Received: 19 December 2024 Accepted: 9 February 2025

DOI: 10.1111/mice.13445

RESEARCH ARTICLE

Multi-hazard probabilistic risk assessment and equitable

multi-objective optimization of building retrofit strategies
in hurricane-vulnerable communities

Abdullah M. Braik1 Himadri Sen Gupta2 Maria Koliou1 Andrés D. González2

1 ZachryDepartment of Civil and

Environmental Engineering, Texas A&M Abstract
University, College Station, Texas, USA Coastal communities are increasingly vulnerable to hurricanes, which cause
2 School
of Industrial and Systems billions of dollars in damage annually through wind, storm surge, and flood-
Engineering, University of Oklahoma,
Norman, Oklahoma, USA ing. Mitigation efforts are essential to reduce these impacts but face significant
challenges, including uncertainties in hazard prediction, damage estimation,
Correspondence
and recovery costs. Resource constraints and the disproportionate burden borne
Maria Koliou, Zachry Department of Civil
and Environmental Engineering, Texas by socioeconomically vulnerable groups further complicate retrofitting strate-
A&M University, College Station, TX gies. This study presents a probabilistic methodology to assess and mitigate
77843, USA.
hurricane risks by integrating hazard analysis, building fragility, and economic
Email: [email protected]
loss assessment. The methodology prioritizes retrofitting strategies using a
Funding information risk-informed, equity-focused approach. Multi-objective optimization balances
US National Science Foundation (NSF),
Grant/Award Number: 2052930; National
cost-effectiveness and risk reduction while promoting fair resource allocation
Institute of Standards and Technology among socioeconomic groups. The novelty of this study lies in its direct inte-
(NIST), Grant/Award Number: gration of equity as an objective in resource allocation through multi-objective
70NANB20H008
optimization, its comprehensive consideration of multi-hazard risks, its inclu-
sion of both direct and indirect losses in cost assessments, and its use of
probabilistic hazard analysis to incorporate varying time horizons. A case study
of the Galveston testbed demonstrates the methodology’s potential to minimize
damage and foster equitable resilience. Analysis of budget scenarios and trade-
offs between cost and equity underscores the importance of comprehensive loss
assessments and equity considerations in mitigation and resilience planning. Key
findings highlight the varied effectiveness of retrofitting strategies across dif-
ferent budgets and time horizons, the necessity of addressing both direct and
indirect losses, and the importance of multi-hazard considerations for accurate
risk assessments. Multi-objective optimization underscores that equitable solu-
tions are achievable even under constrained budgets. Beyond a certain point,
achieving equity does not necessarily increase expected losses, demonstrating

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium,
provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2025 The Author(s). Computer-Aided Civil and Infrastructure Engineering published by Wiley Periodicals LLC on behalf of Editor.

2074 wileyonlinelibrary.com/journal/mice Comput Aided Civ Inf. 2025;40:2074–2097.

 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2075

that more equitable solutions can be implemented without compromising overall

cost-effectiveness.

1 INTRODUCTION ings in the United States, with this percentage increasing

annually (NAHB, 2024).
Hurricanes are among the most devastating natural haz- Researchers have employed probabilistic fragility analy-
ards, inflicting widespread damage on infrastructures, ses for these buildings under wind hazards using various
communities, and economies. Between 2000 and 2018, methodologies, including empirical (e.g., Ibrahim et al.,
the United States experienced 21 hurricanes that incurred 2024), experimental (e.g., Ling et al., 2024; Lopez et al.,
damages exceeding $1 billion each. Notably, Hurricane 2020), and analytical methods (e.g., Abdelhady et al., 2022;
Katrina in 2005 and Hurricane Harvey in 2017 inflicted Li & Ellingwood, 2006; Memari et al., 2018). Addition-
damages that each surpassed $100 billion (NHC, 2018). ally, studies have explored the effectiveness of retrofitting
These storms not only wreak immediate destruction strategies to mitigate vulnerabilities (e.g., Masoomi et al.,
through high winds, storm surges, and floods but also 2018; Yoo et al., 2024). Dong and Li (2016, 2017) extended
impose long-term economic impacts, displacing thousands these analyses to incorporate damage loss and recovery
of individuals, overwhelming local and federal disaster assessments.
response systems, and necessitating years of recovery Other studies have focused on fragility related to storm
efforts. Such catastrophic events highlight the critical need surge and wave effects (e.g., Do et al., 2020; Tomiczek et al.,
for enhanced hurricane preparedness, resilient infrastruc- 2014) and building fragility in flood scenarios (e.g., Nofal &
ture, and equitable recovery strategies to mitigate future van de Lindt, 2021; Nofal et al., 2020; van de Lindt & Tag-
impacts (Koliou et al., 2020). gart, 2009). Li et al. (2012) performed a multi-hazard loss
analysis at the building level, evaluating direct losses from
wind and flood damage while considering the influence of
1.1 Advancements in hurricane risk land-use planning mitigation strategies. Nofal et al. (2021)
assessment proposed a methodology for hurricane multi-hazard reli-
ability assessment of buildings subjected to wind, wave,
Research on community responses to hurricanes dates and flood damage, while Nofal et al. (2024) focused on
back to the 1960s (Moore et al., 1963), yet the inherent the flood hazard when assessing the functionality of build-
complexity of these hazards and the community struc- ings in post-disaster situations. At the community level,
ture as a system of systems pose significant challenges Nofal et al. (2023) expanded this research by evaluating
in accurately modeling resilience. The substantial epis- direct damage to buildings, electric power networks, and
temic and aleatory uncertainties associated with hurricane road infrastructure for a single hurricane scenario, fol-
loads, their interactions with structures, and the resultant lowed by a simplified recovery analysis based on heuristic
losses and recovery necessitate a probabilistic approach rules. Moreover, recent research has integrated advanced
within a risk assessment framework (Moehle & Deierlein, technologies like remote sensing and deep learning for
2004). With the development of Load and Resistance Fac- post-hurricane building damage assessment (e.g., Braik &
tor Design criterion in modern design codes, safety checks Koliou, 2024b; Cheng et al., 2021; Khajwal et al., 2023).
increasingly rely on probabilistic structural reliability the- However, these studies primarily focused on the damage
ory (Ellingwood, 1994). Consequently, performance-based and risk assessment phase without progressing into risk
engineering has gained traction for its ability to consider management, where optimization at both building and
broader social and economic impacts (Rosowsky & Elling- community levels is vital for informed decision-making
wood, 2002). This evolution has paved the way for com- regarding mitigation and retrofitting strategies.
prehensive frameworks prioritizing functional recovery,
re-occupancy, economic losses, and resilience, advancing
beyond traditional safety-focused objectives. 1.2 Optimization in disaster
Numerous studies have concentrated on low-rise wood management
residential buildings, developing fragility functions to eval-
uate their vulnerability to various hurricane hazards. Optimization plays a vital role in various civil engi-
These structures are pivotal to community resilience, as neering applications. Since the late 1990s, significant
they represent over 90% of the current residential build- advancements have been made in the cost optimization
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2076 BRAIK et al.

of structural systems, including reinforced concrete Despite these advancements, most previous studies
(Ahmadkhanlou & Adeli, 2005; Sarma & Adeli, 1998), were limited by their reliance on single-event (i.e., sin-
steel (Sarma & Adeli, 2000a, 2000b), prestressed concrete gle hurricane) scenarios or parametric analyses, without
(Sirca & Adeli, 2005), composite slabs (Kim & Adeli, 2001), integrating comprehensive probabilistic hazard modeling.
and high-rise buildings (Aldwaik & Adeli, 2014). More Additionally, their cost assessments frequently focused
recently, optimization techniques have expanded their exclusively on direct damage costs, overlooking the sub-
scope, addressing challenges in many-objective control stantial indirect costs associated with prolonged disrup-
optimization (Gutierrez Soto & Adeli, 2017), topology tions and restoration efforts (e.g., Gupta et al., 2022; Li
and shape optimization (Xia et al., 2021), transportation et al., 2012). Furthermore, multi-hazard hurricane risk
system optimization (Akhand et al., 2020; Bai et al., 2021; assessments that integrate the impacts of wind, waves, and
Zhao et al., 2021), and infrastructure restoration strategies flood have yet to evolve toward optimization frameworks
(González et al., 2017). that facilitate effective disaster management and informed
Recent studies have increasingly focused on optimizing decision-making.
resource allocation strategies across various types of haz-
ards (e.g., González et al., 2016; N. Zhang & Alipour, 2023; 1.3 Equity in disaster mitigation and
J. Zhang et al., 2023). Ostadi et al. (2021) proposed a model recovery
for optimizing resource allocation in disaster situations to
maximize resilience and business continuity. Optimization Another significant gap in the existing studies is the
techniques have also played a significant role in enhanc- rare consideration of the critical concept of equity, which
ing infrastructure resilience (Hosseini et al., 2016). For pertains to the fairness of resource allocation across dif-
example, Borghei and Ghassemi (2020) proposed a multi- ferent socioeconomic groups within a community. Recent
objective optimization scheme for microgrids’ resilient research has documented disparities in hardship experi-
and cost-effective planning, emphasizing preventive mea- ences among vulnerable populations following hazards
sures such as system hardening and substation elevation. (Coleman et al., 2020, 2023; Highfield et al., 2014; Van
Tiong (2022) developed a resilience optimization model to Zandt, 2019). The integration of fairness in optimization
evaluate interdependent critical infrastructure networks, gained traction as researchers recognized that efficiency-
addressing trade-offs between resilience enhancement and focused models often exacerbate existing inequalities
cost. (Chen & Hooker, 2021; Rodríguez-González, 2024). Tra-
In the context of earthquake mitigation, W. Zhang ditionally, resource allocation models aimed to minimize
and Nicholson (2016) proposed a multi-objective opti- economic losses, yet failed to consider the uneven impacts
mization model designed to support decision-making for of disasters on socioeconomically disadvantaged groups.
retrofitting buildings within communities by balancing This gap has driven interest in frameworks that embed
trade-offs between reducing direct economic losses and equity considerations, with the Gini index emerging as
minimizing population displacement. J. Zhang et al. (2023) a key measure of fairness (Coleman et al., 2023; Dhakal
developed a multi-objective optimization framework for & Zhang, 2023; Seyedashraf et al., 2022). By quantify-
scheduling community recovery and evaluating resilience ing resource disparities across socioeconomic groups, the
following earthquakes, with a focus on balancing perfor- Gini index facilitates the design of solutions that address
mance goals. Gomez and Baker (2019) introduced an opti- systemic inequities.
mization framework to enhance infrastructure resilience The inclusion of equity metrics in optimization extends
specifically against seismic hazards in transportation net- beyond theoretical modeling to actionable policy design.
works, addressing whether to retrofit or repair bridges For instance, Valinejad and Mili (2022) emphasized that
while minimizing costs and ensuring optimal network fairness in energy distribution fosters public trust, espe-
performance. In a broader context, Gupta, Alduri, et al. cially during crises. Similarly, Rahmattalabi et al. (2021)
(2024) developed a multi-objective optimization model highlighted how equitable interventions in disaster pre-
that accounts for earthquake and tsunami hazards, seek- paredness improved social cohesion. These studies demon-
ing to identify optimal retrofitting strategies that bolster strate the broader societal benefits of fairness-aware opti-
community resilience against these multi-faceted threats. mization, supporting the real-world applicability of the
Then, Gupta, Gonzalez, et al. (2024) also used these multi- proposed model. Achieving a seamless balance between
objective modeling techniques for evacuation planning for equity and efficiency remains challenging despite these
multi-hazard scenarios. Meanwhile, Gupta et al. (2022) advancements. Resource constraints often force trade-offs,
concentrated on flooding, proposing an optimal flood mit- requiring models to prioritize one objective over another.
igation strategy for buildings within a community, with Chen and Hooker (2021) addressed this by integrating fair-
a particular emphasis on minimizing expected economic ness as a binding constraint rather than an auxiliary goal,
losses resulting from flood events. ensuring that equity remains central to decision-making.
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2077

