Climate change risk in the Swedish ski industry

ABSTRACT

Tourism industry and government demand for knowledge of the impacts of climate change on ski tourism is growing. Despite the more than 70-year history and large cultural significance of alpine skiing in Sweden, little is known about the industry’s future under a changing climate. This study applies the SkiSim2 model with low to high emission scenarios (RCP2.6 to 8.5) to analyse the implications of climate change for ski operations (season length, snowmaking requirements) at 23 alpine ski areas across Sweden for the early, mid and late twenty-first century. Northern areas of Sweden show much less reduction in average season length compared to central and southern Sweden under the high emission mid- (13% versus 58% and 81%) and late-century scenarios (27% versus 72% and 99%). To limit season losses in these scenarios, snow production increases of over 250% are required in all regions. Such increases will create additional financial and environmental stressors, which may lead to the closure of the most at-risk resorts. With greater impacts projected for much of the European Alps ski market, northern Sweden may represent a ‘last resort’ for the European ski industry under higher emission scenarios by the mid-late twenty-first century.

KEYWORDS:

• Climate change
• Sweden
• ski industry
• adaptation
• snowmaking
• winter tourism

1. Introduction

Sweden represents an under-researched ski market, despite its position as the ninth-largest market globally in terms of skier visits and percentage population participation (Vanat, 2020). Following a long period of stable skier visits, Swedish ski resorts experienced recent growth, with the 2017/2018 and 2018/2019 seasons reaching record lift pass sales of 1709 and 1738 million SEK, respectively (SLAO, 2021a). Like all ski tourism markets, the 2019/2020 season was disrupted by the COVID-19 pandemic. Nonetheless, Swedish lift pass sales reached 1519 million SEK. By contrast, in the 2020/2021 season, lift pass sales increased 21% to record levels of 1842 million SEK as many Swedes were forced to remain in Sweden (SLAO, 2021a).

The Swedish ski industry has made substantial investments that are contributing to this recent growth trend, including expanding accommodation and hospitality capacities in resorts, the renewal of lift systems, and the development of the Scandinavian Mountains Airport (SMA) (Demiroglu et al., 2019). The airport was developed with a leading Swedish ski tourism company as well as state and municipal support including a government grant of 250 million SEK. The SMA aims to increase accessibility to the Swedish ski resorts of Sälen and Idre Fjäll, and Trysil and Engerdal in Norway for domestic and international tourists, especially those from Denmark, the UK, Germany and Russia (Demiroglu et al., 2019). The 2020/2021 season was poised to be the first full season for the SMA; however, international travel restrictions due to COVID-19 have limited international operations.

The paucity of research on climate change and the future of the Swedish ski industry and ski tourism is surprising, given that climate change is observed and projected to occur faster in high latitude areas (IPCC, 2014). Additionally, several governmental reports suggest that northern Sweden may hold a competitive advantage over other ski tourism regions (e.g. European Alps, southern Sweden and Norway), presenting potential market opportunities in the decades ahead as tourists look to substitute their ski holidays to more snow reliable destinations. This has resulted in Sweden’s framing as a potential European ‘last resort’ for skiing in an era of accelerating climate change (Demiroglu & Sahin, 2015, p. 5).

The degree to which this climate change market advantage is taken seriously is highlighted by a report from the Swedish Commission on Climate Change and Vulnerability (2007, p. 395):

After the year 2040, the situation for winter tourism looks more serious. The high season weeks around Christmas and New Year, as well as Easter, will be ‘green’ to an increasing extent. As far as we can judge, this trend will increase towards the end of the century. A structural shift of winter tourism towards areas that are more assured of having snow in the northernmost parts of the country may then become necessary.

The Commission’s report was the first to shape the political discourse on climate change adaptation for Swedish winter tourism, and while it was completed over a decade ago, more recent reports by the Swedish Meteorological and Hydrological Institute (SMHI) have made similar claims (Andersson et al., 2015; Sjökvist et al., 2015). These claims are based on projections of natural snow resources, and the omission of snowmaking in climate change risk assessments (both in Sweden and other competing markets) raises important questions about adaptive capacity. The work of Andersson et al. (2015) and Sjökvist et al. (2015) are also cited by the National Climate Change Adaptation Strategy (Regeringen, 2018) and Sweden’s Seventh National Communication on Climate Change (Swedish Environmental Protection Agency, 2017), which further discuss the challenges Swedish ski tourism may encounter under climate change, including increased water resource demand and conflict, as well as the potential indirect competitive benefits due to reductions in snow reliability in the European Alps. This competitive advantage may, however, be limited to more northerly resorts as tourists visiting more southerly areas of Sweden and the European Alps look to substitute their holidays to more snow reliable destinations. Neither of these discussions around shifts in continental competitiveness appear to have consulted studies on skiing in other European markets (e.g. Spandre et al., 2019; Steiger, 2010; Steiger & Scott, 2020).

