University of Southern California Dissertations and Theses

A robust model for disaster risk and decision analysis

(USC Thesis Other)

A robust model for disaster risk and decision analysis

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

A ROBUST MODEL FOR DISASTER RISK AND DECISION ANALYSIS

by

Carl Erckman Southwell

_________________________________________________________________________________________________________________________________________

A Dissertation Presented to the
FACULTY OF THE USC SOL PRICE SCHOOL OF PUBLIC POLICY
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF POLICY, PLANNING, AND DEVELOPMENT

May 2012

Copyright 2012 Carl Erckman Southwell
ii

Table of Contents

List of Tables iii

List of Figures v

Abstract x

Chapter 1. Introduction 1

Chapter 2: An Intentional Disaster at a Single Unit of Planned Infrastructure: A Case
Study Involving the Long Beach SES LNG Receiving Facility 21

Chapter 3: An Intentional Disaster at a Single Unit of Pre-existing Infrastructure: A
Case Study Involving the Rancho LPG Storage Facility 74

Chapter 4: A Natural Disaster Involving a Pre-existing Infrastructure System: A
Case Study of the New Orleans’ Flood Control System 100

Chapter 5: Mitigating Potential, Intentional Disasters at a Pre-existing Regional
Infrastructure Complex: The Port Security Risk Analysis and Resource Allocation
System (PortSec) 126

Chapter 6: Mitigating Natural Disasters Involving Regional Infrastructure Systems:
A Hurricane Tracking, Consequences, and Mitigation Capabilities Modeling Tool 157

Chapter 7: Feasible Solutions in Allocating Countermeasures Related to Intentional
Disasters on Regional Infrastructure Systems: Preventive Radiological/Nuclear
Detection (PRND) Resource Allocation in California 182

Chapter 8: Overall Observations and Implications 203

Bibliography 211

Appendices 223
Appendix A: Software Description and Reference for HAZUS-MH 223
Appendix B: Software Description and Reference for Hurrevac2010 224
Appendix C: Software Description and Reference for Landview6 with
MARPLOT 226
Appendix D: Hurricanes 227
iii

List of Tables

Table 1 Disaster Types 4

Table 2 Disaster Cases 16

Table 3 Current U.S. LNG Receiving Facilities and their Annual Gas Imports
(in Billions of Cubic Feet) 52

Table 4 Long Beach SES LNG Receiving Facility Decision Tree Variables 54

Table 5 2005 Demographic Profile of Area within Various Distances of Long
Beach SES LNG Facility 57

Table 6 2000 Demographic Profile of Area within Various Distances of
Everett Tractebel LNG N.A. Facility 67

Table 7 2000 Demographic Profile of Area within Various Distances of
Providence KeySpan LNG Facility 70

Table 8 Relative Severities of Potential Terrorist Acts at Selected U.S. LNG
Receiving Facilities 73

Table 9 Demographic Profiles of Areas within Various Distances of Rancho
LPG Facility based on 2000 Census and Corresponding Expected
Consequences 99

Table 10 Relationships between Storm Categories, Wind Speed, Minimum
Surface Pressure and Storm Surge 110

Table 11 Data Dictionary—Summary of PortSec Risk Analysis Model
Variables 146

Table 12 Entry Screen of the Hurricane/Floods Capabilities and Characteristics
Excel Input Spreadsheet 169

Table 13 Entry Screen for Capabilities Percentage Savings or Costs in the Excel
Input Spreadsheet 170

Table 14 Output of the Analytica Model—Qualitative Mitigation Adjustment
Factors 178

iv

Table 15 Output of the Analytica Model—Expected High-end Hurricane Losses
Matrix 179

Table 16 Output of the Analytica Model—Expected Low-end Hurricane Losses
Matrix 180

Table 17 Sample Target Areas with Estimated Population Densities 189

Table 18 Sample Formulae for Probabilities and Consequences of Terrorist RN
Attacks 190

Table 19 Estimated Types, Costs, and Impacts of Mitigation Resources 191

Table 20 Allocation of Mitigation Resources by Target Assuming a $5 million
Budget, 5 kilograms of Radioactive Material, a Value of Life of $5
million, and an Indirect- to Direct-Consequences Factor of 3.0—
Output of the Analytica Model 200

Table 21 Saffir/Simpson Hurricane Scale 228
v

List of Figures

Figure 1 A Robust Model for Disaster Risk and Decision Analysis 12

Figure 2 A Robust Model for Disaster Risk and Decision Analysis—Iterative
and Incremental 13

Figure 3 Long Beach SES LNG Receiving Facility Decision Tree—Scenario 1
Proposed Location 53

Figure 4 Long Beach SES LNG Receiving Facility Decision Tree Scenario 1
Proposed Location (Rolled Back Given Base Assumptions) 55

Figure 5 Simulated Terrorist Attack on Proposed Long Beach LNG Receiving
Facility (during winter, spring, or fall) 56

Figure 6 Tornado Diagram of Long Beach SES LNG Receiving Facility
Decision Tree 58

Figure 7 2 Way Sensitivity Diagram—Long Beach SES Receiving Facility
Decision Tree 59

Figure 8 Simulated Terrorist Attack on Proposed Long Beach LNG Receiving
Facility (during summer) 60

Figure 9 Simulated Terrorist Attack on Long Beach LNG Receiving Facility at
Pier J Alternative Site (during winter, spring, or fall) 61

Figure 10 Simulated Terrorist Attack on Long Beach LNG Receiving Facility at
Pier J Alternative Site (during summer) 62

Figure 11 Simulated Terrorist Attack on Long Beach LNG Receiving Facility at
Offshore Island or Jetty with Pier (3 mi S of Pier J) Alternative Site
(during winter, spring, or fall) 63

Figure 12 Simulated Terrorist Attack on Long Beach LNG Receiving Facility at
Offshore Island or Jetty with Pier (3 mi S of Pier J) Alternative Site
(during winter, spring, or fall) 64

Figure 13 Expected cost given a Terrorist Attack Occurs ($ Millions) 65

Figure 14 Dominant Strategies given Base Assumptions 66

vi

Figure 15 Simulated Terrorist Attack on Boston LNG Receiving Facility (during
summer) 68

Figure 16 Simulated Terrorist Attack on Boston LNG Receiving Facility (during
winter) 69

Figure 17 Simulated Terrorist Attack on Rhode Island LNG Receiving Facility
(during summer) 71

Figure 18 Simulated Terrorist Attack on Rhode Island LNG Receiving Facility
(during winter) 72

Figure 19 Vapor Cloud Explosion Size by Volume of Butane 85

Figure 20 BLEVE 87

Figure 21 “Twin Sisters” Propane Facility, Elk Grove, CA and Rancho LPG
Facility, San Pedro, CA 89

Figure 22 Simulated Terrorist Attack in LPG Storage Facility (first tank
breached) 97

Figure 23 Simulated Terrorist Attack in LPG Storage Facility (other tank
breached) 98

Figure 24 NOLA Flood Control Risk Analysis System—High Level View of the
Decision Analysis Model 105

Figure 25 Mississippi Flood Submodel 106

Figure 26 Lake Pontchartrain Flood Submodel 107

Figure 27 Analysis Submodel 109

Figure 28 Cumulative Distribution Functions of 100-Year Flood Levels for Lake
Pontchartrain for Different Levels of Protection 114

Figure 29 Cumulative Distribution Function of 100-Year Flood Levels at the
Mississippi River for Different Protection Levels and Assuming No
Use of Cut-offs 116

vii

Figure 30 Cumulative Severity Distribution of Hurricane Losses (in $ Millions
and excluding value of lives) 117

Figure 31 Cumulative Severity Distribution of Non-hurricane Flood Losses (in $
Millions and excluding value of lives) 118

Figure 32 Cumulative Severity of Lives (in $ Millions) as a Function of
Evacuation Time 119

Figure 33 Net Annual Flood Costs from all causes assuming a population of
750,000, 48 hour evacuation time, mitigation costs for levee and
floodwall fortification and maintenance, 7% interest, $3.265 million
per vertical foot per mile construction costs, and an imputed value of
$10 million per life 121

Figure 34 PortSec System Architecture 133

Figure 35 Tactical User Interface – PortSec 1.0 135

Figure 36 Example: PortSec Influence Diagram 143

Figure 37 Example: PortSec Operating Modules 145

Figure 38 Ports of Long Beach/Los Angeles: Base Case—Conditional Expected
Losses for each of 56 MAST Targets given 9 MAST Attack Scenarios
(Assumes All Berths are Active and Status Quo Security Allocations) 151

Figure 39 Ports of Long Beach/Los Angeles: Enhanced Base Case—Conditional
Expected Losses for each of 56 MAST Targets given 9 MAST Attack
Scenarios (Assumes All Berths are Active and Selected Enhanced
Security Allocations) 152

Figure 40 Ports of Long Beach/Los Angeles: Preliminary Test Case—
Conditional Expected Losses for each of 56 MAST Targets given 9
MAST Attack Scenarios (Assumes Selected Berths are Active {based
on test Marine Exchange berthing data} and Selected Enhanced
Security Allocations) 153

Figure 41 Entry Display Screen to the Analytica Model 164

viii

Figure 42 Non-hurricane Concurrent Events’ Submodel to the Analytica Model 165

Figure 43 Flood Surge and Wind Damage Submodel to the Analytica Model 165

Figure 44 Land-Use Characteristics Submodel to the Analytica Model 166

Figure 45 Demographics Submodel to the Analytica Model 166

Figure 46 Qualitative Capabilities Matrices Submodel to the Analytica Model 168

Figure 47 Hurricane Loss Analysis Submodel to the Analytica Model 171

Figure 48 HURREVAC Hurricane Andrew Historical Track Prior to Florida
Landfall 174

Figure 49 MARPLOT Hurricane Andrew Historical Track Prior to Florida
Landfall 176

Figure 50 MARPLOT Hurricane Andrew Historical Track Prior to Florida
Landfall (Areas Impacted) 177

Figure 51 Output of the Analytica Model—Expected Hurricane Losses by
Evacuation Time and Assuming All Mitigations Are Applied (Upper
Curve = Lowest Savings %s and Lower Curve = Highest Savings %s) 181

Figure 52 Estimation of RN Damage Incurred by Target 192

Figure 53 Estimation of Optimal Mitigation Resources by Type and by Target 193

Figure 54 Estimated RN Terrorist Attack Consequences by Target Assuming a 5
kilograms of Radioactive Material, a Value of Life of $5 million, and
an Indirect- to Direct-Consequences Factor of 3.0 196

Figure 55 Estimated Savings by Allocation of Mitigation Resources Types by
Target Assuming a $5 million Budget, 5 kilograms of Radioactive
Material, a Value of Life of $5 million, and an Indirect- to Direct-
Consequences Factor of 3.0 198

ix

Figure 56 The Number of Hurricanes Expected to Occur During a 100-year
Period Based on Historical Data 229

Figure 57 Disaster Caused by Hurricanes or Tropical Storms 231
x

Abstract

There are many types of disasters, and it is a challenge to derive a single model of
disaster risk and decision analysis that encompasses them all. At the same time, a
general model for analyzing disaster risks and decisions could make the tasks of
disaster avoidance, prevention, mitigation, rescue, rebuilding, and recovery much
easier. This dissertation formulates a model of risk and decision analysis concerned
with behaviors (decisions or “objectives”) and stimuli (initial “risks and
uncertainties” and their associated sets of choices/utilities/“levers”) that are
chronicled and calibrated by “metrics.” Changes in this model are displayed as
changes in preferences over time or as conditioned by prior choices. In practice, this
dynamic model of risk and decision analysis is usually represented as one or a set of
static “time slices” from the model that serve as exemplars for the policy implications
the specific, practical analysis is attempting to highlight.

This dissertation showcases some specific, practical analyses—six case studies—
whose underlying model is derived from this general model. Each case makes its
own points about policy implications related to its specific disaster type and
circumstances while, at the same time, each demonstrates the robustness and
flexibility of the more general model. Each case employs qualities of a robust model
as follows:
1. It outlines the essential elements and relationships in the disaster process
being considered,
xi

2. It establishes a set of data that describes and measures the disaster process
being considered,
3. It offers opportunities for comparison to actual or hypothetical disasters,
4. It acts as a template to better understand and help prevent recurrences of past
disasters, to react and respond to and recover from unfolding disasters, and to
plan for and mitigate future disasters, and
5. It serves as a tool to communicate the risks and potential consequences of the
disaster being considered and the actions necessary to mitigate its risk.

1

Chapter 1: Introduction

Disasters are rare events that cause substantial harm. They injure and kill people, destroy
property, and disrupt businesses. Intentional and unintentional catastrophes and natural
disasters can be divided into three classes of disasters (see Table 1). Intentional
catastrophes are purposive, non-random acts that include war, acts of terrorism,
insurgencies, and large-scale crimes (for example, an arsonist lighting a wildfire).
Unintentional catastrophes include random human actions or failures of equipment or
systems such as industrial accidents, nuclear reactor meltdowns, plane crashes, and
sinking ships. Natural disasters consist of hurricanes, tornadoes, blizzards, tsunamis,
earthquakes, droughts, floods, volcanoes, pandemics, wildfires, and other usually difficult
to predict natural actions that can trigger devastating harm.

Given the variety of disaster types, it is difficult to analyze and manage them in a
purposive, consistent, and systematic manner. For this reason, models of disaster risks
and decisions are used to make the analysis and organization of disasters’ avoidance,
prevention, mitigation, rescue, rebuilding, and recovery mechanisms tractable. Models
of risk and decision analysis are, at their core, concerned with behaviors (decisions or
“objectives”) and stimuli (initial “risks and uncertainties” and their associated sets of
choices/utilities/“levers”) that are chronicled and calibrated by “metrics.” (see Figure 1)
The dynamics of such models can be displayed as changes in preferences over time or as
conditioned by the choice of prior objectives (this is equivalent to viewing each action or
iteration as a static time slice representing that which would have happened if the
2

decision-maker had stopped deliberating at that point in time). (see Figure 2). Indeed, a
risk and decision analysis often displays one or a set of these static “time slices” as
exemplars of the point(s) to be made.

Analyzing disaster risks and decisions can employ many different analytic methods. The
more generalizable stochastic methods include probabilistic fault tree and event tree
analysis, Bayesian networks, influence diagrams, game theoretic simulations, Markov
modeling, and Monte Carlo simulations. Regardless of the method of analysis used,
robust models can accommodate the analysis of disaster risk and decisions for a wide
variety of disaster types that communicate certain, essential ideas as follows:
1. Robust models must outline the essential elements and relationships in the
disaster processes.
2. Robust models must establish dictionaries of the data to be collected to adequately
describe and measure the disaster processes.
3. Robust models must offer opportunities for comparison to actual disasters and
platforms from which the models may be evolved and improved to incorporate the
complexity of actual disasters.
4. Robust models must act as templates to better understand and help prevent
recurrences of past disasters, to react and respond to and recover from unfolding
disasters, and to plan for and mitigate future disasters.
5. Robust models must serve as tools to communicate the risks and potential
consequences of disasters and the actions necessary to mitigate these risks and
consequences.
3

The negative consequences that define disasters emerge from pre-existing landscapes
wherein unique, complex combinations of circumstances emerge from necessary pre-
conditions. The decisions influencing such consequences have tools that can be used to
ameliorate the negative impacts through iteratively measuring and recalibrating these
levers to seek attainment of the disasters’ objectives (see Figure 1). These objectives
engage unique approaches and solutions to the potential mitigation of disasters,
approaches and solutions that vary depending on whether the disaster is an intentional or
unintentional action or a natural event or involves varying perspectives of location, scope,
and timing.

Various mitigations and countermeasures can influence the probability that intentional
disasters might occur. More security guards and cameras combined with better
monitoring can reduce the chance of a crime or terrorist event at a given location. For
example, Al Qaeda may desire to crash commercial aircraft into new, prominent
buildings, but enhanced security procedures may prevent the group’s desires from
becoming reality. Management techniques such as training and drills and engineered
controls (e.g., strategic retreat as a building strategy in coastal floodplains) or redundant
safety systems and hardened SCADA in infrastructure can also reduce the likelihood that
unintentional disasters might occur.

4
Table 1: Disaster Types

Disaster Type Disaster Characteristics Risks and Uncertainties Disaster Risk and
Decision Analysis
Metrics
Disaster Risk and Decision
Analysis Levers
Disaster Risk and Decision
Analysis Objectives
Crimes,
cyber (e.g.,
Internet
attacks)
• Indefinite or global
infrastructure and
disaster scope
• Infrastructure
network
• Planned/existing
• Usually, little or no
warning
• Intentional disaster
• GDP resiliency
• Internet traffic flows/security
borders
• GDP
• Supply &demand seasonality
• Internet traffic flows
• Regulations
• Adversaries
• Built environment
• Engineering/redundancy/
controls
• Maintenance
• Security
• Funding changes due
to new laws
• Rate of infrastructure
recovery
• Rate of throughput
recovery
• Policy changes
• Operational changes
• Magnitude of
disaster
• Countermeasure
adjustment
• Internet usage
statistics
• Crime rates
• Crime closure rates
• New laws
• Regulatory enforcement
• Funding
• Pre- & post-disaster
coordination efforts
• Interagency coordination
• Response planning
• Management
• Communications
• Intelligence
• Counterintelligence
• Software countermeasures
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Infrastructure hardening
• Security regimen (strategic)
• Optimized throughput
• Improved response time
• Optimized countermeasures
(tactical/incident)
• Reduce societal fear
Crimes,
large-scale
(e.g.,
wildfires lit
by arsonists)
• Local or regional
infrastructure and
disaster scope
• Planned/existing
• Usually, little or no
warning
• Intentional disaster
• Population density
• GDP resiliency
• Traffic flows/borders
• GDP
• Traffic flows
• Regulations
• Adversaries
• Built environment
• Maintenance
• Security
• Funding changes due
to new laws
• Rate of infrastructure
recovery
• Rate of throughput
recovery
• Policy changes
• Operational changes
• Magnitude of
disaster
• Countermeasure
adjustment
• Crime rates
• Crime closure rates
• New laws
• Regulatory enforcement
• Funding
• Pre- & post-disaster
coordination efforts
• Interagency coordination
• Response planning
• Management/Communica
tions
• Intelligence
• Counterintelligence
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Infrastructure hardening
• Security regimen (strategic)
• Optimized throughput
• Improved response time
• Optimized countermeasures
(tactical/incident)
• Reduce societal fear

5
Table 1, continued

Disaster
Type
Disaster Characteristics Risks and Uncertainties Disaster Risk and
Decision Analysis
Metrics
Disaster Risk and Decision
Analysis Levers
Disaster Risk and Decision
Analysis Objectives
Diseases/
Pandemics
• Natural disaster or
intentional or
unintentional disaster
• Regional, national, or
global disaster scope
• Ranges from little or
no warning to emergent
• Random, intentional,
or unintentional
• Population density
• GDP resiliency
• Regulations
• Mutual aid capabilities
• Medical/vaccination capabilities
• Adversaries
• Security
• Virulence
• Technological advances
• Rate of recovery
• Healthcare policy
changes
• Operational changes
• Magnitude of disaster
• Response rate
• Medical care
effectiveness
• Quarantine
effectiveness
• Vaccination rates
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Regulatory enforcement
• Healthcare policy
• Medical stockpiles
• Technological
improvements
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Adequate medical and
pharmaceutical resources
• Reduce societal fear
Droughts • Natural disaster
• Regional disaster
scope
• Emergent
• Random
• Frequencies are
characteristic of
geographic locations
• Population density
• GDP resiliency
• Regulations
• Mutual aid capabilities
• Climate change
• Rate of recovery
• Water policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Water deliveries
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Water policy
• Regulatory enforcement
• Litigation
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure
improvement/well building and
water treatment
Earth-
quakes
• Natural disaster
• Regional infrastructure
and disaster scope
• Little or no warning
• Random
• Frequencies are
characteristic of
geographic locations
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Secondary disaster from
tsunami
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Evacuation rate
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Regulatory enforcement
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Reduce societal fear

6
Table 1, continued

Disaster
Type
Disaster Characteristics Risks and Uncertainties Disaster Risk and
Decision Analysis
Metrics
Disaster Risk and Decision
Analysis Levers
Disaster Risk and Decision
Analysis Objectives
Floods • Natural, intentional, or
unintentional disaster
• Regional infrastructure
and disaster scope
• Usually, little or no
warning
• Random, intentional, or
unintentional
• Infrastructure network
• Planned/existing
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Secondary disaster from storm
surge and/or floods
• Weather
• Climate change
• Maintenance
• Engineering (e.g., dams, dikes,
floodwalls)
• River avulsions
• Adversaries
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Evacuation rate
• Rate of infrastructure
recovery
• Countermeasure
adjustment
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Regulatory enforcement
• Litigation
• Management/Communicatio
ns
• New laws
• Regulatory enforcement
• Interagency coordination
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Improved evacuation time
• Security regimen (strategic)
• Optimized countermeasures
(tactical/incident)
Global
warming
• Natural disaster
• Global infrastructure
and disaster scope
• Warnings are emergent
• Random or
unintentional disaster
• Frequency/severity is
characteristic of global
cycles and/or human
activities
• Population density
• Geographic settlement patterns
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• International cooperation
• First response capabilities
• Mutual aid capabilities
• Secondary disasters from rising
sea levels, changing weather
patterns, and/or social
instabilities
• Weather/Climate change
• Rate of recovery
• Environmental policy
changes
• Operational changes
• Magnitude of disaster
• Response rate
• Sea level rise
• Evacuation rate
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Coastal zone policy and
management
• Regulatory enforcement
• Litigation
• Emissions standards
• Forestry and agricultural
policy
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Improved evacuation time
• Coastal retreat and/or
hardening

7
Table 1, continued

Disaster Type Disaster Characteristics Risks and Uncertainties Disaster Risk and
Decision Analysis
Metrics
Disaster Risk and Decision
Analysis Levers
Disaster Risk and Decision
Analysis Objectives
Industrial
accidents/
techno-
logical
disasters
(e.g., tank
accident,
dam failure,
bridge
collapse,
nuclear
containment
leak)
• Natural, intentional,
or unintentional
disaster
• Local or regional
infrastructure and
disaster scope
• Usually, little or no
warning
• Random, intentional,
or unintentional
• Infrastructure or
infrastructure network
• Planned/existing/
obdurate
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Maintenance
• Engineering (e.g., dams, dikes,
floodwalls)
• Adversaries
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Evacuation rate
• Rate of infrastructure
recovery
• Countermeasures
• Security
• Control device
adjustment
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Regulatory enforcement
• Litigation
• Management
• Communications
• New laws
• Regulatory enforcement
• Interagency coordination
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Improved evacuation time
• Security regimen and/or
SCADA (strategic)
• Optimized
countermeasures/security/contr
ol devices (tactical/incident)
Radioactive/
nuclear
device
• Regional, national, or
global infrastructure
& disaster scope
• Infrastructure network
• Little or no warning
• Intentional or
unintentional disaster
• Population density
• GDP resiliency
• Traffic flows/borders
• First response capabilities
• Mutual aid capabilities
• Adversaries
• Built environment
• Engineering/redundancy
• Controls
• Maintenance
• Security
• Rate of infrastructure
recovery
• Operational changes
• Magnitude of disaster
• Number of Detectors
• Funding
• Pre- & post-disaster
coordination
• Interagency coordination
• Response planning
• Military response
• Management/
Communications
• Intelligence
• Counterintelligence
• Diplomacy
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Improved evacuation time
• Infrastructure hardening
• Security regimen
• Improved response time
• Reduce societal fear

8
Table 1, continued

Disaster Type Disaster Characteristics Risks and Uncertainties Disaster Risk and
Decision Analysis
Metrics
Disaster Risk and Decision
Analysis Levers
Disaster Risk and Decision
Analysis Objectives
Terrorist
acts
• Regional, national, or
global infrastructure
and disaster scope
• Infrastructure network
• Planned/existing
• Usually, little or no
warning
• Intentional disaster
• Population density
• GDP resiliency
• Traffic flows/borders
• GDP
• Supply &demand seasonality
• Traffic flows
• Regulations
• First response capabilities
• Mutual aid capabilities
• Adversaries/Security
• Built environment
• Engineering
• Redundancy/controls
• Maintenance
• Funding changes due
to new laws
• Rate of infrastructure
recovery
• Rate of throughput
recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Countermeasure
adjustment
• New laws
• Regulatory enforcement
• Funding
• Pre- & post-disaster
coordination efforts
• Interagency coordination
• Response planning
• Military response
• Management
• Communications
• Intelligence
• Counterintelligence
• Diplomacy
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Infrastructure hardening
• Security regimen (strategic)
• Optimized throughput
• Improved response time
• Optimized countermeasures
(tactical/incident)
• Reduce societal fear
Tornadoes/
thunder
storms
• Natural disaster
• Local or regional
infrastructure and
disaster scope
• Warnings of between
minutes and hours
• Random
• Frequencies are
characteristic of
geographic locations
and regions and/or
seasonality
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Weather
• Climate change
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Evacuation rate
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Regulatory enforcement
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Improved evacuation time

9
Table 1, continued

Disaster Type Disaster Characteristics Risks and Uncertainties Disaster Risk and
Decision Analysis
Metrics
Disaster Risk and Decision
Analysis Levers
Disaster Risk and Decision
Analysis Objectives
Transpor-
tation
accidents
• Natural, intentional,
or unintentional
disaster
• Local or regional
disaster scope
• Usually, little or no
warning
• Random, intentional,
or unintentional
• Planned/existing/
obdurate
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• Traffic flows
• First response capabilities
• Mutual aid capabilities
• Maintenance
• Engineering (e.g., roads, rail,
bridges)
• Adversaries
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Evacuation rate
• Rate of infrastructure
recovery
• Countermeasure/securi
ty/ control device
adjustment
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Regulatory enforcement
• Litigation
• Management
• Communications
• New laws
• Regulatory enforcement
• Interagency coordination
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Improved evacuation time
• Security regimen and/or
SCADA (strategic)
• Optimized
countermeasures/security/contr
ol devices (tactical/incident)
Tropical
storms/
hurricanes
• Natural disaster
• Regional
infrastructure and
disaster scope
• Warnings of between
24 hours to 5 days or
more
• Random
• Frequencies are
characteristic of
geographic locations
and/or seasonality
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Secondary disaster from storm
surge and/or floods
• Weather
• Climate change
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Evacuation rate
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Coastal zone policy and
management
• Regulatory enforcement
• Litigation
• Management/Communicatio
ns
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Improved evacuation time
• Coastal retreat and/or
hardening

10

Table 1, continued

Disaster
Type
Disaster Characteristics Risks and Uncertainties Disaster Risk and
Decision Analysis
Metrics
Disaster Risk and Decision
Analysis Levers
Disaster Risk and Decision
Analysis Objectives
Tsunamis • Regional infrastructure
and disaster scope
• Little or no warning
• Random
• Frequencies are
characteristic of
geographic locations
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Secondary disaster from
tsunami
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Evacuation rate
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Regulatory enforcement
• Litigation
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Improved evacuation time
• Reduce societal fear
Volcanoes • Regional infrastructure
and disaster scope
• Varies from little or no
warning to somewhat
predictable
• Random or periodic
• Frequencies are
characteristic of
geographic locations
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Secondary disaster from lahar
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Evacuation rate
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Regulatory enforcement
• Litigation
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Improved evacuation time
• Volcanic zone retreat
• Reduce societal fear
War/
insurg-
encies
• Regional, national, or
global infrastructure
and disaster scope
• Infrastructure network
• Planned/existing
• Usually, emergent
• Intentional disaster
• Population density
• GDP resiliency/GDP
• Traffic flows/borders
• Supply &demand seasonality
• Traffic flows
• Regulations
• First response capabilities
• Mutual aid capabilities
• Adversaries/Security
• Built environment
• Engineering
• Redundancy/controls
• Maintenance
• Military capabilities
• Funding changes due
to new laws
• Rate of infrastructure
destruction and
recovery
• Rate of throughput
destruction and
recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Countermeasure
adjustment
• Public reaction
• Morbidity/mortality
• New laws
• Regulatory enforcement
• Funding
• Pre- & post-disaster
coordination efforts
• Interagency coordination
• Response planning
• Military response
• Management
• Communications
• Intelligence
• Counterintelligence
• Diplomacy
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Infrastructure rebuilding
• Political/security regimen
(strategic)
• Optimized throughput
• Improved response time
• Optimized response
(tactical/incident)
• Minimal duration
• Minimal casualties
• Reduce societal fear

11
Table 1, continued

Disaster
Type
Disaster Characteristics Risks and Uncertainties Disaster Risk and
Decision Analysis
Metrics
Disaster Risk and Decision
Analysis Levers
Disaster Risk and Decision
Analysis Objectives
Wildfires • Natural disaster
• Regional disaster scope
• Little or no warning
• Random or intentional
• Frequencies are
characteristic of
geographic locations
and/or seasonality
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Weather
• Climate change
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Shelter rate
• Evacuation rate
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Energy policy and
management
• Regulatory enforcement
• Litigation/prosecution
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning
• Improved evacuation time
Winter
storms/
blizzards
• Natural disaster
• Regional disaster scope
• Warnings of at least 24
hours
• Random
• Frequencies are
characteristic of
geographic locations
and/or seasonality
• Population density
• Zoning enforcement
effectiveness
• GDP resiliency
• Regulations
• First response capabilities
• Mutual aid capabilities
• Weather
• Climate change
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response rate
• Shelter rate
• Evacuation rate
• Funding
• Pre- & post-disaster
coordination
• Response planning
• Energy policy and
management
• Regulatory enforcement
• Litigation
• Management
• Communications
• Improved delivery of rescue,
planning, preparation,
recovery, rebuilding, and
mitigation
• Improved response time
• Infrastructure hardening
• Adequate shelter in place pre-
planning

12
Figure 1: A Robust Model for Disaster Risk and Decision Analysis

13
Figure 2: A Robust Model for Disaster Risk and Decision Analysis—Iterative and Incremental

The model iteratively and incrementally evolves as time passes and as
conditions/phases emerge based on deterioration, mitigation and other factors.

Pre-Planning
Evacuation/
Mitigation
Rescue
Response/
Rebuilding
Recovery
Post Planning

14

Disasters are complex, rare events, usually local or regional, with severe negative
consequences that can result from intentional or unintentional actions or natural events.
1

Combinations of relatively rare triggering events and of the settlement patterns, building
conditions, and socioeconomic environment within these events’ impact zones result in
disasters.
2
Disasters, especially those that involve human choice, are unpredictable, non-
linear cascades of events whose consequences can interact in unforeseen ways that can
result in long-term indirect consequences.

Decisions about the risk of disasters require consideration of their potential range of
consequences and involve related, iterative, and incremental processes. These
processes—(1) fulfilling the disaster objectives from avoidance, prevention, mitigation,
response, and rescue to restoration, rebuilding, and recovery, (2) manipulating the levers
and tools that enhance or speed the accomplishment of these objectives, and (3)
monitoring and measuring these dynamic processes—strengthen our ability to mold the
future with confidence and enable us to determine more precisely the factors we can and

1
The triggering events that lead to disasters can be the result of human or non-human forces.
This distinction is significant from the point of view of intention (whereas one’s intentions can be
prevented by erecting barriers while nature’s “intentions” are largely stochastic). In another
sense, all disasters can be described as man-made insofar as disasters are events that negatively
and severely impact people and their built environment on an immediate and continuing basis.

2
Disasters are non-linear, polychronic, and netcentric events. A non-linear event is an event
whose net impact is not equal to the sum of the impacts of the event’s components considered
individually and may also include unanticipated, cascading impacts and strategic inflections
indicating chaotic behavior. Stated differently, a non-linear event is an event that does not adhere
to the superposition principle. A polychronic event is an event whose perceived duration cannot
simply be divided into a series of uniform sequential actions. A netcentric event is an event
wherein the networks of people, information, and activities that comprise an event can be used to
achieve better results, for example, more efficient disaster resource allocation and less severe
negative consequences in a disaster. The non-linearity, polychronicity, and netcentricity of
disasters are elements of these events’ complexity.

15

cannot control. Properly recognized and exploited, these processes can be used to
explore robust ranges of potential future disasters that can empower us to use our limited
mitigation resources more efficiently, to better plan and prepare for future disasters, and
to more efficiently allocate response, rescue, restoration, rebuilding, and recovery efforts.

Disasters can be triggered by:
• Intentional and unintentional impacts on single infrastructure components at
specific locales (e.g., an explosion of a tank of a flammable substance in Long
Beach, California or Los Angeles),
• Impacts of a natural disaster on an infrastructure system serving a specific locale
(e.g., the levees and floodwalls of New Orleans during a hurricane),
• Intentional impacts involving networks of infrastructure at specific locales (e.g.,
the delay of activities at the Ports of Long Beach/Los Angeles resultant from a
terrorism event), or
• Natural or intentional impacts involving networks of infrastructure over broader
regions (e.g., evacuation planning over a broad swath of the U.S. Gulf or Atlantic
coasts potentially impacted by hurricanes or planning to prevent
radioactive/nuclear attacks in California).

Risk and decision analysis for these types of disasters varies based on whether tractable
infrastructure currently exists, whether such infrastructure is planned, or whether existing
infrastructure suffers from obduracy (see Table 2).

16
Table 2: Disaster Cases

Disaster
Description
Disaster Characteristics Risks and Uncertainties Disaster Risk and Decision Analysis
Metrics
Disaster Risk and Decision Analysis
Levers
Disaster Risk and Decision
Analysis Objectives
Long Beach
SES LNG
Receiving
Facility Attack
(Chapter 2)
• Local infrastructure and
disaster scope
• Single infrastructure
• Planned
• Intentional disaster
• LNG demand/price
• Regulations
• First response capabilities
• Mutual aid capabilities
• Adversaries
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Response/evacuation rate

• Funding
• Pre- & post-disaster coordination
• Response planning
• Regulatory enforcement
• Litigation
• Management/Communications
• Infrastructure hardening
• Improved security
regimen
• Improved response time
• Improved evacuation time
Rancho LPG
Storage
Facility Attack
(Chapter 3)
• Local infrastructure and
disaster scope
• Single infrastructure
• Obdurate
• Intentional disaster
• LPG demand/price
• Regulations
• First response capabilities
• Mutual aid capabilities
• Adversaries
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Number of Injunctions
• Litigation/regulatory enforcement
• Funding
• Pre- & post-disaster coordination
• Response planning
• Management/Communications
• Infrastructure hardening
• Security regimen
• Improved response time
• Improved evacuation time
New Orleans’
Flood Control
System
Hurricane,
Flood, or
Sabotage
(Chapter 4)
• Local/regional
infrastructure and
disaster scope
• Infrastructure system
• Existing
• Non-intentional or
intentional disaster
• Weather/Climate change
• Regulations
• First response capabilities
• Mutual aid capabilities
• Adversaries
• River avulsion
• Funding changes due to new laws
• Rate of recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• New laws
• Regulatory enforcement
• Funding/Insurance
• Litigation
• Pre- & post-disaster coordination
• Response planning
• Management/Communications
• Infrastructure hardening
• Coastal hardening
• Improved response time
• Improved evacuation time
Ports of Los
Angeles and
Long Beach
Attack
(Chapter 5)
• Regional infrastructure
and disaster scope
• Infrastructure network
• Planned/existing
• Intentional disaster
• GDP
• Supply &demand
seasonality
• Traffic flows
• Regulations
• First response capabilities
• Mutual aid capabilities
• Adversaries
• Funding changes due to new laws
• Rate of infrastructure recovery
• Rate of throughput recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• Countermeasure adjustment
• New laws
• Regulatory enforcement
• Funding
• Pre- & post-disaster coordination
efforts/Interagency coordination
• Response planning
• Management/Communications t
• Infrastructure hardening
• Security regimen
(strategic)
• Optimized throughput
• Improved response time
• Optimized
countermeasures
(tactical/incident)
U.S.
Atlantic/Gulf
Hurricane
(Chapter 6)
• Regional infrastructure
and disaster scope
• Infrastructure network
• Planned
• Non-intentional
disaster
• GDP
• Weather
• Regulations
• First response capabilities
• Mutual aid capabilities
• Built environment
• Funding changes due to new laws
• Rate of infrastructure recovery
• Rate of throughput recovery
• Policy changes
• Operational changes
• Magnitude of disaster
• New laws/Regulatory enforcement
• Funding /Insurance
• Pre- & post-disaster coordination
• Interagency coordination
• Response planning
• Management/Communications
• Infrastructure hardening
• Security regimen
• Throughput
• Improved response time
• Improved evacuation time
California
Radioactive/
Nuclear Device
Attack
(Chapter 7)
• Regional infrastructure
& disaster scope
• Infrastructure network
• Planned
• Intentional disaster
• Population density
• Traffic flows/borders
• First response capabilities
• Mutual aid capabilities
• Adversaries
• Built environment
• Rate of infrastructure recovery
• Operational changes
• Magnitude of disaster
• Number of Detectors
• Funding
• Pre- & post-disaster coordination
• Interagency coordination
• Response planning
• Management/Communications
• Infrastructure hardening
• Security regimen
• Improved response time
17

The negative consequences that define disasters emerge from unique, complex
combinations of circumstances. Tools that iteratively measure and recalibrate these
negative consequences (or help prevent the occurrence from ever happening or assist in
softening the blow or helping to prevent cascading effects once the event occurs) can act
like smart levers which minimize the negative consequences influence the decisions that
impact (either prevent, ameliorate, or worsen) the initiating disaster event and the set of
posterior events from which the consequences emerge. The dominant theme of
minimizing negative consequences allows for unique approaches and solutions to the
potential mitigation of disasters, approaches and solutions that vary depending on
whether the disaster is an intentional or unintentional action, a natural event, or involves
varying perspectives of location, scope, and timing.

This study examines six case studies as exemplars of a robust model for disaster risk and
decision analysis whose details reveal the model’s dynamic definition of and approach to
the potential mitigation of disasters involving various perspectives of location, scope, and
timing. Each case study starts as a purposively-selected, representative time slice from a
selected, prospective disaster event. This static slice displays a selected possible future
that highlights certain decision possibilities and policy implications that elucidate the
case’s risk management.

Chapter 2 examines the case of a formerly proposed LNG Receiving facility in the Port
of Long Beach, California. This case is an example of a risk and decision analysis
involving an intentional disaster, a terrorist attack, on a local scope, directed at a single

18
unit of planned infrastructure.
3
Chapter 3 is similar to Chapter 2, except that the
infrastructure it considers is preexisting and an example of obdurate infrastructure.
4

Chapter 4 explores the impacts of unintentional and intentional disasters on an
infrastructure system. It looks at the impact of hurricanes, floods, and sabotage on the
New Orleans’ flood control system as an example of a risk and decision analysis on a
local scope, preexisting infrastructure system.
5

Chapter 5 looks at a massive regional infrastructure complex, the Ports of Long Beach
and Los Angeles, as a network that is responsive to a matrix of potential terrorist attack
risk scenarios that could emerge based on the current use of the complex and on the
current countermeasures used by the complex.
6

3
This chapter first appeared as a paper by Carl Southwell, “An Analysis of the Risks of a
Terrorist Attack on LNG Receiving Facilities in the United States,” Center for Risk and
Economic Analysis of Terrorism Events, University of Southern California. CREATE Report
Number 05-029, November 2005.

