EPA/600/R-21/096 | September 2021
www.epa.gov/emergency-response-research
United States
Environmental Protectior
Agency
oEPA
INTERCONNECTED
INFRASTRUCTURE
MODELING
INVESTIGATION
Office of Research and Development
Homeland Security Research Program
-------
SEPA
INTERCONNECTED INFRASTRUCTURE
MODELING INVESTIGATION
by
Tanvi Joshi*, Barrett Richter*, Dr. Ryan James*, Timothy Boe**, Dr. M. Worth Calfee**, Dr. Sang Don
Lee**, Joseph Wood**, Dr. Paul Lemieux**, Dr. Shawn Ryan**, Leroy Mickelsen***
*Battelle Memorial Institute
Columbus, Ohio
**US EPA Office of Research and Development (ORD)
Center for Environmental Solutions and Emergency Response (CESER)
Homeland Security Research Program (HSRP)
Durham, NC 27709
***US EPA Center for Environmental Solutions and Emergency Response (CESER)
Office of Land and Emergency Management (OLEM)
CBRN Consequence Management Advisory Division (CMAD)
Durham, NC 27709
Contract EP-C-16-014 to Battelle Memorial Institute
-------
ACKNOWLEDGMENTS
Contributions of the following individuals and organizations to this report are acknowledged:
US EPA Technical Reviewers of Report
Dr. Katherine Ratliff (EPA/ORD/CESER/HSRP)
Scott Hudson (EPA/OLEM/CMAD)
US EPA Quality Assurance
Ramona Sherman (EPA/ORD/CESER/HSRP)
11
-------
DISCLAIMER
The U.S. Environmental Protection Agency, through its Office of Research and Development,
funded and managed the research described here under Contract EP-C-16-014, Task Order
68HERC19F0117, with Battelle Memorial Institute. This document has been reviewed in
accordance with U.S. Environmental Protection Agency policy and approved for publication.
Any mention of trade names, manufacturers or products does not imply an endorsement by the
United States Government or the U.S. Environmental Protection Agency. EPA and its employees
do not endorse any commercial products, services, or enterprises. The contractor role did not
include establishing Agency policy.
Questions concerning this document, or its application, should be addressed to:
Timothy Boe
U.S. Environmental Protection Agency
Office of Research and Development
Center for Environmental Solutions and Emergency Response
109 T.W. Alexander Dr. (MD-E-343-06)
Research Triangle Park, NC 27711
Phone 919.541.2617
in
-------
FOREWORD
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the
Nation's land, air, and water resources. Under a mandate of national environmental laws, the
Agency strives to formulate and implement actions leading to a compatible balance between
human activities and the ability of natural systems to support and nurture life. To meet this
mandate, EPA's research program is providing data and technical support for solving
environmental problems today and building a science knowledge base necessary to manage our
ecological resources wisely, understand how pollutants affect our health, and prevent or reduce
environmental risks in the future.
The Center for Environmental Solutions and Emergency Response (CESER) within the Office of
Research and Development (ORD) conducts applied, stakeholder-driven research and provides
responsive technical support to help solve the Nation's environmental challenges. The Center's
research focuses on innovative approaches to address environmental challenges associated with
the built environment. We develop technologies and decision-support tools to help safeguard
public water systems and groundwater, guide sustainable materials management, remediate sites
from traditional contamination sources and emerging environmental stressors, and address
potential threats from terrorism and natural disasters. CESER collaborates with both public and
private sector partners to foster technologies that improve the effectiveness and reduce the cost
of compliance, while anticipating emerging problems. We provide technical support to EPA
regions and programs, states, tribal nations, and federal partners, and serve as the interagency
liaison for EPA in homeland security research and technology. The Center is a leader in
providing scientific solutions to protect human health and the environment.
This report describes preliminary research to simulate interconnected infrastructure systems to
support infrastructure system remediation following a wide area CBRN incident. This research
included 1.) reviewing the literature that describes methods and software tools for infrastructure
modeling, 2.) identifying the infrastructure sectors that will be considered in modeling efforts, 3.)
assessing modeling software that could potentially simulate the interdependence of
infrastructure, and 4.) selecting and implementing a final model. The selected model, the
Stochastic Infrastructure Remediation Model (SIRM), was applied to hypothetical contamination
to validate the results.
Gregory Sayles, Director
Center for Environmental Solutions and Emergency Response
iv
-------
TABLE OF CONTENTS
Acknowledgments ii
Disclaimer iii
Foreword iv
List of TABLES viii
List of Figures x
Acronyms and Abbreviations xiii
Executive Summary
1 Introduction 1
2 Model Selection Literature Review 2
2.1 Quality Assurance/Quality Control 2
2.2 Interconnected Infrastructure Modeling Methodologies 2
2.3 Software Evaluation 8
2.4 Conclusions from Literature Review 15
3 Stochastic Infrastructure Remediation Model (SIRM) 16
3.1 Introduction 16
3.2 Infrastructure Sectors 19
3.3 Model Methodology 19
3.4 Model Assumptions 23
3.5 Adjustments of Dependency Map 24
3.5.1 Literature Search and Subj ect Matter Expert Interactions 24
3.5.2 Coefficient Modification 26
3.6 Decontamination Factor 27
4 GIS Based Analysis of Denver Scenarios 28
4.1 Obtaining Infrastructure Asset Counts 29
4.1.1 Raw Queries of the HAZUS Database 29
4.1.2 Census Block Analysis of the HAZUS Database 30
v
-------
4.2 Determining Efficiency Losses from Denver Scenarios 33
5 Operating Efficiency Measurement and Calculations 34
5.1 Water and Wastewater Efficiency 35
5.2 Energy Efficiency 35
5.3 Transportation Efficiency 36
5.4 Communication Efficiency 36
5.5 Government Efficiency 37
5.6 Food & Agriculture Efficiency 37
5.7 Emergency Services Efficiency 38
5.8 Waste Management Efficiency 38
5.9 Healthcare Efficiency 39
6 Model Verification with Historical Data 39
6.1 Sample Scenario 1: Hurricane Harvey 39
6.2 Sample Scenario 2: Hurricane Maria 44
6.3 Sample Scenario 3: Hurricane Sandy 46
6.4 Sample Scenario 4: Elk River Spill 47
6.4.1 Addition of Backups 48
6.4.2 Scenario Results with Backups 49
6.5 Sample Scenario 5: Hurricane Florence 50
6.5.1 Scenario Results 51
6.5.2 S cenari o with No Infrastructure Reducti on 52
6.6 Conclusions from SIRM Verification 54
7 Sensitivity Analysis 54
7.1 Initial Efficiency 55
7.1.1 Effects of Water 55
7.1.2 Effects of Energy 57
7.1.3 Effects of Waste Management 59
vi
-------
7.1.4 Effects of Healthcare 61
7.2 Repair Factors 63
7.3 Backup Analysis 65
8 Conclusion 67
9 SIRM Improvement Plan 68
10 References 69
Appendix A: Matrix of Reviewed Literature 74
Appendix B: Coordinates of Scenarios 1 and 2 75
Appendix C: Occupancy Class Definitions in HAZUS Database 76
Appendix D: Sensitivity Analysis Results 80
vii
-------
LIST OF TABLES
Table 1: AWARE Results for Denver Biological Contamination Scenarios: Cost and
Timeline Estimates 15
Table 2: Preliminary Set of Reactions Used in Early Versions of the SIRM 18
Table 3: Infrastructure Sectors Selected in the Preliminary Development of the SIRM 19
Table 4: System of Reactions Currently Implemented in the SIRM 23
Table 5: Infrastructure Connections Mentioned in Literature Search 25
Table 6: Infrastructure Equations 27
Table 7: Relevant Raw HAZUS Database Percentage Contaminated by Infrastructure
Sector 30
Table 8: Occupancy Class Mappings of DHSCIS Sectors 31
Table 9: HAZUS Census Block Scenario Data 32
Table 10: DHS Infrastructure Sectors HAZUS Efficiency Data Quality Ratings 33
Table 11: Preliminary Efficiency Loss Estimations of Denver Scenarios 34
Table 12: Breakdown of HAZUS Data 41
Table 13: Percentage Flooded for Each Infrastructure 41
Table 14: Hurricane Harvey Efficiency Values 42
Table 15: Recovery Time Comparison 43
Table 16: Initial Infrastructure Efficiencies for Hurricane Maria [32] 44
Table 17: Infrastructure Repair Factors for Hurricane Maria 45
Table 18: Initial Efficiency Values for Hurricane Sandy 46
Table 19: Availability of Backup Water 49
Table 20: Inputs for Hurricane Florence Based Scenario 51
Table 21: Hurricane Florence Recovery Times [38-41] 52
Table 22: Repair Factors with No Reduction 53
Table 23: Recovery Times Without Infrastructure Reduction [38-41] 53
viii
-------
Table 28: Matrix of Reviewed Literature 74
Table 29: Key of HAZUS Occupancy Classes, Page 1 of 4 [27] 76
IX
-------
LIST OF FIGURES
Figure 1: ADHS preparedness roadmap of interconnected infrastructures, an example
perspective 5
Figure 2: Denver scenarios mapped in PATH/AW ARE 10
Figure 3: User inputs to PATH: prioritization weights 11
Figure 4: Truncated portion of the infrastructure dependency diagram in PATH 11
Figure 5: Example sector functionality and dependency inputs 12
Figure 6: PATH assets results for Denver scenarios 1 and 2 13
Figure 7: Subset of user inputs to AWARE for cost and timeline estimation 14
Figure 8: Side by side comparison of deterministic and stochastic models in early versions
of the SIR VI 18
Figure 9: A visual representation of the preliminary infrastructure dependencies in the
system 21
Figure 10: Revised infrastructure diagram 26
Figure 11: View of the Denver Metropolitan Area in QGIS with Scenarios 1 (purple) and
2 (pink) 29
Figure 12: Closer view of Denver Scenarios 1 (purple) and 2 (pink) mapped out in QGIS 29
Figure 13: Power outage map for Hurricane Sandy [33] 36
Figure 14: Communication outage map for Hurricane Florence. Colors indicate the
percentage of outage 37
Figure 15: Agriculture map of Franklin County 38
Figure 16: Flooding data for Houston during Hurricane Harvey, with red dots indicating
flooded points 40
Figure 17: Example of flooded areas overlaid on HAZUS blocks 40
Figure 18: Efficiency/time profile of Hurricane Harvey scenario 43
Figure 19: SIRM results 46
Figure 20: Hurricane Sandy efficiency chart 47
Figure 21: Recovery time of Water infrastructure 48
x
-------
Figure 22: Histograms of recovery time of Water (top: original, bottom: with backups) 50
Figure 23: Effects of initial Water efficiency 56
Figure 24: Effects of initial Energy efficiency 58
Figure 25: Effects of initial Waste Management efficiency 60
Figure 26: Effects of initial Healthcare efficiency 62
Figure 27: Infrastructure repair factors effects on Waste Management recovery time 64
Figure 28: Infrastructure repair factors effects on Transportation recovery time 65
Figure 29: Sample chart for backup analysis (recovery time of Communication) 66
Figure 30: Recovery time of Food and Agriculture with backup Water 67
Figure 31: Infrastructure repair factors effects on Water recovery 80
Figure 32: Infrastructure repair factors effects on Energy recovery 80
Figure 33: Infrastructure repair factors effects on Transportation recovery 81
Figure 34: Infrastructure repair factors effects on Communication recovery 81
Figure 35: Infrastructure repair factors effects on Food and Agriculture recovery 82
Figure 36: Infrastructure repair factors effects on Emergency Services recovery 82
Figure 37: Infrastructure repair factors effects on Government recovery 83
Figure 38: Infrastructure repair factors effects on Waste Management recovery 83
Figure 39: Infrastructure repair factors effects on Healthcare recovery 84
Figure 40: Recovery time of Water with backup Water 85
Figure 41: Recovery time of Water with backup Energy 85
Figure 42: Recovery time of Energy with backup Water 86
Figure 43: Recovery time of Energy with backup Energy 86
Figure 44: Recovery time of Transportation with backup Water 87
Figure 45: Recovery time of Transportation with backup Energy 87
Figure 46: Recovery time of Communications with backup Water 88
xi
-------
Figure 47: Recovery time of Communications with backup Energy 88
Figure 48: Recovery time of Government with backup Water 89
Figure 49: Recovery time of Government with backup Energy 89
Figure 50: Recovery time of Food and Agriculture with backup Water 90
Figure 51: Recovery time of Food and Agriculture with backup Energy 90
Figure 52: Recovery time of Emergency Services with backup Water 91
Figure 53: Recovery time of Emergency Services with backup Energy 91
Figure 54: Recovery time of Waste Management with backup Water 92
Figure 55: Recovery time of Water with backup Energy 92
Figure 56: Recovery time of Healthcare with backup Water 93
Figure 57: Recovery time of Healthcare with backup Energy 93
xii
-------
ACRONYMS AND ABBREVIATIONS
ABM
agent-based model
ARF
adjusted repair factor
AWARE
Analyzer for Wide Area Effectiveness
BEA
Bureau of Economic Analysis (US Department of Commerce)
CBRN
chemical, biological, radiological and nuclear
CEV
Constant Elasticity of Variance
DHS
(US) Department of Homeland Security
DHSCIS
(US) Department of Homeland Security's Critical Infrastructure Sectors
EPA
(US) Environmental Protection Agency
FCC
Federal Communications Commission
GIS
Geographic Information System
GRiD
Geospatial Repository and Data Management System
GUI
graphical user interface
HCBM
Hierarchical Coordinated Bayesian Model(ing)
HHM
Hierarchical Holographic Model(ing)
HIFLD
Homeland Infrastructure Foundation-Level Data
IIM
Inoperability Input-Output Model
MCHM
methylcyclohexane methanol
PATH
Prioritization Analysis Tool for All- Hazards
RF
repair factor
SIC
Standard Industrial Classification
SIRM
Stochastic Infrastructure Remediation Model
SME
subject matter expert
SoS
systems of systems
SWT
Surrogate Worth Tradeoff
US
United States
WARRP
Wide Area Recovery and Resiliency Program
Xlll
-------
EXECUTIVE SUMMARY
Wide area chemical, biological, radiological, and nuclear (CBRN) incidents, whether products of
terrorism, war, or accidents, have the potential to damage core infrastructure assets. In this
situation, not only are directly affected assets not able to operate, but operations in other
infrastructure sectors might not be able to operate without the services of the affected assets. For
example, in a CBRN incident that affects an electric power plant, the operations of connected
transportation systems, communications facilities, and/or hospitals would likely be impacted,
regardless of whether these sectors were directly affected by the incident. To effectively respond
to and remediate such incidents, information about the interconnectedness of infrastructure
systems is necessary to bring services back online as quickly as possible.
This report describes preliminary research to simulate interconnected infrastructure systems to
support infrastructure system remediation following a wide area CBRN incident. This research
included 1.) reviewing the literature that describes methods and software tools for infrastructure
modeling, 2.) identifying the infrastructure sectors that will be considered in modeling efforts, 3.)
assessing modeling software that could potentially simulate the interdependence of
infrastructure, and 4.) selecting and implementing a final model. The selected model, the
Stochastic Infrastructure Remediation Model (SIRM), was applied to hypothetical contamination
to validate the results.
The literature review documented the methods other researchers used to model interconnected
infrastructure systems. Many of the models of infrastructure remediation found in the literature
review are used to estimate immediate losses after a CBRN incident, but the dynamic process of
restoring infrastructure is not determined effectively. The literature review included an overview
of existing infrastructure applications, many of which were agent-based models (ABMs), models
using a computational method to simulate a system as an assemblage of entities, agents, that are
programmed to follow specific (simple or complex) rules. While these models proved to be
powerful in their ability to update themselves and the results based on using real-time data
(including time and other factors), they were not able to model all 16 sectors in the U.S.
Department of Homeland Security's Critical Infrastructure Sector (DHSCIS) list, and the
methodology used by many of these models is unclear. On top of the literature specific to
infrastructure systems, several more general modeling strategies were considered. Stochastic
modeling processes were studied due to their powerful ability to incorporate statistics that
provide the user with a range of potential outcomes. Although these modeling strategies were not
used for infrastructure modeling in previous literature, they provided the ability to be generalized
to all DHSCIS sectors and allowed for statistics around the effects of CBRN incidents to be
collected.
An ABM critical infrastructure modeling software program titled PATH/AW ARE (which
contains two acronyms that stand for Prioritization Analysis Tool for All-Hazards [PATH] and
Analyzer for Wide Area Effectiveness [AWARE]) was reviewed in detail. This program allows
the user to model a CBRN incident by placing a user-created contaminated area on a map and
then determines the infrastructure assets that are directly affected by the incident based on the
-------
plume. The program also determines other assets affected by the losses in other facilities. Once
the affected assets are determined, the user can run either the PATH or AWARE tool on the
scenario. The PATH tool is used to generate a prioritized list of assets to repair, and lists all the
asset dependencies, while the AWARE tool is used to calculate restoration costs and timelines
based on a set of user-defined inputs (e.g., number of decontamination teams, labor rates,
number of people on a team). Although PATH/ AW ARE provides helpful information for direct
response to CBRN incidents, this package does not dynamically model the systems of interest (as
it is merely doing a static calculation of restoration timelines) and some interconnected response
elements (e.g., waste and decontamination) are not coupled in the model.
A model based on the Gillespie algorithm was developed - a stochastic modeling process
originally intended for modeling a system of coupled chemical reactions. This model allows for
all infrastructure sectors to be modeled and considers the realistic variability of the impact of a
CBRN incident. The SIRM considered each of the infrastructure sectors as an operating
efficiency percentage and models the restoration of services in each sector as a set of reactions
that use resources from one sector (like Energy) to restore services in other sectors (like
Healthcare and Communications). This process dynamically models the time required to repair
an infrastructure sector, while also allowing the user to consider changes in resource allocation
based on user-defined repair factors. The SIRM can dynamically model all the desired sectors
and allows the user to draw statistical conclusions specific to a CBRN incident. Because of these
desirable model characteristics, the SIRM was selected for use in future infrastructure analyses
and a model improvement plan was generated to guide the future development of the model.
An important challenge in this research is defining the impact of a CBRN incident on critical
infrastructure. As most scenarios are defined primarily by the areas of exposure and the people
affected by that exposure, there is no obvious link between the results of the scenario and the
critical infrastructure remediation that is necessary to restore services. Two biological attack
scenarios in the city of Denver, Colorado that were identified by the Wide Area Recovery and
Resiliency Program (WARRP) were considered for analysis and converted to infrastructure
efficiency loss estimates for use in the SIRM. To do this, a Geographical Information System
(GIS) package titled QGIS was used to query FEMA's HAZUS database (a collection of
inventory databases distributed with the software for FEMA's HAZUS standardized risk
modeling methodology) to gather information directly about the affected infrastructure assets.
Queries on the raw HAZUS database as well as shapefile analysis of HAZUS census blocks were
used to estimate counts of infrastructure assets in the 16 sectors of the DHSCIS list, which were
then converted to percentages of infrastructure based on Denver's total metropolitan area.