Two studies have examined the implications of climate change for the Swedish ski industry; neither of which are cited in the aforementioned government reports. Moen and Fredman (2007) examined the potential impact of climate change on the large ski resort of Sälen in mid-Sweden. They predicted a reduction in season length of 64–96 days, from the 162-day baseline (1990–2001) in 2070–2100. This season loss was then curiously extrapolated to the entire Swedish ‘mountain region’ ski industry to estimate economic losses of 946.5 to 1755.3 million SEK by 2070–2100 using the B2 (low emission) and A2 (high emission) climate change scenarios. These findings are uncertain and potentially overestimated due to the use of snowfall rather than snowdepth as an indicator of natural snow reliability and the absence of adaptive snowmaking to estimate ski season length. The extrapolation to the Swedish ski industry across the entire mountain region (∼61 to 68° North, Figure 1), despite the diverse climate and terrain in the country, is also questionable.

Figure 1. Location of ski areas included in the study.

Brouder & Lundmark, 2011 examined intra-regional perceptions of vulnerability for winter tourism businesses in Northern Sweden. Their results showed coastal areas perceived themselves to be more exposed to climate change than inland areas, and a general perception from companies, which has held true, that climate change would not significantly impact the ski tourism industry in the 2010s. This research is limited due to its focus on the Upper Norrland region and the decade of the 2010s, which climate projections indicated would not depart significantly from natural variability in the 2000s.

To address the limited foresight into the future of the Swedish ski industry in an era of accelerating climate change, the aim of this paper is to: (1) provide the first analysis of the impact of climate change for 23 alpine ski areas across all regions of Sweden for early (2030s), mid (2050s) and late (2080s) century, under low, moderate and high emission scenarios (RCP2.6, 4.5 and 8.5); (2) conduct the first climate change impact analysis for Sweden that includes snowmaking, which is a widespread and increasing climate adaptation; and (3) compare regional market impacts, critically examining the proposition that parts of the Swedish ski industry are less at risk and could benefit from a transfer of ski tourism demand from the European Alps.

2. Methodology

2.1. Study area and resort characteristics

The locations modelled in this work are the major Swedish ski areas listed by the Svenska Skidanlanläggningars Organization (SLAO, 2021a). Several areas were omitted from the study due to a lack of data on snowmaking capacity and insufficient time series of local scale climate data (Table 1). The 23 ski areas included in the study represent 42% of hectares of Swedish ski areas compiled by (Demiroglu et al., 2019) as well as 37% of pistes and 26% of lifts listed by Ski Resort Info (2021). The location of resorts in the study are displayed in Figure 1.

Table 1. Characteristics of ski area.

Ski resort number	Region	Latitude	Longitude	Ski area size (ha)	Elevation range (m)	Snowmaking (% terrain covered)	Lifts	Pistes (km)	Season end
1	Norrland	63.40 N	13.08 E	1361	380-1274	60	31	91	5th May
2	Norrland	63.17 N	14.64 E	307	590-1010	18	17	25	7th April
3	Svealand	61.94 N	12.99 E	82	675-985	30	3	20	21st April
4	Svealand	59.23 N	17.88 E	14	3-103	100	3	2.7	7th April
5	Norrland	63.16 N	17.24 E	35	71-319	100	5	5	5th May
6	Svealand	59.30 N	18.11 E	11	10-95	80	2	2	7th April
7	Norrland	62.08 N	16.73 E	90	160-450	100	7	16	21st April
8	Norrland	65.71 N	15.29 E	289	439-789	7	10	30	5th May
9	Svealand	60.15 N	12.99 E	31	403-452	100	6	11	7th April
10	Gotaland	57.44 N	13.62 E	32	159-309	100	8	6	7th April
11	Norrland	65.25 N	14.59 E	33	545-950	0	4	18	5th May
12	Svealand	61.03 N	13.34 E	161	420-665	98	19	32	21st April
13	Norrland	62.11 N	13.30 E	193	633-1125	64	8	27	7th April
14	Norrland	62.70 N	12.39 E	153	700-1000	17	9	25	21st April
15	Norrland	68.43 N	18.13 E	299	522-909	0	6	21	5th May
16	Norrland	66.02 N	21.25 E	46	60-240	33	4	7	21st April
17	Svealand	58.32 N	14.92 E	17	177-276	100	6	3	7th April
18	Norrland	62.57 N	12.24 E	172	743-1038	11	16	38	5th May
19	Gotaland	57.97 N	15.62 E	25	115-245	100	4	4.5	7th April
20	Norrland	63.27 N	13.16 E	83	673-990	14	8	22	21st April
21	Gotaland	57.82 N	13.41 E	18	179-288	100	7	2.6	7th April
22	Svealand	59.83 N	12.26 E	16	100-320	38	5	8.4	7th April
23	Norrland	62.48 N	13.97 E	546	448-946	12	10	18	5th May

Note: Lift, pistes and elevation data from Ski Resort Info (2021). Snowmaking data from personal correspondence with resorts. Ski area size data from Demiroglu et al. (2019). Season end date from personal correspondence with resorts or Ski Resort Info (2021).