Some recent efforts have addressed equity in relief distri- equitable distribution of retrofit resources among socioe-
bution in the post-disaster context. For example, Soghrati conomic groups in hurricane-vulnerable communities.
Ghasbeh et al. (2022) proposed a multi-stage model for A comprehensive loss assessment framework that inte-
the location-distribution-routing problem in disaster grates probabilistic hazard, fragility, and cost analysis is
relief, utilizing a hybrid algorithm of adaptive large employed to estimate expected losses, with Monte Carlo
neighborhood search and multi-dimensional local search. simulation (MCS) used to model uncertainty. Optimiza-
Similarly, Wang and Sun (2022) developed a multi-period tion of retrofit strategies is performed at both the build-
allocation model for emergency resources that balances ing and system levels. The epsilon-constraint method is
efficiency and equity amid uncertain disaster conditions, utilized for the system level, where the objectives are
employing interval numbers and triangular fuzzy numbers loss reduction and equity. The methodology is presented
to represent uncertainties. Additionally, multi-objective through clear steps, each based on mathematical and prob-
methods aimed at balancing efficiency and fairness have abilistic models, to provide a practical approach that can be
been proposed to enhance the resilience of supply chains applied and built upon in future research. The applicabil-
and transportation networks (e.g., Abushaega et al., 2024; ity of the framework is demonstrated using the Galveston
Moshebah et al., 2024). In evacuation modeling, fairness testbed.
has been explored using agent-based modeling and dis-
crete event simulation techniques (e.g., Mostafizi et al.,
2 METHODOLOGY
2017; Oh et al., 2021), while socioeconomic factors were
integrated into recovery-focused agent-based models by
This section outlines the comprehensive probabilistic loss
Esmalian et al. (2022) and Aghababaei and Koliou (2023).
assessment methodology used to assess multi-hazard risks
Gupta, Gonzalez, et al. (2024) incorporated the concept
and optimize retrofit strategies for residential buildings
of fairness into an evacuation optimization model, where
in hurricane-vulnerable communities. The methodology
an objective function aimed to minimize the concentra-
integrates several key components: data collection and
tion of evacuees at a single node. Focusing on retrofit
analysis, probabilistic hazard modeling, vulnerability anal-
optimization, W. Zhang and Nicholson (2016) introduced
ysis, cost analysis, loss assessment, quantification of
constraints in their optimization model to ensure that
equity, and evaluation of retrofit options. Optimization is
retrofitting efforts did not exacerbate existing dispari-
then employed to support decision-making at both build-
ties between economic groups, particularly concerning
ing and system levels, ensuring that retrofit strategies are
population displacement in post-earthquake situations.
both effective and equitable. This integrated approach is
However, their framework did not directly facilitate equi-
designed to enhance resilience and facilitate recovery in
table optimization of resources distributed among various
communities exposed to the multifaceted risks associated
socioeconomic groups and lacked a quantitative basis for
with hurricanes. A flowchart of the proposed methodology
decision-making, such as utilizing metrics like the Gini
is presented in Figure 1, while details are discussed in the
index to assess and promote equity in resource allocation.
following sections.

1.4 Research gap and contribution 2.1 Data collection and analysis

In summary, despite significant advancements in disaster The methodology begins with the collection and analysis
risk assessment and optimization, existing research has of data related to residential buildings and their exposure
largely overlooked equity as a critical component in mit- to hurricane-induced hazards. These data encompass the
igation strategies. Furthermore, there is a notable lack of structural characteristics of buildings and socioeconomic
comprehensive hurricane multi-hazard risk assessments details critical for equity considerations. Building data
that integrate probabilistic hazard and fragility analy- include the geographical coordinates of each structure,
ses into the optimization process. Most studies rely on its elevation above ground level, the number of stories,
scenario-based approaches for single-return-period events, the floor area, and the structural archetype. These fac-
which limit their ability to account for varying storm inten- tors are essential for assessing vulnerability and losses
sities and time horizons. Additionally, disruption costs are to hazards such as wind, flood, and storm surge wave.
often excluded from loss assessments and tend to focus In addition to building characteristics, household data
solely on direct damage costs. are collected, which include the income level, race or
This study addresses these gaps by proposing a method- ethnicity of the occupants, and homeownership. These
ology that integrates probabilistic multi-hazard loss assess- socioeconomic indicators play a crucial role in ensuring
ments with multi-objective optimization, prioritizing the that equity is accounted for in the overall risk assess-
reduction of expected direct and indirect losses and the ment and decision-making process. For research purposes,
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2078 BRAIK et al.

FIGURE 1 Flowchart showing key steps of the proposed methodology.

testbeds provide an environment where researchers can To model hurricanes probabilistically, 𝑅𝑃𝑗 is defined
access relevant data and develop and test their method- as the RP of a hurricane of intensity 𝑗, where 𝑗 ranges
ologies, which can be accessed from platforms such as from 1 to 𝑚, the number of hurricane intensity categories.
IN-CORE (2024a) and DesignSafe (2024). The hurricanes are sorted from the smallest to the largest
RP. It is assumed that each year can experience only one
2.2 Probabilistic hazard analysis hurricane event, and the occurrence of hurricanes is inde-
pendent from year to year. This simplification allows to
In the proposed methodology, a probabilistic hazard anal- assign a probability of 1∕𝑅𝑃𝑗 that a hurricane of intensity
ysis is conducted to account for the multi-hazard nature of index at least 𝑗 will occur in any given year. Hence, the
hurricanes, which include wind, surge wave, and flooding. probability that a hurricane of intensity index between 𝑗
These hazards vary spatially across geographic regions, and 𝑗 + 1 will occur is given by the difference in probabil-
so a comprehensive hurricane model, such as the cou- ities between consecutive RPs, expressed in Equation (1).
pling of the Advanced Circulation and Simulating Waves
Nearshore models (Dietrich et al., 2011), is required to gen-
erate hazard maps for different return periods (RPs). The 𝑝𝑗 = 𝑅𝑃𝑗−1 − 𝑅𝑃𝑗+1
−1
(1)
goal is to capture the intensity of hurricanes over space and
their associated probabilities based on these RPs, providing For the highest intensity category, 𝑝𝑚 = 1∕𝑅𝑃𝑚 , and
a foundation for risk assessment. for the “no hurricane” scenario, which is associated with
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2079

intensities below 𝑅𝑃1 , the probability is: 𝑝0 = 1 − 𝑅𝑃1−1 . To tial. Such insights enable a balanced evaluation of short-
model the occurrence of hurricanes over a study period 𝑡𝑠 , term affordability and long-term resilience, empowering
the number of hurricanes in each intensity category fol- leaders to make informed choices that align with their
lows a multinomial distribution, expressed in Equation (2): community’s unique priorities and challenges.

(𝑁𝐻0 , 𝑁𝐻1 , 𝑁𝐻2 … , 𝑁𝐻𝑚 ) ∼ 𝑀𝑢𝑙𝑡𝑖𝑛𝑜𝑚𝑖𝑎𝑙

2.3 Probabilistic vulnerability analysis

∑
𝑚
This step of the methodology assesses the vulnerability
(𝑡𝑠 , 𝑝0 , 𝑝1 , 𝑝2 , … , 𝑝𝑚 ) , 𝑝𝑗 = 1 (2) of buildings to hurricane-induced hazards such as wind,
𝑗=0
surge wave, and flooding. The objective is to quantify the
The random variables 𝑁𝐻0 , 𝑁𝐻1 , 𝑁𝐻2 … , 𝑁𝐻𝑚 represent likelihood of damage to a structure based on hazard inten-
the number of years with no hurricane and hurricanes sity and structural archetype. Fragility functions are used
of intensity 𝑅𝑃1 through 𝑃𝑚 . Therefore, the probability to determine the probability that a structure will reach or
mass function of the probabilistic hazard model is given in exceed a given damage state, conditional on the intensity
Equations (3) and (4). of the hazard.
The effects of hurricanes on buildings are categorized
𝑃 (𝐻) = 𝑃 (𝑁𝐻0 = 𝑛𝐻0 , 𝑁𝐻1 = 𝑛𝐻1 , … , 𝑁𝐻𝑚 into two distinct vulnerability groups: (i) damage from
𝑡𝑠 ! ∏
𝑚
𝑛𝐻𝑗
wind and surge wave, representing the combined damag-
= 𝑛𝐻𝑚 |𝑡𝑠 , 𝑝0 , 𝑝1 , … , 𝑝𝑚 ) = ∏𝑚 𝑝𝑗 (3) ing influence, and (ii) flooding, representing the impact of
𝑗=0 𝑛𝐻𝑗 ! 𝑗=0 surge depths exceeding the building’s finished floor eleva-
tion (FFE). These classifications support a subsequent cost
subject to:
analysis, as defined by HAZUS Muti-Hazard methodology
∑
𝑚 ∑
𝑚 (FEMA, 2013), which recognizes that the costs associated
𝑛𝐻𝑗 = 𝑡𝑠 , 𝑝𝑗 = 1 (4) with wind damage differ from those caused by flooding.
𝑗=0 𝑗=0

where 𝑛𝐻0 , 𝑛𝐻1 , 𝑛𝐻2 , … , 𝑛𝐻𝑚 represent the sampled counts 2.3.1 Damage fragility
of hurricanes in each intensity category, and 𝑡𝑠 is the study
period in years. The probability that the damage 𝐷 exceeds a certain
Therefore, the probability distribution for the maximum threshold 𝑑𝑑 , given the wind velocity 𝑣, is modeled using
observed storm intensity over a specified study period 𝑡𝑠 a lognormal distribution as given in Equation (6).
can be expressed in Equation (5).
𝑃 (𝐷 ≥ 𝑑𝑑 |𝑣, 𝐴𝑟𝑐ℎ𝑊𝑉 ) = Φ ((ln (𝑣) − 𝜆𝑣 ) ∕𝜉𝑣 ) (6)
( )
𝑃 𝑀𝑎𝑥 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 = 𝑗 ′
where 𝐴𝑟𝑐ℎ𝑊𝑉 is the wind-wave archetype of the build-
⎧ 𝑝 𝑡𝑠 , 𝑗′ = 0 ing, 𝜆𝑣 , 𝜉𝑣 are parameters based on 𝐴𝑟𝑐ℎ𝑊𝑉 and 𝑑𝑑 , and
⎪ 0
= ⎨( ′ )𝑡𝑆 (∑ ′ )𝑡𝑆 (5) Φ(⋅) is the cumulative distribution function of the standard
∑𝑗 𝑗 −1
⎪ 𝑝 − 𝑝 , 𝑗′ > 0 normal distribution. In this study, the fragility parameters
⎩ 𝑗=0 𝑗 𝑗=0 𝑗
developed by Masoomi et al. (2018) are utilized. Moreover,
the archetypes proposed by the IN-CORE development
where 𝑗 ′ represents the index of the maximum observed
team (Memari et al., 2018) and mapped to the HAZUS
intensity.
(FEMA, 2009) archetypes are adopted.
The choice of the study period is a critical task entrusted
The probability that 𝐷 exceeds 𝑑𝑑 , given the storm
to decision-makers, and perspectives on this matter often
surge height 𝑠, significant wave height 𝑤, and the 𝐹𝐹𝐸 is
vary. From the authors’ meetings with community lead-
expressed in Equations (7) and (8).
ers and representatives, it became evident that opin-
ions diverged significantly. Some stakeholders prioritized 𝑃 (𝐷 ≥ 𝑑𝑑 |𝑠, 𝑤, 𝐹𝐹𝐸, 𝐴𝑟𝑐ℎ𝑊𝑉 ) = Γ (𝑠, 𝑤) ∕ (Γ (𝑠, 𝑤) + 1)
short-term benefits, expressing concerns that any plan- (7)
ning horizon beyond 20 years might be too distant to where
address immediate needs or garner sufficient support.
𝛽4
Conversely, others emphasized the importance of long- Γ (𝑠, 𝑤) = ((𝑠∕𝛽1 ) + (𝑤∕𝛽2 ) + 𝛽3 (𝑠∕𝛽1 ) (𝑤∕𝛽2 )) (8)
term sustainability, recognizing the value of investments
that ensure resilience for future generations. Providing and 𝛽1 , 𝛽2 , 𝛽3 , 𝛽4 are parameters based on
decision-makers with accurate predictions on the expected 𝐹𝐹𝐸, 𝐴𝑟𝑐ℎ𝑊𝑉 , and 𝑑𝑑 . In this study, the fragility
consequences of various study periods is therefore essen- parameters developed by Do et al. (2020) are utilized.
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2080 BRAIK et al.