4
This chapter is also an unpublished paper by Carl Southwell, “In My Back Yard : A Case Study
of Institutional Obduracy in the Face of Public Safety or A Consequence Analysis of a Terrorist
Attack on the Rancho LPG Holdings, LLC Facility in San Pedro, California,” March 2011.

5
This chapter first appeared as a book chapter co-authored by Carl Southwell and Detlof von
Winterfeldt, “A Decision Analysis of Options to Rebuild the New Orleans' Flood Control
System,” Natural Disaster Analysis after Hurricane Katrina: Risk Analysis, Economic
Consequences, and Social Implications, Cheltenham, UK: Edward Elgar; July 2008.

6
Many of this chapter’s ideas have appeared in a series of ongoing white papers, seminars, and
software co-authored by Carl Southwell and including Michael D. Orosz (first author), Anthony
Barrett, Jennifer Chen, Petros Ioannou, Afshin Abadi, and Isaac Maya, “PortSec: A Port Security
Risk Analysis and Resource Allocation System,” IEEE Conference on Technologies for Homeland
Security, Waltham, MA, November 2010 and, more recently, contributions by Daniel Salazar and
Samrat Chatterjee.

19
Chapter 6 explores the potential for controlling a specific type of impact of a specific
type of unintentional disaster on a regional infrastructure system.
7
Mapping of real-time
hurricane tracks and their potential storm surge zones is overlaid with settlement
demographics in order to more accurately estimate potential evacuation timing along the
U.S. Gulf and Atlantic coasts.

Chapter 7 describes a method for the optimal allocation of domestic countermeasures for
radiological/nuclear terrorism detection in California.
8
This macro-method of allocation
for a diffuse, extreme risk uses nonlinear optimization to allocate finite resources over
weighted potential targets.

Risk and decision analysis primarily deals with the future. Inevitably, however, we have
no certitude about the future, only opinions.
9
Thus, the results of such analysis are never

7
This chapter also appeared as a paper co-authored by Carl Southwell and Detlof von
Winterfeldt, “Using Risk Assessment, Economic Assessment, and Risk Management to Improve
Preparedness for Terrorist Attacks and Natural Disasters: A Hurricane/Flood Consequence
Model for Estimating the Impacts of Qualitative Mitigations,” FEMA National Preparedness
Directorate, U.S. Department of Homeland Security, FEMA Grant No. 2008-GA-T8-K004, July
4, 2010.

8
This chapter originally appeared a part of a paper co-authored by Carl Southwell and Samrat
Chatterjee (first author), Isaac Maya, and Daniel Salazar, “Development of Risk-Based
Preventative Radiological/Nuclear Detection Resource Allocation Decisions for State of
California,” An unpublished report for the California Emergency Management Agency, February
7, 2011.

9
Taken to its extreme, this could devolve into a state wherein deliberative rhetoric would be the
only method appropriate for stating and discussing opinions. However, risk and decision analysis
offers methods for improvement beyond unsupported option through systematic argumentation
supported by a variety of quantitative and qualitative tools that rely on the information available
to the analyst.

20
beyond reasonable doubt. This tempts us to embrace methods or procedures that purport
to yield results immune to reasonable challenge (see, for example, Aristotle, 2005).

21
Chapter 2: An Intentional Disaster at a Single Unit of Planned
Infrastructure: A Case Study Involving the Long Beach SES LNG
Receiving Facility

Introduction
Our first example involves a case of local, singleton infrastructure in its planning stages.
The infrastructure’s disaster vulnerability may be most evident when considering an
intentional attack on this infrastructure or when considering similar attacks on pre-
existing infrastructure of similar design. Specifically, this chapter examines the case of a
formerly proposed LNG receiving facility in the Port of Long Beach, California. It is an
example of a risk and decision analysis involving a terrorist attack, on a local scope,
directed at a single unit of planned infrastructure. It demonstrates the importance of
limited levers—countermeasures and siting—in delimiting the potential scope of
potential disasters.

Background
Before the successes of the past five years in recovering domestic natural gas reserves
through hydraulic fracturing,
10
natural gas supplies had been low, and natural gas prices
high in the United States. The need for foreign natural gas imports was perceived as a
national priority. Four LNG receiving terminals were built in the contiguous United
States between 1971 and 1980. These facilities are in Lake Charles, Louisiana, Everett,

10
Given the potential for environmental degradation to and human health risks from groundwater
contamination due to released methane or injected fracking chemicals and air pollution via
released methane, the risks of hydaulic fracturing may ultimately spur future interest in LNG
importation. If hydraulic fracturing technology turns out to be relatively risk free, on the other
hand, the increased domestic production of natural gas may ultimately spur future interest in LNG
exportation.

22
Massachusetts, Elba Island, Georgia and Cove Point, Maryland. After reaching a peak
import volume of 253 billion cubic feet (Bcf) in 1979 that represented 1.3 percent of U.S.
gas demand, LNG imports declined in the 1980s and 1990s because of gas surpluses in
North America and price disputes with Algeria (CEC, 2003, p. 7). The increasing
demand for natural gas as a heating source, as a fuel source for electrical generation
facilities, and as an alternative fuel for internal combustion engines, the resurgence in
natural gas prices in the United States (U.S.) since 2000, and the current high cost of
petroleum spurred a renewed interest in establishing liquefied natural gas (LNG)
receiving facilities in the U.S. until about 2007 or 2008. As a result, during this window
from 2000 until 2008, interest in siting and building new U.S. LNG receiving facilities
intensified. During this period, there were six LNG receiving facilities in the United
States (see Table 3). Anticipating greater need, the Energy Policy Act of 2005 paved the
way for streamlined approvals of more, proposed LNG receiving terminals (CEC, 2005,
p. 1).

Nevertheless, the siting of LNG receiving facilities remained a concern because LNG is a
hazardous material. The extreme cold of LNG can cause direct injury or damage such as
dermal contact burns or the cracking of certain metals, for examples, copper and steel.
Exposure to a LNG vapor cloud, although non-toxic, can cause asphyxiation due to the
displacement of oxygen. But, the primary safety and security concern about LNG arises
from the potential consequences of an intentional, large LNG spill (ABSG, 2004, p. 2). .
Since spilled LNG disperses faster on the ocean than on land, water spills provide very
limited opportunity for containment. LNG vaporizes more quickly on water because

23
large bodies of water provide an enormous heat source. For these reasons, the risks
associated with shipping, loading, and off-loading LNG are generally considered much
greater than those associated with land-based storage facilities. Also, due to its practical
requirements as a floating mode of transport, an LNG tanker ships cannot be as fully
secured as a full containment land-based storage tanks.

For purposes of a potential terrorist attack on a LNG receiving facility, an analysis of this
risk focuses on four significant, potential hazards associated with a spill: a vapor cloud
explosion (VCE), a rapid phase transition (RPT), a pool fire, and a vapor fire.

A VCE refers to the detonation and ignition of a dispersed LNG vapor cloud, and a RPT
refers the “silent explosion” caused by the superheating and rapid expansion of LNG in
contact with water. In this analysis, the potential for VCEs will not be addressed since
there is no evidence that sufficient overpressure is created during an open LNG spill to
initiate the necessary conditions for a VCE.

The potential for RPTs are also beyond the scope of this analysis. Although RPTs are a
possibility, their potential severity is orders of magnitude fewer than the expected
maximum severities for the major hazards of interest: pool fires and vapor fires. Other
potential hazardous outcomes such as BLEVE and jet fires (ABSG, 2004, p. 2) were also
considered beyond the scope of this analysis because such outcomes are generally not
considered credible or of sufficient expected magnitude for large LNG releases. A
BLEVE (boiling liquid expanding vapor explosion) is most often associated with a fire
associated with a pressurized liquefied gas (e.g., propane or butane) contained in a

24
pressure vessel. A jet fire is usually associated with compressed or liquefied gases
released from storage tanks or pipelines and discharged through a hole or piping and
forming a gas jet flame.

In a pool fire, a pond of LNG is ignited by an open flame (or even the static electric
discharge from an arcing spark plug) and intensely burns in the atmosphere. A pool fire
produces significant thermal radiation from a quickly burning fire for distances of more
than a mile from the center of the pool of cryogenic liquid (ABSG, 2004, p. 3). Many
experts believe that the force necessary to breach a LNG containment vessel is so great
that it would cause a nearly instantaneous ignition of the vaporizing methane and quickly
form a pool fire, a conflagration of up to 3898 degrees Fahrenheit (NIST, 2005).

There is also the possibility that the vaporizing methane would not immediately ignite
due to varying ignition sparking characteristics, atmospheric turbulence, atmospheric
pressure, and/or the methane-atmosphere mixture. In such cases, the vapor cloud would
drift downwind until it warms up and mixes sufficiently with the air to either ignite in the
presence of an ignition source (when its volumetric percentage when mixed with air is
between 5% and 15%) or harmlessly disperse (ABSG, 2004, p. 3, 5). Depending upon
the local conditions and the volume of spilled LNG, the vapor fire could cover an area of
several square miles.

As a vapor fire burns, its flame could burn back toward the evaporating pool of liquid,
ultimately forming a pool fire. Thus, a vapor fire and a pool fire can be resultant from the
same event.

25

The Question of Costs
“[T]ypically large decisions are not made by a group of like-minded people…they are,
rather, the result of extended negotiations, either implicit or explicit, between
representatives of different points of view.” (de Neufville and Keeney, 1972) While the
economic benefits of natural gas delivered by LNG carriers are largely clear, the costs
associated with this specific vector for natural gas transmission are less clear, and
defined, in part, by risk analysis. As part of the dialogue of siting, an LNG receiving
facility risk analysis should address the most pressing questions of concern to
neighboring residents and businesses, namely: “What is a reasonable worst case pool or
vapor fire as a result of an intentional LNG spill?” and “What are its consequences?”

An attempt to answer these questions can be approached through a risk analysis that is
divided into three phases—hazard assessment, expected loss assessment, and conditional
expected loss assessment. The LNG hazard assessment identifies carrier-related and
land-based risks. Examples of carrier-related hazards are attacks on ships and ship
collisions; land-based hazards include earthquakes, industrial accidents, and terrorist
attacks.

The LNG expected loss assessment attempts to determine the likelihood of specific
events occurring based on the history of mechanical failures, accident rates, and factors
such as the probability of a terrorist event. In practice, some potentially severe risks are
deemed beyond the scope of analysis because the likelihood of these risks occurring is

26
sufficiently small (e.g., less than 10
-6
annually) or beyond the means of a specific project
to mitigate (e.g., the risk of a damaging meteorite strike).

Finally, the LNG conditional expected loss assessment evaluates of the selected worst-
case scenarios. The LNG receiving facility is assessed to determine its potential for
severe consequences such as loss of life, injury, and property based on site- and situation-
specific decisions.
In this chapter, plots of expected losses (which incorporate the probability of attacks at
the various facilities) versus conditional expected losses (i.e., losses assuming that attacks
have occurred) are constructed. These plots demonstrate the preferred risk reductions
options (or dominance relationships) for the three LNG facilities based on various siting
and defense choices which form a ranking of Pareto optimal options. These plots focus
the risk analysis on several risk reduction options, including:
• Alternative sites, the strengthening the containment walls, and aerial defenses for
the then proposed Long Beach SES LNG Receiving Facility (see CEC, 2005(1);
CEC, 2005(2); FERC, 2005(2); Quest, 2005; and Frick, 2005) and
• The use of strengthened containment walls and aerial defenses for the extant
Everett (Boston, Massachusetts) Tractebel LNG North America Receiving
Facility (see FERC, 2006 and Fay, 2003(2)) and the Providence (Rhode Island)
KeySpan LNG Receiving Facility (see Clarke, 2005; FERC, 2005(1)).

Recognizing that LNG safety, security, and siting is a complex problem helps formulate
the two main questions facing decision-makers from a public policy perspective—First, is

27
the siting of each facility appropriate? Second, are there countermeasures or alternative
siting options with respect to each facility that might be more appropriate? Within this
delimited decision arena, a severity ranking of sites can be constructed, a bordereau
concerning LNG containment, siting, and defense countermeasures can be crafted, and
the answers to these questions may begin to emerge. To this end, this chapter is about
how to be more realistic about the assessment of probabilities of future extreme events
and about their consequences.

Defining the Problem
What is LNG ?
LNG is the natural gas used in residences and businesses every day [basically, methane
(CH
4
) with small proportions of other hydrocarbon gases such as ethane (C
2
H
6
), propane
(C
3
H
8
), butane (C
4
H
10
), and even smaller proportions of non-hydrocarbon gases], except
that it has been refrigerated to minus 259 degrees Fahrenheit and become a buoyant,
colorless, and odorless liquid. As a cryogenic liquid, natural gas occupies about one six-
hundredth of its gaseous volume and can be transported relatively economically over long
distances in special, insulated containers (NIST, 2005; CEC, 2003).

When LNG comes in contact with any warmer surface such as water or air, it evaporates
and expands very rapidly to its more stable, gaseous state. If the LNG vaporizes under
relatively calm atmospheric conditions, a vapor cloud resembling ground fog forms. The
vapor cloud is initially heavier than air since it is so cold, but as it absorbs heat, it
becomes lighter than air, rises, and disperses with the wind. An LNG vapor cloud cannot

28
explode in the open atmosphere, but it can burn when it constitutes between five and
fifteen percent of air by volume in the presence of an ignition source (NIST, 2005;
Clarke, 2005).

LNG tanker safety and security design
Ocean-going tankers transport large amounts of LNG from liquefaction plants to
receiving terminals. These ships are equipped with LNG-cargo tanks housed inside
double-walled hulls. Each cargo tank can store thousands of cubic meters of LNG. These
ships are up to 1,000-feet long and require a minimum draft of up to 40 feet (CEC, 2003).

LNG tankers are equipped with specialized systems for handling the cryogenic gas and
for addressing the hazards associated with spills and fire. The ship’s safety systems are
divided into ship handling and cargo system handling. The ship-handling safety features
include radar and GPS that alert the crew to other traffic and hazards around the ship, and
automatic distress systems and beacons that transmit signals if the ship is in trouble.
Cargo-system safety features include an instrumentation package that shuts down the
system if it operates out of predefined parameters and gas- and fire-detection systems
(CEC, 2003).

LNG receiving terminal safety and security design
A shore-based LNG receiving terminal—consisting of a docking facility, LNG-storage
tanks, LNG-vaporization and -odorization equipment and vapor-handling systems—
occupies approximately 25 to 40 acres (CEC, 2003). An LNG terminal requires roads,
electric transmission lines, and gas and water lines. The docking facility must be

29
designed to accommodate the sizes of the anticipated LNG tankers. It usually consists of
a pier or jetty about 1,800-feet long and 30-feet wide with moorings and off-loading
facilities. Moorings connect the tanker securely to the jetty so that the LNG can be
transferred from the ship’s tanks to the onshore piping (Clarke, 2005).

While unloading their cargoes, LNG tankers are subject to tidal, wind, and wave forces
that can jeopardize the integrity of the mooring. To mitigate this, LNG ports and jetties
have built-in safety features to prevent releases of LNG during these transfers. A ship-to-
shore emergency shutdown (ESD) system and associated shut-off valves allow rapid and
safe shutdown of an LNG transfer. An ESD system will stop the ship’s unloading pumps
and close flow valves both on the ship and shore within 30 seconds. Moreover, quick-
release couplings automatically disconnect the unloading arms during emergencies (CEC,
2003).

Transfer piping used to unload the cryogenic liquid from the ship’s tanks can withstand
up to a 360 degree Fahrenheit temperature drop once LNG pump-out begins. Normally,
the cryogenic piping is made of stainless steel with a nine-percent nickel content.
Expansion loops and expansion bellows are built-in safety features that compensate for
pipeline contraction (CEC, 2003).

LNG is stored on land in specially designed storage tanks while it awaits regasification.
The failure of one or more tanks can release an enormous volume of LNG with severe
consequences due to the size of the resulting vapor cloud and potential fire. In response

30
to this hazard, the design of modern storage tanks has improved such that three types of
LNG storage tanks are used today as follows:
1 Single-containment tanks are double-walled. An interior tank is made of nine
percent nickel with an outer tank made of carbon steel.
2 Double-containment tanks have a primary nine percent nickel tank and a
secondary tank. The secondary tank, typically a concrete wall, is located within
twenty feet of the primary tank. In the event of a leak, the secondary tank
contains the cryogenic liquid and limits the surface area and vaporization from the
LNG liquid pool.
3 Full-containment tanks have a nine percent nickel inner tank, plus a pre-stressed
concrete outer tank. The outer tank, which includes a reinforced concrete roof
and floor lined with carbon steel, can be designed to withstand most impacts from
missiles or flying objects. These modern storage tanks have no side or bottom
openings. All penetrations, including those for LNG transfer, are through the
roof. This design reduces the amount of LNG spilled in the event of a rupture or
leakage. (CEC, 2005(2); Clarke, 2005)
In addition, LNG storage tanks contain instruments to monitor the pressure, temperature,
and density of the LNG along the entire height of the liquid column, and in-tank cameras
enable plant operators to assess tank damage, for example, in the event of an earthquake
and to visually inspect the tank contents in the event of unusual instrument readouts
(CEC, 2003).

31
Fire detection and response systems are also in place. Facility operators use low-
temperature, gas, fire, and smoke detectors. Fireproofing of structures and equipment are
mandatory safety features within LNG facilities (CEC, 2003).

LNG facilities are designed to assure adequate distances between LNG storage tanks,
storage areas, jetties and docks, vaporization process areas, and other parts of the facility.
LNG facilities have exclusion zones—the area surrounding an LNG facility in which an
operator legally controls all activities. These zones are intended to assure that public
activities and structures outside the immediate LNG facility boundary are not at risk in
the event of an on-site LNG fire or a release of a flammable vapor cloud. Federal
regulations identify two types of exclusion zones: thermal-radiation protection (from
LNG pool fires) and flammable vapor-dispersion protection (from LNG clouds that have
not ignited but could migrate to an ignition source) (CEC, 2003).
Thermal-radiation exclusion distances are determined by using the National Fire
Protection Association (NFPA) standard for the production, storage, and handling of
LNG, or by using a computer model that accounts for facility-specific and site-specific
factors, including wind speeds, ambient temperature, and relative humidity. The required
distances are intended to ensure that heat from an LNG fire inside the containment dikes,
for example, is not severe enough at the property line to cause death or severe burns
(CEC, 2003).

Safe distances from dispersing LNG vapor clouds are determined by the same NFPA
standards or by a computer model (for example, Areal Locations of Hazardous

32
Atmospheres (ALOHA)) that considers average gas concentration in air, weather
conditions, and terrain roughness. The permitting authority, the Federal Energy
Regulatory Commission (FERC), in cooperation with the U.S. Department of
Transportation’s (DOTs) Office of Pipeline Safety (OPS) and the U.S. Coast Guard
(USCG), determines the exclusion zones for LNG tankers and port facilities.

Risk Analysis—A Conditional Expected Losses Approach
The Long Beach SES LNG Receiving Facility
Sound Energy Solutions (SES), a subsidiary of Mitsubishi Corporation, applied to
construct, install, and operate an LNG receiving facility on 25 acres of Pier T in the port
of Long Beach, California (FERC, 2005(2); Frick, 2005). This facility would have been
about two miles from downtown Long Beach, a coastal city with a population of nearly
500,000. The facility would include a LNG ship berth with LNG unloading arms and
two full containment LNG receiving tanks, each with a gross liquid volume of 160,000
cubic meters (1,006,000 barrels). The receiving tanks are comprised of multiple
protective layers. The first layer or the inner tank is the primary containment barrier that
consists of steel alloy that contains 9% nickel. The second important layer is the outer
tank that is made of three feet of pre-stressed reinforced concrete. A 20-foot outer
reserve wall surrounds the LNG tank. The diameter is approximately 255 feet, and the
height is approximately 176 feet (FERC, 2005(2)). Our assumptions include an attack
from a large jet such as a Boeing 717-class airliner or larger accompanied by a
simultaneous land-based rocket propelled thermobaric grenade or equivalent (for
example, by using a Ruchnoy Protivotankoviy Granatomet (RPG)-TBG-7V or a SMAW

33
man-pad with a shaped charge (Clarke, 2005)) attack used as missiles projected into the
containers and/or the docked LNG ship. To improve the probability of success from the
terrorists’ points of view, two planes might be used in a coordinated attack. Jets of
comparable size and speed would likely be acquired out of Los Angeles International
Airport (LAX) or Long Beach Airport (LGB), both fewer than twenty miles from the
proposed site.

Research that simulates the crash of an aircraft into a concrete structure such as the full
containment storage tanks to be constructed at the Long Beach SES LNG receiving
facility has been conducted (see, for example, Sugano, 1993). Partially as a result of such
work, provisions for aircraft impact on reinforced concrete structures are incorporated
into the Civil Engineering codes used for the design of nuclear containment structures.
Indeed, Sugano’s study provides evidence that an airframe and the skin of an aircraft
alone are not likely to cause major structural damage on impacted reinforced concrete
targets. The most massive impacting element would be the aircraft’s fuel, as evidenced
by the Pentagon impact of September 11, 2001 (see Popescu, 2003). Simulations show
that the structural damage occurs only when the fuel mass hits. A fully fueled aircraft
strike coordinated with a thermobaric RPG or man-pad attack would have a high
probability of catastrophically breaching the LNG containment.

Selecting the frequency of a terrorist event at a specific locale in a given period of time is
difficult, if not impossible, because of the paucity of data. Estimates to model the
expected annual frequency of terrorist events at the proposed LNG facility was derived

34
by using data from the MIPT Terrorism Knowledge Base (MIPT, 2006). From this data,
we estimated that a range from around three to as many as 26 terror incidents that involve
property damage or bodily injury might occur annually in the United States with a mean
of about three or four (and of less than one of significant severity) at locations of
potential similar impact as the twin ports of Long Beach and Los Angeles. Assuming
that these would be among the 100 most probable targets nationwide (with a one percent
conditional probability of an incident) and that the LNG receiving facility and its docked
LNG tanker would be among the five most probable local targets within the port area
(with a twenty percent conditional probability), we further estimated that a binomial
distribution could serve as a reasonable base frequency. The selected, estimated rates of
terrorist attacks (or Bernoulli trial “successes” given a random draw once a year) on the
Port of Los Angeles and Long Beach was 1 in 230 annually, and, for the LNG facility, 1
in 1150 per annum (with sensitivity analysis conducted on a range of between 1 in
100,000 and 1 in 101).

Probabilities of a catastrophic attack by terrorists are influenced by siting (in particular,
the distance from the local built environment and population), hardening (as a barrier to
intrusion by missiles or explosives), and aerial and anti-terrorist defenses (as a deterrent
to air and other assaults) of the facility. In modeling LNG receiving facility options and
their expected consequences, many input quantities can only be estimated, and thus they
have an inherent degree of uncertainty. A model that allows for a range of uncertainty in
its inputs can provide more realistic and informative projections than static assessments.
Decision tree diagrams are a useful tool in mapping out the variables that influence the

35
potential consequences of decisions and events. In this analysis, Scenarios 1 through 6
model an assault by terrorists on this facility utilizing differing decisions with respect to
siting, hardening, and defense.

We modeled six possible fire scenarios resultant from various counter-terrorism
preparedness decisions and planned terrorist attacks perpetrated by an Al Qaeda-like or
domestic terrorist organization. The first two scenarios consider an attack on the
proposed Long Beach SES LNG receiving facility during different seasons, and the
second four scenarios consider an attack at the proposed facility with altered siting,
containment, and/or aerial defense countermeasures in the Long Beach harbor area and
during different seasons.

Based on the specific attack scenarios, simple cause-and-effect models were utilized.
Essentially, each analysis was broken down into the following steps: First, the attack
mode for the target includes a jet aircraft collision with a tank or a ship, a land-based
rocket-propelled thermobaric grenade (RPG) attack, or a similar strike at the same target,
the release of cryogenic liquid, and a vapor fire. Second, prevention and mitigation
efforts including, but not limited to, hardening of the facility, anti-aircraft and anti-
terrorist technology, and remote siting of the facility or its LNG ship dock affected the
probability of success or failure. Third, consequences of an attack included deaths,
injuries, destroyed property, and business interruption economic damage. For example,
in Long Beach, an expected, worst case economic value consequence of $83 billion was
selected by valuing each expected fatality at $5 million, each injury at $100,000, the

36
LNG facility at $400 million, an LNG ship at $240 million, local property damage within
the plume extent at $4.36 billion, and ongoing economic impacts at $50 billion (primarily
in terms of shot- and medium-term reduced port trade).

The procedures for estimating the worst-case consequences of a disaster at the Long
Beach SES LNG receiving facility are relatively simple—(a) assume the event is
intentional (i.e., a terrorist event) and (b) assume it is catastrophic (i.e., a near-
instantaneous complete release of the LNG due to trauma {e.g., using a jet or thermobaric
RPGs as missiles to breach the LNG tanks}). Using the capacity of a single LNG tank as
the spill size (note that the spill size could be twice as large if both tanks or a tank and the
ship were simultaneously breached, and note that the potential magnifying effects of
cascading impacts are ignored) and assuming a catastrophic breach, we employed the
software package known as CAMEO, a suite of software programs for planning and
response to chemical emergencies that was developed by the U.S. Environmental
Protection Agency's Office of Emergency Prevention, Preparedness and Response ( see
www.epa.gov/swercepp) and the National Oceanic and Atmospheric Administration's
Office of Response and Restoration ( see response.restoration.noaa.gov), to model a 5%
methane cloud’s extent or thermal radiation exclusion distance, assuming typical, local
atmospheric and landscape conditions.

CAMEO includes a set of databases, a gas dispersion model, ALOHA, and an electronic
mapping program called MARPLOT. Our model assumes a leak from a ten-foot
diameter hole in a spherical tank containing 200 million pounds of LNG. The resultant

37
plume extends four miles from the LNG tank, covering the entire downtown Long Beach
area and impacting thousands of residents and workers.

The ALOHA dispersion modeling includes the advection (moving) and diffusion
(spreading) of gases. A dispersing vapor cloud will generally advect in a downwind
direction and diffuse in a crosswind and vertical direction (crosswind is the direction
perpendicular to the wind). A gas cloud that is denser than air (known as a heavy gas)
can slightly spread upwind (Spicer, 1989). ALOHA models the dispersion of a gas in the
atmosphere and displays an overhead view of the area (footprint) in which it predicts gas
concentrations typically representative of hazardous levels called the Levels of Concern
(LOC). The footprint represents the area within which the concentration of a gas is
predicted to exceed a LOC at some time during the release. For example, the selected
LOC for methane in this paper is 50,000 ppm (or 5% of the atmosphere) or its Lower
Explosive Limit (LEL) (Spicer, 1989).

This analysis sought to utilize this standard model in a novel manner. ALOHA uses
simplified heavy gas dispersion calculations based on the DEGADIS model. Thus,
ALOHA’s results are unreliable under very low wind speeds, very stable atmospheric
conditions, wind shifts and terrain steering effects, or concentration patchiness,
particularly near the spill source, and ALOHA does not account for the effects of fires or
chemical reactions, particulates, chemical mixtures, and terrain. As such, the analysis
incorporated ALOHA because it uses DEGADIS “which was originally developed for the

38
USCG and the Gas Research Institute primarily for simulation of the dispersion of
cryogenic flammable gases” (Spicer, 1989).

A rough estimate of the potential damage of the modeled vapor fire can be calculated as a
result of estimating property and population demographics under the plume area and
imputing damage factors.

Note that, in our consequence modeling, we subordinated the potential for a pool fire
(despite the fact that its frequency of occurrence should be much higher than that of the
vapor fire) because we estimated that the maximum consequences of a pool fire would be
approximately 65 percent in Long Beach (estimated as 75% of the business interruption
loss and 50% of other losses) of the similar potential for a vapor fire (note that, due to the
facilities’ relative locations, this percentage in closer to 100% with respect to the Boston
and Providence facilities). That said, using Fay’s algorithm (Fay, 2003(1)) for the
derivation of the maximum radius, r, of a pool fire, defined as r = (ηQ/4πq)
.5
where η is
the fraction of the pool fire’s heat release that is emitted as thermal radiation (selected to
equal 0.150), Q is the heat release rate of LNG (selected as 1.450TW), and q is the
thermal radiative flux (selected as 25,000 kW/m
2
), we included the maximum Long
Beach pool fire diameter of about 2.35 miles (equivalent to 5.3 miles at 5000 kW/m
2

radiative flux, the FERC standard) in our modeling.

The tolerable levels for incident radiant thermal flux on off-site targets from LNG
facilities (49 CFR 193.2057) is set to a minimum of 5 kW/m
2
(or 1600 BTU/hr ft
2
) when
the “outdoor areas are occupied by 20 or more persons during normal use, such as

39
beaches, playgrounds, outdoor theaters, other recreational areas, and areas of public
assembly” (Raj, 1993). A 25,000 kW/m
2
flux refers to the border wherein ignition of
wood occurs without direct flame exposure (Clarke, 2005).

In order to evaluate the likeliest course of action by possible terror cells, we modeled a
range of decision trees, sensitivity analyses including tornado diagrams, and CAMEO
plumes to review the outcomes of various decisions. The decision trees represent the
choices involved in an attack given employed countermeasures and, given these choices,
the attacks’ probabilities of success or failure. For example, does choosing full
containment of the LNG tanks matter with respect to the probability of an attack? Given
that a facility’s tanks are fully contained, does a terrorist choose to attack a docked LNG
ship instead? If there’s an air defense battery at the LNG facility, how does this affect the
probability of an attack’s success? Given an air defense battery, what’s the probability of
an effective use of countermeasures given an attack? Given an ineffective use of
countermeasures, what’s the probability of a miss? And, given a hit, what’s the
probability of a direct hit versus a deflected hit (and what’s the impact, in terms of
consequences, of these contingencies)?

Figure 3 displays a decision tree from the terrorists’ perspective with respect to this
attack. The answers to five consecutive decision questions help determine the optimal
solution to minimize damage as follows: A. (from Figure 3, see notation A to note
graphically the portions of the decision tree that correspond to these questions) Are the
tanks full containment vessels? B. Is the facility protected by an air defense battery? C.

40
Given an attack, will there be sufficient time to use effective countermeasures? D.
Regardless, will the jet (or RPG) hit its intended target? E. If the target is hit, will the hit
be direct and most effective, or will it be a less damaging deflected hit?

In examining this decision tree, there are several relative probability-related observations
one can make. If the LNG tanks at the facility are not contained and the facility is
located such that it could potentially cause much damage, an attack on the LNG tanks is
preferred from the terrorists’ perspective. If the LNG tanks at the facility are contained
and the facility is located such that it could potentially cause much damage, an attack on
the LNG ship (close to the time it docks yet is still fully loaded) is preferred from the
terrorists’ perspective. If aerial or other defenses are utilized, the probability of an attack,
regardless of its type, is reduced.

Table 4 displays the base assumptions and ranges for thirty variables used in the Figure
2 decision tree. Judgmentally, we estimated that the increase in the probability of an
attack on an LNG tank would be twenty-fold (with a range between no increase and 100
times). The value of a life was selected as $5 million; the value of an injury, $100,000.
The estimated number of fatalities given an unimpeded, catastrophic attack was selected
as 5000, and the number of injuries was selected as 25,000. Similar costs, factors, and
numbers as well as their ranges were selected largely judgmentally and based on
reasonability constraints given the local built environment and population.

Figure 4 displays Figure 3 in its “rolled back” state using Table 4 base assumptions.
Rolling back a decision tree calculates the optimal decision based on particular outcomes.

41
To roll back a decision tree, from right to left, you multiply the probability of particular
outcomes by those outcomes’ consequences, and then add all of the products for a
particular consequence.

Scenario 1 uses decision trees to displays the discrete expected and cumulative values at
each state of the conditional sequences of events that describe terrorist attacks on the
formerly proposed Long Beach LNG receiving facility. Figure 12, under Chart Labels
D, A, E, and B, summarizes the original site proposal at Pier T assuming:
• fortified tank containment and aerial defenses,
• fortified tank containment and no aerial defenses,
• no fortified tank containment and aerial defenses, and
• no fortified tank containment and no aerial defenses,
respectively. The developer’s suggested alternative was the alternative suggested by
Chart Label A which develops an expected cost given a terrorist attack of about $75
billion or an expected cost of around $120 million per year.

Developing Scenario 1 further, the “optimal” decision from the public’s perspective is to
have full containment tanks with an aerial defense battery (see also Figure 15). Taking
into account factors such as:
• the expected probability of a terrorist attack on the LNG facility given tanks that
are fully contained,
• the expected increase in the probability of such an attack given that full
containment is not in place,

42
• the probability of a defensive counterattack given that defenses are in place,
• the probability of a successful attack despite defenses,
• the probability of a deflected hit on the facility despite a successful attack,
• the costs of defenses and containment,
• the choice between attacking the fixed tanks or the docked tanker, and
• the savings expected by the various mitigating circumstances,
the decision tree graphically displays the discrete expected and cumulative values at each
state of the conditional sequences of events. Within each decision to attack, there is a
sub-decision to attack either the larger, fixed tanks or the more vulnerable tanker. Such
tankers are equipped with LNG cargo tanks housed inside a double walled hull. To
unload, it takes an average of 12 to 15 hours. The probability for a successful breach
using an airplane and/or RPG directed at the tanker during the tanker’s port entry,
docking, and unloading process represents a substantial increase over the probability of a
facility full containment tank breach. Figure 5 displays the maximum methane plume
from this event.

High-end, expected losses from such an event (see Table 5) include 5000 fatalities,
25,000 injuries, and a total economic impact of about $83 billion.

Figure 6 is a tornado diagram of the decision tree. A tornado diagram is a representation
that depicts the most sensitive precedent variables of a decision tree along with their
impacts on the overall result. Figure 6 shows that the probability of an attack on a
facility and the increase in the probability due to non-containment of the fixed storage

43
tanks dominate the sensitivity of its results. The selected annual frequency probability
range of 0.0001 (1 in 10,000) to 0.0099 (1 in 101) accounts for most of the variability
among expected consequences.

As is further illustrated in Figure 9, as long as full containment decreases the probability
of attack by at least half, it is always preferred to non-containment. This supports the
validity of SES’s proposal to fully contain the tanks although this does not account for
the substitution potential of the LNG ship at dockside and before unloading.

Scenario 2 repeats Scenario 1, except that it assumes the event occurs in the
summertime with an expected plume as displayed in Figure 10. For purposes of the
Long Beach facility, we have assumed that total economic impacts are approximately
equal to those incurred in Scenario 1.

Scenarios 3 and 4 repeat the first two’s assumptions, except that the facility siting is
moved from Pier T to Pier J (away from downtown Long Beach, see Figure 9 and
Figure 10). Expected losses from such events include 500 fatalities, 2500 injuries, and a
total economic impact of about $31 billion. The reduction is primarily due to moving
the facility away from the population and densely built central city.

Scenarios 5 and 6 similarly repeat the first two scenarios’ assumptions, except that, in
these scenarios, a three-mile long jetty is built south of Pier J, and the facility is sited at
Pier J with its docking facility located at the ocean end of the jetty (see Figure 11 and
Figure 12). Expected losses from such an event include 50 fatalities, 250 injuries, and a

44
total economic impact of about $6 billion. This further reduction is due to the relative
futility of attacking the contained LNG tank and the further distance from the more
probable LNG ship target to the populated and heavily built infrastructure zones.

Figure 13 summarizes Scenarios 1 to 6. The first three columns of this table outline the
Long Beach site description, whether or not the fixed tanks are fully contained, and
whether or not aerial defenses are in place. The fourth column expresses the annualized,
expected cost of the set of assumptions wherein the cost is the combination of the
annualized expected value of a terrorist attack plus the cost of containment and
countermeasures. The fifth column lists the expected cost of a terrorist attack in the
given scenario, assuming that a successful attack, from the terrorists’ perspective, occurs.
Figure 14 summarizes Figure 13 by presenting its dominant strategies (i.e., the pairings
of lowest annualized costs and lowest expected value of terrorist attack, assuming that a
successful attack, from the terrorists’ perspective, occurs). The dominant strategies
represent the Pareto optimal choices that incorporate both expected and contingent
expected losses. Choices 1 through 8, respectively, represent reasonable siting,
containment, and countermeasure strategy options for policymakers’ consideration.

The Choices 1 through 8 vary based on three parameters as follows:
• Siting—either (1) on Pier J with the berth separated from the tanks by a jetty
perpendicular to the shoreline, (2) on an artificial island that has both a berth and
a tank facility, (3) on Pier T with the berth separated from the port by a jetty and
with a pipeline on port facilities from Pier J to Pier T, or either (4) on Pier J,

45
• Tank containment—either “Yes” or “No,” and
• Aerial defense batteries—either “Yes” or “No.”
The expected cost of an attack given that an attack has occurred varies inversely with the
cost of implementing the three options. Siting option (1) with tank containment and
aerial defenses returns the lowest expected cost of an attack given that an attack has
occurred, but it costs the most. Any of the these eight countermeasure strategies can be
supported as a dominant or preferred strategy based on the policymakers’ trade-off of
additional countermeasure costs for reduction on expected attack costs given a successful
terrorist attack.

Boston Tractebel LNG North America Receiving Facility
The seventh and eighth scenarios consider the Everett, Massachusetts Tractebel LNG
North America receiving facility utilizing the Long Beach SES LNG methodology. Use
of such decision analysis enables the analyst to consider containment and aerial defense
countermeasures that may enhance the security of an extant facility.

The Tractebel LNG North America facility in Everett, Massachusetts is located on the
Mystic River close to Boston Harbor and near the Tobin Bridge. Its proximity to the
Boston area has heavily influenced transportation security since September 11, 2001.
In formulating Scenario 7, the “optimal” decision from the public’s perspective, given
the assumptions presented, is to have full containment tanks with an aerial and anti-
terrorist defense battery at the Tractebel LNG North America facility. Taking into
account factors such as the expected probability of a summertime terrorist attack on the

46
LNG facility given tanks that are fully contained, the expected increase in the probability
of such an attack given that full containment is not in place, the probability of a defensive
counterattack given that defenses are in place, the probability of a successful attack
despite defenses, the probability of a deflected hit on the facility despite a successful
attack, the costs of defenses and containment, the choice between attacking the fixed
tanks or the docked tanker, and the savings expected by the various mitigating
circumstances, expected losses from such an event include 7000 fatalities, 36,000
injuries, and a total economic impact of $51 billion (see Table 6). Figure 15 displays the
expected methane plume from this event. Note that, for this event, we have assumed a
full breach of either the LNG tank or the carrier. Also, note that, in the event of a pool
fire, the expected losses could even exceed the vapor fire projections.

Scenario 8 repeats Scenario 7, except that it assumes the event occurs in the wintertime
with an expected plume as displayed in Figure 16. For purposes of the Boston facility,
the expected losses from such an event include 750 fatalities, 3750 injuries, and a total
economic impact of $16 billion.