2.2. SkiSim

Early vulnerability assessments of the ski industry overestimated climate change impacts due to the absence of snowmaking in models. Therefore, Scott et al. (2003) developed the SkiSim model, initially testing it in Ontario, Canada, with further applications in Quebec (Scott et al., 2007) and New England, USA (Dawson & Scott, 2013). The model was superior for simulating observed season lengths and demonstrated that previous studies without snowmaking had vastly overestimated the impacts of climate change on the ski industry. More recent repeat studies with a refined model reinforce this critical finding and highlight how other studies that continue to omit snowmaking misrepresent climate change risk particularly through the mid-century (Scott et al., 2017).

SkiSim was further refined for application in the European Alps (Steiger, 2010). SkiSim2 allowed for outputs for the entire altitudinal range of a ski area in 100 m intervals and included rules on snow production, which reflected operational decision-making throughout a season. SkiSim2 has been applied in the European Alps (Abegg et al., 2015; Steiger & Abegg, 2018; Steiger & Stötter, 2013), North America (Scott et al., 2017; Scott & Steiger, 2013), Norway (Scott et al., 2020), China (Fang et al., 2019), previous Winter Olympic host locations (Scott et al., 2014, 2017), and a comprehensive study across the eastern European Alps (Steiger & Abegg, 2018).

The version of SkiSim2 utilized in this study contains three subcomponents: a natural snow model, a snowmaking module and ski area operational rules. The natural snowpack modelling is calculated using observed daily minimum and maximum temperatures and precipitation data collected from 1980 to 2019. The following parameters are calibrated for each used weather station: in the first step, a lower and upper-temperature threshold distinguishing rain, snow and mixed precipitation is calibrated; in a second step, the degree-day factor defining the amount of snow water equivalent that is melted per 1°C daily mean air temperature is calibrated. For the first step, total snowfall is used for calibration and performance evaluation. For the second step, the length of the snow season is used for calibration and evaluation of model performance. Two independent sets of multiple years are used for calibration and evaluation of the model.

The SkiSim2 model set up in this study is consistent with the thresholds used to define operational ski days (more than 30 cm snow depth) in previous studies (e.g. Scott et al., 2020) for international comparability, with a uniform assumed state-of-the-art snowmaking system (10 cm/day) (Steiger & Scott, 2020) and snowmaking periods and climatic thresholds defined by Swedish market conditions (based on consultations with ski area operators). The metric of required snow used in this work refers to the required amount of produced snow to maintain a ski season from 15th December to (depending on the ski area) 7th April, 21st April or 5th May. This is regardless of the possibility to produce snow.

2.3. Climatological data and climate change scenarios

Meteorological data for the baseline period of 1981–2010 were obtained from the SMHI climate observation stations (SMHI, 2020a). Climate station selection criteria were the proximity of the station to the ski area, the completeness of the data record and the altitude of the station, which should be within or close to the altitudinal range of the ski area. In some locations, the closest observation station could not be used due to lack of the relevant data over the required timeframe. In such situations, the next nearest station was used. Where possible, each ski area’s mid-lift station altitude is used; however, several smaller resorts have no mid-station, and their base elevation is used as their critical altitude.

This study used climate change scenarios developed for the IPCC Fifth Assessment Report (IPCC, 2014) which were obtained from the SMHI. An ensemble scenario, based on nine global climate models, was used to represent the range of potential climate futures (SMHI, 2020b). Downscaling is achieved using the Rossby Centre’s regional atmospheric model RCA 3/4, which covers Europe with a grid resolution of roughly 50 × 50 km (SMHI, 2020b). Three emissions pathways are considered: RCP2.6 (carbon emissions peak around 2020 and reach net-zero by 2100), RCP4.5 (emissions reductions pledged during the Paris Climate Change Agreement) and RCP8.5 (emissions continue at current trajectories) (IPCC, 2014). The reference climate period is 1981–2010, with the future time periods being 2030s (2021–2050), 2050s (2041–2070) and 2080s (2071–2100).

3. Results

As 92% of the modelled ski areas have snowmaking capacity, the SkiSim model was run in both a natural snow scenario and a scenario incorporating snowmaking. In the natural snow scenario, the average ski season length in the baseline period is 181 days in Norrland, 57 days in Svealand and 40 days in Gotaland (Table 2). Under natural snow conditions, regional differences can be seen in the impact of climate change. The differences between Gotaland and Norrland are stark. Compared to the baseline under all climate change scenarios, there is at least an 83% reduction in operational days in Gotaland. Reductions in Norrland are restricted to less than 31% except under the warmest climate change scenario of RCP8.5 in the 2080s. The drastic reduction in operational days in Gotaland is expected as it is at a lower latitude where temperatures are higher than elsewhere in Sweden, and ski areas are closer to the critical threshold where snowfall occurs less often. Additionally, ski areas modelled in Norrland and Svealand have the climatic advantage of higher altitudes, with average critical elevations of 584 and 460 m compared to 184 m in Gotaland. The flattening of season length reduction witnessed in RCP 2.6 between the 2050s and 2080s can be explained by atmospheric CO₂ concentrations beginning to stabilize after mid-century (Table 2).

