Ghahremani‑Nahr et al.

Journal of International
Journal of International Humanitarian Action (2024) 9:19
https://doi.org/10.1186/s41018-024-00163-8 Humanitarian Action

RESEARCH ARTICLE Open Access

Designing a humanitarian relief logistics

network considering the cost of deprivation
using a robust‑fuzzy‑probabilistic planning
method
Javid Ghahremani‑Nahr1 , Hamed Nozari2   and Agnieszka Szmelter‑Jarosz3*   

Abstract
This study addresses deprivation costs in humanitarian logistics by incorporating direct and indirect impacts
into a robust-fuzzy-probabilistic framework to optimize resource allocation and minimize delays in disaster response.
This model has concepts such as facility location and optimal flow allocation. The results show that with the
increase in the uncertainty rate and as a result of the increase in the number of people affected by the accident
and the demand for relief items at the scene of the accident, the costs related to the supply and transporta‑
tion of the items and the injured have increased. Also, by examining the effect of deprivation cost on the results
of the model, it was observed that with the increase of deprivation cost, the model aims to increase the goods
transferred to satisfy the demand fully. Therefore, the expenses related to deprivation have decreased, but on the
other hand, the costs associated with the transfer of products have increased. As a result, the cost of the entire relief
logistics network has increased. Finally, by examining the effects of the weighting coefficient of the objective function
on the values of the objective function of the problem, it can be seen that with the increase of this weighting coef‑
ficient, the value of the objective function has increased due to the increase in the volume of transfers of relief items.
Keywords Logistics of humanitarian aid, Uncertainty, Cost of deprivation, Logistics network

Introduction according to their intensity and where they occur. For

Natural disasters occur in different parts of the world in this reason, preparation for them, planning, forecast-
other seasons, leaving many victims and causing much ing, and necessary preventive measures are vital in these
financial damage to governments (Cao et al. 2017). Nat- cases (Tatham Christopher 2018). It was said that disas-
ural disasters are floods, earthquakes, tsunamis, intense ters cause many human and financial losses to societies
storms, tornadoes, meteorite collisions, etc. Each of these and nations. Therefore, appropriate emergency logis-
natural events will have numerous harmful consequences tics planning for natural threats and crisis management
is essential. Furthermore, the nature of such incidents is
their unexpectedness and unpredictability, despite the
*Correspondence:
Agnieszka Szmelter‑Jarosz
progress of science and technology, which makes the
[email protected] availability of preventive plans and emergency response
1
Academic Centre for Education, Culture and Research (ACECR), after the occurrence of the incident critical, so the lim-
Tabriz 5154837693, Iran
2
Department of Management, Azad University of UAE, PO Box: 502321,
ited resources available to we should allocate the optimal
Dubai Knowledge ParkDubai, United Arab Emirates form to the affected people so that their needs are met as
3
Department of Logistics, Faculty of Economics, University of Gdańsk, Ul. much as possible (Salehi et al. 2019).
Armii Krajowej 119/121, 81‑824 Sopot, Poland

© The Author(s) 2024. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://creativecommons.org/licenses/by/4.0/.
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 2 of 16

Deprivation costs, a critical aspect of humanitarian and casualties caused by natural and human disasters
logistics, encompass direct impacts, such as health dete- have increased dramatically, so every year around the
rioration and loss of life, and indirect effects, including world, around 70 thousand people are killed, and nearly
long-term societal and economic consequences. These 200 million people are affected by these disasters.
costs are influenced by factors such as response delays, Hence, crisis management, especially effective man-
inequitable resource allocation, and the vulnerability of agement of relief logistics activities and evacuating the
affected populations. By integrating concepts like depri- injured from the affected areas, seems necessary. One of
vation cost indices, marginal costs of delay, and proba- the most important activities during a crisis is the logis-
bilistic outcomes under uncertainty, this study aims to tics of relief items, delivering these items to the injured
capture the nuanced trade-offs in resource distribution and transporting the wounded to temporary and perma-
and develop a robust, equitable framework for minimiz- nent medical centers. For this purpose, in this study, a
ing deprivation in disaster scenarios. Our country is one multi-objective logistic model is used for the delivery of
of the countries where natural disasters such as floods priority goods in crisis relief operations so that appropri-
and earthquakes have occurred with various severities ate decisions can be made for the delivery of relief goods
and have caused many financial and human losses. Plan- in such a way that the total unfulfilled needs of the goods
ning for an emergency response to the postdisaster cri- are minimized. Therefore, the development of last-mile
sis is very important and makes relief operations more distribution and delivery and evacuation operations are
efficient and effective. The existence of plans and plans essential strategic issues in relief logistics. Furthermore,
to deal with the consequences of the disaster, along with it is necessary to provide disaster relief, locate emer-
raising public awareness of the people, will lead to the gency personnel, transfer aid materials after the start of
reduction of human and financial losses, which will be a natural disasters, and evacuate injured people from the
primary goal of relief responses and reactions. The criti- affected areas to the hospital. All these activities have an
cal point that should be considered in designing these extensive effect on the rescue rate of the survivors after
plans is that the nature of accidents, such as earthquakes, the accident. When the damaged places are far away,
is such that the response should be made at high speed or it is difficult to access them, or when other means of
and in a short time. In other words, in such an emer- transportation cannot, helicopters are the most practical
gency and complex situation, the decision-maker must means to reach the accident victims, especially flood and
quickly and effectively carry out rescue operations and earthquake victims.
take care of the condition of the injured and injured. To This research presents a comprehensive model of the
achieve this important goal and take action at the right humanitarian aid logistics network. In this research, the
time, it seems necessary to have access to an efficient and model presented is to minimize the costs of the entire
systematic pre-defined program in which all the required network, including location, routing, allocation, and
activities, their sequence, and connections are specified maintenance, in which the demand for relief items in the
(Hoseininezhad et al. 2021). Therefore, one of these activ- damaged areas and some other essential parameters are
ities, which is also very important, is the logistics plan- considered non-deterministic. Important decisions such
ning for transporting essential items the affected areas as the location of facilities and optimal allocation of flow
need. To achieve the goals of crisis management in acci- between facilities to transfer relief items and injured peo-
dents, providing aid to the accident areas, the importance ple are made in this model. Considering the existence of
of distributing relief goods and picking up the injured is uncertainty in the network, the robust-fuzzy-probabil-
a sensitive and strategic matter because improving the istic programming method has been used to control the
efficiency in carrying and transporting goods and the non-deterministic parameters of the problem.
injured has a significant impact on the rate of rescu-
ers after the accident. Looking at the natural disasters Literature review
of the last few decades, we clearly understand that the Every year, natural disasters and unforeseen events such
occurrence of natural disasters causes high financial and as earthquakes, floods, storms, droughts, and fires plague
human costs for governments and societies, and day by different parts of the world, and the lack of preparation
day, the concerns in this field are increasing (Jamali et al. and proper coping causes heavy losses and damages to
2021). Making the right decisions and taking appropriate nations and their assets, which are sometimes irrepara-
and timely actions in each phase of the crisis manage- ble. Due to the increased number of unexpected events
ment cycle leads to the reduction of possible damages at and land, ocean, and atmospheric disasters in recent
the time of the accident and the reduction of the vulner- years, planning to better respond to these events seems
ability of society. Decision-making tools are needed to necessary. Therefore, one of the most critical actions
make appropriate decisions. In recent years, the losses in responding to a crisis is choosing the right place for
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 3 of 16