Although most sectors had enough data to generate a preliminary loss estimate; a few sectors had
either sparse data sets or their data were inseparably shared with other infrastructure sectors. This
process of estimating the initial conditions for the model, as well as the mechanics of the SIRM
itself, are subject to subject matter expert input as part of an effort to enhance current
infrastructure modeling capabilities.
1
-------
1 INTRODUCTION
Responding to all types of disasters and emergencies is a critical function of government
agencies to maintain the order of society during extenuating circumstances. Among those
circumstances that could cause the most public disruption and alarm would be a wide area
chemical, biological, radiological, or nuclear (CBRN) incident, whether an accidental release or
an intentional attack. Following a CBRN incident, critical infrastructure needs to be restored in a
timely manner. In previous attempts to address the issue of critical infrastructure, disagreements
over the relative importance of different critical infrastructure assets have arisen. However,
infrastructure assets operate with dependence on other infrastructures in a non-trivial way, so
linear decision making that merely ranks the relative importance of assets is inadequate. Instead,
it is necessary to discern which assets need to be restored first in the network of interdependent
assets. A system is needed to evaluate priorities in a systematic manner. Thus, a critical
infrastructure model that could quantitatively assess the infrastructure restoration would be
beneficial in decision making.
Preliminary research on the issue of critical infrastructure restoration defines the context and
scope of the critical infrastructure modeling effort as 1.) reviewing the literature that describes
methods and software tools for infrastructure modeling, 2.) identifying the infrastructure sectors
that will be considered in modeling efforts, 3.) assessing modeling software that could
potentially simulate the interdependence of infrastructure, and 4.) selecting and implementing a
final model. The selected model, the Stochastic Infrastructure Remediation Model (SIRM), was
applied to hypothetical scenarios and compared with historical data to validate the results. The
final model can also be used to evaluate response strategies, assess limitations and bottlenecks in
current response operations, and assess gaps in capability and knowledge.
To better compare models applicable to the dynamics of interconnected infrastructure systems
during a CBRN incident, criteria were established to describe an optimal model. The model
would:
• Capture the inter-connectivity of infrastructure systems over time during a CBRN
incident (i.e., a dynamic model);
• Consider real world systems using obtainable data;
• Lead the user to make conclusions on how to respond to CBRN incidents by providing
informative results;
• Be consistent with or build upon methodologies described in the literature and used in
various applications;
• Minimize computational time;
• Operate within the U.S. Environmental Protection Agency (EPA) Windows 10 Enterprise
computing environment without requiring enhanced user access privileges;
• Be easy to install, use, and understand; and
• Incorporate some Geographical Information System (GIS) analysis.
1
-------
In selecting a model to be used in future infrastructure analysis, these criteria were used to
measure the capability and applicability of the models.
2 MODEL SELECTION LITERATURE REVIEW
To select models that are consistent with methodologies used in the literature, a brief literature
review was conducted. The review included books and academic journals. In addition, it
included an overview of previously developed software programs (both commercially available
and academic) related to the analysis and modeling of interconnected infrastructure systems. The
review used search terms related to infrastructure sectors as well as recommendations by subject
matter experts. The literature review was used in selecting an initial set of infrastructure systems
for preliminary analysis and in selecting one major modeling effort for deeper analysis.
2.1 Quality Assurance/Quality Control
The purpose of this study was to describe preliminary research to simulate interconnected
infrastructure systems to support infrastructure system remediation following a wide area CBRN
incident. The work and conclusions presented as part of this study were empirical and
observational - no scientific experiments were performed. Technical area leads with expertise
evaluated the quality of the information collected by this effort (i.e., secondary data), and
determined what information should be documented within the literature review. All supporting
documentation of the secondary data considered worthy for inclusion were cited. However, no
experimental confirmation of secondary data (e.g., accuracy, precision, representativeness,
completeness, and comparability) was conducted as part of this study.
2.2 Interconnected Infrastructure Modeling Methodologies
The book "Modeling and Managing Interdependent Complex Systems of Systems" by Yacov Y.
Haimes outlines the status quo on modeling systems of systems (SoS), with examples including
analysis of interdependent infrastructures [2], The Inoperability Input-Output Model (IIM) is a
static model that uses linear algebra methods to calculate operating efficiencies of infrastructures
after an applied perturbation, or outage [2], IIM is based on Wassily Leontif s economic input-
output model and holds the same assumptions, which include a set number of infrastructure
sectors that are linearly related to one another and a static economy (equilibrium-competitive)
[3], The model requires a set of initial operating inoperability values (opposite of an operating
efficiency value), a vector of perturbations, c, to each sector (i.e., outages, which are equivalent
to shifts in demand in Leontif s model), and a matrix of values that relate each infrastructure
sector by their dependence on one another, known as A values [2], For example, the energy
sector might have a low A value for the healthcare sector as the healthcare sector is not needed to
generate energy, while the healthcare sector might have a high A value since it requires energy to
operate hospitals. Equation 1 shows the process by which the perturbation vector, c, is applied to
the inoperability values, x.
2
-------
x = Ax + c
Equation 1: Vector Perturbations of IIM
Normally, the hardest part of accurately modeling systems with IIM is obtaining proper^ values.
However, the Bureau of Economic Analysis (BEA) has outlined a set of A values to use for the
IIM for a large set of infrastructure sectors, a subset of which are used in Haimes's book to
generate the A matrix in Equation 2 [4], The represented infrastructures (in order) are 1.) coal; 2.)
petroleum refining; 3.) railroads and related services; 4.) trucking and courier; 5.) water
transportation; 6.) air transportation; 7.) telephone and telegraph; and communication services;
8.) electric services; 9.) water supply and sewage systems; 10.) banking; 11.) eating and drinking
places; and 12.) hospitals. In Equation 2, the^4 value in row i and column j represents the relative
dependency of infrastructure i on infrastructure j.
A =
0.1130
0.0000
0.0000
0.0000
0.0002
0.0000
0.0000
0.0144
0.0000
0.0000
0.0000
0.0000
0.0826
0.0618
0.0308
0.0385
0.0848
0.1307
0.0006
0.0093
0.1635
0.0005
0.0011
0.0015
0.2236
0.0026
0.0617
0.0025
0.0020
0.0014
0.0001
0.0338
0.0313
0.0001
0.0008
0.0009
0.0667
0.0050
0.0020
0.1569
0.0169
0.0047
0.0012
0.0035
0.4487
0.0064
0.0054
0.0040
0.0060
0.0010
0.0002
0.0003
0.1247
0.0006
0.0000
0.0008
0.0000
0.0000
0.0000
0.0000
0.0118
0.0003
0.0019
0.0021
0.0066
0.0614
0.0016
0.0012
0.0455
0.0013
0.0010
0.0014
0.0090
0.0015
0.0010
0.0148
0.0046
0.0245
0.1236
0.0018
1.0000
0.0062
0.0023
0.0047
0.1204
0.0120
0.0013
0.0050
0.0170
0.0050
0.0030
0.0001
0.7701
0.0033
0.0118
0.0058
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0644
0.0134
0.0252
0.0112
0.1203
0.0185
0.0137
0.0194
0.8386
0.0474
0.0074
0.0040
0.0370
0.0036
0.0067
0.0080
0.0151
0.0576
0.0055
0.0046
0.2021
0.0038
0.0150
0.0168
0.0000"
0.0000
0.0000
0.0000
0.0020
0.0000
0.0000
0.0000
0.0006
0.0000
0.0000
0.0000
Equation 2: A values for use in the IIM given by the BEA
The IIM can be made into a dynamic model by applying a set of industry resilience coefficients,
K. A differential form of the dynamic form of IIM is shown in Equation 3. Estimates for these K
values can be obtained for an infrastructure from its self-referential dependence coefficient (A,,)
from the BEA and exponential recovery time to 99% operability (t.99) using Equation 4, where l
denotes time.
x(t + At) = K[Ax(t) + c(t) — x(t)] * At
Equation 3: Differential form of the dynamic IIM
^ _ -In (1-0.99) _ 4.605
' t99*(l-y4jj) t99*(l-y4jj)
Equation 4: K value estimation for the dynamic IIM
The dynamic IIM covers the bases of being able to model the infrastructure systems of interest
while capturing dynamic effects and interdependencies. The dynamic IIM is robust in the
3
-------
scenarios it can consider, as another paper used the model in a case study on the aftermath of
Hurricane Sandy [5], However, there are shortcomings of the dynamic IIM compared to other
models explored in this report. Implementing effects from people who are displaced is more
difficult to implement, and deviations from typical IIM methodology would need to be made.
Additionally, the model is deterministic in nature, yielding only one answer that is based on
point value estimates. This type of model does not naturally provide for any ability to estimate
variability of real-life scenarios. Distributions about the input parameters of the model could be
used to introduce variability, but this would be better handled in a stochastic method. In addition,
generating distributions for the input parameters would require additional data collection or the
use of less accurate estimates.
The Haimes book discusses other modeling efforts of SoS such as Hierarchical Coordinated
Bayesian Modeling (HCBM), Surrogate Worth Tradeoff Method (SWT), and Hierarchical
Holographic Modeling (HHM) [2], While each of these models have their merits in bringing
additional modeling benefits for other types of SoS, none of these provide a fully applicable
modeling technique for a dynamic set of interconnected infrastructure systems. HCBM
introduces the use of statistics to obtain likelihood and posterior probabilities of attacks and can
be used to make links between infrastructures via conditional probabilities. However, data
regarding the probabilities of attacks under a wide variety of conditions need to be obtained to
use the model. The SWT method provides for a way to model the recovery efforts of a system
where multiple objective functions are being optimized at the same time (e.g., repair costs,
safety, impact on environment). However, SWT is also not able to accommodate for the
dynamic system of multiple systems, but rather is focused on the recovery of one system with
multiple objectives. HHM is a graphical modeling approach that represents the dependencies of
interconnected infrastructures and represents the problem from multiple perspectives (e.g., the
sector perspective, the resource perspective). This approach creates an analysis diagram called a
preparedness diagram (Figure 1) [2], However, HHM does not provide a methodology to
quantify the dynamic behavior of the systems.
4
-------
DHS preparedness
V
FEMA and other federal
agencies
State and local emergency
response agencies
1
Other nongovernment
organizations, e.g. Red Cross,
volunteers
Region I: Boston
Maine, New Hampshire, Vermont,
Rhode Island, Connecticut,
Massachusetts
Region II: New York
New York, New Jersey, Puerto
Rico, Virgin Islands
Region X: Seattle
Alaska, Idaho, Oregon, Washington
Equipment
Materials
CoinriKHjilios
toT) •••
Rospondors
..... 1
n/iiaiQeftcy
rustical
bea vices
PubliQ WDlkeiS
Governmental
admin laialivt)
Hazardous
materials
porsonnol
Hoaltli caro
Public hoaltli
transportation 1
iyslonis
— m ' J:—r
Emergency
Services
Poiial and
uiippng
Dams
Govemmonl
ladMw!
Government
facilities
Figure 1: A DHS preparedness roadmap of interconnected infrastructures, an example perspective.
5
-------
Several other methodologies were explored in academic journals on the topic of interconnected
infrastructure systems. A series of papers by David Rehak et al. have outlined graph theory
methods that consider synergistic effects [6][7], Rehak's models use binary links between
different infrastructures to quantify activity coefficients (number of infrastructures that the sub-
infrastructure affects) and passivity coefficients (number of infrastructures that affect the sub-
infrastructure), with the sum of these being the significance of the sub-infrastructure. These
measures do not only serve as a good representation of the roles of each infrastructure in the
system but allow for the calculation of synergistic effects. Different stages of recovery are
defined by this model, but these are not used to specifically calculate dynamic effects (as the
activity and passivity coefficients are constant with time).
Another paper written by Lam et al. uses fuzzy set theory as an alternative to the binary
synergistic effect analysis [8], Both models require the same information about the
interdependencies of each infrastructure, but fuzzy set theory evaluates the weights of each of the
connections of the nodes using linear weighting and membership functions (similar to an
activation function in a neural network and the linear weighting in those). Fuzzy logic could
additionally serve as a useful tool for being able to prioritize goals that are hard to quantify in
infrastructure decontamination, such as minimizing environmental impact, social effects, public
opinions, and so on. While a powerful tool, no dynamic fuzzy set theory methods have been
developed that can specifically measure efficiency. However, fuzzy logic might be a useful tool
in the infrastructure prioritization portion of the model.
Weiping Wang et al. have developed a GIS based method (without the use of agent-based
modeling) that calculates cascading effects with a focus on the transportation sector [9], Their
method places attacks on the map with various attacking strategies (random, malicious, shell-
based, and oriented methods). The model uses a simple arithmetic method of incorporating
tolerance thresholds and capacities of infrastructures after infrastructure removals from the
graph, and offers some additional performance quantifiers. While this method can be transferable
to other sectors, this model is not readily usable for all U.S. Department of Homeland Security's
Critical Infrastructure Sector (DHSCIS) sectors.
An assortment of other papers collected during the literature review process did not contain
modeling methodologies but gave other relevant information to the field of interconnected
infrastructure (such as survey results of infrastructure operators, government surveys that
emphasize importance with respect to national security, and other literature reviews) [10][15], A
listing of all the papers regarding interconnected infrastructures that were observed, as well as a
tally of which infrastructure sectors were considered in each paper, can be found in Appendix A.
Because the models presented in the field of interconnected infrastructure systems presented
shortcomings, more general stochastic modeling strategies were explored. Stochastic models are
desirable for generating meaningful results because they can capture the inherent variability of
real-life behavior and allow for the use of statistics in making conclusions. Because of the
dynamic nature of interconnected infrastructures, stochastic time series models were studied.
Many time series depend on the selection of a stochastic process, such as Brownian motion,
6
-------
which randomly models the erratic movement of microscopic particles in a fluid [16], A number
of tools can be used to access stochastic time series, such as autocorrelation function, which
measures time series movement with respect to its movement history, and cross correlation
function, which measures the movement of other parallel time series.
A significant motivator in the field of stochastic time series modeling is stock market prediction,
which has led to the development of a variety of models such as the Black-Scholes Model and
Constant Elasticity of Variance (CEV) Model. As an example, the differential equation that
defines a stock price (which can be generalized to any stochastic time series) for the CEV Model
is shown in Equation 5 [17],
dSt = |iStdt + aSj dWt
Equation 5: Differential Equation defining the CEV Model
In the CEV Model, St is a stock price at time i, and Wt is a motion that introduces variability into
the model. Parameter |i characterizes the weight by which price drifts (up or down) control the
movement of St, while parameter o characterizes the weight by which volatility controls the
movement of St. Parameter y relates volatility of the movement with the size of the stock price
(when y > 1 volatility increases as price increases, and the opposite occurs when y < 1).
The CEV Model allows for the control of the time series by the selection of its parameters, which
theoretically could be tuned to model critical infrastructure. The issue, however, with the use of
the CEV Model and Black-Scholes Model for infrastructure modeling is that a history of the
movement of the time series is required for the model to be accurate. Additionally, the
relationships between specific infrastructure sectors might be difficult to enforce without
explicitly changing the form of the model (it could be done by providing that information in the
movement history of the time series, but this does not establish an explicit connection).
Outside the field of stock market prediction, another stochastic modeling strategy was found in
the modeling of coupled chemical reactions. The Gillespie algorithm was developed to
stochastically predict concentration profiles of multiple reacting agents [1], The Gillespie
algorithm calculates the rate of each reaction that is modeled, determines a random length of
each time-step based on the magnitude of the reaction rates, and probabilistically chooses one
reaction to occur at each time-step. Over a large set of stochastic runs, the result of the Gillespie
algorithm has been proven to average to the deterministic solution of the set of differential
equations that define the set of chemical reactions.
To model a set of interdependent infrastructures with the Gillespie algorithm, chemical species
would be replaced with infrastructure sectors. In this context, chemical reactions represent the
usage of other infrastructure resources to restore the services of an infrastructure sector. This
model eliminates the requirement of providing movement history of the relevant time series, as it
simply proceeds forward based on a set of defined reactions. A clear link between the
infrastructure sectors is defined by the selected chemical reactions, and the Gillespie algorithm
7
-------
can be readily modified with discrete effects (e.g., capping the production of certain
infrastructure resources in surplus) due to its time-step based nature. The Gillespie algorithm is
further detailed in Section 3.1.
2.3 Software Evaluation
In addition to evaluating a set of specific modeling methods in the literature, the feasibility of
existing software to address the same tasks listed in the desirable modeling criteria was
evaluated. The possibility of using a pre-developed program mitigates the need to set up a
modeling scheme from scratch. However, using another modeling package does not provide as
much flexibility to change the conditions of the model.
In the Haimes book, an HHM program, Adaptive MultiPlayer Hierarchical Holographic Model,
was developed on Microsoft's Groove environment and adopted in a large parallel computing
session at a workshop [2], This program could observe 280 threat scenarios based on outages in
sets of infrastructure sectors or materials, which could then be separately analyzed for risk.
While the program is useful for responders to prepare for different scenarios and determine
likelihoods of scenarios through separate risk analysis methods, this method cannot capture
system dynamics and is not readily available to the public.
In Bhamidipati et al., a GIS analysis was used in agent-based model (ABM) programs such as
Repast and Netlogo [20], Both programs can model a CBRN incident as an autonomous agent
and dynamically observe the interdependencies of the infrastructure. CBRN incidents are
identified on a GIS map by placing layers that indicate the geographical spread of the agent.
Depending on the program, ABMs can even consider additional geographical factors that result
in the displacement of individuals, such as weather conditions. ABMs use a computational
method to simulate a system as an assemblage of entities, or agents, that are programmed to
follow specific (simple or complex) rules. Projects in these ABM programs can be written to
observe all DHSCIS infrastructure sectors and are readily available online. However, the
software is difficult to use, and any simulation tactics need to be implemented by hand within
these programs.
A similar ABM program that was developed to analyze infrastructures is Aspen-EE [21], This
program has a focus on observing the sequence of events that occur during a CBRN incident and
then specifically looking at the energy sector. This program has the dynamic capabilities similar
to those of the other two ABM programs but does not have the capability of looking at all desired
infrastructure sectors, and is no longer actively supported by Sandia National Laboratories.
Autodesk has developed a GIS package focused on infrastructure called AutoCAD Map3D and
AutoCAD Raster. This program is similar to other GIS tools (such as QGIS) but contains built-in
tools for infrastructure analysis for all sectors on the DHSCIS list. AutoCAD is also able to
implement building information modeling in parallel with GIS. Some simulation capabilities are
included in the package using add-ins, but they might not be able to accommodate
8
-------
interconnected infrastructure systems specifically. While AutoCAD's GIS tools might be robust
analysis platform, the AutoCAD license is expensive and not readily available to all clients.
A large set of infrastructure analysis programs have a specific focus on the information
technology sector and cyber-attacks. These programs fall into a separate field of study called
threat modeling. They contain many graph theory-based analysis methods that range in scope
from an organization network to a nationwide cyber-attack, and many tools express dynamic
information that quantifies risk and recovery timelines. Some of these methods are elaborated in
the literature, such as adjacency matrices, betweenness centrality, efficiency weightings, and
efficiency weighted node degree [22], However, these programs do not cover all the necessary
infrastructure sectors.
