Forecasting Alpine snow amounts for the safety of people, infrastructure and transport

The project will develop a system to provide an automated probabilistic forecast of (new) snow depth in the central Alps to support decision-making by avalanche warning services and authorities for keeping people safe and traffic flowing as smoothly as possible. Look-ahead times are from 12 hours for tactical planning to one week for strategical planning. Advanced statistical models will be developed from long time series of measurements and forecasts from numerical weather prediction (NWP) models to forecast snow at measurement locations and - in conjunction with Digital Elevation Model (DEM) data - also at arbitrary locations in the forecasting domain.

Weather plays an important role for large parts of our daily life. While mostly non-hazardous, certain weather situations can be crucial for the safety of people, infrastructure, and transport. For a region like Tyrol, located in the Eastern European Alps, fresh snow plays a major role in winter. On the one hand tourism relies on good snow conditions, on the other hand the white gold can cause obstructions or life-threatening situations. Within the scope of this project a new statistical toolbox of post-processing methods and corresponding estimation and optimization algorithms have been developed to predict fresh snow amounts for Tyrol. The new methods allow for probabilistic forecasts with a high spatial and temporal resolution. Based on weather forecasts from physically-based numerical models and observations from past years, the methods learn the structure of forecast errors made in the past. Corrections obtained from this knowledge are then applied to the forecasts for the next few days to get more reliable weather predictions. Due to the complexity of the alpine topography the weather, or climate, can strongly vary between two locations, even if they lie only few kilometres apart. To account for these small-scale differences and to be able to include all available information, the new methods make use of standardized anomalies. Standardized anomalies are deviations of forecasts and observations to the corresponding specific local climate. This allows to account for site-specific characteristics in space and time without losing the local properties of a specific location and season.The new method allows to provide reliable probabilistic forecasts for temperature, precipitation sums, and fresh snow amounts for any location and time in Tyrol.