Rademacher - Operational low flow forecasts

Transcription

Rademacher - Operational low flow forecasts
Operational low flow forecasts
Silke Rademacher
Department Water Balance, Forecasting and Predictions
Federal Institute of Hydrology (BfG)
Koblenz
KHR-CHR Workshop Low flows and droughts
25 & 26 September 2007, Würzburg
Operational low flow forecasts
Page 1
Responsibility in Germany
• Navigability of ships and maintenance
of large rivers:
Federal Waterway and
Navigation Administration (WSV)
• Flood defence, flood warning
and forecasting service
German Federal States
Operational low flow forecasts
Page 2
Water Level Forecasting System WAVOS
Operational low water forecast
HochwasserVorhersagezentrale
Frankfurt (Oder)
Bundesanstalt für
Gewässerkunde,
Koblenz
Vorhersagezentrale
Elbe, Magdeburg
Hochwassermeldezentrum
Rhein, Mainz
WAVOS Saar
Hochwassermeldezentrum
Saarland, Saarbrücken
Operational low flow forecasts
HochwasserVorhersagezentrale
Main, Hof
Page 3
Low flow forecasting along the Rhine
> 7 gauging stations: Oestrich,
Kaub, Koblenz, Köln, Düsseldorf,
Duisburg-Ruhrort, Emmerich
> Forecasting horizon: 48 hours
> Forecasting time: 7:00
> Daily Forecast: Water level
at gauge Ruhrort ≤ 400cm
Operational low flow forecasts
Page 4
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
Operational low flow forecasts
Page 5
WAVOS Rhine
> Tributaries Rhine model
> Maxau/Oberrhein
> Rockenau/Neckar
> Neubrück/Erft
> Raunheim/Main
> Hattingen/Ruhr
> Grolsheim/Nahe
> Schermbeck/Lippe
> Kalkofen/Lahn
> Trier u. Cochem/Mosel
> Friedrichstal/Wied
> Altenahr/Ahr
> Menden/Sieg
> Opladen/Wupper
Operational low flow forecasts
Page 6
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
Operational low flow forecasts
Page 7
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
Operational low flow forecasts
Page 8
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
>Moselle-Saar-Sauer (425 km)
Operational low flow forecasts
Page 9
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
>Moselle-Saar-Sauer (425 km)
> Statistical model
>Maxau/Rhein
>Rockenau/Neckar
>Fremersdorf/Saar
Operational low flow forecasts
Page 10
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
>Moselle-Saar-Sauer (425 km)
> Statistical model
>Maxau/Rhein
>Rockenau/Neckar
>Fremersdorf/Saar
> External forecasts
Operational low flow forecasts
Page 11
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
>Moselle-Saar-Sauer (425 km)
> Statistical model
>Maxau/Rhein
>Rockenau/Neckar
>Fremersdorf/Saar
> External forecasts
> Forecaster’s estimate
Operational low flow forecasts
Page 12
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
>Moselle-Saar-Sauer (425 km)
> Statistical model
>Maxau/Rhein
>Rockenau/Neckar
>Fremersdorf/Saar
> External forecasts
> Forecaster’s estimate
Operational low flow forecasts
Page 13
WAVOS Rhine - Results
20
100
16
80
14
12
60
10
8
40
6
1-Tages-Vorhersage
20
Standardabweichung [cm]
Anteil der Vorhersagen mit einer Differenz
kleiner 10cm [%]
18
4
2-Tages-Vorhersage
Standardabweichung (1-Tages-VHS)
2
Standardabweichung (2-Tages-VHS)
0
0
Oestrich
Kaub
Koblenz
Köln
D`dorf
Ruhrort
Emmerich
Vorhersagepegel
Operational low flow forecasts
Page 14
WAVOS Rhine - Results
Anteil der Vorhersagen mit einer Differenz zwischen
Vorhersage- und Messwert kleiner 10 cm [%]
100
90
80
70
60
Pegel Oestrich
Pegel Kaub