An ABM critical infrastructure modeling software program, PATH/AW ARE, is split up into two
analysis tools, PATH (standing for Prioritization Analysis Tool for All- Hazards) and AWARE
(standing for Analyzer for Wide Area Effectiveness). PATH is focused on prioritization of
recovery efforts and observing interconnectedness, while AWARE is focused on estimating
recovery timelines and costs. Although neither of these tools run a dynamic simulation of the
scenario, they can statically calculate outputs such as recovery timelines, restoration costs, and
prioritization schedules from the assets affected in the scenario. The U.S. Department of
Homeland Security's Wide Area Recovery and Resiliency Program (WARRP) is a collaborative
program designed to reduce time and resources required to recover large urban areas, military
installations, and associated critical infrastructures following a catastrophic chemical, biological,
and radiological incident. The PATH/AW ARE work employed some scenarios developed under
WARRP. HAZUS is a nationally standardized risk modeling methodology. It is distributed as
free GIS-based desktop software with a collection of inventory databases for every U.S. state and
territory; the database served as a resource to the PATH/AW ARE work.
PATH/AW ARE works by placing contaminated areas on a GIS map interface (with an
associated HAZUS database containing data for other portions of the analysis). These areas can
either be red, yellow, or green to represent highly, moderately, and insignificantly contaminated
areas, respectively. The tool can then find all the infrastructure assets that are within the
contaminated areas and apply interdependencies recursively. An example screenshot of a
scenario set up on PATH/AW ARE showing all relevant assets for the two hypothetical
biological scenarios in Denver, Colorado, created by WARRP, can be seen in Figure 2. Once all
affected assets are gathered, PATH/AW ARE can run its analysis tools.
9
-------
Map
Google Map Type * Layer Detail:
m + vm Preferences legend Export * Add Map to Summary Presentation
flj|§(VER ARTCC 1
MiA
""•ft lit on
'^©tORADO ANTIQUE FHO &N»$526 l I^EID RANCHES
^S/INGS N THINGS AIRPARK®rtUSEUM m ™ — «EID RANCHES
9)(NKA348—. 'J'
r-, re; EH Parkland m
S WOUUDER f^UCKLA FARMS
g- MoulderIffljNi , ^indys airpark
1 'l^ND
^HE^ARM
^Westbei
n
O PuckelU|ARC^
SPfTAL $NKA34fi
tib^co
L-SlRINO
4-. BRIGHT 01
I^ECKENDORF
® ^EASONER
"^RIG^ON \fciN-AIRE ESTATES
RANCHES lr5US
'^Ithanasiou VAILB*)
" Miners mesa
L ^ B
. Vnka2?o
^RAPAHOtpTARMlGAN
M ^APRI
^ -IS Hulk ma^ti
^Ireggs nr 1
A Denver intl Iogan
ife Comanche l
@ IBM-.
TOGOfflkc
^oAJES RUBBERED
P Martinez USARC/AMSA #100
NTER
&NKA220
J» ft
"^RONT R
^OUNT MORRISON
Keystone ^au. timber
M Youngs sAip
IRTER MEMOaajL hospital
diCTtci
SWEDISH MEDICTlTCENTER
BTC NORTH fa
" tiABSHOfll.E^'W'ORM^USE g
J$EYER RANCH '%)WEN FARMS NR 1 ?ARKER Al
%NKA348 , 'tfVERITT
4*ART1N
Harrington ranch
NTENNIAL HEALTHCARE PLAZA
HOSPITAL
&NKR202
*CH.Z rmi&n
0YeSo«» Zone
*
Pi Red Zone
1 1 Ranedalloo Uh*
0CnScal Asset
0Cniica( Asset Text
0NotvCnbcal Buking
BNon-Oteal Buiking Tod
3 0 OicdWrMtnjclue;
0Afrport
0Antrak
BBndpe
0Cefliiar Tower
0 Commercial
0 Ferny
- 0FVe
0Fuei Stated
0Goveirn>ent Buldng
0 Heliport
0Hghway
0 Hospital
0JoM Fence Headquarter
1 |
0taurwy
0Mttay Base ar
0Mktav hatalaOorc
0Msc Ca^e Factty
0MjctearFuei
j
0Phaniacy
0P«lee
0Pon
0P«afc»e Water Treaimen
0 Power
0 Pmnary Care One
0 Public Hsalth One
0R#ilyard
.
0 School
0Tranat Lrfc
0Was!e Water Treatmert
0WMDGvl Support
£71 HnHW*
<
> 1
1 1 izr Citcal Baking Dap lay
Usage
J
Figure 2: Denver scenarios mapped in PATH/AWARE.
PATH is a tool that specifically focuses on the prioritization schedule recovery efforts should
follow based on a set of goals. The user can set weights for different objective goals (such as
maintaining the economy, specifically focusing on bringing certain sectors back online,
minimizing environmental impact) (Figure 3). PATH also considers interconnected sectors in its
prioritization efforts, which provides a visual representation of the interconnections of
infrastructure assets in a dependency graph (Figure 4). Individual sectors can have their
functionality level, dependencies, and services enabled by the sector viewed and edited in PATH
(Figure 5).
10
-------
ie) Maintain Economy
Commerce
Retail services
Industrial services
Tourism services
Postal and shipping
Postal service
Courier service
~ Banking and finance
Transportation
[ A Agriculture and Food
± Minimize Environmental Impact
Maintain Public Safety
Maintain Public Health
Maintain National Security
Maintain Continuity of Operations
Low Q=
Hiqh
Hiqh
Hiqh
Hiqh
Hiqh
Low Q=
Low
Low Q=
Hiqh
Low Q=
Low Q=
Low Q=
Low Q=
Low Q=
Hiqh
Hiqh
Hiqh
Hiqh
Hiqh
_ Hiqh
Low Q J Hiqh
Low P Hiqh
Figure 3: User inputs to PATH: prioritization weights.
Electricity
transmission and
distribution
Temporary shelter
Water treatment
Permanent shelter
Blood, organ and
tissue services
Military training
Wastewater
treatment
General Medical Care
Daycare
End of life services
K-12
Public road transport
Higher education
Rail transport:
Passengers
Inmate housing
Figure 4: Truncated portion of the infrastructure dependency diagram in PATH.
11
-------
PATH
Export Recalculate Recalculate Service Status Change High-Level View
Set High-Level Priorities Input Sen/ices Data Input Asset Data Output: Service Prioritization Output: Asset Prioritization Dependency Viewer
Service Name ^
Function Name
Metric
Current Status
Desired Status
Milestone Day
Agricultural and food di...
Agriculture and Food
Area served
Fully functional
Fully functional
0
Agricultural and food re...
Agriculture and Food
Employees
Fully functional
Fully functional
0
Agricultural and food st...
Agriculture and Food
Capacity
Fully functional
Fully functional
o
Agricultural and food tr...
Agriculture and Food
Throughput capacity
Fully functional
Fully functional
0
Air traffic control
Transportation
Throughput capacity
Fully functional
Fully functional
0
~
Air transport: Goods
T ransportation
Throughput capacity
Fully functional
Fully functional
0
Air transport: Passengers
Transportation
Throughput capacity
Fully functional
Fully functional
0
Blood, organ and tissue...
Health Care
Procedure rate
Fully functional
Fully functional
0
Cable broadcasting
T elecommunications
Population served
Fully functional
Fully functional
0
Commercial banking
Banking and finance
Capacity
Fully functional
Fully functional
0
Commercial lodging
Shelter
Capacity
Fully functional
Fully functional
0
Courier service
Postal and shipping
Delivery rate
Fully functional
Fully functional
0
Credit services
Banking and finance
Capacity
Fully functional
Fully functional
0
Crude oil pipelines
Energy
Throughput capacity
Fully functional
Fully functional
0
Crude oil storage
Energy
Capacity
Fully functional
Fully functional
0
Daycare
Education and child care
Capacity
Fully functional
Fully functional
0
Electricity transmission ...
Energy
Buildings served
Fully functional
Fully functional
0
Emergency Management
Emergency Services
Num of emergency ma...
Fully functional
Fully functional
0
Emergency Medical Ser...
Emergency Services
Capacity
Fully functional
Fully functional
0
End of life services
Health Care
Cadaver Storage capab...
Fully functional
Fully functional
0
Extended care
Health Care
Num of beds
Fully functional
Fully functional
0
Firefighting
Emergency Services
Num Fire Trucks
Fully functional
Fully functional
0
Fossil fuel electric pow...
Energy
Voltage
Fully functional
Fully functional
0
General Medical Care
Health Care
Num of peak staffed beds
Fully functional
Fully functional
0
Government sensor an...
Government services
Information transfer rate
Fully functional
Fully functional
0
Government storage an...
Government services
Capacity
Fully functional
Fully functional
0
Hazardous chemical re...
Hazardous Materials
Employees
Fully functional
Fully functional
0
Hazardous chemical sto...
Hazardous Materials
Capacity
Fully functional
Fully functional
0
Hazardous chemical tra...
Hazardous Materials
Throughput capacity
Fully functional
Fully functional
0
r
Hoalth nrartitinnor
Hpalth Cr\rp
Patient ranaritv
Pnllv fnnr+irmal
Fnllv fnnrtinnal
nl
n
Service Dependencies
0
| Edit Service Dependencies |
Air traffic control
Petroleum storage
Service Enables
01
Industrial services
Courier service
Military deployment
Service Test
0
a*. 9. n
m a
& a m
Figure 5: Example sector functionality and dependency inputs.
12
-------
Information was input into the PATH/AW ARE for the example scenarios displayed in Figure 2; however, full results could not be
produced since the PATH/AW ARE installation only had access to information in the HAZUS database and not the ThinkGeo map
data, which is required to produce full results. PATH results for the Denver scenarios can be found in Figure 6, however the results are
incomplete as all assets are given a prioritization score of zero due to the incomplete user input provided to the program. The software
attempts to prioritize assets, but generates a score of zero for all assets, and subsequently produces a meaningless result.
Figure 6: PATH assets results for Denver scenarios 1 and 2.
PATH
|-a-||-B \\&m\
1
Export Recalculate Recalculate Service Status Change High-Level View
:
Set High-Level Priorities Input Services Data Input Asset Data Output: Service Prioritization Output: Asset Prioritization Dependency Viewer
Priority Asset Name
Infrastructure Category Name
Reason
Restoration S...
Obje...
Dependency Score
1 William T. Fitzsimons USARC
Military Base Str
Contribution To Priorities
0
0
0
2 NMCRC Denver
Military Base Str
Contribution To Priorities
0
0
0
3 ST JOSEPH HOSPITAL
Heliport
Contribution To Priorities
0
0
0
4 KNKA220
Cellular Tovsrers
Contribution To Priorities
0
0
0,
5 Buckley AFB
Military Base Str
Contribution To Priorities
0
0
0
6 BUCKLEY AIR FORCE B
WMD Civil Support
Contribution To Priorities
0
0
0
7 BUCKLEY AFB
Airport
Contribution To Priorities
0
0
0
8 SIMONS
Airport
Contribution To Priorities
0
0
0,
9 THE CHILDRENS HOSPITAL
Heliport
Contribution To Priorities
0
0
0
10 PRESBYTERIAN/ST LUKES MED CTR
Heliport
Contribution To Priorities
0
0
o|
~
Contribution To Priorities
0
0
0
12 VTOL
Heliport
Contribution To Priorities
0
0
0
13 UNIVERSITY HOSPITAL
Heliport
Contribution To Priorities
0
0
0
14 ST LUKES HOSPITAL
Heliport
Contribution To Priorities
0
0
0
15 ROSE MEDICAL CENTER
Heliport
Contribution To Priorities
0
0
0
16 POLICE HEADQUARTERS
Heliport
Contribution To Priorities
0
0
0
13
-------
AWARE is a tool that calculates restoration costs and timelines based on the assets affected by the defined CBRN incident and a set
user inputs that define the recovery effort. Parameters, such as the number of people on restoration teams, labor costs, and sample
collection rate, can be adjusted as shown in Figure 7, which are used to estimate the costs and timelines in the AWARE results.
AWARE
General
Scenario
Zones
Building Infiltration
Resources
Sampling
Lab
Screening
Outdoor
Indoor
Characterization
Outdoor
Indoor
Summary
Waste
S
I Sampling Resources
Outdoor
Indoor
~econ.'Clea ranee
Transport
Summary
Labor rate:
Number of teams:
Sampling labor cost:
Each team collects
Total sampling rate:
Sampling Resources | Outdoor Indoor Decon/Clearance Transport Summary
Waste Sampling Resources
Team size: 2 % persons perteam
Working day length: 8 hours
S/hr per person
teams
50
5:
4.000
dollars per day
2C Z samples per hour
300 samples per day
Figure 7: Subset of user inputs to AWARE for cost and timeline estimation.
14
-------
In a completely functional version of PATH/ AW ARE, the user would be able to split red,
yellow, and green zones into any number of remediation units in which a team works to restore
services. Because of the incomplete installation state of PATH/AW ARE, the current version of
PATH/AW ARE cannot create new remediation units, as the software disables this option.
PATH/AW ARE documentation suggests creating remediation units that are about 4-6 blocks
long in both street directions, a feat that is unattainable with the current condition of the
software. As a result, the AWARE results are calculated as if teams are required to work over the
entire contaminated area. Nevertheless, the AWARE results for the Denver scenarios detailed in
Section 4 are presented in Table 1.
Table 1: AWARE Results for Denver Biological Contamination
Scenarios: Cost and Timeline Estimates
Recovery Stage
Duration (days)
Cost
S creening/C haracterizati on
963
$42,346,400.00
Waste Removal
399
$7,980,000.00
Decontamination
1994
$39,880,000.00
Infrastructure Reuse Clearance
631
$27,749,500.00
Overall, PATH/AW ARE has some powerful capabilities for predicting recovery costs and
timelines after CBRN incidents and provides tools to prioritize some infrastructure assets over
others. These results are derived from parameters relating directly to the decontamination effort
and gives an overall picture of how the response to an incident would likely occur. However,
only one value is reported for every parameter in the PATH/AW ARE, as the PATH/AW ARE
cannot account for any unexpected circumstances where the results might not correlate with the
calculated estimates. Additionally, PATH/AW ARE is only able to assess the incident statically,
based on the user inputs, and does not account for dynamic changes. PATH/AW ARE also does
not account for components that might be numerically coupled with other components that
would need to be accounted for in a real incident, including waste and decontamination. Finally,
the software needs to be updated and maintained to use it on the currently available computing
environment.
2.4 Conclusions from Literature Review
Many of the models and software presented in the literature review were not considered further,
since they had shortcomings related to the inability to consider all DHSCIS sectors, a static
modeling methodology, or ease of use. Looking past issues with the software regarding missing
mapping data and the various software bugs, PATH/AW ARE could be a useful tool for
quantifying recovery efforts from CBRN incidents since it can consider the interdependencies of
critical infrastructure. The PATH portion of PATH/ AW ARE provides a valuable set of
infrastructure dependency mappings that can be used in setting up other interconnected
15
-------
infrastructure models. However, PATH/AW ARE does not meet all the desired model
characteristics set out in Section 2. Although the tool can calculate restoration times, this
functionality is not dynamically modeling the infrastructures, but is simply doing a static
calculation based on the user inputs and remediation units. This makes PATH/AW ARE more of
a relevant calculation tool as opposed to an actual modeling software. While it does consider all
sectors in the DHSCIS list and their dependencies, there is no consideration for the effects of
people becoming displaced by the CBRN incident, potentially a large factor that affects the
workforce's ability to restore services. Additionally, PATH/AW ARE is a deterministic tool that
is based on user inputs. The user would need to provide distributions around the inputs to
properly capture the variability of effects from the CBRN incident. Since PATH/AW ARE does
not have a scripting environment to allow consecutive calls to PATH or AWARE to be made
using a set of distributions in the user inputs, PATH/AW ARE is only able to provide manually
generated results, which limits its ability to run batches of simulation and generate distributions
of results.
Using the Gillespie algorithm to stochastically model the infrastructures as a time series proved
to be the most promising solution. The Gillespie algorithm was specifically chosen as it does not
have the requirement of providing a time series movement history like other stock market
prediction models. Additionally, using the Gillespie algorithm has the benefit of flexibility in the
types of infrastructure and dynamics that can be modeled, and it can use its stochastic modeling
to draw powerful conclusions.
Although the other modeling programs in the literature were not chosen for further development,
other aspects can be integrated into the chosen model. For example, PATH/AW ARE contains a
set of infrastructure dependencies and a set of infrastructure prioritization weightings that could
be useful in choosing the parameters for other mathematical models. The details of the final
model are presented in the next section.
3 STOCHASTIC INFRASTRUCTURE REMEDIATION MODEL
(SIRM)
The Stochastic Infrastructure Remediation Model (SIRM) was developed to specifically address
the desired model characteristics in Section 1 and the shortcomings of the mathematical models
discussed in Section 2. Its mechanics are based on the Gillespie algorithm of stochastically
modeling chemical kinetic systems, with additional changes made to adapt to infrastructure
modeling. The adaptation process of the Gillespie algorithm, the mechanics of the final model in
its current version, and a preliminary set of results for example scenarios are all documented
below.
3.1 Introduction
To dynamically model the efficiencies of many interconnected infrastructures in the event of an
infrastructure outage, a new mathematical model was developed based on a set of chemical
kinetic reactions using the Gillespie algorithm. The Gillespie algorithm was originally developed
16
-------
to stochastically model coupled chemical kinetic reactions, but can be used to model any set of
interacting systems of the form A + B—>C + D[1], This algorithm employs a method for
stochastically solving differential equations by 1) randomly generating time-steps based on the
overall rate of interactions, 2) randomly selecting an interaction to occur during the time-step,
and 3) executing the interaction according to the interaction dependencies. These steps are
repeated for thousands of iterations, until either a given simulation time is reached or no further
interactions are possible.
This structure was set up to model the use of resources in other infrastructure sectors to restore
services in sectors with an outage. As opposed to modeling the concentration of a chemical
involved in a set of series and parallel reactions, SIRM was developed to keep track of the
efficiency percentage of infrastructure sectors, where efficiency is defined as the sector's
capability to deliver the services it is designed to provide to society. As a result, SIRM takes a
conceptual network of interconnected infrastructure and applies mathematical principles to
translate the connections into a numerical result.
This model structure can capture the interconnectivity of each infrastructure, while also naturally
providing realistic bounds on the infrastructure efficiencies and prioritizing the use and
remediation of infrastructure based on the current state of the system. As SIRM was developed,
deviations from the basic kinetic model were made to more accurately reflect the dynamics of
the response that are specific to infrastructure systems.
As an initial effort, both a more simplistic deterministic model, which used Python differential
equation solvers such as odeint, and a stochastic model, which used the Gillespie algorithm of
stochastic chemical kinetics modeling, were developed [23], The deterministic model was
examined at first due to simplicity and potential accuracy. Each model considered six
infrastructure sectors and could accommodate any number of infrastructure outages at any time.
Both the stochastic and the deterministic models examined the interactions of each of the
infrastructures as a system of differential equations, which is based on a system of chemical
reactions, each with a first order rate constant, defined by the number of parent infrastructures.
