Status:Finalized Material from: Ivan
Discussed in the following Daily reports:
http://intra.ecmwf.int/daily/d/dreport/2014/06/23/sc/
http://intra.ecmwf.int/daily/d/dreport/2014/06/24/sc/
1. Impact
An unstable air mass triggered a lot of convection accompanied by heavy thunderstorms, hail and torrential rain across Bulgaria from 15 to 19 June 2014. Climatologically June is one of the wettest months over most parts of the country. Nevertheless rainfall accumulations exceeded monthly means in some regions, especially in eastern Bulgaria. In NE of Bulgaria more than 100 mm fell just in 24 hours on 19 June causing a dozen of fatalities, most of them during a deluge in the residential area of Asparuhovo in the coastal city of Varna.
120-hour rainfall from 06UTC on 15th June to 06UTC on 20th June 2014 (left) and 24-hour rainfall from 06UTC on 19th to 06UTC on 20th June 2014 (right) including observation (numbers) and gridded analysis product with LAM ALADIN and rain-gauge observations. High-density observations and gridded analysis courtesy of the National Institute of Meteorology and Hydrology, Bulgaria.
2. Description of the event
HRES analyses show that there is a strong synoptic forcing that led to severe weather. There was initially a moderate easterly flow of humid air near the surface. The soil itself was very wet as well following few days of heavy convective showers. At the time of the heaviest rain over NE Bulgaria in the evening on 19 and in the morning on 20 June a rapid cyclogenesis occurred over the Black Sea just offshore the Bulgarian coast. The deep cyclone that formed moved to the north and then to the east turning the wind from E to NW over Bulgaria.
HRES analyses of MSLP and 10-metre wind (left) and soil moisture analysis before the flash floods on 19th June 2014 at 00UTC (right).
Satellite RGB Air mass animation showing strong convective activity over the Balkans and cyclonic circulation over the Black Sea (EUMETSAT).
The animation above from the Romanian meteorological radar network shows a long-lasting quasi-stationary convective cell over Varna, just in the region where the deluge occurred. You can see that cell on the right side of the animation just over Varna on the Black Sea coast. It is striking that the cell had been staying over the same place for more than 2 hours producing heavy rain. As a consequence a large amount of water washed away cars, houses and caused the death of about 10 people in Varna. (Animation is created with data from the National Meteorological Administration of Romania available free on the web).
3. Predictability
3.1 Data assimilation
3.2 HRES
The plot above shows comparison of three consecutive short-range HRES forecasts versus the observed 24-hour rainfall (top left panel). Observations are displayed with numbers while the shading represents a combination of LAM ALADIN and observations to create precipitation analysis on a regular grid. The analysis data are kindly provided by the Forecast Department of the National Institute of Meteorology and Hydrology in Bulgaria. It shows huge amounts of rain over NE Bulgaria. Rainfall totals of 155 mm were recorded in two stations and few other with rain above 100 mm. Although this is a 24-h accumulation it actually fell just in few hours. The convective nature of the rainfall makes forecasting of this event very difficult. Indeed, the short-range HRES forecast gives considerable amounts of rain but not surprisingly the largest totals are misplaced. What is a little bit surprising is the fact that the latest forecast seems worse than previous two over NE Bulgaria where the largest amounts were recorded.
Longer-range HRES forecasts (see the figure above) vary much more but they had kept forecasting heavy rain quite far in the medium range.
Synoptic-scale forecast (see the figure above) about the cyclogenesis seems to be good in the HRES.
Looking at the HRES we see that the instability over the country is quite high. CAPE in the forecast initialised at 00UTC on the day of the event exceeded 2500 J/kg in south-eastern parts of the country. Unfortunately maximum values of CAPE were not particularly high over the most affected region around Varna where a quasi-stationary convective cell developed producing huge amounts of rain just in few hours. This can explain why the model underestimated the rainfall totals in that region. Another important remark is related to the frequency of the model output: maximum CAPE during the period is significantly higher in case of hourly model output compared to the operational 3-hourly output.
Maximum CAPE during the specified period is shown as well as maximum CAPE in case the convective inhibition (CIN) is less than 200 J/kg and therefore convection can be triggered. When CIN is greater than 200 j/kg the cap is so strong that it prevents convection from happening.
The tephigram for Varna, the city affected by flash floods, shows that the atmosphere was unstable, humidity quite high coming from the Black Sea with easterly winds at the surface. The black curve in the analysis is drawn taking the maximum temperature in Varna on that day which was close to 23C. The 12-hour forecast profile was a little bit different with a deep saturated layer from the surface implying rainfall at that time.
Tephigrams for Varna: analysis (left), 12-h forecast (right).
3.3 ENS
The flash floods occurred on 19 June in the evening and in early morning on the 20 June when just in few hours more than 100 mm of rainfall were recorded in NE part of the country. Flash floods occurred there. The analyses show that 99th model climate percentile for 24h total rainfall was exceeded in many places in the east of Bulgaria. Some local heavy showers were observed in far western regions of the country as well. If we compare observations with the 95th model climate (M-climate) percentile, we'll notice areas of local torrential downpours in many parts of Northern Bulgaria as well. The shortest range EFI forecast matches quite well the areas affected by heavy rain but the EFI has never reached extremely high values close to 1 over NE Bulgaria. This is more likely due to the convective nature of the rain and the large uncertainty in the forecast that we might expect in such cases. Note that there is 6-hour shift between the validity of the EFI forecast and the accumulations represented by the observations: EFI is valid from 00 to 00UTC whilst accumulations are for periods starting from 06UTC.
Observed 24-hour total precipitation compared with the model climate. Red triangles denote accumulations that exceed certain model climate percentiles (95th or 99th percentile). Green triangles on the EFI forecast plot denote observations that exceed 99th model climate percentile. The shortest range EFI/SOT forecast is also displayed. The CDFs from the most affected regions show that the forecast was quite uncertain and precipitation amounts were underestimated although they appear extreme compared to the M-climate.
The EFI forecast provided good guidance throughout the forecast from the longest to shorter lead times. At the beginning, 7 days in advance, positive SOT over Bulgaria showed a risk of abnormal rainfall whilst the EFI was positive but still values were low. Closer to the event the EFI became higher.
EFI forecasts for 24-hour total precipitation with observations on top. Green triangles denote observations that exceeded 99th model climate percentile.
The same signal we get by looking at longer accumulation periods, 3 and 5 days. The most extreme rainfall is again in NE parts of Bulgaria, with rainfall totals exceeding the 99th percentile of the M-climate in many places in the east of Bulgaria and also more locally in some western parts of the country. EFI provides good guidance throughout the forecast.
Observed 72- and 120-hour precipitation compared with the model climate. Red triangles denote accumulations that exceed certain model climate percentiles (95th or 99th percentile). Green triangles on the EFI forecast plots denote observations that exceed 99th model climate percentile. The shortest range EFI/SOT forecast is also displayed.
Longer lead time EFI/SOT forecasts for 72- and 120-hour total precipitation. Green triangles denote observations that exceed 99th model climate percentile.
The extended-range EFI, for T+240-360 gives a hint of wet weather.
EFI for total precipitation, forecast for T+240-360h. Yellow/red colours mean wetter than normal conditions.
Cyclonic features according to 12-hour maximum precipitation (-6/+6h) in different forecast valid for 19/06/2014 12UTC and 20/06/2014 00UTC.
3.4 Monthly forecasts
Exceptionally wet weather in the Balkans has been predicted well in the medium range by the ENS. Positive precipitation anomalies appeared even in the far medium range for days 8 to 14. For longer ranges in the monthly forecast there is no signal.
Precipitation anomalies from ECMWF monthly forecast compared with the analysis.
3.5 Comparison with other centres
Four control forecast of total precipitation from TIGGE: ECMWF, UKMO, NCEP and LAM Aladin Austria. Having different resolution and different model characteristics they all forecast significant rainfall in the eastern part of Bulgaria.
Forecasts of 24-h accumulated rainfall valid for 19 June 06UTC to 20 June 06UTC from ECMWF HRES (~16 km) and COSMO LEPS control (~10 km). Black numbers denote observed rainfall accumulations. COSMO LEPS is a non-hydrostatic limited-area ensemble which uses initial and boundary conditions from ECMWF ENS. It comprises 16 perturbed members and a control forecast with a horizontal resolution of about 10 km. To obtain initial and boundary conditions an extended ensemble of 2 consecutive ECMWF ENS runs is used (51+51 members) and a cluster analysis is applied to these 102 members. For each of the 16 clusters a representative member is selected to provide initial and boundary conditions. The ensemble is run up to T+132h.
Forecasts of 24-h accumulated rainfall valid for 19 June 06UTC to 20 June 06UTC from ECMWF HRES and UKMO global model.
Probability of rain above 50 mm/24h from ECMWF ENS (~32 km) and COSMO LEPS (~10 km) in the short range. Numbers in the brackets show the horizontal resolution of each ensemble.
Probability of rain above 20 and 50 mm/24h from ECMWF ENS and COSMO LEPS in the medium-range.
4. Experience from general performance/other cases
5. Good and bad aspects of the forecasts for the event
- The EFI/SOT gave a signal of abnormal rainfall well in advance - 4 to 5 days before the event on 19 June.
- Convective nature of the rainfall makes forecasts of its location and intensity quite difficult. Forecasts vary a lot from one run to another considerably altering the location of the most intense rainfall. This increases the uncertainty of the forecast and makes the localisation of the heavy rain extremely difficult even in the short range. The cyclogenesis over the Black Sea is well captured by IFS.