1 Introduction

In recent decades, a significant increase in temperature has been detected in most mountain regions around the world, accompanied by a shift towards earlier snowpack melt and declining snow accumulation (Mote 2003; Barnett et al. 2005; Hantel and Hirtl-Wielke 2007; De Jong et al. 2009; Minder 2010). This change in snowpack dynamics results from snow sensitivity to temperature increase, causing a decreasing proportion of snowfall relative to precipitation, and an increase in available energy for snowmelt (Rood et al. 2008). Thus, a +1 °C change was reported to reduce accumulated snow by 20 % and shorten the snow season in the Pyrenees (López-Moreno et al. 2013a, b). Despite high uncertainties and a large regional variability, climate models project that the temperature will continue to increase in the coming decades (Ganguly et al. 2009). It is expected that mountain areas will be particularly affected by high rates of warming (Nogués-Bravo et al. 2007; Minder 2010; Gobiet et al. 2014), with the consequent impact on the accumulation and duration of mountain snowpacks (Adam et al. 2009; Hamlet 2011; López-Moreno et al. 2013a, b). Many research efforts have been directed at assessing what the environmental and socioeconomic effects of a thinner snowpack with a shorter duration might be, including those effects on water resource availability (Barnett et al. 2005; Adam et al. 2009), the ecology of affected areas (Tague and Dugger 2010; Trujillo et al. 2012), hydropower production (Finger et al. 2012), and ski resort viability (Abegg et al. 2007; Scott et al. 2012; Gilaberte et al. 2015). Although climate and weather are only two of the many factors affecting winter tourism, changes on the seasonal nature and availability of snow may have a significant socioeconomic impact that results in a decrease in the viability and sustainability of winter activities, such as the ski industry (Saurí and Llurdés 2010; Scott et al. 2012).

In the late 1980s and 1990s, following a succession of winters with poor snow conditions, the first academic studies emerged that dealt with the impact of climate change on the ski industry. Since then, the principal winter tourism regions of the world have been analyzed, including the European Alps (Breiling and Charamza 1999; Elsasser and Bürki 2002; König and Abegg 1997; Steiger 2010, 2012; Steiger and Abegg 2013; Steiger and Mayer 2008; Uhlmann et al. 2009; Endler and Matzarakis 2011), Canada (Scott et al. 2003, 2006, 2007), USA (Dawson and Scott 2007, 2010; Dawson et al. 2009; Nolin and Daly 2006; Scott et al. 2008), Sweden (Moen and Fredman 2007), Australia (Hennessy et al. 2003, 2008; Bicknell and McManus 2006), Japan (Fukushima et al. 2003), or New Zealand (Hendrikx et al. 2013). Some of these studies primarily focus on the supply-side impacts (ski operations), model simply natural snow conditions at ski resorts (Uhlmann et al. 2009), or apply indicators that are not relevant to ski-area operations, such as “snow cover,” defined as 2.5 cm of snow (Lamothe and Périard 1988), when in fact ski operators require 30–100 cm of snow to open a ski run. With the exception of a few studies that use statistical relationships between snow depth and other climatological parameters (Moen and Fredmand 2007), most of these studies base their estimations on physical snow models. One of the major limitations to these statistical model-based studies is they omit the effect of snowmaking on the future natural snowpack, as well as the effect of factors such as rain or warm winter days that can affect the quality of snow. This limitation, found not only in statistical models but also in many other studies using physical snow models, is the main drawback in most of the previous literature assessing the vulnerability of ski resorts (Scott et al. 2012; Gilaberte et al. 2015). This is a key point as nowadays snowmaking is responsible for covering huge areas of the ski resorts and the percentage of snow-machine-covered runs increases every year (Steiger and Mayer 2008). Snowmaking is currently the main adaptation strategy used to offset the natural variability of snow. It helps to guarantee enough snow depth, scheduled openings, and stable revenues, but also represents a commercial and image strategy to extend the season in order to increase revenues (Steiger and Mayer 2008). Over the last few decades, ski resorts across the world have invested considerably in snow-production systems and the Pyrenean ski resorts are no exception to this global trend (Saurí and Llurdés 2010). The studies discussing this issue (Scott et al. 2003, 2007, 2008; Scott 2011; Hennessy et al. 2008; Steiger 2010, 2011) found that the impacts of climate change on the various regions analyzed were lower than the impacts reported in previous studies that considered only natural snow. Finally, an alternative approach to the statistical and physical models in analyzing the climate change impacts on the ski industry is the analog approach. Temporal analogs use past and present experiences and responses to climatic variability, change, and extremes to provide insight into vulnerability to possible future climate change (Ford et al. 2010). So far, this approach has only been applied in a few studies in North America (Scott et al. 2006; Dawson et al. 2009) and one in the Austrian region of Tyrol (Steiger 2011).