These reactions were formed using the PATH/AW ARE dependencies defined by the tool, and
are notional equations that were part of a proof of concept model. The number of parent
connections defined the reactants, while the product is the infrastructure sector in question. The
underlying reaction equations that define the deterministic and stochastic solution is shown in
Table 2. Abbreviated symbols for each infrastructure sector are water (W), energy (E), transport
services (T), communications services (C), government facilities (G), information technology (I),
healthcare (H), food/agriculture (F), emergency services (S), and financial/banking (B).
17
-------
Table 2: Preliminary Set of Reactions Used in Early Versions of the SIRM
Reaction
1
kl = 0.000008
E + T —» 2W
2
k2 = 0.0000005
C + G + I —» 3E
3
k3 = 0.00000007
C + E + W -»¦ 3T
4
k4 = 0.01
E —» C
5
k5 = 0.0000000012
C + T + W + E—>• 4G
6
k6 = 0.000025
E + C —~ 21
A side by side comparison of time profiles of the deterministic and stochastic solution for a
sample scenario is depicted in Figure 8, where all the infrastructure efficiencies start at 100%
except for Electric/Energy which starts at 50%.
water
electric
transport
communications
government
rr
Deterministic Solution: Time Profiles Gillespie Solution: Time Profiles
t (days)
100 -
0 50 100 150 200 250
t (days)
Figure 8: Side by side comparison of deterministic and stochastic models in
early versions of the SIRM.
After reviewing the properties and behaviors of each model, the stochastic model was selected
due to its desirable ability to incorporate discrete effects (e.g., rates of repair). The stochastic
model maintained reasonable accuracy to the deterministic solution and provided the ability to
generate distributions from many data points and make statistical conclusions. Due to the
discrete nature of the stochastic calculation, discrete effects such as the removal of stockpiling
effects were also allowed to be applied in the stochastic solution at each time-step (as the Python
differential equation solvers like odeint are more of a black box and cannot be as readily
18
-------
adjusted). The powerful ability to apply discrete effects and make statistical conclusions was
substantial justification to choose the stochastic solution over the deterministic solution.
3.2 Infrastructure Sectors
Selected infrastructure sectors from the DHSCIS list were prioritized for early development
efforts of SIRM using data from the literature review of academic journals and books related to
critical infrastructure modeling. The number of papers in which each infrastructure sector was
analyzed was tabulated and ordered based on prevalence in the literature. Initial model
development efforts with SIRM used the nine sectors shown in Table 3, with waste management
added based on SME feedback.
Table 3: Infrastructure Sectors Selected in the Preliminary Development of the SIRM
Infrastructure Sector
Number of Papers (out of 16)
Energy
15 (94%)
Water and Wastewater Systems
14 (88%)
Transportation Systems
13 (81%)
Communications
11 (69%)
Government Facilities
9 (56%)
Healthcare and Public Health
9 (56%)
Emergency Services
7 (44%)
Food and Agriculture
7 (44%)
Waste Management
-
3.3 Model Methodology
SIRM is based on the stochastic process of the Gillespie algorithm and consists of the following
steps:
1. From the conditions at the previous time-step (or the initial efficiency conditions
after the incident), calculate the rate of each reaction, or interaction between
sectors. A "reaction" involves one infrastructure child sector, as well as all of
parent infrastructures that the child sector relies on. These are used in determining
time-step length and probability of the reaction occurring at the time-step.
2. Add up the rates of each reaction, naming it rtotai.
3. Generate a random number between 0 and 1 to determine the length of the next
time-step (t):
t = -ln(random number)/rt0tai.
19
-------
4. Calculate the probability for each reaction to occur at the current time-step (p = rate
of given reaction/rtotai).
5. Randomly choose a reaction to occur based on the calculated probabilities of each
reaction.
6. Find v, the vector of efficiency value changes for each reaction based on which
reaction was selected in Step 5. This is determined by the strength of the
connection between the parent infrastructures and the child infrastructure sector.
7. Update the efficiency values by adding the v vector to the efficiency values from
the previous time-step.
8. Repeat until simulation ends (either when rtotai = 0, t exceeds the defined time span,
or a defined maximum number of iterations of the algorithm have been completed).
Other changes were made to SIRM to address more of the desirable model characteristics. To
account for more functionality, more input parameters were added to SIRM. The input
parameters can either be entered in a graphical user interface (GUI) or an input text file. The
input values are not finalized but are used to evaluate and test early versions of the SIRM. The
input values can be modified and expanded upon in future iterations. The input parameters are
defined as follows:
• Initial infrastructure sector efficiencies (no): the initial value of the efficiency values of
each sector in percent;
• Repair factors RF): the base value of the repair factor for each infrastructure sector. It
represents the rate society works to repair the infrastructure sector. A higher repair factor
implies that society is working harder to repair that sector;
• Amounts of additional infrastructure outages (nLoss): the percent reduction of each
infrastructure sector in a second outage at t = tLoSS;
• Time of additional infrastructures outages (tLoss): the time at which additional
infrastructure outages occur, if applicable;
• Number of stochastic runs (nRUn): the number of times the defined scenario is
stochastically simulated;
• Maximum number of reactions (nMax): the maximum number of single reactions that can
occur in a simulation (usually set arbitrarily high);
• Simulation length (timeSpan): the number of simulated days in each scenario
• Parameters to be collected (paramTypes): the type of parameter to be collected in each
defined histogram (min, max, average, recovery time to 100% efficiency after minimum
value, and value at the end of the simulation);
20
-------
• Infrastructure indexes of parameters (paramlndexes): the infrastructure sector of which
each defined parameter is to be collected;
• Infrastructure stoichiometric factor (infStoichFactor): the value by which each
stoichiometric coefficient in the reaction equations is multiplied. This varies the level of
stochastic behavior, as well as the stability of the data; and
• Seed Value (seedValue): a value of the seed for the random number generator if it is
desired for a set of data to be reproducible.
The foundation of the SIRM is similar to many of the other mathematical models and approaches
considered in this report. It takes a conceptual network of interconnected infrastructure and
applies mathematical principles to translate the connections into a numerical result. A
visualization of the network for all DHSCIS sectors is depicted in Figure 9. Due to the nature of
the infrastructure reactions in the SIRM, the connections are best represented as one-way arrows,
as infrastructures are used to directly contribute to the increase in efficiency of other
infrastructures. Bidirectional interactions are possible (and present in the current diagram) and
are represented by two arrows in opposite directions connecting the same infrastructures. The
thickness of the arrows represents the number of connections between each of the infrastructures
- a thicker arrow implies that the infrastructures are more tightly linked.
Facilities
Transportation
Systems
Communications
--
and Public
Health
Waste Management
Figure 9: A visual representation of the preliminary infrastructure dependencies in the system.
The infrastructure reactions, which are a representation of infrastructure connections defined by
PATH/AW ARE, are listed in Table 4 along with the associated rate equations and rate constants,
all of which are first order with respect to each reactant. Each reaction rate is directly
proportional to all contributing infrastructures and is divided by the infrastructure sector that is
21
-------
produced by the reaction to add prioritization to infrastructures that are operating at a low
efficiency. Rate constants were normalized to be i/(ioonumberofparentmfrastructuresectors), as opposed to
a hand tuned set of values. This makes the rather arbitrary choice of rate constants more
defensible and equal among infrastructure sectors, as the rates of every reaction, in the case
where all infrastructure sectors are at 100% efficiency, is exactly 1%/day.
Additionally, repair factors (RFs) are added to infrastructure reactions to model the rate at which
society works to restore infrastructure services. The user inputs a set of base RF values for each
infrastructure sector (which can be used to analyze the effects of resource allocation). However,
the rate at which infrastructures are able to repair is not a constant but could be based on other
factors such as decontamination activities and the proportion of the population available to work.
Base RF values are subsequently adjusted in Equation 6 by the infrastructure stoichiometric
factor (the value by which each stoichiometric coefficient in the equations is multiplied) and the
ratio of healthy people that can go to work to total people, obtaining the adjusted repair factor
(,ARF).
a n r? _ nr? total population
ikixr| — rj'i
total population x infStoich
Equation 6: Calculation of Adjusted Repair Factors in the SIRM
22
-------
Table 4: System of Reactions Currently Implemented in the SIRM
Infrastructure Reaction*
Rate (%/day)
Rate constant (l/(100number
of parent infrastructure sectors^
0. E + T —~ (2+ARFw )W
£o = 0.01
1. C + G + I -> (3+^i?F£-)E
ki = 0.001
2. C + E + W + B—~ (4+ARFt )T
k2 = 0.00001
3. E —~ (l+ARFc )C
II
p
4. C + T + W + E + S—~ (5+ARFg)G
k4 = 0.00001
5. E + C —~ (2+ARFi)l
k5 = 0.01
6. E + S + B + F—~ (4+ARFh)K
h = o.oooi
7. W + T + B —~ (3+ARFf )F
ki = 0.001
8. C + G + T + I—~ (4+^i?F^)S
>^8 = 0.0001
9. E + I -> (2+ARFb)B
r9 = (ySr9*E*I)/B
£9 = 0.01
*B, financial/banking; C; communications services; E, energy; F, food/agriculture; G, government facilities; H,
healthcare; I, information technology; S. emergency services; T, transport services; W. water
3.4 Model Assumptions
Taking into consideration the methodologies of the SIRM, several simplifying assumptions were
made and are listed below.
• Infrastructure sectors recover kinetically with the use of other infrastructure sectors.
• It is a closed system, so there was no input from outside resources.
• Infrastructure sectors are either completely connected or not connected at all.
• Infrastructure connections are only one way unless another connection is defined in
another reaction making a connection in the other direction.
23
-------
• Repair factors are independent of each other, and of the other critical infrastructure.
• In the absence of repair factors and operating under normal operating conditions, all
infrastructure sectors are used and restored at the same rate of 1%/day.
3.5 Adjustments of Dependency Map
After determining an initial dependency map, in order to improve SIRM, some additional
literature searches and subject matter interactions were performed. The process and the
adjustments are summarized in the following sections.
3.5.1 Literature Search and Subject Matter Expert Interactions
An additional literature search was performed to identify any infrastructure connections that
were not included in the original dependency map in Figure 9. Seven historical documents [32]
[33][46-50] and several other additional papers of interest were identified [51-64], Each paper
performed a critical infrastructure assessment of either a historical or hypothetical event. The
sources were examined to determine the critical infrastructure sector dependencies listed, and the
total of all dependencies across all sources was tabulated to form a matrix of parent and child
infrastructure sectors. A summary of the results of this analysis are displayed in Table 5. The
child infrastructure sectors are listed across the top, and the numbers in the columns indicate the
number of sources where that child infrastructure sector is cited as being dependent upon each
parent sector (listed in the leftmost column). Red squares indicate connections that were not in
the PATH/AW ARE based dependency map, and green squares indicate existing connections.
24
-------
Table 5: Infrastructure Connections Mentioned in Literature Search
Child (number of sources where connection is
mentioned)
Parent
W
E
T
C
G
A
S
z
H
Water and Wastewater Systems (W)
5
4
3
1
1
1
1
Energy (E)
4
6
6
2
1
2
Transportation Systems (T)
1
2
2
1
1
Communications (C)
2
3
1
1
1
Government Facilities (G)
Food and Agriculture (A)
1
Emergency Services (S)
1
Waste Management (Z)
Healthcare (H)
1
The results indicate that several of the infrastructures, namely water, energy, transportation, and
communications are heavily connected and well-studied in both the PATH/AW ARE dataset and
in the literature. However, the interdependence of government, agriculture, emergency services,
waste management, and healthcare are less studied in the literature, and less heavily connected in
PATH/AWARE.
In addition to connections mentioned in the literature, several more connections were proposed
through conversations with subject matter experts (SMEs). Food and Agriculture was dependent
on both Water and Waste Management, while Waste Management was dependent on
Communication. The project team also worked with DHS Cybersecurity and Infrastructure
Security Agency (CISA) to evaluate SIRM using a preexisting critical infrastructure dataset and
made several coefficient adjustments. A revised version of the infrastructure diagram is
25
-------
displayed in Figure 10. Orange arrows indicate new connections based on the literature, while
blue arrows indicate connections proposed by SMEs that were felt to be missing in the original
mapping. The thickness of the white arrow indicates the strength (number of connections in
Table 5) of the initial connection.
Energy
Transportation
Systems
Water and
Wastewater System"
lealthcare and Public
Health
Food and Agriculture
Communications
Waste Management
^Government Facilities
Emergency Service
Figure 10: Revised infrastructure diagram.
3.5.2 Coefficient Modification
The coefficients for each of the infrastructures was adjusted based on the results of the literature
search and input from a DHS analysis. The coefficient was directly proportional to both the
number of articles found in the literature search that mentioned the connection as well as the
number of times PATH/AWARE mentioned the connection. This way, connections that were
more prevalent in the literature were quantified and treated as "stronger" by the SIRM itself. The
rate constant was calculated in the same way as defined by Section 3.1. The updated system of
equations is listed in Table 6.
26
-------
Table 6: Infrastructure Equations
Infrastructure Reaction*
Rate (%/day)
Rate constant
(1/(10onumbeipa, tnt
infrastructure sectors^ j
E+ T + C + Z —~ (4+ARFW )W
^ro = 0.0001
W + T + C + G—~ (.A+ARFe)E
ki = 0.0001
W + E + C —~ (3+ARFt )T
^2 = 0.001
W + E + T —~ (3 +ARFC )C
h = o.ooi
W + E + T + C + A—~ (5+^RFg)G
h = o.oooooi
E + T + G —~ (3+^i?F.4)A
h = o.ooi
W + E + T + C—~ (4+,4i?Fs)S
^6 = 0.0001
W + E + T —~ (3 +ARFZ )Z
ki = 0.001
W + E + T + C + G—~ (5+ARFtf)H
n = (fo*W*E*T*C*G)/H
>^8 = 0.00001
*A, food and agriculture; C, communications; E, energy; G, government facilities;
H, healthcare; S, emergency
services; T, transportation systems; W. water and wastewater systems; Z, waste management
3.6 Decontamination Factor
SIRM examines overall operating efficiency of each infrastructure sector. This overall operating
efficiency can be impacted by both reductions from the contamination incident and reductions
due to other infrastructure sectors operating at a reduced efficiency. The initial operating
efficiencies are based on the contamination incident. For example, if 20% of the population in an
area does not have cell service, then the initial overall operating efficiency of the associated
sector might be 80%. This could be reduced further throughout the simulation if other sectors are
operating below 100%. Because the process for repairing infrastructure differs depending on the
reason for the decreased operating efficiency, different types of repair factors are used in SIRM.
To address this, SIRM includes a "contamination" input to each infrastructure. This is a separate
percentage input, strictly described by how contaminated an infrastructure was, and improved
27
-------
solely dependent on time. Using the previous scenario, consider that one out of five cell towers
has been contaminated. The total percent contaminated would be 20%.
The contamination is currently reduced linearly with time, at a percentage per day, using the
decontamination factor. For example, in the previous scenario, if a cell tower is 10,000 square
feet and can be decontaminated at a rate of 1,000 feet per day, then the decontamination factor
(rate of decontamination) is (1000/[ 10000*5]) = 0.5%/day. The "decontaminated" percentage
also serves as a ceiling for the percent efficiency of the infrastructure-if a sector is 30%
contaminated, then the highest efficiency it can operate at is 70%.
4 GIS BASED ANALYSIS OF DENVER SCENARIOS
For the results of SIRM to be meaningful, the user inputs need to be mapped to the real-life
scenarios that are being modeled. Because SIRM specifically takes initial infrastructure
operating efficiencies as inputs that describe the CBRN incident, a process for extracting data
from limited scenario information needed to be developed. The Department of Homeland
Security's WARRP is a collaborative program designed to reduce time and resources required to
recover large urban areas, military installations, and associated critical infrastructures following
a catastrophic chemical, biological and radiological incident. SIRM was applied to two
hypothetical biological scenarios set in Denver, Colorado, which had been developed by
WARRP. For the purposes of this assessment, contamination information was limited to the
plume contours and contamination levels. Given this geographical information, the plumes were
then drawn in QGIS on top of census block data from the HAZUS database and building
information from the Geographical Repository and Data Management System (GRiD) database
[27][28], Because SIRM uses data in the form of percentage efficiency of infrastructure sectors,
the Denver metropolitan area (which includes Adams, Arapahoe, Broomfield, Clear Creek,
Denver, Douglas, Elbert, Gilpin, Jefferson, and Park counties) was chosen as the baseline by
which the percentages were based. Two views of the drawn scenarios (referred to as Scenario 1
and Scenario 2) are shown in Figure 11 and Figure 12, where the Denver Metropolitan Area is
mapped in orange, Scenario 1 is mapped in purple, and Scenario 2 is mapped in pink. Raw
coordinates of the polygons that comprise the scenarios can be found in Appendix B.
28
-------
Figure 11: View of the Denver Metropolitan Area in QGIS with Scenarios 1 (purple) and 2 (pink).
Figure 12: Closer view of Denver Scenarios 1 (purple) and 2 (pink) mapped out in QGIS.
4.1 Obtaining Infrastructure Asset Counts
Two separate datasets were used to count the number of infrastructure assets in the geographical
scenario polygons: a raw database file (.mdf) from the HAZUS database that can be accessed by
spatial queries in SQL Server Management Studio and a census block shapefile (.shp, .shx, dbf,
and .prj) from the HAZUS database that contains counts of the number of buildings that fall into
a set of categories. A building shapefile that contains more detailed building information from
the GRiD database was also investigated, however it did not contain any categorical data specific
to infrastructure sectors (as these data are more focused on building geometry and construction).
The analysis of the raw HAZUS database and the HAZUS census block shapefile are described
in the subsequent sections (Sections 6.1.1 and 6.1.2).
4.1.1 Raw Queries of the HAZUS Database
The raw portion of the HAZ US contains many relations, but only the asset relations (ones with a
dbo.hz prefix) contain building information. A subset of the asset relations contains usable data
for the 16 original DHSCIS sectors, and a subset of those relations are populated with enough
data to be usable. Within these relations, the number of assets inside the Scenario 1, Scenario 2,
Scenario 1 + 2, Denver metropolitan, and Colorado state polygons were then counted with a
spatial query. The resulting percentages were calculated for Scenario 1, Scenario 2, and Scenario
1 + 2 in Table 7 and could be similarly calculated for any potential area of interest. The raw
database contained information on 8 of the 9 infrastructure sectors, although some of these only
have sparse data. Total percentages for each infrastructure were also calculated using an
unweighted sum.
29
-------
Table 7: Relevant Raw HAZUS Database Percentage Contaminated by Infrastructure Sector
Infrastructure
Sector
Denver WARRP
Scenario 1
Denver WARRP
Scenario 2
Denver WARRP
Scenario 1 + 2
Water
2.27%
0.00%
2.27%
Energy
2.78%
0.00%
2.78%
Transportation
8.85%
0.28%
9.13%
Communications
0.00%
0.00%
0.00%
Government
8.48%
0.00%
8.48%
Healthcare
32.00%
0.00%
32.00%
Emergency Services
6.62%
0.66%
7.28%
WARRP: Wide Area Recovery and Resiliency Program
4.1.2 Census Block Analysis of the HAZUS Database
In addition to raw database queries, the census block shapefile for Colorado in the HAZUS
database could be used to fetch counts (and square footage) of buildings in various categories
called occupancy classes. The categories are based on Standard Industrial Classification (SIC)
codes. The key of SIC codes that divide the buildings into occupancy classes is provided in
Appendix C, Table 29.
