Status:Ongoing analysis Material from: Linus
Picture
1. Impact
During 26-27(?) heavy rainfall affected Greece, with flash floods in many places.
http://www.keeptalkinggreece.com/2018/06/26/greece-weather-floods-mandra-videos/
2. Description of the event
The plots below show mslp and 6-hour precipitation in short forecasts with valid time starting from 25 June 12UTC and every 12h over thereafter.
The plots below show the same as above but for z500 and t850.
3. Predictability
3.1 Data assimilation
3.2 HRES
The plots below show observations and the subsequent HRES forecasts valid 26 July 06UTC to 27 July 06UTC. From the plots below one gets the impression that the precipitation was overestimated in the forecasts, but we have to bear in mind the spare observation coverage. It is also worth to notice that the HRES consistently predicted very high values around Mount Olympus, and we do not have any observations from that area. There was many reports on flash-floods in different part of Greece on social media so it could well be that it was as severe as predicted.
3.3 ENS
The plot below shows the evolution of ensemble (blue) and HRES (red dot) forecasts for 3-day (26-28 June) precipitation over Greece (37N-42N, 20E-24E). I strong signal appeared in the ensemble from 19 June onwards for a very wet period.
Torrential rain was accompanied by enhanced lightning activity as well. European Severe Weather Database (ESWD) provides some extra reports of excessive rain, large hail and even tornados/waterspouts. It's interesting to note that probability of lightning was pretty accurate while convective EFIs failed to provide a signal for the most affected areas in Greece. It gave good guidance over NW Turkey though. One of the reasons for the EFI to fail provide a signal might be the fact that we use already diagnosed CAPE and CAPE-shear as a model output while lightning density is computed directly within the convection scheme. As a result, it can happen that in case of very active long-lasting convection, CAPE could be released constantly without peaking at any extreme values, especially at discrete time steps where the CAPE is provided as a model output (which is every hour), so that it could be shown on the EFI.
ATDnet lightning density, ESWD reports and maximum CAPE and CAPE-shear EFI (left) and probability of flash density > 1 flash per 100m2 per day.
3.4 Monthly forecasts
The plots below show precipitation anomalies for 25 June to 1 July from forecasts with different initial times.
The plots below shows the same as above but for z500.
3.5 Comparison with other centres
3.6 EFAS
The plot below shows the EFAS forecast for a river in central Greece, in the forecast initialised just before the rainfall event.
4. Experience from general performance/other cases