A Holistic Approach Towards International Disaster Resilient Architecture by ...
Flash floods: Current practice and new developments in quantitative precipitation forecasting
1. Flash floods: Current practice and new
developments in quantitative precipitation
forecasting
Bruno Rudolf
Deutscher Wetterdienst
Dpt. of Hydrometeorology
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 1
2. Overview
1. DWD operational NWP system today
2. Limits of quantiative precipitation forecast
3. Three case studies (meteorol. flood conditions)
4. Ensemble prediction systems, future models
5. Services and summary
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 2
3. DWD operational NWP system
Three nested models: global, Europe, Germany
COSMO-DE: x = 2.8 km, 21 hours
8 runs per day
COSMO-EU: x = 7 km, 78 hours
4 runs per day
GME: x = 30 km, 7 days
2 runs per day
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 3
4. Limits of predictability
Sources of the limits:
chaotic features of the atmospheric circulation
non-steady behaviour of important processes
(e.g. condensation, convection and turbulence)
initial meteorological conditions are partly unknown
boundary conditions are partly unknown
the model equations are non linear on a high grade
limits of computer capacity insuffient spatial resolution
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 4
5. Limits of predictability
Convective cloud dynamics are not sufficiently
simulated by the operational NWP models
Internal structure and
dynamic processes
of a convective cloud
system
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 5
6. Case 1: Elbe flood August 2002
GME LM Observations
Total of 3 daily forecasts (11, 12, 13 Aug.)
The large scale patterns of the precipitation totals were well predicted.
• However, the high precipitation totals were considerably underestimated.
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 6
7. Case 2: Switzerland, Bavaria August 2005
GME LME LM Observations
24h-precip 22 to 23 August , 6:00h
(model run: 22 Aug, 00 UTC)
The spatial structure of daily precipitation was well predicted.
• However, the spatial-temporal sequence was not yet met suffiently.
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 7
8. Case 3: Local flash flood 2005
‘Millennium‘ precipitation 14:50 to 17:20 CEST RADOLAN
event in Dortmund, Combination of
26 July 2008 Dortmund:
radar-based and
200 mm in 2,5 h
raingauge-based
precipitation
The probability catching observations
extreme local rainfall by a
raingauge is less than 10%
even in a dense national
network.
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 8
9. Case 3: Local flash flood 2005
RADOLAN
COSMO-EU (7 km) COSMO-DE (2,8 km) Combination of radar
model forecasts of 24 hours precipitation and gauge observation
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 9
10. Ensemble predictions
Method:
parallel numerical simulation of the same case with modified
initial or boundary conditions or parametrisations within a given
error range
Ensemble predictions deliver:
• Probalistic forecasts
• Quantification of uncertainties
• Risk information
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 10
11. Ensemble predictions
COSMO-DE-EPS
ensemble with
20 members
for Case 3 - Dortmund
26. July 2008
hourly precipitation totals
15:00 -16:00 UTC
COSMO-DE-EPS
is still in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 11
12. Ensemble predictions
COSMO LEPS
basis ECMWF-EPS
Case 2: Flood in
Switzerland, Bavaria
August 2005
Probability of exceeding
the thresholds
50 mm precipitation
Probability of 24h-
precipitation > 50 mm
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 12
13. The future forecast models
ICON
ICOsahedral Non-hydrostatic General Circulation Model
The grid size of the global model
can be regionally improved.
A first operational version is expected
for autumn 2012
resolution: 20 km globaly
5 km for Europe
COSMO DE
is planned to be improved
to 1 km spatial resolution.
Ensemble applications are
planned for both models.
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 13
14. RADVOR-OP
Radar based precipitation nowcasting up to 2 hours
Method:
Identification and extrapolation of
precipitation areas (tracking) and
assessment of cell development
Characteristics:
Spatial resolution: 1 km
Temporal resolution: 5 min / 60 min
Updates every 5 minutes
(The cell processing method is
still in the development stage)
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 14
15. Communication
DWD warnings in the Web
Example: Case 3 (Dortmund)
The text warning 26 July, 10:00 CEST:
… quasi stationary convergency zone from
Emsland via Northrhine-Westfalia, Rhineland-
Palatinate and Saarland to the Black Forest with
Heavy rainfall of more than 40 mm per hour,
Hail and heavy wind gusts possible
Only slow drift of the thunderstorms with
direction north-west. DWD warn map published
26 July 2008, 16:48 CEST
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 15
16. Communication
DWD Hydrometeorological services
Close cooperation and joint projects of DWD
and the water management authorities
Provision of observational data and special
forecast products tailored to the need of the
water management authorities
Consultancy service for water management
Regular and special customer meetings
Participation in relevant committees
Capacity building: Education and training of the
users and of students
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 16
17. Summary
Deterministic models provide very good quantitative forecasts up to
two days lead time.
