BRIEFING

September 2008

CatastRophE MaNaGEMENt sERvICEs

pREpaRING FoR a NEw vIEw oF U.s. EaRthqUakE RIsk

T

The Uniform California Earthquake Rupture Forecast, Version 2 (UCERF 2)

By 2007 Working Group on California Earthquake Probabilities*

USGS Open File Report 2007-1437 CGS Special Report 203 SCEC Contribution #1138 Version 1.0

2008

2008 United States National Seismic Hazard Maps

ogether, the 2008 National Seismic Hazard Maps and the 2007 Uniform California Earthquake Rupture Forecast describe the current scientific view of seismic hazard in the United States. These studies are the product of a massive amount of research and scientific debate that has transpired over the past several years. This academic work will form the foundation of the catastrophe model updates that will be introduced by vendor modelers in 2009 and, ultimately, will have a significant impact on the risk modeled for property and workers compensation portfolios.

he U.S. Geological Surveys 64+ National Seismic Hazard 48-64 Maps are the basis for seismic 32-48 %g design provisions of building 16-32 codes, insurance rate structures, 8-16 4-8 earthquake loss studies, retrofit 0-4 priorities, and land-use planning. Incorporating these hazard Lowest hazard *Edward H. Field, Timothy E. Dawson, Karen R. Felzer, Arthur D. Frankel, Vipin Gupta, maps into designs of buildings, Thomas H. Jordan, Tom Parsons, Mark D. Petersen, Ross S. Stein, Ray J. Weldon II, and bridges, highways, and critiChris J. Wills cal infrastructure allows these structures to withstand earthquake shaking without collapse. U.S. Department of the Interior Properly engineered designs not U.S. Geological Survey only save lives, but also reduce disruption to critical activities California Department of Conservation following a damaging event. By California Geological Survey estimating the likely shaking for a given area, the maps also help engineers avoid costs from Colors on this map show the levels of horizontal shaking that have a 2-in-100 chance of being over-design for unlikely levels of exceeded in a 50-year period. Shaking is expressed as a percentage of g (g is the acceleration ground motion. of a falling object due to gravity).

Highest hazard

Changes to the Maps The Update Process

The U.S. Geological Survey recently updated the National Seismic Hazard Maps by incorporating new seismic, geologic, and geodetic information on earthquake rates and associated ground shaking. These 2008 maps supersede versions released in 1996 and 2002. Updating the maps involved interactions with hundreds of scientists and engineers at regional and topical workshops. USGS also solicited advice from working groups, expert panels, State geological surveys, Federal agencies, and hazard experts from industry and academia. The Pacific Earthquake Engineering Research Center developed new crustal ground-motion models; the Working Group on California Earthquake Probabilities revised the California earthquake rate model; the Western States Seismic Policy Council submitted recommendations for the Intermountain West; and three expert panels were assembled to provide advice on best available science. The most significant changes to the 2008 maps fall into two categories, as follows: 1. Changes to earthquake source and occurrence rate models: In California, the source model was updated to account for new scientific information on faults. For example, models for the southern San Andreas Fault System were modified to incorporate new geologic data. The source model was also modified to better match the historical rate of magnitude 6.5 to 7 earthquakes. The Cascadia Subduction Zone lying offshore of northern California, Oregon, and Washington was modeled using a distribution of large earthquakes between magnitude 8 and 9. Additional weight was given to the possibility for a catastrophic magnitude 9 earthquake that ruptures, on average, every 500 years from northern California to Washington, compared to a model that allows for smaller ruptures.

Over the coming months, Willis will be providing additional details about how the vendor earthquake models are changing to proactively prepare you for how model changes may affect your companys underwriting guidelines, capital requirements, and portfolio management strategies. This report is a first step in this process, and has been designed to be a point of reference you can refer to over this period, as more information becomes available from AIR, EQECAT and RMS about the changes they are implementing. We encourage you to contact your Willis representative if you would like to have a more in-depth discussion on what this new research could mean for your business.
U.S. Department of the Interior U.S. Geological Survey
Printed on recycled paper

Fact Sheet 20083018 April 2008

Three key themes have emerged from Willis research effort: 1. The greatest magnitude changes in seismic risk have occurred in California, with significant but lesser changes in the Pacific Northwest. 2. Measurements from recent large earthquakes around the world indicate that tall buildings in California may experience less shaking in a large earthquake than was previously assumed. 3. The vendor models will be fully recalibrated and therefore the seismic hazard changes presented in this report may be offset or amplified by changes to other modeling components.

Copyright 2008 Willis Re Inc. All rights reserved: No part of this publication may be reproduced, disseminated, distributed, stored in a retrieval system, transmitted or otherwise transferred in any form or by any means, whether electronic, mechanical, photocopying, recording, or otherwise, without the written permission of Willis Re Inc. Some information contained in this report may be compiled from third party sources and we do not guarantee and are not responsible for the accuracy of such information and expressly disclaim any liability based upon, arrising from or in connection with, any such information or the use or application of any such information or the results obtained from the use or application of any such information. This report is for general guidance only and is not intended to be relied upon. Any action based on or in connection with anything contained herein should be taken only after obtaining specific advice from independent professional advisors of your choice. The views expressed in this report are not necessarily those of Willis Re Inc., its parent companies, sister companies, subsidiaries or affiliates (hereinafter Willis). Willis is not responsible for the accuracy or completeness of the contents herein and expressly disclaims any responsibility or liability for the readers application of any of the contents herein to any analysis or other matter, or for any results or conclusions based upon, arising from or in connection with the contents herein, nor do the contents herein guarantee, and should not be construed to guarantee, any particular result or outcome. Willis accepts no responsibility for the content or quality of any third party websites to which we refer.

taBlE oF CoNtENts
Table of Contents.............................................................................................................................ii How to Get the Most From this Report ..................................................................................... iii - iv

Introduction .......................................................................................................................... page 1 The Next Generation Attenuation (NGA) Equations ............................................................ page 4 Significant Changes on the Horizon for California ............................................................... page 7 Earthquake Potential in Northern vs. Southern California ................................................... page 11 Cascadia Advancements Change the Risk Profile in the Pacific Northwest........................ page 14 Small and Varied Changes in the New Madrid Region........................................................ page 20 How Will the AIR, EQECAT and RMS Models Change? ..................................................... page 23 Glossary .............................................................................................................................. page 25 References .......................................................................................................................... page 26

Willis Re Inc.

ii

September 2008

how to GEt thE Most FRoM thIs REpoRt

This report provides Insurance Executives and Risk Managers with a summary of the 2008 National Seismic Hazard Maps, the 2007 Uniform California Earthquake Rupture Forecast, and other scientific documentation. This section is designed to help you efficiently navigate this report and to help you accurately interpret the maps that are presented throughout this document. There are three levels of depth within this report, identifiable by the color in the margin of each page: 1. 2. 3. Executive summary (GOLD) Selected details targeted at the Risk Manager (BLUE) In-depth content for technical staff (GREEN)

Executive = GOLD Margin Risk Manager = BLUE margin Technical = GREEN margin

The details targeted at the Risk Manager are organized by geographic region, to allow you to efficiently focus on the areas of greatest interest to you. Each regional section has been organized in the same manner to allow you to focus on a specific area in a region or contrast between different regions. You will also find symbols in the margins of the regional sections to distinguish these areas. Two descriptors of hazard, peak ground acceleration (PGA) and spectral acceleration (SA) are used throughout this report. Peak ground acceleration describes how abrupt the ground motion is. PGA is commonly measured in units of gravity (g; 9.8 ms-2). For example, when you push on the gas pedal in your car, you experience the increase in velocity as a force pushing you back into your seat. Acceleration is the rate of increase in velocity. That is, how much the velocity changes in a unit of time. So, consider a car increasing in speed from a stop to 60 miles an hour (88 feet per second) in 8 seconds. If the acceleration is uniform (constant) while the car increases speed, the velocity will have changed by 11 feet per second every second: the acceleration of the car is 11 feet per second, i.e., about 0.34 g. If the acceleration were not uniform, but started off small, achieved a maximum, and then decreased as the car approached 60 miles an hour, the maximum acceleration would equal the peak acceleration. Peak Ground Acceleration, as the name states, refers to the movement of the ground, not the movement felt by buildings. PGA is most relevant to property located at ground level, such as in-ground or surface pipelines and railroad tracks. The shaking experienced by a building is dependent on its height (which relates to its resonant frequency). Spectral Acceleration (SA) is used to distinguish the hazard experienced by buildings of differing heights. SA is expressed in units of g at different periods, such as 0.2 sec or 1.0 sec; however, it is more intuitive to translate these periods into building heights. As a rule of thumb, you can approximate the building height by multiplying the time period by 10 0.2 sec period 2 stories and 1.0 sec period 10 stories. The vendor models calculate the earthquake hazard in terms of spectral acceleration at different time periods, to determine the building loss ratio. Therefore, most of the maps in this report are for 0.2 sec (2 story) and 1.0 sec (10 story) spectral accelerations. We provide these two points of reference to show the variation in the changes in modeled seismic hazard for different building types. The schematic below illustrates that the maps shown in this report are for a hypothetical, uniform distribution of buildings. In reality, high-rise buildings will be concentrated in city centers, business parks, and other commercial areas. Therefore, the actual changes in seismic hazard experienced by the industry will be a blend of the 0.2 sec and 1.0 sec maps, as well as other frequencies that are not presented in this report.