Assuming that the damages from wind and surge wave are considered in this methodology: elevation of the
are independent, the probability that 𝐷 exceeds 𝑑𝑑 , building and roof retrofit. The elevation involves raising
conditional on 𝑣, 𝑠, and 𝑤, is given in Equation (9). the structure. This directly modifies the flood and wave
fragility functions by increasing the FFE value and con-
𝑃 (𝐷 ≥ 𝑑𝑑 |𝑠, 𝑤, 𝑣, 𝐹𝐹𝐸, 𝐴𝑟𝑐ℎ𝑊𝑉 ) sequently reducing the building’s exposure to flood and
wave damage. Roof retrofit, in contrast, strengthens the
= 𝑃 (𝐷 ≥ 𝑑𝑑 |𝑠, 𝑤, 𝐹𝐹𝐸, 𝐴𝑟𝑐ℎ𝑊𝑉 ) + 𝑃 (𝐷 ≥ 𝑑𝑑 |𝑣, 𝐴𝑟𝑐ℎ𝑊𝑉 )
structure’s resistance to high wind speeds by reinforc-
−𝑃 (𝐷 ≥ 𝑑𝑑 |𝑠, 𝑤, 𝐹𝐹𝐸, 𝐴𝑟𝑐ℎ𝑊𝑉 ) ing the roof. Two roof retrofit options are considered: the
first involves replacing the roof sheathing and adopting
𝑃 (𝐷 ≥ 𝑑𝑑 |𝑣, 𝐴𝑟𝑐ℎ𝑊𝑉 ) (9) a closer-spaced nailing pattern, while the second option
includes replacing the roof sheathing similar to the first
Therefore, the probability that 𝐷 will be exactly in 𝑑𝑑 is option and then adding a new clay tile roof cover. These
expressed in Equation (10). retrofitting measures adjust the wind fragility function
parameters 𝜆𝑣 , 𝜉𝑣 as outlined by Masoomi et al. (2018).
𝑃 (𝐷 = 𝑑𝑑 |𝑠, 𝑤, 𝑣, 𝐹𝐹𝐸, 𝐴𝑟𝑐ℎ𝑊𝑉 ) These fragility functions were chosen because, as previ-
ously discussed, they encompass a range of commonly
= 𝑃 (𝐷 ≥ 𝑑𝑑 |𝑠, 𝑤, 𝑣, 𝐹𝐹𝐸, 𝐴𝑟𝑐ℎ𝑆𝑊 , 𝐴𝑟𝑐ℎ𝑊𝑉 )
implemented mitigation measures and provide compre-
−𝑃 (𝐷 ≥ 𝑑𝑑+1 |𝑠, 𝑤, 𝑣, 𝐹𝐹𝐸, 𝐴𝑟𝑐ℎ𝑊𝑉 , 𝐴𝑟𝑐ℎ𝑉 ) (10) hensive parameters for both retrofitted and non-retrofitted
scenarios for various archetypes of structures. Although
The assumption of independence between wind and elevating the building could potentially increase wind vul-
surge damage is acknowledged as a simplification. In nerability due to a higher profile, this effect is excluded
reality, some degree of correlation likely exists. However, from the current methodology for simplicity.
modeling such interactions requires detailed hazard cou-
pling, which may not be feasible due to data limitations
and is outside the scope of this study. 2.4 Probabilistic cost analysis

The proposed methodology considers the impact of both

2.3.2 Flood fragility
direct and indirect costs associated with building damage
and flooding.
The probability that the flood impact on the structure 𝐹
exceeds a certain threshold 𝑓𝑓 , given the storm surge
height 𝑠, is represented by Equation (11).
( ) 2.4.1 Direct costs
𝑃 𝐹 ≥ 𝑓𝑓 |𝑠, 𝐴𝑟𝑐ℎ𝑓
{ (( ) ) To estimate direct costs from damage to buildings, the
Φ ln (𝑠 − 𝐹𝐹𝐸) − 𝜆𝑓 ∕𝜉𝑓 , 𝑠 > 𝐹𝐹𝐸 repair cost models developed by Koliou and van de Lindt
= (11)
0, 𝑠 ≤ 𝐹𝐹𝐸 (2020), along with the HAZUS- Multi-Hazard methodol-
ogy (FEMA, 2003), are employed. The loss, conditioned on
where 𝐴𝑟𝑐ℎ𝑓 is the flood archetype, and 𝜆𝑓 , 𝜉𝑓 are param- the damage state, is modeled using a lognormal distribu-
eters based on 𝐴𝑟𝑐ℎ𝑓 and 𝑓𝑓 . In this study, the fragility tion, as shown in Equation (13).
parameters and archetypes developed by Nofal et al. (2020)
and Nofal and van de Lindt (2021) are utilized. 𝐿𝑙,𝑑𝑖𝑟𝑒𝑐𝑡 |𝑑𝑑 , 𝐴𝑟𝑐ℎ𝑊𝑉 ∼ 𝐿𝑁 ( 𝜆𝑙 , 𝜉𝑙 ) (13)
Therefore, the probability that 𝐹 will be exactly in 𝑓𝑓 is
given in Equation (12). where 𝐿𝑙,𝑑𝑖𝑟𝑒𝑐𝑡 is the direct damage loss, and 𝜆𝑙 , 𝜉𝑙 are the
( ) ( ) parameters of the lognormal distribution for damage losses
𝑃 𝐹( = 𝑓𝑓 |𝑠, 𝐴𝑟𝑐ℎ𝑓 = )𝑃 𝐹 ≥ 𝑓𝑓 |𝑠, 𝐴𝑟𝑐ℎ𝑓 that depend on 𝑑𝑑 and 𝐴𝑟𝑐ℎ𝑊𝑉 , obtained from Koliou and
(12)
−𝑃 𝐹 ≥ 𝑓𝑓+1 |𝑠, 𝐴𝑟𝑐ℎ𝑓 van de Lindt (2020) and FEMA (2003).
Similarly, the direct costs from flooding are estimated
using cost models by Nofal et al. (2020). The flood-
ing loss, conditioned on the flooding state, is defined in
2.3.3 Retrofit
Equation (14).
To mitigate the vulnerability of buildings to hurricane-
𝐿𝑙′ ,𝑑𝑖𝑟𝑒𝑐𝑡 |𝑓𝑓 , 𝐴𝑟𝑐ℎ𝑓 ∼ 𝐿𝑁 ( 𝜆𝑙′ , 𝜉𝑙′ ) (14)
induced hazards, two primary types of retrofit strategies
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2081

where 𝐿𝑙′ ,𝑑𝑖𝑟𝑒𝑐𝑡 is the direct flooding loss and 𝜆𝑙′ , 𝜉𝑙′ are where 𝑇𝑑𝑟𝑦 , 𝑇𝑖𝑛𝑠 , 𝑇𝑓𝑖𝑛 , 𝑇𝑐𝑜𝑛 , and 𝑇𝑝𝑒𝑟 represent the dry-
the parameters of the lognormal distribution for flooding out, inspection, financing, contractor mobilization, and
losses that depend on 𝑓𝑓 and 𝐴𝑟𝑐ℎ𝑓 , obtained from Nofal permitting times, respectively. All these components are
et al. (2020). modeled as lognormal random variables.
Finally, the total loss is governed by the larger of the
damage and flooding total costs as demonstrated in the
2.4.2 Indirect costs subsequent loss analysis. This approach aligns with FEMA
(2013), which avoids double counting when a building is
Indirect costs are defined by HAZUS Multi- exposed to both damage and flooding during the same
Hazard (FEMA, 2013) as the sum of both the disruption hurricane.
cost and the rent cost, expressed in Equation (15).

𝐿𝑖𝑛𝑑𝑖𝑟𝑒𝑐𝑡 = 𝐿𝑑𝑖𝑠𝑟𝑢𝑝𝑡𝑖𝑜𝑛 + 𝐿𝑟𝑒𝑛𝑡 (15) 2.5 Probabilistic loss assessment

where 𝐿𝑑𝑖𝑠𝑟𝑢𝑝𝑡𝑖𝑜𝑛 is the disruption cost, based on FEMA The expected total loss for building 𝑧 belonging to the
(2013), while 𝐿𝑟𝑒𝑛𝑡 is the rent cost, as expressed in Equa- socioeconomic group 𝑞 and archetype 𝑟, given the retrofit
tion (16). interventions for roof and flooding 𝑘 , is calculated by
considering the potential impact across multiple hazard
𝐿𝑟𝑒𝑛𝑡 = 𝐿𝑟𝑎𝑡𝑒 𝑇𝑟𝑒𝑠𝑡 (16) scenarios. The total expected loss is the weighted sum of
potential losses under each hazard scenario ℎ, multiplied
where 𝐿𝑟𝑎𝑡𝑒 is the rental cost rate per unit time (FEMA, by the probability of that scenario occurring. Hence, all dif-
2013), and 𝑇𝑟𝑒𝑠𝑡 is the total restoration time, expressed as ferent damage scenarios 𝑑, flooding scenarios 𝑓, damage
Equation (17). loss scenarios 𝑙, and flood loss scenarios 𝑙 ′ are consid-
ered within the probabilistic model. Mathematically, this
𝑇𝑟𝑒𝑠𝑡 = 𝑇𝑟𝑒𝑝𝑎𝑖𝑟 + 𝑇𝑑𝑒𝑙𝑎𝑦 (17) is expressed as shown in Equation (21). In this context,
𝐻ℎ𝑧𝑞𝑟𝑘 represents the hazard with intensity ℎ, while 𝐷𝑑𝑧𝑞𝑟𝑘
where 𝑇𝑟𝑒𝑝𝑎𝑖𝑟 is the repair time. For damage, the repair and 𝐹𝑓𝑧𝑞𝑟𝑘 represent the damage state 𝑑 and flooding
time models developed by Koliou and van de Lindt (2020) state 𝑓, respectively, for building 𝑧 in socioeconomic group
are adopted. Hence, the damage repair time, conditioned 𝑞 and archetype 𝑟, under the specified retrofit interven-
on the damage state, follows a lognormal distribution and tions 𝑘. Since the formula does not have a closed-form
is expressed in Equation (18). solution, MCS is used as a numerical method to estimate
the solution.
𝑇𝑟𝑑 |𝑑𝑑 , 𝐴𝑟𝑐ℎ𝑊𝑉 ∼ 𝐿𝑁 ( 𝜆𝑟𝑑 , 𝜉𝑟𝑑 ) (18) (
∑ ( ) ∑∑( ( )
Ψ𝑧𝑞𝑟𝑘 = 𝑃 𝐻ℎ𝑧𝑞𝑟𝑘 max 𝑃 𝐷𝑑𝑧𝑞𝑟𝑘 |𝐻ℎ𝑧𝑞𝑟𝑘
where 𝑇𝑟𝑑 is the damage repair time, and 𝜆𝑟𝑑 , 𝜉𝑟𝑑 are ℎ 𝑑 𝑙
the parameters of the lognormal distribution for dam- ( ) ) ∑∑( ( )
age repair time that depend on 𝑑𝑑 and 𝐴𝑟𝑐ℎ𝑊𝑉 , obtained 𝑃 𝐿𝑙𝑧𝑞𝑟𝑘 |𝐷𝑑𝑧𝑞𝑟𝑘 𝐿𝑙𝑧𝑞𝑟𝑘 , 𝑃 𝐹𝑓𝑧𝑞𝑟𝑘 |𝐻ℎ𝑧𝑞𝑟𝑘
from Koliou and van de Lindt (2020). For flooding, the 𝑓 𝑙′

repair time ranges provided by FEMA (2013) are adopted. )

