Global reduction of snow cover in ski areas under climate change

We hypothesize that (i) annual snow cover days in ski areas will decrease globally and under all SSP scenarios and that (ii) optimal climate for ski areas will be located in less populated areas in the future. On this basis, we will discuss consequences for alpine biodiversity and aris- ing conflicts between ski areas and nature protection. Methods Data acquisition Snow cover days. Global raster data on natural annual snow cover days (scd) at a resolu- tion of 30 arc sec for a historic baseline (1950–2010), the present (2011–2040), immediate future (2041–2070) and near future (2071–2100) based on a ISIMIP3b model were selected from the CHELSA (Climatologies at high resolution for earth’s land surface areas) database [23, 24]. The projections are based on GFDL’s Earth System Model Version 4 (GFDL-ESM4, ISIMIP3b, www.isimip.org) for SSP1-2.6 (low), SSP3-7.0 (high) and SSP5-8.5 (very high) emis- sions scenarios. GFDL-ESM4 is a chemistry-carbon-climate Earth system model focusing on Earth system interactions with a doubled horizontal resolution [25]. The projected snow cover days on a high resolution are based on a temperature model using statistical downscaling and a precipitation model which consists of various orographic predictors [24]. Global ski areas. We retrieved information on the global distribution of ski areas from OpenStreetMap via the overpass API (overpass-turbo.eu) using the tag search term “piste: type = downhill” (downloaded 30.06.2020). The tagged ski areas are the result of openly avail-

able collaboratively generated citizen science data. Data quality was ensured by combining information on the occurrence of ski areas from OpenStreetMap with global projections of annual scd from 2005 to 2018. For 2005 to 2013, raster files on annual scd were downloaded from the CHELSA Database [26] and annual scd from 2014 to 2018 were calculated using Saga GIS Version 2.3.2 (snow cover tool) using monthly temperature and precipitation data at a res- olution of 30 arc sec [26]. All indicated ski areas were categorized according to the number of annual scd from 2005 to 2018 (from 0 = no snow days to 365 = snow all year long). Those fall- ing into pixels of zero scd were manually checked and subsequently eliminated from the data- set in cases where they were incorrectly tagged as ski area or turned out to be indoor skiing centres and dry or artificial ski slopes. The occurrence of ski areas was translated into a global presence-absence raster on the resolution of the annual scd data. From this global distribution of ski areas, we extracted seven major mountain regions, namely the Andes, Appalachian Mountains, Australian Alps, European Alps, Japanese Alps, Rocky Mountains, and the South- ern Alps (New Zealand). The European Alps account for 69% of all pixels with documented downhill skiing, followed by the Rocky Mountains that sum up to 15%. Japanese Alps and the Appalachian Mountains sum up to 7% each, the Andes and Australia to 1% each, and Southern Alps accounts for 0.5% of the data. Population density. To assess how global ski areas are currently distributed relative to

population density, data on global population density from 2020 was retrieved at a resolution of 30 arc sec from the Socioeconomic Data and Applications Centre (SEDAC) as part of NASA’s Earth Observing System Data and Information System (EOSDIS) as a geo-referenced raster file [27]. Analysis For every pixel for which ski areas have been recorded within the seven mountain regions, we extracted the number of annual natural snow cover days for the periods 1981–2010 (histori- cal), 2011–2040 (present), 2041–2070 (immediate future) and 2071–2100 (near future) given a very high (SSP5-8.5), high (SSP3-7.0) and low (SSP1-2.6) emissions scenario. A nonparametric PLOS ONE Global reduction of snow cover in ski areas under climate change PLOS ONE | https://doi.org/10.1371/journal.pone.0299735 March 13, 2024 3 / 15