Data-assimilation in flood forecasting for the river Rhine
between Andernach and Düsseldorf
COR-JAN VERMEULEN
Introduction
Huge investments in flood prevention, flood early warning, flood mitigation measures and flood management
FloodMan:Near real-time flood forecasting, warning and management
Introduction
• Data-assimilation of hydrological and hydraulic
parameters for flood forecasting
• Independent of the computer models used
• Use of in-situ and satellite data
• Pilot:Rhine river, Germany
Data-assimilation
• Combining model estimates with measured data
• Including measure of uncertainty for estimates
D ü s s e l d o r f [ 7 4 4 . 2 ]
K ö l n [ 6 8 8 . 0 ]
A n d e r n a c h [ 6 1 3 . 8 ]
B o n n [ 6 5 4 . 8 ]
A h r [ 6 2 9 . 5 4 ]
S i e g [ 6 5 9 . 4 ]
W u p p e r [ 7 0 3 . 3 ]E r f t [ 7 3 6 . 5 5 ]
K ö l n - L a n g e l [ 6 7 1 . 1 ]
W o r r i n g e r B r u c h [ 7 0 9 . 5 . 1 ]
N o d e ( G a g e )
B r a n c h ( i n f l u e n c e o f G r o u n d w a t e r )
R e t e n t i o n A r e a
T r i b u t a r y
Flood forecasting system
• Rainfall-runoff Model (HBV)
• Water Transport Model
• Hydraulic Model (Sobek)
• Data-assimilation
actualmeasurements
Hydrologicalmodel
Hydro-meteodatabase
runoff prediction
Data-assimilation
Filtered water levels and flows
Data-assimilation
Filtered model parameters
Hydraulicmodel
Prediction of water levels and flows
Data-assimilation
Flood forecasting system
• Data-assimilation hydrological model
• Sensitivity and uncertainty analysis
– Adaptation soil moisture content
– Adaptation upper zone
• All sub basins treated equally
• Use adaptation factors in forecasting
Flood forecasting system
• Data-assimilation hydraulic model
• Sensitivity and uncertainty analysis
– Adaptation roughness main channel
– Adaptation lateral discharges
• Desired accuracy
• Until calculated water levels at Bonn and
Cologne “agree” with measurements
• Use adaptation factors in forecasting
Parameter Influence Uncertainty
Roughnessmain channel
Large Moderate
RoughnessBank section
Moderate Moderate
Roughnessfloodplain
Moderate Moderate
DischargeSieg
Moderate Large (?)
Groundwater Small Moderate/Large (?)
Data-
assimilation
ResultsSobek with and without data-assimilation (Köln)
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
23-Dec 24-Dec 25-Dec 26-Dec 27-Dec 28-Dec 29-Dec 30-Dec 31-Dec
wa
ter
lev
el d
iffe
ren
ce
s
40
41
42
43
44
45
46
difference measurement and Sobek difference measurement and Sobek assimilatedwater level measured
Conclusions data-assimilation in-situ data
• Large calculation time (10 minutes for a day)
• Relatively small changes parameters indicating:– well calibrated hydraulic model– robust data-assimilation algorithm
• Forecast pattern remains similar
• Average accuracy around 5 cm in water levels
Role of satellite data
• Use of satellite data in deducing water levels
• Additional information is to be used in data-assimilation of hydraulic model
• Satellite ‘measurements’ are, compared to in-situ measurements:– less accurate, but– more detailed
Possible role of satellite data
• No real flood maps based on EO-data available for Rhine river, Germany
• Synthetic flood maps, using hydraulic model and a digital terrain model
• Introducing inaccuracies (‘noise’) by modelling errors in:- geo referencing; and - classification
Conclusions using flood maps
• Results depend on quality of satellite data– high resolution– low noise
• Flood maps to water levels– Area’s instead of cross-sections– stretches long enough (5 – 10 km)– straight river sections– gentle slopes, no steeps banks
• Opportunity– comparison of flood extent calculated and
satellite data.
Conclusions FloodMan
• The flood forecasting system is robust and ready to serve under operational conditions;
• In the pilot small improvement in the flood forecast accuracy;
• Forecast including measure of uncertainty: useful for decision making.
• Use of satellite data is promising, especially for river systems with few gauging stations– BUT high resolution satellite data needed
Further work
• Flood forecast systems with data-assimilation on hydrological and hydraulic model are implemented
• Different data-assimilation algorithms
• Data-assimilation to combine rainfall radar data with in-situ measurements
• Use of satellite data to determine flood extent in case of dike breach for:– estimate width and depth of dike breach– estimate discharge at dike breach