Figures 15 and 16, respectively, represent identical siting, containment, and
countermeasure strategy options but have expected total economic losses that differ more
than three-fold based simply on the relative dates of occurrence. This demonstrates the
additional importance of seasonal wind and weather patterns in risk and decision
analysis. In the case of this facility, security should be bolstered in the summer to take

47
into account the increased potential for more severe consequences in the event of a
terrorism occurrence.

Providence KeySpan LNG Receiving Facility
The final two scenarios consider the Providence, Rhode Island KeySpan LNG facility
that was rejected on June 30, 2005, by FERC (and appealed on August 4, 2005, to FERC
with the denial upheld on January 19, 2006) by once again utilizing the Long Beach SES
LNG methodology. Use of such decision analysis enables the analyst to consider
upgrade countermeasures that may enhance this facility’s security.

KeySpan LNG, L.P. proposed to construct and operate upgrades to its existing
Providence, Rhode Island LNG facility, and Algonquin Gas Transmission, L.L.C.
proposed to site, construct, and operate a new natural gas pipeline and ancillary facilities
in Providence. The activities proposed by KeySpan LNG and Algonquin are known as
the KeySpan LNG Project. The KeySpan LNG Project would allow for the receipt of
marine LNG deliveries at the existing KeySpan LNG facility, augment LNG supplies for
truck deliveries to LNG storage tanks in the region, and supply up to 375 million cubic
feet per day (MMcfd) of imported LNG to the New England region (see FERC, 2005(1)
and Clarke, 2005).

The KeySpan LNG facility would also continue to deliver up to 150 MMcfd of vaporized
LNG to the New England Gas Company distribution system. In order to provide these
services, the KeySpan LNG Project requested authorization to construct, install, and
operate a ship unloading facility with a single berth capable of receiving LNG ships with

48
cargo capacities of up to 145,000 cubic meters, two 16-inch-diameter liquid unloading
arms, a 24-inch-diameter liquid unloading line from the arms to the LNG storage tank,
two vapor return blowers, a 12-inch-diameter vapor arm, an 8-inch-diameter vapor return
line, four boil-off-gas compressors, a boil-off gas condenser, a two-stage LNG pumping
system, an indirect fired vaporizer system with a capacity of 375 MMcfd, operations
control buildings, ancillary utilities, and LNG facilities (FERC, 2005(1)).

For Scenario 9, the Pareto optimal decision from the public’s perspective with respect to
the KeySpan LNG Project is to have full containment tanks with an aerial defense
battery. Taking into account factors such as:
• the expected probability of a summertime terrorist attack on the LNG facility
given tanks that are fully contained,
• the expected increase in the probability of such an attack given that full
containment is not in place,
• the probability of a defensive counterattack given that defenses are in place,
• the probability of a successful attack despite defenses, the probability of a
deflected hit on the facility despite a successful attack,
• the costs of defenses and containment, the choice between attacking the fixed
tanks or the docked tanker, and
• the savings expected by the various mitigating circumstances,
expected losses from such an event include 2000 fatalities, 10,000 injuries, and a total
economic impact of $18.3 billion (see Table 7). This compares to an “implied” estimate

49
of $42.3 billion (calculated as 8000 fatalities at $5 million apiece, 20,000 injuries at
$100,000 apiece, and $293 million in economic damages) derived from Clarke’s study
for Rhode Island (Clarke, 2005). Figure 17 displays the expected methane plume from
this event. Note that, for this event and similar to the Boston facility, we have assumed a
full breach of either the LNG tank or the carrier. Also, note that, in the event of a pool
fire, the expected losses could even exceed the vapor fire projections, especially since an
existing, nearby LPG facility could potentially cascade into this pool fire (Clarke, 2005).

Scenario 10 repeats Scenario 9, except that it assumes the event occurs in the wintertime
with an expected plume as displayed in Figure 18. For purposes of the Providence
facility, the expected losses from such an event include 500 fatalities, 2500 injuries, and a
total economic impact of $10 billion.

Figures 17 and 18, respectively, similar to Figures 15 and 16, represent identical siting,
containment, and countermeasure strategy options but have expected total economic
losses that differ almost two-fold based simply on the relative dates of occurrence. This
demonstrates again the importance of seasonal wind and weather patterns in risk and
decision analysis. In the case of this facility, security should be greater in the summer to
take into account the increased potential for more severe consequences in the event of a
terrorist attack.

Observations and Implications
Post-September 11, 2001, the possibility of a terrorist event at critical infrastructure in
densely populated areas must always be considered. In particular, attention should be

50
paid to scenarios that can cause massive human and economic losses in the event of a
successful attack.

Not unexpectedly, the decision analysis shows that the probability of attack on an LNG
facility is the most important variable in determining expected losses. Therefore, we also
analyzed expected losses conditional on an attack. A useful analysis examines the
dominant options in the expected loss vs. conditional expected loss plot (See Figure 14).
This analysis shows that tank containment is also an important factor because it is likely
to shift the terrorist’s target from the tank to the LNG tanker. Since the containment of
the tank facility decreases it utility as a target, a near perfect substitute for attack exists in
the form of the docked, not-yet-unloaded LNG ship. Thus, containment coupled with
remote siting of LNG ship docks accomplishes a more effective security regimen. And, it
is also clear that countermeasure costs including aerial and other anti-terrorist defense
measures are a major driver of the expected value of a terrorist act.

In Table 8, the first iteration for each locale displays its current (i.e., Boston and
Providence) or proposed (i.e., Long Beach) status. The current Boston facility indicates
about 60 percent the potential severity of the proposed Long Beach facility, and the
current Providence facility has about 20 percent of the Long Beach facility’s potential
(this compares with about 50 percent in the Clarke study). While these estimated average
severities are within the same order of magnitude, modest variations in local siting and
mitigation strategies as well as the localized effects on the extent of pool fires can affect
estimated severities significantly (by more than an order of magnitude).

51
From a policy perspective, this study demonstrates a few points. First, since attack
frequency cannot be determined, control of potential severity becomes a more tangible
goal. Looking at both unconditional and conditional expected losses can lead to
important insights. First, siting of LNG receiving facilities near other critical
infrastructure or near population centers should be avoided. Next, siting of LNG
facilities in areas with low densities at least four miles from population centers and at
least two miles from major shipping channels to minimize the exposure of people and
property to potential harm is strongly preferred in the regulatory approval process.
Moreover, full containment of facility tanks should be mandatory, and the potential for
separation of the LNG receiving facility from the LNG ship dock facilities by use of a
receiving jetty with mooring and berthing facilities should be considered, especially when
siting in higher density areas is preferred for economic or other strategic purposes.
Prevailing winds and precipitation patterns by season should also be considered on a site-
by-site basis in order to allocate additional security during times when expected
consequences would increase in the event of a terrorism event or an industrial accident.
Siting, containment, and defense strategies should incorporate multidisciplinary design
optimization techniques. Indeed, creative siting decisions might seek to be more regional
rather than local in their recommendations toward resolving the messy problem of LNG
receiving facility siting.

52
Table 3: Current U.S. LNG Receiving Facilities and their Annual Gas Imports (in
Billions of Cubic Feet)

Facility Importer(s) Location Annual Current
Import Capacity (Bcf)
Tractebel LNG North America Everett, Massachusetts 265.0
El Paso Corporation Elba Island, Georgia 161.0
Panhandle Energy/Trunkline LNG Lake Charles, Louisiana 230.0
Dominion Cove Point LNG Cove Point, Maryland 365.0
Excelerate Energy Gulf of Mexico (116 miles offshore of Louisiana) 183.0
EcoElectrica Puerto Rico 33.9

53
Figure 3: Long Beach SES LNG Receiving Facility Decision Tree—Scenario 1 Proposed
Location

A.
B.
C.
D. E.

54
Table 4: Long Beach SES LNG Receiving Facility Decision Tree Variables

No.
Variable Base Minimum Maximum
1
Probability of Attack as proposed (A) 0.0009 0.0001 0.0099
2
Probability of Counterattack|Defenses 0.65 0 0.8
3
Increase in Pr(Attack) due to non-containment 20 1 100
4
Cost of Containment ($Mlns) 50 0 150
5
Probability of Deflection|Hit 0.2 0 0.8
6
Cost of Defenses - Antiaircraft ($Mlns) 18 0 200
7
Factor Savings from Deflected Hit 0.2 0 0.4
8
Factor Savings due to Defenses 0.6 0 1
9
Probability of Hit|Ineffective Defenses 0.8 0 1
10
Factor Savings from Missed Antiaircraft Defense 0.3 0 0.5
11
Value of Life ($) 5,000,000 0 10,000,000
12
Value of Injury ($) 100,000 0 250,000
13
Probability of Attack at Pier J as a Factor of A 3 1 100
14
Probability of Attack at Quasi-offshore as a Factor of A 6 1 100
15 Probability of Attack at Pier T/Breakwater as a Factor of
A 6 1 100
16
Probability of Attack at Pier J/Breakwater as a Factor of A 6 1 100
17
Property Damages total costs ($Mlns) 5,000 0 10,000
18
Economic Damages total costs ($Mlns) 50,000 0 200,000
19
Other Damages total costs ($Mlns) 0 0 10,000
20 Other Costs (jetty-pipeline with original) total costs
($Mlns) 200 0 10,000
21
Other Costs (jetty with Pier J) total costs ($Mlns) 135 0 10,000
22
Other Costs (jetty and island) total costs ($Mlns) 225 0 10,000
23
Facility total costs ($Mlns) 400 0 500
24
Ship total costs ($Mlns) 240 0 250
25 Factor Savings of Costs by moving from proposed to Pier
J 0.5 0 1
26 Factor Savings of Costs by moving from proposed to
offshore 0.1 0 1
27
Number of Fatalities 5000 0 100,000
28
Number of Injuries 25,000 0 200,000
29 Factor Savings of Fatalities/Injuries by moving from
proposed to Pier J 0.1 0 1
30 Factor Savings of Fatalities/Injuries by moving from
proposed to offshore 0.01 0 1

55
Figure 4: Long Beach SES LNG Receiving Facility Decision Tree Scenario 1 Proposed
Location (Rolled Back given Base Assumptions)

56
Figure 5: Simulated Terrorist Attack on Proposed Long Beach LNG Receiving Facility

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Long Beach, CA during a typical winter, spring, or fall day
Full breach of either a facility tank or the LNG ship

Atmospheric conditions:
Wind: 6 mph from WNW at 3 meters height; Atmospheric stability Class: C
Air Temperature: 70° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information: Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Mass in Tank: 200,000,000 pounds
Circular Opening Diameter: 10 feet; Opening is 25 feet from tank bottom
Release Duration: 40 minutes (Heavy Gas Model—Two Phase Flow)
Max Average Sustained Release Rate: 7.79e+06 pounds/min
Max Threat Zone: 4.3 miles

57
Table 5: 2005 Demographic Profile of Area within Various Distances of Long Beach
SES LNG Facility

2-mile Radius 3-mile Radius 5-mile Radius
95
th
percentile
Plume extent
Total Population 7,743 85,124 408,860 275,000
Total Households 3,033 29,246 136,051 90,000
Median Household
Income
$26,547 $27,037 $37,150 n/a
Majority Ethnic
Group
Hispanic 55.4% Hispanic 65.2% Hispanic 54.4% n/a
Total Businesses 893 3,822 11,235 8,000
Total Employees 16,085 44,037 113,855 85,000
Total Estimated
Fatalities from
Attack (%
Population)
n/a n/a n/a
5,000
(1.8%)
Total Estimated
Injuries from Attack
(% Population)
n/a n/a n/a
25,000
(9.1%)
Total Economic
Losses from Attack
(excluding
Fatalities and
Injuries)
n/a n/a n/a $54.36 billion

Sources: Safety Advisory Report On the Proposed Sound Energy Solutions Liquefied Natural Gas Terminal at the Port of
Long Beach, California, Staff Paper, California Energy Commission Publication Number CEC-600-2005-033, September 7,
2005 (p. 8) and author.

58
Figure 6: Tornado Diagram—Long Beach SES Receiving Facility Decision Tree

59
Figure 7: 2 Way Sensitivity Diagram—Long Beach SES Receiving Facility Decision
Tree

60
Figure 8: Simulated Terrorist Attack on Proposed Long Beach LNG Receiving Facility
(during summer)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Long Beach, CA during a typical summer day
Full breach of either a facility tank or the LNG ship

Atmospheric conditions:
Wind: 7 mph from S at 3 meters; Atmospheric stability Class: D
Air Temperature: 70° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Mass in Tank: 200,000,000 pounds
Circular Opening Diameter: 10 feet; Opening is 25 feet from tank bottom
Release Duration: 40 minutes
Max Average Sustained Release Rate: 7.79e+06 pounds/min (averaged over a minute or more)
Total Amount Released: 1.95e+008 pounds; Max Threat Zone: 4.2 miles

61
Figure 9: Simulated Terrorist Attack on Proposed Long Beach LNG Receiving Facility
at Pier J Alternative Site (during winter, spring, or fall)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Full breach of either a Facility Tank or the LNG Ship

Atmospheric conditions:
Wind: 6 mph from WNW at 3 meters height; Atmospheric stability Class: C
Air Temperature: 70° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Mass in Tank: 200,000,000 pounds
Circular Opening Diameter: 10 feet; Opening is 25 feet from tank bottom
Release Duration: 40 minutes (Heavy Gas Model—Two Phase Flow)
Max Average Sustained Release Rate: 7.79e+06 pounds/min
Max Threat Zone: 4.3 miles

62
Figure 10: Simulated Terrorist Attack on Proposed Long Beach LNG Receiving Facility
at Pier J Alternative Site (during summer)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Full breach of Either a Facility Tank or the LNG Ship

Atmospheric conditions:
Wind: 7 mph from S at 3 meters; Atmospheric stability Class: D
Air Temperature: 70° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Mass in Tank: 200,000,000 pounds
Circular Opening Diameter: 10 feet; Opening is 25 feet from tank bottom
Release Duration: 40 minutes
Max Average Sustained Release Rate: 7.79e+06 pounds/min (averaged over a minute or more)
Total Amount Released: 1.95e+008 pounds
Max Threat Zone: 4.2 miles

63
Figure 11: Simulated Terrorist Attack on Proposed Long Beach LNG Receiving Facility
at Offshore Island or Jetty with Pier Alternative Site (during winter, spring, or
fall)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Full Breach of either a Facility Tank or the LNG Ship

Atmospheric conditions:
Wind: 6 mph from WNW at 3 meters height; Atmospheric stability Class: C
Air Temperature: 70° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Mass in Tank: 200,000,000 pounds
Circular Opening Diameter: 10 feet; Opening is 25 feet from tank bottom
Release Duration: 40 minutes (Heavy Gas Model—Two Phase Flow)
Max Average Sustained Release Rate: 7.79e+06 pounds/min
Max Threat Zone: 4.3 miles

64
Figure 12: Simulated Terrorist Attack on Proposed Long Beach LNG Receiving Facility
at Offshore Island or Jetty with Pier Alternative Site (during summer)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Full Breach of either a Facility Tank or the LNG Ship

Atmospheric conditions:
Wind: 7 mph from S at 3 meters; Atmospheric stability Class: D
Air Temperature: 70° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Mass in Tank: 200,000,000 pounds
Circular Opening Diameter: 10 feet; Opening is 25 feet from tank bottom
Release Duration: 40 minutes
Max Average Sustained Release Rate: 7.79e+06 pounds/min (averaged over a minute or more)
Total Amount Released: 1.95e+008 pounds
Max Threat Zone: 4.2 miles

65
Figure 13: Expected Cost of a Terrorist Attack Given that a Terrorist Attack Occurs (in $
Millions)

Scenario
Facility
containment?
Aerial
defenses?
Expected value ($
Millions)
Expected cost |
terrorist attack ($
Millions) Chart Label
Pier T (current proposal) Y Y $ 111.0 $ 47,875 D
Pier T (current proposal) Y N $ 117.6 $ 75,178 A
Pier T (current proposal) N Y $ 879.4 $ 48,372 E
Pier T (current proposal) N N $ 1,354.9 $ 75,273 B
Pier J Y Y $ 83.8 $ 17,616 J
Pier J Y N $ 74.8 $ 27,626 F
Pier J N Y $ 334.3 $ 17,589 K
Pier J N N $ 497.7 $ 27,649 G
Artificial Island Y Y $ 295.9 $ 3,480 R
Artificial Island Y N $ 279.5 $ 5,283 N
Artificial Island N Y $ 301.2 $ 3,476 T
Artificial Island N N $ 317.8 $ 5,378 P
Pier J with a jetty Y Y $ 205.9 $ 3,390 Q
Pier J with a jetty Y N $ 189.5 $ 5,193 M
Pier J with a jetty N Y $ 340.0 $ 10,544 L
Pier J with a jetty N N $ 632.7 $ 27,784 H
Pier T with a jetty/pipeline Y Y $ 270.9 $ 3,455 S
Pier T with a jetty/pipeline Y N $ 254.5 $ 5,258 0
Pier T with a jetty/pipeline N Y $ 677.8 $ 25,762 I
Pier T with a jetty/pipeline N N $ 1,554.9 $ 75,473 C

66
Figure 14: Dominant Strategies Given Base Assumptions

Indicated Strategies
Choice Scenario
Facility
containment? Aerial defenses?
Expected cost ($
Millions)
Expected cost |
terrorist attack ($
Millions)
1 Pier J with a jetty Y Y $205.9 $3,390.0
2 Pier J with a jetty Y N $189.5 $5,193.0
3 Artificial Island Y Y $265.9 $3,450.0
4
Pier T with a
jetty/pipeline Y Y $270.9 $3,455.0
5 Artificial Island Y N $249.5 $5,253.0
6
Pier T with a
jetty/pipeline Y N $254.5 $5,258.0
7 Pier J Y Y $83.8 $17,615.0
8 Pier J Y N $74.8 $27,625.0

67
Table 6: 2000 Demographic Profile of Area within Various Distances of Everett
Tractebel LNG N.A. Facility

2-mile Radius 3-mile Radius 5-mile Radius
95
th
percentile
Plume extent
Total Population 162,205 375,690 788,600 50,000
Total Households 66,843 156,778 333,356 20,000
Majority Ethnic
Group
White 75.9% White 73.3% White 73.2% n/a
Total Employees 81,134 198,288 413,723 25,000
Total Estimated
Fatalities from
Attack (%
Population)
n/a n/a n/a
7,000
(14%)
Total Estimated
Injuries from Attack
(% Population)
n/a n/a n/a
36,000
(72%)
Total Economic
Losses from Attack
(excluding
Fatalities and
Injuries)
n/a n/a n/a $12.4 billion
Source: Landview 6 Census 2000 Population Estimator.

68
Figure 15: Simulated Terrorist Attack on Boston LNG Receiving Facility (during
summer)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source During a Typical
Summer Day
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Full breach of either Facility Tanks or LNG Ship at Mooring

Atmospheric conditions:
Wind: 14 mph from SW at 2 meters height; Atmospheric stability Class: D
Air Temperature: 75° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Volume in Tank: 42 million gallons
Circular Opening Diameter: 5 meters; Opening is 77.2 feet from tank bottom
Release Duration: 49 minutes (Heavy Gas Model—Two Phase Flow)
Max Average Sustained Release Rate: 1.39e+07 pounds/min
Total Amount Released: 1.07e+008 pounds
Max Threat Zone: 2.9 miles

69
Figure 16: Simulated Terrorist Attack on Boston LNG Receiving Facility (during winter)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Full Breach of either Facility Tanks or LNG Ship at Mooring during a Typical Winter Day

Atmospheric conditions:
Wind: 11 mph from NW at 2 meters height; Atmospheric stability Class: D
Air Temperature: 40° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Volume in Tank: 42 million gallons
Circular Opening Diameter: 5 meters; Opening is 55.1 feet from tank bottom
Release Duration: 47 minutes (Heavy Gas Model—Two Phase Flow)
Max Average Sustained Release Rate: 1.45e+07 pounds/min
Total Amount Released: 1.25e+008 pounds
Max Threat Zone: 2.7 miles

70
Table 7: 2000 Demographic Profile of Area within Various Distances of Providence
KeySpan LNG Facility

2-mile Radius 3-mile Radius 5-mile Radius
95
th
percentile
Plume extent
Total Population 71,784 162,875 356,832 33,000
Total Households 26,589 61,425 136,782 10,000
Majority Ethnic
Group
White 56.5% White 62.9% White 71.7% n/a
Total Employees 29,266 69,964 157,148 12,000
Total Estimated
Fatalities from
Attack (%
Population)
n/a n/a n/a
2,000
(6.1%)
Total Estimated
Injuries from Attack
(% Population)
n/a n/a n/a
10,000
(30.3%)
Total Economic
Losses from Attack
(excluding
Fatalities and
Injuries)
n/a n/a n/a $7.3 billion
Source: Landview 6 Census 2000 Population Estimator.

71
Figure 17: Simulated Terrorist Attack on Rhode Island LNG Receiving Facility (during
summer)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Full Breach of Facility Tank and LNG Ship at Mooring during a Typical Summer Day

Atmospheric conditions:
Wind: 14 mph from SW at 2 meters height; Atmospheric stability Class: D
Air Temperature: 75° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Volume in Tank: 50 million gallons
Circular Opening Diameter: 5 meters; Opening is 81.8 feet from tank bottom
Release Duration: 56 minutes (Heavy Gas Model—Two Phase Flow)
Max Average Sustained Release Rate: 1.5e+07 pounds/min
Total Amount Released: 1.07e+008 pounds
Max Threat Zone: 3.2 miles

72
Figure 18: Simulated Terrorist Attack on Rhode Island LNG Receiving Facility (during
winter)

Maximum Extent of 5% Methane Plume Assuming No Ignition Source
Unshaded Plume Areas Indicate 95
th
Percentile Expected Plume Range
And Maximum Extent Pool Fire-related Boundary at Thermal Radiative Flux = 25kW/m
2

Full Breach of Facility Tank and LNG Ship at Mooring During a Typical Winter Day.

Atmospheric conditions:
Wind: 11 mph from NW at 2 meters height; Atmospheric stability Class: D
Air Temperature: 40° F; Relative Humidity: 50%
Ground Roughness: urban; Cloud Cover: 50%
Spill information:
Chemical: METHANE; Molecular Weight: 16.04 g/mol
TEEL-3: 50000 ppm
Normal Boiling Point: -258.7° F; Chemical Volume in Tank: 50 million gallons
Circular Opening Diameter: 5 meters; Opening is 58.4 feet from tank bottom
Release Duration: 14 minutes (Heavy Gas Model—Two Phase Flow)
Max Average Sustained Release Rate: 1.92e+07 pounds/min
Total Amount Released: 1.49e+008 pounds
Max Threat Zone: 3.9 miles

73
Table 8: Relative Severities of Potential Terrorist Acts at Selected U.S. LNG Receiving
Facilities

Millions of Dollars
(a) (b) (c)
Locale Option
Full
Containment
Tanks?
Defenses
Implemented?
Annualized
Costs
Expected Size
of Terrorist
Event Given
an Attack
Factor of
L.B.
Proposed
(b)
Long Beach Proposed [Pier T] Yes No $117.60 $75,178 1.00
Boston Current - summer No No $828.90 $46,049 0.61
Boston Optimal (a) Yes No $91.44 $46,099 0.61
Boston Optimal (b) Yes Yes $94.37 $29,371 0.39
Providence Current - summer No No $296.00 $16,448 0.22
Providence Optimal (a) Yes No $64.80 $16,498 0.22
Providence Optimal (b) Yes Yes $77.40 $10,535 0.14
Long Beach Optimal (a) [Pier J] Yes No $74.80 $27,625 0.37
Long Beach
Optimal (b) [Pier J
with jetty] Yes Yes $205.90 $3,390 0.05

74
Chapter 3: An Intentional Disaster at a Single Unit of Pre-existing
Infrastructure: A Case Study Involving the Rancho LPG Storage
Facility

Introduction
The second example involves a case of local, preexisting singleton infrastructure. Its
disaster vulnerability may be most evident when considering the consequences of an
intentional attack on this infrastructure. An existing LPG storage facility near the Port of
Los Angeles, California is considered as an example of a risk and decision analysis
involving an intentional disaster, a terrorist attack, on a local scope, directed at a single
unit of preexisting and obdurate infrastructure. Its lessons are similar to those in Chapter
2, but public policy decisions, zoning, public takings, and physical improvements are the
levers of disaster containment.

Background
Because of its adjacency to residential neighborhoods, the potentially explosive Rancho
LPG Holdings, LLC butane storage facility (Rancho LPG) in San Pedro, California has
been controversial since it was built in the early 1970s (Pryor, 1977). From its beginning,
local residents have opposed its siting and have continued to voice concern about its
operation. This chapter reviews some aspects of this facility, outlines potential
consequences of a catastrophic scenario, and discusses potential mitigations.

The history of Rancho LPG began about thirty years before it was built. After the U.S.
entered World War II, Huber Smutz, the City of Los Angeles’ zoning administrator,

75
observed that scores of Los Angeles area businesses wanted variances to expand where
and how they could conduct their operations:

“Most of these applicants insisted that their matter was absolutely essential
to winning the war—that they had to have their building permit, power
connection permit, or license right now without any foolishness about silly
zoning requirements—that they could get letters from the Army, Navy,
Air Corps, or even in some cases from the President himself to show the
importance of their proposal, and, anyway, what the heck good was our
zoning going to be if the Japs started bombing our city or if we lost the
war, as we probably would if their plant couldn’t expand or be
established.” (Smutz, 1942)

City Council passed a war emergency ordinance that permitted the zoning administrator
to grant variances until six months after the war emergency had ended for businesses
engaged in making materials or providing services that were essential to the war effort
11
.
(Scott, 1995)

During the war, parcels along North Gaffey Street in San Pedro, California
12
were re-
zoned “heavy industrial” under the City of Los Angeles’ war emergency regulations
because vacant land for industry near the Port of Los Angeles was considered a national

11
Unfortunately, these war emergency variances were approved without limitations. Some
municipalities that passed war emergency ordinances, for example, Portland, had built-in
provisions for repeal.

12
San Pedro is the familiar name used for the area of the City of Los Angeles adjacent to the Port
of Los Angeles and mainly located on the Palos Verdes Peninsula.

76
security priority (Murphy, 1989). Such permanent re-zoning allowed facilities such as
the Union Oil Company of California refinery
13
to be built in 1952.

Since liquefied petroleum gas (LPG) is a byproduct of refining, the building of oil
refineries in the port area created a demand for an LPG processing and storage facility.
In 1972, Petrolane constructed what is the current Rancho LPG to meet this demand.
Although this facility, the largest LPG storage facility in the United States,
14
was built to
the most current engineering and safety specifications, local residents opposed the facility
from its start based on its proximity to residential housing.

After the Petrolane LPG facility was completed in 1974, a local resident sued Petrolane,
claiming the facility was a public nuisance that should be removed. The legal case,
Brown v. Petrolane, Inc. (1980) 102 Cal.App.3d 720 8, ruled in favor of the facility’s
owner. The plaintiff had alleged that the potentially explosive facility was located in an
area of recurring seismic activity, factors that combined to create a site and scenario with
significant potential for a catastrophe.

13
This is the current ConocoPhillips Los Angeles refinery that has a crude oil processing
capacity of 139,000 barrels per day and processes mainly heavy, high-sulfur crude oil. It receives
domestic crude oil via pipeline from California and both foreign and domestic crude oil by tanker
through a third-party terminal at the Port of Long Beach. The refinery produces transportation
fuels, such as gasoline, diesel fuel, and jet fuel. Other products include fuel-grade petroleum
coke. The refinery also produces California Air Resources Board (CARB) gasoline using ethanol
to meet government-mandated oxygenate requirements.

14
The scale of the facility is such that it is the largest LPG facility in the U.S. It stores enough
butane to provide for the equivalent of about 15% of California's demand. The cost to build a
new, similar facility at an appropriate location and up to current code would probably approach
$200 million.

77
The court curiously ruled that this plaintiff’s fear was a fear that differed between
individuals within the community, if at all, in degree rather than kind. It was decided
that, under California’s Civil Code Section 3493, a plaintiff must allege a special injury
different in kind from that suffered by the general public in order to seek and obtain a
public nuisance declaration.
15

In 1989, the Federal Clean Air Act Amendments of 1990 mandated a Reid vapor pressure
(RVP) maximum of 9.0 per square inch (psi)
16
from June 1 to September 15 annually.

15
Almost twenty years later, when jet fuel storage tanks were being constructed at John Wayne
Airport near a residential neighborhood, a similar action was filed under Koll-Irvine Center
Property Owners Assn. v. County of Orange (1994). Brown v. Petrolane was cited as an almost
identical action with the same result. However, it was alleged that the facts would support an
abatement action by a public entity, but not a private party action based on either a public or
private nuisance theory. The main point is that the California courts have said that the potential
condemnation of the facility is simply not going to happen based on private parties suing for
nuisance. Instead, the courts have hinted that condemnation of the facility will occur only if a
public entity pursues it. And, of course, this probably hasn't happened because a government
taking is costly.

16
Gasoline is blended, largely to calibrate its RVP (this helps control evaporation) and octane
(this helps control engine knocking). The RVP is the vapor pressure of gasoline (or a gasoline
blend) at 100 ºF. If a liquid has a vapor pressure of greater than normal atmospheric pressure, the
liquid boils. In summer, when temperatures can exceed 100 ºF in many locations, it is important
that its RVP is below 14.7. Otherwise, gas tanks and gas cans can pressurize, and open
containers of gasoline can boil. Also, the boiling gasoline can produce vapors that contribute to
air pollution. Recognizing this, the EPA declared that summer gasoline blends may not exceed
7.8 psi in some locations, and 9.0 psi, in others. The particulars vary, but key considerations are
the altitude and motor vehicle density of a specific location. As gasoline evaporates, volatile
organic compounds (VOC’s) enter the atmosphere and contribute to ozone formation. In order to
control VOC emissions, the Federal Clean Air Act Amendments of 1990 require that all gasoline
be limited to an RVP maximum of 9.0 psi during the summer high ozone season, which the
Environmental Protection Agency (EPA) established as running from June 1 to September 15.
Current California RVP standards are an RVP flat limit of 7.0 psi for oxygenated fuels and a flat
limit RVP of 6.9 psi for non-oxygenated fuels. Butane, which has an RVP of 52 psi, is blended
into gasoline in higher proportions in the winter because the vapor pressure allowance is higher.
There are two advantages in doing this. First, butane is a cheap component compared to most of
gasoline’s other ingredients. It also adds to the total gasoline production potential of the refinery.

78
This immediately increased the demand for the services of Rancho LPG since the
proportion of high RVP butane in gasoline blends significantly impacted gasoline’s RVP,
butane was a refinery byproduct
17
, and places were needed to store the excess butane that
was no longer blended into gasoline, especially during the summer. This law helped
transform the facility into, primarily, a butane export facility—the type of facility it
remains to this day. The current Rancho LPG facility collects excess butane in the
summer, stores it for futures’ pricing/demand purposes, and ships it when prices peak
(during times of greater demand such as the winter). Although butane, normally one of
the less valuable petroleum byproducts that is sold as a liquid, is more valuable to
refinery owners when it can be blended into gasoline (since gas currently retails at around
$3.50 per gallon), it is still valuable enough to keep and ship elsewhere as a standalone
product (because it retails for about $2.00 per gallon
18
). Maximizing butane profits
involves coordinating the timing of its storage and export. There are mainly two
asynchronous curves that describe the facility’s current butane storage and export—one
for production; the other, demand. The exact facility storage and export figures are
proprietary, but, theoretically, 678 railcars of butane are stored at the facility. Based on

17
At refineries, LPG is collected in the first phase of refinement or crude distillation. The crude
oil is run through a distillation column in which a furnace heats it at high temperatures. During
this process, vapors rise to the top, and heavier crude oil components fall to the bottom. As the
vapors rise, cooling and liquefying occur on “bubble trays,” aided by the introduction of naphtha.
The liberated gases are recovered to manufacture LPG.

18
If the butane weren’t exported, the current alternative use would probably be burning it off by
refinery flares. Or, another use might be fueling butane-powered electric generators located at or
near refinery sites with the generated power sold to the grid (perhaps this would be a good reuse
of the site). This has not been seriously considered because it is probably less economically
efficient (in part, perhaps, because the full cost of risk isn't properly factored into the current
“store it and then export it” option.)

79
its “daily average” figures reported to its local Certified Unified Program Agency
(CUPA), the average actual facility volume is approximately 380 railcars of butane.
Assuming one daily work shift at the facility, its maximum LPG railcar turnover is about
twelve with a similar number of trucks. Since the facility’s permit for marine shipping
has expired, all of its export business is currently by rail and truck.

While this profit motive and its associated timing and storage motivations are clear, the
costs associated with butane transmission via large aboveground storage facilities such as
Rancho LPG are less clear, and can be defined, in part, by risk analysis. First, a Rancho
LPG risk analysis should address the most pressing questions of concern to adjacent
residents and businesses, namely: “What are the probable worst case vapor fire, pool fire,
and boiling liquid evaporative vapor explosion (BLEVE) as a result of an LPG spill from
this facility ?” and “What are its consequences ?” This report is not a comprehensive risk
analysis.
19
Rather, it is a consequence analysis of a probable, though rare, event that
could occur at Rancho LPG that seeks to assess realistically the potential probability,
consequences, and other implications of this future extreme event.

19
An adequate, comprehensive quantitative risk analysis (QRA) of Rancho LPG has never been
conducted, in part, because legal and regulatory requirements have never demanded such a study,
and such an undertaking is very large in scope. A QRA would attempt to determine the
likelihood of specific negative consequence events occurring based on expected rates associated
with mechanical failures, accidents, and rare events such as severe earthquakes, tsunamis, or a
terrorist attack on the facility. A QRA would examine these scenarios in terms of their expected
frequencies of occurrence and in terms of their expected consequence severities given that they
had occurred. For purposes of a QRA, expected risk would be defined as the sum of the products
of the expected frequencies and severities. In practice, some potentially severe risks would be
ignored because the likelihood of these risks occurring is sufficiently small (e.g., less than 10
-6

annually) or beyond the means of the facility to mitigate (e.g., the risk of a damaging meteorite
strike or of a war).

80
Since “large decisions are not made by a group of like-minded people …, rather, [they
are] the result of extended negotiations, either implicit or explicit, between
representatives of different points of view,” (de Neufville and Keeney, 1972) this report
seeks supplement what is known about Rancho LPG and its potential negative impact on
its adjacent neighborhood. Objective reflection about the facility’s site and hazards
might lead to improved protection of the local community by continued dialogue and
implementation of additional mitigation measures.
20

Defining a Potential Catastrophic Breach
LPG was first produced in 1910. LPG consists of light hydrocarbons (typically, propane,
butane, propylene, or a mixture thereof) with a vapor pressure of more than 40 psi at 100
ºF. At standard temperature and pressure (STP or 32 ºF and 14.7 psi), LPG is a gas. It
can be liquefied by increasing its pressure or by dropping its temperature. The properties
of LPG allow it to be stored or transported as a liquid and used as a gas. LPG vapors are
heavier than air and tend to collect on the ground and other low spots. If LPG is released
into the atmosphere, it readily mixes with air and, in the correct proportions, can ignite

20
With the possible exception of the behavior of heavy gas vapor clouds, the consequences of
LPG spills contain substantial uncertainties. While catastrophic spills of LPG at facilities such as
Rancho LPG are certainly possible, almost all experimental spills have used fewer than 10 cubic
meters (fewer than 3000 gallons) of liquid. Thus, there is uncertainty regarding the accuracy and
validity of extrapolation of current empirical information and physical models to spills of
catastrophic size. Less likely but probable explosive fireball and BLEVE models are also
uncertain. Existing LPG risk analyses illustrate the difficulties of enumerating and quantifying
the event tree probabilities of sequences of probable actions and their associated consequences.

81
and cause an unconfined vapor cloud explosion
21
(VCE). A VCE occurs when a large
amount of a flammable vaporizing liquid or gas is rapidly released into surrounding air,
usually under confinement, and is ignited before being diluted below its lower flammable
limit (LFL). When such a release occurs, mixing via air currents, convection, and
diffusion of LPG vapors affects the size and extent of the vapor cloud.
A major concern for LPG bulk storage is a massive failure of a vessel containing a large
quantity of LPG. The probability of this type of failure occurring can usually be
mitigated or at least controlled to a reasonable and tolerable degree with appropriately
designed and operated facilities coupled with local fire department preparedness and
response. Most LPG fires originate as smaller fires. These can escalate into flash fires,
pool fires, and pressure fires. Or, in the event the spilled butane does not immediately
ignite due to variations in ignition sparking characteristics, atmospheric turbulence,
atmospheric pressure, or in the methane-air mixture, it will vaporize into a dense cloud
that hugs the ground. In this case, the vapor cloud can drift downwind until it warmed up
and mixes sufficiently with air until it is either ignites once it is presented with an ignition
source or harmlessly disperses. An LPG vapor cloud burns when it constitutes between
1.9 and 8.5 percent of air by volume in the presence of a spark or fire. Given the proper
local conditions and volume of spilled LPG, the vapor fire can cover an area of several

21
Although a flash fire is a more likely result, there have been instances of confirmed,
unconfined VCEs. See, for example, Burgess, DS and Zabetakis, MG), “Detonation of a
flammable cloud following a propane pipeline break: the December 9, 1970, explosion in Port
Hudson, Mo.” (Technical Report BM-RI-7752), Bureau of Mines, Pittsburgh, Pa. (USA).
Pittsburgh Mining and Safety Research Center.

82
square miles. If a vapor cloud is ignited,
22
its VCE will burn back toward the evaporating
pool of liquid, form a pool fire. In conditions wherein an unbreached LPG tank still
stands and is rapidly heated by a pool fire that engulfs it, the unbreached tank containing
LPG may fail suddenly producing an explosive fire. As the tank heats up, the liquid in
the tank rapidly absorbs energy from the surrounding fire. The resulting increased rate of
vaporization produced increases the ullage pressure. When this pressure exceeds the
limit characteristic of the material properties of the tank wall, its thickness, the internal
temperature, and the capacity of all operating pressure relief valves, the tank may fail
catastrophically. If this occurs, the liquid released from the tank boils rapidly and
expands. The resulting fireball and explosion, the BLEVE, can fragment the tank and
propel sections over large distances.
23

For purposes of a potential terrorist attack on Rancho LPG, an analysis of the risk focuses
on three significant, potential hazards associated with a spill: a VCE, a VCE followed by
a pool fire, and a pool fire followed by a BLEVE. This LPG expected loss assessment
evaluates selected scenarios. Rancho LPG is assessed to determine its potential for
severe consequences such as loss of life, injury, and property based on site- and situation-
specific decisions. First, preliminary consequence modeling was performed for the
purpose of estimating the range of offsite consequences of releases of flammable
material. This modeling was based on equations from the EPA’s RMP Off-Site

22
This would be extremely likely due to myriad ignition source (e.g., idling engines, traffic
signals).

23
In 2004, Planas-Cuchi documented an example of a non-pressurized vessel BLEVE. See
http://ecosakh.ru/data/im_docs_62_vzryv_avtocisterny_original_na_angl.pdf .