( ) )
Hence, the flood repair time, conditioned on the flooding 𝑃 𝐿𝑙′ 𝑧𝑞𝑟𝑘 |𝐹𝑓𝑧𝑞𝑟𝑘 𝐿𝑙′ 𝑧𝑞𝑟𝑘 (21)
state, follows a uniform distribution and is represented in
Equation (19).
( ) 2.6 Building-level optimization
𝑇𝑟𝑓 |𝑓𝑓 , 𝐴𝑟𝑐ℎ𝑓 ∼ 𝑈𝑛𝑖𝑓𝑜𝑟𝑚 𝑇𝑙𝑜𝑤𝑒𝑟 , 𝑇𝑢𝑝𝑝𝑒𝑟 (19)

In this step, a building-level optimization model for

where 𝑇𝑟𝑓 is the flooding repair time, and 𝑇𝑙𝑜𝑤𝑒𝑟 , 𝑇𝑢𝑝𝑝𝑒𝑟
retrofitting is developed. The primary objective is to iden-
denote the lower and upper bounds of the flooding repair
tify the optimal retrofitting option for each building that
time, respectively, as per FEMA (2013).
leads to the most cost-effective solution. A summary of
On the other hand, the delay time is defined in Equa-
the notation used in the optimization model is shown in
tion (20) (Almufti & Willford, 2013; FEMA, 2013).
the Appendix (i.e., sets, parameters, and decision variables,
( ) which are shown in Tables A1, A2, and A3, respec-
𝑇𝑑𝑒𝑙𝑎𝑦 = 𝑇𝑑𝑟𝑦 + 𝑇𝑖𝑛𝑠 + max 𝑇𝑓𝑖𝑛 , 𝑇𝑐𝑜𝑛 + 𝑇𝑝𝑒𝑟 (20) tively). The objective of the building-level optimization is
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2082 BRAIK et al.

expressed mathematically in Equation (22): even in scenarios with uneven distribution of buildings
∑( ) across groups. While other fairness measures, such as
minimize Ψ𝑧𝑞𝑟𝑘 + 𝑐𝑧𝑞𝑟𝑘 𝜚𝑧𝑞𝑟𝑘 , ∀𝑧 ∈  , ∀𝑞 ∈  , the Theil index or max–min fairness (Sampat & Zavala,
𝑘∈ 2019), could also be applied, the Gini index was cho-
∀𝑟 ∈  (22) sen for its ubiquity, simple interpretation, and its estab-
lished relevance in assessing disparities across groups.
This optimization is subject to a set of constraints. The Future research could explore comparisons between var-
constraints shown in Equation (23) are the mitigation strat- ious fairness measures, but this is beyond the scope of
egy balance, which ensures that each building is assigned the current study, which focuses exclusively on the Gini
exactly one mitigation strategy 𝑘. index.
The practical use of the Gini index in optimization is not
∑
𝜚𝑧𝑞𝑟𝑘 = 1, ∀𝑧 ∈  , ∀𝑞 ∈  , ∀𝑟 ∈  (23) without challenges. Its inherently non-linear nature com-
𝑘∈ plicates its direct incorporation into models. To address
this, researchers have developed approximations and sur-
The second set of constraints in Equation (24) presents rogate metrics that simplify the Gini index for linear or
the domain of the decision variables, which ensures that mixed integer programming frameworks. For instance,
𝜚𝑧𝑞𝑟𝑘 is binary. Fish et al. (2016) proposed confidence-based approaches
that maintained equity objectives while preserving com-
𝜚𝑧𝑞𝑟𝑘 ∈ {0, 1} , ∀𝑧 ∈  , ∀𝑞 ∈  , ∀𝑟 ∈  , ∀𝑘 ∈  (24) putational feasibility. These adaptations ensure the Gini
index can be effectively applied even in complex scenar-
ios, such as those involving diverse building archetypes
2.7 Equity analysis or retrofitting strategies, as outlined in the proposed
model. Dynamic optimization techniques have further
In the proposed system-level optimization model, the Gini enhanced fairness-focused resilience planning by adapt-
index is employed to evaluate disparities in retrofit assis- ing to changing community needs. Nozhati et al. (2020)
tance as discussed in the next section. The Gini index is a explored stochastic optimization in post-disaster recovery,
widely used statistical tool for measuring inequality, com- emphasizing the role of equity in restoring critical infras-
monly applied to study resource distribution. It ranges tructure. Their approach treated resource distribution
from 0 (representing perfect equality) to 1 (representing as a sequential decision-making problem under uncer-
maximum inequality), offering a quantitative measure of tainty, a methodology well-suited to capturing the evolving
how uniformly a resource is allocated within a population. nature of resilience needs. This perspective is particu-
The Gini index 𝐺 is calculated using Equation (25) (Sen, larly relevant to the proposed system-level optimization
1997; Weiner, 1985): model, where retrofitting strategies must evolve based on
socioeconomic group characteristics and budgetary shifts.
( 2 )−1 ∑
𝑇 ∑
𝑇
Building on the work by Chen and Hooker (2021), the
𝐺 = 2𝑇 𝑥̄ |𝑥𝜏 − 𝑥𝜏′ | (25)
𝜏=1 𝜏′ =1 Gini index in the proposed model serves as both a con-
straint and an objective to guarantee proportional resource
where in the proposed methodology, T is defined as the distribution.
number of groups, 𝑥̄ is the overall mean across all group
means (mean of the means), and (𝑥𝜏 − 𝑥𝜏′ ) is the pairwise
difference between two group means in the dataset. 2.8 System-level optimization
The Gini index was selected as the fairness metric in
this study due to its widespread use and interpretability Building-level optimization focuses on selecting the most
in measuring inequality across various domains, including cost-efficient retrofit strategy for each building individu-
economics and resource allocation. Its ability to quan- ally, aiming to minimize the expected total loss for each
tify disparities in resource distribution aligns with the building. In contrast, the system-level optimization prior-
objectives of disaster risk management, where equitable itizes buildings for repair at the community level, aiming
allocation of resources is critical for promoting resilience to minimize the overall community cost while adhering to
among vulnerable groups. By defining 𝑥𝜏 as the mean a limited retrofit budget.
allocation of resources for the group 𝜏, normalization A comprehensive multi-objective optimization model
is achieved based on the number of buildings within for retrofitting resource allocation in communities
each socioeconomic group. This approach ensures that the subjected to multiple hazards is developed. The pro-
proposed methodology remains applicable and effective, posed optimization model balances two main objectives:
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2083

reducing expected economic losses from hazards and level, the allocation constraints in Equation (31) calculate
ensuring fair distribution of resources across socioeco- the average allocation in each of socioeconomic groups,
nomic groups within a community using Gini index. The with the maximum allocation constraints in Equation (32)
model is designed to tailor mitigation measures to diverse capping allocations to individual buildings.
buildings and prioritize equitable outcomes. This is Feasibility is maintained through the resource feasi-
achieved through a feasibility constraint, which indicates bility constraints in Equation (33), linking allocations to
if a specific mitigation measure is feasible for a particular retrofitting strategies by considering transition costs and
building. Such a parameter increases the model’s practical feasibility conditions. Equity is explicitly modeled using
applicability by ensuring strategies align with real-world the deviation constraints in Equation (34), which cap-
constraints. Set with indices, parameters, and decision ture disparities in group-level allocations. These deviations
variables are listed in Tables A1, A2, and A3, respectively. are aggregated in the Gini index constraint in Equa-
tion (35), which represents the level of inequity in resource
2.8.1 System-level optimization objective allocation.
functions Structural consistency is ensured through the retrofit
balance constraints in Equation (36), which validates the
The first objective of the system-level optimization focuses transitions between retrofitting strategies, and the building
on reducing the total expected economic losses across all conservation constraints in Equation (37), which pre-
buildings within the community. By selecting optimal mit- serve the total number of buildings in the community. As
igation and retrofitting strategies, the model ensures that defined in Table A1, set  consists of pairs of mitigation
buildings are protected against potential hazards, thereby strategies (𝑘, 𝑘 ′ ), which represent valid changes from the
current strategy k to the target strategy k’. A pair (𝑘, 𝑘′ ) ∈
minimizing damage-related costs. This objective expressed
 indicates that transitioning from k to k’ is feasible.
in Equation (26), is critical for disaster risk reduction,
This is often interpreted as a transition between specific
as it directly targets the financial impact of hazards. It
mitigation strategies, such as moving from “No retrofit”
ensures that resources are allocated to mitigation measures
to “No elevation retrofit, replace roof sheathing” or from
that provide the highest economic benefit in terms of loss
“5 ft (1.5 m) elevation, replace roof sheathing” to “10 ft
reduction.
(3.0 m) elevation, replace roof sheathing and add clay tile
∑∑∑∑ cover.” However,  is not symmetric, meaning (𝑘, 𝑘′ ) ∈
minimize 𝑥𝑧𝑞𝑟𝑘 Ψ𝑧𝑞𝑟𝑘 (26)
𝑧∈ 𝑞∈ 𝑟∈ 𝑘∈  does not imply (𝑘′ , 𝑘) ∈ . For instance, transitioning
from “No elevation retrofit, replace roof sheathing” to “5 ft
The second objective of the system-level optimization (1.5 m) elevation, replace roof sheathing” may be feasible,
given in Equation (27), focuses on promoting equity by but transitioning back from “5 ft (1.5 m) elevation, replace
minimizing the Gini index, which measures inequality roof sheathing” to “No elevation retrofit, replace roof
in resource distribution among different socioeconomic sheathing” may not be practically feasible. In some sce-
groups. A lower Gini index indicates a more equitable narios, however, transitioning between two strategies may
allocation of resources. be feasible in both directions, depending on the practical
and financial considerations. For example, transitioning
minimize 𝐺 (27) from “5 ft (1.5 m) elevation, replace roof sheathing and
add clay tile cover” to “10 ft (3.0 m) elevation, replace roof
sheathing,” and vice versa, could be feasible under certain
2.8.2 System-level optimization constraints conditions.
This asymmetry reflects real-world constraints, such as
The optimization model employs a set of constraints to cost, structural limitations, or physical irreversibility of
balance the dual objectives of minimizing expected eco- certain retrofitting measures. The set  thus represents
nomic losses and ensuring equitable resource allocation. all feasible transitions, whether upgrades or alternative
The budget constraint expressed in Equation (28) ensures strategies, ensuring that the optimization model considers
that total resource allocation does not exceed the avail- only valid and meaningful changes in mitigation strate-
able budget. Allocations are structured hierarchically: the gies. Finally, the non-negativity and domain constraints
group-level constraints in Equation (29) compute resource (Equation 38) guarantee that all decision variables, includ-
allocations for each socioeconomic group, while the ing building counts, allocations, and deviations, remain
community-wide allocation constraints in Equation (30) feasible and non-negative. Together, these constraints pro-
aggregate these group-level allocations. At the building vide a rigorous and implementable framework for hazard
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2084 BRAIK et al.

resilience planning, effectively balancing efficiency, equity, manage this conflict, the proposed methodology employs
and practical feasibility. the epsilon-constraint method (Mavrotas, 2009), which
reformulates the multi-objective problem into a series of
Γ≤𝐵 (28)
single-objective optimization problems by treating one
∑ objective as the primary focus and the other as a constraint.
𝛼𝑞 = 𝛼𝑧𝑞 , ∀𝑞 ∈  (29) For each fixed value of 𝜖, the optimization minimizes
𝑧∈𝑞 the first objective while ensuring that the second remains
∑ below the specified threshold. By incrementally varying
Γ= 𝛼𝑞 (30)
𝑞∈
𝜖, a set of solutions is generated, each corresponding to
𝛼𝑞 a specific trade-off between the two objectives. This pro-
𝛼𝑞 = , ∀𝑞 ∈  (31) cess results in a Pareto front, visualizing the trade-offs
𝑁𝑞
and enabling decision-makers to choose solutions that best
𝛼𝑧 ≤ 𝑐max , ∀𝑧 ∈  (32) align with their priorities.
The original problem can be stated as in Equation (39).
∑∑ ∑
𝛼𝑧𝑞 = 𝜙𝑧𝑞𝑟𝑘 𝛽𝑧𝑞𝑟𝑘 𝑐𝑧𝑞𝑟𝑘𝑘′ 𝑦𝑧𝑞𝑟𝑘𝑘′ , min {𝑓1 (𝑥) , 𝑓2 (𝑥)} (39)
𝑟∈ 𝑘∈ 𝑘 ′ ∈∶(𝑘,𝑘 ′ )∈

