Subseasonal Prediction and Predictability of European Temperatures
In the last decade, increasing efforts have been made to improve atmospheric forecasts on time scales of weeks to months. The aim of these so-called subseasonal predictions is to push the limits of traditional weather forecasting and bridge the gap to seasonal predictions as forecast users from many...
| Main Author: | |
|---|---|
| Other Authors: | , , |
| Format: | Doctoral or Postdoctoral Thesis |
| Language: | English |
| Published: |
ETH Zurich
2021
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/20.500.11850/512182 https://doi.org/10.3929/ethz-b-000512182 |
| Summary: | In the last decade, increasing efforts have been made to improve atmospheric forecasts on time scales of weeks to months. The aim of these so-called subseasonal predictions is to push the limits of traditional weather forecasting and bridge the gap to seasonal predictions as forecast users from many sectors (including agriculture, health, humanitarian sector, energy trading, re-insurance) could strongly benefit from improved subseasonal forecasts. However, subseasonal prediction skill is generally low due to the chaotic nature of the atmosphere. It is thus crucial to distinguish between situations for which the skill is enhanced and those for which subseasonal forecasts cannot provide useful information. Knowledge about the temporal and spatial variations of predictability is valuable for users of subseasonal predictions and can improve their confidence in the forecasts. Despite a variety of suggested sources of predictability, subseasonal prediction skill is particularly poor over Europe. In this thesis, we analyze the variations of subseasonal European near-surface temperature predictability. We use operational and retrospective forecasts out to two months lead time that are generated with numerical prediction models, the main tool used for making subseasonal predictions. We show that when evaluating the skill of subseasonal forecasts, deviations from stationarity in the climatology, such as trends, have to be carefully accounted for. When neglecting long-term warming trends during the reference period, estimates of temperature skill tend to be inflated compared to the actual skill. Only when the non-stationary components of the climatology are correctly taken into account, it is possible to make inferences about the temporal variability of predictability. We then use these results in order to analyze the month-to-month variations in subseasonal forecast skill. Using 20 years of retrospective forecasts, we show that there is a distinct seasonal cycle in subseasonal forecast skill: At lead times between 10 and 20 days, European land temperatures in winter, especially in February and March, are predicted with smaller errors and higher certainty than in any other season. The situations that are predicted best during these months feature a large-scale atmospheric flow pattern of increased westerly flow over the North Atlantic and concurrent higher temperatures over Europe, central Asia and Siberia. This zonal flow is accompanied by distinct anomalies in the tropics and polar stratosphere during forecast initialization that can serve as predictors of forecast skill. Especially cold anomalies in the lower stratosphere during initialization are useful indicators of enhanced subseasonal forecast skill for European land temperatures in winter. While subseasonal forecast skill in summer is lower than in winter, it is also possible to distinguish between better and poorer forecasts in summer based on the type of forecast event. Particularly in central to eastern Europe, summer warm extremes are overall predicted better than cold extremes and average temperatures indicating an asymmetry in skill for temperatures in different parts of the climatological distribution. This has potential implications for prediction skill in a future warmer climate. The results of this thesis can allow stakeholders to make more informed forecast-based decisions and can thus help to increase confidence in the products provided by forecasting centers. |
|---|