A Multi-method Simulation Environment for Humanitarian Supply Chains

Abstract

In this paper we present the design, prototype and evaluation of a multi-method simulation environment for humanitarian supply chains. The degree of preparedness of a
logistics setup highly determines the humanitarian impact. In order to construct comprehensive models of the associated humanitarian supply chains and to support the decision making process of humanitarian actors in a significant manner, the application of multi-method simulation environments promise to be an appropriate approach to complement and enhance the existing decision making practices. The presented simulation environment was developed following the design science approach. Practitioner needs and results from research were identified in order to define the scope and design of the simulation environment. The artefact was then applied and
evaluated in a hypothetical tsunami scenario in the Mediterranean Sea. We used the SPHERE standards to determine the needs of the affected population and average transportation as well as warehousing costs to analyze the effects on the relief chains of three involved humanitarian organizations. The conducted simulation study investigates the effect of a horizontal collaboration between the involved actors. The undertaken parametrizations illustrate potential benefits of the
logistics planning approach.

Full text

978-1-5386-2553-8/17/$31.00 ©2017 IEEE
A Multi-method Simulation Environment for
Humanitarian Supply Chains
Adam Widera
Deptartment of Information
Systems,
University of Münster (WWU)
Münster, Germany
adam.widera@ercis.uni-
muenster.de
Christian Konradt
PricewaterhouseCoopers AG
Frankfurt am Main, Germany
konradt.christian@de.pwc.com
Carsten Böhle
Lufthansa Industry Solutions
Norderstedt, Germany
carsten.boehle@lhind.dlh.de
Bernd Hellingrath
Deptartment of Information
Systems,
University of Münster (WWU)
Münster, Germany
bernd.hellingrath@ercis.uni-
muenster.de
Abstract— In this paper we present the design, prototype and
evaluation of a multi-method simulation environment for
humanitarian supply chains. The degree of preparedness of a
logistics setup highly determines the humanitarian impact. In
order to construct comprehensive models of the associated
humanitarian supply chains and to support the decision making
process of humanitarian actors in a significant manner, the
application of multi-method simulation environments promise to
be an appropriate approach to complement and enhance the
existing decision making practices. The presented simulation
environment was developed following the design science
approach. Practitioner needs and results from research were
identified in order to define the scope and design of the
simulation environment. The artefact was then applied and
evaluated in a hypothetical tsunami scenario in the
Mediterranean Sea. We used the SPHERE standards to
determine the needs of the affected population and average
transportation as well as warehousing costs to analyze the effects
on the relief chains of three involved humanitarian organizations.
The conducted simulation study investigates the effect of a
horizontal collaboration between the involved actors. The
undertaken parametrizations illustrate potential benefits of the
logistics planning approach.
Keywords—Humanitarian logistics, humanitarian supply chain
management, multi-method simulation, SPHERE, simulation study
I. INTRODUCTION
Since the raise of the humanitarian logistics research in the
last decade, quantitative methods dealing with logistics
planning and execution have been recognized as promising
approaches to improve the impact of humanitarian operations
(Kunz and Reiner 2012). Due to many distinct reasons, both
the number as well as the humanitarian and financial impact of
natural and man-made disasters has risen over the previous
century and forecasts even outline a further increase for the
next years (Thomas and Kopczak 2005, p. 1). For instance, the
Typhoon Haiyan affected at least 11.3 million people in nine
regions of the Philippines, 6,201 persons were killed and a
widespread destruction of homes and infrastructure transpired
(approximated economic loss of about $518 million) (Swiss
Reinsurance Company 2015, p. 42; United Nations 2014, p. 3).
In order to address the humanitarian needs, collective
efforts of the international community and humanitarian
operations are frequently required. Humanitarian operations
contain different activities with the goal to reduce human
suffering and to save lives. While short-term actions mainly
happen in response to immediate emergencies, medium and
long-term relief concentrates more on the recovery and
reconstruction in post-emergency situations. Because of the
mobilization and deployment of material as well as financial
resources, logistics activities are a pivotal part of humanitarian
operations (Blecken 2010, p. 1; Thomas and Kopczak 2005, p.
2). Their effective and efficient execution is essential for the
success of humanitarian operations (McGuire 2006, p. 1;
Thomas and Kopczak 2005, p. 2; Tomasini and Van
Wassenhove 2009, p. 11). The term effectiveness describes the
fact that logistics is the main driver for delivering the right
relief items and services at the right time, amount and quality
meeting the beneficiary’s needs, whereas efficiency reflects
that logistics can be considered as the biggest cost factor in
disaster relief (Schulz 2008, pp. 3-4; Tomasini and Van
Wassenhove 2009, pp. 11-12).