supplies and support from the affected areas. One of a forbidden search algorithm to solve larger-size prob-
the essential measures in the event of accidents is logis- lems. The results obtained from their model showed
tical measures and quick relief to crisis areas. Review- the high efficiency of the forbidden search algorithm in
ing and evaluating experimental and scientific studies solving problems of different sizes. Yahya and Bozorgi-
of researchers can open the way for a new method for Amiri (Yahyaei and Bozorgi-Amiri 2018) designed a
assessing the concept of relief logistics, which will be dis- supply chain network under uncertainty. The main goal
cussed in this section. of this article was to control the failure of distribution
Mohamadi and Yaghoubi (Mohamadi and Yaghoubi centers to meet the needs of demand centers by reduc-
2017) developed a two-level model for locating transfer ing investment costs. Also, they used the robust optimi-
points and relief goods distribution centers in earthquake zation method to control the uncertain parameters and
conditions. In the first level, relief facilities are located, showed that the total investment costs increase with the
and the location of transfer points is determined. In increase of the uncertainty level. Elçi and Noyan (Elçi
the second level, routing transfers the injured and dead and Noyan 2018) modeled a scenario-based humanitar-
bodies to predetermined points. Also, three scenarios ian supply chain network. Their model investigated the
have been considered according to the conditions of the optimal number and location of facilities, facility capac-
faults in Tehran (Mesha fault, Rey fault, and North fault) ity, inventory levels, and transportation network condi-
with probabilities of 0.35, 0.30, and 0.35. Ultimately, this tions under demand uncertainty after the accident. They
model was solved using the epsilon constraint method used a branch-and-cut algorithm based on the Benders
and Gems software due to its multiple goals. The research decomposition to solve their probabilistic model. They
results indicate 10 points for establishing transfer points demonstrated the effectiveness of their model by study-
along the highways. Rodríguez-Espíndola et al. (Rod- ing hurricane threats in the southeastern USA. Paul and
ríguez-Espíndola et al. 2018) introduced a disaster pre- Wang (Paul and Wang 2019) addressed robust modeling
paredness system based on multi-objective optimization of the location-assignment problem in conditions. This
and geographic information systems for multi-objective research aims to minimize the total social costs, includ-
decision support. The mapping model is used to prevent ing initial setup and procurement and deprivation costs
the selection of cascading facilities, and a dual-objective associated with delays in accessing resources. The models
optimization model is used to determine the location of include various uncertainties in earthquake magnitude,
emergency facilities, stock forecasting, resource alloca- such as facility damage, losses by intensity, and travel
tion, and aid distribution, as well as the number of actors time. Zanganeh et al. (Zanganeh et al. 2019) presented
required to perform these activities. The deprivation a dual-objective model for humanitarian logistics net-
cost is often a key economic indicator of human suffer- works in post-crisis situations. The purpose of this model
ing in emergency logistics (Yu et al. 2018). An improved was to maximize the functional performance of saving
method for adequate and equitable allocation of vital the lives of injured people and to help injured people
resources in emergency logistics is proposed that consid- with a delay. Hong (Hong 2020) addressed humanitar-
ers human suffering using this economic representation. ian logistics network design with facility location deci-
Torabi et al. (Torabi et al. 2018) designed a scenario- sions. He used an ideal planning approach to solve his
based humanitarian supply chain network model under four-objective model. Data coverage research methods
uncertainty. Their model aimed to reduce the costs of the were used to evaluate effective answers. Mansoori et al.
entire supply chain network, including location, trans- (Mansoori et al. 2020) proposed a multi-objective math-
portation, storage of relief items, and the cost shortage ematical model for the humanitarian supply chain net-
penalty. Their model considered parameters such as fixed work design problem that simultaneously minimizes (1)
construction costs, transportation costs, and demand the number of injured not transported to hospitals and
in a non-deterministic manner and used the scenario- the total number of homeless not evacuated from the
based method. They used GAMS software to solve affected area and (2) pays for the total unmet needs of
their model and studied their model in Iran. In a paper, auxiliary goods. This mathematical model used robust
Noham and Tzur (Noham and Tzur 2018) discussed the optimization to control demand parameters and travel
design of the humanitarian supply chain network using time. Ghaffari et al. (Ghaffari et al. 2020) modeled an
real postdisaster decisions. Reviewing previous articles emergency logistics network to optimize the costs of the
showed that no comprehensive model can help with entire network. This problem also considers the schedul-
post-incident decisions. Therefore, they designed a com- ing of multi-stage operations with limited capacity. The
prehensive humanitarian supply chain network and 40 exact method and particle swarm optimization algo-
numerical examples for their model. They used CPLEX rithm have been used to solve the problem. Cheng et al.
software to solve small- and medium-sized problems and (Cheng et al. 2021) examined the issue of humanitarian
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 4 of 16

aid food distribution. In this model, a group of food humanitarian aid logistics model under probabilistic
donors have taken steps to meet the needs of the affected planning optimization. The objectives of this paper are
areas, and the goal is fair distribution among them. They (1) to minimize the expected value of the total costs of
used the robust optimization method to control the non- the relief supply chain, (2) to minimize the maximum
deterministic demand parameter. Hoseininezhad et al. number of unmet demands for relief workers, and (3) to
(Hoseininezhad et al. 2021) presented a multi-objective minimize the total probability of failed evacuations along
mathematical model for pre-positioning a rescue chain the routes. This paper proposes a scenario-based stochas-
for road accidents. They used the NSGA II algorithm to tic multi-objective location-allocation-routing model for
solve their multi-objective model. In this model, using a real humanitarian relief logistics problem, focusing on
the obtained results, they suggested potential locations pre- and postdisaster situations in uncertainty. A simu-
for the relief chain. Akbari et al. (Akbari et al. 2021) mod- lation approach was used to deal with demand uncer-
eled and solved a relief logistics problem by establishing tainty. The presented model has been solved using the
coordination relationships in the supply chain network. epsilon-constraint method for minor and medium-scale
Their main goal in this research was to minimize the total problems and three metaheuristic algorithms for large-
costs of network design. They used a genetic algorithm to scale problems (case study). Experimental results show
solve the model. Boostani et al. (Boostani et al. 2021) pro- that this model can be used to locate shelters and aid
posed a three-level aid supply chain network before and distribution centers, determine appropriate routes, and
after an incident to improve sustainability in humanitar- allocate resources in uncertain and real disaster situ-
ian aid operations. Their goal was to minimize the costs ations. Abazari et al. (Abazari et al. 2022) modeled a
of the entire humanitarian supply chain, maximize social humanitarian supply chain network that includes locat-
welfare by maximizing the minimum level of satisfaction ing relief centers, pre-positioning and distributing relief
in the affected areas, and minimize the amount of green- items, and providing medical treatment in affected areas
house gas emissions. They used CP and LOM methods to to help affected people. The goal of this model is to mini-
solve their three-objective model. mize total costs. The problem is formulated as a complex
Shu et al. (Shu et al. 2022) designed a humanitarian integer linear mathematical model. A two-stage stochas-
relief network with uncertain demand for each affected tic fuzzy hybrid approach is developed to capture uncer-
area. The mathematical model aimed to maximize the tainty. The grasshopper optimization algorithm has been
response to the demand of the damaged points. They used to solve the model on a large scale, and its perfor-
conducted a case study on Typhoon Rammasun in 2014. mance has been investigated through computational
Computational results demonstrate the effectiveness experiments.
of the solution approaches and show that the stochas- Deprivation costs play a critical role in humanitar-
tic programming models with chance constraints are ian logistics, measuring the human suffering caused by
superior to the deterministic model for humanitarian unmet needs in disaster scenarios. These costs, encom-
relief network design. Modiri et al. (Modiri et al. 2022) passing direct impacts like mortality and indirect effects
presented a multi-objective mixed integer mathematical such as long-term economic losses, are key to designing
programming model for the design of relief item distri- effective disaster response frameworks (Holguín-Veras
bution networks in sustainable disaster relief logistics. et al. 2013). Recent studies highlight the importance of
The first objective function minimizes the total network integrating deprivation costs into humanitarian logistics
costs. These are divided into two parts: (1) relief costs, models. For instance, Silva and Leiras (Silva and Leiras
including transportation costs, inventory costs, and fixed 2021) conducted a comprehensive review underscoring
facility costs, and (2) social costs (deprivation cost). The their significance in optimizing resource allocation and
second objective function minimizes the amount of pol- minimizing suffering in disaster zones. Similarly, van
lution produced by the network. To show the validity of Steenbergen et al. (Steenbergen et al. 2023) proposed
the proposed model, a case study based on real data (the a stochastic dynamic model incorporating deprivation
2019 flood in Sari City, Mazandaran province) was evalu- costs, demonstrating its efficacy in improving disaster
ated. Using the proposed model, decision-makers and response operations. Advanced modeling techniques
managers can make strategic and tactical decisions with have been utilized to address uncertainties inherent in
the least cost and time, and in relief planning, they can disaster environments. Boonmee et al. (Boonmee et al.
improve the structure of distribution and inventory net- 2020) introduced a multi-objective fuzzy mathematical
works and reduce the dissatisfaction of victims. Ghasemi programming model, effectively managing the ambiguity
et al. (Ghasemi et al. 2022) presented a multi-objective in demand and supply data. Further, Sharma and Mehta
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 5 of 16