Table 2. Ski season length (days and percentage reduction compared to the baseline) with only natural snow.

Region	Baseline (days)	RCP2.6			RCP4.5			RCP8.5
		2030s	2050s	2080s	2030s	2050s	2080s	2030s	2050s	2080s
Norrland(sample group 1)	181	150 −17%	148 −18%	149 −18%	148 −18%	139 −23%	127 −30%	144 −20%	126 −30%	87 −52%
Svealand (sample group 2)	57	44 −23%	46 −19%	42 −26%	42 −26%	40 −30%	30 −47%	42 −26%	35 −39%	4 −93%
Gotaland (sample group 3)	40	7 −83%	3 −93%	3 −93%	2 −95%	1 −98%	0 −100%	1 −98%	0 −100%	0 −100%
Sweden (samples 1–3)	125	99 −21%	98 −22%	98 −22%	97 −22%	91 −27%	81 −35%	94 −25%	82 −34%	51 −59%

Table 3 displays season length when advanced snowmaking is applied over 100% of skiable terrain, a capacity most ski areas in the study areas do not currently have but could develop. Consistent with the international literature (Abegg et al., 2015; Dawson & Scott, 2013; Scott et al., 2014, 2017, 2020; Scott & Steiger, 2013; Steiger & Abegg, 2013, 2018; Steiger & Stötter, 2013), the length of the ski season when advanced snowmaking is available increases in the baseline and significantly limits reductions in season losses and regional variation under climate change, with Svealand and Gotaland showing less reduction in operational days than for the natural snow scenario. The difference in ski season length reduction regionally becomes more pronounced in the late-century, specifically for higher emission scenarios, where Gotaland becomes the area with the shortest ski season, whilst Norrland in all scenarios maintains ski seasons with over 150 operational ski days (Table 3).

Table 3. Season length (days and percentage reduction compared to the baseline) with snowmaking.

Region	Baseline	RCP2.6			RCP4.5			RCP8.5
		2030s	2050s	2080s	2030s	2050s	2080s	2030s	2050s	2080s
Norrland (sample group 1)	219	205 −6%	201 −8%	204 −7%	202 −8%	197 −10%	190 −13%	201 −8%	190 −13%	159 −27%
Svealand (sample group 2)	156	114 −27%	106 −32%	106 −32%	101 −35%	85 −46%	67 −57%	98 −37%	66 −58%	44 −72%
Gotaland (sample group 3)	140	114 −19%	101 −28%	101 −28%	95 −32%	69 −51%	48 −66%	89 −36%	27 −81%	1 −99%
Sweden (sample group 1–3)	189	166 −12%	159 −16%	160 −15%	157 −17%	146 −23%	134 −29%	155 −18%	131 −31%	104 −45%

The majority of ski areas in Sweden do not currently have the advanced snowmaking capacity modelled in this scenario and financing the upgrades would require significant capital investment. Snowmaking as a climate change adaptation also becomes less efficient under greater warming scenarios, as the technology requires sufficiently cold temperatures to produce cost-effective snow. Therefore, under progressive climate change, warming temperatures will reduce the number of days considered efficient to produce snow and increase the costs to make required snow. Nevertheless, significant advancements in snow production technology cannot be ruled out.

To be considered snow reliable, a resort needs to have 100 operational days (>30 cm of snow) for 70% of seasons over 30 years (Abegg et al., 2015; Morin et al., 2021). Under natural snow conditions, 85% of Norrland’s ski areas can currently be considered snow reliable. The majority of Norrland’s ski areas maintain their natural snow reliability under all scenarios, with at least 69% of resorts remaining snow reliable in all scenarios except in RCP8.5 2080. In Svealand and Gotaland, the percentage of naturally snow reliable resorts does not reach above 29% or 0%, respectively (Figure 2).

Figure 2. Proportion of snow reliable ski areas under natural snow conditions.

As indicated, 92% of ski areas in this study have some snowmaking capacity. Therefore, it is vital to examine snow reliability with snowmaking. All of Norrland ski areas are snow reliable with advanced snowmaking until RCP8.5 2080 scenario where it drops to 92% (Figure 3). For Svealand and Gotaland, all areas during the baseline period are snow reliable, thereafter a maximum of 43% of Svealand’s ski areas are snow reliable, and only 67% of Gotaland’s ski areas are snow reliable until RCP4.5 2050 (Figure 3).

Figure 3. Proportion of snow reliable ski areas with advanced snowmaking.