For the occupancy class counts to be useful, a mapping of occupancy classes to DHSCIS sectors
is needed. Such a mapping has been proposed in Table 8. Some occupancy classes share
facilities in multiple infrastructure sectors and are listed in parentheses. For example, occupancy
class COM4 contains several transportation services like railroads and air travel, financial
services like credit institutions and insurance agents, government facilities like legal services,
and energy services like electric and gas services. COM4 also contains facilities that do not fall
into any of the DHSCIS categories, such as business services, engineering, accounting, research,
and other miscellaneous services. Because it is impossible to split the HAZUS occupancy class
information by specific SIC codes, simplifying assumptions and/or estimates would need to be
made if these occupancy classes were to be used (e.g., transportation systems compose of 50% of
the COM4 occupancy class). In the current absence of that type of information, occupancy
classes that suffer from this problem are not used in the final estimates of infrastructure losses for
the Denver scenarios in Section 6.2.
30
-------
Table 8: Occupancy Class Mappings of DHSCIS Sectors
Sector
Occupancy Classes (Shared Class)
Water and Wastewater Systems
(COM4)
Energy
(COM1), (COM4), (IND3), (IND4)
Transportation Systems
COM2, (COM4)
Communications
(COM8)
Government Facilities
RES5, GOVI, EDU1, EDU2, (COM4)
Healthcare and Public Health
RES6, COM6, COM7
Food and Agriculture
AGR1, (COM1), (COM8), (IND3)
Emergency Services
GOV2
Another significant consideration in these mappings are the categories that are not mapped to any
given sector (e.g., residential categories RES1, RES2, and RES3). This information could be
used in estimates of the number of people displaced by the CBRN incident. Residency
information could also be considered as additional infrastructure losses in DHSCIS sectors like
energy and water. For current loss estimates in Section 6.2, residence information is not used;
however, future estimates might use this information.
The last step of utilizing the census block information in HAZUS is fetching the spatial data
from the census blocks themselves using QGIS. Using the scenario polygons in Appendix D and
taking Denver metropolitan area and the state of Colorado as baselines, the occupancy class
counts and resulting efficiency percentage losses for each sector were gathered in Table 9. The
cells are colored in a gradient from red to green, where red squares indicate infrastructures that
are more affected and green indicates infrastructures that are less affected.
31
-------
Table 9: HAZUS Census Block Scenario Data
RESl
RES2
RES3
A
RES3B
RES3C
RES3D
RES3E
RES3F
RES4
RES5
RES6
COMl
COM2
COM3
COM4
COM5
Scenario 1
46,905
2,805
1,110
966
862
752
428
245
116
214
21
723
528
814
1481
247
Scenario 2
,2224
207
24
44
68
28
4
1
0
4
0
12
12
13
28
0
Scenario 1 and 2
49,003
3,004
1,134
1010
928
780
432
246
116
218
21
735
540
827
1510
247
Denver
731,539
20,793
7,246
8,051
7,959
6,035
2,234
828
639
920
243
7,748
6,255
7,971
15,148
1,166
Colorado
1,546,700
99112
19153
19182
14777
9604
3407
1200
2753
2757
610
18,168
13,370
17,265
31,074
2,191
Denver Seen 1 %
6.4%
13.5%
15.3%
12.0%
10.8%
12.5%
19.2%
29.6%
18.2%
23.3%
8.6%
9.3%
8.4%
10.2%
9.8%
21.2%
Denver Seen 2 %
0.30%
1.0%
0.33%
0.55%
0.85%
0.46%
0.18%
0.12%
0.00%
0.43%
0.00%
0.16%
0.19%
0.16%
0.19%
0.00%
Denver Seen 1+2 %
6.7%
14.4%
15.7%
12.5%
11.7%
12.9%
19.3%
29.7%
18.2%
23.7%
8.6%
9.5%
8.6%
10.4%
10.0%
21.2%
COM6
COM7
COM8
COM9
INDl
IND2
IND3
IND4
IND5
IND6
AGRi
RELl
GOVi
G0V2
EDUl
EDU2
Scenario 1
64
418
1000
24
165
323
53
102
16
494
140
547
282
26
168
22
Scenario 2
0
0
12
0
7
7
0
0
0
18
7
5
4
0
6
0
Scenario 1 and 2
64
418
1012
24
172
330
53
102
16
514
147
552
286
26
174
22
Denver
196
3,225
6,777
255
2,030
2,915
683
642
142
7,839
2,416
3,718
1,102
180
1,522
131
Colorado
484
6,890
14,253
516
4,529
5,824
1,426
1,223
350
17,351
8,357
8,893
2,758
459
3,492
322
Denver Seen 1 %
32.7%
13.0%
14.6%
9.4%
9.4%
8.1%
11.2%
7.8%
15.9%
11.3%
6.3%
5.8%
14.7%
25.6%
14.4%
11.0%
Denver Seen 2 %
0.00%
0.00%
0.18%
0.00%
0.00%
0.35%
0.24%
0.00%
0.00%
0.0%
0.23%
0.29%
0.13%
0.36%
0.00%
0.39%
Denver Seen 1+2 %
32.7%
13.0%
14.9%
9.4%
9.4%
8.5%
11.3%
7.8%
15.9%
11.3%
6.6%
6.1%
14.8%
26.0%
14.4%
11.4%
32
-------
4.2 Determining Efficiency Losses from Denver Scenarios
Infrastructure efficiency losses for use as inputs to SIRM were estimated for the Denver
scenarios with raw HAZUS data and available census block data. A listing of DHSCIS sectors
along with data quality ratings (green, yellow, and red) are presented in Table 10. Not all
infrastructure sectors had valid or sufficient data for each method, but 14 out of 16 and all 10 in
the current SIRM had at least some data that could be used. Green implies high data quality,
yellow implies average data quality, and red implies low data quality.
Table 10: DHS Infrastructure Sectors HAZUS Efficiency Data Quality Ratings
Sector
Data Quality w/ Raw Method
Data Quality w/ Census Blocks
Water and Wastewater
Average
Average
Energy
Average
Average
Transportation
High
High
Communications
Average
Average
Government Facilities
High
High
Healthcare and Public Health
High
High
Food and Agriculture
Low
High
Emergency Services
High
High
The set of rules below were used in generating estimates of infrastructure losses, with Denver
metropolitan data being chosen as the baseline for the percentages. With these guidelines,
estimates for losses of the DHSCIS sectors in Scenarios 1, 2, and 1+2 are presented in Table 11.
The cells are colored in a gradient from red to green, where red squares indicate infrastructures
that experience more loss and green indicates infrastructures that experience less loss.
• Among each set of data, if usable data are present in multiple occupancy classes or
database relations, then the counts were summed up in an unweighted lump sum.
• For sectors with two usable sets of data, the two percentages were averaged.
• Sectors with only one usable set of data simply utilized that data.
• Sectors with two sets of shared or sparse data defaulted to the raw HAZUS database
method (as the census block data are less usable if it is shared with other sectors). Sectors
with only one sparse or shared set of data, with the other method not producing any data,
used the only available set of data.
• Sectors with no usable set of data were set to 0%.
33
-------
Table 11: Preliminary Efficiency Loss Estimations of Denver Scenarios
Sector
Scenario 1
% Loss
Scenario 2
% Loss
Scenario 1+2
% Loss
Water and Wastewater Systems
2.27%
0.00%
2.27%
Energy
2.78%
0.00%
2.78%
Transportation Systems
8.65%
0.24%
8.88%
Communications
0.00%
0.00%
0.00%
Government Facilities
13.57%
0.19%
13.76%
Healthcare and Public Health
22.86%
0.00%
22.86%
Food and Agriculture
6.30%
0.23%
6.56%
Emergency Services
16.11%
0.51%
16.62%
Shortcomings to these estimates exists. For example, looking at the affected assets as
PATH/AW ARE does, Scenario 2 targets the Buckley Airforce Base, yet the HAZUS database
categorizes the base as an airport and not a military base. Additionally, these sums are
unweighted, treating high and low volume/population facilities in each sector equally. This
makes larger facilities like the Buckley Airforce Base have a smaller weight, and, in cases like
Scenario 2 that target one large facility with few smaller facilities around it, the scenario appears
to have an extremely small effect. In both raw HAZUS queries and census block analysis, partial
data could be loosely used as a weighting to address this issue (such as asset value, production
output, or square footage of the facility), but these data are not available for all sectors.
Some sectors have missing data, and some are only based on very sparse sets of data in the raw
HAZUS database. Additional data and SME input would improve estimates, and additional
considerations can be made for shared occupancy classes, residences, and the additional
cascading losses (e.g., adding energy losses to any sector that directly requires the sector's
services to operate).
After considering the infrastructure asset data using GIS tools, SIRM now has a means to
generate inputs for initial infrastructure operating efficiencies. Although a non-trivial amount of
effort was required to translate the Denver scenarios into a set of inputs for SIRM, a framework
has been established that can create a set of input infrastructure efficiencies for other scenarios.
This framework will be leveraged in future efforts to create a GIS enabled tool that can facilitate
the generation of inputs based on user-defined contaminated areas. Subsequent sections of this
report considered generating inputs for various scenarios from past natural disasters or CBRN
incidents and validating results with data reported in literature sources.
5 OPERATING EFFICIENCY MEASUREMENT AND
CALCULATIONS
The operating efficiency of the critical infrastructure sectors considered in SIRM can be
measured or calculated in a variety of different ways depending on the unique properties of the
sectors. The estimated operating efficiencies can be used to both validate SIRM as well as to
determine the initial sector efficiencies. This section discusses how the efficiency is measured for
34
-------
the nine infrastructure sectors considered in the reduced SIRM and the methods for determining
the efficiency using both GIS and available reports.
The methodology for using HAZUS GIS data to estimate sector efficiency is detailed in Section
4.2, with additional GIS datasets available through the DHS's Homeland Infrastructure
Foundation-Level Data (HIFLD). HIFLD is a national geospatial database with publicly
available infrastructure datasets and shapefiles and can be accessed through the internet [45],
However, in scenarios where GIS data are unavailable or might not be representative of the
sector efficiencies, reported data can be used to estimate sector efficiencies during an incident.
5.1 Water and Wastewater Efficiency
The HAZUS dataset contains some water infrastructure location information, however water is
combined with several other infrastructures into one overarching designation. The HIFLD
database contains more detailed water sector information, including locations of water treatment
plants and reservoirs [45],
The percentage of the population that has access to potable water can be used to estimate the
efficiency of the water sector (which could include private sources if such information is
available). Additionally, the water treatment capacity can also be assessed and used for
quantifying efficiency of the water sector. The threshold for a boil water advisory is a water
treatment capacity of at least 80%; if the capacity dips below this threshold, a mandatory boil
water advisory is instigated [30], implying that this information is reported. This however may
cause the efficiency to be underestimated, since it does not consider wastewater.
5.2 Energy Efficiency
Generally, in the case of an incident, power outage maps are quickly available as energy
companies work to restore power. The percentage of the population with power can be used to
estimate the energy sector efficiency, as well as determine efficiency loss for other
infrastructures. A power outage map can also be used, in tandem with the available GIS
information-it could be overlaid on population data to estimate the percentage of people and
businesses without power. An example of a power outage map that could be overlaid on a
population map is given in Figure 13 for Hurricane Sandy [33], The HIFLD database includes
several datasets that are relevant to the energy infrastructure, including locations of oil wells and
electric power lines.
35
-------
I Electricity outage area
I Electricity overload area
Inundated area
Figure 13: Power outage map for Hurricane Sandy [33].
5.3 Transportation Efficiency
Transportation is a more nuanced sector for which efficiency is evaluated, even with GIS
information. Roads or transportation hubs that are more well-traveled will have a higher impact
on the efficiency, and thus an efficiency calculation would ideally factor in these aspects instead
of just examining the percentage of roads and travel paths that are affected. One option for a
reported value that could account for some of this nuance is the percent increase in travel time,
which has been used in theoretical estimations of transportation efficiency.
A GIS analysis of transportation efficiency would most likely need to include traffic pattern
information to more carefully analyze effects of a contamination or debris scenario on
transportation efficiency. The HIFLD database [45] includes a dataset of traffic patterns and
could be leveraged in an analysis of the effects of a scenario on traffic, and subsequently travel
time. Additionally, if state departments of transportation contain any traffic information, delays
and travel time could be used to determine a metric for efficiency.
5.4 Communication Efficiency
The HAZUS dataset contains a communication designation. Additionally, communication
efficiency tends to be reported on by the Federal Communications Commission (FCC). Usually,
cell tower outages and service maps are quickly available, similar to power outage maps. An
example can be seen in Figure 14, published by the FCC for FTurricane Florence [40], These
values are usually tabulated for each county and can be utilized for assessing communication
efficiency.
36
-------
Percent Cell Sites Out-of-Service By County
9,162018 11:18:42 AM
i cuiieuitt^
41 Map Saidbu?
Leairvgion
Richmond
KENTUCKY
Roanoke* VIRGINIA
Norfolko oVirgmla Beach
Nashville
O Knoxvltte
Greensboro >
iNMESSEE
Chattanooga
Charlotte X.
Huntsvilie
o
Atlanta
llrtTungham O
0
tLABAMA
^T^rleston
Montgomery Q GEORGIA J\
Columbus SaViWyrr
+
Mcola TiBahasaee
g C » (1 0
Jacksonville ~
Go gl«
0
Mind#W «20>IOoool«. INEOi T#»rr>i o» Um
1-15 16 • 30 ¦ 31 - 45M 46 - <0 ¦ 61 - 100
Figure 14: Communication outage map for Hurricane Florence. Colors indicate the
percentage of outage
5.5 Government Efficiency
Government efficiency estimation is relatively uncomplicated from a GIS perspective since
government building location is available both through HAZUS as well as through other data
sources. Additionally, government building closure are reported [43] and thus if it is known how
many government buildings are in the area, finding the percentage of buildings that are affected
by a scenario is straightforward.
5.6 Food & Agriculture Efficiency
Using reported data for food and agriculture might be difficult to do, since assessing efficiency
for this sector is complicated. However, with addition to HAZUS, several GIS datasets exist that
could assist in the estimation of efficiency, used in tandem with incident GIS data. One dataset is
the USDA CropScape raster [44], which depicts the agricultural formation of the United States.
A sample portion of this dataset is depicted in Figure 15, with each color corresponding to a
different crop or other category (such as developed land, or forest).
37
-------
USDA
2015 Franklin County, Ohio
hmm
n0Kjl. r,
yj|T£V_
'ojrv-
Land Cover Categories
(by decreasing acreage)
AGRICULTURE
m Soybeans
I Grass/Pasture
I I Cot
I 1 Whiter Wheal
I I Sod/Grass Seed
I I Alfalfa
| Other Hay/Non Alfalfa
| | Fafiow/ldle Cropland
J DM Crop WnWht'SoyOeans
H Apples
| Tomatoes
I I Sorghum
I I Clover,'WtldBowers
Chraimw Trees
NON-AGRICULTURE*
I | Developed/Low Intensity
[=~ Developed/Open Space
I I Developed/Medium Intensity
I Deciduous Forest
~ Developed Hk)H Intensity
I I Open Water
Produced by CrapScupc ¦ http nnssgoodetu gmu cdu/CrctpScapc
* (My hip 6 non-agricuttureuitcjtone* arc luted
Figure 15: Agriculture map of Franklin County.
5.7 Emergency Services Efficiency
Emergency service facility locations are readily designated through HAZUS and thus efficiency
calculation from a GIS perspective is straightforward. However, calculating emergency service
efficiency through reported values is more difficult. One option is to examine response time to
distress calls and compare average response time to the increased response time, in a manner like
the transportation calculation. However, this information might not be readily available, and
thus, a GIS approach is the most convenient for emergency service efficiency calculation.
5.8 Waste Management Efficiency
Unfortunately, the waste management sector does not have a comparable designation in the
HAZUS dataset. The reported capacity for waste management could be utilized to create an
estimate for the efficiency-this methodology is used for calculating the remediation factor, as
detailed in Section 3.6.
38
-------
GIS data of the landfills and other waste management facilities in the United States could also be
used, where the percentage of landfills affected by an incident would be utilized to calculate an
overall percent efficiency. Though these data are not available through HAZUS, several other
datasets, accessed through HIFLD, have landfill and toxic waste disposal location data.
5.9 Healthcare Efficiency
A variety of methods can be used to leverage reported data to estimate the efficiency of the
healthcare infrastructure. Hospitals and other healthcare buildings are designated in the HAZUS
dataset and thus a GIS analysis is straightforward.
In the case of an evacuation, the percentage of hospitals that were evacuated can be used to
estimate the efficiency of the infrastructure. Though straightforward, this methodology lacks the
nuance required to estimate the actual efficiency of the healthcare infrastructure. One option to
mitigate this is to examine the increase in wait time at the hospital. This would be like the
transportation sector efficiency calculation, where the percent increase in wait time could be used
to calculate the efficiency of the healthcare infrastructure. A third approach could examine the
availability of hospital beds, updating this information as it becomes available. The percent
efficiency could be calculated by dividing the number of potential sick people by the number of
available hospital beds, which would be the percent "affected." The remaining percentage, of
theoretically available beds, would be the efficiency.
6 MODEL VERIFICATION WITH HISTORICAL DATA
A literature search was performed to identify historical data for scenarios that could be modeled
using SIRM. The scenarios were run through SIRM to perform model verification as well as
identify appropriate values for the constant inputs, including the repair factors.
Few scenarios existed for biological hazards, and the infrastructure related information in these
sources did not provide enough detail to inform the modeling efforts. However, efficiency values
for several infrastructures were available for some natural disasters including hurricanes Maria,
Sandy, Florence, and Harvey, and a water contamination scenario. The efficiency charts and
recovery times from SIRM were compared to the real-life reported outcomes.
6.1 Sample Scenario 1: Hurricane Harvey
Hurricane Harvey devastated Texas and Louisiana, causing much damage to Houston. The
recovery of Houston can be modeled using SIRM. GIS inundation data for Houston was
available through a FEMA source [29], A visualization is depicted in Figure 16.
39
-------
Figure 16: Flooding data for Houston during Hurricane Harvey, with red dots indicating
flooded points.
The flooding data were combined into one shapefile using the GIS software. HAZUS data were
also available for Houston. The building data were categorized using the same methodology as
Denver (Section 4.1.2). Any block that was flooded was considered nonfunctional, with the
counts of each infrastaicture in the block added to a "flooded" category. This is depicted in
Figure 17.
Figure 17: Example of flooded areas overlaid on HAZUS blocks.
The breakdown of flooded versus overall HAZUS building types is depicted in Table 12. Only
the relevant HAZUS buildings, as defined by Table 8 are included in the breakdown.
40
-------
Table 12: Breakdown of HAZUS Data
COM1
COM2
COM4
COM6
COM7
COM8
RES5
RES6
GOV1
GOV2
EDU1
EDU2
AGR1
Flooded
3771
2809
4726
190
1578
2393
275
116
331
51
606
89
647
Houston
17359
12435
21156
482
5720
10649
1210
469
1223
231
2638
219
2770
Percent Flooded
22%
23%
22%
39%
28%
22%
23%
25%
27%
22%
23%
41%
23%
The percentage flooded was averaged across each critical infrastructure, and used to calculate the
efficiency, as defined in Table 13.
