Adding value to extended-range forecasts in northern Europe by statistical post-processing using stratospheric observations

Abstract

Abstract. The strength of the stratospheric polar vortex influences the surface weather in the Northern Hemisphere in winter; a weaker (stronger) than average stratospheric polar vortex is connected to negative (positive) Arctic Oscillation (AO) and colder (warmer) than average surface temperatures in northern Europe within weeks or months. This holds the potential for forecasting in that timescale. We investigate here if the strength of the stratospheric polar vortex at the start of the forecast could be used to improve the extended-range temperature forecasts of the European Centre for Medium-Range Weather Forecasts (ECMWF) and to find periods with higher prediction skill scores. For this, we developed a stratospheric wind indicator (SWI) based on the strength of the stratospheric polar vortex and the phase of the AO during the following weeks. We demonstrate that there was a statistically significant difference in the observed surface temperature in northern Europe within the 3–6 weeks, depending on the SWI at the start of the forecast. When our new SWI was applied in post-processing the ECMWF's 2-week mean temperature reforecasts for weeks 3–4 and 5–6 in northern Europe during boreal winter, the skill scores of those weeks were slightly improved. This indicates there is some room for improving the extended-range forecasts, if the stratosphere–troposphere links were better captured in the modelling. In addition to this, we found that during the boreal winter, in cases where the polar vortex was weak at the start of the forecast, the mean skill scores of the 3–6 weeks' surface temperature forecasts were higher than average.

Winter Subseasonal Wind Speed Forecasts for Finland from ECMWF

“…Our results are comparable to those of Lynch et al (2014), who concluded that there is statistically significant skill in predicting weekly mean wind speeds over areas of Europe at lead times of at least 14-20 d. Lynch et al (2014) used five years of operational forecasts; their CIs were narrower than ours, and they could make better inference using operational forecasts.…”

“…Including such information in NGR is rather straight-forward. Furthermore, the most advanced non-linear methods (such as deep learning, e.g., Liu et al, 2016) need a large amount of training data to avoid overfitting, and the data sets available in subseasonal forecasting are of relatively modest size. Therefore, NGR, trainable with reasonable small data, is a feasible choice for future applications.…”

“…This study concentrates on subseasonal wind forecasts in winter, as forecasts for winter in northern Europe are known to be more skilful than forecasts for other seasons (e.g., Lynch et al, 2014). Further, we consider only wind speed, not direction.…”

“…Following Lynch et al (2014), we concentrated on winter forecasts and the lead times up to the start of the third week. The CLIPS project covered two winters (2016-2017 and 2017-2018), but two winters of operational forecasts would have been a rather small data set to make meaningful inferences about how operational wind forecasts perform, and we decided to concentrate on the reforecasts.…”

Assimilating atmospheric infrasound data to constrain atmospheric winds in a two‐dimensional grid

“…However, the observational coverage of these two layers is nowhere near as dense as that of the troposphere. Some of the largest uncertainties in current reanalyses occur in these levels, leading the wind representation of mesospheric winds to lag behind other areas (Baker et al ., 2014; Korhonen et al ., 2019). For instance, Duruisseau et al .…”