So far, the biggest focus of humanitarian logistics research
lies on the first response phase of relief operations in
combination with the application of modeling and simulation
(Kunz and Reiner 2012). In the last years also computer-based
simulation has gained some attention in the humanitarian
sector because of its ability to support the planning of disaster
relief operations and its track record in industrial applications
(e. g. Benini 2012). For using simulation in a proper manner,
simulation environments are helpful in order to enable
modularity, reusability and adaptability. They describe the
software environment and its main elements in order to model,
configure and execute specific simulation models (VDI 2000,
p. 18). Thus, a simulation environment can be understood as a
generic model able to be used to generate organization-,
scenario- or case-specific simulation models.
Using simulation in the context of humanitarian logistics
can support the involved decision makers, e.g. program planer,
humanitarian logisticians or crisis management authorities, in
many different ways. From a methodological perspective,
humanitarians can acquire a better foundation for decision

making on strategic and tactical planning horizons by testing
and comparing different logistics network setups and
procedures. While decisions in the practitioners’ daily work
can have serious consequences, similar decisions in simulation
environments can be evaluated with relatively low costs. Such
a learning environment puts humanitarian organizations in a
position to reflect and analyze their positions, to test network
or process changes in a safe, i.e. artificial, environment and to
increase their degree of preparedness and responsiveness
during unexpected events. A concrete application could be the
simulation of distinct transportation modes in order to
understand, analyze and improve the resource allocation as
well as the delivery time of goods. Furthermore, the use of
simulation provides organizations the possibility of studying
new network structures, processes or products before
committing resources for their realization. Former research
indicates that most of the humanitarian actors still make
essential decisions concerning supply chain design and
planning solely based on their experience and intuition, even
though these mental models cannot fully capture the
increasingly complex and dynamic aspects of these tasks
inducing suboptimal performances within the supply chains
(Besiou et al. 2011, p. 81). Especially, in complex dynamic
situations, these models might be error-prone and can cause
shortcomings concerning the documentation and transparency
of decisions (Sterman 2000, p. 15). This of course has to be
seen in the light of the hardly measureable benefits of
practitioner’s skills and “invisible work” based on their
experience and intuition (Star 1991; Mays et al. 2014).
Henceforth, simulation environments offer a complementary
and compatible supplement of existing practices to improve
current decision making processes of the different actors.
Concerning the current use of simulations in the
humanitarian context, it can be assessed that most of the
existing approaches concentrate on the application of a singular
simulation modelling method. These methods provide
techniques for transferring a real world system to a simulation
model (Borshchev 2013, p. 37). Due to the diverse
requirements at the build-up of a humanitarian simulation
environment, it is hardly possible to cover all relevant aspects
with one simulation method without making significant
compromises. The rising capacities of computational times
enable a significant increase of the complexity and dynamic of
simulation models. Hence, in context of a broader research
agenda in the area of improving the ability to manage
humanitarian supply chains, we decided to design a multi-
method environment in order to construct comprehensive
models of the associated humanitarian supply chains which can
fulfill the distinct requirements. In particular, the presented
prototype of the simulation environment primarily contains
facets from the Agent Based and Discrete Event simulation
method, but further extensions by e.g. System Dynamics
elements have already been foreseen. A significant part of this
research, esp. the involvement of practitioners in the design
and evaluation of the simulation environment, was executed
within the EU FP7 demonstration project “Driving Innovation
in Crisis Management for European Resilience” (DRIVER,
Grant Agreement No 607798).
II. RELATED WORKS
The relevance of humanitarian logistics for the research
community has risen significantly over the previous decade.
Particularly, during the span from 2008 to 2011, the number of
journal publications per year increased from around 10 to 40
(Kunz and Reiner 2012, p. 129; Leiras et al. 2014, p. 105).
Furthermore, former literature reviews emphasize the pivotal
role of simulation for humanitarian logistics (cf. Celik et al.
2012; Kunz and Reiner 2012; Leiras et al. 2014). Modelling
and simulation is the most used research methodology in the
analyzed journal publications by Kunz and Reiner ranging
from 1993 to 2011 with 46 % (Kunz and Reiner 2012, p. 129).