Willis Re Inc.

iii

September 2008

GUIdE (CoNtINUEd)

The National Seismic Hazard Maps form the basis of building code provisions. When applying the code, the design engineer evaluates the site conditions to assess the potential amplifying effect of the local soils, as these maps assume uniform soil conditions. The vendor models also include the local soil conditions when estimating the hazard at a given site; therefore, the changes that occur in the vendor models may not directly follow these maps.

Two hazard return periods are consistently presented in this report, 475 years and 2,475 years. These two years are unique points calculated by the USGS in the development of the National Seismic Hazard Maps. The 475-year return period is more commonly expressed as a 10% exceeding probability in 50 years and the 2,475-year return period is more commonly expressed as a 2% exceeding probability in 50 years. The 475-year return period maps can be used to gain insight into changes that might occur to modeled return periods in the 250 to 500 year range. The 2,475-year return period maps can be used to gain insight into changes that might occur to metrics that are heavily weighted on the tail of the curve. The 2,475-year return period is also commonly referenced in the building code.

Willis Re Inc.

iv

September 2008

INtRodUCtIoN
The United States Geological Survey (USGS) released the latest version of its National Seismic Hazard Maps in April 2008. The maps, which were last updated in 2002, incorporate the best available science on fault slip rates, paleoseismic data, earthquake catalogs, and strong motion recordings from global earthquakes. These maps define the latest scientific view of earthquake hazard at varying probability levels across the United States and will be the impetus for the 2009 model updates from AIR, EQECAT and RMS.
PGA at the 475-yr Return Period (10% exceeding probability in 50 years) PGA at the 2,475-yr Return Period (2% exceeding probability in 50 years)

California Pacific Northwest

The maps above show the national pattern of earthquake risk in terms of peak ground acceleration (PGA). These PGA values have also been related to the Modified Mercalli Intensity (MMI) scale in the legend to provide a physical understanding of the PGA values. Changes in Seismic Hazard Between the 2008 and 2002 Maps Building damage is negligible by Region and Building Type below MMI V (green and white colors on the maps). A full SA of 0.2 sec (2-Story Building) SA of 1.0 sec (10-Story Building) description of the MMI scale is 475-year 2,475-year 475-year 2,475-year provided in the glossary at the Return Period Return Period Return Period Return Period end of this report. Moderate to small Moderate Large to moderate Large to moderate
decreases -15% to 0% changes -20% to +15% Moderate changes -15% to +20% Moderate changes -20% to +15% Moderate to small changes -25% to +5% Moderate to small decreases -25% to -10% Moderate to small decreases -15% to 0% decreases -35% to -15% decreases -35% to -15% Moderate to small changes -15% to +5% Moderate changes -25% to +15% Moderate to small decreases -25% to -5% Moderate to small decreases -25% to -5% Moderate to small decreases -20% to -5% Moderate to small decreases -25% to 0% Large to moderate decreases -35% to -15% Moderate to small decreases -15% to 0% Moderate to small decreases -15% to -5% Moderate to small decreases -15% to -5%  Moderate changes -15% to +20% Large to moderate decreases -35% to -15% Moderate changes -15% to +15% Moderate to small decreases -15% to -5% Small decreases -10% to -5%

Intermountain West New Madrid Northeast South Carolina

Willis Re Inc.

The areas of greatest earthquake risk remain along the West Coast, (particularly California) and in the New Madrid region (at the intersection of Missouri, Illinois, Kentucky, Tennessee, and Arkansas). Other areas with notable hazard include Salt Lake City, Utah; Charleston, South Carolina; and portions of New England.
September 2008

INtRodUCtIoN (CoNtINUEd)

The maps below show the ratio of the 2008 and 2002 USGS hazard maps for the 0.2sec spectral accelerations. Warm colors refer to increases in hazard and cool colors reference decreases in hazard. As described in the previous section, 0.2sec SA is a measure of the shaking that would be experienced Change in Hazard for 0.2 sec SA at the 475 yr Return Period by a 2-story Change in Hazard for 0.2 sec SA at the 475 yr Return Period building.
Change in Hazard for 0.2sec SA at the 475-yr Return Period
(10% exceeding probability in 50 years) (10% exceeding probability in 50 years)
Focus on the Focus on the West Coast West Coast

(10% exceeding probability in 50 years)

Sources of Change in the Western U.S. Sources of Change in the Western U.S. - Introduction of Next Generation Attenuation (NGA) - Introduction of Next Generation Attenuation (NGA) Equations Equations - Updated fault parameters -- Updated fault parameters fault rupture scenarios Addition of multi-segment -- Addition ofin occurrence rates inrupture scenarios Reduction multi-segment fault California for M6.0-7.0 - Reduction ingridded seismicity California for M6.0-7.0 events for occurrence rates in events of gridded seismicity - Additionfor a new deep seismic zone near Portland - Addition of a new deep seismic zone near Portland

Sources of Change in the Central and Eastern U.S. Sources of Change in the Central and Eastern U.S. - Updated attenuation equations - Updated attenuation equations - Updated fault parameters and fault geometry - Updated fault parameters and fault geometry

% Change in Hazard Between the 2002 and 2008 Hazard, (+) increase / (-) decrease

Change in Hazard for 0.2 sec SA at the 2,475 yr Return Period Change in Hazard for 0.2 sec SA at the 2,475 yr Return Period (2% exceeding probability in 50 years) (2% exceeding probability in 50 years)

Change in Hazard for 0.2sec SA at the 2,475-yr Return Period

Focus on the Focus on the West Coast West Coast

(2% exceeding probability in 50 years)

Sources of Change in the Western U.S. Sources of Change in the Western U.S. - Introduction of Next Generation Attenuation (NGA) - Introduction of Next Generation Attenuation (NGA) Equations Equations - Updated fault parameters and fault geometry -- Updated to magnitude-frequency distribution on the Updates fault parameters and fault geometry - Updates to Subduction Zone (affects Pacific Northwest) Cascadia magnitude-frequency distribution on the Cascadia a new deep seismic zone near Portland - Addition ofSubduction Zone (affects Pacific Northwest) - Addition of a new deep seismic zone near Portland

Sources of Change in the Central and Eastern U.S. Sources of Change in the Central and Eastern U.S. - Change in recurrence interval for north arm of the - Change in recurrence interval for north arm of the New Madrid fault New Madrid fault - Modifications to New Madrid fault geometry - Modifications to NewCharleston seismic zone - Revised geometry of Madrid fault geometry - Revised geometry of Charleston seismic zone - Updates to attenuation equations - Updates to attenuation equations

Shaking approximately 10% in much of Central and Eastern United States at both the 475 and the 2,475-year return periods. In the Western U.S., % Change in Hazard Between the 2002 and 2008 Hazard Maps shaking experienced by a Hazard Between the 2002 and hasHazard Maps % Change in 2-story building 2008 decreased by 10% in many areas at the 475-year return (+) increase / (-) decrease (+) increase / (-) decrease period, with greater variation and complexity to the pattern of change at the 2,475-year return period.

25 5 25 15 35 5 25 35 50 o -35 o -25 o -15 o -25 -15 0 5 o 15 5 to 25 5 to 35 5 to 50 50 50 0 t - 5 t - 5 t - 5 t - 5 to -1 5 to 0 o <--50 -50 to -35 to -25 to -15 to -5 to -5 to 0 tto 5 5 tto 1 15 to 25 to 35 to <50 0 < -1 -3 -5 3 2 experienced by a2 2-story building 5has1 decreased<by

Willis Re Inc.

