...
Significant flooding, due to a relatively short-lived burst of convective activity, affected both Copenhagen and Malmo early on in the morning of 31 August.
2. Description of the event
Synoptically, there was a striking absence of warm air, indeed the convection appears to have occurred in the vicinity of two occlusions, in a cyclonic upper and surface pattern.
The plot above shows the synoptic chart from 00UTC on 31 August.
The plot above shows 500 hPa height and 850 hPa temperature at 00UTC Sunday - note the lack of warm air in the area of the strong convection.
The plot above shows IR animation, with what is believed to be the key convective outbreak arrowed.
The plot above shows the standard verification chart (lower right) with two values of 50-100mm near Malmo, but only 20-30mm in Copenhagen. The much higher density obs from around Copenhagen (top left), from DMI shows some massive local totals in the city, in excess of 100mm, but also large local variations across small distances. If we then consider the HRES grid size, and average up the plotted values in the HRES grid box shown, we get a gridbox average of about 35mm. Bearing in mind that this was drawn around the largest obs, and that the grid wouldn't necessarily coincide, we can perhaps re-scale this to 25mm - which maybe a suitable target maximum for HRES to 'aim for'.
The plots below shows the 12-hour accumulated rainfall between 00UTC and 12UTC on the 31 August. The fist plot is the observed rainfall and the second accumulated rainfall from radar (provided by SMHI).
3. Predictability
3.1 Data assimilation
3.2 HRES
The animation above shows successive HRES mslp and 12h ppn forecasts for Sat night. Mostly near Copenhagen these runs show about 10mm, though within the pink zone there will be more locally. so HRES forecasts are a bit underdone.
The plots above show 12-hour accumulation for forecasts from 31 August 00UTC from CF (left), HRES (middle) and 2.8 km AROME (right), provided by SMHI. While the both ECMWF forecast missed the event over Malmo-Copenhagen, AROME had intensive rainfall in the region.
3.3 ENS
...
The plots above shows EFI for rainfall valid for 31 August. The first plot is from 28 August 00UTC and the following from 29, 30 and 31 August. 3 days before (top left) there was a hint of heavy rain, whilst the 2 and 1 day before forecasts had a notable EFI and SOT (>1) signal. The 'day 0' forecast backtracked a little, though is of no relevance for forecasting given issue time.
3.4 Monthly forecasts
-
3.5 Comparison with other centres
The plots below shows the accumulated rainfall between 00UTC and 12UTC on the 31 August. The fist plot is the observed rainfall.
Below we compare different ensembles retrieved from the TIGGE archive. For all ensemble, a maximum of 20 members are plotted. These plots should be compared to the 12-hour accumulation shown above.
31 August 00z
ECMWF (32 km)
NCEP (?? km)
HIRLAM-DMI (11 km?)
COSMO-LEPS (10 km?)
...
COSMO-DE (2.8 km)
30 August 12z
ECMWF
NCEP
HIRLAM-DMI
COSMO-LEPS
COSMO-DE
From both these initial times, it is clear that the both global ensembles (ECMWF and NCEP) missed the extreme event. For the next level of resolution (~10 km LAM), the intensity is higher. For the COSMO-DE with 2.8 km the intensity is in the extreme range (>60 mm), but the uncertainty is still large.
4. Experience from general performance/other cases
...
5. Good and bad aspects of the forecasts for the event
- Increased EFI values from 3 days before the event.
- Clearly underestimated values in the global models.
- Extreme rainfall predicted by ~3 km limited area models (both AROME and COSMO)
- A good example of the value of limited area models.