∀𝑧 ∈ , ∀𝑞 ∈  (33) where 𝑓1 (𝑥) represents the total expected economic loss

(Equation 26) and 𝑓2 (𝑥) represents the Gini index mea-
| | suring inequity (Equation 27). Using the epsilon-constraint
Υ𝑞𝑞′ = |𝛼𝑞 − 𝛼𝑞′ | , ∀𝑞, 𝑞′ ∈  (34)
| | method, we transform this problem to Equation (40).
( )
∑ ∑ ∑ min {𝑓1 (𝑥)} subject to 𝑓2 (𝑥) ≤ 𝜖 (40)
𝐺 2𝑀 𝛼𝑞 ) = 𝛾𝑞𝑞′ (35)
𝑞∈𝑄 𝑞∈𝑄 𝑞′ ∈𝑄
where 𝜖 represents a series of threshold values for the Gini
∑
𝑥𝑧𝑞𝑟𝑘 = 𝜙𝑧𝑞𝑟𝑘 𝑦𝑧𝑞𝑟𝑘′ 𝑘 + 𝑏𝑧𝑞𝑟𝑘 index 𝐺, enabling the exploration of trade-offs between the
𝑘 ′ ∶(𝑘 ′ , 𝑘) ∈  two objectives. For each fixed value of 𝜖, the model mini-
∑ mizes expected economic loss while ensuring that inequity
− 𝜙𝑧𝑞𝑟𝑘 𝑦𝑧𝑞𝑟𝑘𝑘′ remains within acceptable bounds. To operationalize this
𝑘 ′ ∶(𝑘 ′ , 𝑘) ∈ 
method, the Gini index (Equation 35) is integrated as a
( )
∀ 𝑧 ∈ , ∀ 𝑟 ∈ , ∀ 𝑘, 𝑘′ ∈ , ∀𝑞 ∈  constraint, as expressed in Equation (41):

(36) 𝐺≤𝜖 (41)

∑ ∑ where 𝐺 is computed as a function of resource alloca-

𝑥𝑧𝑞𝑟𝑘 = 𝑏𝑧𝑞𝑟𝑘 , ∀𝑧 ∈ , ∀ 𝑟 ∈ , ∀𝑞 ∈ 
𝑘∈ 𝑘∈
tions among socioeconomic groups in constraints (35). By
(37) incrementally varying 𝜖, a Pareto-optimal front is gener-
ated, which highlights the trade-offs between economic
𝑥𝑧𝑞𝑟𝑘 ∈ ℤ≥0 , ∀ 𝑧 ∈ , efficiency and equity.
∀ 𝑟 ∈ , ∀𝑘 ∈ , ∀𝑞 ∈  ( ) The epsilon-constraint method ensures that the opti-
𝑦𝑧𝑞𝑟𝑘𝑘′ ∈ ℤ≥0 , ∀ 𝑧 ∈ , ∀ 𝑟 ∈ , ∀ 𝑘, 𝑘′ ∈ , mization remains tractable and interpretable. Each result-
∀𝑞 ∈  ing solution corresponds to a feasible allocation of
(38)
Υ𝑞𝑞′ ≥ 0, ∀𝑞 ∈ , ∀𝑞′ ∈  resources and retrofitting strategies, derived by solving
𝑝𝑞 ≥ 0, ∀𝑞 ∈  the transformed single-objective problem with objective
𝛼𝑡𝑜𝑡𝑎𝑙 ≥ 0 function in Equation (26) and subject to the constraints
𝛼𝑞 , 𝛼𝑞 , 𝛼𝑧𝑞 ≥ 0, ∀𝑞 ∈ , ∀ 𝑧 ∈ 𝑞 outlined in (28−38), including budget limits, resource allo-
cation feasibility, equity measures, and retrofit balance.
This structured approach allows for the generation of
2.8.3 System-level optimization solving actionable solutions that balance hazard resilience and
approach equitable resource distribution. The resulting Pareto front
provides a visual representation of the trade-offs between
The optimization problem addresses two conflicting objec- objectives, offering decision-makers flexibility to prioritize
tives: minimizing expected economic loss and minimizing strategies that align with their goals for equity or economic
inequity in resource allocation through the Gini index. To outcomes.
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2085

F I G U R E 2 Socioeconomic spatial distribution on Galveston Island: (a) racial/ethnic composition, (b) income levels, and (c) property
ownership patterns.

3 ILLUSTRATIVE EXAMPLE: T A B L E 1 Moran’s I test for spatial autocorrelation of

GALVESTON, TX socioeconomic demographics on Galveston Island.
Moran’s I p-value
3.1 Galveston testbed and hurricane Race/ethnicity 0.3487 <.001
models Income level 0.5413 <.001
Property ownership 0.3185 <.001
The proposed probabilistic loss assessment and optimiza-
tion methodology is applied to a case study utilizing the
Galveston testbed (IN-CORE, 2024b), which has been emphasize disparities that could lead to varying levels of
employed in numerous prior research studies (e.g., Amini vulnerability and resilience during natural hazards.
& Padgett, 2023; Braik & Koliou, 2023, 2024a; Darestani Additionally, hurricane spatial models for RPs of 50,
et al., 2021; Nofal et al., 2023). Galveston was chosen due 100, and 500 years are obtained from IN-CORE (2024b),
to its well-documented history of hurricane impacts. Its providing the maximum wind speed, flood depth, and sig-
combination of coastal exposure, flood-prone geography, nificant wave height for every location on Galveston Island
and diverse socioeconomic conditions makes it a rep- to facilitate the probabilistic hazard analysis (Darestani
resentative case for testing the proposed method. Many et al., 2021). Figure 3 depicts the maximum wind speed
other coastal communities, particularly in the Gulf Coast and flood depth for each building location under the dif-
and southeastern United States, share these characteris- ferent RP storms. The spatial distribution of wind speed
tics, making the insights gained from this testbed broadly is relatively uniform across the island, whereas flood
applicable. depth demonstrates significant non-uniformity, with vari-
The building inventory dataset includes 26,535 residen- ations in spatial distribution depending on storm intensity.
tial structures, providing a detailed spatial representation This disparity underscores the differing levels of flood
of Galveston’s built environment. Figure 2 shows the vulnerability across the island.
spatial distribution of key socioeconomic factors, includ-
ing race/ethnicity, income levels, and ownership status
(owner-occupied vs. rented). Figure 2a shows a fifth cat- 3.2 Damage and cost analysis
egory for 7017 vacant buildings used for vocational or
secondary homes. These vacant properties, which are pri- Figure 4 shows the distribution of damage and flooding
marily used as vocational properties or secondary homes, limit states for buildings under (a) a 50-year and (b) a 500-
will be included in building-level optimization, as own- year RP storm. The damage limit state includes combined
ers may choose to retrofit them, but they are excluded wind and surge-wave impacts, while the flooding limit
from system-level optimization since they do not qual- state represents flood effects. For the 50-year storm, most
ify for assistance programs. Table 1 highlights the results buildings experience no or moderate damage, indicating
of Moran’s I test (Moran, 1950), indicating strong spa- higher resilience for lower-intensity events. In contrast,
tial correlations among socioeconomic indicators, which the 500-year storm shows a greater number of buildings
14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al.

Hurricane maximum demand loads for 50, 100, and 500 years return period (RP) storms for (a) wind speed and (b) flood

Damage and flooding limit state distribution for: (a) 50 years RP storm and (b) 500 years RP storm.
FIGURE 4
FIGURE 3
depth.
2086
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2087

FIGURE 5 Cost results considering the 500-year RP storm for (a) total flooding cost, (b) total damage cost, and (c) total hurricane cost.

gesting that indirect costs are significant and should not

be overlooked. In the context of building-level optimiza-
tion (Equation 22), including indirect costs is crucial for
accurately capturing the total losses. Excluding them could
result in suboptimal decisions, as the full economic impact
would not be accounted for. Furthermore, the significant
variance in the indirect-to-direct cost ratios across different
buildings underscores the need for careful consideration in
system-level optimization. When prioritizing buildings for
retrofitting based on expected losses, neglecting indirect
costs could lead to inaccurate comparisons between build-
ings, ultimately affecting the allocation of retrofit resources
within the community.

F I G U R E 6 Ratio of indirect to direct damage cost for the 3.3 Optimization of retrofit strategies
500-year RP storm.
In the optimization process, the cost of roof retrofit strate-
with major damage and destruction, highlighting severe gies is considered based on Rsmeans (2018) and adjusted
consequences despite its lower probability. This empha- for inflation. The cost of replacing the roof sheathing and
sizes the need for probabilistic risk assessment to balance adopting a closer-spaced nailing pattern is estimated at
high consequences against the low likelihood of extreme $2.21/ft2 while replacing the roof cover with a new clay tile
events. cover is assumed to be $10.9/ft2 . The minimum total cost of
Figure 5 shows the cost distribution for the 500-year the above is $288 and $272, respectively. For building ele-
RP storm, with (a) total flooding cost, (b) total damage vation, the cost for elevating single-story residential wood
cost, and (c) total hurricane cost, revealing that most build- buildings is $15/ft2 , and for multi-story residential wood
ings experience significant costs. The total hurricane cost buildings, it is $22.5/ft2 . These values correspond to a 1.5 m
is influenced by wind damage and flooding costs, high- (5 ft) elevation height. For a 3 m (10 ft) elevation, a 30%
lighting the importance of a comprehensive multi-hazard additional cost is applied (RenoTag, 2024).
model to capture the full impact of complex interactions The distributions in Figure 7 illustrate the impact of
between different hazards. study periods on the optimized retrofitting strategies at the
Figure 6 shows the ratio of indirect to direct damage building level, with subfigures (a) representing 5 years, (b)
costs for buildings that experienced some level of damage 10 years, (c) 20 years, (d) 50 years, and (e) 100 years. The
during a 500-year RP storm. The prevalence of build- count axis represents the number of buildings for which
ings with a ratio of zero is primarily attributed to vacant the combination of wind and elevation retrofit strategies
properties, which are assumed to incur only direct costs, is found to be optimal. The analysis, utilizing MCS with
reflecting their status as secondary homes. The analysis 5000 samples, indicates that longer study periods increase
reveals a mean ratio of 16% and a 90% quantile of 35%, sug- the likelihood of encountering stronger RP storms,
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2088 BRAIK et al.