83
Consequence Analysis Guidance document for estimating impact distances for
explosions and BLEVEs and Fay’s Algorithm (Fay, 2003(1)) for estimating pool fire
sizes. The equations for these events were programmed into an EXCEL spreadsheet
and used to determine the size of the impact zone given different spill size. These
equations which summarize the theoretical extent of different scenarios, regardless of
their respective plausibilities, are summarized below.

Vapor Cloud Explosion (VCE) Model
24

For VCEs, the total quantity of flammable substance is assumed to form a vapor cloud.
The entire cloud is assumed to be within the flammability limits, and the cloud is
assumed to explode. Two percent of the flammable vapor in a saturated hydrocarbon
cloud is assumed to participate in the explosion. The distance to the one pound per
square inch overpressure level is determined using equation 1 as follows:
(1)

Where:
X = distance to overpressure of 1 psi (meters)
W
f
= weight of flammable substance (kg)
H
Cf
= heat of combustion of butane (49,100,000 joules/kg)

24
Based on empirical TNT equivalency with yield modified via the Automated Resource for
Chemical Hazard Incident Evaluation (ARCHIE) Method.

84

H
CTNT
= heat of combustion of trinitrotoluene (4,680,000 joules/kg)

The distance to an overpressure of 1 psi translates into the maximum distance at which
the force of the exploding gas will break windows and potentially hurt some of the local
residents. This is generally recognized as the minimum safe distance from the source.
Figure 19 displays a curve defining the radius from the source to the boundary wherein
the overpressure is 1 psi given different volumes of liquid involved in the VCE.

85
Figure 19: Vapor Cloud Explosion(VCE) Size by Volume of Butane

Vapor Cloud Explosion (VCE) Size by Volume of Butane
0.00
0.50
1.00
1.50
2.00
2.50
0 5,000,000 10,000,000 15,000,000 20,000,000 25,000,000 30,000,000
Gallons of Butane
Radius in Miles
Distance to overpressure of 1 psi

Pool Fires

Using Fay’s algorithm
25
, the maximum radius, r, of a pool fire, is defined using equation
2 as follows:
(2)

Where:
R = maximum radius of a pool fire (meters)

25
Fay's algorithm was part of his paper describing pool fires on marine waters. However, his
original title for his article was "Model of Large Pool Spills," and its assumptions present a
simplifying, but reasonable, maximum radius for pool fires.

86
= fraction of the pool fire’s heat release that is emitted as thermal radiation
(selected to equal 0.150)
Q = heat release rate of butane (selected as 1.325 TW
26
)
q = thermal radiative flux (selected as 5 and 25 kW/m
2
)

Based on this algorithm, the maximum pool fire diameter is about 0.5 miles at 25 kW/m
2

or 1.1 miles at 5 kW/m
2
radiative flux.

Boiling Liquid Expanding Vapor Explosion
The equations used by the EPA to estimate impact distances for BLEVEs are summarized
in equations 3 and 4 below:
(3)
75 . 0
6
67 . 0
c a
t
10 x 3.42
4
H R t 2.2
X








Π
=
f
W

Where:
X = distance to the 5 kilowatts per square meter endpoint (meters)
R = radiative fraction of the heat of combustion (assumed to be 0.4)
t
A
= atmospheric transmissivity (assumed to be 1)
H
C
= heat of combustion of butane (49,100,000 joules/kg)
W
f
= weight of flammable substance in the fireball (kg)
t = duration of the fireball in seconds (estimated from the following equations)

For W
f
> 30,000 kg, duration is described by equation 4 as follows:
(4)

26
See http://www.nrc.gov/reading-rm/doc-collections/nuregs/staff/sr1805/ch3-6.pdf .

87
The distance to a 5 kW/m
2
radiative flux translates into the maximum distance at which
the flux will hurt residents. This is generally recognized as the minimum safe distance
from the source. Figure 20 displays a curve defining the radius from the source to the
boundary wherein the 5 kW/m
2
radiative flux is given different volumes of liquid
involved in the BLEVE as well as the estimated burn times at each of these volumes.

The tolerable levels for incident radiant thermal flux on off-site targets from LPG
facilities is set to a minimum of 5 kW/m
2
(or 1600 BTU/hr ft
2
) when the “outdoor areas
are occupied by 20 or more persons during normal use, such as beaches, playgrounds,
outdoor theaters, other recreational areas, and areas of public assembly” (Raj, 1993). A
25,000 kW/m
2
flux refers to the border wherein ignition of wood occurs without direct
flame exposure (Clarke, 2005).

Figure 20: Bleve

BLEVE
-
5
10
15
20
25
30
35
40
45
50
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
Gallons of Butane
See Legend Below
Duration of fireball (in sec.) Distance to 5 kilowatt per square meter impact force (in miles)

88
In order to give more precision to plausible consequences given selected scenarios, more
detail is needed. For example, the Rancho LPG facility is located on a twenty-acre site
that has two major storage tanks in addition to an array of smaller storage tanks,
pipelines, and receiving/loading facilities. The large receiving tanks are comprised of
multiple protective layers. The first layers or inner tanks are the primary containment
barriers that consist of welded and reinforced one-fourth and three-eighths inch steel
alloy plates. The second important layers are the outer tanks that are made of welded
one-fourth inch steel plates. A layer of insulating materials separates the two layers, and
containment basins surround each butane tank. The diameters of the inner tanks are
approximately 171 feet, and their working heights are approximately 80 feet. Our
catastrophic event assumption includes an attack by small domestic- or foreign-based
terrorist group using a land-based rocket propelled thermobaric grenade or equivalent (for
example, by using a Ruchnoy Protivotankoviy Granatomet (RPG)-TBG-7V or a SMAW
man-pad with a shaped charge (Clarke, 2005)) (alternatively, an attack by intentionally
crashing a fully-fueled private jet as a missile projected into one of the large tanks could
produce equivalent consequences).
27
An attack of this type is plausible given the “Twin

27
Jets of comparable size and speed could likely be acquired out of Los Angeles International
Airport (LAX) or Long Beach Airport (LGB), both fewer than twenty miles from the site.
Research that simulates the crash of an aircraft into a concrete structure such as full containment
storage tanks has been conducted (see, for example, Sugano, 1993). Partially as a result of such
work, provisions for aircraft impact on reinforced concrete structures are incorporated into the
Civil Engineering codes used for the design of nuclear containment structures. Indeed, Sugano’s
study provides evidence that an airframe and the skin of an aircraft alone are not likely to cause
major structural damage on impacted reinforced concrete targets. The most massive impacting
element would be the aircraft’s fuel, as evidenced by the Pentagon impact of September 11, 2001
(see Popescu, 2003). Simulations show that the structural damage occurs only when the fuel
mass hits. A fully fueled aircraft strike alone or coordinated with a thermobaric RPG or man-pad
attack would have a high probability of catastrophically breaching the LPG containment.

89
Sisters” plot of 1999.
28
In Twin Sisters, the FBI foiled a domestic terrorist attack on a
facility in Elk Grove, California that is very similar in size and layout to the Rancho LPG
facility.
Figure 21: “Twin Sisters” Propane Facility, Elk Grove, CA and Rancho LPG Facility,
San Pedro, CA

“Twin Sisters” Propane Facility, Elk Grove, CA Rancho LPG Facility, San Pedro, CA
Both images courtesy of Google Maps

The specific attack scenario is based on a simple cause-and-effect model. Essentially, the
analysis is broken down into the following steps: First, the attack mode includes a land-
based rocket-propelled thermobaric grenade (RPG) or similar weapon (i.e., a jet aircraft

28
See http://www2.fbi.gov/hq/nsb/wmd/wmd_cases.htm#Twin . In 1999, three San Joaquin
County Militia members, part of a right-wing militant group, plotted to destroy the “Twin
Sisters,” two propane tanks in Elk Grove, California. The liquid propane tanks were
approximately 12 million gallons apiece and were viewed as a “target of opportunity” by Donald
K. Rudolph, Kevin Ray Patterson, and Charles Dennis Kiles. Their plot was hatched in 1996 in
hopes of causing a disruption that would lead to a declaration of martial law. This was part of a
larger conspiracy by militia groups to destabilize the federal government. Rudolph, Patterson,
and Kiles intended to execute their plan in 1999. An expert estimated that their planned
explosion would have caused 12,000 deaths, widespread fire, and third-degree burns among
individuals within five miles or more of the explosion. During the FBI investigation, Rudolph
agreed to assist in exchange for not being charged in this matter and instead pleaded guilty to a
previous charge of plotting to assassinate a U.S. District Court judge in 1998. In 2002, Patterson
and Kiles were convicted.

90
collision with a tank) striking one of the large butane storage tanks, the liquid butane
releasing, and a pool fire igniting. Second, the pool fire surrounds and heats the other
large liquid butane storage tank’s contents, eventually the tank’s ability to vent off the
excess pressure is overwhelmed, and a BLEVE occurs. Finally, consequences of the
attack—deaths, injuries, destroyed property, and business interruption economic
damage—are calculated. The procedures for estimating the worst-case consequences of
this disaster at the butane storage facility are relatively simple—(a) assume the event is
intentional (i.e., a terrorist event) and (b) assume it is catastrophic (i.e., a near-
instantaneous complete release of the butane in one of the large tanks due to trauma {e.g.,
using a jet or thermobaric RPGs as missiles to breach one of the tanks}).
Using the capacity of a single butane tank as the spill size and assuming a catastrophic
breach, we employed the software package known as CAMEO
29
, a suite of software
programs for planning and response to chemical emergencies that was developed by the
U.S. Environmental Protection Agency's Office of Emergency Prevention, Preparedness
and Response (see www.epa.gov/swercepp) and the National Oceanic and Atmospheric
Administration's Office of Response and Restoration (see response.restoration.noaa.gov),

29
CAMEO includes a set of databases, a gas dispersion model, ALOHA, and an electronic
mapping program called MARPLOT. The ALOHA dispersion modeling includes the advection
(moving) and diffusion (spreading) of gases. A dispersing vapor cloud will generally advect in a
downwind direction and diffuse in a crosswind and vertical direction (crosswind is the direction
perpendicular to the wind). A gas cloud that is denser than air (known as a heavy gas) can
slightly spread upwind (Spicer, 1989). ALOHA models the dispersion of a gas in the atmosphere
and displays an overhead view of the area (footprint) in which it predicts gas concentrations
typically representative of hazardous levels called the Levels of Concern (LOC). The footprint
represents the area within which the concentration of a gas is predicted to exceed a LOC at some
time during the release. For example, the selected LOC for butane in this report is 19,000 ppm
(or 1.9% of the atmosphere) or its Lower Explosive Limit (LEL) (Spicer, 1989).

91
to model a pool fire and a subsequent BLEVE of the other large tank and assuming
typical, local (summertime) atmospheric and landscape conditions.

The analysis utilizes this standard model
30
in a novel manner. In the case of this facility,
an expected, worst case economic value consequence of $23 billion was selected by
valuing each expected fatality at $5 million, each injury at $100,000, the LPG facility at
$200 million, local property damage at $3 billion, and ongoing economic impacts at $9
billion (primarily in terms of short- and medium-term reduced refinery and port trade).
Of course, the estimation of such a severe event is alarming, but it must be viewed in
terms of its expected frequency of occurrence. Selecting the frequency of a terrorist
event at a specific locale in a given period of time is difficult, if not impossible, because
of the paucity of data. Estimates to model the expected annual frequency of terrorist
events at the Rancho LPG facility were derived by using data from the MIPT Terrorism
Knowledge Base (MIPT, 2006). From this data, a range from around three to as many as
26 terror incidents that involve property damage or bodily injury might be expected to
occur annually in the United States with a mean of about three or four (and of less than
one of significant severity) at locations of potential similar impact as the twin ports of
Long Beach and Los Angeles. Assuming that these would be among the 100 most

30
ALOHA uses simplified heavy gas dispersion calculations based on the DEGADIS model.
Thus, ALOHA’s results are unreliable under very low wind speeds, very stable atmospheric
conditions, wind shifts and terrain steering effects, and concentration patchiness, particularly near
the spill source, and ALOHA does not account for the effects of fires or chemical reactions,
particulates, chemical mixtures, and terrain. Nevertheless, the analysis incorporated ALOHA
because it uses DEGADIS “which was originally developed for the USCG and the Gas Research
Institute primarily for simulation of the dispersion of cryogenic flammable gases” (Spicer, 1989).
A rough estimate of the potential damage of a modeled vapor fire can be calculated as a result of
estimating property and population demographics under the impact zones and imputing damage
factor.

92
probable targets nationwide (with a one percent conditional probability of an incident)
and that the LPG storage facility would be among the five most probable local targets
within the port area (with a twenty percent conditional probability), we further estimated
that a binomial distribution could serve as a reasonable base frequency. The selected,
estimated rates of terrorist attacks (or Bernoulli trial “successes” given a random draw
once a year) on the Port of Los Angeles and Long Beach was 1 in 230 annually, and, for
the LPG facility, 1 in 1150 per annum (with sensitivity analysis conducted on a range of
between 1 in 100,000 and 1 in 101). Given this expected Fermi event frequency and
range, then, the equivalent, annualized pure premium
31
related to Rancho LPG is $20
million (with a range from $230,000 to $227.7 million).

Observations and Implications
In contrast to the first chapter, which looks at a planned facility and examined alternative
options for siting and protecting the facility, this chapter examines an old, existing
facility with limited options for risk reduction.

This examination highlights several insights as follows:
1. There is a significant difference in the risk management of planned versus
existing hazardous facilities. In Unbuilding Cities: Obduracy in Urban
Sociotechnical Change, Hommels comments extensively about the persistence
of ill-conceived and unwanted infrastructure in urban environments. Rancho
LPG is an example of obduracy due to embeddedness. The persistency of

31
The “equivalent annualized pure premium” may be thought of as the estimated fair price that
an insurance company would charge yearly to assume the entire risk presented by the facility.

93
Rancho LPG is due, in part, to its relative rigidity and irreversibility with
respect to zoning, legal developments, and the “’deep-rooted [American]
ideological antipathy to government intervention in urban and regional
developments’” (Hommels, p. 13).
2. From a policy perspective, it is a bad idea to exempt facilities from regulations
without a sunset clause. It is improbable that, if built from scratch, Rancho
LPG would be licensed today at its current location given current regulations
and other requirements such as an EIS and a QRA. From a macro-level
perspective, the example of Rancho LPG highlights the absurdity of unlimited
grandfathering in zoning. The original rationale for grandfathering was that
significant, sudden regulatory change hurts existing facilities and discourages
future investment. Arguments centering on “fairness” and “economic
feasibility” (e.g., it is less expensive to implement pollution controls at the
time of new construction rather than as a retrofit) were developed to favor the
owners of infrastructure. In retrospect, however, the obvious problems of
grandfathering emerged. By creating a permanent, regulatory environment
favoring existing facilities, grandfathering established a perverse incentive to
keep aging facilities open. The grandfathered status of Rancho LPG may
have become its most valuable asset. Protecting that asset has meant
defending the facility, even at the potential expense of downplaying public
safety and, in the case of its expired marine shipping permit, operating less
efficiently. When an area’s zoning changes, whether it be economically

94
positive zoning with respect to the property owner (i.e., for “war emergency”
purposes) or economically negative zoning with respect to the property owner
(i.e., new pollution regulations or the requirement of an EIR), the
implementation of grandfathering, at least with respect to public safety and
environmental compliance issues, should either be eliminated or strictly
delimited in scope and time.
3. Siting of aboveground LPG storage facilities near other critical infrastructure or
near population centers should always be avoided and, when present, rectified.
Similar to the similar arguments made in Chapter 2, alternative siting,
hardening, and additional security can be established as very effective tools in
reducing both reducing conditional and unconditional expected losses. Siting
of aboveground LPG storage facilities in areas with low densities at least four
miles from population centers and at least two miles from other significant
commercial enterprises to minimize the exposure of people and property to
potential harm should be strongly preferred in the regulatory approval process,
and full containment of facility tanks should be mandatory.

4. The persistency of Rancho LPG may also be due to its significant sunk costs.
Political decisions often consider sunk costs, and avoidance of this
consideration can only be accomplished absolutely by use of prospective
analysis of proposed sites rather than retroactive analysis of existing sites.

95
The possibility of a terrorist event at critical infrastructure in densely populated areas
must always be considered. Since terrorist attack frequency cannot be determined for an
attractive target, control of potential severity is critical. Siting, containment, and defense
strategies should incorporate multidisciplinary design optimization techniques. All
aboveground LPG storage facilities should have real time leak detection systems (LDS)
with supervisory control and data acquisition (SCADA) industrial control systems that
monitor and control the facility-based processes. In addition, aqueous film-forming foam
concentrate (AFFF) systems
32
can assist in protecting each tank and diked area. Constant
compliance with the various sources of standards and codes exist for dealing with LPG
facilities and proper fire protection should be maintained, including, but not limited to the
National Fire Protection Association (NFPA) 54, National Fuel Gas Code, NFPA 58,
Liquefied Petroleum Gas Code, NFPA 59, Utility LP-Gas Plant Code, American
Petroleum Institute (API) 2510, Design and Construction of LPG Installations, API
2510A, Fire-Protection Considerations for the Design and Operation of Liquefied
Petroleum Gas (LPG) Storage Facilities, and the IP Code of Practice for LPG.

Creative siting decisions might seek to be more regional rather than local in their
recommendations toward resolving the messy problem of LPG storage facility siting.
The facility’s location in San Pedro is partly a consequence of the prior siting of
refineries in Wilmington, Carson, and Torrance. However, the need for storage of this

32
AFFF is an aggregate of air filled bubbles formed from aqueous solutions and is lower in
density than flammable liquids. In firefighting, it can be used to form a cohesive floating blanket
and prevent or extinguish fire by excluding air and cooling the fuel. It also prevents re-ignition
by suppressing the formation of flammable vapors.

96
refinery byproduct (since alternative uses such as fueling butane electric generators exist)
and the need for its proximity to marine shipping (since the marine permit has expired
33
)
no longer apply. Consequently, condemnation of the facility should be considered, and,
if there is a continued need for such a facility, planning for a new facility that is sited in a
more appropriate location, after a thorough and complete environmental review process is
conducted, should be considered.

33
This ironically makes the facility even more hazardous since shipping by rail and truck is more
hazardous than transport by ship—due to more connections and more hands needed per volume
handled.

97
Figure 22: Simulated Terrorist Attack on LPG Storage Facility (first tank breached)

And Maximum Extent Pool Fire-related Boundary at Thermal Radioactive Flux = 10kW/m
2

Full breach of one of the large facility tanks

SITE DATA: Location: RANCHO LPG, CALIFORNIA CHEMICAL DATA: Chemical Name: BUTANE Molecular Weight: 58.12 g/mol, LEL: 15000
ppm UEL: 90000 ppm
Ambient Boiling Point: 31.1° F, Vapor Pressure at Ambient Temperature: greater than 1 atm, Ambient Saturation Concentration: 1,000,000 ppm or 100.0%
ATMOSPHERIC DATA: : Wind: 7 miles/hour from s at 3 meters, Ground Roughness: urban Cloud Cover: 3 tenths, Air Temperature: 80° F, Stability Class:
D
Relative Humidity: 25%
SOURCE STRENGTH: Leak from hole in horizontal cylindrical tank Flammable chemical is burning as it escapes from tank Tank Diameter: 171 feet, Tank
Length: 73.75 feet
Tank Volume: 12,669,979 gallons Tank contains liquid Internal Temperature: -45° F Chemical Mass in Tank: 34,177 tons Tank is 100% full Circular
Opening Diameter: 10 feet

98
Figure 23: Simulated Terrorist Attack on LPG Storage Facility (other tank breached)

And Maximum Extent BLEVE Boundary at Thermal Radioactive Flux = 10kW/m
2

Full breach of the other large butane storage tank

SITE DATA: Location: RANCHO LPG, CALIFORNIA CHEMICAL DATA: Chemical Name: BUTANE Molecular Weight: 58.12 g/mol LEL: 15000 ppm
UEL: 90000 ppm
Ambient Boiling Point: 31.1° F Vapor Pressure at Ambient Temperature: greater than 1 atm Ambient Saturation Concentration: 1,000,000 ppm or 100.0%
TMOSPHERIC DATA: Wind: 7 miles/hour from s at 3 meters Ground Roughness: urban or forest Cloud Cover: 3 tenths Air Temperature: 80° F Stability
Class: D Relative Humidity: 25%
SOURCE STRENGTH: BLEVE of flammable liquid in horizontal cylindrical tank Tank Diameter: 171 feet , Tank Length: 73.75 feet Tank Volume:
12,669,979 gallons
Tank contains liquid Internal Storage Temperature: -45° F Chemical Mass in Tank: 5,000,000 kilograms Tank is 16% full Percentage of Tank Mass in
Fireball: 100% Fireball Diameter: 1085 yards Burn Duration: 43 seconds

99

Table 9: Demographic Profiles of Areas within Various Distances of Rancho LPG
Facility based on 2000 Census and Corresponding Expected
Consequences|Successful Attack (Conditional Expected Losses) and
Unconditional Expected Losses

Phase I – Pool Fire of Tank A Phase II – BLEVE of Tank B caused
by Pool Fire
Mean Annual
Unconditional
Expected
Loss

0.3-mile
Radius
0.41-mile
Radius
0.63-
mile
Radius
1.33-mile
Radius 1.86-mile
Radius
2.95-mile
Radius
2.95-mile
Radius

Pool fire - 524
yards - (10.0
kW/(sq m) =
potentially
lethal within
60 sec)
Pool fire -
722 yards -
(5.0 kW/(sq
m) = 2nd
degree burns
within 60
sec)
Pool fire -
1103 yards
- (2.0
kW/(sq
m) = pain
within 60
sec)
BLEVE -
(10.0
kW/(sq m)
=
potentially
lethal
within 60
sec)
BLEVE -
(5.0 kW/(sq
m) = 2nd
degree burns
within 60
sec)
BLEVE - (2.0
kW/(sq m) =
pain within 60
sec)
BLEVE - (2.0
kW/(sq m) =
pain within 60
sec)
Total
Population
0 451 4808 26,105 65,430 162,606 162,606
Total
Households
n/a 1634 9934 22,751 56,900 56,900
Total
Estimated
Fatalities
from Attack
(%
Population)
n/a
2,500
(1.5%)
2
(0.001%)
Total
Estimated
Injuries from
Attack (%
Population)
n/a
12,500
(7.7%)
11
(0.007%)
Total
Estimated
Economic
Losses from
Attack
(excluding
Fatalities and
Injuries)
n/a $12.0 billion $10.4 million

100

Chapter 4: A Natural Disaster Involving a Pre-existing Infrastructure
System: A Case Study of the New Orleans’ Flood Control System

Introduction
This example involves a case of a preexisting infrastructure system impacting a large
local area. Its disaster vulnerability may be most evident when considering the
consequences of unintentional (and, possibly, intentional) disasters on this infrastructure.
Specifically, this chapter outlines a decision analysis of options for the levee and
floodwall system in and around New Orleans. The impact of hurricanes, floods, and
sabotage on the New Orleans’ flood control system presents an example more complex
than those examples presented in the first two chapters. Here, multiple levers can be
considered, and the interactions of the levers with the multiple possible objectives present
policymakers with a variety of choices.

Background
The levees and floodwalls protecting New Orleans from hurricanes and floods were
designed to withstand a Saffir-Simpson category 3 hurricane (see U.S. Army Corps of
Engineers (USACE), 1984). When making landfall on August 29, 2005, Hurricane
Katrina was estimated to be a category 4 hurricane; later, it was downgraded to a high
category 3. The devastation that followed was much more extensive than predicted by
USACE in 1984, but it was close to predictions made by scientists and emergency
managers in more recent years (see Maestri, 2002 and Laska, 2004). When examining
the analyses conducted to support the 1984 decisions to fortify the levees and floodwalls,
von Winterfeldt (2006, page 31) concluded:

101

“In summary, there were several problems with the analyses and decisions
regarding the development of levees and floodwalls in the New Orleans area: 1)
probabilities and consequences of extreme hurricane events were underestimated;
2) alternatives that provided a higher level of protection were not explored; 3) the
preferred alternative was implemented slowly and with many funding delays.”

Subsequent reports (for examples, Interagency Performance Evaluation Task Force
(IPET), 2006; Seed et al., 2006) came to similar conclusions.

Years later, the United States continues to face decisions about how to fortify and
upgrade the flood protection system of New Orleans. This chapter outlines a decision
analysis of options for the levee and floodwall system in and around New Orleans.
Substantial portions of New Orleans need to be rebuilt and require protection. As such,
we should consider a comprehensive list of options for flood mitigation, of the possible
types of events in terms of precipitation-, overtopping-, and breach-induced floods, and
of the consequences of these types of events. Historical data was developed to obtain
realistic estimates of flood frequencies and consequences and to combine these estimates
with a parametric analysis of events for which little historic data is available (e.g.,
breaches, sabotage) or for which consequences are uncertain (e.g., fatalities as a function
of evacuation speed). This analysis framework was developed in the form of an
influence diagram, a well-established modeling tool in risk and decision analysis
(Clemen, 1997).

102

New Orleans’ System of Levees and Floodwalls
The levees and floodwalls developed by the USACE in the 1970s and 1980s reduced the
risks of flood damage and provided economic development opportunities. At the time the
USACE designed the system, its analysts believed that it protected New Orleans against a
100-year flood (i.e., a flood of such magnitude that it would occur, on average, only once
in 100 years). However, due to many optimistic assumptions (e.g., no levee breaches,
rapid evacuation and resettlement, no consideration of fatalities), the analysts
overestimated the level of protection and underestimated the consequences of such a
major flood. In fact, the New Orleans area had experienced two near misses of category
3 hurricanes (Betsy in 1965 and Camille in 1969) which suggested that the probability of
a category 3 hurricane or more severe hurricane (which would induce a 100-year flood
event) was much higher than one in a hundred years.

Furthermore, the levels of protection decreased over time due to natural and man-made
changes. Natural changes included continued subsidence, lack of sedimentation, and
declining vegetative growth. Land use changes such as road building and increased
residential densities induced hydrologic changes (including faster runoff) that reduced the
level of protection provided by levees and floodwalls. And, while these levees and
floodwalls required regular and extensive maintenance, their record of maintenance
quality was spotty.

Over time, New Orleans’ levees and floodwalls became structurally deficient and
presented an increased risk to public safety and to the region’s economic infrastructure.

103

Minimum standards to regulate and to enforce the design, placement, construction, and
maintenance of levees and floodwalls had been and are critical to the built environment
of New Orleans and its reconstruction. Indeed, the structural integrity and protection
level of southeastern Louisiana’s floodwall and levee system will strongly influence the
extent of resettlement in New Orleans and influence the probability and consequences of
future catastrophic hurricanes and floods.

In urban areas, the federal government has typically designed levees and other flood
damage reduction projects with a 100-year flood threshold as the minimum standard for
identifying, mapping, and managing flood hazards. Participating National Flood
Insurance Program (NFIP) communities are required to adopt building codes and other
types of activities that will reduce losses posed by a 100-year flood as a result of
mandates by the Federal Emergency Management Agency (FEMA) and in order to
maintain eligibility in this program. FEMA also requires levees and floodwalls
protecting flood-prone areas to be certified for structural soundness and proper
maintenance to a 100-year flood level. The USACE performs most of these
certifications. However, its current process does not assess the geotechnical conditions or
the hydrological conditions of the levees, and neither the areas to be protected nor the
structures built behind the protection of 100-year levees are classified as within
“designated floodplains.”

104

The accuracy of maps used by FEMA to define flood hazard areas is also problematic, as
more than three-quarters of these maps are more than a decade old, raising concerns that
hydrologic data has changed since the maps were last reviewed and updated.
Model Overview
In modeling future floods and their expected consequences in New Orleans, many input
quantities can only be estimated, and thus they have an inherent degree of uncertainty. A
model that explicitly specifies the range of uncertainty in its inputs can provide more
realistic and informative estimates than deterministic assessments. Influence diagrams
are a useful tool in mapping out the decisions, events, and variables that influence the
potential consequences of decisions and events (see, for example, Clemen, 1997). In this
analysis, the software tool, Analytica (see www.lumina.com), assisted in modeling an
influence diagram that represents the interrelationships among 58 variables that include
data for wind, rain, wave action, geology, engineering, demographics, and the potential
for negative consequences of hurricanes and floods in the New Orleans’ area.

At the highest level, the risk and decision analysis model shows a “NOLA Flood Control
Risk Analysis System” entry screen (Figure 24) that defines its submodels: Mississippi
River flood frequency modeling, Lake Pontchartrain flood frequency modeling, land use
and mitigation options for the New Orleans’ area, and demographic and consequence
valuations for the New Orleans’ area. The submodel, “Analysis,” aggregates the
expected frequencies of floods with their expected severities and presents their expected
costs as a function of their mitigation options. This model-submodel hierarchy of

105

influence diagrams within Analytica serves as its key organizational tool. Because the
visual layout of an influence diagram is intuitive, readers are able to learn about the
model’s structure and organization more quickly than is possible with less visual
paradigms.
Figure 24: High Level View of the Decision Analysis Model

The influence diagram also serves as a tool for communication. An understanding of
how the results are obtained and how the various assumptions impact the results is often
more important than the specific input and output numbers. In addition to
communicating high level findings, stakeholders can examine lower levels of modeling
when more detail is desired, aided by the visual aspects of the model’s structure. When
stakeholders are able to understand such a model, debate and discussion will often focus
more directly upon specific assumptions and lead to more productive results. Thus, the
influence diagram serves as a tool to help to make the model accessible.

106

Following is a brief description of the influence diagram structure, followed by a
description of the model inputs and calculations:

The Mississippi Flood submodel is shown in Figure 25. Floods are divided into two
classes of chance nodes based on cause: overtopping and breaches caused by overtopping
floods (which include upstream Mississippi River floodwaters compounded by sinking
floodwalls and design errors as well as downstream Mississippi River surges
compounded by sinking floodwalls and design errors) and breaches caused by anything
other than overtopping (this includes terrorist acts, poor workmanship or materials, and
design errors).
Figure 25: Mississippi Flood Submodel

Figure 26 shows the Lake Pontchartrain submodel. Once again, floods are divided into
two classes based on cause: overtopping and breaches caused by overtopping floods

107

(which include Lake Pontchartrain surges, seiches, and waves compounded by sinking
floodwalls and design errors) and breaches caused by anything other than overtopping
(this includes terrorist acts, poor workmanship or materials, and design errors).
Figure 26: Lake Pontchartrain Flood Submodel

The “land use” submodel includes the options considered in this analysis for
improvements of the levee and floodwall system:
Restoring levees and floodwall to the current (base) levels
Increasing the levees and floodwalls by 5 feet
Increasing the levees and floodwalls by 10 feet
Other options that can be explored with this model are improved levee maintenance and
improved pumping systems and channels.

108

The demographics submodel contains the information representative of the housing stock
and population in the New Orleans areas subject to possible flooding. This demographic
information is used in the loss calculations, which determine, for each flood level, two
consequences: lives lost and economic impacts. Lives lost are converted to economic
equivalents by using a value of life of either $5 million or $10 million.

The analysis submodel (Figure 27) shows three decision nodes (rectangles). From the
land use submodel, floodwall and levee heights can be selected. In addition, an option to
allow the use of river flow cut-offs, such as the use of partial rechanneling of the
Mississippi River down the Atchafalalya River during severe floods, is introduced.
Attenuated by these choices, the products of flood and hurricane severities and
frequencies return expected annual flood and hurricane losses and costs (net losses plus
mitigation costs) for the New Orleans’ area. The uncertain quantities are specified using
probability distributions. When evaluated, the distributions are sampled using Monte
Carlo sampling, and the samples are propagated through the computations to the expected
annual flood consequences (in terms of lives lost and economic impacts). These
distributions of consequences can then be analyzed in light of various mitigation
strategies to evaluate the expected costs and benefits of these strategies.

109

Figure 27: Analysis Submodel

Flood Frequency Risk Analysis
The frequency distributions of potential floods in the New Orleans’ area were based on
historical data. As a starting point, we assumed that catastrophic floods could inundate
New Orleans through two major pathways—one, primarily from the south and east via
hurricanes as occurred with Hurricane Katrina and, the other, primarily from the north via
Mississippi River basin flood flows as in the extreme example of the Great Mississippi
Flood of 1927 (see Barry, 1997).

In an attempt to capture accurate historic records of floods along the east side of New
Orleans (i.e., the Lake Pontchartrain shoreline and similar areas) and along the banks of
the Mississippi River, we used the United States Geological Survey (USGS) flood gauge
data of peak annual flood discharges from a flood gauge station on Lake Pontchartrain

110

and from two Mississippi River gauges, one upstream in Baton Rouge, Louisiana, and
one downstream in West Pointe a La Hache, Louisiana. For the Mississippi River, two
gauges were selected to represent maximum flood waters along the Mississippi River in
New Orleans because, historically, cut-offs and intentional levee breaches have been used
to temper rising waters along the banks of the Mississippi in New Orleans. The Baton
Rouge station, then, was used as a proxy for maximum flood potential from Mississippi
basin floods that are resultant from upstream run-off, and the West Pointe a La Hache
station was used as a proxy for downstream surge from western approaching hurricanes.

Table 10 shows the relationship between storm categories, wind speed, minimum surface
pressure, and storm surge. Minimum pressures and surge heights are important in
associating floods with the Saffir-Simpson scale of hurricane intensities (Simpson, 1974
and also see http://www.ncdc.noaa.gov/oa/satellite/satelliteseye/educational/saffir.html ).
Table 10: Relationships between Storm Categories, Wind Speed, Minimum Surface
Pressure and Storm Surge

Saffir-
Simpson
Category
Wind speed Minimum
surface pressure
Storm
surge
mph m/s mb ft
1 74-95 33-42 greater than 980 3-5
2 96-110 43-49 979-965 6-8
3 111-130 50-58 964-945 9-12
4 131-155 59-69 944-920 13-18
5 155+ 69+ less than 920 18+

For bodies of water with water-level gauges such as Lake Pontchartrain, a standard flood
frequency analysis procedure is used. These guidelines, known as Bulletin 17
(Interagency Advisory Committee on Water Data, 1982), are the official procedures of

111

federal agencies in the United States. Bulletin 17 characterizes flood frequency at a
given location as a function based on the sequence of annual data points known as the
peak annual flood discharges that are defined as the annual maximum water levels at the
flood gauge location. These magnitudes are assumed to be independent random variables
that are represented by log-Pearson Type III (gamma) probability distributions. These
distributions give the annual exceedence probabilities, the probability that a flood will
exceed a given magnitude in an annual period.

Bulletin 17 defines the annual peak flows for a site and describes the calculations in
detail. Its steps include data collection, outlier detection and adjustment, skew
adjustment, curve computation, plotting, and confidence limits’ calculation. A flood
frequency curve is typically formulated for each type of hazard that is applicable, for
examples, upstream rainstorm or snow melt run-off inundations and hurricanes. As such,
each hazard curve is a conditional probability curve. The unconditional probability
distribution is obtained by weighting the conditional probability curves in proportion to
the chance that a flood will be of each respective type. A means of expressing the
magnitude of an expected flood is through the use of a term known as a return period or
probability of exceedence. The exceedence probability is not a random event, but a
quantile of the flood frequency distribution. Thus, the probability of an exceedence next
year for a 100-year return period is 1%, regardless of this year’s outcome, the probability
of exceedence in the year after next is 0.99 times 1%, and so forth, such that the average
time to the next exceedence is 100 years.

112

The choice of a simple functional form for flood frequency distributions is problematic.
Three of the more common choices for flood frequency are the extreme value
distribution, the logistic distribution, and the lognormal distribution. We chose the
logistic to represent the flood-surge exceedence curves for the Lake Pontchartrain
floods/surges because it represents a reasonable fit to both the hurricane-induced and
non-hurricane induced floods, it is an available and flexible option within the Analytica
modeling software, and its problematic tails are censored and truncated in the analysis.
The logistic distribution’s cumulative distribution function (cdf) is defined as follows:
F(x,µ,s) = 1/(1+e
-(x-µ)/s
)
The probability density function (pdf) of the logistic distribution is given by:
f(x,µ,s) = e
-(x-µ)/s
/[s(1+e
-(x-µ)/s
)
2
]
The µ (mean) for the selected distribution is 10.487, and its s (shape) is 6.988. The fitted
cdf was based on hurricane flood frequency calculations derived from standard project
hurricane (SPH) frequency analyses for Lake Pontchartrain. This distribution represents
the expected range of maxima of lake depths plus surge heights in feet over a return
period equivalent to the number of years of data. An adjustment factor was used to
convert the 32 available, annual data points to a distribution of measurements whose
return period is 100 years (according to the formula, adjustment factor = (1 - [1/selected
interval])/{1 – [1/actual interval]} or 1.021935484).
In addition to expected surge/seiche/wave maxima, the probability of non-overtopping
related breaches due to design errors, poor workmanship, improper materials, and
intentional sabotage as well as the gradual sinking of existing levees and floodwalls due

113

to subsidence were incorporated. Design errors, poor workmanship, and improper
materials were estimated to cause catastrophic structural failure (without floodwater
assistance) an average of once in 10,000 years. A Poisson distribution represents this
failure rate. Intentional sabotage was estimated at a fixed probability of 1 in 10,000 per
year, due to a lack of specific threat information. Average subsidence was estimated at
0.081 ft per year based on the estimates of subsidence as much as 0.162 ft per year (see
Westerrint, 2003). We also assumed that, once cumulative subsidence reaches one foot,
mitigation occurs.

The product of these distributions returned a distribution of peak water levels for Lake
Pontchartain (see Figure 28). The current average height of the levees and floodwalls
above the lake’s water level was estimated at 17.5 feet. From this measure, we
constructed levee heights for different mitigation options at 17.5 feet, 22.5 feet, 27.5 feet,
30 feet, 32.5 feet, and 37.5 feet . Floods are expected when peak water levels exceed the
levee heights (represented as surge=0.00 ft. in Figures 28 and 29).

114

Figure 28: Cumulative Distribution Functions of 100-Year Flood Levels for Lake
Pontchartrain for Different Levels of Protection

100 Year Max Surge Level of Lake Pont. (ft)
Probability
0 0.2 0.4 0.6 0.8 1
-50.00
-40.00
-30.00
-20.00
-10.00
0.00
10.00
20.00
30.00
40.00
Key Levee Mitigation - Height - Pontchartrain (Current floodwall = 17.5 ft at 0.00)
17.5
22.5
27.5
30
32.5

We determined the cumulative distribution function over flood levels in the Mississippi
River in a similar manner. For purposes of this analysis, we selected the logistic function
to represent the flood-surge exceedence curves for the Mississippi River floods. The µ
(mean) for this selected distribution is 23.483, and the σ (shape) is 15.708. The fitted cdf
was based on the distribution using non-hurricane flood frequency calculations starting
with a standard log-Pearson Type III analysis for Mississippi River at New Orleans,
Louisiana representing river depth in feet over return period in years. This distribution
represents the expected range of maxima of river depths over a return period equivalent
to the number of years of data. An adjustment factor of 0.96785 was used to convert

115

these 122 available annual data points to a distribution of measurements whose return
period is 100 years.
In addition to expected floodwater maxima, the probability of non-overtopping related
breaches due to design errors, poor workmanship, improper materials, and intentional
sabotage as well as the gradual sinking of existing levees and floodwalls due to
subsidence were incorporated. Design errors, poor workmanship, and improper materials
were estimated to cause catastrophic structural failure (without floodwater assistance) an
average of once in 10,000 years. A Poisson distribution represents this failure rate.
Intentional sabotage was estimated at a fixed probability of 1 in 10,000 per year.
Average subsidence was estimated at 0.081 ft per year based on the estimates of
subsidence as much as 0.162 ft per year. We also assumed that, once cumulative
subsidence reaches one foot, mitigation occurs.

