Many of us enjoy going backcountry skiing and lots of us plan and even share our trips online. But did you know that the digital traces that you leave online could help make backcountry touring safer?

Backcountry skiing has become a lot more popular, especially in Switzerland, where avalanches claim about 23 lives each year. But while we know the number of accidents, we know surprisingly little about accident rates: the proportion of people in the backcountry who have accidents.

In cities, we can easily track pedestrian or cyclist activity using sensors and surveys. But in the mountains, traditional methods – like on-site surveys – are costly and impractical for large areas. That’s where digital data comes in. Every time you plan a tour online or upload a GPS track, you leave behind valuable clues about backcountry activity.

We used click data from Skitourenguru, a website for ski tour planning, as well as GPS tracks uploaded to various online platforms to model ski touring activity on a daily basis for 126 regions in the Swiss Alps.

Combining the modelled activity with accident data allows us to estimate accident rates and how they change under different avalanche conditions. It also provides important context for avalanche forecasters, who often face a key challenge: when no avalanches are reported, is it because none occurred – or simply because no one was in the mountains to observe and report them? By understanding when and where people ski, forecasters can better interpret missing avalanche observations, verify avalanche conditions, and ultimately improve the accuracy of avalanche forecasts.

Online and real-world behaviour are similar – yet different!

We found that the click activity on a specific route on the Skitourenguru website strongly predicts real-world activity, as estimated through GPS data, with a one-day delay. This means today’s online planning can help estimate tomorrow’s on-snow activity.

In addition, we linked both GPS and click data to environmental, temporal and avalanche-related variables to model fluctuations in activity. And we examined how different levels of avalanche danger influence behaviour.

Online planning activity increases when the avalanche danger level rises from ‘low’ to ‘moderate’, however, the picture changes when the avalanche danger is elevated from ‘moderate’ to ‘high’:  clicks – which are a proxy for activity – shift from the central Alps to the pre-Alps, favoring lower, gentler terrain (Figure 1).  Ultimately however, weather, season, and accessibility also shape decisions – sometimes more than avalanche risk alone.

Planning versus reality: the digital divide

While click data and GPS activity are driven by similar factors, they reveal different patterns of behavior. People tend to click on tours in areas with a higher avalanche risk and under worse weather conditions than they would actually pursue in real life. This gap shows that planning does not always match reality. However, click data remain a valuable and cost-effective way to estimate backcountry skiing activity across large areas such as the Alps.

If you’d like to know more, check out the full paper.
Schäfer, L., Techel, F., Schmudlach, G., and Purves, R. S.: Tracking the slopes: a spatio-temporal prediction model for backcountry skiing activity in the Swiss Alps using user-generated content Nat. Hazards Earth Syst. Sci., 26, 1537–1559, 2026.

If you do backcountry skiing yourself and use Gipfelbuch.ch for the plannning, we would be happy if you’d take part in a short survey that is currently being conducted on the topic. You can take part in the survey using the link below – and as a thank you for participating, you’ll also have the chance to win a voucher from Bächli Bergsport. Skitouren auf Gipfelbuch.ch: Warum liest oder schreibst du Einträge?

Leonie Schäfer, Geocomputation