Skip to end of metadata
Go to start of metadata

 Status: Finalised Material from: Ivan



1. Impact

A severe thunderstorm was initiated over the Balkan mountain range (Stara planina) in north-western Bulgaria. It moved to the east-northeast and developed quickly to a supercell badly affecting the town of Byala Slatina and Pleven, one of the biggest cities in northern Bulgaria in the afternoon hours on 15 May 2018. The storm caused all possible types of convective hazards: strong winds, large hail, heavy rain and accompanying flash flooding. It inflicted a lot of damage on properties, cars and infrastructure. Strong winds uprooted trees and caused power outages. In the municipalities of Byala Slatina and Pleven has been declared a state of emergency.

The supercell around the city of Pleven.

Damage inflicted by the severe thunderstorm in Byala Slatina and Pleven, northern Bulgaria.

Large hail along the path of the supercell.

The development of the supercell is associated with very intense lightning activity as well. The track of the supercell over the Danube plain can easily be tracked from the reports in the European Severe Weather Database (ESWD).

ATDnet lightning flash density (left) and ESWD reports (right).

ATDnet flashes - animation.

2. Description of the event

The daytime convection RGB depicts the development of the supercell forming over north-west Bulgaria and moving eastwards causing extensive damage along its path.

Daytime convection RGB animation (every 15 min). Severe storms appear in yellow colour on this RGB.

Large-scale processes (an upper low over Italy, a cold front from southern Italy towards north-western Balkans and a divergent flow at 200 hPa) favour strong synoptic lifting over the Balkan Peninsular.

Z500 and T 850 hPa HRES analysis(left) and MSLP and winds at 200 hPa (right).

High values of CAPE-shear composite parameter (high instability coupled with high deep-layer wind shear) are a sign of favourable conditions for well-organised severe convective storms.

Z500, 10-metre wind and CAPE-shear, HRES analysis.

The radar imagery reveals a hook echo which is a sign of the mesocyclone associated with the supercell.

Radar imagery from the Hail Suppression Agency in Bulgaria.

Radar animation reveals that the storm was very long-lived. It formed at around 14 local time (11 UTC) and could be tracked by this radar station till about 19:00 local time (16 UTC), about 5 hours in total.

Radar animation of the storm development. Data are from the radar station Bardarski Geran provided by the the Hail Suppression Agency in Bulgaria. All the times are in local Eastern European summer time (UTC + 3 hours).

3. Predictability


3.1 Data assimilation


3.2 HRES

With IFS 45r1 two new convective parameters have been introduced:

  • maximum CAPE in the last 6 hours
  • maximum CAPE-shear in the last 6 hours.

These new parameters are derived from hourly model output and they will be tested in replacing current CAPE and CAPE-shear in the EFI product. An example that compares these new model outputs to the instantaneous CAPE and CAPE-shear for this case of severe convection are shown below.

Maximum CAPE in the last 6 hours at T+18h compared to CAPE at T+12h, T+15h and T+18h (left).
The same but for CAPE-shear.

3.3 ENS

The EFI for CAPE-shear is the operational ensemble product that is designed to be used for forecasting such well-organised convection. The forecast gave an early indication of convective instability as early as 10 May 12 UTC run. The signal increased and was pretty high in the last 6 consecutive forecasts. In the last two short-range forecasts the maximum values of the EFI are located where the storm originated over NW Bulgaria.

A sequence of consecutive CAPE-shear EFI forecasts starting from the latest one. The blue triangles denotes Pleven, one of the worst affected places in this storm.

ECMWF is about to offer a number of new products to facilitate forecasting deep moist convection.

One of these products is the ensemble tephigrams part of which is the hodograph view. Hodographs of all 50 members, HRES and control forecast are shown with colours denoting wind at different layers in the atmosphere. This provides information about the ensemble spread alongside wind shear at different layers in the atmosphere. For this event, the spread in the medium range (left) is large especially going up in the atmosphere but the deep-layer shear is relatively large. Low-level winds are weak with easterly components in some members. Winds become stronger higher up in the atmosphere with quite large spread in speed and direction but in general they have a strong westerly component. In the short range (right plot below), the uncertainty decreases but still some spread is present at low levels and between 500 and 200 hPa while between 700 and 500 hPa and above 200 hPa the spread is smaller and the winds are south-westerly. The hodographs shown here suggest the presence of quite large wind shear which favours organisation of convection if the latter can be initiated.

Hodographs for T+108h (left) and T+12h (right) all valid for 15 May 2018 at 12 UTC.

To assess the convective instability and whether convection can be initiated, CAPE plot is provided with the ensemble tephigrams as well. CAPE is plotted in three categories depending on the range of CIN values. The fourth category is CAPE=0. In the first category CAPE is relatively easy to be released due to small CIN values less than 50 J/kg. In the second category, some substantial lift would be necessary to bring the air parcel to the level of free convection (LFC). In the third category convection is unlikely because the CIN is too high. In the medium range forecast (T+108h) ensemble members are distributed in all 4 groups. The most populated groups are the second one and  that with CAPE=0. A lot of members fall also in the third group where HRES and Control forecast belong to. In total there are 23 members (9+14) out of 50 which potentially can develop deep moist convection and HRES is not among them. The instability itself is moderate with CAPE values mostly between 300 and 800 J/kg. There are few members in the first category with CAPE greater than 1000 J/kg. In the short range, most members (28 out of 50) plus HRES and control forecast show convective instability in the forecast but most of them including Control forecast are in the second category requiring more substantial lift so that convection could be initiated. Still quite a lot of members (19) have very large CIN and in this group the CAPE is not particularly high.

CAPE plot showing CAPE distributed into 3 categories + CAPE=0 (shown in the title) for T+108h (left) and T+12h (right).

With the IFS cycle 45r1 lightning density is added as a new model parameter. HRES forecast in the medium range (T+114h) does not have much lightning over the affected area because of the very strong capping (large CIN values) while the short-range forecast (T+18h) have the signal in the right place over NW Bulgaria but overpredicted lightning intensity over southern Romania. In this case, one can see the value of having an ensemble system. Probabilistic forecast in the medium range (T+114h) provided very good guidance on the area under greatest threat of intense lightning activity.

Lightning density in flashes.100km-2.h-1 between 12Z and 18Z on 15 May 2018 from HRES T+114 (left) and T+18h (right).

ENS T+114h Probability of lightning density above 0.1 flashes.100km-2.h-1 (left) and above 2 flashes.100km-2.h-1 (right).

Probabilities of the new maximum CAPE parameter to be above 1000 J/kg highlights the affected area in the medium range whilst in the short-range forecasts that area stands out. It’s worth mentioning that the probability of high CAPE is very high over the central part of southern Romania but severe storms fail to develop there.

Probability of maximum CAPE greater than 1000 J/kg at T+114h (left) and T+18h (right).

Maximum CAPE and maximum CAPE-shear parameters have been used to compute the EFI/SOT for this case. Below, the operational EFI (left) and the EFI with the new parameters (right) is presented.

EFI/SOT for CAPE-shear: operational (left) and with new maximum CAPE-shear parameter (right). The navy triangle denotes the city of Pleven.

EFI/SOT for CAPE: operational (left) and with new maximum CAPE parameter (right). The navy triangle denotes the city of Pleven.

3.4 Monthly forecasts

3.5 Comparison with other centres

4. Experience from general performance/other cases

5. Good and bad aspects of the forecasts for the event

6. Additional material

1 Comment

  1. Well done, excellent work !