To protect and prolong the ski season, snow production increases are needed, especially in central and southern Sweden (Table 4). The spatial and temporal patterns of snow production increases are similar to changes in season length. Snow production requirements increase throughout time and particularly under higher emissions scenarios, with the largest increases in snow production seen for the 2080s under the RCP8.5 scenario. Although Norrland’s ski areas have higher percentage increases in snow production requirements, this is owing to the low level of snow production they require during the baseline period. When the depth of snow needing to be made is considered, it is much higher in Svealand and Gotaland (Table 4).

Table 4. Increases in snow production are required under climate change.

Region	Baseline	RCP2.6			RCP4.5			RCP8.5
		2030s	2050s	2080s	2030s	2050s	2080s	2030s	2050s	2080s
Norrland (sample group 1)	16cm	36 cm 125%	39 cm 144%	39 cm 144%	37 cm 131%	50 cm 212%	64 cm 300%	41 cm 156%	61 cm 281%	151 cm 844%
Svealand (sample group 2)	129cm	198 cm 54%	206 cm 60%	218 cm 69%	215 cm 67%	260 cm 102%	297 cm 130%	225 cm 74%	299 cm 132%	451 cm 250%
Gotaland (sample group 3)	70cm	137 cm 96%	154 cm 120%	161 cm 130%	163 cm 133%	207 cm 196%	237 cm 239%	169 cm 141%	257 cm 267%	396 cm 466%
Sweden (sample group 1–3)	58cm	98 cm 69%	105 cm 81%	109 cm 88%	107 cm 85%	134 cm 131%	157cm 171%	114 cm 97%	159 cm 174%	263 cm 353%

The ski season length results without snowmaking show 14 ski areas are currently snow reliable, with only location 3 and 12 outside Norrland. For the baseline period, no ski area south of 12 (61.03 N) is considered snow reliable. By the 2080s, under high emissions (RCP8.5), only seven resorts remain snow reliable, with all of these in Norrland. Of the seven remaining snow reliable areas, all have base altitudes above 400 m and three areas lie in or within a degree latitude of the Arctic circle.

With advanced snowmaking, all 23 areas are considered snow reliable during the baseline period. As early as the 2030s, under even the low emission RCP2.6 scenario, the areas around sea level of locations 4 and 6 and the southerly areas of 9 and 19 are no longer snow reliable, suggesting they are already at the climate margins for ski tourism. In the 2080s, under the RCP8.5 scenario, 15 ski areas (65%) are able to maintain snow reliability with all except 3 and 12 found in Norrland. The 13 largest ski areas modelled are able to maintain seasons with over 100 days throughout, with smaller areas displaying greater impacts from climate change. Because of different terrain capacities, it is important to note that even under the warmest scenario at least 94% of the 4014 ha of pistes included in this study could remain operational with sufficient investment in snowmaking.

4. Discussion

Previously, Moen and Fredman (2007) calculated the loss of ski days for Sälen from their baseline of 162 days to 98 days (−40%) for the period 2070–2100 under the B2 scenario (low emission) and to 66 days (−57%) in the same period for the A2 (high emission) scenario. These two emission scenarios are roughly comparable to RCP4.5 and 8.5 (Pedersen et al., 2020) as used in the current study. Without snowmaking, the closest comparator (site 12) from our study had a 173 day baseline and for the late-century (2071–2100) was reduced to 90 days (−48%) under moderate emissions (RCP4.5) and 11 days (−94%) under a high emission scenario (RCP8.5). With snowmaking 12’s baseline season was 210 days and late-century season lengths were 154 days (−27%) under lower emissions and 124 (−41%) under the high emissions scenario. Although comparability with Moen and Fredman (2007) is limited, as they did not physically model the natural snowpack (instead, they used days with snowfall of any amount as a proxy for operational ski days with snow depth of 30 cm or more), nor did they account for snowmaking. See discussion by Steiger et al. (2019) on the shortcomings of such a methodology. However, our results show that considering snowmaking substantially alters the impact of climate change on season length. This and the fact that Moen and Fredman (2007) extrapolated their case study results to estimate nation-wide economic losses leads to the conclusion that financial resilience of the Swedish ski industry has so far been underestimated.

A more recent study requested in 2019 by the County Administrative Boards of seven northern Swedish counties viewed snow conditions under the RCP4.5 and RCP8.5 future climate scenarios (SMHI, 2021). The work calculated the number of days with >50 cm snow for 16 locations across northern Sweden, including location 8 in our work (SMHI, 2021). For location 8, the number of days >50 cm reduced sharply in all scenarios from 170 days in the baseline period (1963–1992) to around 145 and 140 days for the RCP4.5 and RCP8.5 scenarios for the mid-century (2021–2050) and to around 135 and 110 days for the late-century (2069–2098) (SMHI, 2021).

Although the SHMI (2021) work uses a different baseline period, timeframes and metrics, a comparison is nonetheless useful. In our modelling, the baseline for location 8 had 211 days with over 30 cm of snow which dropped to 193, 180 and 173 days for 2030, 2050 and 2080 under the RCP4.5 scenario. For the RCP8.5 scenario, there were 191, 172 and 118 days for the 2030s, 2050s and 2080s, respectively. Therefore, overall our work supports the notion in the SMHI (2021) report that the number of days with significant depths of snow will decrease under both the RCP4.5 and 8.5 scenarios, and that these decreases will be more pronounced both during the late-century and for the RCP8.5 scenario.

