THE WILLIS HURRICANE INDEX

1
THE WILLIS HURRICANE INDEX

• Brian Owens, Willis Re
• Greg Holland, NCAR

3rd June 2009
2nd International Summit on Hurricanes Climate
Change
2
Willis Research Network

• A partnership between Willis and over twenty
  academic research organisations around the world
• Developed a research programme working with
  clients and academics to tackle clients key risk
  issues
• Research organised along six themes including
  atmospheric sciences
• Fund postdoctoral research at cornerstone
  universities

3
Willis Hurricane Index (WHI)

• Index for the transfer of property insurance risk
  and settlement of related losses
• Initially set up to cover offshore energy
  facilities
• Universal application (landfall) but initial
  focus on Gulf of Mexico

4
Background

• Replaces a previous simpler index used by Willis
  that was based on the Saffir Simpson scale
• Hurricanes of 2004 2008, in particular
  Hurricane Ike, highlighted problems in this
  approach
• In particular basis risk, or lack of good
  correlation between the calculated index value
  and actual insured property loss
• Other indices from which to choose including the
  ACE, PDI, IKE, and the Carvill Hurricane Index
  (CHI)
• CHI designed to cover risks along all
  hurricane-exposed U.S. coasts, not just the Gulf
  of Mexico

5
Objective

• To develop a comprehensive hurricane damage index
  under the following guiding principles
• Use only information readily available in
  advisory reports from the relevant hurricane
  warning center
• Initial development for offshore facilities in
  part of the Gulf of Mexico bounded by the coast,
  north of 26N and west of 88W
• Useable world-wide

6
Data Method

• Data
• Willis Energy Loss Database (WELD) for the Gulf
  of Mexico (7 hurricanes)
• Inflated using CEPCI (Arnold and Chiltern 1963)
• Critical hurricane parameters
• The amount of energy dissipated at the surface by
  the maximum winds
• The radial extent and character of the surface
  wind field (nautical miles)
• The translational speed of the hurricane (kts)
• Combined these are a proxy for waves, currents
  and storm surge

7
Data Method

• Method
• Examine existing indices and parameters
• Base development largely on physical principles
  with care not to over specify the use of the
  small available data base
• Independent testing to come from comparison with
  existing catastrophe models and from application
  during 2009 hurricane season
• Combine the three critical factors in an additive
  formula

8
Parameter Contributions
Intensity0.23
Size0.48
Translation Speed0.71
9
Parameter Determination

• Intensity Term
• Set aa3 following Emanuel (2005) Power
  Dissipation Index
• Normalise by setting vm065 kt
• Constrained to storms vmgt65 kt
• Set a1 through experimentation and sensitivity
  analysis

10
Parameter Determination

• Size term
• Set bb1 for consistency with areal coverage of
  storm winds, given translation
• Normalise by setting Rh050 nautical miles
• Set b5 through experimentation and sensitivity
  analysis

11
Parameter Determination

• Translation term
• Set cc2 from regression on available data
• Normalise by setting vt015 kts
• Set a5 through experimentation and sensitivity
  analysis
• Also set vt lt25kt on physical grounds and gt7kt
  based on a posteriori checking

12
The WHI for Gulf of Mexico
13
Sensitivity Analysis

• There is no marked sensitivity to any one
  parameter

14
Application to Gulf Losses
Maximum WHI vs. Losses
R2 0.96
Average WHI vs. Losses
R2 0.91
15
WHI vs. Other Indices
16
Business Applications

• Use in purchasing or selling insurance cover for
  off-shore energy facilities in the Gulf of
  Mexico
• Due to complexities in energy insurance policies,
  very difficult to assess how much exposure you
  have at a point in time with reasonable certainty
• Index approach separates out exposure monitoring
  from loss settlement
• Insurance contracts are simple
• Quick and straightforward settlement
• Avoids insurance loss creep
• Potentially allows for other market participants
  to be involved

17
Climate Change Link

• Climate change of major interest to insurance and
  reinsurance industry
• Climate change signal in hurricane activity
• Impact on hurricane tracks, intensities and
  teleconnections
• Many of the hurricane loss modelling tools we use
  are developed based on historical event data
• May not be reflective of future climate and
  events
• Specifically for the WHI
• Limited data from 7 storms, all since 1992
• Increase in hurricane frequency and intensity
  since 1995
• Calibrated against catastrophe model event sets

18
Thank You
R2 0.96