Despite the heterogeneity of approaches, most of these studies are congruent in their results, indicating that climate change will lead to a reduction in ski-season length, a loss of skiable areas, and a drop in visitors, mainly in low-altitude and low-latitude ski resorts.

The Pyrenees is a 450-km-long mountain range comprising the northern part of Spain, the southern part of France, and Andorra (Fig. 1). Altitude ranges from 300 m to over 3000 m. The climate of the Pyrenees is subject to an eastward transition from Atlantic to Mediterranean conditions. Moreover, its macro relief introduces a significant variability in the distribution of precipitation and temperature. The Pyrenees is the most important winter tourism region in Europe after the Alps (Vanat 2014). Alpine skiing is the most important winter sport in this region in terms of visitors and revenues. Encompassing 49 alpine ski resorts, the region receives approximately 11 million skiers a year (average for the seasons 2009–2010 to 2012–2013 in Pons et al. 2014) (Fig. 1). The elevation of the Pyrenean ski resorts ranges from 1350 to 2700 m with a mean elevation of approximately 1950 m.

Fig. 1
figure 1

Area of study and location of the ski resorts. Size of the bullets represents the average skier attendance at each ski resort from 2009–2010 to 2012–2013

Full size image

The aim of this paper is to analyze the vulnerability of the Pyrenean ski resorts to projected changes in the snowpack under various future climate scenarios. By means of a physical snow model, the changes in the natural snowpack are projected for each ski resort in two different future scenarios. Subsequently, the changes in the ski-season length of the various Pyrenean ski resorts are projected. Finally, the capacity to produce snow under these two future scenarios is assessed to analyze the suitability of snowmaking to offset climate variability of natural snow cover.

2 Data and methods

The study uses regional climate change projections in order to simulate the future natural snowpack in the various ski resorts of the Pyrenees. A snowmaking module simulating the effect of snow-production systems has been included in order to analyze the effect of these systems to enhance natural snow depth. A conceptual map with the main components of the model is shown in Fig. 2. Data from the smallest ski resorts of the Pyrenees (less than a couple of runs, often presented as a complementary activity to a hotel, or a Nordic ski resort) was not available. Only 41 of the total 49 ski resorts were hence taken in this study, representing more than 92 % of the total skiers in the region (DSF 2012; Botti et al. 2013; ATUDEM 2013; SkiAndorra 2013).

Fig. 2
figure 2

Conceptual map of the model linking climate change projections of future changes in the snowpack with the vulnerability of the Pyrenean ski resorts

