ews
The automatically calibrated forecasts provide reliable warnings for unusual events that exceed a threshold, which may result in impacts on livelihoods or health. These warnings account for the uncertainty in the calibrated forecast and are classified into three categories:
- Red (High Risk): 70% or more of the forecast probability indicates an exceedance of the threshold.
- Yellow (Moderate Risk): Between 40% and 70% of the forecast probability exceeds the threshold.
- Green (Low Risk): The probability of exceeding the threshold is below 40%.
These warning levels are displayed on the map for each district (based on the spatial forecast) and for each station selected within a predefined warning window.
The thresholds for warnings can be configured within the Shiny application of the warning platform, allowing users to define three thresholds each for low and high levels. These thresholds can be set separately for each district, enabling customization to account for varying impact dynamics. Additionally, each district includes an indication of the topographic elevation limit up to which the spatial forecasts are considered. By default, this limit is set to 3,000 meters, under the assumption that no assets above this elevation are significantly impacted. This consideration is important, as higher elevations naturally reduce the average temperature within a district, thereby influencing the forecast.
The accuracy of the warnings is typically assessed using two key metrics: the False Alarm Ratio (FAR) and the Probability of Detection (POD). The FAR measures the proportion of cases where a warning was issued for a threshold but no event occurred. Conversely, the POD evaluates the proportion of cases where an event was predicted and actually occurred.
The verification results for both metrics are shown in the provided figure, focusing on cold events observed between February and May 2022. The dashed lines represent the requirements set by MeteoSwiss: an FAR below 0.3 and a POD of at least 0.9. As illustrated in the figure, the warning service meets these thresholds for at least the first five forecast days.
The warning information can be used by national or regional administrations to inform through their channels their custodians. It is advised that warnings are validated by a human and disseminated by an authoritative voice in order to ensure liability. The service also includes a interface with the Telerivet platform to send automatically SMS to larger number of users. This has been tested in 2024 by sending the warning of an exceptional cold event to around 2'000 farmers to protect their orachards or livestock (e.g. beecomps).
