2 COPLEY PL ACE • BOSTON, M A 02116 • T: 617.423.2800 • F: 617.423.2808
The Great New England Hurricane of 1938
What a 100 Year Northeast Hurricane Will Do Today
K A R E N C L A R K & C O M P A N Y
September 16, 2013
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 2
The Great New England Hurricane of 1938 and What a 100 Year Northeast Hurricane Will Do Today
Seventy-five years ago, on September 21, 1938 a deadly and devastating hurricane came ashore near
Bellport, Long Island. Believed to be a Category 3 hurricane at landfall with sustained winds of 120
mph, the storm caused unprecedented destruction and casualties for the Northeastern US. Significant
wind damages were experienced throughout the region, and many coastal towns were completely
wiped out by storm surge heights exceeding 10 feet.
Because there was no advance warning, nearly 700 people died in this hurricane with an equal number
injured. Thousands of homes and other buildings were destroyed, and over 3,000 ships were sunk or
wrecked. The hurricane felled millions of trees throughout the region and destroyed entire forests. It
downed power lines and caused outages over most of New England.
When this type of storm occurs again, the insured damages and economic losses could be larger than
any storm seen to date because of the enormous concentrations of property values along the coasts of
New York, Connecticut, Rhode Island, and Massachusetts. KCC estimates that over $15 trillion of
property value would be in the storm’s footprint today. While better forecasting systems will likely
prevent the large loss of life, the right track could result in insured losses exceeding $100 billion and
economic losses more than double that number.
As we approach the seventy-fifth anniversary of the Great New England Hurricane of 1938 this report
examines what this storm would do today and more importantly, what a 100 year Characteristic Event
(CE) looks like for the Northeast region. What type of event should insurers be prepared for and what
could be worse than the 1938 storm? This report shows the primary determinant of likely losses in the
Northeast is the storm track, and insurers should not be lulled into a false sense of security by looking
only at what’s happened in the historical record.
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 3
A Historical Perspective – The Storm in 1938
The Great New England Hurricane of 1938 started out as a typical Cape Verde hurricane with a west-
northwesterly track that seemed destined for a Florida landfall. On September 19th the Jacksonville
office of the US Weather Bureau issued a warning that the storm might impact the state. But a deep
trough over the Appalachian region caused the hurricane to turn northward on September 20th and to
pick up speed on a more northerly track.
After the track change, the responsibility for the forecast shifted to the weather bureau’s headquarters
in Washington DC. Senior meteorologists in that office did not believe the storm would make landfall in
the Northeast but rather predicted it would veer out to sea as was typical for hurricanes. They also
forecast a weaker storm due to the colder ocean waters north of Cape Hatteras.
Only one young meteorologist named Charles Pierce was concerned about the track of the hurricane.
Observing the high pressure system over the continental US and a Bermuda High further north than
usual, Pierce predicted the storm would remain sandwiched between these two systems and take a path
due north into New England rather than turning to the northeast and out to sea.
Unfortunately, Pierce was overruled by the senior staff who had never seen a hurricane make landfall in
the Northeast. Consequently, no hurricane warnings were issued for New England, and the storm was
downgraded to a tropical storm. Only gale force warnings were issued for the region.
The great storm continued on its northerly track, picking up speed and intensity. When it made landfall
across central Long Island on the afternoon of the 21st, it was traveling at over 50 mph with a defined
eye of over 40 miles. Maximum winds at landfall were an estimated 120 mph. The hurricane sped
through the region moving due north until turning to the northwest and tracking into Canada.
Track of Great New England Hurricane
September 21, 1938 Weather Map
Source: intellicast.com
Source: Track Data from “Reanalysis of Ten U.S.
Landfalling Hurricanes”
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 4
With no warning of the imminent danger, coastal residents along the Long Island, Connecticut and
Rhode Island shores were taken by surprise when the skies darkened, the winds picked up, and the
rising waters started flooding their homes. Before they could even imagine what was happening, waves
and walls of water over 10 feet high came crashing across the shore destroying homes and well built
structures as if they were matchsticks.
Personal accounts of survivors can only hint at the horror of that afternoon. The images below reveal
the extent of the destruction.
Location: Westhampton Beach, NY
Source: Time
Location: Island Park, RI
Source: NOAA Photo Library
Location: Cranston, RI
Source: Rhode Island Historical Society
Location: Providence, RI
Source: Boston Globe
Location: New London, CT
Source: Rhode Island Historical Society
Location: Williston Park, Long Island
Source: New York Daily News
Location: Amherst, MA
Source: Library of Congress Collections
Location: Worcester, MA
Source: NOAA Photo Library
Location: Canton, MA
Source: Canton Historical Society
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 5
The damage toll was extensive:
Over one third of New England’s forests affected with over two billion board feet of trees felled
by the storm
Significant structural wind damage experienced as far inland as Worcester, MA
700 fatalities; 63,000 homeless
Tens of thousands of buildings destroyed
What a Similar Hurricane Would Do Today
While there is less actual data, historical records indicate several major hurricanes impacted the
Northeast region before 1900. Dr. Nicolas Coch from Queens College on Long Island has found evidence
of notable events in 1635, 1821 and 1893. His work is supported by the scientific papers of Donnelly
and Boose referenced at the end of this paper.