Full size image

In order to estimate future natural snow conditions at each ski resort, the model uses the regional projections of the daily snowpack in the Pyrenees during the twenty-first century of López-Moreno et al. (2009). In the study conducted by López-Moreno et al. (2009), the seasonal evolution of snowpack was simulated at 20 individual points across the entire Pyrenees, each of them representing the average conditions over a surface area typical of a regional climate model (RCM) employed during the EU PRUDENCE project, with a resolution of 50 km2 (Christensen et al. 2002). The snow depth and the snow duration were simulated by running GRound ENergy Balance for NaturaL Surfaces (GRENBLS), a surface energy balance model (Keller and Goyette 2005), with climatic inputs provided by the HIRHAM regional climate model (Christensen et al. 1998). Surface energy balance models (SEBM) for snow simulate the evolution of the snowpack based on the thermal fluxes affecting the snow. GRENBLS is a single-layer physically based model driven by hourly input data of air temperature, dew-point temperature, anemometer-level wind velocity, precipitation, surface pressure, and incident solar radiation. As GRENBLS employs hourly data and RCMs usually provide outputs on a daily basis, these variables were disaggregated into hourly series following the procedures of Jansson and Karlberg (2004). The model computed the radiative fluxes from cloudiness data and the surface turbulent latent and sensible heat fluxes. The bulk heat and moisture transfer coefficients were parameterized according to Benoît (1977) based on the Monin–Obukov similarity theory. Surface temperature, soil moisture, and snow mass are prognostic variables. The energy budget also considered the energy change associated with the melting of frozen soil moisture and snow. The temperature of the snowpack was computed via heat storage using a force-restore method (McFarlane et al. 1992). Precipitation is considered as solid if air temperature is less than that of the triple point of water. Liquid precipitation on a snowpack induces snowmelt, and the melt water enters the soil directly in liquid form. Snow was modeled as an evolving one-layer pack characterized by temperature T snow (K), mass M snow (kg/m2), and density q snow (kg/m3). The surface energy budget was computed over the snow cover at each model time step. The radiative and turbulent fluxes were computed first, followed by heat storage in the snowpack; if the latter was positive and the snow temperature was below the melting point, the excess energy was first used to raise the temperature of the pack. Once the temperature reaches the melting point, any additional excess energy is used to melt the snow. The age effect of the snow on snow density was adopted following Verseghy (1991). The snow density of the bulk snow layer increases exponentially over time from the fresh-fallen snow value, q snow,min = 100 kg/m3, to a maximum of q snow,max = 300 kg/m3. In a similar manner, changes in snow albedo that accompany snow aging were parameterized as a time-decay function from an initial fresh snow albedo of 0.80. GRENBLS also incorporates total cloudiness as an input parameter. The model was run for a control period (1961–1990) and for two future emissions scenarios, the IPCC Special Report on Emissions Scenarios (SRES) A2 and B2 (IPCC 2007), and for various altitudinal levels, 1500, 2000, 2500, and 3000 m. Outputs were snow water equivalent (SWE) and snow depth series at hourly intervals at 20 points of the Pyrenees and at 4 different altitudinal levels (1500, 2000, 2500, and 3000 m). Between these altitudinal levels, the snow depth was interpolated in order to simulate the snowpack every 150 m. The model was calibrated with a control period, and its performance was validated, finding good agreement in the overall snow depth and duration between the observed and simulated snowpack in the Pyrenees (see detail of validation in López-Moreno et al. 2009). Only Formigal (central-western part of the Pyrenees) presented significantly different behavior to the simulated data. In this case, the snowpack was modeled using the Cold Regions Hydrological Model (CRHM) (Pomeroy et al. 2007) with historical data from the nearest meteorological station to the ski resort, Izas (López-Moreno et al. 2013a, b). The CRHM platform uses a modular modeling object-oriented structure to simulate a comprehensive range of hydrological phenomena in mountainous and cold regions (including blowing snow, intercepted snow, energy balance snowmelt, infiltration to frozen soils, etc.).

The 30-cm threshold is commonly used as the average minimum snow depth to operate a ski resort (Witmer 1986; Abegg et al. 2007; Scott et al. 2008; Steiger 2010). Thus, for each ski resort, season lengths have been analyzed by taking into account the number of days during an average winter season in which this lower boundary is reached. The ski-season length has been calculated by applying this 30-cm threshold to the natural snowpack projections obtained from the snow model at the mean elevation of each ski resort. This elevation is used as an indicator of the average snowpack available at each ski resort (Abegg et al. 2007; Scott et al. 2003; Scott and McBoyle 2007; Steiger 2010).

