Flood Risk Assessment Models in Europe

1
Flood Risk Assessment Models in Europe
Dr. Jens Mehlhorn IUA/RAA Conference 20 June 2007
2
Agenda

• Flood Losses in Europe
• Types of Flood Risk Assessment Models
• Model Availability
• Model Comparison
• Conclusions

3
Large Flood Events in Europe
Losses in million EUR inflated to 2006
Source Swiss Re Sigma
4
Flood and wind losses in Europe
Total Loss in million EUR inflated to 2006
Number of Events
Observation Period 1992 - 2006
Source Swiss Re Sigma
5
Flood Zones versus Flood Event

• Flood Zones
• zones for definedreturn periods along all rivers
  in one country
• -gt Flood zone lookup for single location UW
• Flood Event
• areas affected during one and the same event
• spatial correlation of high water
• -gt probabilistic models for Risk Management incl.
  Reinsurance

Single Location
Portfolio
6
Flood Zone Model
7
Flood Zone Models in Europe

• many different tools
• mostly driven by Insurance Associations
• no consistent flood zone categories
• partly outdated

Environment Agency Several Proprietary Tools
(PIU)
ZUERS (GDV)
FRAT
HORA (ÖVV)
SIGRA (ANIA)
8
Probabilistic Fresh Water Flood Models in Europe

• many tools
• Consultants, Brokers, Reinsurers and Associations
• so far no cross-country /basin-wide
  /Europe-wide model

• RMS
• Swiss Re (ZUERSII)
• EQE CAT

• Benfield
• Guy Carpenter

• RMS
• Swiss Re
• AIR (planed)

• Benfield
• Swiss Re /MMC
• Guy Carpenter

• RMS
• Swiss Re

• EQE CAT
• HORA (ÖVV)
• Guy Carpenter

• SIGRA (ANIA)

9
Model Comparison

• Modellers A, B and C
• Countries
• United Kingdom
• Germany
• Belgium
• (Austria)
• Normalized Market Portfolios
• Residential, Commercial and Industry
• Building, Contents and Business Interruption
• Comparison
• between countries
• between modellers in one country
• between single LoBs

10
Comparison between countries

• agreement between A, B and C about relative
  exposure of markets GE gt BE gt UK
• modeller C is always more conservative than
  modeller A
• modeller A is always more conservative than
  modeller B

11
Germany Market Loss Curves

• all models are less than 2 years old
• shape of all curves are similar
• overall only small difference between A and B
• C more conservative
• small difference between all for low return
  periods (lt 50 years)
• event of 2002 had return period of about 50 years

12
Germany LoBs only for A and B

• shapes of all curves are similar
• A Res and Com almost identical, Ind more
  exposed
• B Res lt Com ltlt Ind
• A vs B market lower losses of B for Res and Com
  are compensated by more conservative curve for
  Ind

13
UK Market Loss Curves

• both models are one of the oldest probabilistic
  fresh water models
• overall A much more conservative than B
• however, high degree of consistency for return
  periods lt30 years
• Autumn 2000 event had return period of 20-30 years

14
Summary of model comparison

• probabilistic flood models are very complex and
  sources of uncertainty are larger than for other
  perils
• However, there seems to be some consistency
  between modellers which markets are more exposed
  then others
• recently released models in one market seem to
  converge
• closest match between models are achieved for low
  return periods
• however, different LoBs still deviate
  substantially
• local portfolios deviate also substantially (not
  shown)
• all modellers calibrate against the same market
  losses

15
Conclusions

• Flood losses in Europe are comparable to
  windstorm losses
• more events but less severe
• low flood cover penetration might change in
  future causing overproportional increase of
  losses
• Flood models are complex and expensive
• however, many flood risk assessment tools for
  primary insurance and reinsurance UW are now
  available
• Probabilistic models show some degree of
  consistency on a market level
• however, larger differences remain for regional
  portfolios and for different LoBs
• Multi-Model use helps to reduce uncertainty