The product of these distributions returned a distribution of peak water levels for the
Mississippi River. The current average height of the levees and floodwalls above the
river’s bottom near its banks was estimated at 43.8 feet. From this measure, we
constructed theoretical levee/floodwall heights at 43.8 feet, 48.8 feet, and 53.8 feet (see
Figure 29). Floods are expected when peak water levels exceed the floodwall heights
(represented as flood=0.00 ft. in the graph).

116

Figure 29: Cumulative Distribution Function of 100-Year Flood Levels at the
Mississippi River for Different Protection Levels and Assuming No Use of
Cut-offs

Cumulative Probability
100 Year Max Flood Level of MS River (ft)
-100.00 -80.00 -60.00 -40.00 -20.00 0.00 20.00
0
0.2
0.4
0.6
0.8
1
Key Levee Mitigation - Height - MS (Current floodwall = 43.8 ft at 0.00)
43.8
48.8
53.8

Evaluation of the Consequences of Floods and Hurricanes
We estimated both economic consequences of floods and the number of lives lost,
depending on surge and flood levels, breaches, and evacuation times. For the expected
flood level for hurricanes, economic (excluding the value of lives) consequences were
estimated in this analysis by utilizing historic economic consequences’ data collected by
the National Oceanic and Atmospheric Administration (NOAA) (see Blake et al., 2006
and Landsea et al., 2003) adjusted to current levels (see Pielke et al, 2002). Historic
hurricane losses were trended to current loss expectation levels by adjusting past losses
for the cumulative effects of economic inflation, the growth of infrastructure, and
population change. The economic inflation adjustment was accomplished by using the

117

annual Consumer Price Indices (CPI) from the U.S. Bureau of Labor Statistics (see
www.bls.gov ). Infrastructure changes were quantified by using the annual indices
measuring investments in fixed assets available from the U.S Bureau of Economic
Analysis (see www.bea.gov ). Finally, annual population estimates were derived from
the U.S. Bureau of the Census (see www.census.gov ). The adjusted losses (from 1955 to
current) were then fitted to a cumulative size-of-loss distribution as a gamma distribution
(Figure 30) with an α of 0.1305 and a β of 62,500.
Figure 30: Cumulative Severity Distribution of Hurricane Losses (in $ Millions
and excluding value of lives)

Cumulative Probability
Hurricane Loss ($ Millions)
$0 $10,000 $20,000 $30,000 $40,000 $50,000 $60,000 $70,000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1

For non-hurricane floods, we estimated the non-hurricane flood economic (excluding the
value of lives) consequences by utilizing historic economic consequences’ data collected
by the National Weather Service (NWS, a part of NOAA) (see Pielke, Jr., R.A. et al,
2002) adjusted to current levels (Figure 31). Historic flood losses were trended to

118

current loss expectation levels by adjusting past losses for the cumulative effects of
economic inflation, the growth of infrastructure, and population change similar to the
hurricane consequences’ data. The adjusted economic losses (from 1955 to current) were
then fitted to a cumulative loss distribution fitted as a log-logistic distribution (Figure 32)
with a µ (mean) of 3.622 and an σ (shape) of 2.996. In addition, we included an option
for the use of cut-offs, such as the Atchafalaya River, during floods to decrease the peak
flows. When the use of cut-offs is allowed, it was assumed to reduce the floodwater
peaks by fifty percent and vastly reduce the potential for a Mississippi River inundation
in the city of New Orleans.
Figure 31: Cumulative Severity Distribution of Non-hurricane Flood Losses (in $
Millions and excluding value of lives)

Cumulative Probability
Non-hurricane Loss ($ Millions)
$0 $10,000 $20,000 $30,000 $40,000 $50,000 $60,000 $70,000
0
0.2
0.4
0.6
0.8
1

119

The economic value of losses of lives for both hurricane and non-hurricane floods was
estimated in this analysis as a function of the population at risk, the evacuation time, and
the assigned economic value of a lost life, namely, economic value of lives lost = selected
value of life times estimated lives lost where estimated lives lost is assumed to be the
ratio of the population at risk of dying to the product of the number of hours of
evacuation time and 36.236466 (from Brown and Graham, 1988, also see Stedge et. al,
2006). Economic value of lives lost, then, is the product of the selected value of a life
and the number of lives lost (Figure 32 uses $10 million as the value of a life and
displays economic values of lives lost as a function of evacuation time).
Figure 32: Cumulative Severity of Lives (in $ Millions) as a Function of Evacuation Time

Lives (Economic Value) ($ millions)
Evacuation Time (hrs)
20 25 30 35 40 45 50 55 60
0
2000
4000
6000
8000
10K
12K
14K
16K
Key Lives
0
750K
1.106M

120

The aggregate economic value of an event, then, is derived simply as the products of the
frequencies of these events and the sums of their independent economic severity
distributions and the value of life distribution.
Some Preliminary Results
We consider first a base case analysis, comparing the expected costs of several options to
reduce the risk of floods in the New Orleans area, followed by several sensitivity
analyses. Figure 34 shows, in the form of bar charts, how the expected costs stack up
against each other. There are three major messages of this figure: First, the expected
consequences of a flood are dominated by the economic impacts rather than by the
potential fatalities; second, the mitigation costs are commensurate with the economic
costs of floods; third, there appear to be three contenders that minimize the total expected
costs:
1. Status quo with cutoffs;
2. Increased levees at the Mississippi with cut-offs;
3. Increased levees at Lake Pontchartrain with cutoffs.
Note that all three options include cutoffs.

121

Figure 33: Net Annual Flood Costs from all causes assuming a population of
750,000, 48 hour evacuation time, mitigation costs for levee and
floodwall fortification and maintenance, 7% interest, $3.265
million per vertical foot per mile construction costs, and an
imputed value of $10 million per life

Mean Annual Flood Costs
453
1428
453
453
453
1428
1428
99
99
99
99
150
150
150
0
0
1959
1959
2233
0
2233
690
1596
690
690
1596
1596
1596
100
100
100
100
100
100
100
1823
1432
1432
1823
1739
1348
1348
0 1000 2000 3000 4000 5000 6000 7000 8000
Lake Pontchartrain and MS River + 10
ft w/ cut-offs
Mississippi River + 10 ft w/ cut-offs
Mississippi River + 10 ft
Lake Pontchartrain and MS River + 10
ft
Lake Pontchartrain + 10 ft
Status Quo w/ cut-offs
Status Quo
Options
Millions of Dollars
Hurricane Losses
River (non-hurricane) Flooding Losses
Add'l River Flooding Losses w/o Cut-off
Value of Lives
Floodwall/levee maintenance costs
Floodwall/levee building costs

Lake Pontchartrain mitigation options are substantially more expensive than Mississippi
River options, but the savings in terms of economic losses avoided tend to more than
make up for the expense. In this analysis, mitigation of floodwalls and levees by
increasing height is assumed to cost $3,265,000 per vertical foot per mile. The
Mississippi side of the levee and floodwall system is approximately 100 miles long; the
Lake Pontchartrain side (including the interior fortifications), 250 miles.

122

Interestingly, the Atchafalaya River and other potential “relief valves” that serve as flood
flow cut-offs in the event of upstream flooding provide an estimated mitigation value of
$1.9 billion annually. Indeed, New Orleans has depended on such cut-offs historically to
avoid Mississippi River inundations. In the future, given the physical characteristics of
the upstream region that is home to the Old River Control Structure and similar upstream
areas with significant sandy deposits and gradients steeper than the current Mississippi
River bed, it may be prudent to consider carefully this traditional mitigation strategy.
Potentially, such a use could catalyze the necessary initial conditions for the avulsion of
the Mississippi to the Atchafalaya. An avulsion of the Mississippi River has the potential
to irreparably doom the economy and future welfare of New Orleans and Baton Rouge.

This analysis is most sensitive, respectively, to the economic value we impute to a human
life, to mandatory evacuation time, and to the combined levee-floodwall height on the
Lake Pontchartrain side of New Orleans. Simply stated, the most immediate, significant
flood/hurricane mitigations in New Orleans can be accomplished by increasing
minimum, mandatory evacuation times for hurricanes to 48 hours and by giving first
priority to repairs and fortifications of levees and floodwalls on the Lake Pontchartrain
side of New Orleans.

Observations and Implications
Hurricane Katrina was a major natural disaster whose impacts were exacerbated by a
poorly performing flood protection system due to engineering and institutional failures,
questionable judgments, and errors involved in the design, construction, operation, and

123

maintenance of the system. The organizational and institutional problems associated with
the response and recovery efforts for this combined natural and man-made disaster
resulted in one of America’s most severe catastrophes.

The risk analyses of the 1970s and 1980s which led to the design of the levees and
floodwalls were flawed, leading to underestimation of the probabilities of severe events,
an optimistic assessment of evacuation effectiveness, and a very low estimate of the
losses of lives and damages to properties. This risk analysis provides more realistic
estimates, leading to quite different conclusions.

To avoid repeating the mistakes of the past, this analysis can serve as an example of
being more realistic about the assessment of probabilities of these future extreme events
and about their consequences. Some preliminary results include the following:
1. Increasing the floodwall and levee heights by 10 feet can be cost effective,
2. Continuing to provide a Mississippi River cut-off option seems to be very cost
effective, and
3. Increasing mandatory hurricane evacuation periods to at least 48 hours and, on a
prescribed basis, for up to 60 hours can be very cost effective and save many
lives.
Figure 33 (see the bottom three options) details the cost effectiveness of these options.
These options’ impacts are cumulative, and even better solutions for reducing negative
consequences can be obtained by combining these mitigations.

In addition, it may be worthwhile to consider other options, for examples:

124

• Assigning flood plains that are enclosed by levees and floodwalls within
prescribed bounds the status of “designated flood plain,” regardless of the
engineering standards of the levees and floodwalls,
• Reworking flood plain maps on a regular basis to more accurately reflect
elevation changes due to natural (e.g., subsidence, erosion) and man-made
impacts (e.g., global warming, subsidence caused by oil and gas extraction,
hardscaping effects).,
• Addressing floodwall and levee improvement in creative ways (e.g.,
considering use of slurry walls in levees, trading-off additional floodwall
height for marsh restoration [e.g., one foot of floodwalls is approximately
equivalent to 2.7 miles of restored marshlands] or other sustainable
improvements),
• Continuing a state-of-the-art program of continuous levee and floodwall
monitoring and maintenance, and
• Considering the probable avulsion of the Mississippi River in considering the
refortification and rebuilding of New Orleans.
Finally, this analysis can be improved in several important ways:
• Improving the breach and overtopping model,
• Using formal expert elicitation methods to improve the assessment of
probability distributions and their parameters,
• Using uncertainty analysis to account for changes in frequencies and severities
of flooding due to climate change phenomena,

125

• Using uncertainty analysis to assess the impacts of subsidence and the benefits
of subsidence mitigation options,
• Using uncertainty analysis to assess the feasibility and effects of varying
evacuation times,
• Developing and analyzing a more complete set of consequence measures,
including impacts on the ecology, habitat, and environmental justice, and
• Involving stakeholders in the design and modification of the analysis and in
the interpretation and communication of its results.
Perhaps, an optimal allocation would involve a multiple lines of defense strategy that
incorporates the principles and lessons of integrated coastal zone management (ICZM)
and includes the combined buffering impacts of the offshore shelf within the Gulf of
Mexico, the Louisiana barrier islands, the Louisiana sounds, marsh land bridges, natural
ridges, man-made soil foundations, flood gates, flood protection levees, flood protection
pumping, elevated homes and businesses, and enhanced and more timely evacuation
procedures (see IPET, 2006).

126

Chapter 5: Mitigating Potential, Intentional Disasters at a Pre-existing
Regional Infrastructure Complex: The Port Security Risk Analysis and
Resource Allocation System (PortSec)

Introduction
This chapter describes a case of a preexisting, regional infrastructure system and an
ongoing project to design a system that applies countermeasures to mitigate
consequences of intentional disasters on this infrastructure.

The Ports of Long Beach and Los Angeles are a massive regional infrastructure complex.
This complex can be modeled as a network that is responsive to a matrix of potential
terrorist attack risk scenarios that emerge based on the current use of the complex and on
the current countermeasures used by the complex. This model is part of an ongoing,
software development project that considers a system of systems approach to risk and
decision analysis with an ability to alter mitigation measures on a near real-time basis
within the problem domain.

Background
Seaports, airports, and other freight and cargo nodal points face many challenges.
Maximizing operational efficiency, minimizing risk from terrorism or other disaster
events, and minimizing impacts to the environment are among these tests. Often these
challenges are at odds with one another – increasing one often comes at the expense of
achieving others. For example, in a seaport environment, increasing port security (by
reducing the probability of attack) by adding additional container inspection stations

127

often causes a corresponding slowdown in container movements through the port which
also has an impact on the environment (e.g., increased emissions). The challenge is
finding the right balance between operational efficiency, security, safety, and the
environment. The Port Security Risk Analysis and Resource Allocation System (or
PortSec) is designed to address these challenges.

Trade-offs among operational efficiency, security, safety, and the environment can be
either tactical or strategic. Tactically, there is a need to adjust in near real-time resource
allocations to maximize security (i.e., reduce risk from an attack) while simultaneously
minimizing impact on day-to-day operations and the environment. For example, if a VIP
is visiting the port, what is the assessed increase in attack risk and the anticipated
reduction in that risk resulting from reallocating patrol cars and boats to accommodate
the event?

Strategically, there is a need to forecast the impact that future facility expansions and new
technologies will have on operations, security, safety, and the environment. Included in
this strategic analysis is the benefit of implementing new and upcoming security
technologies on current operations. For example, is the cost in implementing a new
container inspection technology (including the cost of removing the existing system)
worth the anticipated reduction in attack risk?

The challenge in undertaking these trade-off analyses is that transport nodal points such
as seaports are complex, composed of many different components with varying degrees

128

of dynamic interaction. Such systems-of-systems
34
are difficult to model and, therefore,
difficult to understand and to use in answering such basic resource allocation questions as
to what is the impact that a particular countermeasure, when implemented, will have on
facility operations and throughput.

PortSec seeks to provide both tactical and strategic risk assessment and resource
allocation support that simultaneously addresses business continuity, security, safety, and
the environment. Currently, the Ports of Los Angeles and Long Beach (POLA/LB) are
the focal environment/domain of the project. Combined, the POLA/LB are the nation’s
largest container port and the world’s sixth largest container port. Any disruption to
normal business from a successful terrorist attack or from the implementation of a
counter-terrorism countermeasure will ripple through the nation’s economy.

Background/State of Art
Currently, there are a number of tools and utilities that address port security risk
assessment and resource allocation, however, the USCG’s Maritime Security Risk
Analysis Model (MSRAM) is perhaps the most well-known and the current gold
standard. MSRAM serves as a terrorism risk analysis software tool and is used to
perform scenario risk assessments on critical infrastructure and key resources. MSRAM
aggregates its assessments and provides analyses to support risk management. Its
methodology captures the security risk facing different targets spanning industry sectors

34
Jamshidi, M., “System-of-Systems Engineering - A Definition,” IEEE SMC 2005, 10-12 Oct.
2005.

129

and allows comparisons among different targets and geographic areas at the local,
regional, and national levels.

Advantages of PortSec
Although MSRAM helps identify high risk areas of the port complex and similar
“situational awareness” applications help inform decision-makers, these solutions
typically model only portions of the overall port complex and do not adequately address
business continuity (especially from the local perspective) and usually only focus on
either tactical or strategic operations and rarely on both. In addition, these solutions
rarely address costs to the environment. PortSec helps port security personnel minimize
risk from attack while maximizing business continuity with minimum or no cost to the
environment. PortSec has been designed to address these concerns.

The complex operations and interactions within and between the various physical and
non-physical components that make up a port complex are difficult to model and make
risk assessment and management difficult. The purpose of PortSec is to capture this
dynamic “system of systems” environment and to provide port authorities with a risk
assessment and management system that can be used for both tactical and strategic risk
management and resource allocation.

State of Development
A prototype that supports tactical day-to-day use has been implemented. External data
sources such as the Marine Exchange (which provides maritime vessel movements) and

130

DOT traffic flow information are simulated. This prototype has been demonstrated to the
POLA/LB and to DHS Science & Technology Directorate.

Technical Description
There are a number of challenges in developing the PortSec system, including:

• Modeling of a complex system of systems that represents port infrastructure
operations,
• Modeling of port risk,
• Development of consequence models,
• Methods for communicating risk to stakeholders,
• Combining tactical, real-time interactive, and strategic support,
• Developing an infrastructure that focuses on reducing risk from attacks or
consequence mitigation from a successful attack while maintaining port
operations (business constancy/resiliency),
• Interfaces to real-time external data systems, including GIS, traffic, rail, and cargo
movement information,
• Sensitivity of the intelligence,
• Verification and validation,
• Generalizing the system to support other port complexes and similar operations,
• Developing methods of scoring mitigation impacts through data analysis and/or
expert elicitation, and

131

• Predicting probabilities of attack or, as a proxy, developing methods for modeling
reasonable distributions of attack scenarios.

In addressing these challenges, two simultaneous parallel efforts are being undertaken: 1)
model the port complex as a “system of systems” where each terminal, pier, rail line,
roadway, and other identified “areas of interest (AOIs)” of the port are treated as systems
with their own operations and interfaces to other systems and 2) undertake the necessary
research to develop the underlying risk assessment and resource allocation algorithms for
each of these “systems” as well as the overall port complex.

System of Systems
The overall PortSec system architecture consists of a user interface, modeling
infrastructure, and data stores (Figure 34). The graphical user interface (Figure 35)
provides the user with access to either the tactical or strategic components of the PortSec
system. The Modeling Infrastructure is composed of the risk assessment, terminal and
port operations, transportation, economic and environmental assessment models.
Middleware consisting of both APIs and simple file transfers are used to link these
models and internal/external data systems together in a “system of systems” software
architecture. The middleware also provides the link between this infrastructure and the
graphical user interface.

The data stores are composed of databases and system that are both internally and
externally maintained. The internally maintained databases and systems hold the model
parameters, scenario definitions, and user-specified data. The external databases and

132

systems store information such as port resource allocations and port operation
environment. In the working demonstration prototype (prototype 1.0), many of these
external data sources are currently simulated.

The system design is intentionally modular to support future upgrades to system
components (e.g., a new terminal model) and expansion into supporting other port and
non-port infrastructures.
35

35
Another aspect of PortSec is port operations modeling. In this module, port operations are
simulated via a system of systems network of three modules. The first module is a macroscopic
model of the flows of containers and trucks in and out of terminals and ports within the capacity
constraints of each terminal and its characteristics. The second module is a microscopic traffic
simulator of the trucks and traffic on the port’s surrounding roadway network. The third module
is a terminal cost model that assesses the impact of cost on the terminal due to changes in
operation, equipment characteristics etc. The three modules are integrated into a system of
systems solution that is accessed in PortSec by middleware (Figure 33). The environmental
model is an extension of the Comprehensive Modal Emissions Model (CMEM). This model
predicts fuel consumption and gas emissions of vehicles based on the velocity and acceleration of
the vehicles from the microscopic traffic simulator. Once fully implemented, PortSec will not
only provide an estimate of the change in assessed risk associated with a resource and/or
infrastructure modification, it will also provide an estimate of the cost to the environment
resulting from that change.

133

Figure 34: PortSec System Architecture
Concept of operations
Tactical operations

The port complex is divided into user definable AOIs (colored regions in Figure 35).
For each AOI, the PortSec software automatically and periodically calculates a risk
assessment based on assessed threat, vulnerability and consequences of various AOIs and
reflects current collected intelligence and countermeasures (CM) allocated to each AOI.
Collected intelligence includes current and anticipated movements of maritime vessels,
movement of truck freight, cargo contents, highway traffic flow, and anticipated events
(e.g., port visit by a VIP). Currently implemented CMs include location of law
enforcement resources, scanning technologies, inspection stations and surveillance
technologies.
User Interface

Requests Results
Dynamic
Data
Parameters
Parameters
Countermeasure
Allocations
Operational
Environment
No. of police, patrol
cars, boats, cameras,
etc.,
Number of ships in berths,
events, auto traffic
conditions, etc.
Internally maintained databases
Externally maintained databases
ACTA, e-Modal, PierPass,
Marine-Exchange, AVL,
ATMIS, Sonar/Radar Info
Model
Parameters
Parameters
Middleware

Modeling
Infrastructure
Parameters
Results
Parameters
Risk Assessment & Resource
Models
Port/Transport models
Environmental cost models

134

In addition to individual AOI risk calculations, an overall port complex risk assessment is
calculated based on aggregation of the various AOI risk assessments. If a user makes a
resource allocation adjustment or if the operating environment changes (e.g., new
maritime vessel has docked, etc.), the PortSec software automatically recalculates the
assessed risk of each AOI as well as the overall port complex. Areas of interest are
painted green (low risk), yellow (moderate risk), and red (high risk) based on the risk
assessment calculated. In addition, since many of the external data sources (e.g., Marine
Exchange) provide advanced notice of pending vessel movements and resource
allocations, PortSec also calculates anticipated risk into the future for all AOIs as well as
the overall port complex (top of Figure 35). Finally, port security personnel can also
enter anticipated future events (lower right side of Figure 35) to help with anticipating
future risk.

The port security officer can allocate port resources by either “dragging” a resource to a
desired AOI or adjusting sliders allocated to each available port resource (see Figure 35).
After each adjustment the PortSec software recalculates the assessed risk and repaints the
various AOIs to reflect those assessments. In its next iteration, the system will support
optimization algorithms that will suggest resource allocations based on the current
environment.

Strategic Operations
The focus of the next phase of the project is to extend the system to support strategic risk
assessment and resource allocation. Once strategic support is implemented, decision-

135

makers can use PortSec to evaluate 1) the trade-offs (costs vs. anticipated effectiveness)
of implementing new and/or improved counter-measures against terrorism and/or natural
hazards, 2) the impact infrastructure improvements or modifications will have on both
AOI and overall port security and 3) the impact such improvements will have on
environment. As with tactical operations, the PortSec software periodically re-calculates
the risk assessment for all AOIs.
Figure 35: Tactical User Interface – PortSec 1.0. Regions in green, yellow, and red
represent areas of interest (AOI) and are user defined. Colors represent
assessed level of risk (green – low, yellow – moderate, red – high).

Risk modeling and resource allocation
A significant challenge faced in developing the PortSec system was how to model risk
assessment and resource allocation in a port complex composed of many different areas
Highlighted
AOI
Calculated
risk for
highlighted
AOI and
overall port
l
Anticipated
events
Tactical
resource
allocation
(sliders)
Tactical
resource
allocation
(drag &
drop))
Calculated risk into the future for
selected AOI and overall port complex

136

of interest such as cargo-handling terminals and rail systems, business entities, and
various local, state, and federal law enforcement agencies—each with its own priorities,
operations, and concerns. This challenge was compounded by the ability to model—
with a specified degree of accuracy—the predicted impact that various terrorism-based
scenarios might have on the overall operation of the port complex and the economic costs
resulting from interdicting, responding, and recovering from these terrorism events.
Specifically, what predicted disruptions to operations that ripple throughout the port
complex resulting from a terrorist attack would occur? In addition, for each attack
scenario, what countermeasures might be implemented to prevent or reduce these impacts
(both direct and indirect) while simultaneously minimizing impacts to day-to-day
operations (i.e., throughput) within the port complex?

A terrorism risk analysis of the Ports of Long Beach and Los Angeles is a decision
analysis of port security alternatives in terms of scenarios, infrastructure hardening, and
defenses. Such an analysis begins with an assessment of the multiple layers of security
directed at reducing the potential exposure to and consequential severity of conditions
that may induce potential targeting by terrorists—the threats that induce the selected
security activities, the vulnerabilities to these threats, the flexibility in terrorists’
potential attack modes and targeting, the consequences of attacks, and the ability to
reproduce attacks—and in neutralizing the advantages of any specific modes of attack,
whether by water, air, road, rail, perimeter intrusion, deceptive action, pipeline, or insider
activity. This can be accomplished by expert elicitation that subjectively weighs the

137

targeting selection criteria and the specific target preferences to create more efficient
allocations of the ports’ security resources.

After delineating the factors involved in targeting—the existence of viable targets,
potential vulnerabilities in these targets, flexibility in attacking these targets—and
eliciting the target categories—inbound maritime, road, rail, pipeline, insider threat,
deceptive activity, and air—and locations within the ports, judgments can be made by
ranking expert elicitations over individual target characteristics. Ranks can be derived
for each characteristic, and diverse and otherwise potentially incommensurable
characteristics can be compared to one another in a rational and consistent way.
36
From
these inputs, an implied allocation of security resources by port area-type to be defended
based on specific scenarios can be calculated.

One way of thinking about the risk analysis embedded in PortSec is decision trees. The
security costs by port area-types to be defended can be derived based on analyses using
decision trees. Decision trees can be crafted for each representative attack scenario. For
example, a “bomb hidden in an inbound container scenario” decision tree could follow a
four-stage decision that occurs when the container ship approaches the Ports of Long
Beach and Los Angeles. First, a decision is made with respect to the ship and its

36
In the current analysis, inputs were judgmentally selected by the authors. More representative
inputs can be obtained by expert solicitation. For example, distribution of a survey form to
various military, security, maritime, technical, and administrative port personnel familiar with
Ports of Long Beach and Los Angeles security concerns have been and are being conducted. The
results of these surveys will be aggregated and, as needed, re-solicited using Delphi or other
expert elicitation techniques. In such a manner, the inputs will be more representative of an
expert consensus.

138

containers if the U.S. Coast Guard wishes to board the ship and inspect containers while
the ship is still at sea. The tree assigns probabilities and costs to this contingency. The
next decision is made once the containers are off-loaded at the port. In this decision, a
sample of containers is passed through a nuclear/radiation portal. Based on a negative
result (i.e., no detection of radiation), strata of containers are then either physically
inspected or scanned with personal radiation detectors. Finally, those containers that
successfully pass the radiation detector screening are either physically inspected or not.

Based on this specific attack scenario, simple cause-and-effect models are utilized. In
order to evaluate the likeliest course of action by possible terror cells, a range of decision
trees and sensitivity analyses that includes tornado diagrams can be modeled to review
the outcomes of various decisions. The decision trees represent the choices involved in
attacks assuming employed countermeasures and, given these choices, the attacks’
probabilities of success or failure. The scenarios developed, for example, could target or
exploit a cruise ship, a container ship, a tanker ship, a harbor truck, a barge, a rail yard,
port industrial facilities, or a bridge. Other scenarios could include terrorist stowaways
on an inbound hazardous cargo vessel, an explosion at a fuel receiving terminal, a
suspicious package at a port facility, surveillance of petrochemical terminals, an
improvised explosive device (IED) attached to the hull of a freighter, the theft of gasoline
tanker truck, an explosives attack on a chlorine storage tank, hostage-taking and
executions aboard a vessel in port, underwater explosive devices planted on multiple
vessels in port, a nuclear device aboard an incoming vessel in a 55-gallon drum, an attack
on the ports with a biological disease agent, detonation of a “dirty” bomb in a shipping

139

container in port, an aircraft attack on a cruise ship, ammonium nitrate bombs shipped by
rail to a port, a sarin gas attack on a cruise ship, an explosives attack on a ship in port,
underwater and fishing boat explosives attacks, bombing and sinking of a liquefied
propane gas (LPG) tanker, hijacking of a tanker for use as a “floating bomb,” ramming
and “dirty” bombing a cruise ship with a hijacked cargo ship, coordinated bombing of
docks and bridges, and mining of the harbor.
37
Plotting of expected consequence versus
conditional additional expected costs based on various countermeasure and defense
options defines a border of Pareto optimal port security options based on the dominance
relationships of selected target/mitigation scenarios.

Given the extremely large number of targets, threats, vulnerabilities, and potential
consequences in this problem, the use of decision trees quickly becomes impractical and
nearly intractable. For this reason, we use influence diagrams in this risk analysis. The
influence diagram approach serves as a tool for the communication of high-level findings
and as a means of examining the impact and value of individual mitigation decisions. An
understanding of how the results are obtained and of how the various assumptions impact
the results is often more important than the specific input and output numbers. In
addition to communicating high-level findings, PortSec stakeholders can examine lower
levels of modeling when more detail is desired, aided by the visual aspects of the model’s
structure. As stakeholders are able to understand the model, debate and discussion can

37
The source of these DHS scenarios is Paul W. Parfomak and John Frittelli (2007), Maritime
Security: Potential Terrorist Attacks and Protection Priorities, Resources, Science, and Industry
Division, Congressional Research Service.

140

focus more directly upon specific assumptions and lead to more productive results. Thus,
the influence diagram serves as a tool to help to make the model accessible.

Specifically, we used the software tool, Analytica (see www.lumina.com), to assist in
modeling an influence diagram that represents the interrelationships among variables that
include data for significant port locations, various manner of threats, and the potential for
negative consequences as well as potential prevention or mitigations with their impacts
and costs.

The draft influence diagram depicted in Figure 36 calculates expected consequences
given various threat and mitigation combinations. Once the actual ship movement data
and expert elicitations enable more accurate mitigation effect factors and mitigation
choices, the influence diagram enables the optimization of mitigation resource allocations
based on the input criteria.

This tool, in coordination with the graphical user interface and other modules of PortSec,
enables short-, medium-, and long-range security planners to manipulate existing security
assets and to coordinate future security assets in a more efficient and speedy manner.

The risk calculation is based on the Department of Homeland Security’s (DHS’s) risk
assessment methodology. This methodology separates risk into three components: threat,
vulnerability and consequence (Risk = Threat x Vulnerability x Consequence). Threat is
defined as the likelihood that an attack occurs, vulnerability is the likelihood of that an
attack succeeds if attempted, and consequence is a measure of damage inflicted on the

141

port as a result of the attack. This framework is similar to that used in the United States
Coast Guard’s (USCG’s) Maritime Security Risk Analysis Model
38
(MSRAM).
Maritime Security Risk Analysis Model (MSRAM)
MSRAM assesses risk based on scenarios—combinations of targets and attack modes—
in terms of threat
39
, vulnerability
40
, and consequence
41
. MSRAM considers the response
capability of the owner or operator, local first responders, and the USCG and other
federal agencies to mitigate the consequences of terrorist attacks.

Within MSRAM, a target is defined in terms of class, location, potential maximum
consequences, geographic location, and other key factors. Based on class and
consequences, a selection of attack modes
42
is specified for analysis of consequences and
vulnerabilities. Consequence scoring is based on a value system that considers morbidity

38
A major tool in the conceptual development of PortSec and its risk analysis module is the
USCG’s MSRAM. MSRAM is a MS Access-based terrorism risk analysis software tool used by
the USCG to perform scenario risk assessments on critical infrastructure and key resources.
MSRAM is based on the premise that risk is a function of threat, vulnerability and consequences;
it aggregates scenario-based risk assessments and provides discrete analyses to support risk
management mitigations. Its methodology captures the security risk facing different targets
spanning industry sectors and allows comparisons among different targets and geographic areas at
local, regional, and national levels of interest.

39
The intent and capability of terrorists to deliver an attack on a specific target.

40
The probability of a successful attack based on attack difficulty, ability of the owner/operator,
local law enforcement, and the USCG to interdict an attack, and ability of the target to withstand
the attack.

41
The negative impact of a successful attack on the United States in terms of deaths/injuries,
primary economic impact, environment, national security impacts, symbolic impacts, and
secondary impacts to the US economy.

42
Threat data is provided by the USCG Intelligence Coordination Center for all combinations of
attack modes and target classes.

142

and mortality, primary and secondary economic impacts, environmental impacts, national
security significance, and symbolic significance. The values for these consequences are
determined by expert elicitation. Vulnerabilities are segregated into factors: attack
achievability, system security, and target hardness.

MSRAM produces separate valuations of primary and secondary economic impacts of
terrorist attacks to improve the USCG’s understanding of response and recovery
capabilities, dynamic analysis and reporting to review assessment results, an “as if” case
module to assess and compare the impact of alternative strategies, and a three-tiered
(local, regional, and national scopes) review of risk assessments. MSRAM serves as the
generalized template for the development of PortSec. MSRAM’s outputs are directed
with respect to the USCG’s mission; PortSec’s outputs are directed with respect to the
mission of the port operator/owner while maintaining the realization that integration of
the PortSec system’s treatment of security concerns into the more comprehensive scopes
of higher level systems’ treatment of security concerns is a necessary prerequisite and
part of an exogenous (as well as endogenous) “system of systems’ approach” to port risk
management.

When a counterterrorism resource is moved from one Area of Interest (AOI) to another
emergent AOI, the overall vulnerability is reduced. In addition, since vulnerability is
reduced (e.g., extra patrol boats are positioned next to a potentially targeted cruise ship),
the threat of an attack on that target is also reduced. The end result is that the overall risk
of an attack on that target is reduced.

143

Figure 36: Example PortSec Influence Diagram

The terminal, port, and transportation models are used to calculate the cost of
implementing a counter-measure to reduce risk. In many cases, there may not be any
cost increase (e.g., moving a patrol boat from one pier to another). In other cases, there
may be a cost due to anticipated slow-down of throughput (e.g., implementing an
inspection station to search all trucks will slow truck traffic and, therefore, add costs to
the operation of the terminal operator). Factors influencing threat estimates for each
attack scenario include attacker type (reflecting the intent and capability of each
attacker), available target areas (including port terminals and several non-terminal areas
such as bridges), weapons (IED, dirty bomb, etc.), and approach vector types (road,
water, rail, etc.). Vulnerability estimates, as well as the current list of targets, incorporate
available risk analyses. Consequences of attack depend on the target, weapon,
mitigations, and port conditions.

144

Ultimately, PortSec will operate in three modes—tactical, incident response, and strategic
(see Figure 37). Tactical mode is the default mode. It is the mode used by a security
analyst in allocating security countermeasures for day-to-day scheduling. Incident
response is tactical mode enhanced with real-time geospatial parameters of
countermeasures for reallocation on the fly—particularly, during the time immediately
after an incident. Strategic mode is a long-range planning mode. It combines tactical
mode’s features with the ability to enhance or delete infrastructure and countermeasure
components for planning purposes.

Maritime Assessment and Strategy Toolkit (MAST)
Another major data source tool and the primary source for initial scenario input data in
the development of PortSec’s risk analysis module is the Maritime Assessment and
Strategy Toolkit (MAST).
43
This report focused on “providing risk management
solutions to assets located in the ports.” (MAST, p. xvii) This report divides target assets
into four risk severity groups. The report identifies six of-interest classes of
countermeasures—closed circuit television, employee alerts/notification/signage,
patrols/guards, fences/gates, hardened perimeters, and vehicle check points—in addition
to Homeland Security Presidential Directive 8 response capability factors
(staffing/personnel, training, equipment/systems, planning/preparedness, exercises,
evaluations, and corrective actions, and organization/leadership (STEPCO)). PortSec
directly utilizes the combined criticality factors within its algorithms as a means to

43
Port of Los Angeles/Long Beach Needs Assessment Project, Draft Project Report, Volume 2,
Summary Results, February 21, 2008.

145

overlay relative risks to target assets by threat type. Threats considered include large
conventional explosives on land, large conventional explosives in water (by scuba diver,
by swimmer, and by boat), small conventional explosives in transit, small conventional
explosives in buildings, ramming a boat into a target, use of biological weapons on land
or in a building, use of chemical weapons on land or in a building, and use of radiological
weapons on land or in a building.

Figure 37: Example PortSec Operating Modes

146

Preliminary Data and Results
Table 11 displays a table showing the input, derived, and output variables and their
respective descriptions and/or methods of calculation for PortSec’s risk analysis and
countermeasure allocation module.
Table 11: Data Dictionary—Summary of PortSec Risk Analysis Model Variables

Number
Variable Values or Description
1 Target areas
These values are omitted since the exact targets are SSI.
2 Coordinates of target areas
Latitudes and Longitudes corresponding to each Target Area
3 Attack modes
['BIO Bldg','CHEM Bldg','LCE','LCE water ','Ramming','SCE ','SCE Bldg','SCE
scuba','SCE swimmer'] (These are from MAST.)
4 Baseline Threats
An array of (Target Areas, Attack Modes) 9 (All values are from the MAST study
at the Ports of LA & LB; see the MAST study for more details.)
5 Baseline Vulnerabilities
An array of (Target Areas, Attack Modes) 9 (All values are from the MAST study
at the Ports of LA & LB; see the MAST study for more details.)
6 Baseline Consequences
An array of (Target Areas, Attack Modes) 9 (All values are from the MAST study
at the Ports of LA & LB; see the MAST study for more details.)
7 Dependence of
Consequences on Ship
Movements
An array of (Target Areas, Attack Modes) 9 (All values are judgmental. Ideally,
these should be derived from expert elicitations.). Value = “Yes” or “No”
8 Defensive resource types
['Patrol boat unit','Patrol car unit','Pan & tilt camera']. Values may include any
type of countermeasure.
9 Risks before defense
resource allocations
Baseline Vulnerabilities* Baseline Threats * Baseline Consequences
10 Exchange data on ship
movements
Not active. At each time update, a new set of data is produced. The update
changes the Dependence of Consequences on Ship Movements Variable. This
is the position for the Marine Exchange data.
11 Consequences as a function
of ship movements
This variable is the restatement of Risks before defense resource allocations
after each time interval based on Exchange data on ship movements.
12 Port risk as a function of
time before defense
resource allocations
Sum(Sum(Risks before defense resource allocations, Target areas), Attack
modes)
13 Risk before defense
resource allocations
breakdown by target, time
Sum(Risks before defense resource allocations, Attack modes)
14 Risk before defense
resource allocations
breakdown by attack mode,
time
Sum(Risks before defense resource allocations, Target areas)
15 Defensive resource
allocations
Table(Target areas, Defensive resource types) Optimize this resource
allocation decision. Does not currently optimize.
16 Nominal effects of
defensive resource
allocations on
vulnerabilities
Not accurate. These are currently subjective estimates. These should be
determined by expert elicitation.
17 Vulnerabilities as a
function of defensive
resource allocations
Baseline Vulnerabilities * Product( (1 - Nominal effects of defensive resource
allocations on vulnerabilities)^ Defensive resource allocations, Defensive
resource types)

147

Table 11, continued

Number
Variable Values or Description
18 Defensive resources
available
Table(Defensive resource types); An integer for each type. This is a cross-check
variable such that allocations cannot exceed available resources.
19 Port risks after defensive
resource allocations
Vulnerabilities as a function of defensive resource allocations * Baseline
Threats * Consequences as a function of ship movements
20 Baseline risk before
defense resource
allocations breakdown by
target area
Sum(Risks before defense resource allocations, Attack modes)
21 Baseline risk before
defense resource
allocations breakdown by
attack mode
Sum(Risks before defense resource allocations, Target areas)
22 Port risk before defense
resource allocations
Sum(Sum(Risks before defense resource allocations, Target areas),
Attack modes)
23 Port risk as a function of
time after defense resource
allocations
(Sum(Port risks after defensive resource allocations, Target
areas), Attack modes)
24 Risks after defense
resource allocations
breakdown by target, time
Sum(Port risks after defensive resource allocations, Attack
modes)
25 Risks before defense
resource allocations
breakdown by attack mode,
time
Sum(Port risks after defensive resource allocations, Target areas)
26 Region risk threshold
values
Table(Region risk color code threshold elements); Integer values
27 Region risk color code
threshold elements
['region med threshold','region high threshold','port med threshold','port high
threshold']

Much of this data¸ including the target data and the threats, vulnerabilities, and
consequences to each target, is derived from the MAST study. Some of this data, in
particular, variables 7, 10, and 16 from Table 11, is problematic. Variable 7 adjusts
expected consequences based on ship movement. In other words, it applies a factor to the
MAST consequences, which increases the expected consequences for an event at a berth
or other target whenever a ship is berthed at or near the target. The current variable
contains judgmental estimates that are essentially placeholders. An expert elicitation of
values (or an elicitation of a range of estimates that can be manipulated by a user defined

148

slider over a reasonable range) would be necessary to calibrate this more accurately.
Variable 10 refers to data on the geographic coordinates of specific ships at specific times
on a real-time basis. Currently, real-time data is not captured within the model, and
dummy variables serving as placeholders are used instead. Variable 16 assigns
countermeasure reduction factors which lower expected consequences by target-attack
mode scenario based on cumulative reduction factors developed from the cumulative use
of countermeasures. These measures, as well as measures of maximum numbers of
countermeasures and measures of diminishing utility of accumulated countermeasures,
are currently judgmentally estimated. Once again, an expert elicitation of a value (or an
elicitation of a range that can be manipulated by a user defined slider over a reasonable
range) would be necessary to calibrate this more accurately.
Despite these limitations, an Analytica
®
model within PortSec was developed to allocate
and measure countermeasure impacts based on current port use. This Analytica risk
module contains summary desciptions of Security-Sensitive Information (SSI) about real-
world targets and their vulnerabilities from the MAST report. The SSI data are the list of
targets at the Ports of Los Angeles and Long Beach and the associated Threat,
Vulnerability and Consequence scores as derived from the Port of Los Angeles/Long
Beach Maritime Assessment and Strategy Toolkit (MAST) study.