(Sharma and Mehta 2021) developed a probabilistic fuzzy Due to the uncertainty of the amount of demand at
goal programming model for optimizing postdisaster the place of the accident and the severity of the crisis,
supply chain, accommodating demand and transporta- the parameters of demand and cost of transmission
tion cost uncertainties. Integrating robust, fuzzy, and have been considered uncertain in different scenarios.
probabilistic approaches enhances the resilience of relief Therefore, if any of the distribution centers are out of
operations, ensuring the timely and equitable allocation reach due to the disruption, it is impossible to transfer
of resources and reducing deprivation costs. relief items from that center, and other centers should
Based on the studied literature, research innovation be used to meet the demand of the disaster areas.
can be expressed in the following cases: In the designed model, there are two different types
of decisions. First, strategic decisions are essential,
• Development of a relief logistics network model in such as choosing the optimal number and location of
pre- and post-crisis conditions relief item distribution centers and potential emer-
• Using robust-fuzzy-probabilistic programming gency centers. Also, tactical decisions, such as how to
method to control non-deterministic parameters transfer, buy, and store environmentally compatible and
non-compatible relief items and how to transport the
Problem definition and modeling injured by vehicle, are of great importance.
This section discusses the modeling of the humanitarian Considering the above, the assumptions of the prob-
aid logistics problem. According to Fig. 1, in this prob- lem can be stated as follows:
lem, a set of suppliers have the task of providing envi-
ronmentally friendly and non-environmentally friendly • The severity of the crisis is considered based on dif-
relief items. There is also a set of potential locations for ferent scenarios, each with a different probability of
aid distribution centers, which should be located accord- occurrence.
ing to the severity of the crisis and be considered in dif- • The location of suppliers and hospitals is already
ferent scenarios. In this case, the distribution centers, as known.
an intermediary, transfer relief items from the supplier • Demand and transmission costs are considered non-
to the demand points after the crisis. In the meantime, deterministic and trapezoidal fuzzy numbers.
due to the size of the crisis, there is a possibility of injury • Aid items are considered to be compatible with the
to some people, who should be transferred to temporary environment and non-compatible with it.
emergency centers in case of minor injuries and to hospi- • The capacity of the facility and the vehicles is known
tals in case of serious injuries. There is also the possibility in advance.
of a severe crisis after transferring people to emergency
centers, which will also be addressed in the upcoming According to the above assumptions and stated
modeling. On the other hand, some people need to go to objective functions, the symbols used in the modeling
shelters after an accident. are as follows.

Fig. 1 Proposed humanitarian aid logistics network

Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 6 of 16

Sets Sets

I Set of demand points i ∈ I Tr ljcs The cost of transporting relief

goods c ∈ C from the supplier l ∈ L
J The set of potential distribution to the relief distribution center j ∈ J
centers j ∈ J after the disaster in the scenario
K Capacity set of aid distribution cent‑ s∈S
ers k ∈ K Tr lics The cost of transporting relief
L The set of suppliers l ∈ L goods c ∈ C from the supplier
C Relief goods set c ∈ C l ∈ L to the demand point i ∈ I
after the disaster in the scenario
C1 A set of items with environmentally s∈S
friendly packaging c1 ∈ C1
Tr jics The cost of transporting relief goods
C2 A set of items with non-environ‑ c ∈ C from the relief distribution
mentally friendly packaging c2 ∈ C2 center j ∈ J to the demand point
S The set of probability events i ∈ I after the disaster in scenario
of the scenario s ∈ S s∈S
N The set of potential emergency Tr ihs The cost of transporting the injured
centers n ∈ N from the demand point i ∈ I
to the hospital h ∈ H after the disas‑
H Hospital set h ∈ H
ter in the scenario s ∈ S
G The set of shelters g ∈ G
Tr ins The cost of transporting the injured
Parameters from the demand point i ∈ I
ps The probability of scenario s ∈ S to the temporary emergency center
n ∈ N after the disaster in the sce‑
dics Demand amount of relief items
nario s ∈ S
c ∈ C at demand point i ∈ I in sce‑
nario s ∈ S Tr nhs The cost of transporting the injured
from the temporary emergency
wis The number of injured per‑
center n ∈ N to the hospital h ∈ H
sons required to be treated
after the disaster in the scenario
at the demand point i ∈ I in the sce‑
s∈S
nario s ∈ S
Tr igs The cost of transporting the injured
θc Space required to store relief goods
from the demand point i ∈ I
c ∈ C in cubic meters
to the shelter g ∈ G after the disas‑
Fjk The fixed cost of establishing a relief ter in the scenario s ∈ S
distribution center in location j ∈ J
hjc The cost of maintaining a unit
with capacity level k ∈ K
of relief goods c ∈ C in the relief
Fn The fixed cost of establishing a tem‑ distribution center j ∈ J
porary emergency center in place
vlc1 The amount of relief goods c1 ∈ C1
n∈N
that can be provided from supplier
blc1 The cost of buying a unit of relief l∈L
goods with environmentally friendly
vlc2 The amount of relief goods c2 ∈ C2
packaging c1 ∈ C1 from supplier
that can be provided from supplier
l∈L
l∈L
blc2 The cost of buying a unit of relief
vlc1 s The amount of relief goods c1 ∈ C1
goods with non-environmentally
that can be provided from supplier
friendly packaging c2 ∈ C2
l ∈ L in scenario s ∈ S
from supplier l ∈ L
vlc2 s The amount of relief goods c2 ∈ C2
blc1 s The cost of purchasing a unit
that can be provided from supplier
of relief goods with environmentally
l ∈ L in scenario s ∈ S
friendly packaging c1 ∈ C1 from sup‑
plier l ∈ L in scenario s ∈ S Vk The amount of capacity for each aid
distribution center k ∈ K
blc2 s The cost of purchasing a unit
of relief goods with non-environ‑ Cc ic The cost of denying demand point
mentally friendly packaging c2 ∈ C2 i ∈ I from receiving relief goods
from supplier l ∈ L in scenario s ∈ S c∈C
Tr ljc The cost of transporting relief δn The percentage of injured people
goods c ∈ C from the supplier l ∈ L who should be transferred
to the relief distribution center j ∈ J to the hospital from potential emer‑
gency center n ∈ N
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 7 of 16

Sets Sets
ωi Percentage of people in need Qljcs Amount of relief items c ∈ C pur‑
of shelter from demand point i ∈ I chased from supplier l ∈ L by relief
ωP i The percentage of people who distribution center j ∈ J in the post‑
need an emergency center disaster stage in scenario s ∈ S
from the demand point i ∈ I XN ins The number of people
CapH h Hospital capacity h ∈ H with minor injuries transferred
from the demand point i ∈ I
CapGg Shelter capacity g ∈ G to the temporary emergency center
CapN n Potential temporary emergency n ∈ N in the scenario s ∈ S
center capacity n ∈ N XH ihs The number of people with serious
Decision variables injuries transferred from demand
Ijc The amount of relief goods point i ∈ I to hospital h ∈ H in sce‑
c ∈ C stored in the relief distribu‑ nario s ∈ S
tion center j ∈ J in the stage NH nhs The number of seriously injured
before the crisis people transferred from temporary
Ijcs The amount of relief goods c ∈ C emergency center n ∈ N to hospital
stored in the relief distribution h ∈ H in scenario s ∈ S
center j ∈ J in the post-crisis phase XGigs Number of people transferred
in the scenario s ∈ S from demand point i ∈ I to shelter
Xjk It takes the value of 1 if the relief g ∈ G in scenario s ∈ S
distribution center with capacity Un It takes the value of 1 if the tempo‑
k ∈ K is opened at the potential rary emergency center is opened
location j ∈ J , and it takes the value at the location n ∈ N and otherwise
of 0 otherwise it takes the value of 0
Yjics The amount of relief goods c ∈ C
transferred from the aid distribution According to the symbols defined in the previous sec-
center j ∈ J to the demand point
i ∈ I in the scenario s ∈ S tion, the mathematical model of humanitarian aid logis-
Ylics The amount of relief goods c ∈ C tics network under uncertainty conditions will be as
transferred from the supplier l ∈ L follows.
to the demand point i ∈ I in the sce‑
nario s ∈ S J K N J C J L C1
MinZ1 = Fjk .Xjk + Fn .Un + hjc .Ijc + blc1 .Qljc1
Ylic1 s The amount of relief goods c1 ∈ C1 j=1 k=1 n=1 j=1 c=1 j=1 l=1 c1 =1
transferred from supplier l ∈ L J L C2 J L C
to demand point i ∈ I in scenario + blc2 .Qljc2 + Tr ljc .Qljc
s∈S j=1 l=1 c2 =1 j=1 l=1 c=1