The papers mainly deal with topics related to infrastructure
decisions, as, for example routing problems (cf. Özdamar
2011; Trautsamwieser et al. 2011), scheduling problems (cf.
Hu 2011) or facility location problems (cf. Mete and Zabinsky
2010).
These results have been justified by the findings of our
systematic literature review. The following four research
databases were used to find relevant scientific sources:
EBSCOhost, Google Scholar, JSTOR and ScienceDirect. The
literature search was limited to peer reviewed journal
publications within the period 2004 and 2015. For the initial
literature search the terms “humanitarian logistics” or “disaster
relief” were included in each the combination (“and”) of
“simulation”, “supply chain management”, “simulation
environment”, “agent based” discrete event” and “system
dynamic”. The search terms were defined reflecting interviews
with experts covering both the humanitarian practitioner
domain and the research community on simulation. According
to the Cooper Taxonomy (cf. Cooper 1988) the literature
search investigated the coverage of those search terms and was
concentrated on the central and pivotal category. Following
these search criteria we have identified 28 relevant
publications.
The literature search results were then analyzed and
categorized according the (i) the applied methodologies and (ii)
the application areas. The applied methodologies contained:
optimization models, OR/MS modeling, simulation, agent-
based modelling, discrete-event modelling, system dynamics,
multi-method modelling. The application areas covered:
analysis of beneficiaries, assessment and analysis of disaster
spread, cascading effects, emergency/disaster relief logistics
(different aspects concerning response & planning), food
management and distribution, inventory planning, medical
supply location and distribution, needs assessment,
prepositioning of facilities and supplies, resource allocation,
risk management in disaster relief; routing problem and vehicle
fleet allocation/management; scheduling problem. It should be
mentioned that these categories are all a result of a simulation-
driven perspective. This means that a search backwards must
lead to different results, e.g. publications on “optimization
model” and “routing problem”.
We detected in the 28 analyzed sources a nearly equal
distribution between the topics optimization models,
Operational Research/Management Science (OR/MS)
modelling and simulation, whereas decision support systems
(DSS) have only been addressed in three articles.

Simulation modelling frameworks have been used
concretely only in a few cases as the categorization of the
sources depicts. Besiou et al. illustrated the appropriateness of
a SD methodology as a tool for humanitarian decision makers
to understand the effect of their decisions on humanitarian
operations (cf. Besiou et al. 2011). By developing distinct
cases, they showed that the SD paradigm has the capacity to
represent the dynamic complexity of humanitarian operations.
Peng and Chen developed a SD model to describe the
processes of delivering emergency supply (cf. Peng and Chen
2014). They highlighted that simulation results can have a
significant impact on the supply chain risk management of
humanitarian organizations. Das and Hanaoka introduced an
ABM framework for integrating stakeholders’ interests (cf.
Das and Hanaoka 2014). They conclude that ABM is a good
tool for analyzing the effects of resource allocation. It allows
actors to investigate the effects of transport measures and to
understand the mechanisms of demand management in a
dynamic environment (Das and Hanaoka 2014, p. 282). For a
multi-method model only one source was found. Xanthopoulos
and Koulouriotis used discrete-event and agent-based facets to
investigate the vehicle routing problem during relief
distribution operations in a post-disaster environment (cf.
Xanthopoulos and Koulouriotis 2013). The usage of this
simulation tool showed that significant insights can be obtained
concerning the routing selection (Xanthopoulos and
Koulouriotis 2013, p. 181).
Regarding the application categories it can be stated that
the following areas are mainly covered in the identified works:
Routing problem and vehicle fleet allocation/management (9),
emergency/disaster relief logistics (8), and prepositioning of
facilities and supplies (5). As the conceptualization of a
simulation environment considering all possible application
areas would become too complex and would be inefficient
from a user perspective, we first analyzed the application areas
concerning their relevance, existing solutions, and secondly we
prioritized them based on the results of our expert interviews.
In result, the focus of the conceptualized simulation
environment was laid on the areas: collaboration, route
planning, prepositioning of facilities as well as supplies and
bottleneck analysis because they cover important business
needs within humanitarian supply chains and offer immense
potential for improvements in conjunction with simulation
which will be illustrated in the following.
A. Collaboration
The focus of collaboration was selected because it is
inherent to all humanitarian operations. This is proven by
research and was also highlighted during the interviews. One
interviewee indicated that especially the merging process of
local branches in Germany is inevitable for the success of the
operations. This is caused by the fact that local branches
command over different machines, vehicles and specialization
groups. Therefore, it is necessary to combine the needed means
before starting a coordinated help in the affected regions.
