ECMWF Newsletter #163

Forecasting February’s wet and stormy weather in parts of Europe

Linus Magnusson
Laura Ferranti
David Lavers
Fredrik Wetterhall


February 2020 was dominated by a strong westerly flow over northwest Europe. Several severe cyclones affected the region. These include storm Ciara on 8–9 February and storm Dennis on 15–16 February (as named by Met Éireann and the UK Met Office), and further systems on 22–23 February and 28–29 February. The cyclones brought strong winds and heavy rainfall leading to extreme precipitation accumulations over the month. New monthly precipitation records were set in England and Wales, Denmark and parts of southern Sweden, and there was some flooding in those countries. In this article we will focus on various aspects of the predictions for storm Dennis.

Main features of Dennis

Dennis originated from a wave that appeared on 13 February along the northeast US and southeast Canada while there was an intense extratropical cyclone in the North Atlantic, which brought hurricane force winds to Iceland. This wave developed into an extremely rapidly intensifying cyclone, which replaced the previous system in the North Atlantic. Dennis’s remarkable development was likely favoured by the baroclinic environment associated with the remnants of the previous system. Cyclone Dennis formed over the central North Atlantic on 15 February. During its deepest stage, it was located to the south of Iceland with a central pressure below 920 hPa on 16 February 00 UTC. This makes it one of the most intense extratropical cyclones on record. During its mature stage, it grew bigger than the European continent, dominating the weather across the North Atlantic and Europe. Dennis pushed severe weather conditions into western Europe and moved a very warm air mass further east. Although the cyclone centre was located far from the western areas of Britain, Dennis brought extreme precipitation to those areas and later to southern Scandinavia. In Wales and central England, several villages were flooded after the passage of the cyclone.

Precipitation anomalies for February 2020. The plot shows average daily precipitation anomalies in February 2020 (mm/day) relative to the reference period 1981–2010, as provided by the Copernicus Climate Change Service implemented by ECMWF, based on ERA5 reanalysis data.

Atmospheric river forecast

The extreme precipitation events during Ciara and Dennis were associated with intense advection of water vapour from across the North Atlantic in the warm sector of the cyclones. This feature is often referred to as an atmospheric river. Atmospheric rivers are known to cause extreme rainfall, especially when forced to rise over orography. At longer lead times, they are in general more predictable than precipitation. That is why ECMWF has recently begun to produce an Extreme Forecast Index (EFI) for the vertically-integrated water vapour flux. For both storms, ECMWF’s ensemble forecasts captured the risk of heavy precipitation 7 days in advance. Initial findings suggest that the EFI signal for water vapour flux towards northwest Europe was stronger and appeared earlier than the EFI signal for precipitation. This added confidence and process understanding to the precipitation forecast.

Extreme Forecast Index maps for storm Dennis. The maps show the Extreme Forecast Index (EFI, shading) and Shift of Tails (SOT, contours) for total precipitation (left) and vertically-integrated water vapour flux (right) during storm Dennis valid on 16 February in day 7 forecasts. The green lines indicate areas where EFI values were larger than 0.9 in day 1 forecasts.

Extended-range predictions

The strong cyclonic activity over the North Atlantic was associated with a strong positive phase of the North-Atlantic Oscillation (NAO). An NAO+ signal for the week from 10 to 16 February was present in the extended-range forecast from 23 January and became stronger in the forecast from 30 January. The predicted flow-regime anomaly is visualised in our new two-dimensional plot of NAO-Blocking regime phases for the northeast Atlantic (see Newsletter No. 158). Although this product was originally designed to give early warnings about cold spells, it also highlights the likelihood of a westerly flow across the Atlantic with warm and wet weather over western Europe. The 30 January flow-regime anomaly forecast was shifted towards a combination of NAO+ and a trough over Scandinavia (Blocking–). The verifying analysis was in the same region of the diagram but somewhat more extreme.

Weather regime outlook for the northeast Atlantic. The plot shows the NAO-Blocking weather regime probability density function for an ensemble forecast starting on 30 January 2020 for the week from 10 to 16 February. Daily values of the verifying analysis are represented by dots: yellow (first day) to brown (last day of the target period).

Flooding in Sweden

During February 2019, parts of southern Sweden saw record precipitation with more than four times the normal monthly total. The situation was made more severe by the fact that almost all precipitation fell as rain instead of snow due to the anomalously mild weather. From a hydrological perspective, although the rivers in the region are relatively small, the large storage capacity in many lakes makes them respond relatively slowly. However, at the end of February the situation became very difficult, with rivers and lakes close to bursting their banks. The Copernicus Emergency Management Service (CEMS) has recently begun to produce regional river discharge forecasts as part of the European Flood Awareness System (EFAS) based on ECMWF extended-range forecasts. As discussed above, as early as the end of January a wet signal appeared in atmospheric forecasts for the middle of February. This signal translated into a high river discharge anomaly for southeast Sweden in the EFAS extended-range forecasts from 30 January for a lead time of 3–4 weeks ahead.

Regional river discharge prediction for southeast Sweden. The plot shows the area average river discharge prediction for southeast Sweden in the Copernicus Emergency Management Service (CEMS) forecast from 30 January. According to the ensemble median, average discharge was predicted to be above the 90th percentile in weeks three and four. The analysed outcome was that the discharge peaked just below 10 m3/s at the end of February (week five in the plot).


A period of high cyclonic activity over northwest Europe can bring forecasting challenges on various aspects and timescales. In this article, we have shown how several recently added products could help users to interpret the forecast information at different lead times. Early signals of the potential for severe weather were seen in the large-scale flow patterns three weeks ahead, while more information about the scale and intensity of the event were available as the event approached.

Feedback from Member States

“At DMI (Danish Meteorological Institute), warnings for strong wind and storms are normally first issued about 36 hours before the event is expected, while a pre-warning is issued up to 5 days ahead. This winter in Denmark has been rather windy, and the ECMWF/IFS model has been very useful when deciding if a pre-warning should be issued. However, we sometimes found that gusts from the Integrated Forecasting System (IFS) seemed too high, especially in warm sectors ahead of approaching cold fronts, based on experience of measured gusts in past similar weather situations over Denmark. There was a particular case on 9 February when the IFS predicted gusts of 30–36 m/s widespread over Denmark while the typical measured values were gusts of 20–25 m/s. However, locally in the western part of Denmark 30–36 m/s was measured.” (Jesper Eriksen, Forecaster, DMI)

“Iceland got hit with a ~930 mb low on 14 February. The skill of the ECMWF forecast was extremely impressive for this cyclone, the +84 hours ensemble had all members agreeing on the position and depth. But even before that the low was very consistent in the majority of members and the deterministic forecast. For the conditions on Friday, we were able to issue the first yellow warnings on Tuesday. Warnings became orange on Wednesday and upgraded to red on Thursday for some regions, giving all stakeholders a long lead time to prepare. Winds were locally over 30 m/s in the mean with gusts of up to 60 m/s. Damage to infrastructure was quite substantial but no fatalities.” (Elín Björk Jónasdóttir, Group leader of weather services, Icelandic Meteorological Office)