The goal of this project is to develop a new binary snow stability classification scheme that incorporates terrain-based metrics alongside traditional stability tests. The new scheme is evaluated for its effectiveness in reducing false-stable assessments compared to previous stability-test-only schemes.
This repository contains code and data processing workflows for:
- Integrating field snow pit data with the gridded snow depth dataset.
- Converting Swiss geographic coordinates from LV03 to LV95 and interpolating snow depth values at pit locations.
- Calculating local snow depth variance (and CV) using a 3×3 grid window around each pit.
- Merging meteorological data (daily mean wind speed) with pit observations.
- Linking stability classifications from the original schemes and avalanche activity records for analysis.
- Preparing feature sets for testing the proposed binary stability classification scheme.
-
Load and Normalize Field Data Extract observation dates, convert coordinates from LV03 to LV95, and prepare a working DataFrame.
-
Snow Depth Extraction Retrieve daily snow depth values from SPASS at pit coordinates, using bilinear interpolation to account for coarse grid spacing.
-
Variance Calculation Compute the variance of snow depth (and CV) in a 3×3 grid window around each pit, also via linear interpolation, to measure local snowpack variability.
-
Wind Data Integration Match each pit observation date to its corresponding daily mean wind speed from WFJ station data.
-
Stability and Avalanche Labels Attach traditional stability classifications and avalanche activity data to each pit observation.
SPASS data is relatively coarsely gridded, so direct nearest-neighbor sampling can miss sub-grid variability. Linear interpolation provides a smoother and more accurate estimate by blending values from surrounding cells in proportion to distance. This is applied both to depth and to variance calculations.
-
Stability test dataset: 589 snowpit samples from the eastern Swiss Alps, Davos region, from winters of 2002-2019.
https://opendata.swiss/en/dataset/field-observations-of-snow-instabilities
-
Snow depth dataset (SPASS): long-term daily 1 km gridded data of snow depth for Switzerland spanning 1962-2023.
https://envidat.ch/#/metadata/spass---new-gridded-snow-datasets-for-switzerland
-
Wind dataset: daily average wind speeds taken from MeteoSwiss (Federal Office for Meteorology and Climatology), specifically the WFJ station.
https://www.meteoswiss.admin.ch/services-and-publications/service/open-data.html
- numpy: Efficient numeric operations.
- pandas: Data cleaning, time series handling, and merging multiple datasets.
- xarray: Accessing and interpolating snow depth data from SPASS.
- matplotlib: Creating visualizations of snow depth, as well as bar charts for final results.