Figures 38 through 40, inclusive, display conditional expected losses for each of fifty-six
selected MAST Targets for each of the nine defined MAST terrorist attack scenarios.
Note that these figures are intentionally displayed at a scale that obscures detail because

149

of the sensitive nature of the data. This scaling, however, does not impact the general
points of these comparative figures.
• Figure 38 displays these losses assuming that all berths are active (i.e., occupied
by ships in port) and that security is consistent with the status quo security
allocations at the time the MAST study was performed (this is assumed to be the
base security level of the ports).
• Figure 39 displays these losses assuming that all berths are active (i.e., occupied
by ships in port) and that security is enhanced to a level above the status quo
security allocations at the time the MAST study was performed (this is an
assumed, enhanced security level of the ports). The horizontal spikes on Figure
39 are shorter than the analogous spikes on Figure 38 to indicate the reduction in
expected losses due to the additional security allocations.
• Figure 40 replays Figure 39 and displays its losses assuming that only selected
berths are active (i.e., occupied by ships that are currently in port) and that
security is enhanced to a level above the status quo security allocations at the time
the MAST study was performed (this is an assumed, enhanced security level of
the ports). There are fewer horizontal spikes on Figure 40 than the analogous
spikes on Figure 39 since all berths are not active. This figure displays a
snapshot in time of security allocations and port activity that can be compared to
other security configurations and/or activity scenarios.

The draft influence diagram depicted in Figure 36 calculates expected consequences
given simulated, realistic threat and mitigation combinations as displayed in Figure 40.

150

Once the actual ship movement data and expert elicitations enable more accurate
mitigation effect factors and mitigation choices, the influence diagram will be able to
calculate real threat and mitigation combinations and to optimize resource allocations
based on input criteria.

This tool, in coordination with the graphical user interface and other modules of PortSec,
will enable short-, medium-, and long-range security planners to manipulate existing
security assets and to coordinate future security assets in a more efficient and speedy
manner.

151
Figure 38: Ports of Long Beach/Los Angeles: Base Case—Conditional Expected Losses for each of 56 MAST Targets given 9
MAST Attack Scenarios (Assumes All Berths are Active and Status Quo Security Allocations)

Queen Mary (POLB asset list)
Vincent Thomas Bridge (POLA A4)
Queens Way Bridge (POLB asset list)
Heim & Badger Bridges (POLB asset list)
Chemoil Company (POLB asset list)
General Petroleum (POLA asset list)
Gerald Desmond Bridge (POLB asset list)
BP Marine Terminals (POLB asset list)
Shell Oil Products US-Mormon Island (POLA asset list)
Kinder Morgan (Terminal 7 Berths 118-120)
Petro Diamond Liquid Bulk Terminal (POLB B3)
Westway Terminal Company (POLA asset list)
Exxon Mobil Oil Corp. (POLA A9)
Shell LB84 (POLB asset list)
Conoco Phillips (Terminal 10 Berths 148-151)
APM Terminals/ Pier 400 Container Terminal (Terminal 25 Berths 401-406)
Matson Container Terminal (POLB asset list)
Vessel Traffic Service (VTS)
Vopak (Terminal 17 Berths 187-191)
National Gypsum (POLB asset list)
International Transport Service (POLB asset list)
LB Container Terminal (POLB asset list)
Pacific Container Terminal (SSA) (POLB asset list)
Pacific Coast Recycling (POLB asset list)
Borax (Terminal 14 Berths 165-166)
Baker Commodities (POLB asset list)
LB Aquarium of the Pacific (POLB asset list)
Ports O Call (POLA asset list)
TraPac Container Terminal (Terminal 9 Berths 135-139)
Sea Launch (POLB asset list)
Catalina Passenger Ferry Terminal (POLA asset list)
Evergreen Container Terminal (STS) (POLA asset list)
ZIM Container Terminal (SSA)
Yusen/NYK Container Terminal (POLA asset list)
Hanjin Container Terminals (TTI) (POLB asset list)
APL Container Terminal/ Global Gateway South (Terminal 24 Berths 302-305)
Morton Salt Co. (POLB asset list)
Ga-Pacific Gypsum (POLB asset list)
Hugo Neu Proler (POLB asset list)
SSA Containers, INC., Pier F LB (POLB asset list)
CA United Terminals (POLB asset list)
BP Pier T-121
SSA (POLA Terminal 2 Berths 54-55)
POLA Admin. Bldg (POLA asset list)
Jankovich Company (POLA asset list)
POLB Admin. Building (POLB asset list)
Pasha (Terminal 16 Berths 174-181)
Mikes Main Channel Marine (POLA asset list)
Island Platforms (POLB asset list)
Offshore Platforms
Cruise Ship Anchored at Avalon (asset list)
Tanker at Anchor
0 10K 20K 30K 5000 15K 25K 35K

Risks before defense resource allocations
Attack modes
BIO Bldg CHEM Bldg LCE LCE water Ramming SCE SCE Bldg SCE scuba SCE swimmer

152
Figure 39 Ports of Long Beach/Los Angeles: Enhanced Base Case—Conditional Expected Losses for each of 56 MAST
Targets given 9 MAST Attack Scenarios (Assumes All Berths are Active and Selected Enhanced Security
Allocations)

Queen Mary (POLB asset list)
Vincent Thomas Bridge (POLA A4)
Queens Way Bridge (POLB asset list)
Heim & Badger Bridges (POLB asset list)
Chemoil Company (POLB asset list)
General Petroleum (POLA asset list)
Gerald Desmond Bridge (POLB asset list)
BP Marine Terminals (POLB asset list)
Shell Oil Products US-Mormon Island (POLA asset list)
Kinder Morgan (Terminal 7 Berths 118-120)
Petro Diamond Liquid Bulk Terminal (POLB B3)
Westway Terminal Company (POLA asset list)
Exxon Mobil Oil Corp. (POLA A9)
Shell LB84 (POLB asset list)
Conoco Phillips (Terminal 10 Berths 148-151)
APM Terminals/ Pier 400 Container Terminal (Terminal 25 Berths 401-406)
Matson Container Terminal (POLB asset list)
Vessel Traffic Service (VTS)
Vopak (Terminal 17 Berths 187-191)
National Gypsum (POLB asset list)
International Transport Service (POLB asset list)
LB Container Terminal (POLB asset list)
Pacific Container Terminal (SSA) (POLB asset list)
Pacific Coast Recycling (POLB asset list)
Borax (Terminal 14 Berths 165-166)
Baker Commodities (POLB asset list)
LB Aquarium of the Pacific (POLB asset list)
Ports O Call (POLA asset list)
TraPac Container Terminal (Terminal 9 Berths 135-139)
Sea Launch (POLB asset list)
Catalina Passenger Ferry Terminal (POLA asset list)
Evergreen Container Terminal (STS) (POLA asset list)
ZIM Container Terminal (SSA)
Yusen/NYK Container Terminal (POLA asset list)
Hanjin Container Terminals (TTI) (POLB asset list)
APL Container Terminal/ Global Gateway South (Terminal 24 Berths 302-305)
Morton Salt Co. (POLB asset list)
Ga-Pacific Gypsum (POLB asset list)
Hugo Neu Proler (POLB asset list)
SSA Containers, INC., Pier F LB (POLB asset list)
CA United Terminals (POLB asset list)
BP Pier T-121
SSA (POLA Terminal 2 Berths 54-55)
POLA Admin. Bldg (POLA asset list)
Jankovich Company (POLA asset list)
POLB Admin. Building (POLB asset list)
Pasha (Terminal 16 Berths 174-181)
Mikes Main Channel Marine (POLA asset list)
Island Platforms (POLB asset list)
Offshore Platforms
Cruise Ship Anchored at Avalon (asset list)
Tanker at Anchor
0 10K 20K 30K 5000 15K 25K 35K

Risks after defensive resource allocations
Attack modes
BIO Bldg CHEM Bldg LCE LCE water Ramming SCE SCE Bldg SCE scuba SCE swimmer

153
Figure 40 Ports of Long Beach/Los Angeles: Preliminary Test Case—Conditional Expected Losses for each of 56 MAST
Targets given 9 MAST Attack Scenarios (Assumes Selected Berths are Active {based on test Marine Exchange
berthing data} and Selected Enhanced Security Allocations)

Queen Mary (POLB asset list)
Vincent Thomas Bridge (POLA A4)
Queens Way Bridge (POLB asset list)
Heim & Badger Bridges (POLB asset list)
Chemoil Company (POLB asset list)
General Petroleum (POLA asset list)
Gerald Desmond Bridge (POLB asset list)
BP Marine Terminals (POLB asset list)
Shell Oil Products US-Mormon Island (POLA asset list)
Kinder Morgan (Terminal 7 Berths 118-120)
Petro Diamond Liquid Bulk Terminal (POLB B3)
Westway Terminal Company (POLA asset list)
Exxon Mobil Oil Corp. (POLA A9)
Shell LB84 (POLB asset list)
Conoco Phillips (Terminal 10 Berths 148-151)
APM Terminals/ Pier 400 Container Terminal (Terminal 25 Berths 401-406)
Matson Container Terminal (POLB asset list)
Vessel Traffic Service (VTS)
Vopak (Terminal 17 Berths 187-191)
National Gypsum (POLB asset list)
International Transport Service (POLB asset list)
LB Container Terminal (POLB asset list)
Pacific Container Terminal (SSA) (POLB asset list)
Pacific Coast Recycling (POLB asset list)
Borax (Terminal 14 Berths 165-166)
Baker Commodities (POLB asset list)
LB Aquarium of the Pacific (POLB asset list)
Ports O Call (POLA asset list)
TraPac Container Terminal (Terminal 9 Berths 135-139)
Sea Launch (POLB asset list)
Catalina Passenger Ferry Terminal (POLA asset list)
Evergreen Container Terminal (STS) (POLA asset list)
ZIM Container Terminal (SSA)
Yusen/NYK Container Terminal (POLA asset list)
Hanjin Container Terminals (TTI) (POLB asset list)
APL Container Terminal/ Global Gateway South (Terminal 24 Berths 302-305)
Morton Salt Co. (POLB asset list)
Ga-Pacific Gypsum (POLB asset list)
Hugo Neu Proler (POLB asset list)
SSA Containers, INC., Pier F LB (POLB asset list)
CA United Terminals (POLB asset list)
BP Pier T-121
SSA (POLA Terminal 2 Berths 54-55)
POLA Admin. Bldg (POLA asset list)
Jankovich Company (POLA asset list)
POLB Admin. Building (POLB asset list)
Pasha (Terminal 16 Berths 174-181)
Mikes Main Channel Marine (POLA asset list)
Island Platforms (POLB asset list)
Offshore Platforms
Cruise Ship Anchored at Avalon (asset list)
Tanker at Anchor
0 10K 20K 30K 5000 15K 25K 35K

Risks after defensive resource allocations
Attack modes
BIO Bldg CHEM Bldg LCE LCE water Ramming SCE SCE Bldg SCE scuba SCE swimmer

154

Observations and Implications
A risk and decision analysis model that dynamically applies countermeasures to mitigate
consequences of intentional disasters on an existing, regional infrastructure system is
complex. When considering a risk and decision analysis as part of a large-scale, “system
of systems” approach to project risk management that also can:
• alter mitigation types and magnitudes on a near real-time basis,
• measure changes in transportation flows resultant from security measures,
• monitor air pollution changes resultant from security measures, and
• map results in an interactive graphical user interface which incorporates a
geographic information system that tracks real-time countermeasure and target
movement within its problem domain,
there are several points that must be considered:
1. Decision and risk analysis used for reducing the risks in a complex infrastructure
system with multiple assets that pose different risks and need different protective
measures is particularly complex. Some of the complexities of this type of
analysis include:
a. The ability to track and map objects in real-time,
b. The ability to alter the risk calculations for each potential target in real-
time,
c. The interplay between the security countermeasures and the trade flows in
real-time, and
d. The desired trade-offs between unimpeded trade and tight security.

155

2. The need for expert elicitation must not be underestimated. PortSec requires
expert knowledge in areas such as:
a. Wait times associated with the implementation of various security
countermeasures,
b. Effectiveness metrics associated with the implementation of additional
units of each type of security countermeasure,
c. Measures of the maximum units of each type of countermeasure (marginal
utilities by type) and overall (marginal utilities),
d. Target selection (beyond MAST),
e. Attack mode selection (beyond MAST),
f. Expected delays associated with various scenarios, and
g. Seasonality/weather adjustments.
3. From a broader perspective, efforts such as PortSec potentially offer many
benefits and challenges as follows:
a. PortSec is potentially portable. Although each new port (or other
infrastructure system) would require its own set of maps, trade flows,
expert elicitations, etc., the idea of overseeing allocation and management
of security resources within a defined domain has potential.
b. Projects such as PortSec would benefit from a phased approach. The
planning process for a large-scale project might be served better if
development phases based on functionality were established. In other
words, the first phase of the project could consist of soliciting and

156

engaging the buy-in of project proponents which would include members
of the user community (e.g., security, terminal operators, port authorities,
USCG). Project proponent buy-in is essential in order to have a co-
operative source of data, to have a co-operative pool of experts for beta
test users, to have an initial user/client, to have an initial source to help
market the project to other potential project users, and to have a potential,
continuing project funding source. The next phase could be the
development of attack scenarios. Attack scenarios would include targets,
modes, countermeasures, impacts of countermeasures, constraints on
countermeasures, sensitivities of attack modes to target selection,
countermeasure deployment, and target conditions (e.g., whether or not a
ship is berthed). The impacts, constraints, and sensitivities would be
derived, in part, from expert elicitation. Once this stage was developed, it
would be tested with sample data. Subsequent phases would add layers of
functionality successively until the desired overall functionality was
obtained. Successive, iterative rounds of testing would be followed by
live use (shadowed with status quo procedures). The testing and live use
would be monitored and rated (internally, by the proponent, and
independently) until the beta system was installed. After a reasonable beta
phase, the system would be implemented and monitored for usability,
reliability, and effectiveness.

157

Chapter 6: Mitigating Natural Disasters Involving Regional
Infrastructure Systems: A Hurricane Tracking, Consequences, and
Mitigation Capabilities Modeling Tool

Introduction
This example involves a case of a preexisting, regional infrastructure system. Its
potential for disaster vulnerability is considered by the design of a system that applies a
countermeasure to mitigate consequences of unintentional disasters on this infrastructure.
In this chapter, the potential for controlling a specific type of impact of a specific type of
unintentional disaster on a regional infrastructure system is explored. Mapping of real-
time hurricane tracks and their potential storm surge zones is overlaid with settlement
demographics in order to more accurately estimate potential evacuation timing along the
U.S. Gulf and Atlantic coasts. Herein, the levers of evacuation and evacuation timing are
considered as means for mitigating the consequences of hurricanes.

Background
In this chapter, we develop a methodology for evaluating the risk associated with a
hurricane/flood and assessing the value of mitigating activities. The hurricane scenario
used in this chapter is derived from DHS National Planning Scenario 10, “Natural
Disaster – Major Hurricane” (DHS 2005). In this scenario, a Saffir-Simpson Category 5
hurricane makes landfall at a major metropolitan area (MMA). Sustained winds are 160
miles per hour with a storm surge greater than 20 feet above normal. As the storm moves
closer to land, a massive evacuation of one million people is ordered 24 hours before
landfall. Certain low-lying escape routes are inundated approximately five hours before

158

the eye of the hurricane reaches land. The hurricane ultimately causes 1000 fatalities and
5000 serious injuries.

Using this DHS scenario as a template, the similarly-sized Miami MMA was chosen as
the site for the hypothetical hurricane in this study. Major portions of the DHS MMA
flood scenario are similar to the 1926 Miami Hurricane, but on a scale correlated with
current population and infrastructure levels. Structures in low-lying areas are inundated.
Many older facilities collapse due to the swift moving water and structural degradation.
Newer facilities and structures sustain heavy damage to contents on their lower levels.
Most vegetation within the storm’s path are damaged or destroyed. Massive amounts of
debris accumulate. Service disruptions are numerous. Shelters throughout the region are
filled to capacity. Hundreds of people are trapped and require search and rescue. Until
debris is cleared, rescue operations are difficult because much of the area is reachable
only by helicopters or boats. Wind and downed trees have damaged nearly all of the
electric transmission lines within the area. Most communications systems within the
impacted area are not functioning due to damage and lack of power.

Thousands of residents are left homeless. There is a great need for drinking water, food,
ice, and medicine. Sewage treatment plants are damaged. Hazardous substances have
spilled into the floodwaters. A 95,000-ton tanker ship has struck a bridge, breached, and
is leaking oil into waters within the MMA.

The 20-foot storm surge has breached and overtopped flood control and hurricane
protection works. Transportation routes and port facilities are damaged. Many hospitals

159

are damaged and overwhelmed. Schools are being used as shelters for the disaster
victims. Many pets, domesticated animals, and wild animals have been killed or injured.
There are severe economic repercussions for the entire region.

An Analytica influence diagram risk and decision analysis model was developed that
utilizes DHS National Planning Scenario 10 as a major input. We used the software
tool, Analytica (see www.lumina.com), to assist in modeling an influence diagram that
represents the interrelationships among variables that include data for wind,
demographics, and the potential for negative consequences of hurricanes and floods as
well as potential mitigations with their impacts and costs in the Miami MMA.

The specific scenario modeled is similar to an updated replay of the 1926 Miami storm.
The model estimates, depending on mitigations’ selected, current negative economic
consequences of between approximately $56 and $144 billion. Other authors have
estimated this hurricane’s economic consequences at between $101 and $157 billion.

Overview
Accurate and timely prediction of the location, magnitude, and impact of natural hazards
on human populations is difficult, and human development compounds that difficulty.
Whereas natural disasters disrupt an area’s economic life, development can alter an area’s
vulnerability to natural disaster either positively (e.g., more human resources are
available), or negatively (e.g., building on beaches or other high risk areas has occurred).
Because the occurrence of natural hazards is a certainty and the human resources devoted

160

to their mitigation are finite, careful planning and preparation for response and recovery
are critical. Risk analysis offers an important tool in this planning.

Risk analysis, based on quantitative and qualitative data, assists in the development
of mitigation plans and operational decisions. Qualitative risk analysis uses
descriptive scales to evaluate the likelihood and magnitude of risks or mitigations.
It can be used where data or resources adequate for quantitative analysis are
insufficient or inappropriate. In this chapter, qualitative analysis takes the form of a
capabilities matrix that incorporates mitigation capability/breadth and cost matrices.
Quantitative risk analysis, on the other hand, is conducted based on numerical values that
use geographic, meteorological, and physical observations and historic data to produce
estimates of the consequences and frequencies of risks.

Disaster risk reduction using a systematic approach to identifying, assessing, and
reducing risks associated with disaster hazards and with human activities can optimize
the use of scarce resources. This approach recognizes that a wide range of geological,
meteorological, environmental, technological, and socio-political hazards, individually
and collectively, threaten our communities. As such, disasters are not observed as
random events to which one can only respond, but are contemplated as long-term,
complex, natural and man-made problems subject to anticipation, preparation, and
mitigation.

161

The following activities are essential for effective disaster planning:
1. Prevention,
2. Preparedness,
3. Mitigation,
4. Response,
5. Recovery, and
6. Communication.
In this chapter, our model incorporates a methodology for addressing activities 2 and 3,
and relating the qualitative capabilities matrix
44
to the mitigation impact.

Hurricane/Flood Model Development
To help in developing a methodology for hurricane/flood planning and disaster mitigation
efforts, we developed an Analytica influence diagram risk and decision analysis model
that utilizes DHS National Planning Scenario 10 as its major input. We used the software
tool, Analytica (see www.lumina.com), to assist in modeling an influence diagram that
represents the interrelationships among variables that include data for wind,
demographics, and the potential for negative consequences of hurricanes and floods as
well as potential mitigations with their impacts and costs in the Miami MMA.

Using this DHS scenario as a template, a similarly-sized Miami MMA was chosen as the
site for the hypothetical hurricane for this study. The influence diagram serves as a tool

44
These refer to the thirty-seven key capabilities as outlined in the DHS FEMA Target
Capabilities List (see http://www.fema.gov/pdf/government/training/tcl.pdf ). These are part of
the U.S. National Preparedness Guidelines. The guidelines outline 1600 tasks that support these
37 capabilities that are essential in disaster preparedness, mitigation, response, and recovery.

162

for communication of high-level findings and examining the impact and value of
individual mitigation decisions. An understanding of how the results are obtained and of
how the various assumptions impact the results is often more important than the specific
input and output numbers. In addition to communicating high-level findings,
stakeholders can examine lower levels of modeling when more detail is desired, aided by
the visual aspects of the model’s structure. As stakeholders are able to understand the
model, debate and discussion can focus more directly upon specific assumptions and lead
to more productive results. Thus, the influence diagram serves as a tool to help to make
the model accessible.

At the highest level, the model displays an entry screen (Figure 38) that provides access
to its submodels (Figures 42 through 46, inclusive). This model-submodel hierarchy of
influence diagrams within Analytica serves as its key organizational paradigm. Because
the visual layout of this influence diagram is intuitive, its users are able to learn about the
model’s structure and organization quickly through its visual cues. Clicking on the six
rectangular icons on the left side of Figure 41 opens the overall model’s submodels. The
blue icons open input models, and the red icon opens the output model.

Referring to Figure 41, the top blue icon is the Non-hurricane Concurrent Events
Submodel (Figure 42). This module enables the analyst to adjust for local estimates of
crime as a percentage of total hurricane damages and of design error or product liability
as a percentage of total hurricane damages.

163

The next blue icon refers to the Hurricane Flood Surge and Wind Damage Submodel
(Figure 43). This submodel adjusts a selected lognormal curve of hurricane severity
based on a Hurricane Severity Index (HSI)
45
approximation of the hurricane. The HSI is
calculated based on a variant of the CHI index
46
that is calibrated by sustained wind
speed at landfall and by the radius from the center of the hurricane to the edge of
hurricane-force winds and normalized to the HSI promulgated for Hurricane Andrew.
The expected damages are calculated based on the population and number of housing
units at risk for the calibrated HSI storm. These numbers are derived from U.S. Census
estimates derived from Landview 6 and MARPLOT software.

45
The Hurricane Severity Index (or HSI) is a hurricane rating system which defines the size and
destructive potential (intensity) of a storm. It was devised in 2005 by Chris Hebert and Bob
Weinzapfel of ImpactWeather.

46
The SHI index used in this model is a hybrid version of the ImpactWeather HSI based on the
Carvill Hurricane Index (CHI), an index which describes the potential for damage from an
Atlantic hurricane and that is used as the basis for trading hurricane futures and options on the
Chicago Mercantile Exchange (CME).

164
Figure 41: Entry Display Screen to the Analytica Model

165

Figure 42: Non-hurricane Concurrent Events’ submodel to the Analytica Model

Design/
material error
Fire/Other
Perils
Non-hurricane
Property
Damage
Arson/Other
Intentional

Figure 43: Hurricane Flood Surge and Wind Damage Submodel to the Analytica Model

Design/ material
error
Expected Economic
Consequences of the
Hurricane
Hurricane Property
Damage
SHI
Approximation
of Hurricane
Wind Damage
Only
Wind Intensity
at Landfall
Radius from
Center of
Hurricane to
Edge of
Hurricane-Force
Winds
Number of
Housing Units
(Input)
Population
(Input)
StandDev

166

Figure 44: Land-Use Characteristics Submodel to the Analytica Model

Breadth -
Prevent Mission
Capabilities
Hurricane
Property
Damage
mitigated with
Land Use
Capabilities

Figure 45: Demographics Submodel to the Analytica Model

Lives
Lives Lost
"Value of
a Life"
Evacuation Time
Lives (Economic
Value)
Life Value

The Land-use Characteristics Submodel (Figure 44) stores the infrastructure and population
counts that were derived from the Landview 6 Population Estimator for the modeled storm. The
Demographic Information Submodel (Figure 45) includes an evacuation time algorithm that
estimates the number of deaths due to an evacuation delay based on evacuation time before
landfall and population count being evacuated.

The next blue icon, the Qualitative Capabilities Matrices Submodel to the Analytica model, is
shown in Figure 46. This module incorporates data from the U.S. DHS National Preparedness

167

Guidelines. As part of the guidelines, the “Target Capabilities List (TCL) defines 37 specific
capabilities that communities, the private sector, and all levels of government should collectively
possess in order to respond effectively to disasters.”
47
Input to this module comes from a
standalone Excel spreadsheet matrix; snapshots of this spreadsheet are shown in Figures 47 and
48. Within this matrix of mitigation capabilities, at the Main Entry Screen shown in Figure 47,
the analyst selects from low, medium, and high impacts with respect to the breadth of the
mitigation’s application and from low, medium, and high costs with respect to these respective
mitigations’ cost to implement. These selections emulate an expert elicitation of a triangular
distribution of the mitigation impact percentages associated with each capability. The completed
spreadsheet matrix is automatically input into the Analytica model via OLE linking.

Within this matrix of mitigation capabilities, the analyst selects Percentage Impacts/Costs that
correlate to low, medium, and high selections with respect to respective mitigations in the screen
shown in Figure 48. The completed matrix is automatically input into the Analytica model via
OLE linking. This allows the analyst to void the impact of mitigations (by setting percentages to
zero) or to subjectively differentiate mitigations (by varying percentages of cost reduction
impact).

All the modules are linked to produce the analysis results as shown in Figure 49.

47
See page 3 of http://www.dhs.gov/xlibrary/assets/National_Preparedness_Guidelines.pdf .

168

Figure 46: Qualitative Capabilities Matrices Submodel to the Analytica Model

Q p
Breadth -
Protect Mission
Capabilities
Breadth -
Respond
Mission
Capabilities
Breadth -
Recover Mission
Capabilities
Breadth -
Common
Mission
Capabilities
Cost - Prevent
Mission
Capabilties
Cost - Protect
Mission
Capabilities
Cost - Respond
Mission
Capabilities
Cost - Recover
Mission
Capabilities
Cost - Common
Mission
Capabilties
Emergency
Planning
Risk
Management
General
Planning and
Building Code
First Responders
Insurance
Schemes
Type - Prevent
Mission
Capabilties
Type - Protect
Mission
Capabilities
Breadth -
Prevent Mission
Capabilities
Type - Respond
Mission
Capabilities
Type - Recover
Mission
Capabilities
Type - Common
Mission
Capabilties

169
Table 12: Entry Screen of the Hurricane/Floods Capabilities and Characteristics Excel Input Spreadsheet

170

Table 13: Entry Screen for Capabilities Percentage Savings or Costs in the Excel Input
Spreadsheet

171
Figure 47: Hurricane Loss Analysis Submodel to the Analytica Model

Background
Lives (Economic
Value)
Expected
Hurricane Loss
($ Millions)
Non-hurricane property
damage
Hurricane Loss
FEMA
Qualitative
Adjustment
(sample -
planning
mitigations)

Generating Capabilities’ Impact Results
The model works as follows using these five main submodeling stages:

1. Based on historic hurricane tracks derived by utilizing HURREVAC2010
software (or judgment), the analyst selects (a) a latitude-longitude describing the
center of the hurricane’s eye at landfall, (b) a storm radius from the eye’s center to
the edge of hurricane force winds, and (c) a maximum sustained wind speed at
landfall. For historic tracks, HURREVAC SHP
48
files are converted to
MARPLOT MIE files, and these MIE files are loaded as layers into MARPLOT

48
Shapefiles are a public domain mapping format developed by Environmental Systems
Research Institute, ESRI
®
. Shapefiles are an output format for ESRI mapping products and can
exchange information with some commercial Geographic Information Systems (GIS). The Shape
to MIE (shp2mie) utility downloaded as part of the LandView installation allows converting
shapefiles to a MARPLOT import format. As such, the output of most GIS systems can be
brought into MARPLOT and related to LandView/MARPLOT topological and population
information.

172
software. Landview’s Population Estimator summarizes U.S. Census data by
census block within the estimated impact zone. (See Figure 48).

2. Incorporating data from Step 1 as inputs and by using Landview 6 and
MARPLOT software, the analyst quantitatively determines the scopes of the
population and infrastructure impacted by the modeled hurricane event. These
determinations are based on creating an impact zone estimated as the circle
described as radiating from the central point of the hurricane’s eye at landfall with
a radius equal to the storm radius described in Step 1. (See Figures 49 and 50).

3. Using a matrix of qualitative hurricane mitigation capabilities to estimate their
impact breadths (benefits) and costs from an Excel spreadsheet outlining these
capabilities, the analyst subjectively selects relative benefits and costs for the
scenario to be generated from Steps 1 and 2. (See Figures 45 and 46).

4. After opening the Analytica model, the FEMA analyst inputs the data from Steps
1 and 2 as well as selects an economic value of a human life and an evacuation
lead time. The data from Step 3 is automatically updated within the model via
OLE links.

5. When Step 4 is run through a modeled scenario, the analyst obtains estimates of
the prospective costs of the modeled hurricane based on no mitigations, on
selected evacuation scenarios, and on selected matrices of mitigation capabilities.

173
From these various estimates, the analyst may be able to better discuss and select
those mitigations that have the most potential for cost and loss of life reductions.

The scenario modeled in this example is similar to an updated replay of the 1926 Miami
storm. The results are shown in Tables 15 and 16. The model returns, depending on
mitigations’ selected, current economic consequences of between approximately $54 and
$144 billion. Other authors have estimated this hurricane’s economic consequences at
between $101
49
and $157
50
billion.

49
See AIR Worldwide’s modeled estimate from http://www.air-
worldwide.com/PublicationsItem.aspx?id=18484

50
See Pielke, Roger A., Jr.; et al. (2008). Normalized Hurricane Damage in the United States:
1900–2005. Natural Hazards Review 9 (1): 29–42.

174
Figure 48: HURREVAC Hurricane Andrew Historical Track Prior to Florida Landfall.

Observations and Implications
Figure 51 shows the range of expected hurricane losses in a repeat of the 1926 Miami
hurricane given the current demographics by various evacuation times and assuming all
37 capabilities advocated by FEMA are applied. This table shows three curves of losses
that range in losses from $50 billion (lower right curve) to more than $153 billion (upper
left curve). The highest losses assume the lowest range of savings due to no disaster
reduction capabilities being implemented, and the lowest losses assume the highest range
of savings due to all forms of disaster reduction capabilities being implemented.

This chapter reveals several interesting observations as follows:

175
1. It is difficult to conduct precise quantitative risk analyses in some important
cases, especially when the problem requires prioritization of activities and
estimation of their impacts across a large range of mission elements.
2. Evacuation time is highlighted as a critical disaster mitigation. A model is
developed that enables analysts to more precisely estimate optimal evacuation
times for approaching hurricanes in near real-time and to adjust expected
disaster losses based on these modified evacuation times.
3. A qualitative analysis based on a low-medium-high scoring elicitations using
expertise by FEMA staff and other emergency managers is introduced to help
estimate ranges of disaster consequences.

176
Figure 49: MARPLOT Hurricane Andrew Historical Track Prior to Florida Landfall
This graphic shows the Hurricane Andrew pre-landfall track/1926 Miami
hurricane proxy overlaid with U.S. Census data within MARPLOT as part of
Landview 6.

This hurricane track was loaded in HURREVAC 2010. At this mapped
tracking point, the hurricane’s estimated landfall was approximately the same
as the 1926 Miami hurricane. For that reason, this shapefile was converted to
an MIE file for layering use within MARPLOT for modeling purposes.

177
Figure 50: MARPLOT Hurricane Andrew Historical Track Prior to Florida Landfall

The shaded area on this graphic represents the selected area wherein
infrastructure and population counts were derived from the Landview 6
Population Estimator.

178
Table 14: Output of the Analytica Model—Qualitative Mitigation Adjustment Factors

179
Table 15: Output of the Analytica Model—Expected High-end Hurricane Losses Matrix

180
Table 16: Output of the Analytica Model—Expected Low-end Hurricane Losses Matrix

181
Figure 51: Output of the Analytica Model—Expected Hurricane Losses by Evacuation Time and Assuming All Mitigations
Are Applied (Upper Curve = Lowest Savings %s and Lower Curve = Highest Savings %s)

182

Chapter 7: Feasible Solutions in Allocating Countermeasures related to
Intentional Disasters on Regional Infrastructure Systems: Preventive
Radiological/Nuclear Detection (PRND) Resource Allocation in
California

Introduction
This example involves a case of a preexisting, regional infrastructure system. Its
potential for disaster vulnerability is robustly considered by the design of a system that
applies countermeasures to mitigate consequences of intentional disasters on this
infrastructure. Chapter 7 describes a method for the optimal allocation of domestic
countermeasures for radiological/nuclear terrorism detection in California. This macro-
method of allocation for a diffuse, extreme risk uses nonlinear optimization to allocate
finite resources over weighted potential targets. Simultaneously considering multiple
levers over multiple targets given multiple potential disasters presents an example of a
more complex scenario subject to a robust risk and decision analysis.

Goals and Objectives
The goal of this chapter is to develop and apply a decision and risk analysis model and
quantitative methodologies to support Preventive Radiological/Nuclear Detection
(PRND) risk management and resource allocation deployment and capability-building
decisions.

The objectives of this chapter are to provide a strategic approach for evaluating and
optimizing the effectiveness of programs to protect the State of California against

183
radiological and nuclear (RN) risks, and also provide the tools, models and
methodologies to assess the relative value of program alternatives.
Project Background
In 2005, the Domestic Nuclear Detection Office (DNDO) was created to protect the
United States against such a radiological and nuclear threat. A key mission for DNDO is
to build state and local capabilities through equipment and training systems. Recently,
individual states have also begun to join this effort. Recognizing the need to align state
preventive radiological and nuclear (RN) resources, California is looking to develop
quantitative methodology for application to the States’ Preventive Radiological/Nuclear
Detection (PRND) program to (1) establish a prioritization for funding; and (2) determine
how these assets will be distributed.

Methodology is needed to give quantitative, analytical guidance to the decision-making
process for building a preventative RN process for the allocation of limited resources. By
applying a strategic decision-making methodology it will allow the state to adopt a
comprehensive risk management strategy.

A systematic approach to intelligently build prevention capabilities can enhance efficient
funding and security decisions. Because the state has to protect against numerous threats
and balances the interests of multiple jurisdictions and disciplines, decisions must be
made in order to maximize finite resources. A model for quantitative analysis was
developed for Critical Infrastructure Protection (CIP) asset protection, which works by
giving numerical values to threats, vulnerabilities and consequences and calculating a risk

184
score. By applying a risk assessment model in a novel manner to PRND assets, a
comprehensive picture of where assets and capabilities will be most effectively deployed
can be obtained. In the PRND arena, the state needs to quantify risks across multiple,
non-congruent categories—pathways (i.e., land borders, highways, and waterways),
targets (i.e., ports, critical infrastructure, high- profile targets, special events, and transit
systems), and interdiction methods (i.e., intelligence-driven missions, passive events, and
fixed or mobile detection systems)—all competing for the same sources of funding.

A tool is needed to cross-compare these risks and to make funding decisions based upon
prioritization. Absent such a methodology, prioritizations are less systematic and,
perhaps, less defensible. For instance, if you ask a port official to rank priorities, you will
get a very different answer than if you ask a highway patrol officer who will more than
likely anticipate a different target or vulnerability. Similarly, if you poll a group of state,
local, and federal subject-matter experts (SME’s) to determine priorities, you will get
varying opinions largely based upon their experience, discipline, or jurisdiction.
Currently, such target “ranking” is done without any empirical data or threat analysis.
Without a common methodology, a comprehensive statewide picture of where risk exists
is difficult or impossible to determine. However, when risk methodology is applied,
priorities begin to emerge.
The CIP risk model integrates multiple factors into a risk matrix to determine the
allocation of resources to reduce those risks. Similarly, a PRND-specific methodology
can apply the key criteria risk score model to:
1. Identify assets and determine which are most critical,

185
2. Identify, characterize and assess RN threats,
3. Access the vulnerability of critical assets to specific RN threats,
4. Determine the risk (i.e. the expected consequences of RN specific types of attacks
on specific assets compared to other types of attacks),
5. Identify ways to reduce those risks,
6. Prioritize risk reduction measures based on a comprehensive PRND strategy,
7. Focus on detection along the transportation system,
8. Identify that state-level detection supplements federal efforts,
9. Account for the sensitivity parameters associated with radiation detection,
10. Measure the impact to commerce (both passenger and commercial),
11. Provide characteristics-based site prioritization (i.e. population, population
density, economic activity, iconic value, transportation systems),
12. Provide site-specific effectiveness of detection systems (e.g., how good is mobile
detection in a choke/constraint point?),
13. Include effect of collaboration (first layer: sensitivity and second layer:
specificity) among different authorities (e.g., lack of collaboration reduces
effectiveness), and
14. Initially, use a sample budget of $5 million.
Risk scores are determined by giving a numerical value to threats, vulnerabilities, and
consequences and by calculating the numerical value into a comprehensive risk score.
Calculating quantitative risk scores based upon threat, vulnerability, and consequence
(and adopting the same formula of calculation for THREAT X VULNERABILITY X

186
CONSEQUENCE = RISK) will allow the state (or any jurisdiction) to merge multiple
disciplines and multiple agencies into a common prioritization standard. This
methodology works when applied to a PRND program because it is supported by a
common statewide framework based upon a comprehensive strategy for protection and
asset allocation when compared across multiple sectors and PRND missions.