Table 13: Percentage Flooded for Each Infrastructure
Infrastructure Sector
Percent
Flooded
Efficiency
Water
22.30%
77.70%
Energy
22.0%
77.97%
Transport
22.0%
77.97%
Communication
22.5%
77.53%
Government
28.4%
71.65%
Food
22.5%
77.46%
Emergency
22.1%
77.92%
Waste Management
-
100.00%
Healthcare
30.6%
69.42%
Some reported information was available about infrastructure efficiency values during Hurricane
Harvey. These values were compared to the values calculated by GIS flooding data and are
summarized in Table 14.
41
-------
Table 14: Hurricane Harvey Efficiency Values
Infrastructure
% Efficiency
through GIS
% Efficiency
Reported
Reasoning
Water
77.70%
80%
Water treatment capacity dipped to at
least 80%, the minimum for a boil water
advisory [30]
Energy
77.97%
83%
350,000 people lost power [31]
Transport
77.97%
Communication
77.53%
98.60%
FCC report on the county [34]
Government
71.65%
Food & Agriculture
77.46%
Emergency Services
77.92%
Waste Management
100.00%
Healthcare
69.42%
81.82%
20 out of 110 hospitals were evacuated
[35]
The estimated water and energy infrastructure efficiencies were comparable to the values found
in report. However, communication reports tended to give a higher efficiency than the GIS data.
This is logical as cell towers and other communication infrastructure tend to be able to function
even in the case of some flooding. Similarly, hospitals that experienced minor flooding were less
likely to evacuate, and thus the reported efficiency was higher than the GIS estimate of
efficiency.
The scenario was simulated in SIRM, using the repair factors defined in the Denver scenario and
the population of Houston as the baseline population. The efficiency versus time graph is
depicted in Figure 18.
42
-------
food
emergency
20 waste management
healthcare
°0 20 40 60 80 100
Time (days)
Figure 18: Efficiency/time profile of Hurricane Harvey scenario.
The recovery times of several of the infrastructures was available in the literature. These were
compared to the results produced by SIRM in Table 15. SIRM tended to overestimate recovery
time. However, since the reported recovery data were imprecise, the recovery times produced by
SIRM were still logical and passed a visual inspection/validation.
Table 15: Recovery Time Comparison
Infrastructure
Initial %
Efficiency
Recovery Time in Days
(Literature) [37]
Recovery Time
in Days
(SIRM)
Water
77.70%
>21
39
Energy
77.97%
>21
57
Transport
77.97%
28
Communication
77.53%
40
Government
71.65%
Food & Agriculture
77.46%
Emergency Services
77.92%
Waste Management
100.00%
>21
Healthcare
69.42%
>21
22
43
-------
6.2 Sample Scenario 2: Hurricane Maria
Hurricane Maria devastated Puerto Rico, causing massive power outages, and severely affecting
many of the other infrastructures. Data were available for the infrastructure efficiencies over
time, though no detailed information was available for how these efficiencies were initially
calculated [32], The initial efficiencies are summarized in Table 16.
Table 16: Initial Infrastructure Efficiencies for Hurricane Maria [32]
Infrastructure
Initial % Efficiency
Water
40%
Energy
0%
Transport
75%
Communication
30%
Government
100%
Food & Agriculture
49%
Emergency Services
100%
Waste Management
100%
Healthcare
15%
The scenario was simulated in SIRM, with several possible repair factors, and using the
population of Puerto Rico as the baseline population. Since efficiency/time data were available
through the original paper [32], the repair factors were adjusted to best match the efficiency
curves in the paper. The repair factors that were ultimately chosen are displayed in Table 17.
Since these were the only repair factors that could be compared to a reported efficiency chart,
these are the repair factors suggested for SIRM and were used in most of the remaining
scenarios.
44
-------
Table 17: Infrastructure Repair Factors for Hurricane Maria
Infrastructure
Repair Factor
Water
0.95
Energy
0.95
Transport
0.95
Communication
0.9
Government
0.9
Food & Agriculture
0.9
Emergency Services
0.9
Waste Management
0.9
Healthcare
0.9
The efficiency results are depicted in Figure 19 and compared to the results from the original
paper in Figure 20. The results were comparable based on a visual inspection. Since only the
overall chart of the efficiency values was provided in the paper, a more analytical inspection was
not possible.
45
-------
Figure 19: SIRM results.
Recovery ,pf Puerto Rica's CI Systems FolldvingA Hu'ticjne M;»ri«»
0
-Power
= Telecom
»Banks
-ATMs
- Water
- Supermarket/Grocery Store
* Hospitals (w/ elec.)
-Commercial Flights
140 160
"CHI Antertnas/Sites
-Gas Stations
* Paris
- AMA Routes (Bus)
Figure 20: Paper results. [32]
6.3 Sample Scenario 3: Hurricane Sandy
Hurricane Sandy information was available through a source found in the literature review. The
initial infrastructure efficiencies described in the paper are displayed in Table 18 [33] The direct
damages were defined as the "physical damages caused by Sandy in each sector" [33], The
indirect damages were caused by functional problems, estimated using incident GIS data, and
included power outage, overload and impacts of failures in other sectors [33 |.
Table 18: Initial Efficiency Values for Hurricane Sandy
Infrastructure
Direct Damage (%)
Indirect Damage (%)
Initial %
Efficiency
Electricity
9.9%
2.8%
87.3%
Transportation
10.7%
19.4%
86.9%
Healthcare
7.5%
2.4%
90.1%
Waste Management
10%
-
90%
46
-------
The efficiency values were calculated using open source information about the buildings in New
York City and flooding and power outage GIS data (following a similar strategy as described in
Section 1), and combined with reported efficiency gathered by the City of New York. The GIS
data were used to estimate the indirect damage, while the City of New York reported data were
used to estimate the direct damage.
The repair factors for energy and transportation were set to 0.1, while the repair factor for
transportation was set to 0.5. The remaining repair factors were set to 0.9. New York City was
used as the bounds for the scenario. The results are depicted in Figure 20.
water
energy
transport
communications
government
food
emergency
waste management
healthcare
°0 10 20 30 40 50
Time (days)
Figure 20: Hurricane Sandy efficiency chart.
Some data were available for the recovery of the energy infrastructure; sources claimed that it
took several weeks for the power to return to everyone in New York City [33], This corresponds
with the length of time the energy infrastructure required to recover when simulated in SIRM.
6.4 Sample Scenario 4: Elk River Spill
In 2014, methylcyclohexane methanol (MCHM) was released in the Elk River in West Virginia.
Elk River was used as the source of the water for the Charleston, West Virginia drinking water
utility. The chemical passed through the treatment plant and was spread throughout the
distribution system. In total, 17.6% of residents reported having rainwater and 5.6% reported
well water available, however the rest of the population was assumed to rely on the Elk River
water [46], The efficiency of the water infrastructure was thus assumed to be 23.3%. This is
likely a lower efficiency than what was observed in reality, since wastewater is included in the
water efficiency but may not have been affected in this particular scenario.
Six hundred people complained of illness. It took around six days to decontaminate. These
values were used to set the percentage of the water infrastructure contaminated, with the
assumption being that the entire water infrastructure was contaminated. The remediation factor
47
-------
was set to 16.7% per day, to capture that the infrastructure was decontaminated in six days. The
scenario was simulated in SIRM, using the population of the area defined in the original source
and the repair factors used for the Denver and Hurricane Harvey scenarios [36], Since the water
infrastructure was the only one seriously affected, a histogram of the recovery time for the water
infrastructure is displayed in Figure 21.
Parameter Histogram: Sector Recovery Time (Days) of Water Sector Efficiency
0 10 20 30 40 50
Parameter Value
Figure 21: Recovery time of Water infrastructure.
The median recovery time (i.e., the time to reach 100% operating efficiency as per SIRM
simulation) was around 33 days. No clear information was provided as to when the infrastructure
was completely recovered, only some information that the water tested negative for MCHM a
year later [36],
6.4.1 Addition of Backups
Some infrastructures might have backups available. For example, a communications facility
might have a backup power generator, in which case it could subsist on this while the energy
sector focuses on recovery and providing support to the other sectors. Though this might not be
relevant for the majority of scenarios, in the case of the Elk River Spill, an analysis can be done
with the assumption that all infrastructures have some level of stored water that can be used for
the needs of the sector while the water sector is recovering.
DHS CISA provided a preexisting critical infrastructure dataset that included information about
the backups available for the water infrastructure. An examination of the same Elk River
scenario with backup water infrastructure would be informative as to the impact of a backup. A
summary of the availability of backup water (tanks, water power), as per the DHS dataset, is
detailed in Table 19. The percent of the sector with the backup parent was translated to the
percent efficiency of the backup infrastructure for input into SIRM.
48
-------
Table 19: Availability of Backup Water
Parent: Water
Child Sector
Percent of
Mean days
sector with
backup is
backup parent
available
Communications
31.0%
3.81
Emergency
9.0%
2.83
Energy
26.86%
2.57
Food
46.5%
3.46
Healthcare
42.0%
3.44
Transport
14.5%
1.79
Water
10.5%
1.67
Government
3.75%
2.60
6.4.2 Scenario Results with Backups
The Elk River Spill, with the same initial inputs, was again simulated in SIRM, with the addition
of the backups detailed in Table 19. A histogram of the recovery time for the water infrastructure
is displayed in Figure 22, compared to the original histogram of water recovery.
49
-------
Figure 22: Histograms of recovery time of Water (top: original, bottom: with backups).
As can be seen by the histograms, the recovery time of the water infrastructure is reduced to a
median of 26.63 days with the addition of backup water. This makes sense due to the nature of
SIRM, as each infrastructure is "used" to improve the efficiency. Since the water sector is highly
connected to the other infrastructure sectors, the efficiency is reduced to improve the efficiencies
of the other infrastructures. In the case where all the other infrastructures have backup water, the
water sector can focus efforts to decontaminate and increase efficiency without having to provide
to other infrastructures.
6.5 Sample Scenario 5: Hurricane Florence
Hurricane Florence was a powerful and long-lived hurricane that caused catastrophic damage in
the Carolinas in September 2018. Information for North Carolina was available through sources
found in literature searches [38]-41], The initial infrastructure efficiencies described in several
pieces of literature are displayed in Table 20. Based on these available values, a hypothetical
scenario for a portion of North Carolina, influenced by some of Florence's impacts, was
developed and simulated in SIRM.
50
-------
Table 20: Inputs for Hurricane Florence Based Scenario
Infrastructure
Initial %
Efficiency
(post
impact)
Reasoning
Water
100%
Very little information about the water sector was
available in literature.
Energy
83%
Taken from a DOE daily situation reports describing
the percentage of people in the state with and
without power [38],
Transport
0%
All roads were closed in several areas and
transportation was essentially impossible according
to the NCDOT [39],
Communication
86.4%
FCC published situation reports on the
communication infrastructure, based on reports of
cell sites out of service by county [40],
Government
100%
Food & Agriculture
100%
Though food and agriculture was affected, a
percentage estimate was difficult to find in the
literature.
Emergency Services
100%
Waste Management
100%
Healthcare
68.3%
It was estimated that 40 hospitals [41] in North
Carolina were in the path of Florence and thus
evacuated. North Carolina has a total of 129
hospitals[41], This proportion was used to calculate
overall percent efficiency.
6.5.1 Scenario Results
The scenario was simulated in SIRM, using the population of North Carolina and the repair
factors used for the Denver and Hurricane Harvey scenarios. The results of the recovery times
for the relevant infrastructure sectors are tabulated in Table 21 and compared to the cited times
of recovery.
51
-------
Table 21: Hurricane Florence Recovery Times [38-41].
Infrastructure
Recovery Time in
Literature
Recovery Time using SIRM
(days)
Water
-
-
Energy
Around 7 days
56
Transport
A few weeks - the last road
opened about 6 months later
50
Communication
8 days
18
Government
-
-
Food & Agriculture
-
-
Emergency Services
-
-
Waste Management
-
-
Healthcare
Uncertain since Hurricane
Michael hit soon after
27
SIRM tended to overestimate recovery time, which was cause for further investigation, detailed
in the following section.
6.5.2 Scenario with No Infrastructure Reduction
The same scenario was simulated in SIRM again, however the efficiencies were not reduced.
Depending on the infrastructure reaction that was "chosen," the parent infrastructure efficiencies
would not be reduced and instead kept at the same level. To calibrate these values, the repair
factors were reduced, and adjusted until the visual inspection of the results seemed logical. The
chosen repair factors are listed in Table 22.
52
-------
Table 22: Repair Factors with No Reduction
Infrastructure
Repair Factor
Water
0.9
Energy
0.5
Transport
0.5
Communication
0.5
Government
0.5
Food & Agriculture
0.5
Emergency Services
0.5
Waste Management
0.5
Healthcare
0.5
SIRM and literature recovery times are tabulated in Table 23.
Table 23: Recovery Times Without Infrastructure Reduction [38-41].
Infrastructure
Recovery Time
Recovery Time using
SIRM (days)
Water
-
-
Energy
Around 7 days
14
Transport
A few weeks - the last
road opened about 6
months later
41
Communication
8 days
11
Government
-
-
Food & Agriculture
-
-
Emergency Services
-
-
Waste Management
-
-
Healthcare
Uncertain since
Hurricane Michael hit
soon after
26
53
-------
The results align more closely with the real-life results, with transportation understandably
taking a long time to recover while other infrastructures are quicker to recover. Future SIRM
enhancements will explore the need for infrastructure efficiencies reduction.
6.6 Conclusions from SIRM Verification
Overall, the model verification showed that the results tended to be reasonable and correspond
with actual real-life results. Several considerations would need to be addressed in future
modeling efforts and real-life use cases. One consideration is the determination of accurate repair
factors to use in a scenario. Appropriate repair factors are difficult to determine, but are
important in further decision making, as they provide a method for prioritizing and quantifying
recovery efforts.
Another consideration is the appropriateness of stoichiometric infrastructure efficiency reduction
as opposed to not reducing the infrastructure efficiency. Due to the nature of the Gillespie
algorithm, which is used as the backbone of SIRM, it is advisable to reduce the parent
infrastructures when a child infrastructure equation is implemented. However, this might not be
relevant in cases where the infrastructures are drawing on independent outside forces, such as in
the case of Hurricane Florence (Section 6.5.2). This could be modeled by using backup
infrastructures, which would keep the basic framework of the initial SIRM and allow for only
certain infrastructures to be used for support, as exemplified in the Elk River Spill (Section
6.4.1).
Finally, GIS data might not be representative in certain scenarios. For some sectors, such as
communications and healthcare, the infrastructure sector can operate at a higher efficiency than
could be depicted by the GIS data. For example, a cell tower that is flooded or contaminated can
still be operational, and in cases where hospitals are closed, the healthcare infrastructure can
continue to function at high capacity. Thus, examining these infrastructures in more detail during
a simulation might be logical.
7 SENSITIVITY ANALYSIS
Analysis is performed to determine how different values of each variable or input to a model
affect the model's output. To further evaluate the effect of the inputs to SIRM, sensitivity
analyses were performed on several of the parameters, using Hurricane Harvey inputs as the
baseline, which are in Table 14 (repeated below).
54
-------
Infrastructure
% Efficiency
through GIS
% Efficiency
Reported
Reasoning
Water
77.70%
80%
Water treatment capacity dipped to at
least 80%, the minimum for a boil water
advisory [30]
Energy
77.97%
83%
350,000 people lost power [31]
Transport
77.97%
Communication
77.53%
98.60%
FCC report on the county [34]
Government
71.65%
Food & Agriculture
77.46%
Emergency Services
77.92%
Waste Management
100.00%
Healthcare
69.42%
81.82%
20 out of 110 hospitals were evacuated
[35]
The effects of three parameters were assessed: (1) the initial efficiency of designated "core"
infrastructures, (2) the repair factors, and (3) backups. The recovery time (i.e., days for an
infrastructure sector to reach 100% efficiency) was calculated and analyzed for each sector, for
each set of inputs, and used as the marker for assessing the influence of each of the parameters.
7.1 Initial Efficiency
Four previously determined "core" infrastructure sectors that were especially relevant to the EPA
mission - Water, Energy, Waste Management and Healthcare - were analyzed. The initial
efficiencies were varied from 10% to 90%, one at a time. The other inputs remained constant,
and the subsequent effects on recovery time were observed for the other infrastructures.
7.1.1 Effects of Water
The initial efficiency of the Water sector was varied while the other infrastructures kept an initial
efficiency level the same as the original efficiencies for Hurricane Harvey. The recovery time for
each infrastructure is charted in Figure 23.
55
-------
Water Percent Efficiency Sensitivity
200
180
160
__ 140
"i/T
ra
2. 120
o
u
0)
cc
80
60
40
20
0
l-J ¦
1 ' ^
1
-Communications
Emergency Services
Energy
Food and Agriculture
Government Facilities
Healthcare
-Transportation Systems
-Water and Wastewater Systems
-Waste Management
20 40 60 80
Water Initial Percent Efficiency
100
Figure 23: Effects of initial Water efficiency.
In summary, Energy, Transport, Healthcare, and Communications sectors were most heavily
affected by the initial efficiency of the Water sector. This is most likely due to the heavy linkages
between these infrastructures and the Water sector in the infrastructure map as evidenced in
Figure 10. The slopes of the subsequent trendlines are displayed in Table 24. The darker the
green, the higher the effect of the initial Water efficiency on the sector recovery time.
56
-------
Table 24: Effect of Initial Water Efficiency on Recovery Time
Sector
Slope - decrease in recovery time (days) for
every 1% increase in initial efficiency
Communications
-0.62
Emergency Services
-0.08
Energy
-0.67
Food and Agriculture
-0.04
Government Facilities
-0.35
Healthcare
-0.56
Transportation Systems
-0.66
Waste Management
-0.04
Water and Wastewater Systems
-0.70
7.1.2 Effects of Energy
The recovery time for each infrastructure as Energy is varied is charted in Figure 24.
57
-------
180
Energy Percent Efficiency Sensitivity
160
140
120
a;
jl 100
>-
1—
QJ
8 80
01
Cd
60
40
20
0
0 20 40 60 80 100
Energy Initial Percent Efficiency
Figure 24: Effects of initial Energy efficiency.
In summary, Water, Transport, and Communications sectors were most heavily affected by the
initial efficiency of the Energy sector. This is most likely due to the heavy linkages between
these infrastructures and the Energy sector in the infrastructure map as evidenced in Figure 10.
The slopes of the subsequent trendlines are displayed in Table 25. The darker the green, the
higher the effect of the initial Energy efficiency on the sector recovery time.
-Communications
Emergency Services
Energy
Food and Agriculture
Government Facilities
Flealthcare
-Transportation Systems
-Water and Wastewater Systems
-Waste Management
58
-------
Table 25: Effect of Initial Energy Efficiency on Recovery Time
Sector
Slope - decrease in recovery time (days) for
every 1% increase in initial efficiency
Communications
-0.74
Emergency Services
-0.10
Energy
-0.72
Food and Agriculture
-0.05
Government Facilities
-0.25
Healthcare
-0.42
Transportation Systems
-0.78
Waste Management
-0.02
Water and Wastewater Systems
-0.71
7.1.3 Effects of Waste Management
The recovery time for each infrastructure as Waste Management is varied is charted in Figure 25.