September 2008

INtRodUCtIoN (CoNtINUEd)

The maps below show the ratio of the 2008 and 2002 hazard maps for the 1.0sec spectral acceleration. As on the previous page, warm colors refer to increases in hazard and cool colors reference decreases Change in Hazard measure of the shaking that would be experienced by a 10-story building. in hazard. A 1.0sec SA is afor 1.0 sec SA atthe 475 yr Return Period
Change in Hazard for 1.0 sec SA at thein 50 yr Return Period (10% exceeding probability 475 years) (10% exceeding probability in 50 years)
Focus on the West Coast Focus on the West Coast

Change in Hazard for 1.0sec SA at the 475-yr Return Period

(10% exceeding probability in 50 years)

Sources of Change in the Western U.S. Sources of Change in the Western U.S. - Updated fault parameters and fault geometry
-- Updated attenuation equations Updated fault parameters and fault geometry - Updated attenuation equations

Sources of Change in the Central and Eastern U.S. Sources of Change in the Centralfault geometry - Updated fault parameters and and Eastern U.S. --Updated attenuation equationsfault geometry Updated fault parameters and - Updated attenuation equations

% Change in Hazard Between the 2002 and 2008 Hazard, (+) increase / (-) decrease

Focus on the West Coast Focus on the West Coast

Change in Hazard for 1.0 sec SA at the 2,475 yr Return Period Change in Hazard for 1.0 sec SA at thein 50 years) (2% exceeding probability 2,475 yr Return Period (2% exceeding probability in 50 years)

Change in Hazard for 1.0sec SA at the 2,475-yr Return Period

(2% exceeding probability in 50 years)

Sources of Change in the Western U.S.

Sources of Change in the Western U.S. - Introduction of Next Generation Attenuation (NGA) Equations - Introduction of Next Generation Attenuation (NGA) - Updated fault parameters and fault geometry Equations -- Updates to magnitude-frequency distribution on the Updated fault parameters and fault geometry Cascadia magnitude-frequency distribution on the - Updates toSubduction Zone (affects Pacific Northwest) - Addition of Subduction Zone (affects near Portland Cascadia a new deep seismic zone Pacific Northwest)
- Addition of a new deep seismic zone near Portland

Sources of Change in the Central and Eastern U.S.

Sources ofin recurrence interval for north arm U.S. - Change Change in the Central and Eastern of the New Madrid fault - Change in recurrence interval for north arm of the - Modifications to New Madrid fault geometry New Madrid fault --Updates to attenuation equations geometry Modifications to New Madrid fault - Updates to attenuation equations

Shaking approximately 35% in most of the Western U.S. at the 475-year and the 2,475-year return periods, with the exception of Oregon and % Change in Hazard Between the 2002 and in Hazard Maps southern Washington. The major changes 2008 the hazard for this building type in the Western U.S. (+) increase / (-) decrease % Change in Hazard Between the 2002 and 2008 Hazard Maps are mainly due to the introduction of decrease (+) increase / (-) the Next Generation Attenuation (NGA) equations for crustal sources and revised attenuation equations for the Cascadia Subduction Zone. Seismic hazard in Central and Eastern U.S. for a 10-story building has decreased by about 10%.

5 5 15 5 5 0 5 -2 15 -2 -3 5 15 o 2 o 3 o 5 to 5 to 5 to 5 to to - o 0 0 5 0 5 -5 -50 --35 --25 --15 -25 -1-5 t 0 0 to 5 to 515 t 225 t 335 t 5<50 5 5 1 3 2 < 0 5 1 to to 5 to 5 to 5 to - to o o t t to to 5 50 experienced -by a 10-story buildingtohas decreased0by 3 -5 -5 0 <-1 -2 5 15 25 35 <5

Willis Re Inc.

September 2008

thE NExt GENERatIoN attENUatIoN (NGa) EqUatIoNs

INTRODUCTION Of ATTENUATION EqUATIONS
Attenuation equations predict how ground motion decays with increasing distance from an earthquakes epicenter. Attenuation equations are used to determine the size of the earthquake footprint. Earthquake models utilize a suite of different attenuation equations. For example, separate equations are used to describe how the energy that affects 2-story buildings (0.2sec SA) decays as compared to the energy that affects 10-story buildings (1.0sec SA). Some buildings will experience damage at far distances from the epicenter, while other buildings may not experience any damage. The graphs below illustrate how attenuation equations can differ by earthquake magnitude and time period (i.e., building height). In this example, which is representative of what might happen in a California earthquake, the energy decays rapidly in the first 20 km (12.4 miles) from the epicenter.
0.2sec Spectral Acceleration (Strike slip faulting, uniform soft rock site)
1 .2 1 .0
Spectral Acceleration (g)

1.0sec Spectral Acceleration (Strike slip faulting, uniform soft rock site)
0 .6

M w=6 .5
Spectral Acceleration (g) Special Acceleration (g)

M w=6 .5 0 .4 M w=7 .5

Special Acceleration (g)

0 .8 0 .6 0 .4 0 .2 0 .0 0 20

M w=7 .5

0 .2

40 60 D ista nce to S ite (km )

80

100

0 .0 0 20 40 60 D ista nce to S ite (km ) 80 100

Campbell & Bozorgnia 2006, Next Generation Attenuation Equation

Attenuation equations also vary based on the fault type, the fault rupture characteristics, and the ground-motion modifications that occur along the path between the source and the site (e.g., soil type). For example, different equations are used to define the relatively small earthquake footprints from the San Andreas Fault, compared to the relatively large earthquake footprints from the New Madrid Seismic Zone.

KEy CHANGES TO THE ATTENUATION EqUATIONS USED IN THE 2008 NATIONAL SEISMIC HAzARD MAPS
1. Implementation of Next Generation Attenuation (NGA) equations for crustal faults (e.g., California) and modeling of additional ground-motion epistemic uncertainty in the NGA equations 2. Revisions to the subduction zone attenuation equations (the Cascadia subduction zone in the Pacific Northwest) . Updates for the Central U.S. attenuation equations by adding new published attenuation equations

Willis Re Inc.

September 2008

attENUatIoN (CoNtINUEd)

NExT GENERATION ATTENUATION (NGA) EqUATIONS

The implementation of the Next Generation Attenuation (NGA) equations is the most significant change that has occurred to the seismic hazard maps. The NGA research project was initiated by the Pacific Earthquake Engineering Research (PEER) center. The project had two phases: the development of a global strong motion database, followed by the development of crustal ground motion prediction equations based on this database. A global database was required to understand the behavior of strong earthquakes because there is insufficient data of strong earthquakes in the U.S. (due to their low frequency). In total, the database includes records of 173 earthquakes containing 3,500 multi-component strong-motion recordings with over 100 parameters describing earthquake source, path, site Change in Attenuation of 0.2sec SA conditions and other parameters. Similar to Shaking for a 2-Story Building In the second phase of the project, five different modelers/ research groups were identified to interpret the global strong motion database and develop independent sets of attenuation equations. The NGA modelers applied their own selection criteria for using earthquake ground motion records from the global strong-motion database. The modelers were not completely isolated; they interacted extensively with the each other and with the broader research community in developing the models. Despite this interaction, the modelers developed several different ground motion attenuation models, which is reflective of the (epistemic) uncertainty associated with this modeling parameter. Each researcher published their work to document the judgements they made in the development of their attenuation equations. Based on the recommendation by an expert panel, the USGS used three of the five NGA attenuation equations to calculate the ground motion from crustal earthquake sources in the western United States: Boore & Atkinson (2006), Campbell & Bozorgnia (2006), and Chiou & Youngs (2006). Ground motions were calculated for each of the three attenuation relations separately, and then combined using a weighted logic tree approach. The USGS assigned equal weights to each of the three sets of NGA equations in the logic tree as recommended by the expert panel.

(Strike slip faulting, uniform soft rock site)

40% M w=6 .5 (NG A / O ld ) M w=7 .5 (NG A / O ld )

0% 0 20 40 60 D ista nce to S ite (km ) 80 100

-4 0 %

Change in Attenuation of 1.0sec SA Similar to Shaking for a 10-Story Building (Strike slip faulting, uniform soft rock site)
40% M w=6 .5 (NG A /O ld ) M w=7 .5 (NG A /O ld ) 0% 0 20 40 60 D ista nce to S ite (km ) 80 100

The new NGA equations are significantly different from previous equations (especially for tall buildings). One of the three sets of equations used in the USGS seismic hazard maps is used in the charts on the left to illustrate -8 0 % the difference between the new NGA equations and the old attenuation equations. Negative values indicate a decrease Comparison of one of the three NGA equations used in predicting the ground shaking in in the shaking felt by buildings at the distance from the the new maps with the older version of the attenuation equation. (Campbell & Bozorgnia earthquake epicenter shown on the horizontal axis. These 2003 vs. Campbell & Bozorgnia 2006) charts show that there have been large decreases in the shaking experienced by 10-story buildings for M6.5 and M7.5 events at all distances away from the earthquake epicenter. This is the reason for significant decreases in the seismic hazard experienced by tall buildings. The new NGA equations have caused a 20-30% decrease in the shaking experienced by 2-story buildings at distances greater than 10 km (6 miles) for M7.5 event.
-4 0 %
Willis Re Inc.