Application of this model in a novel way will allow for a quantitative methodology to
support deploying PRND resources and building capability. This not only gives a
strategic approach to protecting the State of California, it also helps to give policy makers
justification in funding PRND programs. By determining a quantitative value to RN
risks, the state can begin to determine how to effectively distribute assets and to
strategically consider how to leverage finite preventive resources for the benefit of all
state and local jurisdictions

Problem Statement
The consequences of RN attacks in the State of California would be catastrophic.
However, defensive resources are finite, and not all potential defensive resource
allocations have the same value. California seeks a comprehensive and quantitative
methodology to aid its PRND program in: (1) establishing prioritization for funding; and
(2) determining how these assets will be distributed. To facilitate the comprehensive
comparison of risk and resource values, the state seeks a methodology that is compatible
with and builds upon the previous quantitative CIP risk analyses methodology, and which

187
uses numerical scores for cross-comparison of attack threats, vulnerabilities, and
consequences.

An Introduction to the Model
Traditionally, the United States has implemented four lines of defense against RN threats.
The first defense is prevention and entails activities such as:
• the strategic deactivation of many existing nuclear warheads,
• the maintenance of extant nuclear warheads under strict military control,
• the in situ security of nuclear weaponry and fissile materials, and
• diplomatic efforts to stop the spread of weapons of mass destruction.
The second defense is deterrence against the capabilities and actions of terrorists and
“rogue states.” The third defense is the interdiction and involves international efforts to
stop the trade of nuclear weapons and radioactive materials at national borders. The
fourth defense, the line of defense dealt with in this report, is crisis and consequence
management and involves efforts at domestic preparedness.
51

Since the terrorist attacks of September 11, 2001, homeland security analysts have
detailed many potential RN terrorist attack scenarios at variable levels of detail, and they
have suggested a number of domestic mitigating responses. The set of RN attack
possibilities and countermeasures is immense, and the effectiveness of these various
countermeasures is difficult to assess. In this chapter, risk reduction benefits of different

51
This is a summarization from Cirincione, Joseph. “Senator Richard Lugar on Threat Reduction
and Defense,” Proliferation Brief, Vol. 4, No. 14, July, 2001.

188
countermeasures and their economic costs are assessed, and a method for an optimal
allocation of selected countermeasures is suggested.

This chapter presents a risk analysis model of the radioactive/nuclear terrorist threats to
California. Its objective is to describe the links among the key elements of different RN
attack scenarios. This model can be used to assess the benefits of threat reduction
measures and to allocate these measures. Prior to this study, it was generally assumed
that the optimum benefit for non-federal government-directed, domestic RN threat
reduction could be obtained by the strategic deployment of radiation detectors at key
transportation nodes and along key transportation links.

“How do we protect cities, major urban areas in this country from a
nuclear or a radiological bomb that was fabricated inside the country?
How do we prevent someone from getting a hold of radioactive material in
the U.S., making a dirty bomb, and then trying to detonate it in … Los
Angeles? …[T]o address that … concern, we are unveiling our Securing
the Cities Initiative, which is a program to see how we can deploy this
kind of detection equipment, not only at seaports and ports of entry by
land, but in cities and around cities, so we could detect a truck coming into
a city with a dirty bomb, even if it didn’t cross an international border.”
52

This probabilistic model is presented as an influence diagram. The probabilistic
description of most events and variables of this model can be represented as the result
either of a much more detailed model or of the encoding of expert opinions.

52
“Remarks by Homeland Security Secretary Michael Chertoff and DNDO Director Vayl
Oxford at a Press Conference to Announce Spectroscopic Portal (ASP) Program Contracts,” U.S.
Department of Homeland Security, Office of the Press Secretary, July 14, 2006.

189
Alternatively, state policy analysts and other experts could rank threats and
countermeasures directly.
53

Overview of the Model Structure and Objectives
Several levels of analysis are incorporated in this model to support decisions to adopt and
allocate countermeasures. First, different potential targets, including infrastructure
systems, symbolic structures, and specific population center targets are input in the
model. Currently, sample targets are used (see Table 17); ultimately, the selection of
these targets will be enhanced by expert elicitation. This could ultimately include a
representation of the effects of interdependencies among the networks and systems that
constitute potential targets. More robust target selection would aid in the identification of
the most effective terrorist countermeasures to increase California’s overall ability to
mitigate RN terrorist attacks.
Table 17. Sample Target Areas with Estimated Population Densities
Los
Angeles
San
Francisco
San
Diego
Sacramento Fresno Long
Beach
San
Jose
Golden
Gate
Bridge
Port
of
LB/LA
Santa
Barbara
Population
density
per square
mile
23,900 17,300 4,200 4,200 4,100 9,800 5,100 10,000 9,200 4,900

Second, the probabilities and consequences of the different attack scenarios are estimated
(see Table 17). The dimensions or attributes of these consequences are measured as
functions of economic losses, including direct and immediate losses but also the

53
Of course, such results might reflect classic biases grounded in knowledge of past events or in
the experts’ professional familiarity with some of the scenarios.

190
secondary effects of primary damage to different parts of the infrastructure including
interruption of economic production or consumption and other threats to California.

Table 18. Sample Formulae for Probabilities and Consequences of Terrorist RN Attacks

Nuclear explosion
54
severity mean = Indirect cost factor x Value of a life x
Target population density x 4.8∏ x (0.397678 x (Amount of fissile
material)
0.33
)
2
(Lognormal distribution)
55

Nuclear explosion annual probability mean = 0.004 (Poisson distribution)

Expected annual nuclear explosion consequences mean = Nuclear
explosion severity x Nuclear explosion annual probability
RDD
56
severity mean = Indirect cost factor x Value of a life x Target
population density x 0.08∏ x (0.397678 x (Amount of fissile
material)
0.33
)
2
(Lognormal distribution)

RDD annual probability mean = 0.08 (Poisson distribution)

Expected annual RDD consequences mean = RDD severity x RDD annual

Third, mitigations to reduce the probabilities and/or consequences of the different attack
scenarios are estimated (see Table 19). The dimensions or attributes of these mitigations
are measured as functions of expected annual consequence means.

54
A terrorist nuclear explosion could result from a stolen nuclear weapon (or “loose nuke”) or an
improvised nuclear weapon (IND). An IND is intended to cause a yield-producing nuclear
explosion. An IND might consist of nuclear weapon components, a modified nuclear weapon, or
an indigenous-designed device.

55
Derived from Glasstone, Samuel and Dolan, Philip J. The Effects of Nuclear Weapons. United
States Department of Defense and United States Department of Energy; 3rd Edition (1977).

56
An RDD is a conventional bomb designed to disperse radioactive material to cause destruction,
contamination, and injury from the radiation produced by the material. An RDD can be almost
any size, defined only by the amount of radioactive material and explosives.

191
Table 19. Estimated Types, Costs, and Impacts of Mitigation Resources

Total annual cost (initial
cost plus labor and
maintenance)
Incremental impact of one
unit of the mitigation (as a
factor of overall loss
reduction)
Maximum impact of the
mitigation (as a factor of
overall loss reduction)
Patrol boat unit $475,000 0.001 0.05
Patrol car unit $200,000 0.0005 0.01
Pan & tilt camera unit $20,000 0.0005 0.01
Security check point $250,000 0.0075 0.05
Fixed detector portal $1,000,000 0.001 0.01
Unmanned aerial systems $250,000 0.002 0.05
Foot patrol with handheld
radiation detector
$125,000 0.0005 0.01
Retrofit patrol car with
radiation detector
$25,000 0.0005 0.01
Internet surveillance $25,000 per target 0.01 0.05
Kill switch radiation
detector devices
(embedded
semiconductors) in vehicle
$250 per vehicle 0.000002 0.25

The overall model is based on the engineering risk analysis method (Apostolakis, 1990,
Modarres et al., 1999).
57
In an influence diagram such as that shown in Figure 52, an
oval represents a random variable and the probabilities assigned to each. A rectangular
node represents a decision and contains the possible options. Each attack scenario is
presented by a combination of nodes, i.e., the combination of a target/target density (e.g.,
a city) and a type and size of weapon (e.g., a radioactive dispersion device (RDD or
“dirty bomb”) or a “loose nuke” or improvised nuclear device (IND/LN)). The arrows
in Figure 52 represent probabilistic dependencies among the various nodes.

57
The Bayesian theory of probability is the framework within which expert opinions are
combined with conditioned observations to produce quantitative measures of the risks.

192
Figure 52. Estimation of RN Damage Incurred by Target

Influence Diagram of Expected Improvised/Loose Nuclear Device
and RDD Terrorism Incidents and their Potential Mitigations
IND/LN rate
Number of nuke
incidents
IND/LN Mean
damage
Rad/Nuc
Damage
incurred
RDD rate
RDD Mean
damage
Number of
RDD incidents
Ability to
Obtain Fissile
Materials
Ability to
Deliver the
Weapon
Ability to
Decipher PAL
Codes
Nuclear
Power Plants
Events
Yield (kTons
released per Ton
of Fissile
Material)
Amount of
Fissile Material
Target
population
density
Value of a Life
Multiplier to
Estimate
Indirect
Economic
Impacts
Targets

Once the potential for damages is estimated, the potential for prevention and/or reduction
of damages is incorporated in the model (see Table 19 and Figure 53).

193
Figure 53. Estimation of Optimal Mitigation Resources by Type and by Target

Coordinates of
targets -
latitudes and
longtidudes
Mitigation
Resource Costs
Mitigation
resource types
Nominal effect
of Mitigation
Allocations on
Vulnerabilities
Maximum effect
of Mitigation
Allocations on
Vulnerabilites
Target
population
density
Average
population
density
Optimal
Mitigation
Resources
Optimal
Solution
Solution Status
Mitigation
Resource
Allocations
Total Resource
Allocations
Mitigation
Savings Factor
Mitigation
Savings
Savings
Total Mitigation
Costs
Sum of MRC
Maximum
Mitigation Costs Grand Sum of
MRC
No. of mit.
resource types

Figure 53 presents the portion of the model that performs an optimization of the
allocation of countermeasures by target and type. Performing this optimization requires
formulating the problem in four parts: a set of decision variables, an objective function,
bounds on the decision variables, and constraints. Since this is a nonlinear problem, a
nonlinear program (NLP), whose objective can be any expression or variable that
depends on the decision variables, is encoded into the model.

194
A NLP is the most general formulation for an optimization.
58
The objective and the
constraints are functions of the decision variables, in this case, whole number variables.
Potential solutions pass through the arrays as parameters to the Analytica Optimizer,
which operates on them directly to find a solution without further interaction with the rest
of the Analytica model. The objective function is defined as the Analytica variable
dependent upon the decision variables. The Optimizer repeatedly evaluates the objective
function as it assigns different values to the decision variables in its search for a solution.
It does the same with expressions passed to left-hand side of the constraints.

An Analytica model of PRND resource allocation has been developed that overviews
California’s risk at a high level. The model assumes that various PRND devices can be
deployed. The model uses new patrol boat units with radiation detectors, new patrol car
units with radiation detectors, pan & tilt cameras, security check points with hand held
detectors, fixed radiation detector portals, unmanned aerial systems, foot patrol with
handheld radiation detectors, patrol cars retrofitted with radiation detectors, enhanced
internet surveillance, and a test program to install kill switch detector devices on vehicles
as potential PRND devices. Table 20 displays the results of the Optimizer. Note that
this may represent a local minimum over the search space rather than an optimum.

58
The Analytica Optimizer uses the Premium Solver Platform licensed from Frontline Systems,
Inc. The NLP methods offer hybrid methods using classical gradient-search and evolutionary
(genetic) algorithms for smooth and discontinuous objective functions.

195
Modeling scenarios of RN terrorist attacks on California
This influence diagram model is based on risk analysis (Apostolakis, 1990) and decision
analysis (Raiffa, 1968; Keeney and Raiffa, 1976). Its objective is to quantify the
elements of California’s infrastructure that are potential RN terrorist attack targets and to
identify the type and number of countermeasures to enable the most effective means of
reducing the overall threat to these targets given a budget.
59
Ultimately, such a
countermeasure capability-based, target-based approach to set countermeasures’ priorities
can help ensure the effectiveness of the efforts involved in California.

The first step in the analysis is to combine probabilistic modeling of the actions of
terrorists with the consequences for the California targets. This part of the analysis
permits identification of attack scenarios, assessment of the likelihood of occurrence of
these scenarios, and prioritization of these attack scenarios based both on their blended
likelihood and expected damage to California if they occur.

59
Other important factors such as the use of intelligence and the potential redundancy of
countermeasures already implemented by the federal government are not considered in this
model.

196
Figure 54: Estimated RN Terrorist Attack Consequences (IND/LN/RDD mixed by selected frequencies) by Target Assuming a
5 kilograms of Radioactive Material, a Value of Life of $5 million, and an Indirect- to Direct-Consequences Factor
of 3.0

197

Risk analysis is appropriate in this case since it allows for the combination of data about
different aspects of the problem from different sources of information. For example,
Figure 54 represents the aggregated results of a Monte Carlo sampling of 10,000
iterations of the selected attack dynamics. Note that this figure is based on product of the
mean annual frequencies and the mean annual severities. For example, this incorporates
an improvised nuclear weapon in downtown Los Angeles with a 5 kilogram HEU
payload (i.e., a “suitcase” bomb) that would cause approximately $2.5 trillion in total
economic damages as well as a “dirty bomb” at the Ports of Long Beach/Los Angeles
that would cause about $20 billion in total economic damages, among other possibilities.

198

Figure 55: Estimated Savings by Allocation of Mitigation Resources Types by Target Assuming a $5 million Budget, 5
kilograms of Radioactive Material, a Value of Life of $5 million, and an Indirect- to Direct-Consequences Factor
of 3.0 (assumes mixed IND/LN/RDD scenarios)

Los Angeles
San Francisco
San Diego
Sacramento Fresno
Long Beach
San Jose
Golden Gate Bridge
Port of LB/LA Santa Barbara
0
100M
20M
40M
60M
80M
120M
Targets

Mitigation resource types
Patrol boat unit
Patrol car unit
Pan & tilt camera
Security check point
Fixed detector portal
UAS
Foot patrol with handheld
Retrofit patrol car with detector
Internet surveillance
Install kill switch detector device on vehicles

199

Figure 55 overlays the Figure 54 sampling with the expected savings as assumed in
Table 19 resultant
60
from various mitigations. In this simple illustration, the various
mitigations as allocated within Table 20 return differing expected savings dependent
upon the target. The Table 20 allocations are a result of the nonlinear optimization
performed within the Analytica model. This allocation of mitigation resources generates
an expected annual savings (reduction in disutility or positive impact to the State of
California) of $931.5 million for a $5 million
61
annual investment in mitigation
resources. Thus, a net annual savings of $926.5 million is realized for the modeled
scenarios. In such a manner, the disutilities (negative impacts) to California of various
successful attack scenarios are assessed, and these assessments can be used to compute
and measure the expected disutility of an attack in the absence of additional
countermeasures or assuming the budgeted, allocated countermeasures as displayed in
Table 20. The countermeasures can affect either the probability of an attack (e.g., by
decreasing the perceived or real probability of success from the terrorists’ view point) or
the consequences of an attack, therefore the disutility of the consequences to California.
The countermeasures that yield the largest decrease in disutility to California can then be
identified based both on costs and benefits.

60
These estimates for incremental and maximum savings (as a result, mainly, of reduction of the
expected probability of occurrence) could be improved by the use of expert elicitation.

61
This amount was a provided budget constraint. Given the expected savings of this investment,
it might make sense to invest more in these mitigation resources.

200
Table 20: Allocation of Mitigation Resources by Target Assuming a $5 million Budget,
5 kilograms of Radioactive Material, a Value of Life of $5 million, and an
Indirect- to Direct-Consequences Factor of 3.0—Output of the Analytica
Model

Locale/ Counter-
measure
Los
Angeles
San
Francisco
San
Diego
Sacramento Fresno Long
Beach
San
Jose
Golden
Gate
Bridge
Port of
LB/LA
Santa
Barbara
Patrol boats 0 0 0 0 0 0 0 0 0 0

Patrol cars 0 0 0 0 0 0 0 0 0 0

Cameras 0 0 0 0 0 0 0 0 0 0

Check points 0 0 0 0 0 0 0 0 0 0

Fixed detector
portal
0 0 0 0 0 0 1 1 0 1

Unmanned aerial
systems (UAS)
1 1 0 0 0 0 0 0 0 0

Foot patrol with
handheld
0 1 0 0 1 1 1 2 1 1

Detector in patrol
car
5 5 0 0 0 0 0 0 0 0

Internet
surveillance
1 1 1 1 1 2 2 2 2 2

Vehicle kill switch
detector
1728 749 9 9 9 9 9 9 9 9

Observations and Implications
In designing and implementing strategies of countermeasures to potential terrorist
attacks, a systematic analysis that permits identification and analysis of a large spectrum
of possible scenarios is essential. Such a quantitative approach permits a comparison of
the net effects of different threats (both in terms of probabilities and consequences) and
the impact of and overall allocation of countermeasures. In the model that was developed
as part of this chapter, the user/analyst can:
• define her or his own sets of targets (with corresponding population densities),
• attack severities (by adjusting the amount of fissile materials),

201
• types of mitigations (e.g., these can be anything from hand-held radiation
detectors to satellite surveillance and these must be tied to a mitigation per unit
cost),
• overall annual mitigation budget, and
• per unit and overall mitigation effectiveness factors (i.e., estimates of the factor
reduction each unit of mitigation reduces the probability of an attack as well as
estimates of the limits of overall effectiveness for each type of mitigation
resource).
After adjusting any or all of these variables, the model will find a feasible allocation of
whole number units of each mitigation by target that conforms to a budget.

The policy implications with respect to the RN problem analyzed are as follows:
• Relatively small countermeasure investments can yield very large benefits
whenever conditional expected losses are catastrophically large. When
countermeasures can, for example, cut an already small frequency in half, a
corresponding reduction in the expected losses occurs.
• Countermeasures that may be useful for one mitigation strategy may be less
useful for another. This change in utility can be a result of budgetary constraints
(for example, you may not be able to afford enough of an otherwise good
countermeasure if your budget is limited) or because of tactical and/or
engineering concerns. For example, use of fixed radiation detector portals as a
primary strategy was rejected within this model for two reasons in addition to

202
budgetary constraints: First, current technological limitations
62
related to
variations of time, distance, and shielding with respect to the detection of fissile
materials severely limit the feasibility of reliable (and relatively quick) detection.
Second, even if reliable, fast detection is possible, the ability to identify
accurately and intercept successfully specific source terrorist suspects, especially
in moving vehicles (and boats and aircraft), might be limited in near real-time. Of
course, with larger budgets and/or new/unknown technology, fixed portals might
be a more feasible option. Recognizing this, this model incorporates a variety of
potential mitigations that depend on more passive, less expensive, and more
numerous methods of detection. However, this type of model has the flexibility
to change as budgets and technologies evolve.
• Probably the most sensitive variables and the weakest links in this analysis are the
frequency estimates of an attack (see Table 18) and the effectiveness assessments
of the countermeasures (see Table 19).

Hopefully, this model will contribute to more rational allocations of countermeasures
against RN terrorist attacks.

62
See, also GAO, “Combating Nuclear Smuggling: DHS’s Decision to Procure and Deploy the
Next Generation of Radiation Detection Equipment Is Not Supported by Its Cost Benefit
Analysis,” Statement of Gene Aloise, Director Natural Resources and Environment. March 17,
2007. GAO-07-581T.

203
Chapter 8: Overall Observations and Implications

Each of the models of risk and decision analysis presented in the preceding six chapters
was concerned with behaviors (decisions or “objectives”) and stimuli (initial “risks and
uncertainties” and their associated sets of choices/utilities/“levers”) that were chronicled
and calibrated by “metrics.” Each of these cases demonstrated that, within this general
framework, different methods for disaster risk and decision analyses can be
accommodated and return useful results. Despite differences in methods, each disaster
risk and decision analysis started with this simple framework, and each case’s model
communicates certain, essential ideas as follows:
1. It outlines the essential elements and relationships in the specific case examined.
2. It establishes the data that describes and measures the specific case examined.
3. Its model is based on and applied to an actual disasters or plausible disaster
scenario from which general and specific policy implications may be drawn.
4. Robust models must act as templates to better understand and help prevent
recurrences of past disasters, to react and respond to and recover from unfolding
disasters, and to plan for and mitigate future disasters.
5. Robust models must serve as tools to communicate the risks and potential
consequences of disasters and the actions necessary to mitigate these risks and
consequences.

The negative consequences that define disasters emerge from unique, complex
combinations of circumstances. The decisions that impact (either prevent, ameliorate, or

204
worsen) the initiating disaster event (and the set of posterior events from which its
consequences emerge) evolve through tools that iteratively measure and recalibrate like
consequence-minimizing levers. This dominant theme of minimizing negative
consequences allows for unique approaches and solutions to the potential mitigation of
disasters, approaches and solutions that vary depending on whether the disaster is
intentional or unintentional and involves varying perspectives of location, scope, and
timing. As such, each of the six cases examined adheres to a simple objective-stimuli-
lever model that is monitored by metrics and that maintains the five criteria outlined in
the preceding bullets.

From a policy perspective, Chapter 2 demonstrates that, if attack frequency cannot be
determined, control of potential severity becomes a more tangible goal. It also shows that
looking at both unconditional and conditional expected losses can lead to insights such as
follows:
1. Siting of LNG receiving facilities should not be near other critical infrastructure
or population centers,
2. Siting of LNG facilities should be in areas with low densities at least four miles
from population centers and at least two miles from major shipping channels to
minimize the exposure of people and property,
3. Full containment of facility tanks should be mandatory,
4. The separation of the LNG receiving facility from LNG ship dock facilities by use
of a receiving jetty with mooring and berthing facilities should be considered, and
5. Prevailing winds and precipitation patterns by season should be considered.

205

Chapter 3 examines a planned facility and considers alternative options for siting and
protecting an old, existing LPG facility with limited options for risk reduction. This
chapter highlights that:
1. There is a significant difference in the risk management of planned versus
existing hazardous facilities,
2. It is a bad idea to exempt facilities from regulations without a sunset clause,
3. Similar to the arguments in Chapter 2, alternative siting, hardening, and additional
security should be considered, and
4. The persistency of existing, dangerous facilities may also be due to their
significant sunk costs.

Chapter 4 is an analysis whose message is that, to avoid repeating the mistakes of the
past, we need to be as realistic as possible in our assessment of probabilities of future
extreme events and their consequences. With respect to the New Orleans’ flood control
system, the analysis’ preliminary results included the following:
1. Increasing the floodwall and levee heights by 10 feet,
2. Continuing to provide a Mississippi River cut-off option, and
3. Increasing mandatory hurricane evacuation periods to at least 48 hours and, on a
prescribed basis, for up to 60 hours.

206
In addition, other options considered include:
1. Assigning flood plains that are enclosed by levees and floodwalls within
prescribed bounds the status of “designated flood plain,”
2. Reworking flood plain maps on a regular basis to more accurately reflect
elevation changes due to natural and man-made impacts,
3. Addressing floodwall and levee improvement in creative ways,
4. Continuing levee and floodwall monitoring and maintenance, and
5. Considering the probable avulsion of the Mississippi River in considering the
refortification and rebuilding of New Orleans.

Chapter 5 presented a risk and decision analysis model that dynamically applied
countermeasures to mitigate consequences of intentional disasters on an existing, regional
infrastructure system, the Ports of Long Beach/Los Angeles. When considering a risk
and decision analysis as part of a large-scale, “system of systems,” two major points must
be considered:
1. The need for expert elicitation must not be underestimated. For example, PortSec
requires expert knowledge in areas such as:
a. Wait times associated with countermeasures,
b. Effectiveness metrics associated with the countermeasures,
c. Measures of the maximum units of each type of countermeasure and
overall,
d. Target selection,
e. Attack mode selection,

207
f. Expected delays, and
g. Seasonality and weather adjustments.
2. From a broader perspective, efforts such as PortSec potentially offer many
benefits and challenges as follows:
a. PortSec is potentially portable, and the idea of overseeing allocation and
management of security resources within other defined domains has
desirable potential and
b. Large-scale projects such as PortSec should benefit from a phased
development approach.

Chapter 6 presents a model for estimating hurricane evacuation times by focusing on the
range of expected hurricane losses in a repeat of the 1926 Miami hurricane given the
current demographics by various evacuation times and assuming all 37 capabilities
advocated by FEMA are applied. This chapter reveals several interesting observations as
follows:
1. It is difficult to conduct precise quantitative risk analyses in some cases,
2. Evacuation time is a critical disaster mitigation, and
3. Qualitative analysis based on expert elicitation can help estimate ranges of
disaster consequences.

Chapter 7 demonstrates a systematic radioactive/nuclear risk and decision analysis that
permits identification and analysis of a large spectrum of possible scenarios involving
“looses” nuclear weapons and “dirty” bombs. This analysis permits a comparison of the

208
net effects of different threats (both in terms of probabilities and consequences) and the
impacts of and overall allocations of countermeasures. In the model that was developed
as part of this chapter, the user/analyst can:
1. Define her or his own sets of targets (with corresponding population densities),
2. Attack severities (by adjusting the amount of fissile materials),
3. Types of mitigations (e.g., these can be anything from hand-held radiation
detectors to satellite surveillance and these must be tied to a mitigation per unit
cost),
4. Overall annual mitigation budget, and
5. Per unit and overall mitigation effectiveness factors (i.e., estimates of the factor
reduction each unit of mitigation reduces the probability of an attack as well as
estimates of the limits of overall effectiveness for each type of mitigation
resource).
After adjusting any or all of these variables, the model finds feasible allocations of whole
number units of each mitigation by target and which conform to a budget.

With respect to the radioactive/nuclear devise terrorism problem analyzed, the policy
implications are as follows:
1. Relatively small countermeasure investments can yield very large benefits
whenever conditional expected losses are catastrophically large,
2. Countermeasures that may be useful for mitigation strategy may be less useful for
another, and

209
3. The most sensitive variables and the weakest links in this type of analysis may be
the frequency estimates of attacks and the effectiveness assessments of the
countermeasures.

Lindblom asserts that there are two primary approaches to decision-making, the rational
comprehensive method (root) and the successive limited comparison method (branch)
(Lindblom, 1959). Although policymakers tend to use the branch method, the literature
tends to focus on the root method. However, the root method is more difficult because
human comprehension (of all possible alternatives) is bounded, evaluation (of all possible
factors) is bounded, and there is always the potential for the economic and/or political
infeasibility of an option. As such, the branch method generally allows for a more
feasible approach to solving complex problems. Successive limited comparison—or
“muddling through”—builds on existing paradigms, tweaking them here and there in a
continuous, evolutionary process.

In making decisions about the risk of disasters, consideration of the potential range of
consequences involving the related, iterative, and incremental processes of (1) fulfilling
the objective of complete disaster risk management, (2) manipulating the levers that alter
the attainment of this objective, and (3) monitoring and measuring these dynamic
processes requires much more than “muddling through.” Properly recognized and
exploited, robust models can be used to explore ranges of potential future disasters that
potentially can empower us to use our limited mitigation resources more efficiently, to

210
better plan and prepare for future disasters, and to more efficiently allocate our disaster
risk management efforts.

211
Bibliography

(ABSG, 2004) ABSG Consulting Inc. Consequence Assessment Methods for Incidents
Involving Releases from Liquefied Natural Gas Carriers. Federal Energy
Regulatory Commission contract number FERC04C40196, May 13, 2004.
(Ahrens, 2003) Ahrens D. 2003. Meteorology today. Pacific Grove, CA: Brooks/Cole-
Thomas Learning.
(Apostolakis, 1990) Apostolakis, G. 1990. “The Concept of Probability in Safety
Assessments of Technological Systems,” Science, Vol. 250, 7 December, 1990.
(Aristotle, 2005) Aristotle, Rhetoric, Translated by W. Rhys Roberts, Digireads.com,
Stiwell, Kansas, 2005.
(Barry, 1997) Barry, John M. Rising Tide: The Great Mississippi Flood of 1927 and
How It Changed America. New York: Touchstone Books, 1997.
(Beatley et al., 2002) Beatley, Timothy, Brower, David J. and Schwab, Anna K. (2002)
An Introduction to Coastal Zone Management (Second Edition). Washington:
Island Press.
(Bell and Dallas, 2007) Bell, William C. and Dallas, Cham E. 2007. “Vulnerability of
populations and the urban health care systems to nuclear weapon attack –
examples from four American cities,” International Journal of Health Geographics
2007, 6:5 doi:10.1186/1476-072X-6-5. http://www.ij-
healthgeographics.com/content/6/1/5.
(Bister and Emanuel, 1997) Bister, Marja and Kerry Emanuel. “Dissipative Heating
and Hurricane Intensity,” Meteorology and Atmospheric Physics, Volume 62,
Numbers 3-4, Vienna: Springer Wien, September 1997, p. 223-240.
(Bister and Emanuel) Bister, Marja and Kerry Emanuel, “Potential Intensity Estimates
for Individual Cities and Regions,” Hurricane Climatological Potential Intensity
Maps and Tables, Program in Atmospheres, Oceans, and Climate, Massachusetts
Institute of Technology. [ from http://wind.mit.edu/~emanuel/pcmin/climo.html ]
(Blake et al., 2005) Blake ES, Rappaport EN, Jarrell JG, and Landsea CW. 2005. “The
deadliest, costliest, and most intense United States tropical cyclones from 1851 to
2004 (and other frequently requested hurricane facts),” NOAA Technical
Memorandum NWS TPC-4; August.

212
(Blake et al., 2006) Blake, Eric S., Jerry D. Jarrell, Edward N. Rappaport, and
Christopher W. Landsea. The Deadliest, Costliest, and Most Intense United
States Tropical Cyclones From 1851 to 2005, NOAA Technical Memorandum
NWS TPC-4, The Deadliest, Costliest, and Most Intense United State Tropical
Cyclones from 1851 to 2005 (and Other Frequently Requested Hurricane Facts).
Miami, Florida: NOAA, Updated July 2006. [ From
http://www.nhc.noaa.gov/Deadliest_Costliest.shtml ]
(BRAG, 2004) Bear River Association Governments [BRAG]. “Part IV: Risk
Assessment”. 22 Jan 2004. (p. 32)
http://www.brag.utah.gov/pdf/PDM%20Hazard%20Planning/Part%20IV%20Intro.pdf
(Broad. 2005) Broad, William J. (2005, September 9) “In Europe, High-Tech Flood
Control, with Nature’s Help,” The New York Times.
(Brown and Graham, 1988) Brown, Curtis A., and Wayne J. Graham, “Assessing the
Threat to Life from Dam Failure,” Water Resources Bulletin, Volume 24, No. 6,
page 1303-1309, December 1988.
(Bulletin 17-B) Bulletin 17-B: U.S. Interagency Advisory Committee on Water Data,
1982, Guidelines for determining flood flow frequency, Bulletin 17-B of the
Hydrology Subcommittee: Reston, Virginia, U.S. Geological Survey, Office of
Water Data Coordination, [ Available from National Technical Information
Service, Springfield VA 22161 as report no. PB 86 157 278 or from
http://www.fema.gov/mit/tsd/dl_flow.htm ].
(CEC, 2003) California Energy Commission. “Liquefied Natural Gas in California;
History, Risk and Siting.” Staff White Paper 700-03-005. July 2003.
(CEC, 2005(1)) California Energy Commission, B.B. Blevins, Executive Director .
Safety Advisory Report on the Proposed Sound Energy Solutions Liquefied
Natural Gas Terminal at the Port of Long Beach, California. State of California
Energy Commission. Publication # CEC-600-2005-033, September 7, 2005.
http://www.energy.ca.gov/lng/documents/.
(CEC, 2005(2)) California Energy Commission, B.B. Blevins, Executive Director .
International and National Efforts to Address the Safety and Security Risks of
Importing Liquefied Natural Gas: A Compendium, Consultant Report. Publication
# CEC-600-2005-002. March 23, 2005.
http://www.energy.ca.gov/lng/documents/.
(Chatterjee at al., 2011) Chatterjee, Samrat, Carl Southwell, Isaac Maya, and Daniel Salazar,
“Development of Risk-Based Preventative Radiological/Nuclear Detection Resource
Allocation Decisions for State of California,” An unpublished report for the California
Emergency Management Agency, February 7, 2011.

213
(Cirincione, 2001) Cirincione, Joseph. 2001. “Senator Richard Lugar on Threat
Reduction and Defense,” Proliferation Brief, Vol. 4, No. 14, July, 2001.
(Clarke, 2005) Clarke, Richard A. LNG Facilities in Urban Areas: A Security Risk
Management Analysis for Attorney General Patrick Lynch Rhode Island. May
2005.
(Clemen, 1997) Clemen, Robert T. Making Hard Decisions: An Introduction to
Decision Analysis. Pacific Grove, California: Duxbury Press, 1997.
(Coastal Ocean Policy Roundtable, 1992) Coastal Ocean Policy Roundtable. (1992) The 1992
Coastal Status Report: A Pilot Study of the U.S. Coastal Zone and Its Resources.
Newark, DE: Coastal Ocean Policy Roundtable.
(Coleman, 1988) Coleman, J.M. (1988) “Dynamic changes and processes in the Mississippi River
Delta,” Geological Society of America Bulletin. v. 100, p. 999-1015.
(Colton, 2005) Colten, Craig E. (2005) Unnatural Metropolis. LSU Press.
(Committee, 1985) Committee on Safety Criteria for Dams, Water Science and
Technology Board (Committee), Commission on Engineering and Technical
Systems, National Research Counci. Safety of Dams: Flood and Earthquake
Criteria. Washington, DC: National Academy Press, 1985.
(Committee, 2006) Committee on the Restoration and Protection of Coastal Louisiana
(Committee), Ocean Studies Board, Division on Earth and Life Studies. (2006) Drawing
Louisiana’ s New Map: Addressing Land Loss in Coastal Louisiana. W ashington, D.C.:
The National Academies Press.
(Considine et al., 1982) Considine, M., G.C. Grant and P.L.Holden. “Bulk: Storage of
LPG - Factors Affecting Offsite Risk,” Institution of Chemical Engineers
Symposium Series No. 71, pp. 291-320.
(Cutter and Scott, 1997) Cutter SL, Mitchell JT, Scott MS. 1997. Handbook for
conducting GIS-based hazard analysis at the county level.
http://www.cla.sc.edu/geog/hrl; November.
(De Souza, 2004) De Souza, RM. 2004. In harm’s way: hurricanes, population trends,
and environmental change. Population Reference Bureau.
http://www.prb.org/Articles/2004/InHarmsWayHurricanesPopulationTrendsandE
nvironmentalChange.aspx.
(Dilley et al., 2005) Dilley M, et al. 2005. Natural disaster hotspots: a global risk
analysis. Washington DC: The World Bank and Columbia University.
http://publications.worldbank.org/ecommerce/catalog/product?item_id=4302005.

214
(DOE, 2005) “Liquefied Natural Gas: Understanding the Basic Facts,” DOE/FE -0489.
Washington, DC: U.S. Department of Energy, August 2005.
http://www.fe.doe.gov/programs/oilgas/publications/lng/LNG_primerupd.pdf .
(Fay, 2003(1)) Fay, J.A. “Model of spills and fires from LNG and oil tankers,” Journal of
Hazardous Materials.;96(2-3):171-88, 2003 Jan 31.
(Fay, 2003(2)) Fay, J.A. “Spills and Fires from LNG and Oil Tankers in Boston
Harbor,” Cambridge, MA 02139, March 26, 2003.
(FERC, 2005(1)) Federal Energy Regulatory Commission. KeySpan LNG Facility
Upgrade, Algonquin Gas Transmission, L.L.C. Docket Nos. CP04-223-000,
CP04-293-000, andCP04-358-000,
http://www.ferc.gov/industries/lng/enviro/eis/05-20-05-eis.asp#skipnavsub
(FERC, 2005(2)) Federal Energy Regulatory Commission. Staff of FERC and Port of
Long Beach CA Issue Joint Draft Environmental Document on Long Beach LNG
Proposed Project Docket No. CP04-58-000, et al.,
http://www.ferc.gov/industries/lng/enviro/eis/10-07-05-eis.asp#skipnavsub
(FERC, 2006) Federal Energy Regulatory Commission. Distrigas (now Tractebel)
Boston LNG documents. http://elibrary.ferc.gov/idmws/search/advResults.asp
(Search criteria “Everett, MA”)
(Frick, 2005) Frick, Suzanne. City of Long Beach response to Draft Environmental
Impact Statement/Environmental Impact Report, Long Beach LNG Import
Project, FERC Docket No. CP04-58-000, et al, POLB Application No. HDP 03-
079, December 8, 2005.
(Fuller et al., 1995) Fuller, D.A., J.G . Gosselink, J. Barras, and C.E. Sasser. (1995) Physical
Setting. pp. 9-23. In: D.J. Reed (ed.) Current Status and Historical Trends of Hydrological
Modification, Reduction in Sediment Availability, and Habitat Loss/Modification in the
Barataria-T errebonne Estuarine System. BTNEP No. 20. Barataria T errebonne National
Estuary Program, Thibodaux, LA.
(GAO, 2007) GAO, “Combating Nuclear Smuggling: DHS’s Decision to Procure and
Deploy the Next Generation of Radiation Detection Equipment Is Not Supported
by Its Cost Benefit Analysis,” Statement of Gene Aloise, Director Natural
Resources and Environment. March 17, 2007. GAO-07-581T.
(Glasser and Grunwald, 2005) Glasser, Susan B. and Michael Grunwald, “The Steady
Buildup to a City’s Chaos: Confusion Reigned At Every Level Of Government,”
The Washington Post, Sunday, September 11, 2005, [ From
http://www.washingtonpost.com/wp-
dyn/content/article/2005/09/10/AR2005091001529_pf.html ]

215
(Glasstone and Dolan, 1977) Glasstone, Samuel and Dolan, Philip J. 1977. The Effects
of Nuclear Weapons. United States Department of Defense and United States
Department of Energy; 3rd Edition.
(Guidelines, 2003) Guidelines on Participatory Planning and Management for Flood
Mitigation and Preparedness (Guidelines). (2003) United Nations Publications.
(Hicking, 2002) Hicking, Alan. “Gambling with Frozen Fire?” Rational Analysis for a
Problematic World Revisited: Problem Structuring Methods for Complexity,
Uncertainty and Conflict. Jonathan Rosenhead and John Mingers, editors. West
Sussex, UK: John Wiley & Sons, Ltd., 2002.
(Hightower, 2004) Hightower, Mike, Louis Gritzo, Anay Luketa-Hanlin, John Covan,
Sheldon Tieszen, Gerry Wellman, Mike Irwin, Mike Kaneshige, Brian Melof,
Charles Morrow, and Don Ragland. Guidance on Risk Analysis and Safety
Implications of a Large Liquefied Natural Gas (LNG) Spill Over Water. Sandia
Report SAND2004-6258, December 2004.
(Hommels, 2005) Hommels, Anique. Unbuilding Cities: Obduracy in Urban
Sociotechnical Change. MIT Press, 2005.
(Hoppe, 1993) Hoppe, Rob and Aat Peterse. Handling Frozen Fire: Political Culture
and Risk Management .Westview Press, 1993.
(IPET, 2006) Interagency Performance Evaluation Task Force (IPET), Performance
Evaluation of the New Orleans and Southeast Louisiana Hurricane Protection
System—Draft Final Report, United States Army Corps of Engineers (USACE),
June 1, 2006 [ see also https://ipet.wes.army.mil/ ].
(International Commission, 2009) International Commission on Nuclear Non-
proliferation and Disarmament Report, “Eliminating Nuclear Threats: A Practical
Agenda for Global Policymakers,” Canberra/Tokyo. First published November
2009. Reprinted December 2009, http://www.icnnd.org/reference/reports/ent/part-
ii-4.html .
(Jarrell and Mayfield, 1992) Jarrell JD, Hebert PJ and Mayfield M. 1992. Hurricane
experience levels of coastal county populations from Texas to Maine. NOAA
Technical Memorandum NWS NHC-46; 1992 August.
(Jefferson, 1809) Jefferson, Thomas, (1809) “Letter to the Republican Citizens of
Washington County, Maryland, March 3, 1809,” The Writings of Thomas
Jefferson. (1871) H.A. Washington (ed.) Volume 8. New York: G.P. Putnam.

