Status: Ongoing analysis Material from: Linus, Zied, Tim
New Addition (4 Oct 2022): An examination of model performance for two ECMWF model cycles, centring on the fatal flash floods that occurred around the High Fens area of Germany / Belgium (by Tim Hewson).
1. Impact
On 14 July parts of western Germany and Belgium was hit by extreme rainfall, leading to flash-floods and rapid raise of the water in middle-sized rivers like Meuse. Approximately 200 people were killed in the event. Also Switzerland, France, the Netherlands, and Luxembourg was hit.
https://en.wikipedia.org/wiki/2021_European_floods
https://floodlist.com/europe/2021-floods-germany-forecast-warnings
2. Description of the event
The event started around 12 July with heavy rainfall in southern Germany and Switzerland. The rainfall moved later to the north on the north-western side of the upper-level trough.
The plots below show precipitation observations from DWD, shared via Twitter,
The evaluation below will be based on two regions and period: 3-day (12-14 July) in a box covering the extended Rhine valley (46.5N-52N, 5E-8.5E) and 1-day (14 July) centred on the Ardenne mountains (49.5N-51.5N, 6E-7.5E).
The plots below show analyses of z500 (contour) and T850 (shade) from 9 July 00UTC to 20 July 00UTC, every 24 hour. The rainfall was associated with an upper-level cut-off that moved from the west into central Europe. After the peak in rainfall on 15 July the feature moved further to the south-east and continued to produce heavy rainfall in southern Germany, Austria and the Balkans.
The plots below show analyses of MSLP and 6-hour forecasts of precipitation from 12 July 00UTC to 15 July 00UTC, every 12 hour.
The plots below shows the accumulated precipitation since 60 days before the event for the Rhine valley box (left) and the Ardenne box (right). For the larger Rhine valley box (marked "central"), the rainfall anomaly started in the end of June. For the smaller bon over the Ardenne (marked "RhineNW"), only the main rainfall event contributed according to the available observations.
The plots below shows the soil-moisture index for the top metre for the Rhine valley box (left) and the Ardenne box (right), for the past 3.5 years. The three last summers have been very dry in the region, but for this year the soil has become wet over the two weeks leading up to the flooding event and with a sharp rise during the main rainfall event.
3. Predictability
3.1 Data assimilation
3.2 HRES
14 July
The plots below show observations and HRES forecasts of 24-hour precipitation valid 14 July 06UTC - 15 July 06UTC, from different initial dates. The black rectangle marks the Ardenne box.
12-14 July
The plots below show observations and HRES forecasts of 72-hour precipitation valid 12 July 06UTC - 15 July 06UTC, from different initial dates. The black rectangle marks the extended Rhine valley box.
3.3 ENS
14 July
The plots below show EFI for 1-day precipitation 14 July, from different initial times.
The plot below shows the forecast evolution plot for 24-hour precipitation valid 14 July 06UTC - 15 July 06UTC in the box over the Ardennes. Observation – green, HRES –red, ENS CF – purple, ENS blue box-and-whisker, Model climate – red box-and-whisker. Triangle marks the maximum in the model climate based on 1200 forecasts. The wet signal started to appear on 9 July and grow gradually stronger. From 11 July all ensemble median was above 99th percentile and from 12 July about the model climate maximum.
The plots below show the crossing-point forecast (CPF) for 1-day precipitation 14 July, from different initial times.
12-14 July
The plots below show EFI for 3-day precipitation 12-14 July, from different initial times.
The plot below shows the forecast evolution plot for 72-hour precipitation valid 12 July 06UTC - 15 July 06UTC in the box around the Rhine valley. Observation – green, HRES –red, ENS CF – purple, ENS blue box-and-whisker, Model climate – red box-and-whisker. Triangle marks the maximum in the model climate based on 1200 forecasts. A signal started to appear from 5 July, but started to increase from 9 July. For 10 July the ensemble median was above 99th percentile of the model climate.
3.4 Monthly forecasts
The plots below show weekly precipitation anomalies valid 12-18 May.
5 Comments
Edouard Goudenhoofdt
The second region (49.5N-51.5N, 6E-7.5E) corresponds to the Eifel moutains and a larger impacted area in Germany. It does not correspond to the Ardennes mountains, which are centered in Belgium, and have also been impacted. I would like to see the nice forecast evolution figure for this region. The rainfall event is best captured, taking into account manual observation limitation, from 13 July 06UTC to 15 July 06UTC there.
Does the apparent spatial accuracy of the forecasts not allow studying the smaller but most affected areas related to the rivers Ahr, Erft and Vesdre?
Linus Magnusson
Thanks for the suggestion! Could you suggest coordinates of such a region and I can run the evolution plot?
Best Regards,
Linus
Edouard Goudenhoofdt
Thank you Linus for the quick reply !
We can look at the most extreme rainfall on Figure 6 (left) of the attribution study. For studying the forecasted precipitation from 13 July 06 UTC to 15 July 06UTC, I suggest this box: 49.5N-51N, 5E-7E.
Best regards,
Edouard
Linus Magnusson
Hi Edouard,
Here are plots for the suggested 48-hour precipitation. The plots are the observations, HRES from 13 July 00UTC and the evolution plot for the precipitation in 49.5-51N, 5E-7E. The evolution in the ensemble is coming out good for this box as well.
Best Regards,
Linus
Edouard Goudenhoofdt
This is perfect.