) method was used to estimate the benefits generated by the accommodation units for different economic scenarios.
The simulations show that tephra deposition from a VEI 4 sub-Plinian eruption during the summer period occurs to the eastsoutheast of Fogo caldera, while during the winter period the deposition is to the east of the caldera. For a VEI 5 Plinian 455 eruption the dispersion patterns are similar but with a larger dispersion area and thicker tephra deposition. The simulations of PDCs show that the central part of São Miguel is the most affected, as currents flow down the flanks of the volcano, reaching the sea on both coasts.
The assessment of the economic impact on the tourism sector shows that economic scenario 1 (tephra fallout from a VEI 4 sub-Plinian eruption) has the lowest LPV when compared to economic scenarios 2 and 3 (PDCs from VEI 4 sub-Plinian and 460 VEI 5 Plinian eruptions, respectively), which have similar LPVs. Although economic scenario 3 is not the most likely, as it represents near total destruction of Vila Franca do Campo municipality, it corresponds to the higher economic loss, with approximately 145 million euros over 30 years.
Tourism is a growing industry worldwide and in the Azores has had an increasing importance in the economy since 2015.
However, the Azores and other active volcanic regions are vulnerable to future eruptions, which may have long-term economic 465 consequences. Volcanic hazard and risk assessment is therefore essential in areas where people live side by side with active volcanoes, in order to provide the competent authorities with appropriate strategies to mitigate volcanic risk, such as land use planning, emergency management and post-disaster economic recovery planning.