September 2008

attENUatIoN (CoNtINUEd)

To further illustrate the changes associated with the NGA equations, the following maps contrast a M7.0 earthquake scenario on the southern San Andreas Fault as predicted by the new and old attenuation equations. The southern San Andreas Fault is the most likely fault to generate a strong earthquake in California in the next 30 years. This M7.0 scenario is representative of an earthquake that is in-between the two magnitudes shown in the charts on the preceding page.
0.2sec Spectral Acceleration (2-story building) New NGA equation 1.0sec Spectral Acceleration (10-story building) New NGA equation

Old equation

Old equation

Comparison of Campbell & Bozorgnia 2003 attenuation equation with Campbell & Bozorgnia 2006, NGA. M7.0, strike slip faulting, soft rock site conditions.

ATTENUATION EqUATIONS fOR OTHER REGIONS

USGS use three ground motion equations to predict the shaking from Cascadia Subduction Zone earthquakes, using a logic tree approach. In the 2008 National Seismic Hazard Maps, a new equation (Zhao et al.,2006) has replaced one of the three equations used previously (Sadigh et al., 1997). In addition, the weight on each branch of the logic tree has changed. Attenuation equations for intermediate depth earthquakes (depth >40 km) and intraslab earthquakes have been updated by replacing or adding one or more equations. The new maps also include several new simulation-based attenuation relations that were not available in 2002 to predict ground shaking from earthquakes in the central U.S.
Willis Re Inc.

September 2008

sIGNIFICaNt ChaNGEs oN thE hoRIzoN FoR CalIFoRNIa

tIME-INdEpENdENt vIEw

KEy IMPLICATIONS fOR CATASTROPHE RISK MANAGERS

The amount of USGS modeled shaking experienced by high-rise (>8 story) buildings in large earthquakes has decreased at all distances from an earthquakes epicenter. If the vendor modelers fully adopt the NGA equations, there could be significant reductions (20% or more) in modeled risk associated with high-rise properties. The modeled risk from large earthquakes to low-rise properties in the immediate vicinity of a fault (i.e., within 5 miles) may increase by 5 to 10%. Business rules that are based on the distance to a fault, such as exposure aggregate thresholds, underwriting guidelines or insurance rates, will need to be reviewed.

SUMMARy Of CHANGES IN THE USGS SHAKE HAzARD ExPERIENCED By LOW AND HIGH RISE BUILDINGS
SA of 0.2 sec (2-Story Building) 475-year Return Period 2,475-year Return Period Moderate changes - 20% to +15% Small changes - 5% to +5% Moderate changes -15% to +15% SA of 1.0 sec (10-Story Building) 475-year Return Period Large to moderate decreases -35% to -15% Large to moderate decreases -35% to -15% Large to moderate decreases -35% to -15% 2,475-year Return Period Large to moderate decreases -35% to -15% Large to moderate decreases -35% to -15% Large to moderate decreases -35% to -15%

California (Statewide) San Francisco Bay Area Los Angeles Area

Moderate to small decreases -15% to 0% Moderate to small decreases -15% to 0% Moderate to small decreases -20% to -0%

The primary reason for the large decreases in the modeled hazard is the implementation of the Next Generation Attenuation (NGA) equations. The change to the attenuation methodology overshadows the smaller effects from changes to earthquake source parameters and their magnitude-frequency relationships.

GAINING INSIGHT INTO POTENTIAL MODELED LOSS CHANGES

Based on the shape of a building damage function for earthquakes, the amount of damage a building incurs rapidly decreases as the ground motion attenuates from the fault (all other components remaining constant). Therefore, as an example, a 20% decrease in hazard can equal a 30-50% decrease in expected damage. This means that modeled damage for 2-story and 10-story buildings may decrease by a much larger amount than the change in modeled hazard shown in the maps on the following two pages. The exception to this rule is the immediate vicinity of faults (which have been shaded in gray), where marginal changes in hazard have little effect on modeled damages. At this point, we can not predict how the commercial model losses will change. However, USGS modeled hazard decreases for high-rise buildings is so substantial, significant decreases in modeled losses are likely to occur if the vendors fully adopt the NGA equations. The change in the amount of shaking experienced by low-rise (e.g., 2-story) and high-rise (e.g., 10story) buildings at the 475-year and 2,475-year return periods are shown in the following pages. Cool colors indicate decreases in hazard, while warm colors indicate increases in hazard. Only those areas where hazard is significant enough to result in damage at these return periods are shown.
Willis Re Inc.

September 2008

CalIFoRNIa REGIoN (CoNtINUEd)

tIME-INdEpENdENt vIEw

CHANGE IN USGS HAzARD fOR 2-STORy BUILDINGS

475-yr Hazard Return Period (10% exceeding probability in 50 years)

% Change in Hazard Between the 2002 and 2008 Hazard, (+) increase / (-) decrease

2,475-yr Hazard Return Period (2% exceeding probability in 50 years)

The gray shaded areas along many of the fault traces designate areas where changes in modeled damage could be lower than the changes in modeled hazard presented in these maps. The shading is based on a representative 2-story wood frame building damage function. Outside the shaded areas, it is possible for the change in modeled damage to exceed the changes in modeled hazard.
Willis Re Inc.

September 2008

CalIFoRNIa REGIoN (CoNtINUEd)

tIME-INdEpENdENt vIEw

CHANGE IN USGS HAzARD fOR 10-STORy BUILDINGS

475-yr Hazard Return Period (10% exceeding probability in 50 years)

% Change in Hazard Between the 2002 and 2008 Hazard, (+) increase / (-) decrease

2,475-yr Hazard Return Period (2% exceeding probability in 50 years)

The gray shaded areas along many of the fault traces designate areas where changes in modeled damage could be lower than the changes in modeled hazard presented in these maps. The shading is based on a representative 10-story steel frame building damage function. Outside the shaded areas, it is possible for the change in modeled damage to exceed the changes in modeled hazard.
Willis Re Inc.

September 2008

CalIFoRNIa REGIoN (CoNtINUEd)

tIME-INdEpENdENt vIEw

LIST Of SCIENTIfIC ADvANCEMENTS DRIvING THE USGS CHANGES IN CALIfORNIA

The USGS incorporated many new methods in its 2008 National Seismic Hazard Maps, in addition to the Next Generation Attenuation (NGA) equations. The key technical updates are listed below. 1. Reduced the moment rate on faults by 10% to account for aftershocks, foreshocks, after slip, and smaller earthquakes (influences the results for all locations in California). 2. Implemented four new recurrence models for Type-A faults (well understood active faults, e.g., San Andreas, San Jacinto, Elsinore etc.) in Southern California, based on work by the Working Group on California Earthquake Probabilities (influences the results for all locations in Southern California). 3. Revised slip rates for sections of the San Andreas fault, San Jacinto fault, and nine other Type-B faults (i.e., faults with observed slip rates and with known other characteristics of faulting, e.g., Owens valley, Imperial valley, Northridge fault etc.) (influences the results for all locations in California). 4. Developed new multi-segment ruptures for several California Type-B faults (influences the results for all locations in California). 5. Implemented the Southern California Earthquake Center (SCEC) Community Fault Model (CFM) for the geometry of faults in Southern California (influences the results for all locations in Southern California). 6. Reduced the rate of M6.5 and larger earthquakes by 1/3rd in smoothed gridded seismicity, since this seismicity is now accounted directly on faults (influences the results for all locations in California).

Willis Re Inc.

