Status: Finalised Material from: Esti, Linus, Ivan
Picture
1. Impact
On 11th August a severe wind storm hit north Poland causing the death of 5 people and significant damages to trees and cutting power to 340,000 households. Rescuers in northern Poland said they had to force their way through "kilometers" of fallen trees in the Tuchola Forest to reach the site of Friday night's tragedy. This severe storm was caused by a mesoscale convective system (MSC) and the observations between 00 and 23 UTC show wind gusts over 30 m/s, and in some points even greater than 35
2. Description of the event
IR satellite image VT:11/08/2017 21UTC
Air mass RGB animation during the outbreak of severe convection over Poland.
European Severe reports valid from 11/08/2017 at 00 UTC to 24 UTC
The plot below shows observations of 24-hour maximum wind gusts on 11 August. Note the observations of above 30 m/s in northern Poland 
3. Predictability
3.1 Data assimilation
3.2 HRES
The plot below shows the 24-hour maximum wind gusts from the last HRES forecast before the event. The maximum wind gusts in the forecast are nowhere near the observed values in northern Poland.
The plot below shows observations of 24-hour precipitation and the last HRES forecast and DWD COSMO 2.5km model.
3.3 ENS
The plots below show CAPE/SHEAR EFI valid for 11 August.
The plot below shows the test product for convection risk from Ivan.
Severe convective storms (predicted by convective EFIs well in advance) brought strong gusty winds (not captured well by the EFI) that killed at least 5 people.
a)
b)
T+48-72h EFI for a) CAPESHEAR and b) wind gusts.
3.4 Monthly forecasts
3.5 Comparison with other centres
The plot below shows the 3-hour wind gusts from COSMO-DWD 2.5 ensemble (member 1) from 11 August 12z 6-9h.The model generated gusts up to 49 m/s.
4. Experience from general performance/other cases
5. Good and bad aspects of the forecasts for the event
- Early detection of the risk from the CAPE/SHEAR EFI
- Clear miss in the convective wind gusts also in the short range (model deficiency), captured by limited-area models.