SEASONAL PREDICTION OF HURRICANE ACTIVITY REACHING THE COAST OF THE UNITED STATES

1
SEASONAL PREDICTION OF HURRICANE ACTIVITY
REACHING THE COAST OF THE UNITED STATES
Mark Saunders and Adam Lea Department of Space
and Climate Physics University College London
RMS Conference 2005 Exeter University 16
September 2005
2
Tropical Storm Risk (TSR)

• Founded in 2000, Tropical Storm Risk (TSR) offers
  a leading resource for forecasting the risk from
  tropical storms worldwide.
• The current TSR consortium comprises experts on
  insurance, risk management and seasonal climate
  forecasting.
• Industry partners Benfield, Royal
  SunAlliance, Crawford Company.
• Scientific partner UCL/Benfieldhrc.

3
US Hurricane Impacts
Hurricane Katrina 1100 UT 29 August 2005
(Courtesy NOAA)
Annual total (economic) and insured losses
1950-2004 from hurricanes striking the U.S. are
estimated to be US 5.6 bn and US 3.0 bn
respectively at 2004 prices and exposures.

4
Seasonal Forecast Relevance

• Strong correlation link (0.68 1900-2004) exists
  between U.S. hurricane activity and loss.
• Skillful seasonal forecasts of U.S. landfalling
  hurricane activity would benefit society,
  business and government by reducing risk,
  uncertainty and financial volatility.
• However, significant seasonal U.S. landfalling
  skill has not been shown to date.

5
What We Have Done

• We show that seasonal US landfalling hurricane
  wind energy 1950-2003 is predictable from the 1
  August start of the main Atlantic hurricane
  season with useful and significant (rrank 0.67
  p lt 0.001) skill.
• Predictability arises from tropospheric
  height-averaged wind anomalies present over the
  North Atlantic, North America and east Pacific
  during July which establishes persistent steering
  winds that either favour or hinder US hurricane
  landfall.
• Hindcasts are linked significantly (rrank 0.48
  p lt 0.01) to U.S. hurricane economic and insured
  losses 1950-2003.

6
Monthly Distribution of U.S. Hurricane Strikes
97 of intense (Cat 3 to 5) hurricane strikes on
the U.S. and 87 of hurricane hits on the U.S.
occur after 1 August.
(1950-2004 Data)
7
Data

• Use the NOAA Accumulated Cyclone Energy (ACE)
  index as our measure of landfalling hurricane
  wind energy and define this as the US ACE
  index.
• U.S. National Hurricane Centers North Atlantic
  hurricane database.
• Monthly gridded wind data averaged between 925
  and 400 mb from the NCEP/NCAR global reanalysis
  project 1950-2003.
• U.S. hurricane economic and insured loss data
  are obtained from Pielke and Landsea (1998)
  and from Collins and Lowe (2000) respectively.

8
Height-Averaged Winds and US
ACE Index

Figure shows height-averaged wind anomalies in
(a) July and (b) August-September-October linked
significantly to years of above-median US ACE
index 1950-2003.

9
Hindcast Skill for US ACE Index
Comparison of strength, significance and
stationarity in skill from three models.
Hindcast skill is significant to
p 0.001 over each of the sub-periods
1950-1976 and 1977-2003.

10
Hindcast Skill Comparison
Hindcast skill from the July wind model is double
that possible from knowing the observed North
Atlantic total ACE index at the hurricane season
end on 30th November.

11
Hindcast Link to Losses (1)
Link between hindcast US ACE index and U.S.
hurricane economic and insured losses 1950-2003.

12
Real-Time Performance

• The July 925-400mb wind index model performed
  well in real-time operation in 2004 and 2005.
• 2004
• US ACE Index forecast (4th August, 2004) to be in
  upper quartile.
• US hurricane losses estimated to be US 45 bn
  (economic) and US 23 bn (insured).
• 2005
• US ACE Index forecast (5th August, 2005) to be in
  upper decile.
• Insured losses from hurricane Katrina estimated
  to be US 40bn to US 60bn.

13
Business Application
US hurricane total insured loss contingent on the
TSR 1st August forecast. The chance of a large
total loss is much higher in those years when the
forecast is high.
14
Summary

• To our knowledge this is the first example of
  useful skill for predicting seasonal US
  landfalling hurricane activity and damage.
• The model has a sound physical basis.
• The model will benefit risk awareness and offers
  good potential for application in business
  decision making.

Further Information Saunders, M. A. and A. S.
Lea, Seasonal prediction of hurricane activity
reaching the coast of the United States, Nature,
434, 1005-1008, 2005.