216
(Keeney and Raiffa, 1976) Keeney, R.L. and H. Raiffa. 1976. Decision Analysis with
Multiple Objectives: Preferences and Value Trade-Offs. John Wiley and Sons,
New York.
(Kelman, 2003) Kelman, Ari. (2003) A River and Its City: The Nature of Landscape in
New Orleans. University of California Press. Berkeley, CA.
(Keynes, 1936) Keynes, John Maynard, (1936) “General Theory of Employment, Interest and
Money ,” Collected W ritings of John Maynard Keynes, V ol. 7—The General Theory (1973),
Moggridge, Donald (ed.), London: Macmillan for the Royal Economic Society.
(Klapp, 1992) Klapp, Merrie G. Bargaining with Uncertainty: Decision-Making in
Public Health, Technological Safety, and Environmental Quality. Auburn House,
1992.
(Kundzewicz, 2002) Kundzewicz, Z. (2002) Risk, Reliability, Uncertainty, and
Robustness of Water Resource Systems. Cambridge University Press.
(Kupperman and Smith, 1975) Kupperman, Robert H., Wilcox, Richard H., Smith,
Harvey A., (1975, Feb. 7) “Crisis Management: Some Opportunities,” Science.
New Series, V ol. 187, No. 4175., pp. 404-410.
(Landsea et. al, 2003) Landsea, Christopher W., Craig Anderson, Noel Charles, Gilbert
Clark, Jason Dunion, Jose Fernandez-Partagas, Paul Hungerford, Charlie
Neumann, and Mark Zimmer. The Atlantic Hurricane Database Re-analysis
Project Documentation for 1851-1910 Alterations and Addition to the HURDAT
Database. 2003. [ from
http://www.aoml.noaa.gov/hrd/hurdat/Documentation.html ].
(Laska, 2004) Laska, S. “What if Hurricane Ivan Had Not Missed New Orleans?”
Natural Hazards Observer, Vol. XXIX, November, 2004.
(Lempert et al., 2003) Lempert, Robert J., Steven W. Popper, and Steven C. Bankes.
Shaping the Next One Hundred Years: New Methods for Quantitative, Long-Term
Policy Analysis. Santa Monica, CA: RAND, 2003.
(Levin, 1993) Levin, D.R. (1993) “Tidal inlet evolution in the Mississippi River
delta plain,” Journal of Coastal Research 9.2:462-480.
(Linnerooth, 1983) Linnerooth, J. “Risk Analysis in the Policy Process,” in: Kunreuther
H. C. et al. (eds.), Risk Analysis and Decision Processes: The Siting of Liquefied
Energy Gas Facilities in 4 Countries, Springer, Berlin: 178-208, 1983.

217
(Liu, 2004) Liu, K.-b. (2004) “Paleotempestology: Principles, Methods, and Examples
from Gulf Coast Lake Sediments,” in R. J. Murnane and K.-b. Liu (Eds.),
Hurricanes and Typhoons: Past, Present, and Future, pp. 13-57. New York:
Columbia University Press.
(Lindblom, 1959) Lindblom, Charles E. “The Science of "Muddling Through,"” :
Public Administration Review, Vol. 19, No. 2 (Spring, 1959).
(Loomis, 1967) Loomis, Charles P ., (1967, Dec.) “In Praise of Conflict and Its Resolution,”
American Sociological Review. V ol. 32, No. 6., pp. 875-890.
(Louisiana Authority , 1998) Louisiana Coastal Wetlands Conservation and
Restoration Task Force and the Wetlands Conservation and Restoration
Authority (Louisiana Authority). (1998) Coast 2050: Toward a
Sustainable Coastal Louisiana. Louisiana Department of Natural
Resources. Baton Rouge, La. 161 p.
(Maestri, 2002) Maestri, W. Interview with American RadioWorks, 2002. [from
http://americanradioworks.publicradio.org/features/wetlands/hurricane2.html ].
(Maralek, 2000) Maralek, Jiri. (2000) Flood Issues in Contemporary Water Management.
Springer.
(Marks, 1984) Marks G., and von Winterfeldt, D. “Not in my back yard: Influence of
motivational concerns on judgments about a risky technology,” Journal of
Applied Psychology, 3, 408-415, 1984.
(May, 1986) May , Peter J., (1986) “Politics and Policy Analysis,” Political Science Quarterly.
V ol. 101, No. 1, pp. 109-125.
(McBride et al., 1992) McBride, R.A., S. Penland, M.W. Hiland, S.J. Williams,
K.A. Westphal, B.E. Jaffe, and A.H. Sallenger, Jr. (1992) "Analysis of
barrier shoreline change in Louisiana from 1853 to 1989." pp. 36-97 In:
S.J. Williams, S. Penland, and A.H. Sallenger (eds.) Louisiana Barrier
Island Erosion Study, Atlas of Shoreline Changes in Louisiana from 1853
to 1989. U.S. Geological Survey Miscellaneous Investigations Series I-
2150-A.
(McQuaid, 2005) McQuaid, John. “Anatomy of a Disaster,” Newhouse News Service,
September 8, 2005. [ From
http://www.newhousenews.com/archive/mcquaid090805.html ].
(MIPT, 2006) The National Memorial Institute for the Prevention of Terrorism (MIPT).
MIPT Terrorism Knowledge Base http://www.tkb.org/Home.jsp

218
(Modarres et al, 1999) Modarres, M., M. Kaminsky, and V. Krivtsov, 1999. Reliability
Engineering and Risk Analysis: A Practical Guide, Marcel Dekker, New York.
(Moore, 1956) Moore, Harry Estill, (1956, Dec.) “T oward a Theory of Disaster,”
American Sociological Review. V ol. 21, No. 6., pp. 733-737.
(Morrow, 1999) Morrow BH. 1999. Identifying and mapping community vulnerability.
Disasters 23(1)1–18.
(Mudan, 1984) Mudan, K.S. “Thermal radiation hazards from hydrocarbon pool fires,”
Prog. Energy. Combust. Sci., 10, 59-80, 1984.
(Murphy, 1989) Murphy, Dean. “Building Curb Urged for North San Pedro Strip,” Los
Angeles Times, July 21, 1989.
(National Atlas, 2006). Spatial Hazard Events and Losses for the United States, 1995-2000,
National Atlas of the United States (National Atlas).
(Neumann et al., 1999) Neumann CJ, Jarvinen BR, McAdie CJ, Hammer GR. 1999.
Tropical cyclones of the North Atlantic Ocean, 1871–1998. NOAA, Historical
Climatology Series 6-2. Available at:
http://www5.ncdc.noaa.gov/pdfs/tropcyc/HCS6-2-intro.pdf
(NOAA, 2002) National Oceanic and Atmospheric Administration (NOAA). Coastal
Services Center. 2002. Case studies for vulnerability assessment techniques
(VAT) I, II, III Workshops. http://www.csc.noaa.gov/vata/index.html.
(De Neufville and Keeney, 1972) De Neufville, R. and R.L. Keeney. “Systems
Evaluations through Decision Analysis: Mexico City Airport,” Journal of Systems
Engineering, 3, 34-50, 1972.
(Nicolet-Monnier and Gheorghe, 1996) Nicolet-Monnier M and Gheorghe AV. 1996.
Quantitative Risk Assessment of Hazardous Materials Transport Systems. Kluwer
Academic Publishers. Dordrecht, The Netherlands.
(NIST, 2005) NIST Chemistry WebBook, NIST Standard Reference Database Number
69, June 2005 Release http://webbook.nist.gov/chemistry/.
(OECD, 1995) OECD Environmental Performance Reviews: Netherlands (OECD).
(1995) Organisation for Economic Co-operation and Development.
(Okuma, 1997) Okuma, Takashi. (1997, February) “A Brief History of Japanese Flood
Control,” World Rivers Review. v. 12, no. 1.

219
(Orosz et al., 2010) Michael D. Orosz, Carl Southwell, Anthony Barrett, Jennifer Chen,
Petros Ioannou, Afshin Abadi, and Isaac Maya, “PortSec: A Port Security Risk
Analysis and Resource Allocation System,” IEEE Conference on Technologies for
Homeland Security, Waltham, MA, November 2010.
(Pelling, 2003) Pelling, Mark (editor). (2003) Natural Disasters and Development in a
Globalizing World. Routledge (UK).
(Pielke et al., 2002) Pielke, Jr., R.A., M.W. Downton, and J.Z. Barnard Miller. Flood
Damage in the United States, 1926-2000: A Reanalysis of National Weather Service
Estimates. Boulder, CO: UCAR, 2002 [ From
http://www.flooddamagedata.org/full_report.html ].
(Popsecu, 2003) Popescu, V., Hoffmann, C., Kilic, S., and Sozen, M. “Producing high-
quality visualizations of large-scale simulations,” Purdue University Computer
Science Dept., Tech. Report 03-011, West Lafayette, Indiana, 2003.
(Pryor, 1977) Pryor, Larry. “Two San Pedro LPG Tanks Worry Officials,” Los Angeles
Times; April 4, 1977.
(Quest, 2005) Quest Consultants Inc. Hazards Analysis of a Proposed LNG Import
Terminal in the Port of Long Beach, California. August 30, 2005.
(Raiffa, 1968) Raiffa, H. 1968: Decision Analysis, Addison-Wesley, Reading,
Massachusetts.
(Raj, 1993) Raj, Phani K. and Clayton K. Turner. Hazardous Materials Transportation
in Tank Cars: Analysis of Risks - Part I. U.S. Department of Transportation,
FRA/ORD-92/34 Washington, DC: Federal Railroad Administration, 15 May
1993.
(Rosoff and von Winterfeldt, 2007) Rosoff, Heather and von Winterfeldt, Detlof, 2007.
“A Risk and Economic Analysis of Dirty bomb Attacks on the Ports of Los
Angeles and Long Beach,” Risk Analysis, 27(3).
(Rittel and Webber , 1973) Rittel, Horst, and W ebber, Melvin, (1973) “Dilemmas in a General
Theory of Planning,” Policy Sciences. 4, pp. 155-59. Elsevier Scientific Publishing Company ,
Inc., Amsterdam.
(Schattschneider, 1960) Schattschneider, E. E., The Semi-Sovereign People: A Realist's
View of Democracy in America, Holt, Rinehart, and Winston, New York, 1960.
(Schumpeter , 1942) Schumpeter , Joseph A. (1942) Capitalism, Socialism and Democracy. New
Y ork and London: Harper and Brothers.

220
(Schwarz, 1983) Schwarz, M. Controversial Decisionmaking: LNG Technology in The
Netherlands, Umweltbundesamt, Bonn, 1983.
(Scott, 1995) Scott, Mel. American City Planning Since 1890. American Planning
Association, 1995.
(Seed et al., 2006) Seed, R. B., R. I. Abdelmalak, A. G. Athanasopoulos, R. G. Bea, G.
P. Boutwell, J. D. Bray, J.-L. Briaud, C. Cheung, D. Cobos-Roa, J. Cohen-
Waeber, B. D. Collins, L. Ehrensing, D. Farber, M. Hanenmann, L. F. Harder, M.
S. Inamine, K. S. Inkabi, A. M. Kammerer, D. Karadeniz, R.E. Kayen, R. E. S.
Moss, J. Nicks, S. Nimala, J. M. Pestana, J. Porter, K. Rhee, M. F. Riemer, K.
Roberts, J. D. Rogers, R. Storesund, A. Thompson, A. V Govindasamy, X. Vera-
Grunauer, J. Wartman, C. M. Watkins, E. Wenk, and S. Yim, “Investigation of the
Performance of the New Orleans Flood Protection Systems in Hurricane Katrina
on August 29, 2005,” Report No. UCB/CCRM – 06/01, University of California,
May 22, 2006.
(Sen, 1995) Sen, Amartya, “Rationality and Social Choice,” The American Economic Review,
V olume 85, Issue I, pp. 1-24, March 1995.
(Simpson, 1974) Simpson, R.H.: “The Hurricane Disaster Potential Scale,”
Weatherwise. 27,169,186, 1974.
(Smith, 1776) Smith, Adam. (1776) An Inquiry Into the Nature and Causes of the W ealth of
Nations. Edinburgh.
(Smith, 1989) Smith, S.K. “Toward a Methodology for Estimating Temporary
Residents,” J. Amer. Statistical Assoc. 84 (46): 430-436, 1989.
(Smutz, 1942) Smutz, Huber E. “Zoning Experiences During Wartime Hysteria,” ASPO
News Letter, VIII (October 1942), p. 82.
(Southwell, 2005) Southwell, Carl. “An Analysis of the Risks of a Terrorist Attack on LNG
Receiving Facilities in the United States," Center for Risk and Economic Analysis of
Terrorism Events, University of Southern California. CREATE Report Number 05-029,
November 2005.
(Southwell, 2011) Southwell, Carl. “In My Back Yard : A Case Study of Institutional
Obduracy in the Face of Public Safety or A Consequence Analysis of a Terrorist
Attack on the Rancho LPG Holdings, LLC Facility in San Pedro, California,”
unpublished, March 2011.
(Southwell and von Winterfeldt, 2008) Southwell, Carl and Detlof von Winterfeldt, “A
Decision Analysis of Options to Rebuild the New Orleans' Flood Control System,”
Natural Disaster Analysis after Hurricane Katrina: Risk Analysis, Economic
Consequences, and Social Implications, Cheltenham, UK: Edward Elgar; July 2008.

221
(Southwell and von Winterfeldt, 2010) Southwell, Carl and Detlof von Winterfeldt,, “Using
Risk Assessment, Economic Assessment, and Risk Management to Improve Preparedness
for Terrorist Attacks and Natural Disasters: A Hurricane/Flood Consequence Model for
Estimating the Impacts of Qualitative Mitigations,” FEMA National Preparedness
Directorate, U.S. Department of Homeland Security, FEMA Grant No. 2008-GA-T8-
K004, July 4, 2010.
(Spicer and Havens, 1989) Spicer, Tom and Jerry Havens, User's Guide for the
DEGADIS 2.1 Dense Gas Dispersion Model, EPA-450/4-89-019, November
1989.
(Stedge et al., 2006) Stedge, Gerald, Mark Landry, and Maged Aboelata. Estimating
Loss of Life from Hurricane-Related Flooding in the Greater New Orleans Area-
Loss-of-Life Modeling Report: Final Report. Alexandria, Virginia: U.S. Army
Corps of Engineers Institute for Water Resources, May 22, 2006.
(Sugano, 1993) Sugano, T. et al. “Full-scale aircraft impact test for evaluation of impact
force,” Nuclear Engineering and Design, Vol. 140, 373-385, 1993.
(Thompson and Sultana, 1996) Thompson, Paul M. and Parvin Sultana. (1996, March)
“Distributional and Social Impacts of Flood Control in Bangladesh,” The Geographical
Journal, V olume 162, Part 1.
(Townsend et al., 2006) Townsend, Frances Fragos, Kenneth P. Rapuano, Joel B.
Bagnal, S. Michele, L. Malvesti, Kirstjen M. Nielsen, Thomas P. Bossert, Daniel
J. Kaniewski, Marie O’Neill Sciarrone, Joshua C. Dozor, Michael J. Taylor,
Stuart G. Baker, Richard W. Brancato, Donovan E. Bryan, Christopher Combs,
Theodore M. Cooperstein, William T. Dolan, Michael O. Forgy, Douglas J.
Morrison, Richard L. Mourey, David C. Rutstein. The Federal Response to
Hurricane Katrina: Lessons Learned, The White House, Washington, DC,
February 23, 2006.
(USACE, 1984) United States Army Corps of Engineers. Lake Pontchartrain,
Louisiana, and Vicinity Hurricane Protection Project – Reevaluation Study.
New Orleans, LA: U.S. Army Corps of Engineers, 1984.
(USACE, 2004) U.S. Army Corps of Engineers Water Resources Planning. (2004)
National Academies Press.
(USC, 2006) University of South Carolina. 2006. Spatial Hazard Events and Losses
Database for the United States (SHELDUS). http://www.sheldus.org .
(USGS, 2006) United States Geological Survey. 2006. Spatial hazard events and losses
for the United States, 1995–2000. National Atlas of the United States. Available
at: http://www.nationalatlas.gov/.

222
(Van de Vink et al., 1998) Van der Vink GR, Allen M, Chapin J, Crooks MM, Fraley
WJ, Krantz AM, et al. 1998. “Why the United States is becoming more vulnerable
to natural disasters.” in EOS Transactions, American Geophysical Union. Vol.
79: No. 44: November 3, 1998, Pages 533-537.
(Westerrint and Luettich, 2003) Westerrint, J.J. and R.A. Luettich, “The Creeping
Storm,” Civil Engineering Magazine, June 2003
(White and Cahill, 2006) White, G.W.; Fox, M.H., Rooney, C.; and Cahill, A. 2006.
Assessing the impact of Hurricane Katrina on persons with disabilities.
http://www.nobodyleftbehind2.org.
(von Winterfeldt, 2006) von Winterfeldt, Detlof. “A risk and decision analysis
framework to protect New Orleans against another Katrina,” R. J. Daniels, D. F.
Kettl, and H. Kunreuther (Eds.) On risk and disaster. Pittsburgh, PA:
Pennsylvania State University Press, 2006.

223
Appendix A: Software Description and Reference for HAZUS-MH

FEMA developed HAZUS-MH with the National Institute of Building Sciences (NIBS).
FEMA describes HAZUS-MH as “a powerful risk assessment methodology for analyzing
potential losses from floods, hurricane winds and earthquakes. In HAZUS-MH, current
scientific and engineering knowledge is coupled with the latest geographic information
systems (GIS) technology to produce estimates of hazard-related damage before, or after
a disaster occurs.”
63
It is a planning tool for estimating the potential loss of life and
infrastructure, and the economic impact of an earthquake, flood, or hurricane. It may be
used to assess damages for scenarios at the county, state, or regional level. The software
package functions with three models and several included data layers. Support software
known as ARCGIS is required, and, for the flood model, the additional spatial analyst
extension is required.

HAZUS-MH is available for no cost from
http://msc.fema.gov/webapp/wcs/stores/servlet/CategoryDisplay?catalogId=10001&store
Id=10001&categoryId=12013&langId=-1&userType=G&type=14 .

63
See http://www.fema.gov/plan/prevent/hazus/hz_orderform.shtm .

224
Appendix B: Software Description and Reference for Hurrevac2010

Hurricane Evacuation or Hurrevac2010 was developed with support from FEMA and the
U. S. Army Corps of Engineers (USACE). It is an evacuation decision support tool for
emergency management officials facing the threat of a hurricane. Hurrevac has many
capabilities. It accesses weather advisories from the National Weather Service’s National
Hurricane Center and plots active storms’ projected and observed tracks and positions. It
also posts traffic information for major hurricanes. For certain areas, Sea, Lake, and
Overland Surge from Hurricanes (SLOSH) model maps are available. These are
graphics depicting along coastal areas the projected surge of water brought on by
hurricane-associated high winds and low pressure. River flood outlook predictions are
also downloadable. Hurrevac can download past data from historical cyclones going
back to the beginning of record keeping on such storms. This capability allows for a
limited degree of analysis on past storms for determining where and when storms have
traveled in the Atlantic and Pacific Ocean basins.

Hurrevac does not provide GIS functionality beyond the plotting of the weather
information, but the forecasted storm and weather data are exportable to shape file
format. This enables performance of more powerful spatial analysis and overlay onto
other themes pertinent to the user’s focus. In addition to its live tracking capabilities,
Hurrevac2010 is a useful tool for evaluating historical and hypothetical storms. When
viewing and analyzing storms from the Archives Tab, you have access to all of the same

225
STORM FEATURES, ANNOTATION, and REPORTS functions employed when
tracking a live storm.

Registered, approved users can apply to download Hurrevac2010 at
http://www.hurrevac.com/register.php ,

226
Appendix C: Software Description and Reference for Landview 6 with
MARPLOT

LandView 6 has its roots in the CAMEO software (Computer-Aided Management of
Emergency Operations). CAMEO was developed by the EPA and the NOAA to facilitate
the implementation of the Emergency Planning and Community Right-to-Know Act.
This product contains both database management software and mapping software used in
the CAMEO system to create a simple computer mapping system involving two
programs—MARPLOT
®
and LandView. The MARPLOT mapping program allows
users to map Census 2000 legal and statistical areas, EPA Envirofact sites, and USGS
Geographic Names Information (GNIS) features. The LandView database system allows
users to retrieve Census 2000 demographic and housing data, EPA Envirofacts data and
USGS GNIS information. The GNIS contains over 1.2 million records which show the
official federally recognized geographic names for all known places, features, and areas
in the United States that are identified by a proper name. It also uses a Population
Estimator function to calculate Census 2000 demographic and housing characteristics for
user defined radii.

227
Appendix D: Hurricanes

Hurricanes are natural hazards with great destructive potential that kill thousands of
people every year worldwide. Hurricanes are categorized by the Saffir-Simpson scale
based on air pressure, sustained winds, and height of storm surge (Table 21). A
hurricane is defined as a tropical storm whose sustained wind speeds exceed 74 miles per
hour. Hurricanes of Category 3 or greater are considered major hurricanes (Neumann,
1999) and cause the greatest damage and loss of life. Hurricane winds can exceed 155
miles per hour. The dangers from hurricanes arise from their winds, storm surges,
rainfall, and internally-generated tornadoes. Hurricanes have wide storm paths that range
in diameter from less than 100 miles to more than 600 miles. A hurricane’s area of most
intense rainfall is its eye wall which surrounds the relatively calm eye of the storm. The
entire storm usually moves at speeds of 10 to 25 miles per hour. When it makes landfall,
it pushes immense amounts of seawater inland in a phenomenon known as storm surge.
In addition to rain, wind, and storm surge, a hurricane produces waves. The strongest
winds of a hurricane occur in its right-front quadrant (when viewed from above). It is
here that tornadoes often form (Christopherson, 2006). Hurricanes originate in and
receive their energy from warm tropical waters. Hurricanes lose their energy when they
move into cooler waters or over land (Ahrens, 2003).

228
Table 21: Saffir/Simpson Hurricane Scale

All U.S. Gulf Coast and Atlantic Coast counties as well as Hawaii are at various risks of
hurricane landfall. More than half of the U.S. population live in coastal areas, and this
number is growing. Thus, many coastal residents have little or no experience with
hurricanes, and this inexperience raises the potential for additional risk (Jarrell 1992).
Development in low-lying areas and certain construction styles make some areas
especially vulnerable. As infrastructure development proceeds in susceptible areas, the
potential for property damage and loss of life increases in spite of improvements in
hurricane forecasting and preparation.

The International Hurricane Research Center at Florida International University has
identified the most vulnerable areas of the United States for hurricane damage. Based on
twelve criteria that include physical factors, socioeconomic indicators, and storm history,
the nation’s most vulnerable areas are New Orleans, Louisiana, Lake Okeechobee,
Florida, the Florida Keys, coastal Mississippi, Miami/Ft. Lauderdale, Florida,

229
Galveston/Houston, Texas, Cape Hatteras, North Carolina, eastern Long Island, New
York, Wilmington, North Carolina, and Tampa/St. Petersburg, Florida.
64

Figure 56: The Number of Hurricanes Expected to Occur During a 100-year Period
Based on Historical Data

Hurricane-prone areas such as coastal wetlands naturally and dynamically adapt to
changes in weather, water levels, and the built environment. As such, it seems reasonable
that these areas would readily adapt to hurricane mitigations. Nevertheless, in areas akin
to the densely populated Miami metropolitan area wherein infrastructure development is

64
Source: http://www.ihc.fiu.edu/media/docs/10_Most_Hurricane_Vulnerable_Areas.pdf

Legend
Light blue area: 20 to 40
dark blue area: 40 to 60
red area: more than 60

Map not to scale.
Source: the National Atlas
and the USGS.

230
extensive, the most potentially dramatic mitigations such as strategic retreat
65
are very
difficult and may not be possible, except for the rebuilding or relocation of capital assets
at the end of their useful lives. In areas such as this where extensive planned retreat are
less viable options, accommodation
66
and coastal reinforcement using soft structures
(e.g., beach maintenance and wetlands reconstruction) is more feasible (Beatley et al., p.
70, 2002).

Historically, many of the modifications and activities meant to increase the productivity
and attractiveness of this area (e.g., dikes, canals, groins, and jetties) and its economic
viability (e.g., agricultural production) have contributed to rapid wetland losses. In
Florida, for example, such poor coastal zone land use choices contributed heavily to
wetland loss that is now in the process of being reversed by the Comprehensive
Everglades Restoration Plan.

65
Strategic or planned retreat or, at least, partial planned retreat, is an option sometimes
available after extensive flood and wind damage resultant from a major hurricane. While partial
retreat in the form of prohibited or controlled rebuilding in certain of the low-lying and flooded
areas is more likely, retreat from an entire devastated area is highly unlikely due to historic,
political, and economic reasons.

66
Accommodation strategy within the city of Miami may be limited to building on pilings,
elevated foundations, and earthen mounds, a strategy that may meet resistance for economic and
aesthetic reasons.

231
Figure 57: Disaster Caused by Hurricanes or Tropical Storms

As of result of hurricanes and their cumulative impacts on these coastal areas, decision
makers, the scientific community, developers, businesses, and individuals must
increasingly collaborate to develop sustainable, long-term management policies and
strategies that seek to find a balance among infrastructure repair and improvement,
sustained resource exploitation, restored ecosystem health, and the continuation of a
unique quality of life. The need to repair and improve infrastructure after hurricanes and

232
the need for the use of integrated coastal zone management
67
(ICM) principles to ensure
that the most critical needs of the coastal zone are addressed present a wicked problem.

Geographic zones such as the Miami MMA tend to have the greatest concentrations of
natural resources (e.g. wetlands, estuaries, fisheries, habitat for coastal dependent
fisheries, and waterfowl), greatest transportation advantages (e.g. ports and transfer
points for multi-mode transportation), greatest frequency and severity of hazards (e.g.
flooding and wind damage from hurricanes, coastal erosion, subsidence, and relative sea
level rise), and greatest populations and built environments at risk from hazards and
greatest intensity of conflicts among all the stakeholders that seek to exploit, conserve, or
preserve coastal resources and environments.
68
In such geographic zones, ICZM tackles
the balance between coastal dependent economic issues such as fisheries, maritime trade,
and tourism and environmental and social issues such as the exploitative development of
coastal areas with its resultant population growth and urbanization, the environmental
degradation of marine ecosystems, and the increasing vulnerability of development to
natural hazards. ICZM is important for the stakeholders who seek to optimize the
cumulative impacts or carrying capacity within the coastal area.

67
ICZM is the long-term interdisciplinary process and goal wherein natural and social scientists,
coastal managers, stakeholders, and decision-makers focus on how to manage the diverse
problems of coastal areas. Qualitative hazard mitigation practices are part of ICZM.

68
See, for example, the proceedings contained in the Report of the Workshop on the Planning
and Management of Modified Mega Deltas, 24 – 26 September 2001, The Hague, The
Netherlands.

233
ICZM as a planning problem can be characterized as a conflict between the dominance of
efficiency (Rittel and Webber, 1973) and exigency. Achieving ICZM planning goals
expeditiously and effectively may be considered optimal. However, such attempts at
rational planning are often flawed insofar as they assume unbiased, objective
stakeholders with perfect information. Lacking this idealized state, decision makers are
tasked with the need to identify and view the goals and aims of the various stakeholders
in a coastal zone before even seemingly straightforward concepts like efficiency and
exigency can be contextualized. For example, efficiency can be defined to include
elements of allocational efficiency (the choices of coastal zone land uses) (Adam Smith,
1776), the marginal efficiency of capital (the marginal rate of return over cost of the
specific coastal zone land use investment) (John M. Keynes, 1936), the efficiency of
renewal (the “creative destruction” of change as expressed in coastal zone land use
redevelopment) (Joseph Schumpeter, 1942), and distributive efficiency (the efficiency
derived from reasoned social welfare as realized within the coastal zone land use issues
such as coastal access) (Amartya Sen, 1995). And, exigency is viewed as self-evident in
some coastal zone planning choices such as the continuing reconstruction of New Orleans
and its levees after Hurricane Katrina. At the same time, Rittel and Webber argue that
such goals are impossible to attain since the “plurality of objectives held by pluralities of
politics makes it impossible to pursue unitary aims” (Rittel and Webber, 1973.) The
practicing coastal disaster planner’s roles, then, must include negotiation and mediation
to effect change and to reach resolution (not solution) of such wicked problems.

234
Wetlands’ loss has resulted from both natural and human-induced causes. Natural causes
include the wind and wave action of storms and sea level rise. Human-made causes
include the channelization of estuaries and dredging of canals through wetlands for
agricultural and urban development purposes. Every manipulation of the environment
makes an impact on a region’s social geography, and each change, in turn, requires
continual maintenance and improvement. The maintenance of the carrying capacity of
the region is a continuing, wicked problem. The balance necessary for sustainable
development is a continuing challenge that is perhaps best approached through the lens of
ICZM and its application.

The role of coastal wetlands as an important buffer against the damage that would
otherwise be caused by hurricanes has historically been underestimated. In the absence
of wetlands, artificial barriers to flooding and storm surges would be devastated by direct
exposure to hurricane-generated wind and waves. Water management has many goals
besides flood control—and these goals vary with distance from the individual mitigations
utilized as well as over time. Selected “fixes” might be analyzed with respect to expected
(cost and benefit) consequences associated with geographic proximity, time horizon,
impact on biodiversity, maintenance of wetlands, environmental degradation, and
individual preferences, among other consequences.

In Japan’s Edo era (1603-1868), people attempted to coexist with low-lying rivers and
wetlands. Accepting infrequent flooding, they allowed water to overflow where little
damage would be done. In this way, major overflows were avoided at locations where

235
flooding could breach levees and cause extensive damage. Forests were planted beside
rivers to further mitigate flood damage. When high waters overflowed, the forest belt
weakened the flow. As a result, overflows usually contained only slow moving, fine
sediment. Although farmland was submerged, farmers in the Edo era welcomed a flood
if it occurred only infrequently because its sediments left fields enriched (Okuma, 1997).

To cope with possible house flooding, elevated houses were built on mounds, and
evacuation boats were kept available for emergencies. Such mitigation measures
minimized flood damage and loss of life.

Flood control and river improvement can be based on the knowledge of nature,
recognition of the limitations of technology, and a willingness to coexist with low-lying
areas. If such a policy were implemented, flooding might be more frequent, but less
damaging, in affected areas. People would have to recognize the inherent inequality in
flood damage due to the differences in geographical location and accept a certain amount
of overflow. House flooring would need to be elevated, basements water-proofed or
eliminated, and appropriately-rated flood damage insurance introduced as practical
measures to mitigate damage. The construction of retention basins and permeable
pavements could also be promoted along the entire system. At the same time, the
maximum size of flood flow to be contained could be decided while taking measures to
ensure that overflow runs gently and can be returned to main drainage channels.

Since the flood of 1953, the Dutch have also developed a sense for ICZM. Safety for
people living in coastal areas is a primary focus, and factors such as natural resources,

236
recreation, and comfortable habitation are also taken into account.. Water quality, the
environment, nature, fishery, recreation, agriculture, shipping, and industry are all
incorporated into its national hazard mitigation strategy.
69

The Padma, Jamuna, and the lower Meghna Rivers are the widest in Bangladesh, helping
form a river system outflow only surpassed by the Amazon and Congo river systems.
Regular flooding is a chronic problem in poverty-stricken Bangladesh. An effective life-
saving mechanism successfully utilized in Bangladesh has been the construction of
concrete shelters built on stilts as emergency shelters for flood and monsoon victims
(Thompson, 1996). Emulation of this concept in south Florida and other flood-prone
areas could also be life-saving.

Reflecting on the conflict between political will and the technical, physical, and
socioeconomic challenges with respect to repair, mitigation, and enhancement
possibilities after a hurricane and within this framework, several observations can be
made. First, there is no definitive technical, socioeconomic, or physical solution to most
dilemmas. There almost always exist sets of solutions or resolutions to problems,
especially when comprehensive socioeconomic, physical, and political variables are
incorporated into robust solutions. Indeed, the choice of a technical, socioeconomic, or
physical solution to a planning problem is often a political solution. One should be
cognizant of the political probability of success of the various proposed strategies, aware
of the political barriers and opportunities created by these various strategies, and mindful

69
See www.delatawerken.com for background information on the Delta Works.

237
that other strategies (that have their own and comparable opportunity costs) exist (May,
1986). Regardless, the relative desirability of an ICZM problem being resolved within
the political or apolitical dimension is undeniable. A synthesis of the techniques may
serve as a starting point in formulating a solution to this problem. A model such as that
presented herein serves as a tool toward approaching sets of potential solutions for this
slice of the many problems of this type.

Abstract (if available)

Abstract There are many types of disasters, and it is a challenge to derive a single model of disaster risk and decision analysis that encompasses them all. At the same time, a general model for analyzing disaster risks and decisions could make the tasks of disaster avoidance, prevention, mitigation, rescue, rebuilding, and recovery much easier. This dissertation formulates a model of risk and decision analysis concerned with behaviors (decisions or “objectives”) and stimuli (initial “risks and uncertainties” and their associated sets of choices/utilities/“levers”) that are chronicled and calibrated by “metrics.” Changes in this model are displayed as changes in preferences over time or as conditioned by prior choices. In practice, this dynamic model of risk and decision analysis is usually represented as one or a set of static “time slices” from the model that serve as exemplars for the policy implications the specific, practical analysis is attempting to highlight. ❧ This dissertation showcases some specific, practical analyses—six case studies—whose underlying model is derived from this general model. Each case makes its own points about policy implications related to its specific disaster type and circumstances while, at the same time, each demonstrates the robustness and flexibility of the more general model. Each case employs qualities of a robust model as follows: 1. It outlines the essential elements and relationships in the disaster process being considered, 2. It establishes a set of data that describes and measures the disaster process being considered, 3. It offers opportunities for comparison to actual or hypothetical disasters, 4. It acts as a template to better understand and help prevent recurrences of past disasters, to react and respond to and recover from unfolding disasters, and to plan for and mitigate future disasters, and 5. It serves as a tool to communicate the risks and potential consequences of the disaster being considered and the actions necessary to mitigate its risk.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Using decision and risk analysis to assist in policy making about terrorism

PDF

Critical factors in evaluating compliance to United Nations Security Council Resolution 1540: developing a methodology for compliance evaluation

PDF

A strategic talent management retention model: an effective way to shape the United States Space Force

PDF

Threat assessment of physical attacks of electrical infrastructure: how do terrorist groups select their targets?

PDF

A multiattribute decision model for the selection of radioisotope and nuclear detection devices

PDF

A risk analysis methodology to address human and organizational factors in offshore drilling safety: with an emphasis on negative pressure test

PDF

Life without nuclear power: a nuclear plant retirement formulation model and guide based on economics: San Onofre nuclear generating station case: economic impacts and reliability considerations ...

PDF

The cost of missing objectives in multiattribute decision modeling

PDF

Homegrown violent extremism: designing a community-based model to reduce the risk of recruitment and radicalization

PDF

Information sharing, deliberation, and collective decision-making: A computational model of collaborative governance

PDF

Emerging technology and its impact on security policy

PDF

Water security, national security and MCIWest: a grounded theory for operationalizing risk management

PDF

Making terrorism risk assessments more useful for decision-making

PDF

A stakeholder approach to reimagining private departments of public safety: the implementation of a community advisory board recommendation report

PDF

An examination of the law enforcement response to active shooter incidents in U.S. cities: findings and implications for improving future response efforts

PDF

Housing consumption-based asset pricing and residential mortgage default risk

PDF

Cryptographic currency and economic security: threats, opportunities, and regulatory challenges

PDF

The impact of social capital: a case study on the role of social capital in the restoration and recovery of communities after disasters

PDF

Macroeconomic conditions, systematic risk factors, and the time series dynamics of commercial mortgage credit risk

PDF

Use of a clinical decision model to support the transition of services for individuals with autism spectrum disorder

Asset Metadata

Creator Southwell, Carl Erckman (author)

Core Title A robust model for disaster risk and decision analysis

School School of Policy, Planning and Development

Degree Doctor of Policy, Planning & Development

Degree Program Policy, Planning, and Development

Publication Date 05/08/2012

Defense Date 02/27/2012

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag decision analysis,disaster,flood,hurricane,LNG,LPG,Model,OAI-PMH Harvest,Port,risk,risk analysis,Security,terrorist attack

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Richardson, Harry W. (committee chair), von Winterfeldt, Detlof (committee chair), Maya, Isaac (committee member), Pryor, Lawrence (committee member), Southers, Erroll (committee member)

Creator Email carl.southwell@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-36559

Unique identifier UC11290257

Identifier usctheses-c3-36559 (legacy record id)

Legacy Identifier etd-SouthwellC-810.pdf

Dmrecord 36559

Document Type Dissertation

Rights Southwell, Carl Erckman

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA