201608 - Rainfall Macedonia (FYROM)

 Status:Finalised Material from: Linus, Calum, Tim

1. Impact

Between 6 and 7 August Skopje, the capital of Macedonia, was hit by flash floods and at least 21 people were killed. According to the reports, Skopje got more than 90 mm of rain in two hours. It is unclear when the rain fell, but satellite images suggest Saturday evening .

Reports about the damages can be found here:

http://www.bbc.co.uk/news/world-europe-37002364

Google translate's version of the text from Miha's link below:

"This afternoon regiion Skopje was hit by the strongest storm this year, and one of the stronger over the years. Strong electric discharges were observed between 17 and 19 hours and reached number 800 (for the entire Skopje area).

The storm was accompanied by strong winds, intense rain raining continuously for about 1 hour. The nature of this storm cloud is such that cells are replenished behind one another, and therefore is the longevity of this storm."

2. Description of the event

The plots below shows the analysis for z500 and t850 (left) and short forecast for MSLP and 12-hour precipitation (right), both valid 7 August 00z. Skopje is located on the eastern edge of the upper-level trough.

The list below shows the metars from Skopje with observations of thunderstorms (TS) highlighted.

The plot below show satellite image (from Dundee) from 16UTC (left) 20UTC (right) on 6 August. In the left panel Skopje is marked with a tiny black symbol.

The plot below shows the 24-hour precipitation accumulations from observations received via GTS. The black dot is from Skopje and has more than 90 mm. Unfortunately we do not have any observations from Kosovo and cannot verify the northward extent of the cell.

3. Predictability

3.1 Data assimilation

3.2 HRES

The plots below show HRES forecasts for 24-hour precipitation (6 August 06z - 7 August 06z) and MSLP.

The left plot below shows the 2-metre dew point forecast (shade) and observations (symbols) for the 12-hour forecast from 6 August 00z. The right plot shows the time-series of the dew point for Skopje from HRES (red) and two nearby stations (black). For Skopje the dew point was clearly underestimated just before the convection started.

Examining a Skopje model sounding for 15UTC, and applying observed dewpoints and temperatures (at the true, lower altitude of Skopje) for a parcel curve construction, one sees the potential to release much more CAPE (maybe 5x as much as in the model) and no CIN. The nearest actual sounding from Nis, to the N, has similar Td values to the light green parcel curve, also suggesting this curve is not unreasonable. In addition we have very substantial wind shear on the sounding (conducive to storm longevity) and perhaps some dynamically forced ascent as evening arrived, from an upper trough to the S (to also help keep storms going).

3.3 ENS

The plots below show EFI and SOT for precipitation on 6 August from different initial times (all 00z).

The plots below show the EFI for CAPE (left) and CAPE/Shear from the last forecast before the event. None of the products gave additional information about this event.

3.4 Monthly forecasts

3.5 Comparison with other centres

The plots below show 16 ensemble members from 6 August 00z for 24-hour precipitation (6 August 06z - 7 August 06z).

ECMWF

COSMO-LEPS

3.6 EFAS

Calum had a look at the EFAS forecast for the Macedonia floods, specifically at the 2016080512 forecast. Firstly the soil moistures were quite dry, in NW Macedonia the values were ~20% in all the ensemble members (this is calculated by: (forecast soil moisture - minimum climatology soil moisture)/(maximum climatology soil moisture - minimum climatology soil moisture)). Secondly in EFAS the flash flood warnings are generated from the surface runoff rather than the precipitation, so the dry soils mean that not much of the rain was translated into runoff (less than 20%). We generate flash flood warnings when there is a high probability of exceeding the mean annual maximum climatology runoff value, in the attached map the smaller dark-red pixels show the probability of exceeding this threshold, typically they were very low ~25%. The system has automatically generated 4 warning points (3 in Macedonia and 1 in Serbia) but as you can see from the graphs the probabilities of exceeding any of the return period levels is quite low. EFAS only display points on the website which have >20% probability of exceeding the 5 year return period, because this condition wasn't met at these locations they therefore weren't displayed on the website.

This case might raise an issue about how we generate runoff during an extreme rainfall event, in the flash floods we're using an empirical relationship between soil moisture and the runoff coefficient but the parameters are derived from the LISFLOOD climatology run so the same problem might emerge from the original model results. LISFLOOD uses the Xinanjiang model (aka VIC/ARNO) to generate surface runoff, maybe we should test how suitable it is for extreme flash flood events.

4. Experience from general performance/other cases

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

• The extreme rainfall was missed over Skopje, but present further to the north in the forecasts
• Negative influence of very dry soil in the EFAS forecast?
• Underestimation of the convection due to too low dew point?

6. Additional material