F I G U R E 7 Evaluation of building-level retrofit optimization strategies across varying study periods of (a) 5 years, (b) 10 years, (c) 20
years, (d) 50 years, and (e) 100 years.

leading to a higher number of buildings for which ing multiple strategies to identify the most effective and
retrofitting is cost-efficient. Notably, roof retrofitting cost-efficient solution.
through sheathing replacement proves effective even Moreover, Figure 9 illustrates the distribution of the
within shorter study periods, while adding clay tile cover strongest RP storm from the 5000 MCS runs and compares
is rarely cost-efficient. This suggests that roofs are vul- these results against the theoretical distribution obtained
nerable to damage even in low RP events, but adequate using Equation (5). The analysis reveals an almost perfect
sheathing replacement can significantly reduce losses. match between the empirical and theoretical distribu-
In contrast, building elevation strategies are heavily tions, demonstrating the reliability of the probabilistic
influenced by the length of the study period. Higher RP model. Additionally, it indicates that for longer study
storms capable of causing severe flooding have a pro- periods, the likelihood of encountering strong storms
nounced effect when assessed over extended durations, increases, which explains the necessity for more build-
underscoring the need for robust elevation retrofitting in ings to undergo retrofitting for effective mitigation in these
areas prone to long-term flood risks. scenarios.
Figure 8 evaluates the cost-efficiency of different combi- The next step in the optimization process is to incorpo-
nations of elevation and roof retrofit strategies over study rate equity considerations. To do this, 16 distinct socioeco-
periods of 5, 10, 20, 50, and 100 years, comparing each to the nomic groups are defined by combining four race/ethnicity
“do-nothing” approach. Unlike Figure 7, which identifies categories with four economic income levels as outlined
the most optimal strategy, Figure 8 highlights the number in Figure 2. The income levels are classified as follows:
of buildings where a given retrofit strategy is cost-efficient very low income (less than $25K), low income ($25K–
relative to doing nothing. For example, while adding a clay $45K), medium income ($45K–$100K), and high income
cover may be cost-efficient compared to the “do-nothing” (more than $100K). While ownership/rental status was not
approach, it is rarely the optimal strategy. Replacing roof directly included in the socioeconomic categorization, it
sheathing alone often provides sufficient reliability at a played a role in estimating disruption cost parameters as
lower cost. This underscores the importance of evaluat- outlined by FEMA (2013). The distribution of buildings
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2089

F I G U R E 8 Evaluation of building-level retrofit strategies cost-efficiency across varying study periods of (a) 5 years, (b) 10 years, (c) 20
years, (d) 50 years, and (e) 100 years.

across these socioeconomic groups is depicted in Figure 10. nificantly reduce expected losses across all study periods.
Importantly, the utilization of the mean allocation for each This emphasizes the critical need to balance short-term
group, as defined in Equation (25), normalizes the allo- affordability with long-term resilience planning. It fur-
cated resources based on the number of buildings within ther suggests that higher investments in mitigation yield
each socioeconomic group. This ensures that the method- long-term cost savings. Such visualizations provide valu-
ology remains practical and generalizable, even in realistic able tools for decision-makers, enabling them to effectively
scenarios where the distribution of buildings is highly assess the long-term implications of various mitigation
uneven across socioeconomic groups. strategies.
Python was utilized as the programming language Figure 12 presents Pareto fronts that illustrate the trade-
to model the optimization framework when performing offs between reducing the total losses and promoting
system-level optimization, while Gurobi 12.0 served as the equity, represented by the Gini index, across various study
solver to efficiently handle complex mathematical formu- periods and budget levels. In a multi-objective optimiza-
lations. Figure 11 presents a spider chart illustrating the tion problem such as this, trade-offs between competing
relationship between study periods (10, 20, 50, and 100 objectives are expected: improving one objective often
years) and allocated budgets ($20 million to $100 million) requires sacrificing the other. This is evident in all plots,
for retrofitting buildings to mitigate hazard risks when where moving from the left (perfect equity) to the right
the Gini index is fixed at 0. The radial axis indicates (greater inequality) results in decreased economic losses.
the allocated budget, while the circular axis indicates the The extent of these trade-offs is particularly sensitive to
expected total losses. Longer study periods correspond to budget availability. Under limited budgets, the reduction
higher expected losses due to the increased likelihood of in economic losses achieved by sacrificing equity is rela-
hurricane exposure as highlighted in Figure 9. Moreover, tively small. Conversely, with larger budgets, the trade-offs
the chart highlights that higher budget allocations sig- become more pronounced, and prioritizing equity entails
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2090 BRAIK et al.

F I G U R E 9 Comparing empirical to theoretical distributions of the strongest RP for a study period of (a) 5 years, (b) 10 years, (c) 20
years, (d) 50 years, and (e) 100 years.

FIGURE 10 Distribution of buildings across socioeconomic

groups.

F I G U R E 1 1 Risk exposure (economic loss) across different

a greater sacrifice in loss reduction potential. This high- budget allocations ($20, $40, $60, $80, and $100 million) and study
lights a key challenge for decision-makers: larger budgets periods for building retrofitting.
enable retrofitting a broader set of buildings, which can
better address the needs of socioeconomically disadvan-
taged groups. However, spreading resources to improve
equity may reduce the overall effectiveness of risk miti- trends between equity and loss remain consistent across
gation, increasing costs and lowering economic efficiency. all study periods. Notably, all plots exhibit a flattening of
The study period also affects expected losses, with longer the curve at a certain point when moving from left to
periods leading to higher losses. Nonetheless, the trade-off right. This indicates that beyond a certain level of equity,
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2091

F I G U R E 1 2 Risk exposure (economic loss) versus Gini index, for different budget allocations and study periods of (a) 100 years, (b) 50
years, (c) 20 years, and (d) 10 years.

further improvements can be achieved without additional fer substantially in their equity levels, showcasing that
economic loss, effectively eliminating the trade-off. strategies incorporating fairness considerations have the
Figure 13 provides an alternative perspective to the potential of having significant positive effects on society,
trade-off analysis in Figure 12 by focusing on the total num- with relatively small added costs. Ignoring equity in these
ber of buildings included in retrofit assistance. Specifically, cases may lead to selecting highly inequitable solutions,
it illustrates the distribution of hazard mitigation strategies even when more equitable alternatives are available at
for a 100-year study period with a $40 million budget. The comparable costs. Second, in some cases—particularly for
results reveal that promoting equity in resource allocation limited budgets, which are common scenarios for socially
can lead to a greater number of buildings receiving retrofit vulnerable communities—the trade-off can achieve higher
assistance. This finding complemented the insights from levels of equity without significantly sacrificing loss reduc-
Figure 12 by demonstrating that, while prioritizing equity tion. For example, Figure 12 demonstrates that at a budget
may result in higher expected economic losses, it could level of $20 million for all study periods, more equi-
result in a broader distribution of resources. Consequently, table solutions can be achieved with minimal impact
more households benefit from the assistance, emphasiz- on loss reduction. Third, equity-focused strategies ensure
ing the value of equity in achieving widespread community that socioeconomically vulnerable communities, which
resilience. are disproportionately impacted by hazards, receive the
While these results highlight the challenges of balanc- necessary support to enhance their resilience. Although
ing equity and loss reduction, it is crucial to emphasize prioritizing equity may reduce overall cost efficiency, it
the importance of pursuing equity. First, each Pareto front delivers long-term societal benefits by reducing disparities,
in Figure 12 reveals a balance point where solutions with improving recovery outcomes for marginalized groups,
similar loss reductions can exhibit varying equity levels. and fostering a more inclusive approach to disaster mit-
For instance, at a budget level of $20 million across dif- igation. This is further exemplified in Figure 13, which
ferent study periods, the Pareto front flattens beyond a highlights that prioritizing equity results in more build-
Gini index of approximately 0.2 to 0.25. This indicates that ings being included in retrofit assistance, even if the
several solutions with comparable loss reductions may dif- total expected losses increase. Understanding how budget
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2092 BRAIK et al.

F I G U R E 1 3 Distribution of mitigation strategies across buildings: trade-offs between loss reduction and equity promotion for a 100-year
study period and a $40 million budget.

levels and study periods influence these trade-offs is essen- in the cost evaluation over the entire lifespan of the struc-
tial for informed decision-making that promotes both ture. Additionally, for long study periods, the effect of
fairness and resilience. climate change and sea-level rise can be integrated into
the proposed framework, enabling a more robust anal-
ysis of future risk scenarios. Moreover, a study focused
4 LIMITATIONS AND FUTURE on sensitivity analysis of various hazards, fragility, cost
WORKS function, socioeconomic vulnerability, optimization, and
equity parameters would provide valuable insights into
Despite its contributions, the current study has some the robustness and reliability of multi-objective optimiza-
limitations. Long-term predictions, such as over 50 or tion for disaster resilience. Future research could also
100 years, may require incorporating climate change and explore the cumulative impact of multiple, closely spaced
sea-level rise to better capture evolving risks. Moreover, hurricane events, including partially damaged buildings’
while the study provides insights into cost-efficiency and fragility and damage analysis. Furthermore, the proposed
equity trade-offs, actionable recommendations would ben- framework can be implemented using a participatory
efit from a more comprehensive social science perspective. research approach, where communities are engaged as
Additionally, although Galveston shares characteristics partners to emphasize direct collaboration in addressing
with many coastal locations in the United States, relying local populations’ specific vulnerabilities and priorities.
on a single testbed may limit the framework’s validation Integrating a social science perspective is also essential to
across diverse regions and hazards. The assumption of guide the selection of reasonable fairness levels and to sup-
independence between wind and flood impacts is also port decision-making in balancing the trade-off between
acknowledged as a simplification. Furthermore, the cur- economic efficiency and equity.
rent study’s fragility analysis is unsuitable for modeling
scenarios involving multiple, closely spaced hurricanes.
The study also focuses on only two retrofit strategies, leav- 5 CONCLUSION
ing unexplored other potential measures at the building
and community levels. This paper presents a probabilistic loss assessment
The proposed framework can be expanded to address methodology to evaluate multi-hazard risks and optimize
other hazards, such as earthquakes, tornadoes, and wild- retrofit strategies for residential buildings in hurricane-
fires, or adapted to a multi-hazard context, such as prone communities. The optimization process facilitates
combined earthquake and tsunami scenarios. Moreover, decision-making at both building and system levels,
the cost analysis can be extended to a life cycle cost ensuring retrofit strategies are effective, equitable, and
framework, incorporating the aging of structures into the aligned with long-term resilience goals. The methodology
fragility assessment and including periodic maintenance incorporates the multi-hazard effects of wind, surge wave,
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2093

and flooding, alongside both direct and indirect losses, to of Standards and Technology (NIST) under the Finan-
accurately capture the full economic impact of disasters. cial Assistance Award Number (FAIN) #70NANB20H008
The generalizability of the proposed methodology stems for the second, third, and fourth authors. These financial
from its modular and adaptable design, which allows it to supports are gratefully acknowledged. Any opinions, find-
be applied to diverse regions and hazard types. The frame- ings, conclusions, and recommendations presented in this
work can address varying community characteristics and paper are those of the authors and do not necessarily reflect
resilience goals by updating probabilistic hazard models, the views of NSF or NIST.
fragility curves, and socioeconomic data. The epsilon-
constraint optimization approach and resource allocation
REFERENCES
strategies are scalable to communities of different sizes
Abdelhady, A. U., Spence, S. M., & McCormick, J. (2022). Risk and
and complexities.
fragility assessment of residential wooden buildings subject to
Key findings highlight that the effectiveness of mitiga- hurricane winds. Structural Safety, 94, 102137.
tion efforts depends on the length of the study period, with Abushaega, M. M., González, A. D., & Moshebah, O. Y. (2024). A
longer durations increasing the likelihood of severe storms fairness-based multi-objective distribution and restoration model
and enhancing the benefits of cost-effective retrofitting. for enhanced resilience of supply chain transportation networks.
Roof retrofitting, such as sheathing replacement with Reliability Engineering & System Safety, 251, 110314.
closer nailing spacing, is effective for shorter study periods, Aghababaei, M., & Koliou, M. (2023). Community resilience assess-
ment via agent-based modeling approach. Computer-Aided Civil
while elevation strategies become critical for mitigating
and Infrastructure Engineering, 38(7), 920–939.
flood risks over longer durations. For instance, elevat- Ahmadkhanlou, F., & Adeli, H. (2005). Optimum cost design of rein-
ing buildings was found to be cost-efficient for less than forced concrete slabs using neural dynamics model. Engineering
2% of the buildings over a 5-year study period, com- Applications of Artificial Intelligence, 18(1), 65–72.
pared to 33% over a 100-year period. In contrast, roof Akhand, M., Ayon, S. I., Shahriyar, S., Siddique, N., & Adeli,
retrofitting was cost-efficient for 25% of buildings over H. (2020). Discrete spider monkey optimization for travelling
5 years, which increased to 94% over 100 years. Balanc- salesman problem. Applied Soft Computing, 86, 105887.
ing short-term affordability with long-term resilience at Aldwaik, M., & Adeli, H. (2014). Advances in optimization of highrise
building structures. Structural and Multidisciplinary Optimiza-
the system level demonstrates that higher budgets lead
tion, 50, 899–919.
to significant savings and improved outcomes over time. Almufti, I., & Willford, M. (2013). REDi™ rating system: Resilience-
For instance, increasing the budget from $20 million to based earthquake design initiative for the next generation of
$40 million reduces expected community losses by $30, buildings. Arup Co. https://doi.org/10.13140/RG.2.2.20267.75043
$60, $115, and $165 million, for study periods 10, 20, 50, Amini, K., & Padgett, J. E. (2023). Probabilistic risk assessment
and 100 years, respectively. The study also emphasizes of hurricane-induced debris impacts on coastal transporta-
the importance of equitable resource allocation, address- tion infrastructure. Reliability Engineering & System Safety, 240,
109579.
ing the varied impacts of retrofitting strategies on different
Bai, Q., Miralinaghi, M., Labi, S., & Sinha, K. C. (2021). Methodol-
socioeconomic groups. Trade-offs between loss reduction
ogy for analyzing the trade-offs associated with multi-objective
and equity highlight the complexity of developing optimal optimization in transportation asset management under uncer-
mitigation strategies that minimize economic losses while tainty. Computer-Aided Civil and Infrastructure Engineering, 36(4),
ensuring fairness, particularly for vulnerable communi- 381–401.
ties. However, this study reveals an important balance Borghei, M., & Ghassemi, M. (2020). A multi-objective optimization
point beyond which equity can be achieved at no cost. scheme for resilient, cost-effective planning of microgrids. IEEE
By integrating probabilistic risk assessment, optimiza- Access, 8, 206325–206341.
Braik, A. M., & Koliou, M. (2023). A novel digital twin framework
tion techniques, and equity considerations, this research
of electric power infrastructure systems subjected to hurricanes.
offers decision-makers a comprehensive tool to evalu- International Journal of Disaster Risk Reduction, 97, 104020.
ate the consequences of various mitigation strategies Braik, A. M., & Koliou, M. (2024a). A digital twin framework
and effectively balance competing objectives. This holis- for efficient electric power restoration and resilient recovery in
tic approach fosters sustainable, resilient communities the aftermath of hurricanes considering the interdependencies
by protecting vulnerable populations, promoting fairness, with road network and essential facilities. Resilient Cities and
and enhancing long-term disaster recovery and risk man- Structures, 3(3), 79–91.
agement. Braik, A. M., & Koliou, M. (2024b). Automated building dam-
age assessment and large-scale mapping by integrating satellite
imagery, GIS, and deep learning. Computer-Aided Civil and Infras-
AC K N OW L E D G M E N T S tructure Engineering, 39, 2389–2404.
Financial support for this work was provided by the US Chen, V. X., & Hooker, J. (2021). Fairness through social welfare
National Science Foundation (NSF) under Award Num- optimization. arXiv preprint arXiv:2102.00311. https://doi.org/10.
ber 2052930 for all authors and by the National Institute 48550/arXiv.2102.00311
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2094 BRAIK et al.