Another interviewee highlighted the relevance of collaboration
for the international context. He particularly pointed out the
platform Reliefweb and the program Logistics Cluster which
represent fundamental aspects for information sharing and
collaboration between the organizations (OCHA 2015).
Reliefweb is a specialized digital service of the United Nations
Office for the Coordination of Humanitarian Affairs (OCHA)
which provides humanitarian information on global crises and
disasters. Taken the crisis in Nepal as an example, 28 different
organizations reported to the progress report released by the
OCHA. The Logistics Cluster program concentrates on the
coordination and information management between
humanitarian organizations to ensure an effective and efficient
logistics response. It is hosted by the World Food Programme
(WFP) and is activated when there are response and
coordination gaps in addressing humanitarian needs (WFP
2016). In general, a cluster is a group of organizations working
together on specific topics to improve humanitarian response
(WFP 2016). Exemplary tasks include the filling of gaps in
logistics capacities or the meeting of needs for logistics
coordination services. Henceforth, it is important to analyze
different collaboration mechanism during the transportation of
goods from the supplier to the beneficiary considering effective
and efficient aspects. Although, collaboration is a vital problem
of humanitarians, it has not been investigated extensively yet
by using operations research technologies. This was also an
important criterion for the selection of collaboration as one of
the thematic focuses of the simulation environment.
B. Route planning
Route planning has become a complex task for
humanitarian organizations due to the variety of possibilities
and influencing factors. Especially, in this context simulation
environment can support the decision making process by
offering data concerning diverse performance measures. Two
interviewees acknowledged that the route planning is a
complicated task which influences the whole traffic in the
affected areas. While exemplary decisions include in which
form the aid workers travel (e.g. singular units or convoys),
influencing factors for the route planning are rest periods for
the auxiliaries or the consideration of gas stations. Therefore, it
is important that simulation environments model these multi-
faceted constraints to provide the humanitarian organizations
the opportunity to test different route combinations before the
actual release process takes place. Although various
approaches already exist, it has to be acknowledged that these
solutions mostly do not consider the whole supply chain as
well as different means of transport. For instance, Özdamar
concentrated only on the use of helicopters (cf. Özdamar
2011). Due to its relevant business needs and possible
advancements considering existing approaches, it was selected
as application area for the simulation environment.
C. Prepositioning of facilities and supplies
Since humanitarian organizations have limited financial
capacity, an efficient and effective prepositioning of facilities
and supplies is a fundamental task. While too low inventory
levels could evoke bottlenecks within a supply chain and
ultimately lead to the situation of not fulfilling the demand of
the beneficiaries, too high inventory levels are not ideally
concerning opportunity costs (e.g. increased costs for
warehousing) (Pettit et al. 2015, p. 2). Additionally, the
prepositioning of facilities is a highly complex process. Due to
the fact that in most case, multi-tier supply chains are relevant,

the amount of decision possibilities, influencing variables and
dependence rises significantly. For instance, the THW has to
define meeting points where the merging of local branches
takes place which can hardly be done by considering only a
map and their respective expert knowledge. Henceforward, the
simulation environment should provide the possibility to
position the facilities and define the inventories which can then
later be assessed in form of a simulation study. Similar to the
field of routing problems, different optimization models and
simulation approaches have been developed for this application
context in the previous years. However, the consideration of
multi-tier supply chains with their respective components
including multiple products would also extend most of the
existing approaches and provide humanitarians a better
foundation for decision making.
D. Bottleneck analysis
Finally, the field of bottleneck analysis is chosen because it
has a significant effect on the overall performance of the
supply chain. Bottlenecks can result into extensive delays or
break-downs of the associated processes (Van Wassenhove and
Pedraza Martinez 2012, p. 310). For instance, Van
Wassenhove and Pedraza Martinez outlined bottlenecks in the
relief operations after the major earthquake in Haiti.
Infrastructural limitations, such as airports or roads, lead to
difficult problems during the distribution of goods (Van
Wassenhove and Pedraza Martinez 2012, p. 310). Bottlenecks
are also pretty common in humanitarian logistics due to
concise planning of organizations and a lack of support
mechanisms. Looking into the neighboring domain of
commercial logistics, it can be stated that computer-based
simulation has successfully been used to identify and fix
bottlenecks over the years. Secondly, the literature review
showed also that almost no specific solution approach is
available for the detection of bottlenecks at distinct levels of
the supply chain. Thus, it is important that the simulation
environment encompasses the relevant methods and
performance measures in order to support the bottleneck
analysis.