Ylic2 s The amount of relief goods c2 ∈ C2 S J C I J L C1

+ Ps hjc Ijcs − Yjics + blc1 s .Qljc1 s
transferred from the supplier l ∈ L s=1 j=1 c=1 i=1 j=1 l=1 c1 =1
to the demand point i ∈ I in the sce‑
J L C2 J L C
nario s ∈ S + blc2 s .Qljc2 s + Tr ljcs .Qljcs
Qljc1 Order amount of a relief item j=1 l=1 c2 =1 j=1 l=1 c=1

with environmentally friendly pack‑ I L C I J C

aging c1 ∈ C1 from supplier l ∈ L + Tr lics .Ylics + Tr jics .Yjics
by relief distribution center j ∈ J i=1 l=1 c=1 i=1 j=1 c=1

I H I N
Qljc2 Order amount of a relief item + Tr ihs .XH ihs + Tr ins .XN ins
with non-environmentally friendly i=1 h=1 i=1 n=1
packaging c2 ∈ C2 from supplier N H I G
l ∈ L by relief distribution center + Tr nhs .NH nhs + Tr igs .XG igs
j∈J n=1 h=1 i=1 g=1

Qljc1 s Order amount of a relief item J I C L I C

Ccic . dics − Yjics + Ccic .(dics − Ylics )
with environmentally friendly pack‑ +
j=1 i=1 c=1 l=1 i=1 c=1
aging c1 ∈ C1 from supplier l ∈ L
by relief distribution center j ∈ J (1)
in scenario s ∈ S s.t. :
Qljc2 s Order amount of a relief item
with non-environmentally friendly
packaging c2 ∈ C2 from supplier Qljc = Qljc1 + Qljc2 , ∀j, l, c, c1 , c2 (2)
l ∈ L by relief distribution center
j ∈ J in scenario s ∈ S
Qljc Amount of relief items c ∈ C pur‑ Qljcs = Qljc1 s + Qljc2 s , ∀j, l, c, c1 , c2 , s (3)
chased from supplier l ∈ L by relief
distribution centerj ∈ J in the pre-
disaster phase
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 8 of 16

L
 N

Ijc = Qljc , ∀j, c (4) XN ins = ωpi .wis , ∀i, s (17)
l=1 n=1

L
 L
 L
 I
 H

Ijcs = Qljcs + Qljc1 + Qljc2 , ∀j, c, c1 , c2 , s δn XN ins = NH nhs , ∀n, s (18)
l=1 l=1 l=1 i=1 h=1
(5)
I
 H

Yjics ≤ Ijcs , ∀j, c, s (6) NH nhs ≤ BM.Un , ∀n, s (19)
i=1 h=1

C
 K
 I

θc .Ijc ≤ Vk .Xjk , ∀j (7) XG igs ≤ CapG g , ∀g, s (20)
c=1 k=1 i=1

C
 K
 N
 I

θc .Ijcs ≤ Vk .Xjk , ∀j, s (8) NH nhs + XH ihs ≤ CapH h , ∀h, s (21)
c=1 k=1 n=1 i=1

K I


Xjk ≤ 1, ∀j (9) XN ins ≤ CapN n .Un , ∀n, s (22)
k=1 i=1

J Ijc , Ijcs , Yjics , Ylics , Ylic1 s , Ylic2 s , Qljc1 , Qljc2 ,


Qljc1 s , Qljc2 s , Qljc , Qljcs , XN ins , XH ihs , NH nhs , XG igs ≥ 0, integer
Qljc1 ≤ vlc1 , ∀l, c1 (10) (23)
Un , Xjk ∈ {0,1}
j=1

Equation (1) shows the costs of the entire humani-

J
 tarian logistics network. These costs include the total
Qljc2 ≤ vlc2 , ∀l, c2 (11) cost of location, purchase, transfer, allocation, and
j=1 exclusion costs. Equations (2) and (3), respectively,
show the total amount of goods purchased from the
J
 I
 supplier by the relief distribution center in the condi-
Qljc1 s + Ylic1 s ≤ vlc1 s , ∀l, c1 , s (12) tions before and after the crisis. Equation (4) shows
j=1 i=1 the inventory level of aid items purchased before the
crisis, and Eq. (5) shows the inventory level that is
J
 I
 compatible/incompatible with the environment after
Qljc2 s + Ylic2 s ≤ vlc2 s , ∀l, cs , s (13) the crisis. Equation (6) shows that the number of
j=1 i=1 relief items transferred from the distribution centers
to the demand points should be less than the amount
Ylics = Ylic1 s + Ylic2 s , ∀i, l, c, c1 , c2 , s (14) of stock in the distribution center’s warehouse. Equa-
tions (7) and (8) show the capacity limitation of relief
items distribution centers in all capacity levels. Equa-
N H G
   tion (9) guarantees that each distribution center can
XN ins + XH ihs + XG igs = wis , ∀i, s (15) use at most one capacity level. Relationships (10) to
n=1 h=1 g=1
(13) show the constraints related to the capacity to
provide environmentally friendly/unfriendly relief
G
 items. Equation (14) shows the total relief items trans-
XG igs = ωi .wis , ∀i, s (16) ferred from the supplier to the demand points. The
g=1
relation (15) guarantees that all injured people should
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 9 of 16

be transferred to temporary hospitals or emergency function and the pessimistic action for the non-determin-
rooms. The relation (16) shows the percentage of peo- istic constraints. To do this, first consider the abbreviated
ple with the demand points that should be transferred form of the presented model:
to the shelter. Equation (17) shows the number of peo-
ple needing the emergency center from the demand minZ = F .y + C s .xs
point. The relation (18) shows the percentage of the s.t. :
injured who must be transferred from the emergency Axs ≥ D s (24)
center to the hospital for treatment. Relationship (19) Exs ≤ Sy
shows that if an emergency center is established, that y ∈ {0,1}, xs ≥ 0
center can be used to transfer the injured to hospitals. where the vectors D s , F , and S represent the fixed
s, C
Equation (20) shows the capacity of using shelters. construction cost, variable cost (transportation), demand
Relationships (21) and (22) show the capacity of emer- amount, and facility capacity. Also, A and E are the
gency centers and hospitals. The relation (23) shows matrix of coefficients, and finally xs and y are continuous
the type and gender of the decision-making variables. variables and zero and one under scenario s, respectively.
Due to the dynamic and fluctuating nature of some Now, it is assumed that the vectors D  s and C
s in the above
important parameters (including transportation costs model are presented as non-deterministic parameters.
and demand) that are beyond planning, as well as the According to the general form of non-deterministic lim-
unavailability and unattainability of historical data ited programming, the expected value of the objective
required in the design phase, these parameters are function and the pessimistic fuzzy should be obtained
mainly based on experts’ opinions and subjective expe- to deal with the objective function and the non-deter-
riences. It is estimated. Therefore, the above fuzzy ministic constraint, respectively. Now, according to the
parameters are formulated as non-deterministic data abbreviated form, the pessimistic fuzzy basic model is as
in the form of trapezoidal fuzzy numbers. It is worth follows:
noting that evaluating the transportation cost and exact
demand for long-term decisions is complicated and minZ = E[Z] = F .y + E[C s ].xs
sometimes even impossible. Even if one can estimate a s.t. :
probability distribution function for these parameters,
 
