The beginning of 2026 was characterised by cold weather in the Nordic countries, while south-western Europe was extremely wet and hit by a number of storms. In January, Storm Harry impacted several Mediterranean countries with extreme precipitation, strong winds and destructive waves. The hardest-hit regions included southern Italy, where persistent gale-force winds brought significant wave heights exceeding 8 metres, and some parts of Sicily observed nearly 500 mm of rainfall in 72 hours (19–21 January). Greece and Tunisia also experienced extreme rainfall. In this short report, we focus on the prediction of the large-scale flow and ocean waves.
On 16 January, an upper-level trough amplified over the eastern Atlantic and formed a cut-off low on 18 January over the Iberian Peninsula. The cut-off low then propagated towards the south-east, settling over Tunisia on 20–21 January (Figure 1a). At the surface, a strong pressure gradient developed over the Ionian Sea and the central Mediterranean between a surface low associated with the upper-level cut-off low and a high-pressure system over the Balkans.
To investigate the predictability of the upper-level trough, we applied a curvature diagnostic, CURV (Curvature Using Radial Variation), recently developed at ECMWF. The diagnostic was computed for 500 hPa geopotential height, with values at each grid point ranging between -16 (closed cyclone) to 0 (no curvature) to +16 (closed anticyclone). For both the Integrated Forecasting System (IFS) and Artificial Intelligence Forecasting System (AIFS) ensembles, the signal for cyclonic conditions (negative CURV) started to appear around 10 January, and both confidently predicted a very cyclonic pattern in forecasts from 12 January onwards (Figure 1b). This indicates that the large-scale flow was well predicted for this case.
The Extreme Forecast Index (EFI) for maximum significant wave height from the IFS ensemble indicates an early signal of high waves across a large part of the central Mediterranean on 20 January (Figure 2), already evident six days before. In a corresponding forecast evolution plot for waves south of Sicily (Figure 1c), the signal starts to appear around 12 January. The gain of the signal in the waves corresponds well with the appearance of the signal for the cut-off low as seen in the forecast evolution for z500 CURV (Figure 1b).
The upcoming AIFS version 2 will introduce several wave parameters (see article by Clare et al. in this Newsletter). The forecast maps in Figure 3 show significant wave height on 20 January at 12 UTC in the IFS analysis 36 h forecasts from the IFS control and AIFS Single v2, with the AIFS model predicting waves reaching above 8 metres south of Sicily. The structure of the forecast field from AIFS agrees well with the IFS forecast and the analysis.
This is supported by the altimeter observations, with the altimeter to the west of the storm reporting waves of 7.5 metres. Measurements from a buoy south of Sicily, operated by the Italian National Institute for Environmental Protection and Research (ISPRA), also reported significant wave height of around 8 metres at this time. Looking at the forecast evolution plot for significant wave height for AIFS Single v2, we also see a consistent prediction of high waves (7–8 metres) from 12 January, in line with the prediction of the cyclonic conditions.
In summary, the high predictability of the upper-level trough in the medium-range forecasts translated well into the ocean wave forecast. This was probably due to the large-scale nature and slow movement of the low-pressure system. While not covered here, the predictability of extreme precipitation varied more, with relatively high predictability over southern Italy, but lower predictability, even in the short range, over northern Tunisia.