flood risk models: reinsurers’ perspective dr. gerry lemcke deputy head catastrophe risk unit for...
TRANSCRIPT
Flood Risk Models: Reinsurers’ Perspective
Dr. Gerry LemckeDeputy Head Catastrophe Risk Unit for Americas, Swiss Re
Financing the risks of natural disasters World bank Headquarters, June 2-3, 2003
Flood Insurance - a ‘burning’ issue
recent flood events
State budget deficits
Climate Change: Prediction for the 21 Century:„more intensive precipitation are very likely over many areas” (IPCC, 3rd Report, 2001)
insurance industry holds a small stake in the coverage of flood damages in many European countries
More recent events
0 10 20 30
Bangladesh
China
Mississippi/USA
NW-Europe
Piemont/Italy
NW-Europe
China
Poland/Czechia
China
Tokai/Japan
Oratia/UK
Allison
Europe
Loss insuredTotal
bn USD
06/98
07/97
06/96
01/95
11/94
12/93
06/93
06/91
04/91
10/00
09/00
06/01
08/02
3 000
100
2 800
27
64
14
45
1 700
Death
18
16
33
38
Date
Hypothesis
Flood has been underestimated by the reinsurance industry as a risk for two major reasons:
The risk of flood is the most difficult one to assess on a large scale (country-wide, entire portfolios). Talking flood modelling one is talking detailed loss modelling (DLM).
The individual often knows whether or not his property is at risk with respect to flooding, i.e. whether or not to buy insurance. Consequently there is the risk of large scale anti-selection, which undermines one of the most important principles of insurability: having of a large community of individuals taking risk.
Example Loss, insured vs. uninsured or economic loss: Flood Event 8/2002
© Macon Data Professional World Set
€ 2-3bn€ 1bn
€ 10bn€ 1.8bn
€ 2-3bn€ 400mn
Example: anti selection or the Effect of a large risk community
Risk premium (weightedaverage), when all propertyin the relevant zone(s)is insured for the samepremium
Risk premium zone
Hazard: affected every 100-200 years100-300 y.
300-400 y.
400-500 y.
500-1000 y.
3.5 ‰ 1.6 ‰ 0.8‰0.5 ‰0.2 ‰ 0.05‰
3.5 ‰
2.8 ‰
2.4 ‰
2.2 ‰
1.9 ‰
0.2 ‰ Compulsory insurance
Distribution of property
Only exposed to torrential rainfall and/or backwater
source: Swiss Re, 1998
0
10
20
30
40
50
60
70
80
90
100
110
120
Storm
Europe
USD bn
EQCaliforni
a
300
estimates: Swiss Re
Economic viability
River Flood
Europe?*
* assuming full flood insurance penetration
adequate premium level
sufficient capacity for large losses
Uninsured part of thetotal economic loss
Direct insurance(excess of reinsurance)
Reinsurance
Direct insurance
How to come up with a rate, finally
Modern methods from science and geo-informatic facilitate quantification of both:
annual expected losses (Risk Premiums)
peak accumulation losses (EML, PML)
Zone 1
Zone 2
Flooded areas can range from a few square meters to hundreds of square kilometers
– risk assessment requires high spatial and temporal resolution
Sparse information on historical flood events
Unlike earthquakes and cyclones, flood risk is influenced by human activity
Challenge
Rationale
– covering the entire range of possible events according to scientifically derived extreme value distribution of hazard parameters
Hazard Modeling: Probabilistic Approach
Generate a probabilistic set of discharge regimes
Model flood wave propagation using a hydraulic model
Calculate flood footprints
00:00:001-1-2001
03:00:00 06:00:00 09:00:00 12:00:00 15:00:00 18:00:00 21:00:00 00:00:002-1-2001
03:00:00 06:00:00 09:00:00 12:00:00 15:00:00
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
15.0
16.0
17.0
18.0
19.0
20.0
21.0
[m^3/s] Time Series Discharge (extreme.res11)
Probabilistic Flood Hazard Modeling
Create a set of new events
Given the covariance properties of the historical events at gauged and interpolated (ungauged) stations
By means of Monte Carlo on a multivariate normal distribution, create new events (return periods) with the same covariance properties as the original events
Adjust time lags between the stations
calculate hydrographs depending on return periods and catchment characteristics
Probabilistic Flood Hazard Modeling
Fully hydro-dynamic modeling with MIKE11 (DHI)Input data:
00:00:001-1-2001
03:00:00 06:00:00 09:00:00 12:00:00 15:00:00 18:00:00 21:00:00 00:00:002-1-2001
03:00:00 06:00:00 09:00:00 12:00:00 15:00:00
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
15.0
16.0
17.0
18.0
19.0
20.0
21.0
[m^3/s] Time Series Discharge (extreme.res11)
river networkhydrographs
depending on return periods and catchment characteristics
Probabilistic Flood Hazard Modeling
Detailed DTM
– 50 m hor. resolution
– 0.1 m vert. resolution
216 gauging stations
– more than 25 years of data
– daily mean flows
– monthly maximum flows
1081 stream branches
2750 pour points
Example: Event Model UK
Spatial Correlation of RPs Accuracy of interpolation
Example: Event Model UK
Probabilistic Flood Modeling: the result
100-year flood zones along rivers
minimum data requirements
globally applicablehumid and semi-arid regions
Geomorphologic Regression
Flood Zonation: Geomorph approach
MARS (Multivariate Automated Regression Splines)
– nonlinear estimation of flood zones for a defined set of return periods (e.g. 50-100-250-500y)
Patent Pending
Flood Hazard Zoning
The model - assessment tools & private/public partnership
Insured
– takes preventive action to minimize damage
– participates in losses via significant self-retention
Insurance industry
– agrees to automatically provide flood cover
– exception: individual highly exposed objects
– charges risk-adjusted premiums
The model - assessment tools & private/public partnership
State
– raises risk awareness of the population
– considers flood risk for regional planning
– issues construction codes
– guarantees investment in flood control works
– allows insurance companies to build up loss reserves
– denies compensation in case of disaster
There is no technical obstacle to comprehensive flood cover:
– a large risk community can be created
– adequate accessibility is given
– economic viability can be guaranteed
… provided that all stakeholders assume their respective responsibilities.
Conclusion: Floods are insurable!