It is also important to compare our results with work using the same methodology conducted in competing markets both in the European Alps (Steiger & Abegg, 2018), and perhaps more pertinently due to their geographical closeness and likelihood of immediate substitution, Norway (Scott et al., 2020). When the loss of ski season length, including adaptive snowmaking, is compared to Scott et al.’s (2020) work in Norway, reductions in ski season length in Norrland for the RCP4.5 scenario fit within the range of the Norwegian regions modelled. For the same scenario in both Svealand and Gotaland, the reduction in season length is considerably higher in all scenarios and timeframes. For the RCP8.5 scenario in the 2030s, reductions in ski season length for Norrland (8%) again fit within the range projected for the Norwegian regions (6–15%). A similar pattern is also found in the 2050s when Norrland’s (13%) reduction in season length is at the lower end of the reductions in Norway (12–27%). By the 2080s, there is greater variation in the reduction from Norwegian regions between 18% in southern Norway up to 60% in northern Norway. Norrland’s reduction in season length (27%) is like that of eastern Norway (27%) and Trøndelag (26%). In the RCP8.5 scenario, both Svealand and Gotaland have greater reductions in ski season length than all the Norwegian areas, apart from the 2080s.

At this point, it must be acknowledged that SkiSim has been applied for existing ski areas only. Both Norrland and northern Norway have vast areas of high altitude, snow reliable terrain without ski areas, which in accordance with future demand levels may be developed for alpine ski tourism.

In contrast to results from northern Norway where relative changes in season length are larger than those in southern, eastern and western Norway (Scott et al., 2020), changes in Norrland are significantly lower than southern and central Sweden. In Norway, this differential can be explained by the lower altitudes of ski areas in northern Norway, northern areas being closer to the coast, and projected temperature changes being greater at higher latitudes (Scott et al., 2020). Conversely, in Norrland, the areas are located away from the coast and are typically at higher altitudes.

Isolating Norrland, it can be shown that climate change risk from mid-century under the RCP8.5 scenario is like many ski areas in the European Alps (Steiger & Abegg, 2018). In the European Alps, 96% of ski areas are projected to be snow reliable under a 1°C rise (∼2030s) and 85% are predicted to be snow reliable under 2°C rise (∼2050s). This compares favourably with areas in Norrland where 100% of resorts are projected to be snow reliable under the +1°C and +2°C future. In the longer term (+4°C ∼2080s), the climate change risk is higher in the European Alps where only 42% of ski areas remain snow reliable compared to 92% in Norrland. Therefore, Northern Sweden’s potential to be a future ‘last resort’ for the European ski industry may not emerge as a competitive advantage until after the 2050s. Norrland will, however, still face competition from Norwegian resorts with similar climatic advantages.

Although northerly Swedish ski areas may remain snow reliable for longer than many Central European and Southern Swedish competitors, the travel behaviour of ski tourists is uncertain. Previous research has indicated tourists may become less loyal to resorts and seek more snow reliable resorts (Behringer et al., 2000; Rutty et al., 2015a; Steiger et al., 2020; Unbehaun et al., 2008), or delay decision-making and booking until snow conditions are acceptable, meaning they may only travel short distances for shorter stays (Luthe, 2009; Pickering et al., 2010). Such inconsistencies in the results of demand-side studies are partially the result of different methods and different questions being asked of tourists. This suggests that more research into customers’ perceptions and reactions to climate change and snow conditions is needed to provide a more comprehensive understanding (Steiger et al., 2019). It is also unclear whether tourists would continue to ski as frequently if resorts are only kept operational through snowmaking. A demand simulation based on a choice experiment with day trippers in Austria shows that a lack of natural snow could cause demand losses of up to 19% despite good skiing conditions on ski slopes with man-made snow (Steiger et al., 2021). Such sensitivities need to be investigated more deeply and in more markets, perhaps incorporating novel methods such as the use of virtual reality to show the aesthetics of potential future scenarios.

Attracting a more international market to Sweden’s ski resorts goes further than the projected advantages of snow reliability from Northern Sweden beyond the 2050s and relies on cultural and socioeconomic factors (Demiroglu et al., 2020). Typically, Swedish resorts are smaller compared to the most popular destinations in the European Alps. France, Italy, Austria and Switzerland have more average lifts per resort, with France, Austria and Italy having more than double the number of resorts with four or more lifts (Vanat, 2020). Additionally, temperatures in Norrland are cooler than the European Alps, and although work in Arctic Finland has predicted a future decrease in extremely cold days (≤ −20°C) (Tervo-Kankare et al., 2018), these conditions will nonetheless be different to the European Alps. For areas like 15, the most northerly resort in this work, there is 24-hour darkness from mid-December to early January, and very little daylight until February. Although the season extends longer into spring than most of Europe, it nonetheless poses the question of whether tourists would want to substitute their holiday for a colder and darker one in northern Sweden. In this context, the success of skiing in Arctic Finland should be considered. Levi, Ruka and Ylläs, the three largest Finnish ski areas, are all found in Finnish Lapland (Demiroglu et al., 2020) and account for 40% of Finland’s skier days (Suomen Hiihtokeskusyhdistys ry, 2018). Therefore, northern Swedish ski areas may consult the template provided in Arctic Finland in order to increase their attractiveness for skiers looking to substitute to more snow reliable resorts.