NEC Axs ≥ D  s ≥ αs (25)
it is possible that these parameters do not behave the Exs ≤ Sy
same as the past data. Therefore, these parameters that y ∈ {0,1}, xs ≥ 0
change in a long-term planning horizon are considered
fuzzy data. where αs controls the minimum degree of certainty of
In addition, for the non-deterministic (probable) con- establishing the non-deterministic limit with a (pessimis-
straints with non-deterministic data on the left or right tic) decision-making approach. According to the trap-
side of their equality, the non-deterministic constraint pro- ezoidal probability distribution for fuzzy parameters, the
gramming method is usually used. Suppose this method general form of relations (25) is as follows:
controls the confidence level of establishing these uncer-  1 2 3 4
C +C +C +C
tain limits. In that case, the concept of decision-making minZ = F .y + s s 4 s s xs
can obtain the minimum confidence level as a suitable safe s.t. :
margin for establishing each of these limits. To do this, two Axs ≥ (1 − αs )Ds3 + αs Ds4 (26)
measures of the standard fuzzy method, optimistic and Exs ≤ Sy
pessimistic fuzzy, are commonly used. It is worth mention- y ∈ {0,1}, xs ≥ 0
ing that the optimistic fuzzy represents the level of opti-
mistic probability of the occurrence of an uncertain event, In probabilistic models, the minimum confidence level
including uncertain parameters. In contrast, the pessimistic to establish an uncertain constraint should be deter-
fuzzy represents a pessimistic decision about an uncertain mined by decision preferences. However, as seen in the
event. However, it is more conservative to use a pessimistic presented models, the objective function is not sensitive
hybrid fuzzy; we assume that decision-making has an opti- to the deviation from its expected value, which means
mistic attitude to establish non-deterministic limits. Now, that achieving stable solutions in the basic model is not
an axiomatic equivalent of the original non-deterministic guaranteed. In such cases, a high risk may be imposed
model can be formulated based on the mentioned fuzzy on decision-making in many real cases, especially in
parameters and using the expected value for the objective
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 10 of 16

strategic decisions where the stabilization of the solution J 

 K N
 J 
 C J 
 C1
L 
Zmin = Fjk .Xjk + Fn .Un + hjc .Ijc + blc1 .Qljc1 +
is critical to a large extent. Therefore, to deal with this j=1 k=1 n=1 j=1 c=1 j=1 l=1 c1 =1

inefficient situation, the non-deterministic robust-prob- J 

 C2
L  J 
 L  V1
C 
blc2 .Qljc2 + Tr ljcv1 .Qljc +
abilistic programming approach is used. This approach j=1 l=1 c2 =1 j=1 l=1 c=1 v1 =1
takes advantage of the advantages of both robust and S
 J 
 C
 I

 S
 J 
 C1
L 

probabilistic planning, which clearly makes it differ- s=1

Ps
j=1 c=1
hjc Ijcs −
i=1
Yjics +
s=1
Ps
j=1 l=1 c1 =1
blc1 s .Qljc1 s +

ent from other uncertainty planning approaches. In this S J  C2

L  S J 
L  V1
C 

research, uncertainty planning is applied based on the

   
Ps blc2 s .Qljc2 s + Ps Tr 1ljcv1 s .Qljcs +
s=1 j=1 l=1 c2 =1 s=1 j=1 l=1 c=1 v1 =1
presented model, which is as follows: S L 
I  C S J 
I  C
   
   Ps Tr 1lics .Ylics + Ps Tr 1jics .Yjics +
minω1 = E[Z] + ξ ′ Z(max) − Z(min) + ω s Ps {E[Z] − E[Zs ] + 2θs } s=1 i=1 l=1 c=1 s=1 i=1 j=1 c=1
  4 (α −)Ds +(1−αs )Ds3
4

η s Ps Ds − s 1− + S
  H
I  S
 I 
 N
Ps Tr 1ihs .XH ihs + Ps Tr 1ins .XN ins +
s.t. :
Axs ≥ (1 − αs )Ds3 + αs Ds4
(27) s=1 i=1 h=1 s=1 i=1 n=1
S
  H
N  S
 I 
 G
Exs ≤ Sy Ps Tr 1nhs .NH nhs + Ps Tr 1igs .XG igs +
s=1 n=1 h=1 s=1 i=1 g=1
y ∈ {0,1}, xs ≥ 0, αs ∈ [0.1,0.9]
S
 J 
 C
I    
3 4
Ps Ccic . (1 − αs )dics + αs dics − Yjics +
where M is a vast non-negative number and Z(max) and s=1 j=1 i=1 c=1

Z(min) and E[Zs ] can be expressed as follows: S

  I 
L  C 
3 4
 
Ps Ccic . (1 − αs )dics + αs dics − Ylics
s=1 l=1 i=1 c=1
Z(max) = F .y +  s Ps Cs4 xs

1
(31)
Z(min)
 = F .y + s Ps Csxs J 
K N J 
C J  C1
L 
1− 
    
E[Zs ] = F .y + 2 Cs + Cs + 2 Cs3 + Cs4 .xs
1 2 Zmax =
j=1 k=1
Fjk .Xjk +
n=1
Fn .Un +
j=1 c=1
hjc .Ijc +
j=1 l=1 c1 =1
blc1 .Qljc1 +

(28) J 
 C2
L  J 
 L 
C
blc2 .Qljc2 + Tr ljc .Qljc +
In the first objective function of Eq. (27), the first term j=1 l=1 c2 =1 j=1 l=1 c=1

refers to the expected value of the first objective func- S J 

C I S J  C1
L 
 
    
Ps hjc Ijcs − Yjics + Ps blc1 s .Qljc1 s +
tion using the average values of the uncertain parameters s=1 j=1 c=1 i=1 s=1 j=1 l=1 c1 =1

of the model. The second term refers to the penalty cost S

 J 
 C2
L  S
 J 
 L  V1
C 
Tr 4ljcv1 s .Qljcs +
for deviation from the expected value of the first objec- s=1
Ps
j=1 l=1 c2 =1
blc2 s .Qljc2 s +
s=1
Ps
j=1 l=1 c=1 v1 =1
tive function (stability of optimality). The third sentence S
  L 
I  C S
 J 
I 
 C
Tr 4lics .Ylics + Tr 4jics .Yjics +
also shows the total cost of the penalty for deviation from s=1
Ps
i=1 l=1 c=1 s=1
Ps
i=1 j=1 c=1
the demand (non-deterministic parameter). Therefore, S H
I  S N
I 

the parameter ξ ′ of the weighting coefficient of the objec-

   
Ps Tr 4ihs .XH ihs + Ps Tr 4ins .XN ins +
s=1 i=1 h=1 s=1 i=1 n=1
tive function, η is the penalty cost of not estimating the S H
N  S G
I 

demand. αs parameters as correction coefficients in the

   
Ps Tr 4nhs .NH nhs + Ps Tr 4igs .XG igs +
s=1 n=1 h=1 s=1 i=1 g=1
value of fuzzy levels show the numbers, which should be S J  C
I 
a number between 0.1 and 0.9.
    
3 4
Ps Ccic . (1 − αs )dics + αs dics − Yjics +
s=1 j=1 i=1 c=1

MinZ1 =ξ ′(Zmax − Zmin) + ω Ps {E[Z] − E[Zs ] + 2θs } S
  I 
L  C   
3 4
s Ps Ccic . (1 − αs )dics + αs dics − Ylics
C 
I  S
 
(29) s=1 l=1 i=1 c=1
(32)
4 + (1 − α )d 3
(αs − )dics s ics

4
+η dics −
1−
i=1 c=1 s=1

s.t. :

Eqs(2) − (23) (30)

Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 11 of 16

J 
 K N
 J 
 C J 
 C1
L 
Table 1 Numerical example size
E[Z] = Fjk .Xjk + Fn .Un + hjc .Ijc + blc1 .Qljc1 +
j=1 k=1 n=1 j=1 c=1 j=1 l=1 c1 =1
Set Size Set Size Set Size
J 
 C2
L  J 
 L 
C S
 J 
 C2
L 
blc2 .Qljc2 + Tr ljc .Qljc + Ps blc2 s .Qljc2 s +
I 5 C1 2 N 3
j=1 l=1 c2 =1 j=1 l=1 c=1 s=1 j=1 l=1 c2 =1

S
 J 
 C
 I

 S
 J 
 C1
L  J 3 C2 2 H 3
Ps hjc Ijcs − Yjics + Ps blc1 s .Qljc1 s + K 4 S 2 G 4
s=1 j=1 c=1 i=1 s=1 j=1 l=1 c1 =1
L 4 C 4
S J 
L 
C Tr 1ljcs + Tr 2ljcs + Tr 3ljcs + Tr 4ljcs
 
 
Ps .Qljcs +
4
s=1 j=1 l=1 c=1

S I  L  C
 
Tr 1lics + Tr 2lics + Tr 3lics + Tr 4lics
Analysis of the numerical example
 
Ps .Ylics +
4
s=1 i=1 l=1 c=1
S I  J  C

Tr 1jics + Tr 2jics + Tr 3jics + Tr 4jics
 In this part of the research, a numerical example has
been solved in a small size corresponding to the sizes
 
Ps .Yjics +
4
s=1 i=1 j=1 c=1

S I  H
  presented in Table 1.
Tr 1ihs + Tr 2ihs + Tr 3ihs + Tr 4ihs
This study presents a robust-fuzzy-probabilistic plan-
 
Ps .XH ihs
4
ning approach to optimize humanitarian relief logistics
s=1 i=1 h=1
S I  N
 
Tr 1ins + Tr 2ins + Tr 3ins + Tr 4ins
networks while considering the cost of deprivation. The
 
+ Ps .XN ins +
4
proposed model accounts for the inherent complexities
s=1 i=1 n=1
S N  H
 
Tr 1nhs + Tr 2nhs + Tr 3nhs + Tr 4nhs
and uncertainties in disaster scenarios, including fluctu-
 
Ps .NH nhs +
4
s=1 n=1 h=1
 1 ating demand, unpredictable supply chains, and the criti-
S I  G Tr igs + Tr 2igs + Tr 3igs + Tr 4igs

cal need to minimize human suffering. The methodology
 
Ps .XG igs +
4
s=1 i=1 g=1
encompasses model formulation, data generation, and
S J  C
I 
solution strategies, which are elaborated below.
    