Concluding, it can be stated that the simulation
environment covers four important key tasks of humanitarian
logistics. This design offers on the side new analysis tool for
vital problem fields and on the other side improvements of
enhancement for existing approaches. All of the four aspects
encompass significant business needs.
III. METHODOLOGY AND THE CONCEPTUALIZATION OF THE
SIMULATION ENVIRONMENT
A. Methodology
Our research design consisted of several steps and
orientated at the design science paradigm of the information
systems research framework. The procedure of this paradigm
can be summarized as follows: Based on a clear problem
formulation, insights from the environment (relevance cycle)
and knowledge (rigor cycle) base are utilized for developing a
new artefact. At this point, it is important to identify relevant
business needs and the applicable knowledge from the
knowledge base. Therefore, we conducted a literature review
and expert interviews. After the product is completed, it is
assessed for justification and evaluation purposes. In order to
ensure the relevance of the solution approach, we based the
logic and content of the simulation model on the Reference
Task Model (RTM) for Humanitarian Supply Chains (cf.
Blecken 2010). The main advantage of the application of a
reference model is seen in the modularity, reusability and
adaptability, which is a major requirement to design a generic
humanitarian logistics environment instead of one, probably
more precise, model (Widera et al. 2013). Up to now, the RTM
was applied and evaluated within four relief organizations,
including UNICEF, German Red Cross, the NGO humedica
and the Germany’s Federal Agency for Technical Relief
(THW). By these means, an adequate consideration of real
world logistics processes within the targeted simulation model
can be ensured. However, in order to validate developed the
simulation environment, several semi-structured interviews
with logistics practitioners from humanitarian organizations as
well as simulation experts have been executed. From the
humanitarian domain three experienced practitioners have been
invited to review the simulation environment. Two of them
were representing the THW. The third expert was chosen
because of a long-standing experience in humanitarian logistics
with NGO work (World Vision International) and UN agencies
(UNICEF and UNOPS). From the simulation domain, two
experienced researcher carrying out similar approaches (multi-
method simulation) for commercial supply chain management
research have been involved during the design phase and the
evaluation of the final artifact.
One last additional note from the methodological
perspective has to be made regarding the scenario, both in
terms of disaster scenario and the involved relief
organization(s). The very first evaluation of the simulation
environment was illustrating a hypothetical relief agency
having a rather small network. The generic simulation
environment was exemplary applied for the case of a Seaquake
in the Mediterranean Sea causing a Tsunami for the South of
France. Here, we modeled the structure and processes of the
hypothetical relief chain and ran a first simulation study. This
study was presented to the DRIVER consortium, including the
end users, in order to gather feedback on the functionalities and
results of the simulation environment. After the first round of
experiments in November 2014 and based on the current
seismological risk analysis a specific disaster scenario –a
Tsunami was defined the Mediterranean Sea, Italy – was
defined for the next Joint Experiment. This scenario enabled a
second summative evaluation of the artefact and is presented in
this paper. We have increased the complexity of the model, but
the provided network data (like average transportation costs) is
still fictional. In the meantime further iterations were executed,
esp. a flooding scenario with the THW based on real event logs
(see Detzer et al. 2016, Widera et al. 2017). However, in this
paper we decided to present the hypothetical case in order to
describe the existing functionalities.
B. Requirements Catalogue
After selecting the four application areas, it is necessary to
outline a requirement catalogue so that an adequate simulation
environment can be build which is displayed in Tab. 2. This

identification of requirements followed an argumentative-
deductive research approach which is primarily based on the
characteristics of humanitarian logistics and the chosen
application areas for this simulation environment (Wilde and
Hess 2007, p. 282).
Category Requirement
Actors
The modelling of different humanitarian
organizations with their respective supply
chain structures should be enabled.
Beneficiary
It should be possible to model the diverse
needs of a beneficiary and its dynamic
behavior (demand for a certain product
portfolio, changing health conditions,
entering and leaving of the affected areas).
Collaboration
The simulation environment should allow
for the modelling of various collaboration
mechanisms between the organizations.
Disaster &
Environment
Dynamic and uncertain aspects need to be
considered within the simulation
environment as well as interactive
possibilities for the user to regard these
effects .Certainly, it is also important that
multiple disaster occurrences with different
phases can be modelled.
Mode of
Transportation
Different transportation modes should be
available to the user.