The model includes a snowmaking module in order to simulate the effect of snowmaking systems in enhancing the snowpack in order to achieve a more realistic projection of the future ski-season length. Moreover, by including the capacity to produce snow in future climate change scenarios, the suitability and the sustainability of this adaptation strategy can also be assessed. In this model, only snowmaking to ensure the minimum snow conditions has been simulated. Following the experience of technical staff in ski resorts, a daily minimum temperature threshold of −2 °C has been used as a proxy to compute the potential snowmaking days during a typical winter season. In line with previous studies (Scott el al. 2008; Steiger, 2011), it is assumed that during these potential snowmaking days, a maximum of 10 cm per day can be produced to reach the 30-cm threshold. Thus, the natural snow depth is complemented with snowmaking according to these criteria. The resulting season length will depend on the natural snowpack available and the potential snow produced by these systems. Once the season length of each resort has been projected based on the natural snow availability and snowmaking production, the season lengths have been analyzed by taking into account the number of days during an average winter season for each scenario. Using the “100-day” criterion (Abegg et al. 2007; Dawson and Scott 2007, 2010; Scott et al. 2003; Scott and McBoyle 2007; Steiger 2010; Witmer 1986), the ski resorts reaching the 30-cm threshold during at least 100 days per winter season are considered as being “reliable” and any that do not are considered “unreliable.”

3 Results

3.1 Projected changes in the snowpack in Pyrenean ski resorts

The model simulations project that the snowpack in the Pyrenees will be strongly affected by projected climate changes, with a marked decrease in snow depth and duration of the snowpack (Fig. 3). Various greenhouse gas emission scenarios (SRES) lead to significant differences in the severity of expected changes in the snowpack, with these differences being twice as pronounced under scenario A2 compared to B2. Noticeable spatial differences in the magnitude of simulated changes in snowpack have been detected. The snowpack in the central and eastern areas of the Spanish Pyrenees is clearly the most affected by climate change. The impact of climate change on the snowpack is highly sensitive to altitudinal gradient. A decrease in snow depth at 3000 m is just 25 % of that simulated at 1500 m. In the highest sectors under SRES A2 and B2, snow depth is predicted to decrease by 11 and 5 %, respectively. In the lowest sectors, snow depth is predicted to decrease by up to 70 and 32 %, respectively, and the duration of the snowpack by 78 and 44 %.

Fig. 3
figure 3

Simulated changes in the duration of the snowpack according to climate change projected by the HIRHAM model under SRES B2 (blue) and A2 (red) at different altitudinal planes: 1500, 2000, 2500, and 3000 m

Full size image

Three different scenarios have been evaluated in order to analyze the vulnerability of the Pyrenean ski resorts to projected changes in the future snowpack. The first scenario simulates the current average winter season, while the other two simulate future snow depth, assuming a 2 °C increase in the winter mean temperature (equivalent to SRES B2 at the end of the twenty-first century) as a moderate climate change scenario, and a 4 °C increase (equivalent to SRES A2 at the end of 21st century) as a strong climate change scenario (López-Moreno et al. 2008). These two scenarios consider future changes in air temperatures as well as changes in precipitation, wind, dew-point temperature, atmospheric pressure at the surface, and incident solar radiation. In order to assist the communication of results to the region’s various stakeholders (ski resorts, policy makers in regional and local administrations) who are usually not familiar with climate scenarios and who often have a short-term vision, the nomenclature of these two scenarios was simplified to a +2 and +4 °C increase in the winter mean temperature for the Pyrenean region. Winter is taken to be the ski-season period (from December 1st to April 20th). Hence, the comparison of the projected scenario with a similar past situation (i.e., the 2006–2007 season was similar to the +2 °C scenario) had a stronger impact on local stakeholders’ awareness of the situation when communicating results and engaging them in the study.

Figure 4a shows the mean control period (1961–1990) and the future snowpack simulated at average elevation (1900 m) of Ax 3 Domaines (a French ski resort located in the Midi-Pyrénées region with a mean elevation of 1850 m), assuming a +2 °C increase (equivalent to the B2 scenario for the period 2070–2100) and +4 °C increase (equivalent to the A2 scenario for the 2070–2100 period) in winter mean temperature. Figure 4b shows the contribution of snowmaking in enhancing the natural snowpack at 1900 m in the Andorran ski resort of Pal according to the defined parameters for a +2 °C climate change scenario. The figure shows the importance of including this factor in the analysis. Should only natural snow be considered, the resort presented would only have a short period with snow depth over 30 cm, whereas with snowmaking, for most of the winter season, the depth will be above this threshold.