3 4
Ps Ccic . (1 − αs )dics + αs dics − Yjics +
s=1 j=1 i=1 c=1

S
 L 
 C
I    
3 4
Ps Ccic . (1 − αs )dics + αs dics − Ylics
s=1 l=1 i=1 c=1 Model formulation
(33) The model integrates deprivation costs into a nonlin-
E[Zs ] =
J 
 K
Fjk .Xjk +
N

Fn .Un +
J 
 C
hjc .Ijc +
J 
 C1
L 
blc1 .Qljc1 +
ear optimization framework. The objective function is
j=1 k=1 n=1 j=1 c=1 j=1 l=1 c1 =1 designed to minimize a weighted combination of depri-
J 
 C2
L 
blc2 .Qljc2 +
J 
 L  V1
C 
Tr ljcv1 .Qljc +
J 
 C2
L 
Ps .blc2 s .Qljc2 s +
vation costs, transportation costs, and facility operation
j=1 l=1 c2 =1 j=1 l=1 c=1 v1 =1 j=1 l=1 c2 =1 costs. Constraints ensure equitable resource distribution,
J 
 C
 I

 S
 J 
 C1
L  adherence to budgetary limitations, and timely deliv-
Ps .hjc Ijcs − Yjics Ps blc1 s .Qljc1 s +
ery of aid. The model incorporates robust optimization
+
j=1 c=1 i=1 s=1 j=1 l=1 c1 =1

J 
 L 
C  
1−  1      techniques to address demand uncertainties, fuzzy logic
Tr ljcs + Tr 2ljcs + Tr 3ljcs + Tr 4ljcs .Qljcs +
j=1 l=1 c=1
Ps
2 2 to handle imprecise data, and probabilistic components
 L 
I  C  
1−  1      to model scenario-dependent variability. This hybrid
i=1 l=1 c=1
Ps
2
Tr lics + Tr 2lics +
2
Tr 3lics + Tr 4lics .Ylics +
approach ensures flexibility and adaptability in real-world
 J 
I  C  
1−  1      applications.
Ps Tr jics + Tr 2jics + Tr 3jics + Tr 4jics .Yjics +
2 2
i=1 j=1 c=1

H
I        
1−  1
Table 2 Interval limits of deterministic parameters of the

Ps Tr ihs + Tr 2ihs + Tr 3ihs + Tr 4ihs .XH ihs +
2 2
i=1 h=1
N
I   
problem
 1−     
Ps Tr 1ins + Tr 2ins + Tr 3ins + Tr 4ins .XN ins +
i=1 n=1
2 2 Parameter Interval limits Parameter Interval limits
N 
H       
 1− θc ∼ U(0.5 − 1) Fjk ∼ U(10000 − 12000)
Ps Tr 1nhs + Tr 2nhs + Tr 3nhs + Tr 4nhs .NH nhs +
2 2
n=1 h=1 δn 0.3 Fn ∼ U(5000 − 9000)
G
I        
 1−  1 ωi 0.4 blc1 ∼ U(0.1 − 0.15)
Ps Tr igs + Tr 2igs + Tr 3igs + Tr 4igs .XG igs +
2 2
i=1 g=1 Vk ∼ U(10 − 50) blc2 ∼ U(0.1 − 0.15)
J 
I 
C
 
3
Ps .Ccic . (1 − αs )dics 4
+ αs dics
 
− Yjics +
CapH h ∼ U(1000 − 1200) hjc ∼ U(0.05 − 0.15)
j=1 i=1 c=1 CapGg ∼ U(1000 − 1200) Tr ljc ∼ U(0.2 − 0.25)
I 
L  C
 
3 4
  CapN n ∼ U(1000 − 1200) vlc1 ∼ U(40 − 45)
Ps .Ccic . (1 − αs )dics + αs dics − Ylics , ∀s
l=1 i=1 c=1 σlc1 ∼ U(0.1 − 0.15) vlc2 ∼ U(40 − 45)
(34) σlc2 ∼ U(0.1 − 0.15) ωP i 0.2
Cc ic ∼ U(70 − 80)
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 12 of 16

Table 3 Interval limits of uncertain parameters of the problem

Scenario 1
Parameter Interval limits Parameter Interval limits
cjirs ∼ (0.07 − 0.10) vlc1 s ∼ (30 − 35)
blc1 s ∼ (0.08 − 0.12) vlc2 s ∼ (30 − 35)
blc2 s ∼ (0.08 − 0.12) wis ∼ (150 − 200)
ps 0.5
Level 1 Level 2 Level 3 Level 4
dics ∼ (200 − 250) ∼ (250 − 300) ∼ (300 − 350) ∼ (350 − 400)
Tr ljcs ∼ (0.2 − 0.25) ∼ (0.25 − 0.3) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40)
Tr lics ∼ (0.2 − 0.25) ∼ (0.25 − 0.3) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40)
Tr jics ∼ (0.2 − 0.25) ∼ (0.25 − 0.3) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40)
Tr ihs ∼ (0.18 − 0.22) ∼ (0.22 − 0.28) ∼ (0.28 − 0.32) ∼ (0.32 − 0.35)
Tr ins ∼ (0.18 − 0.22) ∼ (0.22 − 0.28) ∼ (0.28 − 0.32) ∼ (0.32 − 0.35)
Tr nhs ∼ (0.18 − 0.22) ∼ (0.22 − 0.28) ∼ (0.28 − 0.32) ∼ (0.32 − 0.35)
Tr igs ∼ (0.18 − 0.22) ∼ (0.22 − 0.28) ∼ (0.28 − 0.32) ∼ (0.32 − 0.35)

Scenario 2
Parameter Interval limits Parameter Interval limits
cjirs ∼ (0.10 − 0.13) vlc1 s ∼ (35 − 40)
blc1 s ∼ (0.12 − 0.16) vlc2 s ∼ (35 − 40)
blc2 s ∼ (0.12 − 0.16) wis ∼ (200 − 250)
ps 0.5
Level 1 Level 2 Level 3 Level 4
dics ∼ (250 − 300) ∼ (300 − 350) ∼ (350 − 400) ∼ (400 − 450)
Tr ljcs ∼ (0.25 − 0.3) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40) ∼ (0.40 − 0.45)
Tr lics ∼ (0.25 − 0.3) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40) ∼ (0.40 − 0.45)
Tr jics ∼ (0.25 − 0.3) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40) ∼ (0.40 − 0.45)
Tr ihs ∼ (0.22 − 0.26) ∼ (0.26 − 0.30) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40)
Tr ins ∼ (0.22 − 0.26) ∼ (0.26 − 0.30) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40)
Tr nhs ∼ (0.22 − 0.26) ∼ (0.26 − 0.30) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40)
Tr igs ∼ (0.22 − 0.26) ∼ (0.26 − 0.30) ∼ (0.3 − 0.35) ∼ (0.35 − 0.40)