1938 Track and Wind Footprint
Source: RiskInsight®
While there are few reliable wind
speed measurements for the Great
New England Hurricane of 1938,
the wind footprint can be
reconstructed using the
meteorological data and analyses
conducted by various research
organizations, including the
National Hurricane Center (NHC)
and the Risk Prediction Initiative
(RPI). Using this information, KCC
experts reconstructed the wind
footprint. KCC estimates this storm
would cause insured losses in
excess of $35 billion today and
economic losses of more than twice
that amount.
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 6
Given the historical record, it’s reasonable to assume the Great New England Hurricane of 1938 is a 100
year type event for the Northeast. While a storm of this intensity, i.e. maximum sustained winds of 120
mph, has an estimated one percent probability of occurring, it could make landfall anywhere along the
Long Island, Rhode Island, or Massachusetts coastline.
The various landfall points will result in dramatically different industry losses and damages because
hurricanes are “right handed” in the northern hemisphere, with the strongest winds occurring from a
few miles to 50 miles to the right of the storm center. In the Northeast, hurricanes tend to be large with
an eye radius in excess of 20 miles. Hurricanes that make landfall further to the west will cause more
damage in the Northeast because more of the right side of the storm will be over highly populated
areas.
To account for the risk along the entire coast, KCC has developed the Characteristic Event (CE)
methodology of “floating” the 100 year CE and estimating the resulting losses. There are 22 landfall
points spaced at 10-mile increments in the Northeast region. The CEs have the return period wind
speeds and other characteristics, such as track, forward speed, and filling rate that are typical for the
region and selected return period.
The figures below show the wind footprints for three of the 100 year CEs.
Industry Loss
Landfall: Queens, NY
Landfall: Hampton Bays, NY
Landfall: Westport, MA
Distribution of Loss Distribution of Loss Distribution of Loss
NY
CT
RI
MA
NH
ME
Other
NY
CT
RI
MA
NH
ME
Other
NY
CT
RI
MA
NH
ME
Other
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 7
To estimate the industry losses for the floated CEs, KCC utilized a detailed proprietary database of
property values (KPD). The KPD includes estimates of building, contents, and time element values by
occupancy class and five-digit ZIP code. The chart below shows the industry losses from the 100 year CE
making landfall from western Long Island to Cape Cod.
With respect to the losses, the worst tracks are those in which the hurricane makes landfall across
western Long Island. The westernmost landfall points result in insured losses exceeding $100 billion.
Historically, no major hurricane has taken this track, but given the length of the historical record this
doesn’t mean it’s not possible or probable. Tools that rely on the historical record alone can
significantly underestimate the chances of this type of an event and potential Northeast hurricane
losses. Insurers should not make the same mistake as the senior forecasters who did not believe the
1938 storm was going to make landfall on Long Island because they hadn’t seen it happen before.
Two recent tropical cyclones, Irene
(2011) and Floyd (1999), made
landfall close to New York City
and had they been major hurricanes
would have caused losses much
greater than any historical Northeast
event. Irene’s track in particular would
have caused insured losses in excess
of $100 billion had it been the intensity
of the 1938 event.
150
100
50
100 Year CE Losses for the Industry by Landfall Point
Esti
mat
ed L
oss
($
Bill
ion
s)
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 8
Managing Hurricane Risk Using the CE Methodology
For the past 20 years, insurance companies have been using exceedence of loss probabilities, such as
one percent and 0.4 percent, to quantify and manage hurricane risk. One in 100 and one in 250 year
“probable maximum losses” (PMLs) are currently used by rating agencies and regulatory bodies. While
these numbers are interesting mathematically and are readily derived from the catastrophe model-
generated exceedence probability (EP) curves, they do not provide intuitive or robust risk metrics for
managing hurricane risk.
The PML approach can give a false sense of security by masking exposure concentrations that could lead
to solvency-impairing events. Because the model-generated EP curves are developed through
simulations of random tracks and other hurricane characteristics, exposure concentrations can be
missed entirely—that is, there may be no or insufficient events in geographical areas where a company
has significant exposure concentrations. (Alternatively, other areas can be over sampled.) That is one
reason why there have been so many surprise events and “black swans” over the past several years.