Fig. 4
figure 4

a Mean control period (CP), future natural snowpack (+2 and +4 °C) for Ax 3 Domaines (France) ski resort at 1900 m, and the 30-cm threshold assumed as minimum snow depth to consider a resort reliable. b Snowpack in Pal (Andorra) assuming a 2 °C increase of the winter average temperature enhanced with snowmaking when the 30-cm threshold is not achieved. c Mean control period (CP) for Ax 3 Domaines (France) ski resort at 1900 m, future natural snowpack (+2 and +4 °C), enhancement of the future natural snowpack due to the snowmaking, and the 30-cm threshold assumed as minimum snow depth to consider a resort reliable

Full size image

3.2 Vulnerability of the ski resorts

The vulnerability of each ski resort considering both the natural snow and the capacity to produce snow, hereafter referred to as “technical reliability,” has been analyzed. Figure 5 shows reliable ski resorts with only natural snow as blue dots, those considered reliable thanks to snowmaking systems as yellow dots, and those ski resorts considered as unreliable as red dots, during the two future climate change scenarios, assuming an increase of 2 and 4 °C. The total percentage of ski resorts that are reliable under natural snow conditions and under technical conditions (i.e., with snowmaking) in the Pyrenees for the control period (1961–1990), B2 (+2 °C), and A2 (+4 °C) are shown in Fig. 6. At present, 83 % of the ski resorts are considered naturally reliable in an average winter season. However, taking into account snowmaking capacity, 98 % of the ski resorts in the Pyrenees are considered reliable during a current average winter season. In a future scenario, assuming an increase of 2 °C in the winter mean temperature, this share would be reduced to 44 %. In a strong climate change scenario, that is, a scenario assuming an increase of 4 °C, the total share of reliable ski resorts in the Pyrenees would be dramatically reduced to only 7 %. When analyzing the capacity of snowmaking to offset the variability of natural snow, i.e., technical reliability, it has been observed that these systems can largely enhance and extend the season length in a moderate climate change scenario (i.e., 2 °C increase). In this case, the share of technical reliable ski resorts is increased to 85 %. However, in a more intense warming scenario (4 °C), only a residual effect from the snowmaking systems is observed, where no significant increase in the share of reliable ski resorts is projected.

Fig. 5
figure 5

Reliability of the Pyrenean ski resorts considering only natural snow and natural snow enhanced with snowmaking systems (technical reliability) in two future climate change scenarios: +2 and +4 °C. The size of the points represents the present average attendance in number of skiers during the control period

Full size image
Fig. 6
figure 6

Share of naturally and technically reliable ski resorts in the Pyrenees in the present and under two climate change scenarios: +2 and +4 °C

Full size image

A high variability at short distances has been observed in the level of impacts and therefore in the vulnerability of the different Pyrenean ski resorts. Along this line, it was found that within the same region, two resorts located at a very short distance can have significantly different levels of vulnerability. Factors such as aspect or proximity to the Atlantic Ocean are the main contributors to this high variability.

4 Discussion

The results achieved in this first approach for the Pyrenean region are in line with the majority of studies published to date that analyze climate change impacts on the ski resorts in different areas around the world. To cite some of them, reductions in the season length in northeast USA have been projected as being around 15 and 41 % (Scott et al. 2008), 15 and 50 % in Ontario (Scott et al. 2008), 5 and 35 % in Quebec (Scott et al. 2007), and 14 and 41 % in Tyrol, Austria (Steiger 2010), taking into account moderate- and high-emissions scenarios, respectively. The projections of the snow model simulations show that the snow depth and the duration of snowpack in the Pyrenean ski resorts may be notably affected by future climate change. Different climate change scenarios lead to significant differences in the severity of the expected changes in the snowpack. Even though there is a high geographic variability in the climate change vulnerability, the reduction of the ski-season length is projected in both a moderate (+2 °C) and a strong climate change scenario (+4 °C).