Data generation for solving nonlinear programming problems, as it com-

Due to the lack of access to real-world disaster relief data, bines branch-and-bound techniques with deterministic
this study employs synthetic data generated using a uni- global optimization algorithms. The solver identifies optimal
form distribution function. The uniform distribution is or near-optimal solutions by efficiently exploring the solu-
selected for its simplicity and capacity to simulate various tion space, ensuring the results are feasible and high-quality.
plausible scenarios. Key parameters, including demand, The solution process begins by inputting the gener-
supply availability, transportation costs, and facility ated data into the model and running the BARON solver
capacities, are generated within predefined ranges to under various disaster scenarios. The results are analyzed
mimic realistic disaster conditions. These parameters are to evaluate the trade-offs between cost minimization and
calibrated to ensure a balanced representation of low-, the timeliness of aid delivery. Sensitivity analysis assesses
medium-, and high-demand scenarios. the impact of key parameters, such as demand variability
and transportation delays, on the model’s performance.
Solution strategy Additionally, robustness tests are performed to validate
The optimization model’s nonlinear nature necessitates the model’s effectiveness under worst-case scenarios.
using advanced solvers capable of handling global optimi- Tables 2 and 3 show the range limits of determinis-
zation problems. For this purpose, the BARON solver in tic and non-deterministic parameters of the developed
GAMS 23.5.2 software is employed. BARON is well-suited mathematical model.
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 13 of 16

Table 4 The number of people affected by accidents and Table 6 Costs of the entire logistics network under the scenario
transferred to temporary shelters, hospitals, and emergency and different uncertainty rates
rooms
α2 = α1 Total costs P1 = 1 − P2 Total costs
Demand point Temporary Scenario 1 Scenario 2
0.1 327,091.440 0.1 480,670.6
emergency
0.2 353,683.256 0.2 467,488.3
XN ins 1 2 30 47
0.3 370,095.073 0.3 449,255.3
2 1 41
0.4 392,175.312 0.4 428,288.6
2 3 35
0.5 413,098.706 0.5 413,098.7
3 1 40 46
0.6 433,512.671 0.6 400,348.7
4 2 35
0.7 458,774.201 0.7 387,443.7
4 3 41
0.8 479,684.340 0.8 376,958.9
5 1 40
0.9 493,514.672 0.9 364,473.6
5 3 42

Demand point Hospital Scenario 1 Scenario 2

XH ihs 1 2 62 96
2 1 70 83 Table 7 Costs of the entire logistics network under different
values of deprivation cost
3 1 94
3 2 80 Cc % Z1
4 3 73 84
50 − % 408,682.5
5 1 80
30 − % 409,745.2
5 3 86
10 − % 411,279.6
Temporary emer- Hospital Scenario 1 Scenario 2 0 413,098.7
gency 10 + % 418,328.7
NH nhs 1 1 24 26 20 + % 425,675.3
2 2 19 14 30 + % 429,744.6
3 2 10
3 3 24

Demand point Shelter Scenario 1 Scenario 2 Table 8 Costs of the entire logistics network under different
XGigs 1 1 95 values of model robustness
1 4 61 ξ′ Total cost η Total cost
2 1 82
1 413,098.7 2 413,098.7
2 2 70
10 426,841.7 4 418,223.8
3 2 80 92
20 462,846.8 10 429,476.6
4 1 83
100 497,842.0 15 433,674.8
4 2 71
5 3 80 84

After designing the numerical example and determin-

ing the interval limits of the problem’s deterministic
and non-deterministic parameters, the numerical exam-
ple analysis was carried out. Also, due to the use of the
Table 5 Amount of relief items purchased in the pre-disaster
phase robust-fuzzy-probabilistic optimization method to con-
trol the non-deterministic parameters of the problem,
Supplier Distribution Eco-friendly Environmentally the effective technical coefficients are considered as
center incompatible
ξ ′ = 1, ω = 2, η = 2,  = 0.4. After solving the problem,
Product Product Product Product 4 the objective function value equal to 413,098.71 has been
1 2 3
obtained, which is discussed in the following decision
1 1 4.912 4.912 4.912 4.912 variables.
1 3 4.932 4.932 4.932 4.932 Table 4 shows the number of people injured and trans-
2 2 4.881 4.881 4.881 4.881 ferred to other levels of the supply chain network.
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 14 of 16

Table 5 also shows the number of relief items pur- Therefore, the change in the most critical parameters
chased from suppliers by the relief distribution center in related to the robustness of the model will lead to a
the pre-disaster phase. change in the total costs and decision variables. Table 8
Based on this, supplier numbers 1 and 2 supplied prod- shows the relief logistics network’s total costs in different
ucts before the disaster. Also, three centers, Nos. 1, 2, and optimization parameter values.
3, were built as temporary emergency centers and pro- By examining the effects of the weighting coefficient of
vided services to people affected by accidents. the objective function on the values of the objective func-
The following discusses the problem’s sensitivity anal- tion of the problem, it can be seen that with the increase
ysis under different parameters. Due to the problem’s of this weighting coefficient, the value of the objective
nonlinearity, the process of changing the values of the function has increased due to the rise in the volume of
objective functions of the problem under the uncertainty transfers of relief items. Also, the results of Table 8 show
rate has been investigated. The previous section solved that by increasing the penalty cost of demand estimation,
the numerical example, assuming an uncertainty rate of the model tries to estimate the maximum demand of the
0.5. Table 6 shows the total costs of the logistics network affected points by increasing the costs of distribution,
under different uncertainty rates and scenarios. transmission, etc. This has led to increased transfer of
Increase in the uncertainty rate, and due to the increase relief items between different centers.
in the number of people affected by the accident and the Pandemics pose significant challenges to humanitar-
demand from the accident site for relief items, the costs ian logistics, particularly in the timely distribution of
related to the supply and transportation of the items and medical supplies, protective equipment, and vaccines.
the injured have increased. Therefore, it can be seen that In such scenarios, the proposed model can play a critical
with the increase of the uncertainty rate of as much as role in mitigating deprivation costs, which are exception-
90%, the total costs have increased by 19.46%. On the other ally high due to the direct impact of delayed medical aid
hand, the results of Table 6 show that with the increase in on mortality and morbidity rates. The model can simu-
the probability of the second scenario, the costs of pro- late varying infection rates across regions, incorporat-
viding, producing, and distributing relief products have ing probabilistic elements to account for uncertainties
increased due to the rise in demand in the network. like sudden demand spikes or supply chain disruptions
On the other hand, in the presented model, it was caused by overwhelmed infrastructure or labor shortages.
impossible to meet all the demands of the affected areas Using a robust-fuzzy-probabilistic framework, the model
from relief items due to the disruption. Therefore, the prioritizes allocating limited resources to regions where the
model tries to send the most relief items to the affected deprivation costs are most severe, such as areas with high
areas by considering the balance of the problem. Thus, if population densities, low healthcare capacity, or delayed
the aid distribution centers do not meet the demand of the response times. By optimizing supply distribution routes
affected areas, the exclusion fee will be applied. In this sec- and facility operations, the model ensures equitable access
tion, by examining the effect of increasing and decreasing to critical resources while minimizing total costs. For
the cost of deprivation on the network, its impact on the instance, vaccine distribution during a pandemic can ben-
objective functions of the problem has been measured. efit from the model’s ability to handle fluctuating demand,
Therefore, the deprivation cost value is considered 10, 30, ensuring that vulnerable populations are prioritized.
and 50% lower and higher than the base value applied in This application also allows sensitivity analysis to assess
random data. Table 7 shows the total cost of the logistics the impact of various constraints, such as supply delays
network under the change in the deprivation cost values. or resource shortages. This provides actionable insights
According to Table 7’s results, the model aims to for policymakers to improve resilience and effectiveness
increase the number of goods transferred to fully satisfy in managing pandemic logistics.
the demand by increasing the deprivation cost. There-
fore, the expenses related to deprivation have decreased, Conclusion
but on the other hand, the costs associated with the The importance of rescue and relief of the injured during
transfer of products have increased. As a result, the cost an accident caused this research to model humanitarian aid
of the entire relief logistics network has increased. logistics under uncertainty. The designed logistics network
In this research, a robust-fuzzy-probabilistic optimi- model minimizes the costs of the entire relief logistics net-
zation method is used to control the non-deterministic work. The most important decisions taken in this model are
parameters of the model. In this method, to prevent fuzzy determining the optimal location of temporary emergen-
numbers and the uncertainty rate to justify the prob- cies and distribution centers for the distribution of relief
lem and guarantee the maximum demand of damaged items, determining the optimal amount of accumulated
points, the robust method is used in different scenarios. relief goods in the distribution centers, and determining the
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 15 of 16

optimal amount of allocation of the transfer flow of goods Authors’ contributions

