Perception of uncertainty in water management by stakeholders and researchers
Prague, 14 -16 May 2007
NeWater Project
Comparison of different
approaches for flood damage
and risk assessment
Annegret H. Thieken1, Heiko Apel1, Heidi
Kreibich1, Guiseppe Aronica2
GeoForschungsZentrum
Potsdam, Germany
Risk definition
Source: Merz & Thieken (2004) ÖWAW.
Types of flood damage
damage
direct
tangible
intangible
property
damage
fatalities,
evacuees
indirect
tangible
disruption
of trade
and traffic
intangible
psychological
damage,
migration
damage to
residential
property
modified after Smith & Ward (1998): Floods.
Flood loss estimation: Input data
Land use and assets
(exposed values)
Relative loss model
(susceptibilty)
Example: Loss ratio at
residential buildings
Loss ratio of building L [-]
Inundation scenarios
0.6
L = 0.02 h
L = (2 h² + 2h)/100
L = (27 SQRT(h))/100
0.4
0.2
0.0
0
1
2
3
4
Water level h [m above ground surface]
Loss estimates [Euro]
5
Scales in flood loss estimation
Meso-scale estimation / data…
ƒ is based on land use units
(e.g. CORINE land cover),
ƒ provides no damage estimate
per object, only per region.
Micro-scale estimation / data…
ƒ is based on single objects/buildings,
ƒ demands very detailed input data
The model matrix
complexity
hazard
vulnerability
complexity
simple
damage
functions (I)
meso-scale
damage
models (II)
micro-scale
damage
models (III)
linear
interpolation
(A)
1D/2Dhydraulics
(B)
2Dhydraulics
(C)
Test area: Eilenburg (Saxony, Germany)
serious damage during flood of August 2002
ƒ inundation depths at about 400
buildings (Bauhaus-University Weimar)
ƒ mapped inundation extent (LfUG)
ƒ Repair costs (Saxonian Reconstruction
Bank)
⇒ model validation
Model description – hazard model A and B
A: Linear interpolation
Linear interpolation of absolute water levels between gauging stations;
derivation of inundation maps by cut and fill procedure of planar water
surface with DEM
Required input data:
DEM (25m resolution)
B: 1D/2D-hydraulics (LISFLOOD-FP; Bates & de Roo, 2000)
1D-hydraulics in channel, simple geometry, kinematic; 2D- hydraulics
in the floodplain, raster based, diffusion wave
Required input data:
DEM, channel information (cross sections)
Model description – hazard model C
C: 2D- hydraulics
(Aronica et al., 1998)
2D parabolic model (St.Venant without convective
acceleration) for channel and
floodplain, finite elements
Required input data:
DEM, channel information
(detailed), building inventory
Model description – vulnerability I: Stage- damage curves
Functional relation of inundation depth to loss ratio
0.8
Loss ratio of buildings d [-]
MURL (2000): d = 0.02 h
ICPR (2001): d = (2 h² + 2h)/100
0.6
Hydrotec (2001, 2002): d = (27 SQRT(h))/100
0.4
0.2
0.0
0
1
2
3
4
Inudation depth h [m above ground surface]
5
Flood Loss Estimation MOdel FLEMO
35
Data basis:
Interviews of 1697
households after the
flood in August 2002
Loss ratio of a building [%]
one-family house
30
(semi-)detached
multifamily house
high building
quality
25
poor/average
building quality
20
15
10
5
FLEMO+
0
< 21 cm
2. Modellstufe:
> 150 cm
Water level (above groundZu-/Abschläge
surface)
21-60 cm
61-100 cm
101-150 cm
Contamination
Source: Büchele, Kreibich et al. (2006) –
NHESS 6: 485-503.
Private precaution
none
good
very
good
none
0.92
0.64
0.41
moderate
1.20
0.86
0.71
severe
1.58
---
---
Vulnerability model II: FLEMO+ on the meso-scale
ƒ Calculation of a mean loss model per zip code or municipality
ƒ Consideration of a typical composition of building types and building
quality per zip code based on census data (INFAS Geodaten, 2001)
Typical composition of building types
Mean composition of residential buildings
Dominated by multifamily houses
Mixed: high share of multifamily homes
Mixed: high share of detached houses
Mixed: high share of one-family homes
Dominated by one-family homes
Source: Thieken et al. – submitted to J. Hydrol.
Average building quality
Example: Eilenburg
FLEMO
mean composition of residential buildings: Cluster 2
31 % one-family homes
25 % (semi-)detached houses
44 % multifamily houses
mean building quality: slightly below average
mean damage ratios DRmean in Eilenburg:
DRmean = 0.31 * DREFH + 0.25 * DRRDH + 0.44 * DRMFH
with:
DREFH: damage ratio for one-family homes; average building quality
DRRDH: damage ratio for semi-detached houses; average building quality
DRMFH: damage ratio for multifamily houses; average building quality
FLEMO+
No precaution, severe contamination: 1.58
Vulnerability model I and II for Eilenburg
Loss ratio of a building [-]
0.6
MURL
ICPR
0.5
Hydrotec
FLEMO
0.4
FLEMO+
0.3
0.2
0.1
0.0
0
1
2
3
Water level [m above ground surface]
Additionally required data:
Land cover data and associated building assets
4
5
Disaggregation of assets: Example of Dresden
Asset Database of Kleist et al. (2006), NHESS
EDIM
Vulnerability model III: FLEMO+ on the micro-scale
#1 mean loss function and mean property value per affected building
#2 building type specific loss functions and asset values
Loss ratio of a building [-]
0.4
0.3
0.2
FLEMO+ Eilenburg
One-Fam-Homes (adapted)
0.1
(semi-)detached (adapted)
Multi-Fam-Homes (adapted)
0.0
0
1
2
3
Water level [m above ground surface]
4
5
Required input data:
detailed building inventory (ALK) incl. type, use, quality and assets
Results: Inundation simulation
interpolation
Source: Apel et al.(2007) – IAHR
2D
1D/2D
Results: Inundation simulation – Model performance
performance
measured inundation depths
flood extent
bias [m]
mean
absolute
difference [m]
root mean
square error
[m]
flood area
index [%]
interpolation
0.28
0.60
0.83
96.43
1D/2D
-0.07
0.63
0.88
96.05
2D
-0.62
0.80
0.93
93.36
model
Source: Apel et al.(2007) – IAHR
Inundation simulation – Model performance (Bias)
Interpolation
h simulated [m]
8
6
4
2
0
0
0.5
1
1.5
2
2.5
3
2
2.5
3
2
2.5
3
1D/2D-model
h simulated [m]
8
6
4
2
0
0
0.5
1
1.5
2D-model
h simulated [m]
8
6
4
2
0
0
0.5
Source: Apel et al.(2007) – IAHR
1
1.5
h surveyed [m]
Loss estimation
Official repair costs: 77.12 Mio. Euro (SAB, 2005)
damage
models
inundation
models
Interpolation
1D/2D
2D
FLEMO+
meso
FLEMO+
micro #1
FLEMO+
micro #2
39.75 11.47 105.88 53.78
(-48%) (-85%) (37%) (-30%)
84.97
(10%)
76.59
(-1%)
109.02
(41%)
48.68
(-37%)
76.92
(0%)
67.50
(-12%)
94.67
(23%)
6.04
69.32 35.03
16.82
(-78%) (-92%) (-10%) (-55%)
55.35
(-28%)
46.47
(-40%)
67.42
(-13%)
ICPR
MURL
HYDRO
-TEC
95.03
34.50
9.78
(-55%) (-87%) (23%)
FLEMO
meso
Source: Apel et al.(2007) – IAHR
Assessment of loss estimates by resampling (bootstrap)
Input data
Repair costs of 765 buildings in Eilenburg (SAB, 2005)
Sum of repair costs: 77.12 Million Euro
Variation of total repair costs by resampling
Statistics of the sum of
10000 samples
Mean
77.11 Million Euro
2.5-percentile 72.00 Million Euro
97.5-percentile 83.39 Million Euro
Evaluation of loss estimates
Estimates that are within the 95% range of resampled sum:
damage
models
inundation
models
Interpolation
1D/2D
2D
FLEMO+
meso
FLEMO+
micro #1
FLEMO+
micro #2
39.75 11.47 105.88 53.78
(-48%) (-85%) (37%) (-30%)
84.97
(10%)
76.59
(-1%)
109.02
(41%)
48.68
(-37%)
76.92
(0%)
67.50
(-12%)
94.67
(23%)
6.04
69.32 35.03
16.82
(-78%) (-92%) (-10%) (-55%)
55.35
(-28%)
46.47
(-40%)
67.42
(-13%)
ICPR
MURL
HYDRO
-TEC
95.03
34.50
9.78
(-55%) (-87%) (23%)
FLEMO
meso
Relevance: Comparison of Average Annual Losses (AAL)
T = 10
Input:
T = 10a
Flood scenarios HQ10 … HQ1000
T = 50
T = 50a
LISFLOOD-FP (1D/2D)
Various loss models
T = 100
T = 100a
Estimates of AAL:
T = 500
T = 500a
Min. (MURL):
0.3 Million Euro
Max. (Hydrotec): 3.0 Million Euro
FLEMO+ (meso): 2.3 Million Euro
ICPR:
1.0 Million Euro
Conclusions
ƒ Differences between inundation models are comparatively small (in this
case).
ƒ Variability of results by loss models is considerably larger than by
inundation models.
ƒ The GFZ model FLEMO+ which includes additional factors
(contamination, preparedness) improves loss estimation remarkably.
ƒ Inundation simulation by linear interpolation works well in this case, but
fails in lowland areas as well as in mountainous terrain.
ƒ 1D/2D simulations are the best compromise between data and
simulation effort and required accuracy.
ƒ Hazard and vulnerability (loss) models need to be calibrated and
validated independently.
Acknowledgements
Funding:
More information:
German Ministry of Education and
Research (BMBF)
Dr. Annegret Thieken
e-mail: [email protected]
MEDIS
Methods for the Evaluation of
Direct and Indirect flood losseS
Additional funding:
Engineering Hydrology Section
Telegrafenberg
D-14473 Potsdam
References and further reading I
APEL, H., G. ARONICA, H. KREIBICH, A.H. THIEKEN: Evaluation of Different Modelling Strategies for
Flood Risk Assessment in Urban Areas. In: Proceedings of the 32nd Congress of IAHR
“Harmonizing the Demands of Art and Nature in Hydraulics”, Venice, Italv, 1-6 July 2007 (in press)
ARONICA, G., NASELLO C., TUCCIARELLI, T. (1998): 2D multilevel model for flood wave propagation
in flood-affected areas. Journal of Water Resources Planning and Management, 124(4): 210-217.
ARONICA, G., LANZA, L. (2005), Drainage efficiency in urban environment. Hydrological Processes, 19
(5), 1105-1119.
BATES, P.D., DE ROO, A.P.J. (2000): A simple raster-based model for flood inundation simulation.
Journal of Hydrology, 236(1-2): 54-77.
BÜCHELE, B., H. KREIBICH, A. KRON, A. THIEKEN, J. IHRINGER, P. OBERLE, B. MERZ, F. NESTMANN
(2006): Flood-risk mapping: contributions towards an enhanced assessment of extreme events and
associated risks. – NHESS 6(4): 485-503, (http://www.copernicus.org/EGU/ nhess/6/nhess-6-485.pdf).
HYDROTEC (2001): Hochwasser-Aktionsplan Angerbach. Teil I: Berichte und Anlagen. Studie im
Auftrag des StUA Düsseldorf, Aachen.
HYDROTEC (2001): Hochwasser-Aktionsplan Lippe. Teil I: Berichte und Anlagen. Studie im Auftrag
des StUA Lippstadt, Aachen.
ICPR (2001): Atlas on the risk of flooding and potential damage due to extreme floods of the Rhine.
International Commission for the Protection of the Rhine (ICPR).
KLEIST, L., A.H. THIEKEN, P. KÖHLER, M. MÜLLER, I. SEIFERT, D. BORST, U. WERNER (2006):
Estimation of the regional stock of residential buildings as a basis for comparative risk assessment
for Germany. – NHESS 6(4): 541-552 (http://www.copernicus.org/EGU/nhess/6/nhess-6-541.pdf).
KREIBICH, H., A.H. THIEKEN, TH. PETROW, M. MÜLLER, B. MERZ (2005): Flood loss reduction of
private households due to building retrofitting - Lessons learned from the Elbe floods in August
2002. – NHESS 5: 117-126.
References and further reading II
MERZ, B., A.H. THIEKEN (2005): Separating Natural and Epistemic Uncertainty in Flood Frequency
Analysis. - Journal of Hydrology 309: 114-132.
MERZ, B., A.H. THIEKEN (2004): Flood risk analysis: Concepts and challenges. - Österreichische
Wasser- und Abfallwirtschaft 56(3-4): 27-34.
MURL (2000): Potentielle Hochwasserschäden am Rhein in NRW - Ministerium für Umwelt.
Raumordnung und Landwirtschaft des Landes Nordrhein-Westfalen. Düsseldorf.
SMITH, K., WARD, R. (1998): Floods: Physical processes and human impacts. John Wiley and Sons,
Chichester.
THIEKEN, A.H., M. MÜLLER, L. KLEIST, I. SEIFERT, D. BORST, U. WERNER (2006): Regionalisation of
asset values for risk analyses. – NHESS 6(2): 167-178,
(http://www.copernicus.org/EGU/nhess/6/nhess-6-167.pdf)..
THIEKEN, A.H., M. MÜLLER, H. KREIBICH, B. MERZ (2005): Flood damage and influencing factors:
New insights from the August 2002 flood in Germany. – Water Resources Research 41(12), W12430.
THIEKEN, A.H., H. KREIBICH, T. NELTCHINOVA, B. MERZ: A new model for the estimation of flood
losses on the meso-scale. – Journal of Hydrology (submitted).