Pegel Koblenz
Pegel Köln
Pegel Düsseldorf
Pegel Ruhrort
Pegel Emmerich
50
40
Auswertezeitraum: 01.07.2005 - 30.06.2007
30
12
Operational low flow forecasts
24
36
Vorhersagedauer [h]
48
Page 15
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
>Moselle-Saar-Sauer (425 km)
> Statistical model
>Maxau/Rhein
>Rockenau/Neckar
>Fremersdorf/Saar
> External forecasts
> Forecaster’s estimate
Operational low flow forecasts
48h-Forecast
☺
Page 16
Low flow forecasting along the Rhine
> 7 gauging stations: Oestrich,
Kaub, Koblenz, Köln, Düsseldorf,
Duisburg-Ruhrort, Emmerich
> Forecasting horizon: 48 hours
> Forecasting time: 7:00
96 hours
> Daily Forecast: Water level
at gauge Ruhrort ≤ 400cm
> Dissemination via email to river
headquarter and Internet (ELWIS)
Operational low flow forecasts
Page 17
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
>Moselle-Saar-Sauer (425 km)
> Statistical model
>Maxau/Rhein
>Rockenau/Neckar
>Fremersdorf/Saar
> External forecasts
> Forecaster’s estimate
Operational low flow forecasts
96h-Forecast
/
Forecasting Range to short
Only for 2 tributaries available
....
Page 18
WAVOS Rhine
> 1D hydrodynamic model
>Rhine and Moselle (685 km)
>Main (380 km)
>Moselle-Saar-Sauer (425 km)
> Statistical model
>Maxau/Rhein
>Rockenau/Neckar
>Fremersdorf/Saar
> External forecasts
> Precipitation runoff model
Operational low flow forecasts
To obtain
> a consistent forecast
> with a adequate leadtime
for all tributaries
Page 19
NL
HBV Rhine
D
Precipitation runoff model
> Catchment upstream of Lobith
( more than 160.000 km²)
L
> 134 subbasins (each between 500
and 2.000 km²)
> time step: 1 hour
F
A
CH
Operational low flow forecasts
Page 20
NL
HBV Rhine
D
Development in cooperation with
RIZA:
> Flood forecast RIZA
> case studies
>Transnational effects of
extreme floods at the Lower
Rhine
> Climate change
L
F
A
CH
Operational low flow forecasts
Page 21
WAVOS Rhine – FEWS DE
Operational low flow forecasts
Page 22
External
forecasts
Hydro
measurements
Meteo
forecasts
Meteo
measurements
WAVOS Rhine – FEWS DE
Hydro
measurements
FEWS – DE
WAVOS Rhine
forecast for
tributaries
Rhine forecast
Operational low flow forecasts
Page 23
FEWS DE
Precipitation runoff model - input data
Forecast
Initialisation
Phase
Data
Parameter
Spatial
Timestep Leadtime
Resolution
Measured
TTRR station data
P, T
45 stations
1h
Measured
synop station data
P, T
204 stations 6 h / 12 h
W
47 gauges
≤1h
DWD weather forecast
Local Model (LM)
P, T
~ 7 km
1h
78 h
DWD weather forecast
Global Model (GME)
P, T
~ 40 km
3h
174 h
W
gauge
Rheinfelden
1h
67 h
Measured
water level
BAFU
external forecast
Operational low flow forecasts
Page 24
WAVOS Rhine – FEWS DE Results
Pegel Köln (10 cm – Schwelle)
90
80
70
Pegel Köln (20 cm – Schwelle)
100
Pegel Köln
60
2-Tages-VHS (operat. 2007)
GME + manuelle Anpassungen
GME automatisch
LM + manuelle Anpassungen
LM automatisch
HVZ + LM
WAVOS + LM
HVZ + WAVOS + LM
50
40
30
12
24
36
48
Anteil der Vorhersagen mit einer Differenz kleiner 20 cm
Anteil der Vorhersagen mit einer Differenz kleiner 10 cm
100
90
80
70
Pegel Köln
60 60
Vorhersagedauer [h]
50
40
72
84
2-Tages-VHS (operat. 2007)
GME + manuelle Anpassungen
GME automatisch
LM + manuelle Anpassungen
LM automatisch
HVZ + LM
WAVOS + LM
HVZ + WAVOS + LM
96
30
12
Operational low flow forecasts
24
36
48
60
Vorhersagedauer [h]
72
84
Page 25
96
FEWS DE