All authors contributed to the study’s conception and design. JGN, HN, and ASJ
and casualties between facilities, optimal routing. Since the performed material preparation, data collection, and analysis. All authors wrote
model’s parameters are defined as fuzzy numbers in differ- the first draft of the manuscript, and all authors commented on previous ver‑
ent scenarios, the new robust-fuzzy-probabilistic method sions of the manuscript. All authors read and approved the final manuscript.
was used to control the parameters and transform the non- Funding
deterministic model into a solvable deterministic model. By The authors did not receive support from any organization for the submitted work.
analyzing the sensitivity of various parameters of the prob-
Data availability
lem, it was also observed that with the increase in the uncer- This is not applicable. We do not analyze or generate datasets because our
tainty rate and as a result of the increase in the number of work proceeds using a theoretical and mathematical approach.
people affected by the accident and the demand for relief
items at the accident site, the costs related to the supply and Declarations
transportation of the items and the injured have increased.
Competing interests
Also, by examining the effect of deprivation cost on the The authors have no relevant financial or non-financial interests to disclose.
results of the model, it was observed that with the increase
of deprivation cost, the model aims to increase the amount
Received: 30 August 2024 Accepted: 6 December 2024
of goods transferred to satisfy the demand fully. Therefore,
the expenses related to deprivation have decreased, but on
the other hand, the costs associated with the transfer of
products have increased. As a result, the cost of the entire
References
relief logistics network has increased. Finally, by examining Abazari SR, Jolai F, Aghsami A (2022) Designing a humanitarian relief network
the effects of the weighting coefficient of the objective func- considering governmental and non-governmental operations under
tion on the values of the objective function of the problem, uncertainty. Int J Syst Ass Eng Man 13(3):1430–1452
A. Silva and A. Leiras, "Deprivation costs in humanitarian logistics: a systematic
it can be seen that with the increase of this weighting coef- review," in Proc. 17th Int. Conf. on Hum Log, 2021, pp. 45–56.
ficient, the value of the objective function has increased due Akbari F, Valizadeh J, Hafezalkotob A (2021) Robust cooperative planning of
to the increase in the volume of transfers of relief items. relief logistics operations under demand uncertainty: a case study on a
possible earthquake in Tehran. Int J Syst Sci: 1–24
While offering a robust framework for humanitarian relief Boostani A, Jolai F, Bozorgi-Amiri A (2021) Designing a sustainable humanitar‑
logistics, this research has certain limitations. First, the reli- ian relief logistics model in pre-and postdisaster management. Int J Sust
ance on synthetic data generated via uniform distribution Transp 15(8):604–620
Boonmee C, Arimura S, Asada P (2020) A multi-objective fuzzy mathematical
limits the realism of the analysis, as actual disaster scenar- programming model for humanitarian relief logistics. Oper Res in Hum
ios often exhibit complex, region-specific demand patterns Con 3(2):89–102
and supply constraints. Second, the nonlinear nature of the Cao C, Li C, Yang Q, Zhang F (2017) Multi-objective optimisation model of
emergency organisation allocation for sustainable disaster supply chain.
optimization model, despite being handled by the BARON Sust 9(11):2103
solver, can lead to increased computational complexity, Cheng J, Feng X, Bai X (2021) Modeling equitable and effective distribution
particularly for large-scale problems with numerous deci- problem in humanitarian relief logistics by robust goal programming.
Comp Ind Eng 155:107183
sion variables and scenarios. Third, the focus on probabil- Elçi Ö, Noyan N (2018) A chance-constrained two-stage stochastic pro‑
istic and fuzzy methods introduces model assumptions that gramming model for humanitarian relief network design. Trans Res B
may oversimplify real-world uncertainties, such as dynamic 108:55–83
Ghaffari Z, Nasiri MM, Bozorgi-Amiri A, Rahbari A (2020) Emergency supply
changes in transportation networks or sudden shifts in chain scheduling problem with multiple resources in disaster relief
demand. Additionally, the model does not fully account for operations. Transportmetrica A 16(3):930–956
behavioral factors like community responsiveness or organi- Ghasemi P, Goodarzian F, Abraham A (2022) A new humanitarian relief logistic
network for multi-objective optimisation under stochastic programming.
zational collaboration challenges. Lastly, while deprivation Appl Intell: 1–34
costs are central to the model, quantifying them accurately Hoseininezhad F, Makui A, Tavakkoli-Moghaddam R (2021) Pre-positioning of a
in practice can be challenging due to the lack of universally relief chain in humanitarian logistics under uncertainty in road accidents:
a real-case study. S Afr J Ind Eng 32(1):86–104
accepted metrics and data. Hong JD (2020) Application of integrated multiple criteria data envelopment
The researchers offer the following suggestions to analysis to humanitarian logistics network design. J Sys Sci Syst Eng
improve the model and develop it: 29(6):709–729
Holguín-Veras J, Jaller M, Van Wassenhove LN, Pérez N, Wachtendorf T (2013)
On the appropriate objective function for post-disaster humanitarian
• Using reliability in the correct transfer of relief items logistics models. J of Oper Man 31(5):262–280
Jamali A, Ranjbar A, Heydari J, Nayeri S (2021) A multi-objective stochastic
between centers
programming model to configure a sustainable humanitarian logistics
• Using the routing of helicopters to distribute relief considering deprivation cost and patient severity. Ann oper res 1–36
items instead of vehicles Mansoori S, Bozorgi-Amiri A, Pishvaee MS (2020) A robust multi-objective
humanitarian relief chain network design for earthquake response,
• Using precise and innovative methods in problem-
with evacuation assumption under uncertainties. Neu Comp App
solving 32(7):2183–2203
Ghahremani‑Nahr et al. Journal of International Humanitarian Action (2024) 9:19 Page 16 of 16

Mohamadi A, Yaghoubi S (2017) A bi-objective stochastic model for emergency

medical services network design with backup services for disasters under
disruptions: an earthquake case study. Int j of Dis Risk Red 23:204–217
Modiri M, Eskandari M, Hasanzadeh S (2022) Multi-objective modeling of relief
items distribution network design problem in disaster relief logistics
considering transportation system and CO2 emission. Scientia Iranica
Noham R, Tzur M (2018) Designing humanitarian supply chains by incorporat‑
ing actual postdisaster decisions. Eur J Oper Res 265(3):1064–1077
Paul JA, Wang XJ (2019) Robust location-allocation network design for earth‑
quake preparedness. Trans Res B 119:139–155
Rodríguez-Espíndola O, Albores P, Brewster C (2018) Disaster preparedness in
humanitarian logistics: a collaborative approach for resource manage‑
ment in floods. Eur J Oper Res 264(3):978–993
Salehi F, Mahootchi M, Husseini SMM (2019) Developing a robust stochastic
model for designing a blood supply chain network in a crisis: a possible
earthquake in Tehran. Ann Oper Res 283(1):679–703
Shu J, Song M, Wang B, Yang J, Zhu S (2022) Humanitarian relief network
design: responsiveness maximisation and a case study of Typhoon Ram‑
masun. IISE Trans: 1–13
Sharma R, Mehta K (2021) Probabilistic fuzzy goal programming for emer‑
gency relief supply. Ann of Oper Res 293(2):345–360
Torabi SA, Shokr I, Tofighi S, Heydari J (2018) Integrated relief pre-positioning
and procurement planning in humanitarian supply chains. Trans Res E
113:123–146
Tatham P, Christopher M. (Eds.) (2018) Humanitarian logistics: meeting the
challenge of preparing for and responding to disasters. Kogan Page
Publishers
van Steenbergen J, Pieters R, Van der Vaart C (2023) Stochastic dynamic
models for deprivation cost optimization in disaster response. J of Oper
Res 12(4):334–348
Yu L, Zhang C, Yang H, Miao L (2018) Novel methods for resource allocation
in humanitarian logistics considering human suffering. Comp Ind Eng
119:1–20
Yahyaei M, Bozorgi-Amiri A (2018) Robust reliable humanitarian relief network
design: an integration of shelter and supply facility location. Ann Oper
Res 1–20
Zanganeh M, Ebrahimnejad S, Moosavi A, Tayebi H (2019) A bi-objective
model for humanitarian logistics network design in response to postdis‑
aster. Int J Log Syst Man 33(2):256–272

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in pub‑
lished maps and institutional affiliations.