The snowpack in the eastern areas of the Spanish Pyrenees is clearly the most affected by climate change. Resorts in low-elevation areas, closer to the Mediterranean Sea, and/or with a predominance of south-oriented slopes were identified as being the most vulnerable (season length reduced by around 38 % under the +2 °C scenario). In the particular case of the Pyrenees, those ski resorts closer to the Atlantic Ocean, located at higher elevations, and/or with northerly orientation were identified as being the most resilient (season length reduced by 15 % under the +2 °C scenario) (Fig. 7). Higher altitudes guarantee low temperatures, both for the occurrence of snow precipitation as well as for snowmaking production. The Atlantic influence affects the precipitation pattern, with greater effects on the perturbations associated with polar fronts and enhanced by the mountain ranges that oppose the westerly flows (Vicente-Serrano and Cuadrat 2007). These areas are characterized by more frequent and more equally distributed precipitation throughout the season as compared to a more erratic pattern in the Mediterranean-influenced area, which is usually concentrated in the autumn and spring (Vicente-Serrano and Cuadrat 2007).

Fig. 7
figure 7

Reliable ski resorts compared to their elevation and distance to the Atlantic Ocean. Red dots show unreliable ski resorts, blue dots show those reliable with only natural snow, and yellow dots those reliable thanks to snowmaking

Full size image

The snowmaking analysis showed that under the +2 °C scenario, snowmaking can significantly enhance the ski-season length in many of the Pyrenean ski resorts. However, this measure can only act as a robust adaptation strategy in the region provided climate change is limited to +2 °C snowmaking. In the +4 °C scenario, the capacity of snowmaking significantly reduced due to an increase in the minimum temperature, a determining factor in efficiently producing snow. Therefore, the effect of snowmaking systems is residual in this scenario. The results of this study can be used as the basis for better characterization and understanding of the geographical differences in the vulnerability to impacts due to future climate changes in the region. However, the results illustrated the need for better geographical characterization and higher spatial resolution of the spatial variability of the snowpack in the area.

The spatial distribution of snow in mountain areas is characterized by high variability within very short distances. This variability results from complex interaction between mesoscale meteorology, local topography, and weather factors. Aspect, slope or wind-blown effects (Green and Pickering 2009), or forest density (Lundquist et al. 2013) are crucial factors that affect the spatial distribution of snow. For example, due to the complex topography of mountain areas, slope angle and aspect are also very likely to influence the sensitivity of snowpack to temperature change (Uhlmann et al. 2009). Snowpack dynamics are strongly influenced by aspect (Hinckley et al. 2012), affecting snow accumulation and melting, especially in areas with a marginal snowpack (McNamara et al. 2005). Along this line, it was found that as temperature increases, the effect of aspect on accumulation and melting increases, resulting in greater differences in the maximum snow accumulation and snowpack duration (López-Moreno et al. 2013a, b). Therefore, the inclusion of local topography effects, as well as technical operations on snow (i.e., grooming), when analyzing the future snowpack will be key in achieving a better vulnerability assessment of the Pyrenean winter tourism industry. This information could be complemented in future research with snow cover models with higher spatial resolution, which are better able to capture the regional variability of the snow depth for each ski resort.

4.1 Conclusion

Future climate-induced changes in snowpack of the Pyrenean ski resorts have been projected by means of the GRENBLS model, driven with climatic inputs from the HIRHAM regional climate model. The study shows that snow depth and duration of snowpack in the Pyrenean ski resorts may be strongly affected by future climate change. Based on two future scenarios, the vulnerability of the Pyrenean ski resorts to climate change has been assessed. Different climate change scenarios lead to significant differences in the severity of the expected changes in the snowpack. Even though there is high geographic variability in the vulnerability to climate change, a reduction of the ski-season length is projected both in a moderate (+2 °C) and a strong climate change scenario (+4 °C). Moreover, an analysis of the effects of snowmaking showed that this measure can only act as a robust adaptation strategy in the region provided climate change is limited to +2 °C snowmaking. Finally, the study remarks on the need to improve the inclusion of local topography effects, as well as technical snow operations (i.e., grooming), when analyzing the future snowpack at ski resorts. This issue, complemented with snow cover models with accurate spatial resolution that are better able to capture the regional variability, could significantly improve the climate change vulnerability assessment of ski resorts.