Cheng, C. S., Behzadan, A. H., & Noshadravan, A. (2021). Deep González, A. D., Chapman, A., Dueñas-Osorio, L., Mesbahi, M.,
learning for post-hurricane aerial damage assessment of build- & D’Souza, R. M. (2017). Efficient infrastructure restoration
ings. Computer-Aided Civil and Infrastructure Engineering, 36(6), strategies using the recovery operator. Computer-Aided Civil and
695–710. Infrastructure Engineering, 32(12), 991–1006.
Coleman, N., Esmalian, A., Lee, C.-C., Gonzales, E., Koirala, P., & González, A. D., Dueñas-Osorio, L., Sánchez-Silva, M., & Medaglia,
Mostafavi, A. (2023). Energy inequality in climate hazards: Empir- A. L. (2016). The interdependent network design problem
ical evidence of social and spatial disparities in managed and for optimal infrastructure system restoration. Computer-
hazard-induced power outages. Sustainable Cities and Society, 92, Aided Civil and Infrastructure Engineering, 31(5), 334–
104491. 350.
Coleman, N., Esmalian, A., & Mostafavi, A. (2020). Equitable Gupta, H. S., Adluri, T., Sanderson, D., González, A. D., Nicholson,
resilience in infrastructure systems: Empirical assessment of dis- C. D., & Cox, D. (2024). Multi-objective optimization of mitigation
parities in hardship experiences of vulnerable populations during strategies for buildings subject to multiple hazards. International
service disruptions. Natural Hazards Review, 21(4), 04020034. Journal of Disaster Risk Reduction, 100, 104125.
Darestani, Y. M., Webb, B., Padgett, J. E., Pennison, G., & Gupta, H. S., Gonzalez, A. D., Jnad, R., & Kameshwar, S. (2024).
Fereshtehnejad, E. (2021). Fragility analysis of coastal roadways Fairness-driven multi-objective optimization for evacuation plan-
and performance assessment of coastal transportation systems ning in natural disasters. International Conference on Transporta-
subjected to storm hazards. Journal of Performance of Constructed tion and Development 2024, Atlanta, GA (pp. 144–152).
Facilities, 35(6), 04021088. Gupta, H. S., Nofal, O. M., González, A. D., Nicholson, C. D., & van de
DesignSafe. (2024). NSF NHERI DesignSafe. https://www.designsafe- Lindt, J. W. (2022). Optimal selection of short-and long-term miti-
ci.org/ gation strategies for buildings within communities under flooding
Dhakal, S., & Zhang, L. (2023). A social welfare–based infrastruc- hazard. Sustainability, 14(16), 9812.
ture resilience assessment framework: Toward equitable resilience Gutierrez Soto, M., & Adeli, H. (2017). Many-objective control opti-
for infrastructure development. Natural Hazards Review, 24(1), mization of high-rise building structures using replicator dynam-
04022043. ics and neural dynamics model. Structural and Multidisciplinary
Dietrich, J., Zijlema, M., Westerink, J., Holthuijsen, L., Dawson, C., Optimization, 56, 1521–1537.
Luettich, R. Jr., Jensen, R., Smith, J., Stelling, G., & Stone, G. (2011). Highfield, W. E., Peacock, W. G., & Van Zandt, S. (2014). Mitiga-
Modeling hurricane waves and storm surge using integrally- tion planning: Why hazard exposure, structural vulnerability, and
coupled, scalable computations. Coastal Engineering, 58(1), 45–65. social vulnerability matter. Journal of Planning Education and
Do, T. Q., van de Lindt, J. W., & Cox, D. T. (2020). Hurricane Research, 34(3), 287–300.
surge-wave building fragility methodology for use in damage, loss, Hosseini, S., Barker, K., & Ramirez-Marquez, J. E. (2016). A review
and resilience analysis. Journal of Structural Engineering, 146(1), of definitions and measures of system resilience. Reliability Engi-
04019177. neering & System Safety, 145, 47–61.
Dong, Y., & Li, Y. (2016). Risk-based assessment of wood residential Ibrahim, H. A., Elawady, A., & Prevatt, D. O. (2024). Empirical
construction subjected to hurricane events considering indirect hurricane fragility assessment of elevated and slab-on-grade res-
and environmental loss. Sustainable and Resilient Infrastructure, idential houses. International Journal of Disaster Risk Reduction,
1(1-2), 46–62. 110, 104663.
Dong, Y., & Li, Y. (2017). Risk assessment in quantification of hurri- IN-CORE. (2024a). Interdependent networked community resilience
cane resilience of residential communities. ASCE-ASME Journal modeling environment. https://www.ncsa.illinois.edu/research/
of Risk and Uncertainty in Engineering Systems, Part A: Civil project-highlights/in-core/
Engineering, 3(4), 04017027. IN-CORE. (2024b). Galveston testbed. https://incore.ncsa.
Ellingwood, B. R. (1994). Probability-based codified design: Past illinois.edu/doc/incore/notebooks/Galveston_testbed/
accomplishments and future challenges. Structural Safety, 13(3), Galveston_testbed.html
159–176. Khajwal, A. B., Cheng, C. S., & Noshadravan, A. (2023). Post-disaster
Esmalian, A., Wang, W., & Mostafavi, A. (2022). Multi-agent mod- damage classification based on deep multi-view image fusion.
eling of hazard–household–infrastructure nexus for equitable Computer-Aided Civil and Infrastructure Engineering, 38(4), 528–
resilience assessment. Computer-Aided Civil and Infrastructure 544.
Engineering, 37(12), 1491–1520. Kim, H., & Adeli, H. (2001). Discrete cost optimization of compos-
FEMA. (2003). HAZUS-MHMR methodology, multi-HAZARd loss ite floors using a floating-point genetic algorithm. Engineering
estimation. Technical Manual, NIST. Optimization, 33(4), 485–501.
FEMA. (2009). Earthquake model (HAZUS-MH MR5): Technical Koliou, M., & van de Lindt, J. W. (2020). Development of building
Manual. Federal Emergency Management Agency. restoration functions for use in community recovery planning to
FEMA. (2013). Multi-hazard loss estimation methodology, flood model, tornadoes. Natural Hazards Review, 21(2), 04020004.
HAZUS, technical manual. FEMA, Mitig Div Washington. Koliou, M., van de Lindt, J. W., McAllister, T. P., Ellingwood, B. R.,
Fish, B., Kun, J., & Lelkes, Á. D. (2016). A confidence-based approach Dillard, M., & Cutler, H. (2020). State of the research in commu-
for balancing fairness and accuracy. Proceedings of the 2016 SIAM nity resilience: Progress and challenges. Sustainable and Resilient
International Conference on Data Mining, SIAM, Miami, FL (pp. Infrastructure, 5(3), 131–151.
144–152). Li, Y., & Ellingwood, B. R. (2006). Hurricane damage to residen-
Gomez, C., & Baker, J. W. (2019). An optimization-based deci- tial construction in the US: Importance of uncertainty mod-
sion support framework for coupled pre-and post-earthquake eling in risk assessment. Engineering Structures, 28(7), 1009–
infrastructure risk management. Structural Safety, 77, 1–9. 1018.
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2095