A 100% certainty will, however, never by possible, especially for the
point in time and space of single small scale precipitation and
thunderstorm events.
Observational data as radar products disseminated (near) realtime
provide disaster mangement with quantiative information in order to
early react and to reduce damage.
New technics:
Ensemble forecast provide probabilities.
Decision makers need to learn to use this information.
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 17
18. Additional material
1. Ensemble movie for Case 3 – Dortmund, 26. July 2008
hourly precipitation totals 15:00 -16:00 UTC
2. Weather chart for Case 3 – Dortmund, 26. July 2008
Convergency line as a cause of the extrem rainfall
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 18
19. COSMO-DE-EPS
ensemble predictions with 20 members
No. 1
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 19
20. COSMO-DE-EPS
ensemble predictions with 20 members
No. 2
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 20
21. COSMO-DE-EPS
ensemble predictions with 20 members
No. 3
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 21
22. COSMO-DE-EPS
ensemble predictions with 20 members
No. 4
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 22
23. COSMO-DE-EPS
ensemble predictions with 20 members
No. 5
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 23
24. COSMO-DE-EPS
ensemble predictions with 20 members
No. 6
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 24
25. COSMO-DE-EPS
ensemble predictions with 20 members
No. 7
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 25
26. COSMO-DE-EPS
ensemble predictions with 20 members
No. 8
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 26
27. COSMO-DE-EPS
ensemble predictions with 20 members
No. 9
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 27
28. COSMO-DE-EPS
ensemble predictions with 20 members
No. 10
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 28
29. COSMO-DE-EPS
ensemble predictions with 20 members
No. 11
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 29
30. COSMO-DE-EPS
ensemble predictions with 20 members
No. 12
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 30
31. COSMO-DE-EPS
ensemble predictions with 20 members
No. 13
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 31
32. COSMO-DE-EPS
ensemble predictions with 20 members
No. 14
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 32
33. COSMO-DE-EPS
ensemble predictions with 20 members
No. 15
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 33
34. COSMO-DE-EPS
ensemble predictions with 20 members
No. 16
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 34
35. COSMO-DE-EPS
ensemble predictions with 20 members
No. 17
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 35
36. COSMO-DE-EPS
ensemble predictions with 20 members
No. 18
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 36
37. COSMO-DE-EPS
ensemble predictions with 20 members
No. 19
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 37
38. COSMO-DE-EPS
ensemble predictions with 20 members
No. 20
COSMO-DE-EPS
One hour precipitation
total 15:00 -16:00 UTC
Dortmund 26. July 2008
Method in development
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 38
39. Case 3: Local flash flood 2005
DWD Weather chart of 26. July 2008 00 UTC (02 MESZ)
DWD -Dpt of Hydrometeorology, B. Rudolf – IGRC Davos 2010 39
Notas del editor
Auch RADOLAN, die Aneichung quantitativer Radarbeobachtungen an die Niederschlagsdaten von Stationsmessnetzen, wurde in Zusammenarbeit mit den für den Hochwasserschutz zuständigen Landesbehörden entwickelt. Bei diesem Verfahren liefern das Radarkomposit (16 Geräte in Deutschland) die hoch auflösenden zeitlichen und räumlichen Strukturen und die Niederschlagsmessdaten die quantitative Genauigkeit. Das RADOLAN-Verfahren wird in Echtzeit angewendet. Alle erforderlichen Daten werden umgehend nach Messung elektronisch an die Zentrale des DWD gesandt und dort ausgewertet. Die angeeichten Produkte wie das oben gezeigte stehen den Hochwasserschutzzentralen innerhalb von 20 Minuten in digitaler Form zur Verfügung. Warndienste erhalten Vorprodukte bereits innerhalb von wenigen Minuten. Neben den DWD-Messdaten werden zur Verdichtung des Gesamtnetzes auch die Daten aus den Landesmessnetzen von Bayern, Baden-Württemberg und Rheinland-Pfalz eingebunden. Vorangegangen war auch eine Abstimmung der Messnetzkonfigurationen des DWD und dieser Länder. Zur besseren quantitativen Erfassung der Niederschlagshöhe ist die Einbeziehung der Daten weiterer Länder geplant. Darüber hinaus sollen RADOLAN entsprechende Produkte für Zentraleuropa in internationaler Zusammenarbeit entwickelt werden.
The radar-based nowcasting procedure is based on pattern recognition in subsequent radar images and (partly) MSG. 5-min qualitative-quantitative radar images are extrapolated in 5-min steps up to max. 2 hours lead time and summed-up to hourly sums for the 2 forecast hours. Hourly sums are quantified under the presupposition of a constant frequency distribution since the latest adjustment procedure. (Method: Mapping of Weibull type of precipitation frequency distributions in forecast and latest adjusted product.) Verification of selected flood cases underway. Ongoing work: Development of methods to derive additional estimates on convective potential and phase of precipitation