In recognition of the fact that PMLs don’t provide answers or complete information, companies can be
tempted to go back to the more straightforward approach of simply adding up limits to monitor
accumulations. While providing important information, this can lead to overly conservative strategies
and is not totally satisfying because a billion dollars of limit is a lot different in Florida versus the
Northeast, for example. Likewise, simple scenario-based approaches, such as Lloyd’s Realistic Disaster
Scenarios (RDS), do not capture exposure concentrations or probabilities effectively.
In light of these problems and risk management challenges, KCC introduced the Characteristic Event (CE)
loss estimation methodology. In the CE approach, the hazard probabilities are quantified and then the
losses are calculated for different return period events by landfall point. It’s the flip side of the EP curve
approach. The probabilities are based on the hazard versus the loss.
There are several advantages of the CE approach. While providing probability information, it also clearly
identifies exposure concentrations and “hot spots” and provides transparent and intuitive information
for decision makers, including boards and CEOs. The CEs stay the same from year to year, providing
consistent metrics for measuring and monitoring risk over time. CEs are operational risk metrics that
can be drilled down to individual policies for marginal impact analyses, pricing, and portfolio
management.
The 100 year CE loss chart is shown below for a hypothetical company. The landfall points are shown
along the x-axis, and the red bars show the losses from the 100 year events at each point. The black
horizontal line shows the company’s model-generated PML in a position that’s typical for most insurers.
The CE chart illustrates much richer and more complete information than the simplistic PML number. It
shows clearly where the company could have a loss well above the 100 year PML from the 100 year CE.
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 9
Because the CEs are stable from year to year, companies can use them to measure and monitor loss
potential over time and to develop more optimal portfolios of business while reducing the “spikes.”
Model-generated PMLs provide valuable information on potential losses, but they are less effective in
helping companies understand and manage the risk. Companies want more consistent risk metrics, full
transparency on the loss calculations and key drivers, and flexibility to customize assumptions to better
reflect their specific books of business. The CE approach delivers the solution to these risk management
challenges.
The CE approach is not limited to hurricanes, but can be applied to any peril in any region of the world.
KCC is collaborating with several scientific organizations to create CEs for European windstorms,
earthquakes, floods, and other hazards.
300
200
100
Esti
mat
ed L
oss
($
Mill
ion
s)
100 Year CE Losses by Landfall Point for Hypothetical Company
Model-generated PML
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 10
Validating the Wind Footprint for the 1938 Hurricane
The wind footprints shown in this document were developed using RiskInsight®, the advanced open
platform for catastrophe risk management. KCC experts devote considerable resources to developing
and validating the components of RiskInsight®. Because there were very few recordings of the wind
speeds for the 1938 Great New England Hurricane, scientists have studied this storm and inferred the
likely wind speeds from other data, including the damage. Dr. Emory Boose extensively studied the tree
damage for the 1938 event and converted his findings to the Fujita scale used for tornados. There is high
correlation between the observed damage pattern and the RiskInsight® footprint.
Fully transparent and highly credible wind footprints for all significant US hurricanes since 1900 are
available in RiskInsight®. Over 2,000 CE footprints are also available for all major return periods and
coastal regions. These footprints can be used to estimate losses for individual policies, portfolios of
policies, and the industry as a whole.
For more information contact: [email protected]
RiskInsight® is a registered trademark of Karen Clark & Company
Source: RiskInsight®
1938 Wind Footprint from RiskInsight®
1938 Wind Footprint from Dr. Boose
Source: Boose, E.R.
Copyright©2013 Karen Clark & Company, Inc. The Great New England Hurricane of 1938 11
References
Boose, Emery R. et al., 2001. Landscape and Regional Impacts of Hurricanes in New England. Ecological
Monographs 71:27–48.
Coch, Nicholas. “The Unique Vulnerability of the Northeast U.S. to Hurricane Damage.” Geological
Society of America. Abstract. National G.S.A. Philadelphia, 2006
Coch, N. K., 1994, “Hurricane hazards in the Northeast U.S.” Journal of Coastal Research. Spec Issue No.
12 - Coastal Hazards, Chapter 9 (pp. 115-147).
D’Aleo, Joe (2008). “The September Surprise – The Great New England Hurricane of ‘38”. Retrieved from
http://www.intellicast.com/Community/Content.aspx?a=135
Donnelly, Jeffrey et al, “700 Year Sedimentary Record of Intense Hurricane Landfalls in Southern New
England” Geological Society of America. June 2001.
Goudsouzian, Aram. The Hurricane of 1938.Commonwealth Editions, 2004.
Landsea, C., M. Dickinson, and D. Strahan, 2008: Reanalysis of Ten U.S. Landfalling Hurricanes. Final
report submitted to the Risk Prediction Initiative, 120pp.
Monthly Weather Review - September 1938. National Hurricane Center. Volume 66 Issue 9.
NOAA Hurricane Research Division of AOML (HURDAT). Retrieved from
http://www.aoml.noaa.gov/hrd/hurdat/