Li, Y., van de Lindt, J. W., Dao, T., Bjarnadottir, S., & Ahuja, A. (2012). Nofal, O. M., van de Lindt, J. W., & Do, T. Q. (2020). Multi-variate and
Loss analysis for combined wind and surge in hurricanes. Natural single-variable flood fragility and loss approaches for buildings.
Hazards Review, 13(1), 1–10. Reliability Engineering & System Safety, 202, 106971.
Ling, C., Yazdani, N., Beneberu, E., & Koliou, M. (2024). Experimen- Nofal, O. M., Van De Lindt, J. W., Do, T. Q., Yan, G., Hamideh, S., Cox,
tal evaluation of elevated home slabs for flood mitigation. Journal D. T., & Dietrich, J. C. (2021). Methodology for regional multihaz-
of Coastal Research. https://doi.org/10.2112/JCOASTREWS-D-24- ard hurricane damage and risk assessment. Journal of Structural
00009.1 Engineering, 147(11), 04021185.
Lopez, J. M., Roueche, D. B., & Prevatt, D. O. (2020). Wind resis- Nozhati, S., Ellingwood, B. R., & Chong, E. K. (2020). Stochastic opti-
tance and fragility functions for wood-framed wall sheathing mal control methodologies in risk-informed community resilience
panels in low-rise residential construction. Journal of Structural planning. Structural Safety, 84, 101920.
Engineering, 146(8), 04020139. Oh, W. S., Yu, D. J., & Muneepeerakul, R. (2021). Efficiency-fairness
Masoomi, H., Ameri, M. R., & van de Lindt, J. W. (2018). Wind trade-offs in evacuation management of urban floods: The effects
performance enhancement strategies for residential wood-frame of the shelter capacity and zone prioritization. Plos One, 16(6),
buildings. Journal of Performance of Constructed Facilities, 32(3), e0253395.
04018024. Ostadi, B., Seifi, M. M., & Husseinzadeh Kashan, A. (2021). A multi-
Mavrotas, G. (2009). Effective implementation of the ε-constraint objective model for resource allocation in disaster situations to
method in multi-objective mathematical programming problems. enhance the organizational resilience and maximize the value of
Applied mathematics and computation, 213(2), 455–465. business continuity with considering events interactions. Proceed-
Memari, M., Attary, N., Masoomi, H., Mahmoud, H., van de Lindt, ings of the Institution of Mechanical Engineers, Part O: Journal of
J. W., Pilkington, S. F., & Ameri, M. R. (2018). Minimal build- Risk and Reliability, 235(5), 814–830.
ing fragility portfolio for damage assessment of communities Rahmattalabi, A., Jabbari, S., Lakkaraju, H., Vayanos, P., Izenberg,
subjected to tornadoes. Journal of Structural Engineering, 144(7), M., Brown, R., Rice, E., & Tambe, M. (2021). Fair influence max-
04018072. imization: A welfare optimization approach. Proceedings of the
Moehle, J., & Deierlein, G. G. (2004). A framework methodology AAAI Conference on Artificial Intelligence, Virtual (pp. 144–152).
for performance-based earthquake engineering. 13th World Con- RenoTag. (2024). Cost to raise a house. https://renotag.com/costs/
ference on Earthquake Engineering, WCEE, Vancouver, Canada. cost-to-raise-a-house
Moore, H. E., Bates, F. L., Layman, M. V., & Parenton, V. J. (1963). Rodríguez-González, S. (2024). On the multifaceted complexities of
Before the wind: A study of the response to Hurricane Carla. fair resource allocation problems in network systems and how
National Academy of Sciences—National Research Council. to address them via mathematical programming. (Doctoral dis-
Moran, P. A. (1950). Notes on continuous stochastic phenomena. sertation). University of Oklahoma. https://hdl.handle.net/11244/
Biometrika, 37(1/2), 17–23. 340471
Moshebah, O. Y., Rodríguez-González, S., & González, A. D. (2024). Rosowsky, D. V., & Ellingwood, B. R. (2002). Performance-based engi-
A max–min fairness-inspired approach to enhance the perfor- neering of wood frame housing: Fragility analysis methodology.
mance of multimodal transportation networks. Sustainability, Journal of Structural Engineering, 128(1), 32–38.
16(12), 4914. Rsmeans. (2018). Contractor’s pricing guide 2018: Residential repair
Mostafizi, A., Wang, H., Cox, D., Cramer, L. A., & Dong, S. (2017). & remodeling costs with RSMeans data. Gordon Group Incorpora-
Agent-based tsunami evacuation modeling of unplanned net- tion.
work disruptions for evidence-driven resource allocation and Sampat, A. M., & Zavala, V. M. (2019). Fairness measures for decision-
retrofitting strategies. Natural Hazards, 88, 1347–1372. making and conflict resolution. Optimization and Engineering, 20,
NAHB. (2024). Wood framing continues to dominate single- 1249–1272.
family market. The National Association of Home Builders. Sarma, K. C., & Adeli, H. (1998). Cost optimization of concrete
https://www.nahb.org/blog/2024/08/share-of-wood-framed- structures. Journal of Structural Engineering, 124(5), 570–578.
homes-dips-in-2023#:∼:text=Wood%20framing%20remains% Sarma, K. C., & Adeli, H. (2000a). Cost optimization of steel
20the%20most,a%20percent%20were%20steel%2Dframed structures. Engineering Optimization, 32(6), 777–802.
NHC. (2018). Costliest U.S. tropical cyclones tables updated. National Sarma, K. C., & Adeli, H. (2000b). Fuzzy discrete multicriteria cost
Oceanic and Atmospheric Administration (NOAA). https://www. optimization of steel structures. Journal of Structural Engineering,
nhc.noaa.gov/pdf/nws-nhc-6.pdf 126(11), 1339–1347.
Nofal, O., Rosenheim, N., Kameshwar, S., Patil, J., Zhou, X., van de Sen, A. (1997). On economic inequality. Oxford University Press.
Lindt, J. W., Duenas-Osorio, L., Cha, E. J., Endrami, A., & Sutley, Seyedashraf, O., Bottacin-Busolin, A., & Harou, J. J. (2022). A design
E. (2024). Community-level post-hazard functionality methodol- framework for considering spatial equity in sustainable urban
ogy for buildings exposed to floods. Computer-Aided Civil and drainage infrastructure. Sustainable Cities and Society, 85, 103960.
Infrastructure Engineering, 39(8), 1099–1122. Sirca, G. F. Jr., & Adeli, H. (2005). Cost optimization of pre-
Nofal, O. M., Amini, K., Padgett, J. E., van de Lindt, J. W., Rosenheim, stressed concrete bridges. Journal of Structural Engineering, 131(3),
N., Darestani, Y. M., Enderami, A., Sutley, E. J., Hamideh, S., & 380–388.
Duenas-Osorio, L. (2023). Multi-hazard socio-physical resilience Soghrati Ghasbeh, S., Pourmohammadzia, N., & Rabbani, M.
assessment of hurricane-induced hazards on coastal communities. (2022). Equitable post-disaster relief distribution: A robust multi-
Resilient Cities and Structures, 2(2), 67–81. objective multi-stage optimization approach. Journal of Humani-
Nofal, O. M., & van de Lindt, J. W. (2021). Fragility-based flood risk tarian Logistics and Supply Chain Management, 12(4), 618–651.
modeling to quantify the effect of policy change on losses at the Tiong, A. (2022). Multi-objective resilience optimization of interde-
community level. Civil Engineering Research Journal, 11(5), 555822. pendent critical infrastructure networks (Doctoral dissertation).
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2096 BRAIK et al.

Oregon State University. https://ir.library.oregonstate.edu/ Zhang, J., Li, G., & Zhang, M. (2023). Multi-objective optimiza-
concern/graduate_thesis_or_dissertations/zs25xh48c tion for community building group recovery scheduling and
Tomiczek, T., Kennedy, A., & Rogers, S. (2014). Collapse limit state resilience evaluation under earthquake. Computer-Aided Civil and
fragilities of wood-framed residences from storm surge and waves Infrastructure Engineering, 38(12), 1657–1676.
during Hurricane Ike. Journal of Waterway, Port, Coastal, and Zhang, N., & Alipour, A. (2023). A stochastic programming approach
Ocean Engineering, 140(1), 43–55. to enhance the resilience of infrastructure under weather-related
Valinejad, J., & Mili, L. (2022). Community resilience optimization risk. Computer-Aided Civil and Infrastructure Engineering, 38(4),
subject to power flow constraints in cyber-physical-social systems. 411–432.
IEEE Systems Journal, 17(2), 2904–2915. Zhang, W., & Nicholson, C. (2016). A multi-objective optimization
van de Lindt, J. W., & Taggart, M. (2009). Fragility analysis method- model for retrofit strategies to mitigate direct economic loss and
ology for performance-based analysis of wood-frame buildings for population dislocation. Sustainable and Resilient Infrastructure,
flood. Natural Hazards Review, 10(3), 113–123. 1(3-4), 123–136.
Van Zandt, S. (2019). Impacts on socially vulnerable populations. Zhao, D., Li, X., & Cui, J. (2021). A simulation-based optimization
In M. Lindell (Ed.), The Routledge handbook of urban dis- model for infrastructure planning for electric autonomous vehi-
aster resilience (pp. 33–51). Routledge. https://doi.org/10.4324/ cle sharing. Computer-Aided Civil and Infrastructure Engineering,
9781315714462 36(7), 858–876.
Wang, Y., & Sun, B. (2022). Multiperiod equitable and effi-
cient allocation strategy of emergency resources under uncer-
tainty. International Journal of Disaster Risk Science, 13(5), 778–
792. How to cite this article: Braik, A. M., Gupta, H.
Weiner, J. (1985). Size hierarchies in experimental populations of S., Koliou, M., & González, A. D. (2025).
annual plants. Ecology, 66(3), 743–752. Multi-hazard probabilistic risk assessment and
Xia, Y., Langelaar, M., & Hendriks, M. A. (2021). Optimization-based
equitable multi-objective optimization of building
three-dimensional strut-and-tie model generation for reinforced
retrofit strategies in hurricane-vulnerable
concrete. Computer-Aided Civil and Infrastructure Engineering,
36(5), 526–543. communities. Computer-Aided Civil and
Yoo, Y., Koliou, M., & Yazdani, N. (2024). Assessing elevated home Infrastructure Engineering, 40, 2074–2097.
slab retrofitting method for residential coastal buildings under https://doi.org/10.1111/mice.13445
hurricane loads. Journal of Building Engineering, 86, 108994.
 14678667, 2025, 15, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mice.13445, Wiley Online Library on [02/09/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRAIK et al. 2097

APPENDIX TA B L E A 3 Variables used in the optimzation model.

Binar decision variable used in building-level
optimization, representing the number of
TA B L E A 1 Sets and indices used in the optimization model.
buildings in block 𝒛 ∈  belonging to
Set of all buildings/blocks in the community, socioeconomic group 𝒒 ∈  of archetype 𝒓 ∈ 
 indexed by 𝒛 𝝔𝒛𝒒𝒓𝒌 with mitigation strategy 𝒌 ∈ 
 Set of all building archetypes, indexed by 𝑟 Decision variable used in system-level optimization,
 Set of all possible building mitigation strategies, 𝑥𝑧𝑞𝑟𝑘 representing the number of buildings in block 𝑧 ∈ 
indexed by 𝑘, 𝑘 ′ belonging to socioeconomic group 𝑞 ∈  of
 Set of valid changes in mitigation strategies, (𝑘, 𝑘 ′ ) ∈  archetype 𝑟 ∈  with mitigation strategy 𝑘 ∈ 
 Set of socioeconomic groups within the community, Number of buildings in block 𝑧 ∈  belonging to
indexed by 𝑞, 𝑞′ 𝑦𝑧𝑞𝑟𝑘𝑘′ socioeconomic group 𝑞 ∈  of archetype 𝑟 ∈  that
change from mitigation strategy 𝑘 to 𝑘 ′
𝑞 Set of all buildings in the socioeconomic group 𝑞,
defined as 𝛼𝑞 Total investment for socioeconomic group 𝑞 ∈ 
1 , 𝑧 ∈ 𝑞 𝛼𝑞 Average investment per buildings in socioeconomic
𝑞 =  ∩ {𝑧| 𝜔𝑞 = 1}, where 𝜔𝑞 = {
0 , 𝑧 ∉ 𝑞 group 𝑞 ∈ 
𝛼𝑧𝑞 Investment for buildings in block 𝑧 ∈  belonging to
socioeconomic group 𝑞 ∈ 
TA B L E A 2 Parameters used in the optimization model.
Γ Total amount of investment in the community
Initial number of buildings in block 𝒛 ∈ 
𝐺 Gini index
belonging to socioeconomic group 𝒒 ∈  of
𝒃𝒛𝒒𝒓𝒌 archetype 𝒓 ∈  with mitigation strategy 𝒌 ∈ 𝕂
Expected hurricane and flood loss for building 𝑧 ∈ 
Ψ𝑧𝑞𝑟𝑘 belonging to socioeconomic group 𝑞 ∈  of
archetype 𝑟 ∈  with mitigation strategy 𝑘 ∈ 
Cost of retrofitting building 𝑧 ∈  belonging to
𝑐𝑧𝑞𝑟𝑘 socioeconomic group 𝑞 ∈  of archetype 𝑟 ∈  to
mitigation strategy 𝑘
Cost of retrofitting building 𝑧 ∈  belonging to
𝑐𝑧𝑞𝑟𝑘𝑘′ socioeconomic group 𝑞 ∈  of archetype 𝑟 ∈  from
mitigation strategy 𝑘 to 𝑘 ′
𝐵 Total budget available for building mitigation
𝑐max Maximum allowable resource allocation per building
𝑁 Total number of buildings in the community
(𝑁 = || )
𝑀 Total number of socioeconomic groups (𝑀 = || )
𝑁𝑞 Total number of buildings in socioeconomic group 𝑞
( 𝑁𝑞 = |𝑞 | )
The proportion of retrofit costs supported by the
𝛽𝑧𝑞𝑟𝑘 belonging to socioeconomic group 𝑞 of archetype 𝑟
from strategy 𝑘 to strategy 𝑘 ′ , 𝛽𝑧𝑞𝑟𝑘 ∈ [0, 1]
Binary parameter indicating that the mitigation
𝜙𝑧𝑞𝑟𝑘 strategy 𝑘 ∈  is feasible for building block 𝑧 ∈  in
group 𝑞 ∈ 