For non-domestic tourists, visiting Sweden compared to France, Germany or Austria is more expensive. Highlighting this is the cost of food and alcoholic drinks. Consumer price levels for food in Sweden are 121 compared to 126 in Austria, 111 in Italy and 116 in France (Eurostat, 2020). Additionally, for alcoholic beverages, the price index in Sweden is 166 compared to 107 in Austria, 104 in Italy and 101 in France (Eurostat, 2020). This is only one example, but it is important in highlighting the price difference of food and beverages between ski holidays in Sweden and Central Europe. Cultural aspects of trips such as Aprés Ski and gastronomy are often key determinants for tourists, and therefore, should this be more expensive or non-existent in Sweden, then tourists may not substitute their trips. In a Nordic context, however, the price index for food and alcoholic drinks in Norway is 151 and 251, respectively (Eurostat, 2020). Such a price disparity may be advantageous to Swedish areas over Norwegian ski areas with similar climatic advantages. It is also not clear if ski tourists from the European Alps would be willing to visit northern Sweden in a net-zero emission economy if low/zero carbon transportation options are not available or very costly. With these differences, it cannot be assumed that tourists no longer able to visit their preferred destinations in the European Alps will replace ski holidays with trips to Northern Sweden because of a deterioration of snow conditions.

The increases in snow production demonstrated will have consequential impacts on water demand, potentially resulting in water usage conflicts with other stakeholders. This is a sensitive topic for ski areas as conflicts have previously occurred around indigenous Sami livelihood rights during the expansion planning for the ski areas of Sälen and Idre Fjäll into the Städjan National Park (Vail & Heldt, 2000). These considerations are critical given that potential ski areas of Arctic Sweden with the best snow reliability, such as the Kebnekaise Massif and the Sarek and Padjelanta national parks, remain underdeveloped (Demiroglu et al., 2020). Potential future conflict could occur as these areas lie in the Laponia UNESCO World Heritage Site and are preserved for their ecological uniqueness and to protect the traditional livelihoods. Should a sustained trend towards development in these more snow reliable areas occur, care must be taken to have a sustainable and inclusive discussion with indigenous populations and local stakeholders such as that witnessed in Ylläs in Finnish lapland (Kulusjärvi, 2017). Additionally, the aforementioned areas have existing, successful nature-based tourism, and as such land use changes in order to develop or expand skiing capabilities may result in conflict with existing industries (Demiroglu et al., 2019).

Given that the future ski industry will be increasingly reliant on snowmaking, it must be considered whether the consequent increases in energy consumption and potential GHG emissions can be considered maladaptive (Aall et al., 2016). To assess this, the issue must be considered at a tourism system level rather than a ski operation/destination scale. Many of the ski areas at the most significant risk from climate change are located in central and southern Sweden. The travel distances to reach snow reliable areas from Sweden’s three biggest cities Stockholm (1,500,000), Gothenberg (600,000) and Malmö (300,000) (World Population Review, 2020) was found to increase under climate change, most notably under the RCP8.5 climate change scenario. Such increases are notable as currently, 85% of visitors to Swedish ski areas are domestic (Vanat, 2020). In Stockholm, the inner-city ski areas of 4 and 6, although small, are vital for beginners and day trips. Even with advanced snowmaking, these areas have no snow reliable seasons for any timeframe under either RCP4.5 or 8.5. Furthermore, of the other nearest areas in Svealand and Gotaland only 3, 10, 12, 21 and 22 would be considered snow reliable in the 2080s under the RCP2.6 scenario. By the 2050s, under the RCP4.5, only 3 and 12 are snow reliable and remain 100% snow reliable for all modelled scenarios.

Should these ski areas become unviable, skiers from these areas would have to increase their travel distances by car, train or plane to continue to ski. Previous research has shown that willingness of tourists to travel longer distances is limited (Rutty et al., 2015b; Steiger et al., 2021; Unbehaun et al., 2008). Therefore, it cannot be ruled out that some Swedish tourists may give up skiing altogether under increased travel distances. The closure of the smaller, more accessible areas is also a threat to the Swedish ski industry demographically. These areas often act as ‘feeder’ resorts where beginners, younger people and urban populations get their first taste of skiing (Steiger & Abegg, 2013). The closure of these ski areas brings a possibility that the supply chain of generating new skiers for the larger, more snow reliable resorts, both in Sweden and further afield, may diminish. As such, the potential climatic advantage of the more northerly Swedish ski areas may be limited due to their reliance on the smaller ski areas to keep the flow of domestic tourists. Additionally, the more snow reliable ski areas may also be hampered by the ‘backyard effect’ (Hamilton et al., 2007), where a reduction in snowfall and snow cover in urban areas produces reduced interest and motivation to go skiing (Demiroglu et al., 2019).