0

September 2008

EaRthqUakE potENtIal IN NoRthERN vs. soUthERN CalIFoRNIa

PrevioustIME-dEpENdENt s
vIEw

WORKING GROUP ON CALIfORNIA EARTHqUAKE PROBABILITIES (WGCEP)

The California section of National Seismic Hazard Maps, which were described in the prior section, is based on a time-independent forecast. In time-independent forecast, the probability of each earthquake rupture is completely independent of the timing of all others. It is generally accepted, however, that a time dependent model provides a more accurate representation of the risk in California, where most faults have been well studied. Time-dependent models are based on the concept of stress renewal; the probability of a fault rupture decreases immediately after a large earthquake releases tectonic stress on the fault and rises again as the stress is regenerated by continuous tectonic loading. In a time-dependent earthquake forecast, the probabilities of a future event is conditioned on known previous earthquakes having occurred. The Working Group on California Earthquake Probabilities (WGCEP) is the name for the multidisciplinary team of scientists and engineers that develops the time dependent earthquake forecasts for California. The latest time-dependent model, the 2007 Uniform California Earthquake Rupture Forecast (UCERF), was released in early 2008. This new study updates the work done by the previous two working groups in 1995 and 2003. The 2007 Working Group revised earlier forecasts for Southern California (WGCEP 1995) and the San Francisco Bay Area (WGCEP 2003) by incorporating new data on active faults and an improved scientific understanding of how faults rupture to produce large earthquakes. In the new 2007 WGCEP study, the earthquake forecast is extended to the entire state using a uniform methodology, allowing for the first time, meaningful comparisons of earthquake probabilities in major metropolitan areas like Los Angeles and San Francisco Bay Area, as well as comparisons among the large faults in different parts of the state for the first time. The UCERF study is distinct from the NSHM; it describes the probability of an earthquake of various magnitudes occurring across California. The UCERF study does not describe the likelihood of amount of shaking caused by these quakes (seismic hazard). This is an important distinction between NSHM and UCERF, because even areas with a low probability of a local fault rupture can experience strong shaking and damage from distant, powerful earthquakes.

KEy IMPLICATIONS fOR CATASTROPHE RISK MANAGERS

The 2007 Uniform California Earthquake Rupture Forecast (UCERF) is the first study to apply a uniform methodology to the entire state, allowing meaningful comparisons between the probability of an earthquake occurring in Northern and Southern California. According to this study: New probabilities for the Elsinore and San Jacinto Faults in Southern California are about half of the previous predictions. UCERF study indicates a near certainty (99.7% probability) that there will be a M 6.7 earthquake in California in the next 30 years. There is a 46% probability of a M 7.5 earthquake in California in next 30 years. The chance of a M7.5 earthquake in Southern California (37% chance in 30 years) is more than double the chance in Northern California (15% chance in 30 years). Time-dependent probabilities for M 6.7 earthquake to occur on the southern San Andreas Fault (near Los Angeles) and Hayward-Rodgers Creek (near Oakland) are 23% and 34% higher than the time-independent probabilities, respectively. The time-dependent probability for a M 6.7 earthquake to occur on the northern San Andreas Fault (near San Francisco) is about 13% lower than time-independent view.

September 2008

Willis Re Inc.

CalIFoRNIa REGIoN (CoNtINUEd)

tIME-dEpENdENt vIEw

THE 2007 UCERf EARTHqUAKE TIME DEPENDENT PROBABILITIES

The UCERF framework comprises a sequence of four model types: 1) a fault model that gives the physical geometry of the larger, known faults; 2) a deformation model that gives slip rates and aseismicity factors to each fault section; 3) an earthquake rate model that gives the longterm rate of all earthquakes of magnitude five or greater (M 5) throughout the region and; 4) a probability model that gives a probability of occurrence for each earthquake during a specified (future) time interval. The latest versions of these models are used in developing the time-independent earthquake rate model in the revisions of 2008 NSHM and UCERF time-dependant model for California. The probability of a M 6.7 earthquake occurring in the next 30 years is shown in the table below relative to the time independent probability. In addition, the results from the last two working group studies are also presented in the below table for comparisons. Earthquake probabilities for many parts of California are similar to those in previous studies; however, the new probabilities calculated for the Elsinore and San Jacinto Faults in southern California are about half of the previous predictions.

Ref: The Uniform California Earthquake Rupture Forecast, Version 2 (UCERF 2): U.S. Geological Survey Open-File Report 2007-1437

Statewide earthquake probabilities are almost the same between the time-dependent and the time-independent (Poisson) models (time-dependent probabilities are 2% to 3% higher at M7.5). At a state level, these differences are insignificant relative to the overall modeling uncertainties. For portfolios with concentrations in specific regions or near specific faults, the difference between time-dependent and time-independent probabilities can be very significant. The difference between the time-dependent probabilities and the time-independent probabilities for M 6.7 earthquakes is greatest for the southern San Andreas and the HaywardRodgers Creek faults.
new probabilities calculated for the Elsinore

fault Name and San Jacinto faults in

southern California are about half of the previous

Time-Independent (Poisson1) Probability 48.3% 30.6% 12.7% 5.7% 23.3% 23.6% 7.6%
1

Time-Dependent (UCERf 07) Average Probability (Min - Max) 59.2% (22% - 94%) 31.3% (14% - 54%) 11.4% (5% - 25%) 6.1% (3% - 12%) 31.1% (12% - 67%) 20.6% (6% - 39%) 7.4% (1% - 22%)

Increase (+), Decrease (-) over Time-Independent Probabilities +23% +2% -10% +7% +33% -13% -3%

Prior Study Time-Dependent Probabilities 53% 61% 24% n/a 27% (10% - 58%) 23% (3% - 52%) 11% (3% - 27%)

% Change from Prior Study -9% -50% -47% n/a -14% +3% -31%

(previous Study in 1995) (previous Study in 2003)

predictions. (near

Southern California Northern California

San Andreas South LA) San Jacinto (near San Bernadino) Elsinore (near Lake Elsinore) Garlock (near Mojave) Hayward-Rodgers Creek (near Oakland) San Andreas North (near San Francisco) Calaveras (near San Jose)

90% segmented +10% un-segmented Poisson probabilities

September 2008

Willis Re Inc.

2

CalIFoRNIa REGIoN (CoNtINUEd)

tIME-dEpENdENt vIEw

The new forecast shows that California has a 99.7% chance to experience M 6.7 earthquake in the next 30 years and the likelihood of M 7.5 earthquake in the next 30 years is 46%. The occurrence probability of a M6.7 earthquake in the next 30 years striking the greater Los Angeles area is 67% and in the San Francisco Bay Area is 63%. The chance of an M7.5 earthquake occurring in Southern California (37% chance in 30 years) is more than double the chance in Northern California (15% chance in 30 years). The southern San Andreas Fault has the highest probability (59%) in California of generating at least one M6.7 earthquake in the next 30 years. In Northern California, the Hayward-Rodgers Creek Fault has the highest probability (31%) of generating at least one M6.7 earthquake in the next 30 years. Events of this size can be major loss causing events for the insurance industry, such as the 1989 Loma Prieta earthquake (M=6.9) which occurred on the northern San Andreas fault and the 1994 Northridge earthquake (M=6.7) The Cascadia Subduction Zone (CSZ) extends about 150 miles into northwest California, and is a major source of earthquakes in this region. There is 10% chance in the next 30 years the Cascadia Subduction Zone will generate a M= 8-9 event along the subduction zone somewhere between Northern California and Washington state.

Ref: The Uniform California Earthquake Rupture Forecast, Version 2 (UCERF 2): U.S. Geological Survey Open-File Report 2007-1437

Willis Re Inc.



September 2008

CasCadIa advaNCEMENts ChaNGE thE RIsk pRoFIlE IN thE paCIFIC NoRthwEst

KEy IMPLICATIONS fOR CATASTROPHE RISK MANAGERS
The tail risk in the Pacific Northwest region may change considerably. The amount of shaking experienced by low, mid-rise and high-rise buildings around Salem may increase at all return periods. The modifications made by the USGS may cause moderate reductions (15% or more) in key return period losses for high-rise buildings located in the Seattle metropolitan area.

SUMMARy Of CHANGES IN THE USGS SHAKE HAzARD ExPERIENCED By LOW AND HIGH RISE BUILDINGS
SA of 0.2 sec (2 Story Building) 475-year Return Period 2,475-year Return Period Moderate changes -15% to +20% Moderate to small decreases -15% to -5% Small changes -5% to +5% Moderate increases +15% to +20% SA of 1.0 sec (10 Story Building) 475-year Return Period Moderate to small decreases -25% to 0% Moderate decreases -25% to -15% Small decreases -5% to 0% Small increases 0% to +5% 2,475-year Return Period Moderate changes -15% to +20% Moderate to small decreases -15% to -5% Moderate to small increases +5% to 15% Moderate to small increases +10% to +20%

Pacific Northwest (Region wide) Seattle Portland Salem

Moderate to small changes -15% to +5% Moderate to small decreases -15% to -5% Small decreases -5% to 0% Small increases 0% to +5%

The primary reason for large decreases to the modeled hazard is the implementation of the Next Generation Attenuation (NGA) equations. In addition, the increasing view of risk along the coast of the Pacific Northwest at > 500-year return period is due to the changes to the magnitude-frequency relationship of the Cascadia Subduction Zone. Changes to the attenuation equations overshadow the smaller effects from changes to the other earthquake source parameters and their magnitudefrequency relationships.

GAINING INSIGHT INTO POTENTIAL MODELED LOSS CHANGES

Based on the shape of a building damage function for earthquakes, the amount of damage a building incurs rapidly decreases as the ground motion decreases (all other components remaining unchanged). Therefore, a 20% decrease in hazard can equal a 30-50% decrease in expected damage. This means that modeled damage for 2-story and 10-story buildings may decrease by a much larger amount than the change in modeled hazard that is shown in the maps on the following two pages. The change in the amount of shaking experienced by low-rise (e.g., 2-story) and high-rise (e.g., 10story) buildings at 475-year and 2,475-year return periods are shown below. Cool colors refer decreases in hazard, while warm colors reference increases in hazard. Only those areas where hazard is significant enough to result in damage are shown.

Willis Re Inc.

14

September 2008

paCIFIC NoRthwEst REGIoN (CoNtINUEd)

CHANGE IN USGS HAzARD fOR 2-STORy BUILDINGS

475-yr Hazard Return Period (10% exceeding probability in 50 years)

% Change in Hazard Between the 2002 and 2008 Hazard, (+) increase / (-) decrease

2,475-yr Hazard Return Period (2% exceeding probability in 50 years)

The gray shaded area along the West Coast at the 2,475-year return period designates the region where changes in modeled damage could be lower than the changes in modeled hazard presented in these maps. The shading is based on a representative 2-story wood frame building damage function. Outside this shaded area, it is possible for the change in modeled damage to exceed the changes in modeled hazard.
Willis Re Inc.

15

September 2008

paCIFIC NoRthwEst REGIoN (CoNtINUEd)

CHANGE IN USGS HAzARD fOR 10-STORy BUILDINGS

475-yr Hazard Return Period (10% exceeding probability in 50 years)

% Change in Hazard Between the 2002 and 2008 Hazard, (+) increase / (-) decrease

2,475-yr Hazard Return Period (2% exceeding probability in 50 years)

No shaded areas appear on these maps; therefore, the changes in modeled damage may exceed the changes in modeled hazard in all areas for 10-story buildings at these return periods.
Willis Re Inc.



September 2008

paCIFIC NoRthwEst REGIoN (CoNtINUEd)

LIST Of SCIENTIfIC ADvANCEMENTS DRIvING THE USGS CHANGES IN PACIfIC NORTHWEST

The USGS incorporated many new methods in its 2008 National Seismic Hazard Maps, in addition to the Next Generation Attenuation (NGA) equations. The key technical updates are listed below. 1. Revised magnitude-frequency distribution (Magnitude 8-9) of earthquakes on the Cascadia Subduction Zone (CSZ); size of possible maximum magnitude of an event from the CSZ was updated to 9.2 from 9.0 2. Introduction of deep seismicity zone near Portland, Oregon . Updates to the earthquake ground shaking attenuation equations for the Cascadia Subduction Zone. 4. Addition of three new crustal faults: Lake Creek-Boundary Creek fault located near Port Angeles, Washington Kendall fault scrap of the Boundary Creek fault located near the Canadian border near Bellingham Stonewall anticline located about 30 km west-southwest of the City of Newport, Oregon

Willis Re Inc.

7

September 2008

thE oRphaN tsUNaMICoNFIRMatIoN oF M8.7-9.2 CasCadIa EaRthqUakEs

In the 1960s, the theory of plate tectonics began to explain the concepts of continental drift and seafloor spreading. By the early 1980s, geophysists had shown that the Juan de Fuca Plate was subducting below the North American Plate at a rate of approximately 4 meters per century. Within this environment, scientists knew that a zone (known as the Cascadia Subduction Zone) where the plates met off the Pacific Coast was locked, and would only be freed by large earthquakes. There was an active debate, however, of how strong these earthquakes could be.
Ref: The Orphan Tsunami of 1700 Japanese Clues to a Parent Earthquake in North America

Researchers looked to the past to provide evidence of what could occur in the future. In the 1980s and 1990s, many studies attempted to reconstruct prior Cascadia earthquakes. One of the clues that was found related to the geophysics of the plates in the region. As the Juan de Fuca Plate descends below the North American Plate, the locked section between the plates causes the Pacific Coast to bulge upward in elevation by up to five feet. When a large earthquake occurs, the Pacific Coast actually drops back to its original level. This abrupt drop in the land triggers a westward propagating tsunami. This process has been the key to recent research included in the new USGS view of hazard in the Pacific Northwest.

Ref: The Orphan Tsunami of 1700 Japanese Clues to a Parent Earthquake in North America

The story was unfolding during the prior two updates to the National Seismic Hazard Maps. Researchers saw clear evidence in the geologic record of periods of abrupt shifts in the tidal patterns along the Pacific Coast. Ghost forests, for example, were discovered in the regions bays and estuaries. These forests were from spruce trees that had been overtaken by tidal waters after a drop in the coasts elevation. These decayed stumps, buried under sediment, provided evidence of the last great earthquake. Radiocarbon dating, however, could not pinpoint the time when the last major earthquake occurred, providing only a range from 1695 to 1720. This uncertainty made it difficult to link the various signs to estimate the true intensity of historic events.

Willis Re Inc.

8

September 2008

Ref: The Orphan Tsunami of 1700 Japanese Clues to a Parent Earthquake in North America

Independent of the work in the states, Japanese researchers had been documenting a tsunami of unknown origin an orphan tsunami that occurred in January of 1700. At a time when the Pacific Northwest was still unsettled, the Japanese had a stable bureaucracy in place, and the high level of literacy promoted extensive record-keeping. This orphan tsunami was documented to have flooded sites along over 500 miles of Japans Pacific coast. Tidal levels were recorded to have been 10 to 15 feet above their normal levels when the tsunami arrived. Although it is not technically depicting a tsunami wave, one of Hokusais Thirty-six views of Mount Fuji paintings has become an icon of such tsunamis. The orphan tsunami of 1700 was linked to the last great earthquake along Cascadia in the late 1990s through tree-ring dating. Additional research since the 2002 update of the USGS National Seismic Hazard Maps provided strong evidence that the 1700 earthquake was of magnitude 8.7-9.2, and most likely was M9.0. The intensity of the event has been explained through computer simulations of the orphan tsunami. A M8.08.5 earthquake would only produce 3 to 5 foot tidal surges along Japans Pacific coast, significantly underestimating what was observed. A Cascadia earthquake of M9.0, however, produces flood depths that provide a good match with history. These simulations have played a key role in the USGS decision to place increased weight on likelihood of M8.7-9.2 Cascadia earthquakes.
Ref: The Orphan Tsunami of 1700 Japanese Clues to a Parent Earthquake in North America Willis Re Inc.

September 2008

19

sMall aNd vaRIEd ChaNGEs IN thE NEw MadRId REGIoN

KEy IMPLICATIONS fOR CATASTROPHE RISK MANAGERS
The tail risk in the New Madrid region may change. The amount of shaking experienced by low, mid-rise and high-rise buildings close to New Madrid fault may decrease at return periods below 500 years. The modeled risk to low-rise properties in the immediate vicinity of the seismic zone (i.e., within 5 miles) may increase at return periods above 500 years.

SUMMARy Of CHANGES IN THE USGS SHAKE HAzARD ExPERIENCED By LOW AND HIGH RISE BUILDINGS
SA of 0.2 sec (2 Story Building) 475-year Return Period 2,475-year Return Period Moderate to small changes -25% to +5% SA of 1.0 sec (10 Story Building) 475-year Return Period Moderate to small decreases -15% to 0% 2,475-year Return Period Moderate changes -15% to +15%

New Madrid Region

Moderate to small decreases -25% to -5%

The changes to modeled hazard are due to various changes, such as changes to attenuation equations and changes to the modeling parameters of the northern arm of New Madrid fault.

