Evaluating the predictive performance of human avalanche forecasts and model predictions in Switzerland

This data set was used in the analysis by Techel et al. **Can model-based avalanche forecasts match the discriminatory skill of human danger level forecasts? A comparison from Switzerland**. Please note that the title of the preprint was *Forecasting avalanche danger: human-made forecasts vs. fully automated model-driven predictions*, submitted to *Natural Hazards Earth System Sciences* on 20 Aug 2024.
The final manuscript used data from three forecasting seasons (2022/2023, 2023/2024, 2024/2025) , the preprint two forecasting seasons (2022/2023, 2023/2024) .
Currently, the repository contains data from two avalanche forecasting seasons in Switzerland. The third season will be added (to be done, 30 June 2025).
**Interpolated predictions**
- The .zip file contains the interpolated predictions for the three models in nowcast- and forecast- mode. This data is needed to reproduce the figures and tables in the submitted preprint.
The other data are the **raw data** underlying the interpolations:
- Avalanche forecast by WSL Institute for Snow and Avalanche Research SLF, published at 17.00 local time, valid for the following 24 hours and relating to dry snow avalanche conditions.
- Model predictions in *nowcast*- and *forecast*-mode for three models (*danger level*, *instability*, *natural avalanche*), valid for 12.00 local time
- Subset of points extracted from GPS tracks (courtesy of Skitourenguru GmbH)
- Avalanche observations - natural avalanches and human-triggered avalanches
- Estimates of the snowline
- Randomly chosen subset of grid points used for generating reference distributions
For details regarding the data sets refer to the publication.