One of the main sources for the increasing of
uncertainty is the uncertainty in the weather
forecast
⇒ Ensemble weather predictions
⇒ Project
“Verification of Ensemble forecasting in the
Rhine basin”
⇒ RIZA – BfG Cooperation /CHR -Project
⇒ WL | Delft hydraulics
Operational low flow forecasts
Page 26
Verification of Ensemble forecasting in the
Rhine basin
• ECMWF-EPS ensemble (re-sampled as used in FEWS-NL)
• 50 ensemble members + 1 control run
• timestep 12h
07
0
2
/
• leadtime 240h
06
/
8
0
4
0
0
• Spatial resolution ~ 80km
6/2
0
/
7
0
• COSMO-LEPS ensemble
• 16 ensemble members
• timestep 3h
• leadtime 5 1/2d
• Spatial resolution ~ 10km
09/0
Operational low flow forecasts
7
0
0
1/2
07
0
2
/
5
21/0
Page 27
FEWS DE – Ensemble
Operational low flow forecasts
Page 28
FEWS DE - Ensemble
Lahn Catchment Area
Operational low flow forecasts
Page 29
FEWS DE - Ensemble
Operational low flow forecasts
Page 30
FEWS DE - Ensemble
Operational low flow forecasts
Page 31
FEWS DE - Ensemble
Operational low flow forecasts
Page 32
FEWS DE - Ensemble
Operational low flow forecasts
Page 33
Verification of Ensemble forecasting in the
Rhine basin
•What is the skill, reliability and potential utility of
the ensemble in predicting events at different lead
times?
•What are the effects of scale on the skill of the
ensemble predictions?
•Do the high resolution ensembles available
through COSMO-LEPS indicate a higher skill
than the low resolution ECMWF-EPS ensemble
where the scale issue becomes apparent?
Operational low flow forecasts
Page 34
Verification of Ensemble forecasting in the
Rhine basin
•Several statistics commonly applied in
verification of probabilistic forecasts were used
• the Brier (Skill) Score,
• the Ranked Probability
• (Skill) Score
• verification diagrams
• Relative Operating Characteristic
• Attributes diagram.
Operational low flow forecasts
Page 35
Verification of Ensemble forecasting in the
Rhine basin - Results
• At low flows it is shown that the ensemble forecasts
are a reliable indicator, with a high forecast skill for
long lead times.
• For the smaller catchments the skill of the ensemble
forecast was clearly shown to deteriorate faster than
in the larger catchments.
• In all cases the ensemble forecast were demonstrated
to have positive skill, meaning that there is utility in
using ensemble in forecasting for the Rhine basin.
Operational low flow forecasts
Page 36
Verification of Ensemble forecasting in the
Rhine basin - Results
• The period over which the ensembles have been verified is
limited, with few events of significance occurring in the
Rhine basin. As a result the skill in predicting rare events for
which the forecasting system has been constructed has not
been assessed, but can only be inferred on the results found
for the evaluation period.
• Relatively short period for which the COSMO-LEPS
ensemble is currently available
=> Repeat the analysis in 1 to 1½ years when a more
substantial set of ensembles is available, and if available a
larger number of (extreme) events
Operational low flow forecasts
Page 37
Thank you for your attention !
Silke Rademacher
Referat Wasserhaushalt, Vorhersagen und Prognosen
Bundesanstalt für Gewässerkunde
Am Mainzer Tor 1
56068 Koblenz
Tel.: 0261/1306-5858, Fax: 0261/1306-5280
E-Mail: rademacher@bafg.de
www.bafg.de
Operational low flow forecasts
Page 38