Status: Finalised Material from: Ivan


 

Picture

1. Impact

In the evening hours – between 18 and 21 UTC (UTC+3h local time) on 6 July 2020 a multi-cellular thunderstorm brought torrential rain and caused flooding in the Bulgaria's capital of Sofia. The heaviest rain was recorded in the western and especially south-western parts of the metropolitan area. High wind gusts also accompanied the thunderstorm.

Rainfall totals in Sofia metropolitan area (yellow numbers). Rainfall data are from a set of automated meteorological stations at the National Institute of Meteorology and Hydrology and from Stringmeteo. Mountains and the highest point in each around Sofia valley are also given as a reference; Sofia valley is between 500 and 600 m of altitude.

2. Description of the event

On 6 July the main frontal zone was far from the Bulgaria – from the British Isles towards Central Europe and then to the north across the Baltic States and NW Russia. A cut-off low was centred over the Ionian Sea and in unstable air mass thunderstorms initiated in the afternoon hours. Vertical sounding from Sofia at 12 UTC reveals relatively humid air near the surface (Td=16°C) and steep lapse rates in the mid troposphere resulting in mixed-layer 50-hPa CAPE  of the order of 1000 J/kg. This instability was almost uncapped – CIN was of the order of few J/kg. Vertical wind shear was weak – weak north-easterly, easterly winds near the surface and relatively weak southerly winds in 700-500 hPa layer. With this vertical profile one can expect thunderstorms to initiate especially over complex orography as in this case (Sofia valley is a high valley surrounded by steep mountains to the south). These thunderstorms would tend to be slow-moving due to the weak steering flow and therefore they could produce high amounts of precipitation. Synoptic-scale processes are not particularly supportive to strong, organised severe thunderstorms.

EUMETSAT satellite imagery and ECMWF Z500 analysis at 12 UTC (left) and 18 UTC (right).

Airmass RGB animation.

  

SkewT-logP diagram of the sounding at 12 UTC from the central meteorological station at the National Institute of Meteorology and Hydrology situated in the eastern part of Sofia metropolitan area.

Radar data show multi-cellular convection which initiates over the mountains to the south of Sofia with convective cells moving slowly to the north, northeast. Some of these cells become quasi-stationary and strong (high radar reflectivity factor) producing torrential rain.

DMRL Vakarel_50km_2.5min.mp4DMRL Vakarel_100km_5min.mp4

Radar reflectivity animation from the radar station of the Bulgarian Air Traffic Services Authority (BULATSA) 50 km around the Sofia airport at 2.5 min temporal resolution (left) and  100 km around Sofia airport at 5 min temporal resolution (right). Observation circles are centred at Sofia airport while the radar station is paced to the south-east of Sofia in Vakarel. Sofia airport is situated in the north-eastern part of Sofia metropolitan area.  

Lightning activity shows thunderstorm activity over south-western parts of Bulgaria during the afternoon and evening hours. 

ATDnet lightning flashes – animation.

There is evidence that the thunderstorm over Sofia became so intense because of the specific vertical wind profile and the compex orography of the area – easterly, north-easterly in the boundary layer directly hitting northern foots of Vitosha mountain, constantly initiating new convective cells and their slow motion to the north, north-west.

3. Predictability

 It is striking that NWP guidance was so poor. Even convection-allowing model AROME at 2.5 km of grid spacing failed to predict any rain in the area. ALADIN at 5 km and ECMWF HRES also has literary 0 rain during the torrential rain.

3-hourly precipitation for 15-18 UTC (top) and 18-21 UTC from ALADIN, AROME and ECMWF HRES.

3.1 Data assimilation

 

3.2 HRES

ECMWF HRES in the short range shows afternoon showers over the area but the timing was wrong – all the precipitation in the model occurred before 18 UTC; the shift in afternoon convection was more than 3 hours – convection in the model precipitated more than 3 hours earlier. Precipitation amounts in the forecast were considerably underestimated.

3.3 ENS

The post-processed product ecpoint rainfall increased precipitation amounts at the upper tail but even these amounts clearly underestimated the extremity of the event.

 

99th percentile of 12-hour precipitation from the raw ECMWF ENS (left) and ecpoint rainfall (right).


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