59
-------
Waste Percent Efficiency Sensitivity
180
160
140
~120
>¦
ro
~o
<¥ 100
0)
>
0
u
01
EE
80
60
40
20
*
>
9 ,
: \ .
-Communications
Emergency Services
Energy
Food and Agriculture
Government Facilities
Healthcare
-Transportation Systems
-Waste Management
-Water and Wastewater Systems
20 40 60 80
Waste Management Initial Percent Efficiency
100
Figure 25: Effects of initial Waste Management efficiency.
The Waste Management sector recovery had the most effect on Transportation, Water, and
Energy sectors. The slopes of the subsequent trendlines are displayed in Table 26. The darker the
green, the higher the effect of the initial Waste Management efficiency on the sector recovery
time. Note that the sensitivity values for Waste Management were significant lower, indicating
that the initial efficiency of the Waste Management infrastructure was less influential on the
recovery time of the other sectors.
60
-------
Table 26: Effect of Initial Waste Efficiency on Recovery Time
Sector
Slope - decrease in recovery time (days) for
every 1% increase in initial efficiency
Communications
-0.15
Emergency Services
-0.01
Energy
-0.26
Food and Agriculture
-0.001
Government Facilities
-0.01
Healthcare
-0.07
Transportation Systems
-0.31
Waste Management
-0.77
Water and Wastewater Systems
-0.28
7.1.4 Effects of Healthcare
The recovery time for each infrastructure as Healthcare is varied is charted in Figure 26.
61
-------
Healthcare Percent Efficiency Sensitivity
180
160
140
120
V)
>
ro
~o
OJ 100
|
i-
i 80
>
o
u
-------
Table 27: Effect of Initial Healthcare Efficiency on Recovery Time
Sector
Slope - decrease in recovery time (days) for
every 1% increase in initial efficiency
Communications
-0.35
Emergency Services
0.004
Energy
-0.56
Food and Agriculture
-0.01
Government Facilities
-0.18
Healthcare
-1.06
Transportation Systems
-0.62
Waste Management
0.01
Water and Wastewater Systems
-0.60
7.2 Repair Factors
Repair factors are used to model the rate society works to repair the infrastructure sector. A
higher repair factor implies that society is working harder to repair that sector. For the repair
factor analysis, the repair factor was set to a constant value of 0.5, while a single infrastructure's
repair factor was varied from 0.1 to 0.9. The effects on recovery time for each infrastructure was
observed. This was done for all the infrastructures.
Detailed charts for each infrastructure are available in Appendix D. The charts displayed the
effects of each sector's repair factors on a single sector's recovery time (i.e., Figure 27 displays
the Waste Management sector's recovery time changing as the other infrastructure repair factors
changed). Generally, the repair factor tended to only affect the relevant infrastructure (i.e., the
Waste Management repair factor mainly affected the Waste Management sector recovery time).
An example of this pattern is displayed in Figure 27.
63
-------
Repair Factor Effect on Waste Management
40
1/5
>-
ra
¦o
O
<_>
a;
Q£
35
30
25
20
15
c
-------
160
V)
>-
ro
~o
O
<_>
a;
fl£
C
o
ro
t
o
CL
c
ro
80
60
40
20
Repair Factor Effect on Transportation Systems
0.2 0.4 0.6 0.8
Repair Factor of Infrastructure Sectors
-Water
- Energy
Transport
Communications
Gove rnment
Food and Agriculture
¦Emergency Services
¦Waste Management
¦Healthcare
Figure 28: Infrastructure repair factors effects on Transportation recovery time.
7.3 Backup Analysis
The effect of adding backups (i.e., a generator, water coolers) for the core infrastructures (Energy
and Water) was observed. A baseline scenario was simulated, and then backup energy was added
to all the sectors for varied numbers of days. The same process was followed for backup water.
The days that the backup infrastructure was available was set at 1 day, 5 days, and 10 days. The
efficiency of the backup parameter was varied between 10%, 50%, and 90%.
Detailed charts of the recovery times for all sectors based on backups are available in Appendix
D. An example chart is displayed in Figure 29 for the Communication sector. The recovery time
is displayed on the y-axis, while the days that the backup infrastructure is available is on the x-
axis. Each curve depicts a backup efficiency level (of 10%, 50% and 90%).
65
-------
120
Communications
-sr no
>-
60
0 2 4 6 8 10 12
Days of Backup Energy
Figure 29: Sample chart for backup analysis (recovery time of Communication).
The backup analysis showed that backup infrastructures functioning at 10% tended to have little
effect on the recovery time, while 50% and 90% backup had larger effects on the recovery time.
Most infrastructures, at the 50% backup level, did have a point of "diminishing returns," where
backups stopped influencing the recovery time. For 50% backup, this point of diminishing
returns tended to be around five days. This can be seen in Figure 30.
Several infrastructures did not appear to benefit from backups, with charts essentially showing a
flat curve. An example of this chart is displayed in Figure 30.
66
-------
Food and Agriculture
40
35
I/)
>-
03
30
£D
-
—
QJ
15
>
O
(J
10
cc
5
0
- 1
•
•
: 1
10%
50%
90%
4 6 8
Days of Backup Water
10
12
Figure 30: Recovery time of Food and Agriculture with backup Water.
Infrastructures that followed this pattern included Government, Food and Agriculture,
Emergency Services, and Waste Management. For these infrastructures, backup Water and
Energy were thus less beneficial and might not be as crucial in remediation efforts.
Infrastructures that did see high benefits from backup Water and Energy included
Communications, Energy, and Transportation. Healthcare recovery time were positively affected
by Energy, but backup Water had less effect on recovery time.
8 CONCLUSION
After performing a literature review, exploring the capabilities of infrastructure software
packages, and developing preliminary versions of SIRM, an educated model selection for future
infrastructure analysis was made. Many of the mathematical models found in the literature
review were not dynamic or could not accommodate the desired infrastructure sectors. Agent-
based models proved to be a promising modeling option based on the literature review. However,
the software selected in preliminary analysis, PATH/AW ARE, did not have the dynamic
modeling capability, one of the desirable model criteria. Software issues existed with ThinkGeo
data (and PATH/AW ARE itself), stochastic modeling was not possible, and the underlying
mathematical model did not track the condition of people in the workforce. Since
PATH/AW ARE is a calculation tool rather than a model, PATH/AW ARE could not serve as a
complete infrastructure model.
SIRM, while requiring a significant amount of effort to create and verify for accuracy, features
most of the desirable modeling criteria. It has stochastic modeling methodology that considers
best-case and worst-case scenarios after a CBRN incident, it can be dynamically written to
67
-------
consider any number of infrastructure sectors, and it can consider any number of dynamic
infrastructure dependencies. As a result, SIRM has been selected for use in future modeling
efforts.
SIRM can assist in emergency response and decision making in several ways. The estimated
recovery time can be assessed based on the initial scenario parameters and can be adjusted and
recalculated as new information becomes available. The repair factors can also be adjusted to
model the prioritization of infrastructure recovery by setting higher relative repair factors for
higher priority infrastructure sectors. Additionally, the benefits of adding backups can be
assessed for decision making with initial incidents - deciding which infrastructures would
benefit the most from backups can be assessed.
9 SIRM IMPROVEMENT PLAN
Because SIRM is a proof-of-concept model, a large amount of SME input and additional data is
required to make SIRM results as accurate and useful as possible. Specifically, the following list
of items require verification or adjustment:
• The infrastructure connectivity network (i.e., reaction formulas)
• Values of input parameters (e.g., nO, pO, repair factors)
• An assessment of the societal impact of infrastructure shutdown and a response variable
for analyzing infrastructure remediation strategies
• Improvements of the implementation of the GIS-based inputs to SIRM
• Exploration of the appropriateness of infrastructure efficiency reduction
• Other changes to the fundamentals of SIRM, as needed
The values of the input parameters are scenario-specific and largely rely upon analysis of known
scenario information via tools such as GIS. Groundwork on the determination of efficiency
parameters based on reported and GIS data has been done in Section 5. SIRM was further
verified with existing historical scenarios in Section 6. The verification processes laid out in
Section 6 can be used as new scenarios and real-life data becomes available. Other input
parameters like the repair factors and infrastructure stoichiometric factor could also be further
verified or adjusted based on data. Future iterations of the SIRM tool will include a methodology
for loading in GIS data of the CBRN incident, thereby allowing for these calculations to be
automated. The HAZUS and other infrastructure datasets will be preloaded into the SIRM tool,
and the affected percentage will be used to calculate infrastructure efficiencies.
68
-------
10 REFERENCES
[1] Gillespie DT. 1977. Exact stochastic simulation of coupled chemical reactions. J Phys
Chem. 81(25):2340-2361.
[2] Haimes YY. 2018. Modelingandmanaginginterdependentcomplexsystemsofsystems.
Wiley-IEEE Press.
[3] Hohn G. 2018. Application to economics: Leontief Model. Kansas State University.
[4] U.S. Bureau of Economic Analysis (US BEA). 2018. Input-Output Accounts Data. Nov
2018.
[5] Cimellaro GP, Crupi P, Kim HU, Agrawal A. 2019. Modeling interdependencies of critical
infrastructures after Hurricane Sandy. Int J Disaster Risk Reduct. 38:101191.
[6] RehakD, Markuci J, Hromada M, Barcova K. 2016. Quantitative evaluation of the
synergisticeffectsoffailuresinacriticalinfrastructuresystem.Int J Crit Infrastruct Prot.
14:3-17.
[7] Rehak D, Senovsky P, Hromada M, Lovecek T. 2019. Complex approach to assessing
resilience of critical infrastructure elements. Int J Crit Infrastruct Prot. 25:125-138.
[8] Lam CY, Tai K. 2018.Modeling infrastructure interdependencies by integrating network and
fuzzy set theory. Int J Crit Infrastruct Prot. 22:51-61.
[9] Wang W, Yang S, Hu F, Stanley HE, He S, Shi M. 2018. An approach for cascading effects
within critical infrastructure systems. Physica A: Stat Mech Appl. 510:164-177.
[10] Laug'e A, Hernantes J, Sarriegi JM. 2015. Critical infrastructure dependencies: A holistic,
dynamic and quantitative approach. Int J Crit Infrastruct Prot. 8:16-23.
[11] Chappin EJL, van der Lei T. 2014. Adaptation of interconnected infrastructures to climate
change: A socio-technical systems perspective. Util Policy. 31:10-17.
[12] Bollinger LA, Bogmans CWJ, Chappin EJL, Dijkema GPJ, et al. 2014. Climate adaptation
of interconnected infrastructures: A framework for supporting governance. Reg Environ
Change. 14(3):919-931.
[13] Roe E, Schulman P. 2016. Managing interconnected critical infrastructures: Challenges to
practice policy. California Public Utilities Commission.
[14] Presidential Policy Directive-Critical Infrastructure Security andResilience. 2013. National
Archives and Records Administration, Feb 2013.
[15] Keller N. 2014. Cybersecurity framework. NIST.
69
-------
[16] Jumarie G. 1993. Stochastic differential equations with fractional Brownian motion input.
IntJSyst Sci. 24(6): 1113-1131.
[17] Linetsky V, Mendoza R. 2010. Constant elasticity of variance (CEV) diffusion model. In:
Encyclopedia of Quantitative Finance, https://doi.org/10.1002/9780470061602.eqf08015
[18] Department of Homeland Security. 2019. Critical Infrastructure Sectors.
https://www.cisa.gov/critical-infrastructure-sectors
[19] Gillespie DT. 1977. Exact stochastic simulation of coupled chemical reactions. J Phys
Chem. 81(25):2340-2361.
[20] Bhamidipati S, van der Lei T, Herder P. 2016. A layered approach to model interconnected
infrastructure and its significance for asset management. Eur J Transp Infrastruct Res. 16(1).
[21] Thompson JR,Frezza D,Necioglu B, Cohen ML,Hoffman K, Rosfjord K. 2019.
Interdependent Critical Infrastructure Model (ICIM): An agent-based model of power and
water infrastructure. Int J Crit Infrastruct Prot. 24:144-165.
[22] Milanovi'c JV, Zhu W. 2017. Modeling of Interconnected Critical Infrastructure Systems
using Complex Network Theory. IEEE Trans on Smart Grid. 9(5):4637-4648
[23] Scipy.integrate.odeint Documentation. 2019. The SciPy Community.
[24] Colorado Hospital Association. 2017. Colorado Hospital Utilization Data.
https://cha.com/data-reporting/colorado-hospital-utilization-data/
[25] Lloyd-Smith JO. 2013. Vacated niches, competitive release and the community ecology of
pathogen eradication. Philos Trans R Soc B: Biol Sci. 368(1623):20120150.
[26] Althaus CL. 2014. Estimating the reproduction number of Ebola virus (EBOV) during the
2014 outbreak in West Africa. PLoS Currents. 6.
[27] US Department of Homeland Security. 2013. Hazus- MH 2.1 User manual. Jul 2013.
[28] US Army Corps of Engineers Cold Regions Research & Engineering Laboratory (CRREL).
2012. Geospatial Repository and Data Management System (GRiD).
[29] FEMA Modeled Building Damage Assessments Harvey 20170829. 2017. Hub.arcgis.com,
hub.arcgis.com/datasets/lac0b7c856a047e6bba2c66b32982fD0_0/data
[30] Sensorex. 2018. Water Quality After Hurricane Harvey: What You Need to Know.
sensorex.com/blog/2017/08/28/water-quality-hurricane-harvey-need-know/
[31] Bubenik T. 2017. Though power outages were limited, Harvey revealed new challenges for
the grid. Houston Public Media, 2 Nov. 2017.
[32] Sarker P, Lester HD. 2019. Post-disaster recovery associations of power systems dependent
70
-------
critical infrastructures. Infrastruct. 4(2):30, doi:10.3390/infrastructures4020030
[33] Haraguchi M, Soojun K. 2016. Critical infrastructure interdependence in New York City
during Hurricane Sandy. Int J Disaster Resil Built Environ. 7:133-143.
https://doi.org/10.1108/IJDRBE-03-2015-0015
[34] Federal Communications Commission (FCC). 2018. Hurricane Harvey. Federal
Communications Commission, 24 Aug. 2018, www.fcc.gov/harvey
[35] Advisory Board. 2017. How Houston Hospitals fared during Harvey-and what they're
preparing for next. Advisory Board Daily Briefing, www.advisory.com/daily-
briefing/2017/09/05/houston-hospitals-during-harvey
[36] National Library of Medicine (NLM). 2016. Rapid Response: West Virginia 2014 Elk
River Spill | NLM in Focus. U.S. National Library of Medicine, National Institutes of
Health, infocus.nlm.nih.gov/2016/03/09/rapid-response-west-virginia-elk-river-2014-spill/
[37] Amadeo K. 2019. Hurricane Harvey shows how climate change can impact the economy.
The Balance, 25 June 2019, www.thebalance.com/hurricane-harvev-facts-damage-costs-
4150087
[38] US Department of Energy (US DOE). 2018. Hurricane Florence situation reports.
Energy.gov, www.energv.gov/ceser/downloads/hurricane-florence-situation-reports
[39] North Carolina Department of Transportation (NCDOT). 2018. NCDOT honored for
Hurricane Florence response, www.ncdot.gov/news/press-releases/Pages/2018/2018-11-29-
ncdot-honored-hurricane-florence-response.aspx
[40] Federal Communications Commission (FCC). 2018. Communications status report for
areas impacted by Hurricane Florence. https://docs.fcc.gov/public/attachments/DOC-
354106Al.pdf
[41] How Hospitals Scrambled to Respond to Hurricane Florence-and 5 Ways to Mitigate the
Health Impacts of a Disaster. 2018. Advisory Board Daily Briefing,
www, advi sory. com/dailv-briefing/2018/09/17/florence
[42] American Hospital Directory, Inc. 2018. Individual hospital statistics for North Carolina.
American Hospital Directory - Individual Hospital Statistics for North Carolina,
www.ahd.com/states/hospital NC.html
[43] U.S. General Services Administration, Office of Facilities Management. (US GSA). 2020.
Facilities Status and Information Overview. GSA, 20 Mar. 2020,
www.gsa.gov/governmentwide-initiatives/emergencv-response/facilities-status-and-
information-overview
[44] US Department of Agriculture (USDA). 2019. CropScape and Cropland data layers.
USD A, http s://nassgeodata. gmu. edu/C rop S cape/
71
-------
[45] US Department of Homeland Security (US DHS). 2017. ArcGIS Hub. ArcGIS Hub,
http s: //hub. arcgi s. com/
[46] Riedman D. 2016. Questioning the criticality of critical infrastructure: A case study
analysis. Homeland Security Affairs 12, Essay 3,
May 2016, https://www.hsai.org/articles/10578
[47] Miller RI. 1969. Hurricane Katrina: Communications & infrastructure impacts. Homeland
Security Digital Library, Army War College (U.S.). Center for Strategic Leadership, 1 Jan.
1969, www.hsdl.org/?abstract&did=236103
[48] Harrison K, Levitan M. 2018. NWIRP Study of Hurricane Maria impacts and recovery in
Puerto Rico: Preliminary project plan for evaluating infrastructure support of critical
buildings and emergency communications. NCST Advisory Committee Meeting, 30 Aug.
2018.
[49] Krishnamurthy V, Kwasinski A. 2016. Modeling of communication systems dependency
on electric power during nuclear attacks. 2016 IEEE International Telecommunications
Energy Conference (INTELEC), pp. 1-8, http://dx.doi.org/10.1109/INTLEC.2016.7749029
[50] Murdock H, de Bruijn K, Gersonius B. 2018. Assessment of critical infrastructure
resilience to flooding using a response curve approach. Sustainability. 10(10):3470,
doi:10.3390/sul 0103470
[51] Arab A, Khodaei A, Khator S, Ding K, Emesih V, Han Z. 2015. Stochastic pre-hurricane
restoration planning for electric power systems infrastructure. IEEE Trans on Smart Grid.
6(2): 1046-1054, doi:10.1109/tsg.2015.2388736
[52] Barabadi A, Ayele YZ. 2018. Post-Disaster Infrastructure recovery: Prediction of recovery
rate using historical data. Reliab Engin Syst Safe. 169:209-223,
doi:10.1016/j.ress.2017.08.018
[53] Bellagamba X, Bradley BA, Wotherspoon LM, Lagrava WD. 2019. A decision-support
algorithm for post-earthquake water services recovery and its application to the 22
February 2011 Mw6.2 Christchurch Earthquake. Earthquake Spectra. 35(3): 1397-1420,
doi: 10.1193/052218eqs 119m
[54] Canter DA, Gunning D, Rodgers P, O'Connor L, Trauneo C, Kempter C. 2005.