GAINING INSIGHT INTO POTENTIAL MODELED LOSS CHANGES

Based on the shape of a building damage function for earthquakes, the amount of damage a building incurs rapidly decreases as the ground motion decreases (all other components remaining unchanged). Therefore, a 20% decrease in hazard can equal a 30 to 50% decrease in expected damage. This means that modeled damage for 2-story and 10-story buildings may decrease by a much larger amount than what is shown in the maps on the following two pages. The change in the amount of shaking experienced by low-rise (1-3 story) and high-rise (>8 story) buildings at 475-year and 2,475-year return periods are shown below. Cool colors indicate decreases in hazard, while warm colors indicate increases in hazard. Only those areas where hazard is significant enough to result in damage are shown.

LIST Of SCIENTIfIC ADvANCEMENTS DRIvING THE USGS CHANGES IN NEW MADRID REGION
The key technical updates specific to New Madrid region are listed below. 1. Implementation of an event cluster model for New Madrid earthquakes 2. Reduced magnitude in the northern New Madrid fault by 0.2 units and assigned this part of the fault a recurrence rate of 1/750 years 3. Modified fault geometry for New Madrid to include five hypothetical strands and an increased weight of 70% on the central strand 4. Revised dip of the Reelfoot fault to 38
Willis Re Inc.

20

September 2008

NEw MadRId REGIoN (CoNtINUEd)

CHANGE IN USGS HAzARD fOR 2-STORy BUILDINGS

475-yr Hazard Return Period (10% exceeding probability in 50 years)

% Change in Hazard Between the 2002 and 2008 Hazard, (+) increase / (-) decrease

2,475-yr Hazard Return Period (2% exceeding probability in 50 years)

The gray shaded area along the New Madrid seismic zone at the 2,475-year return period designates the region where changes in modeled damage could be lower than the changes in modeled hazard presented in these maps. The shading is based on a representative 2-story wood frame building damage function. Outside this shaded area, it is possible for the change in modeled damage to exceed the changes in modeled hazard.
Willis Re Inc.

2

September 2008

NEw MadRId REGIoN (CoNtINUEd)

CHANGE IN USGS HAzARD fOR 10-STORy BUILDINGS

475-yr Hazard Return Period (10% exceeding probability in 50 years)

% Change in Hazard Between the 2002 and 2008 Hazard, (+) increase / (-) decrease

2,475-yr Hazard Return Period (2% exceeding probability in 50 years)

No shaded areas appear on these maps, therefore it is possible for the change in modeled damage to exceed the changes in modeled hazard in all areas for 10-story buildings
Willis Re Inc.

22

September 2008

how wIll thE aIR, EqECat aNd RMs ModEls ChaNGE?

IMPLICATIONS Of NEW USGS NSHM ON vENDOR MODELS
The work coordinated by the USGS and the Working Group on California Earthquake Probabilities is very comprehensive and, as such, they have wide scientific and catastrophe modeler adoption. These studies are the impetus for AIR, EQECAT and RMS to make periodic updates to their U.S. Earthquake models. For example, RMS RiskLink v4.3 model, which resulted in significant changes in modeled risk for many portfolios, was motivated by the 2002 National Seismic Hazard Maps. The vendor modeling companies, however, cannot directly implement the National Seismic Hazard Maps and the Uniform California Earthquake Rupture Forecast into their models. This information must be translated into an event-based catastrophe model that is suited for the insurance industry. Therefore, although the three modeling companies may start their development activities from a similar place, their individual implementation of these studies will result in different answers to this question.

HOW WILL THE vENDOR MODELS CHANGES DIffER fROM THE USGS CHANGES?
There are three ways the changes in the vendor models generally will differ from USGS: 1. The vendor models are broader in scope to fulfill the need of insurance portfolio risk assessment than the USGS. (e.g., site-specific amplification, basin effects, fire following, damage model, demand surge, time dependency etc.). 2. The vendor model developers may selectively differ in their scientific assumptions than the USGS. 3. The vendor modelers have indicated they will recalibrate their models. It is plausible that changes to the engineering components of the models will offset or amplify changes to the seismic hazard. The vendor models are broader in scope than the USGS maps, accordingly, this point has significant implications for how one interprets changes to the USGS hazard information presented in this report. For example, the modeled risk to the structure coverage for 10-story buildings may decrease in the new models, however, new methods for modeling demand surge may offset some of these changes. In addition, some of the modeling assumptions made in the new maps might have already existed in the current version of the vendor models. For example, the known problem in the previous version of USGS maps related to earthquake rates in California (over estimation of M6-7 earthquake rates; known as the bulge issue). That has been updated in the new maps and already exists in the current version of the vendor models. Therefore, the changes in the USGS seismic hazard maps cannot be used to precisely predict the changes that will occur in the vendor models. It may seem counter-intuitive that the modeling companies would choose to differ from the consensus of the scientific community. Academic research is constantly progressing, some of which was not available in time to be included in the National Seismic Hazard Maps. The modelers assumptions may also differ from the weights assumed by the USGS in logic trees at different stages of the model. A choice of different weights at various stages of the hazard model could result in different changes in modeled risk. Finally, vendor modelers often take the opportunity to upgrade many of their models components, in addition to seismic hazard. Ultimately, the insurance industry is most interested in the product of all these components working together the financial liability calculated for a portfolio not each component in isolation. Therefore, the modelers will recalibrate their models to ensure that the final results are well validated, ensuring that each component is scientifically defensible. As such, changes in the seismic hazard component of the model may lead to refinements in the engineering model. These multiplicative changes could result in offsetting or amplifying effects.
Willis Re Inc.

2

September 2008

WHAT CAN WE CONCLUDE NOW ABOUT EARTHqUAKE RISK IN THE WESTERN AND CENTRAL U.S.?
At this point, we can conclude that model changes will be significant for many portfolios, and the patterns of change will be complex and multifaceted. These changes will affect underwriting guidelines, capital requirements, and portfolio management strategies. Also, these changes will affect the Workers Compensation portfolios. Changes to portfolio loss estimates in the Western U.S. will be highly influenced by the new NGA equations, especially for mid-rise and high-rise buildings. Changes to loss estimates in the Central and the Eastern U.S. will be relatively low for many portfolios compared to changes to the Western U.S. Willis will continue to closely follow information related to AIR, EQECAT and RMS 2009 U.S. Earthquake model upgrades. We will provide regular updates as appropriate. In the communications, we will continue to offer balanced advice, based on our range of skills from model development to the practical implementation of portfolio management and underwriting objectives. We encourage you to contact us to explore how we can help your company develop and execute new action plans, which reflect a deep understanding of the changing landscape of risk in the vendor models.

CONTACT US
Prasad Gunturi Director of U.S. Modeling Research prasad.gunturi@willis.com Phone: +1 952 841 6638
Willis Re Inc.