Setting aside the question of whether tourists would accept longer travel times to go skiing, the increase in travel distances should be viewed as potentially detrimental to climate change goals as they currently rely primarily on fossil fuel-based transport. Importantly, this is changing rapidly, with passenger plug-in electric cars reaching a market share of 32.2% of newly registered cars in Sweden for 2020 (Bil Sweden, 2020). As the market share of electric vehicles continues to grow and charging infrastructure in resorts expand, the projected losses of smaller, more accessible resorts may not have a meaningful impact on GHG emissions related to travel. Indeed, if snowmaking keeps regional ski areas open and prevents the need to fly to more distant destinations, snowmaking from a low carbon electricity grid may reduce net winter tourism emissions in Sweden.

5. Conclusion

This paper has investigated the future of the ski industry in Sweden under potential climate change scenarios and whether Sweden may represent a ‘last resort’ for the European ski industry. Results show a notable shortening of the ski season in Svealand and Gotaland, as early as the 2030s, where snowmaking becomes increasingly challenging. As Norrland retains its snow reliability under all scenarios into the 2080s to a greater extent than areas in Norway and the European Alps previously modelled using SkiSim, the meteorological conditions to become a ‘last resort’ clearly exist. What remains unclear is whether cultural and economic barriers such as limited daylight would be attractive to tourists looking to spatially substitute their ski holiday.

The need to increase the Swedish ski industry’s snowmaking infrastructure will continue in the near future. A situation might arise where the more southerly areas, mainly in Gotaland, will need to consider whether increased reliance on this kind of climate adaptation is sustainable. These increases and associated economic and environmental costs will challenge smaller ski areas’ economic viability in a warmer world. This is not to mention that the projected increases in snow production may reduce the ski area’s aesthetic value and, therefore, their attractiveness for visitors. Such challenges could be explored in future research through novel visual methods like virtual reality.

From a modelling perspective, some limitations exist in this study. A temperature-dependent snowmaking capacity, which increases efficiency with decreasing temperatures, is not considered in our analysis. Inclusion of this factor would facilitate a more precise calculation of the snowmaking potential of areas and an estimate of the power consumption and GHG emissions required for the increase in snowmaking. As snowmaking is around five times more efficient at its optimal temperature of ∼10°C compared to producing snow near the freezing threshold (Olefs et al., 2010), this is a critical aspect for future model development, as well as the assessment of GHG and water usage resulting from snowmaking.

A second limitation pertinent for this work is the core requirement of weather station data, which is frequently not available near Sweden’s ski areas. Furthermore, in several stations, not all parameters required have been recorded, the required time series, or the data collection contains prolonged blank periods. A related limitation is that snowmaking potential is calculated based on ambient air temperature only, neglecting the importance of air humidity (Hartl et al., 2018). However, including air humidity would further increase data requirements for suitable weather stations.

The third limitation is the assumption that all ski areas have 100% of pistes covered with the most advanced snowmaking systems. Although this work has updated the model to allow each area’s operational strategies to be considered, this nonetheless represents a notable limitation. In reality, 92% of the areas modelled have snowmaking capacities, and within this, the ski areas have an average terrain coverage of 56%. In the future, should more information on each ski area’s actual snow production capacities become available, the model parameterization could be improved to reflect better the current and adaptive capacities of the resorts studied.

Demands from a diverse range of stakeholders for more detailed information on climate change risk for the ski industry is likely to increase in the coming decades. Indeed, in June 2021, the G7 countries committed to support mandatory climate change risk disclosure for publicly traded companies. Decision-makers such as investors and lenders, ski area managers, real estate developers, insurance companies and snowmaking and lift equipment manufacturers are increasingly asking for data on the consequences of climate change at both a regional and individual ski area scale. Therefore, in response to Demiroglu et al.’s (2020) call, this work has provided the first climate change impact assessment inclusive of snowmaking for the Swedish ski industry. This is a vital step to more accurately reporting climate change risk.

The importance of this work is further highlighted by it being the first to empirically examine the notion of Sweden being a ‘last resort’ for the European ski industry under climate change. Although it seems northern Sweden may hold a climatic advantage over much of the European Alps, it would continue to face competition from Norway, and future research should focus on assessing the substitution likelihood and perceptions potential customers have of the Swedish ski industry. Such research will aid the evaluation of ski tourism for regional development in areas of Sweden with the greatest snow reliability under future climate change.

Disclosure statement

No potential conflict of interest was reported by the author(s).