Remediation of Bacillus cmthracis contamination in the U.S. Department of Justice Mail
Facility. Biosecur Bioterror. 3(2): 119—127, doi:10.1089/bsp.2005.3.119
[55] Choi J, Yoo DG, Kang D. 2018. Post-earthquake restoration simulation model for water
supply networks. Sustainability. 10(10):3618, doi:10.3390/sul0103618
[56] Greenberg M, Mantell N, Lahr M, Felder F, Zimmerman RS, et al. 2007. Short and
intermediate economic impacts of a terrorist-initiated loss of electric power: Case study of
New Jersey. Energy Policy. 35(l):722-733, doi:10.1016/j.enpol.2006.01.017
72
-------
[57] Hallegatte S. 2008. An adaptive regional input-output model and its application to the
assessment of the economic cost of Katrina. Risk Anal. 28(3):779-799, doi: 10.1111/j .1539-
6924.2008.01046.x
[58] Lau E, Chai KK, Chen Y, Loo J. 2018. Efficient economic and resilience-based
optimization for disaster recovery management of critical infrastructures. Energies.
11 (12):3418, doi: 10.3390/enl 1123418
[59] Matisziw TC, Murray AT, Grubesic TH. 2009. Strategic network restoration. Netw Spat
Econ. 10(3):345—361, doi:10.1007/sll067-009-9123-x
[60] Miles SB, Jagielo N, Gallagher H. 2016. Hurricane Isaac power outage impacts and
restoration. J Infrastruct Syst. 22(1):05015005, doi: 10.1061/(asce)is.l943-555x.0000267
[61] Panteli M, Pickering C, Wilkinson S, Dawson R, Mancarella P. 2017. Power system
resilience to extreme weather: Fragility modeling, probabilistic impact assessment, and
adaptation measures. IEEE T Power Syst. 32(5):3747-3757,
doi: 10.1109/tpwrs.2016.2641463
[62] Pregnolato M, Ford A, Wilkinson SM, Dawson RJ. 2017. The impact of flooding on road
transport: A depth-disruption function. Transp Res PartD: Transp Environ. 55:67-81,
doi:10.1016/j.trd.2017.06.020
[63] Yi P, George SK, Paul JA, Lin L. 2010. Hospital capacity planning for disaster emergency
management. Socio-EconPlan Sci. 44(3): 151—160, doi:10.1016/j.seps.2009.11.002
[64] Zhu J-J, Anderson PR. 2017. Effect of long-term successive storm flows on water
reclamation plant resilience. Water Res. 111:1-9, doi: 10.1016/j.watres.2016.12.026
73
-------
APPENDIX A: MATRIX OF REVIEWED LITERATURE
Table 28 lists the books and journal articles that were evaluated in the literature review. Included
is an abbreviated citation of each source.
Table 28: Matrix of Reviewed Literature
Abbreviated Citation
Chemical
Commercial Facilities
Communications
Critical Manufacturing
Dams
Defense Industrial Base
Emergency Services
&
=
W
Financial Services
Food and
Agriculture
Government
Facilities
Healthcareand
Public Health
Information
Technology
Nuclear Reactors
Transportation
Systems
Water and
Wastewater Systems
Haimes, Yacov Y.,
2018 [21
Y
Y
Y
Y
Y
Y
Y
Rehak, David, et al.,
2016 [61
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Laug, Ana, Josune
Hemantes, and Jose
M. Sarriegi. [10]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Lam, C. Y., and K.
Tai., 2015 [81
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Milanovi, Jovica V.,
and Wentao Zhu.,
2018122]
Y
Y
Y
Bhamidipati, Srimma 1
et al., 2016 [20]
Y
Y
Y
Y
Y
Y
Y
Thompson, James R
et al., 2019 [211
Y
Y
Y
Y
Chappin, Emile JL,
and Telli van der Lei,
2014 [111
Y
Y
Y
Y
Y
Bollinger, L. A., et al.,
2014 [121
Y
Y
Y
Y
Roe, Emery and
Schulman, Paul, 2016
[13]
Y
Y
Y
Y
Y
Presidential Policy
Directive PPD-21,
2013 [141
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
National Institute of
Standards and
Technology, 2014 [15]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Rehak, David, et al.,
2019. [71
Y
Y
Y
Y
Y
Y
Y
Wang, Weiping, et al.,
2018 [91
Y
Y
Y
Y
Y
Y
Y
Qmellaro, Gian
Paolo, et al., 2019 [5]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Sum of Papers
Covering
Infrastructure
3
5
11
3
5
6
7
15
8
7
9
9
9
3
13
14
74
-------
APPENDIX B: COORDINATES OF SCENARIOS 1 AND 2
Notes: All coordinates are in decimal form and not HMS. Coordinates are formatted as
'longitude latitude', with °N and °E being the respective positive direction. Denver Metropolitan
Area contains about 7,000 vertices that are not listed here, but the area consists of Adams,
Arapahoe, Broomfield, Clear Creek, Denver, Douglas, Elbert, Gilpin, Jefferson, and Park
counties.
Scenario 1
-105.014573 39.740366, -105.017921 39.744788, -105.008350 39.749325, -104.98730039.754847,
-104.940690 39.762339, -104.887132 39.771642, -104.820446 39.778402, -104.74955539.773528,
-104.618435 39.765803, -104.617984 39.715585, -104.718921 39.724828, -104.825378 39.732065,
-104.886896 39.726778, -104.941095 39.725578, -104.986242 39.727201, -105.01284939.730766,
-105.012678 39.733604, -105.016197 39.733538, -105.014573 39.740366
Scenario 2
-104.790911 39.706530, -104.791254 39.714850, -104.751686 39.722838, -104.71915639.725743,
-104.678301 39.726403, -104.640764 39.726437, -104.603170 39.725183, -104.60297639.678482,
-104.678542 39.688100, -104.697253 39.690147, -104.719397 39.692921, -104.788920 39.703141,
-104.789435 39.706245, -104.790911 39.706530
Scenario 1 + Scenario 2
-104.7889133 39.70308688, -104.7894554 39.70633943, -104.7908558 39.70652013,
-104.791217239.71478704, -104.7516445 39.72287325, -104.7269758 39.72505995,
-104.8253690 39.73199847, -104.887077239.72657754, -104.9411961 39.72558371,
-104.9862801 39.72711964, -105.0128426 39.73064324, -105.012661939.73344404,
-105.0160951 39.73362474, -105.0145592 39.74049124, -105.0179021 39.74464728,
-105.008415539.74934542, -104.9873643 39.75494704, -104.9406540 39.76244598,
-104.8871675 39.77166155, -104.820490239.77834735, -104.7494761 39.77346852,
-104.6184705 39.76569853, -104.6181094 39.72569214, -104.603111239.72522231,
-104.6030209 39.67851203, -104.6784620 39.68799865, -104.6972546 39.69016701,
-104.719299639.69269678, -104.7889133 39.70308688
75
-------
APPENDIX C: OCCUPANCY CLASS DEFINITIONS IN
HAZUS DATABASE
Table 29: Key of HAZUS Occupancy Classes, Page 1 of 4 [27]
Label
Occupancy Class
SIC
Code(s)
SIC Code Definition
RE S I
Single Family Dwelling
RES2
Mobile Home
RES3
Multi Family Dwelling
RES4
Temporary Lodging
70
Hotels, Rooming Houses, Camps, And Other
Lodging Places
RES 5
Institutional Dormitory
RES6
Nursing Home
8051
8052
8059
Skilled Nursing Care Facilities
Intermediate Care Facilities
Nursing And Personal Care Facilities, Not
Elsewhere Classified
52
Building Materials, Hardware, Garden
Supply, And Mobile Home Dealers
COM1
Retail Trade
53
54
55
56
57
59
General Merchandise Stores
Food Stores
Automotive Dealers and Gasoline Service
Stations
Apparel and Accessory Stores
Home Furniture, Furnishings, And
Equipment Stores
Miscellaneous Retail
COM2
Wholesale Trade
42
50
51
Motor Freight Transportation and
Warehousing
Wholesale Trade -durable Goods
Wholesale Trade -non-durable Goods
COM3
Personal/ Repair Services
72
75
76
83
88
Personal Services
Automotive Repair, Services, And Parking
Miscellaneous Repair Services
Social Services
Private Households
76
-------
Table 29: Key of HAZUS Occupancy Classes, Page 2 of 4 [27]
Label
Occupancy Class
SIC
Code(s)
SIC Code Definition
COM4
Prof./ Technical
Services
40
41
44
45
46
47
49
61
62
63
64
65
67
73
78 (except
7832)
81
87
89
Railroad Transportation
Local And Suburban Transit And Interurban
Highway Passenger Transportation
Water Transportation
Transportation By Air
Pipelines, Except Natural Gas
Transportation Services
Electric, Gas, And Sanitary Services
Non-depository Credit Institutions
Security And Commodity Brokers, Dealers,
Exchanges, And Services
Insurance Carriers
Insurance Agents, Brokers, And Service
Real Estate
Holding And Other Investment Offices
Business Services
Motion Pictures
Legal Services
Engineering, Accounting, Research,
Management, And Related Services
Miscellaneous Services
COM5
Banks
60
Depository Institutions
COM6
Hospital
8062
8063
8069
General Medical and Surgical Hospitals
Psychiatric Hospitals
Specialty Hospitals, Except Psychiatric
80
(except
Health Services
COM7
Medical Office/ Clinic
8051,
8052, 8059,
8062, 8063,
8069)
COM8
Entertainment & Rec.
48
58
79 (except
7911)
84
Communications
Eating And Drinking Places
Amusement And Recreation Services
Museums, Art Galleries, And Botanical And
Zoological Gardens
COM9
Theaters
7832
7911
Motion Picture Theaters, Except Drive-In
Dance Studios, Schools, and Halls
COM 10
Parking
77
-------
Table 29: Key of HAZUS Occupancy Classes, Page 3 of 4 [27]
Label
Occupancy
Class
SIC
Code(s)
SIC Code Definition
22
Textile Mill Products
24
Lumber And Wood Products, Except Furniture
26
Paper And Allied Products
32
Stone, Clay, Glass, And Concrete Products
IND1
Heavy
34
35 (except 3571,
3572)
37
Fabricated Metal Products, Except Machinery
And Transportation Equipment
Industrial And Commercial Machinery And
Computer Equipment
Transportation Equipment
23
Apparel And Other Finished Products Made
From Fabrics And Similar Materials
25
Furniture And Fixtures
27
30
Printing, Publishing, And Allied Industries
Rubber And Miscellaneous Plastic Products
IND2
Light
31
Leather And Leather Products
36 (except 3671,
3672, 3674)
38
39
Electronic And Other Electrical Equipment And
Components, Except Computer Equipment
Measuring, Analyzing, And Controlling Instruments;
Photographic, Medical And Optical
Goods; Watches And Clocks
Miscellaneous Manufacturing Industries
20
Food And Kindred Products
IND3
Food/Drugs/
Chemicals
21
28
29
Tobacco Products
Chemicals And Allied Products
Petroleum Refining And Related Industries
10
Metal Mining
12
Coal Mining
IND4
Metals/Minerals
13
Oil And Gas Extraction
Processing
14
33
Mining And Quarrying Of Nonmetallic
Minerals, Except Fuels
Primary Metal Industries
3571
Electronic Computers
IND5
High Technology
3572
3671
3672
3674
Computer Storage Devices
Electron Tubes
Printed Circuit Boards
Semiconductors and Related Devices
15
Building Construction General Contractors And
Operative Builders
IND6
Construction
16
17
Heavy Construction Other Than Building
Construction Contractors
Construction Special Trade Contractors
78
-------
Table 29: Key of HAZUS Occupancy Classes, Page 4 of 4 [27]
Label
Occupancy Class
SIC
Code(s)
SIC Code Definition
AGR1
Agriculture
01
02
07
08
09
Agricultural Production Crops
Agricultural Production Livestock and
Animal Specialties
Agricultural Services
Forestry
Fishing, Hunting, and Trapping
RE L I
Church/ N.P. Offices
86
Membership Organizations
GOV1
General Services
43
91
92 (except
9221,
9224)
93
94
95
96
97
United States Postal Service
Executive, Legislative, And General
Government, Except Finance
Justice, Public Order, And Safety
Public Finance, Taxation, And Monetary
Policy
Administration Of Human Resource
Programs
Administration Of Environmental Quality
And Housing Programs
Administration Of Economic Programs
National Security And International Affairs
GOV2
Emergency Response
9221
9224
Police Protection
Fire Protection
EDU1
Schools
82
(except 8221,
8222)
Educational Services
8221
8222
Colleges, Universities, and Professional
Schools
Junior Colleges and Technical Institutes
79
-------
APPENDIX D: SENSITIVITY ANALYSIS RESULTS
The following charts (Figure 31-Figure 39) contain the results of the repair factor analyses.
Water Recovery Time (days)
ooooooooo
Repair Factor Effect on Water
> Water
• Energy
• Transport
Communications
—•—Government
—•— Food and Agriculture
• Emergency Services
• Waste Management
) 0.2 0.4 0.6 0.8 j
Repair Factor of Infrastructure Sectors
—•—Healthcare
L
Figure 31: Infrastructure repair factors effects on Water recovery.
200
__ 180
"oo"
160
73
100
>
0
S 80
SB 60
1 40
LXJ
20
0
(
Repair Factor Effect on Energy
"V
> Water
—•— Energy
—•—Transport
Communications
—•— Government
• Food and Agriculture
—•— Emergency Services
—•—Waste Management
—•— Healthcare
L
) 0.2 0.4 0.6 0.8 :
Repair Factor of Infrastructure Sectors
Figure 32: Infrastructure repair factors effects on Energy recovery.
80
-------
Repair Factor Effect on Transportation Systems
160
00
>•
CD
~o
140
o
u
-
o
u
-------
_ 50
ro 45
-a
m 40
£
H 35
£~
u 30
Repair Factor Effect on Food and Agriculture
• Water
—•—Energy
* Transport
o
S 25
cc
£ 20
lis
u
M 10
<
"O c
Communications
• Government
• Food and Agriculture
• Emergency Services
fC
"§ 0
£ <
• Waste Management
• Healthcare
) 0.2 0.4 0.6 0.8 1
Repair Factor of Infrastructure Sectors
Figure 35: Infrastructure repair factors effects on Food and Agriculture recovery.
__ 30
>-
ro
— 25
-20
o>
>
o
& 15
en
10
-
y c
Repair Factor Effect on Emergency Services
• Water
• Energy
—•—Transport
Communications
• Government
~ Food and Agriculture
—•—Emergency Services
i— D
-------
Repair Factor Effect on Government Facilities
40
>•
03
"O
35
25
>
O
u
0J
20
Cd
>•
15
<3
re
u_
10
c
a>
E
5
c
QJ
>
0
o
0.2 0.4 0.6 0.8
Repair Factor of Infrastructure Sectors
•Water
- Energy
Transport
Communications
Government
-Food and Agriculture
-Emergency Services
-Waste Management
- Healthcare
Figure 37: Infrastructure repair factors effects on Government recovery.
Repair Factor Effect on Waste Management
_ 40
'vT
3 35
Ol
E
30
¦- 25
>
o
£ 20
a:
£ 15
£
ao 10
tc
c
TO
2 5
aj
tS o
ro u
3
0.2 0.4 0.6 0.8
Repair Factor of Infrastructure Sectors
-Water
-Energy
-Transport
Communications
Government
- Food and Agriculture
-Emergency Services
-Waste Management
-Healthcare
Figure 38: Infrastructure repair factors effects on Waste Management recovery.
83
-------
50
VT 45
I 40
-------
120
110
100
90
80
70
60
120
110
100
90
80
70
60
Water and Wastewater System
10%
50%
90%
4 6 8
Days of Backup Water
10 12
Figure 40: Recovery time of Water with backup Water.
Water and Wastewater System
10%
50%
90%
2 4 6 8 10 12
Days of Backup Energy
Figure 41: Recovery time of Water with backup Energy.
85
-------
Figure 42: Recovery time of Energy with backup Water.
1 /in
Energy
_ 135
i/>
ra 130
~
cu 125
p 120
w 115
§ no
^ 105
100
C
I
w
—•—10%
—•—50%
-•-90%
2
) 2 4 6 8 10 1
Days of Backup Energy
Figure 43: Recovery time of Energy with backup Energy.
86
-------
120
Transportation
^ no
>-
60
0 2 4 6 8 10 12
Days of Backup Water
Figure 44: Recovery time of Transportation with backup Water.
120
no
>
OJ
£. ioo
(D
p 90
S 80
o
u
CD
££ 70
60
C
Transportation
-•-10%
—•—50%
•
•
—•—90%
) 2 4 6 8 10 12
Days of Backup Energy
Figure 45: Recovery time of Transportation with backup Energy.
87
-------
120
Communications
^ no
>-
60
0 2 4 6 8 10 12
Days of Backup Water
Figure 46: Recovery time of Communications with backup Water.
Communications
120
-ST 110
>-
ro
60
0 2 4 6 8 10 12
Days of Backup Energy
Figure 47: Recovery time of Communications with backup Energy.
88
-------
Government
40
_ 35
Kf)
ra 30
~
o> 25
F 20
a 15
§10
^ 5
10%
50%
90%
2 4 6 8
Days of Backup Water
10
12
Figure 48: Recovery time of Government with backup Water.
>-
ro
Q
CD
E
40
35
30
25
20
15
l 10
Government
10%
50%
90%
4 6 8
Days of Backup Energy
10
12
Figure 49: Recovery time of Government with backup Energy.
89
-------
Food and Agriculture
l/>
>-
nj
Q
OJ
E
40
35
30
25
20
15
QJ
l 10
^ 5
- 1
•
•
: t
10%
50%
90%
4 6 8
Days of Backup Water
10
12
Figure 50: Recovery time of Food and Agriculture with backup Water.
>-
ro
Q
CD
E
40
35
30
25
20
15
QJ
l 10
^ 5
Food and Agriculture
10%
50%
90%
2 4 6 8
Days of Backup Energy
10
12
Figure 51: Recovery time of Food and Agriculture with backup Energy.
90
-------
40
_ 35
to
ra 30
~
o> 25
>
a5 15
§ 10
^ 5
0
c
Emergency Services
•
•
-•-10%
—•—50%
—•—90%
) 2 4 6 8 10 12
Days of Backup Water
Figure 52: Recovery time of Emergency Services with backup Water.
/in
Emergency Services
_ 35
to
ra 30
Q
0) 25
I20
>-
-------
Waste Management
-•-1 o%
—•—50%
—•—90%
2 4 6 8 10 12
Days of Backup Water
Figure 54: Recovery time of Waste Management with backup Water
Water and Wastewater System
120
110
I/)
>-
m
IU
100
CD
E
i—
T 1
90
•
#
-•-10%
CD
>
O
u
CD
rv
80
70
—•—50%
—•—90%
60
0 2 4 6 8 10
12
Days of Backup Energy
Figure 55: Recovery time of Water with backup Energy.
92
-------
50
_ 45
If)
ra 40
£3
aj 35
jz 30
>
a> 25
I 20
^ 15
10
C
Healthcare
-•
-•-10%
—•—50%
#
—•—90%
) 2 4 6 8 10 12
Days of Backup Water
Figure 56: Recovery time of Healthcare with backup Water.
50
_ 45
I/)
ra 40
cu 35
p 30
>
iu 25
§ 20
^ 15
10
C
Healthcare
-•-10%
—•—50%
——*
—•—90%
) 2 4 6 8 10 12
Days of Backup Energy
Figure 57: Recovery time of Healthcare with backup Energy.
93
-------
vvEPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
------- |