Kyle Beatty vice President kyle.beatty@willis.com Phone: +1 952 841 6635

September 2008

24

GlossaRy
Attenuation: Decrease in size, or amplitude, of the waves is called attenuation. Seismic waves also become attenuated as they move away from the earthquake source. Creep: Slow, more or less continuous movement occurring on faults due to ongoing tectonic deformation. Faults that are creeping do not tend to have large earthquakes. Earthquake: Earthquake is a term used to describe both sudden slip on a fault, and the resulting ground shaking and radiated seismic energy caused by the slip, or by volcanic or magmatic activity, or other sudden stress changes in the earth. Epicenter: The epicenter is the point on the earths surface vertically above the hypocenter (or focus), point in the earth crust where a seismic rupture begins. fault: A fault is a fracture along which the blocks of crust on either side have moved relative to one another parallel to the fracture. Ground motion: Ground motion is the movement of the earths surface from earthquakes or explosions. Ground motion is produced by waves that are generated by sudden slip on a fault or sudden pressure at the explosive source and travel through the earth and along its surface. Intensity: The intensity is a number (written as a Roman numeral) describing the severity of an earthquake in terms of its effects on the earths surface and the built environment. Several scales exist, but Modified Mercalli (MMI) scale is most commonly used in the United States. There are many intensities for an earthquake, depending on where you are, unlike the magnitude, which is one number for each earthquake. MMI I II III IV V Description Instrumental Feeble Slight Moderate Rather Strong Observable Effects Not felt. Marginal and long-period effects of large earthquakes. Felt by persons at rest, on upper floors, or favorably placed. Felt indoors. Hanging objects swing. Vibration like passing of light trucks. Duration estimated. May not be recognized as an earthquake. Hanging objects swing. Vibration like passing of heavy trucks; or sensation of a jolt like a ball striking the walls. Standing motor cars rock. Windows, dishes, doors rattle. Glasses clink. Crockery clashes. In the upper range of IV wooden walls and frames creak. Felt outdoors; direction estimated. Sleepers wakened. Liquids disturbed, some spilled. Small unstable objects displaced or upset. Doors swing, close, open. Shutters, pictures move. Pendulum clocks stop, start, change rate. Worst effects include some windows broken out; a few instances of fallen plaster or damaged old masonry chimneys on single-family houses; large cracks in interior walls; many small objects overturned and fallen; many items thrown from store shelves; many glassware items or dishes broken; light furniture overturned and moderately heavy furniture displaced. Effects on people not used to define intensities of VI or above. Worst effects include significant damage to unreinforced masonry buildings, including cracks in bearing walls and out-ofplane movement or fall of upper walls and parapets; many old masonry chimneys fallen or broken at the roofline on singlefamily homes; some masonry fences fallen or destroyed; heavy furniture overturned. Worst effects include considerable damage to old, unreinforced masonry buildings, with partial collapse; many cases where wood-frame houses are moved on their foundation if not anchored and braced; damage to wood-frame apartment buildings having open first-stories, with some cases of apartments being destroyed; significant damage to reinforced, lined, masonry chimneys on single-family homes, and widespread damage to old masonry chimneys; structural damage to some reinforcedconcrete structures built when a seismic code was in effect; very heavy furniture moved conspicuously or overturned. Worst effects include multiple cases of structural damage to reinforced-concrete buildings and parking structures built when a seismic code was in effect, with some cases of partial or complete collapse; collapse of elevated freeway sections; widespread damage to unreinforced masonry buildings (e.g., old brick buildings), with total collapse; widespread incidence of wood-frame houses shifted off foundations where not securely anchored and braced; widespread destruction of woodframe apartment buildings having large open areas in their first stories; widespread collapse of masonry (brick, block or stone) chimneys, whether reinforced or not, on single-family homes; furniture and building contents generally overturned and thrown across room. As originally defined and as modified in 1931, these intensity levels described earthquake effects that involve permanent changes in the shape of the ground (fault rupture, landsliding, liquefaction, etc.). Nowadays, however, Intensities X, XI and XII are increasingly regarded as approximately the same level of shaking as Intensity IX. The many phenomena originally associated with intensities X and above are apparently related less to the level of ground shaking than to the presence of ground conditions susceptible to spectacular failure, or to the ease with which seismic faulting of different style and depth can propagate to the ground surface. 25
September 2008

Interplate: Interplate pertains to processes between the earths crustal plates. Intraplate: Intraplate pertains to processes within the plates Magnitude: Magnitude depict the relative size of an earthquake. Magnitude is based on measurement of the maximum motion recorded by a seismograph. Several scales have been defined, but the most commonly used are (a) local magnitude (ML), commonly referred to as Richter magnitude, (b) surface-wave magnitude (Ms), (c) body-wave magnitude (Mb), and (d) moment magnitude (Mw). The moment magnitude (Mw) scale, based on the concept of seismic moment, is uniformly applicable to all sizes of earthquakes but is more difficult to compute than the other types. Peak Ground Acceleration (PGA): PGA is the largest acceleration experienced by the ground at a particular point away from the epicenter. Spectral Acceleration (SA): Spectral acceleration is approximately what is experienced by a building, as modeled by a particle on a massless vertical rod having the same natural period of vibration as the building. Subduction zone: The subduction zone is the place where two plates come together, one riding over the other. Tsunami: A tsunami is a sea wave of local or distant origin that results from large-scale seafloor displacements associated with large earthquakes, major submarine slides, or exploding volcanic islands.

VI

Strong

VII

Very Strong

VIII

Destructive

IX

Ruinous

X-XII

Disastrous

Willis Re Inc.

REFERENCEs
2007 Working Group on California Earthquake Probabilities, 2008, The Uniform California Earthquake Rupture Forecast, Version 2 (UCERF 2): U.S. Geological Survey Open-File Report 2007-1437 and California Geological Survey Special Report 20 Atwater, B. F., The Orphan Tsunami of 1700- Japanese Clues to a Parent Earthquake in North America, U.S. Geological Survey Professional Paper 1707, 133 p. Campbell, K.W., and Bozorgnia, Y., 2008, Ground motion model for the geometric mean horizontal component of PGA, PGV, PGD and 5% damped linear elastic response spectra for periods ranging from 0.01 to 10.0 s: Earthquake Spectra, v. 24, no. 1. Dewey, J.W., Dengler, L., Reagor, B.G., and Moley, K., 1995, Spatial variations of intensity in the Northridge earthquake, in The Northridge, California, Earthquake of 17 January 1994, Calif. Div. Mines & Geology Special Publication 116, p. 39-46. FEMA-351, 2000, Recommended Seismic Evaluation and Upgrade Criteria for Existing Welded Steel Moment-Frame Buildings, prepared by the SAC Joint Venture for the Federal Emergency Management Agency, Washington, DC. FEMA-355E, 2000, State of the Art Report on Past Performance of Steel Moment-Frame Buildings in Earthquakes, prepared by the SAC Joint Venture for the Federal Emergency Management Agency, Washington, DC. Frankel, A., Mueller, C., Barnhard, T., Perkins, D., Leyendecker, E., Dickman, N., Hanson, S., and Hopper, M., 1996, National Seismic Hazard MapsDocumentation June 1996: U.S. Geological Survey Open-File Report 96532, 110 p. Frankel, A.D., Petersen, M.D., Mueller, C.S., Haller, K.M., Wheeler, R.L., Leyendecker, E.V., Wesson, R.L., Harmsen, S.C., Cramer, C.H., Perkins, D.M., Rukstales, K.S., 2002, Documentation for the 2002 update of the National Seismic Hazard Maps: U.S. Geological Survey Open-File Report 2002420, 39 p. Petersen, Mark D., Frankel, Arthur D., Harmsen, Stephen C., Mueller, Charles S., Haller, Kathleen M.,Wheeler, Russell L., Wesson, Robert L., Zeng, Yuehua, Boyd, Oliver S., Perkins, David M., Luco, Nicolas, Field, Edward H., Wills, Chris J., and Rukstales, Kenneth S., 2008, Documentation for the 2008 Update of the United States National Seismic Hazard Maps: U.S. Geological Survey Open-File Report 20081128, 61 p. Porter, K.A., J.L. Beck, H.A Seligson, C.R. Scawthorn, L.T. Tobin, R. Young, and T. Boyd, 2002, Improving Loss Estimation for Woodframe Buildings, Vol. 1, Report, Consortium of Universities for Research in Earthquake Engineering, Richmond, CA, 136 p. Porter, K.A., J.L. Beck, H.A Seligson, C.R. Scawthorn, L.T. Tobin, R. Young, and T. Boyd, 2002, Improving Loss Estimation for Woodframe Buildings, Vol. 2, Appendices, Consortium of Universities for Research in Earthquake Engineering, Richmond, CA, 300 p. Richter, C.F. Elementary Seismology. San Francisco CA: W. H. Freeman Co., 1957. SAC 95-04, 1995, Analytical and Field Investigations of Buildings Affected by the Northridge Earthquake of January 17, 1994, prepared by the SAC Joint Venture for the Federal Emergency Management Agency, Washington, DC. SAC 95-07, 1995, Technical Report: Case Studies of Steel Moment-Frame Building Performance in the Northridge Earthquake of January 17, 1994, prepared by the SAC Joint Venture for the Federal Emergency Management Agency, Washington, DC. Sadigh, K., et. al., Attenuation relationships for shallow crustal earthquakes based on California strong motion data: Seismological Research Letters, 1997, v. 68, p. 180189 Zhao et al., 2006 Zhao J.X., et. al., 2006 Attenuation relations of strong ground motion in Japan using site classification based on predominant period: Bulletin of the Seismological Society of America, v. 96, p. 898913.

The contents herein are provided for informational purposes only and do not constitute and should not be construed as professional advice. Any and all examples used herein are for illustrative purposes only, are purely hypothetical in nature, and offered merely to describe concepts or ideas. They are not offered as solutions to produce specific results and are not to be relied upon. The reader is cautioned to consult independent professional advisors of his/her choice and formulate independent conclusions and opinions regarding the subject matter discussed herein. Willis is not responsible for the accuracy or completeness of the contents herein and expressly disclaims any responsibility or liability for the readers application of any of the contents herein to any analysis or other matter, nor do the contents herein guarantee, and should not be construed to guarantee, any particular result or outcome.

Willis Re Inc.

2

September 2008