Performance
Measures &
Costs
Charts as well as KPI are needed in order
to evaluate the performance. Additionally,
costs are necessary for a multi-faceted
evaluation of the performance.
Processes
A consideration and the integration of user-
specific processes with the certain
parameters should be realizable.
Products
Distinct products with their respective
processes (e.g. cooling process) should be
modelled in the simulation environment.
Route
Planning
The route planning should take into
account real road networks and available
mode of transports to select a suitable route
and transport mode combination.
Simulation
Environment
Design
A flexible and adaptable simulation
environment which can be used to model
diverse scenarios and configuration
settings.
Supply Chain
Structure
Diverse elements need to be available in
order to model a multi-tier supply chain
structure.
Supplier Consideration of suppliers with specific
procurement processes.
C. Description of the Simulation Environment
AnyLogic was selected as simulation software tool for the
design of this simulation environment because it supports a
multi method-modelling approach by offering the basic
functionalities of the three main outlined simulation modelling
paradigms. Furthermore, the modular build-up of the
simulation environments in form of object classes enables
extension possibilities in the future and the adaptability of the
environment in general.
Regarding the technical point of view, the simulation
environment is constructed primarily by ABM and DES facets.
The distinct components (e.g. suppliers or logistics centers) are
represented as agents and DES elements are utilized for the
disaster modelling and the multiple event constraints within the
procedures of a humanitarian supply chain. The dynamic
behavior and the processes are formed by state charts. While,
the statuses of state charts are alternative which means that the
object can only be in one state at a time, transitions between
the states may be triggered by user-defined conditions
(Grigoryev 2012, p. 76). Since state charts are simple to
modify and can be comprehended well, this methodology was
chosen instead of SD components. Nevertheless, this does not
neglect the implementation of SD components in the future.
The developed simulation environment consists of six
fundamental components: stationary agents, moveable agents,
disaster components, collaboration, communication and finally,
performance measures which will be explained in the
following.
Stationary agents can be used to model a multi-tier supply
chain structure and the beneficiaries of the humanitarian
logistics supply chain. AnyLogic’s connectivity feature to
Microsoft Excel is utilized for loading the relevant parameters
and characteristics of the agents. Excel sheets allow for a user-
friendly data input. Additionally, it is possible to expand and
modify the Excel sheets easily. The AnyLogic GIS API is
another important instrument. During the simulation, positions,
routes and distances are requested automatically from the
OpenStreetMap-Server in order to regard real-time data. An
overview of the stationary agents and their interactions is
pictured in Fig. 3.
Fig. 3: Stationary agents of the simulation environment
The element supplier represents external suppliers which
provide the requested goods to the logistics center which
correlates to a headquarters of a humanitarian organization in
reality. Cross docking points form transshipment points within
the supply chain, for example airports. Storage center embody
warehouses, whereas field distribution center (FDC) are
created in the crisis region to hand over the goods to the
beneficiaries. Beneficiaries occur in response to a disaster and
are persons demanding certain goods or services. Suppliers
provide the ordered goods to the humanitarian organizations.
Logistics center illustrate the core of the logistics activities in
this simulation environment. Their main tasks are the ordering
of goods from the suppliers and distributing these goods on the
best and fastest way to the cross docking points and storage
centers. Here it has to be mentioned that a linear flow of goods

is modelled in the simulation environment. Back orders are not
considered explicitly and lower tiers cannot deliver goods to
upper tiers. The material and information flow within in the
supply chain is illustrated in Fig. 4.
Fig. 4: Flows within the supply chain
Analyzing the material flow in more detail, it can be stated
that cross docking points are currently the only elements which
can deliver goods between themselves (e.g. from an airport to a
port and then to a storage center). Furthermore, the logistics
center decides whether a cross docking point is utilized for the
transport or a direct transport to a storage area is chosen.
Certainly, an enhancement of the flexibility of the underlying
supply chain structure is an interesting aspect for future work.
The transportation and route business logic examines all
possible routes (direct and indirect) concerning their distances
and approximate delivery time. Furthermore, costs, weights,
loading and discharging times are considered in this
calculation. Afterwards, the best fitting option is chosen. In
order to get an initial understanding of the procedure and
complexity of this logic, Fig. 5 illustrates the input and output
data of the transportation mechanism.
Fig. 5: Overview of the transportation business logic
Moveable agents are essential for moving the goods within
the supply chain. For this purpose, the simulation environment
contains the five most common vehicles in humanitarian
logistics: airplanes, jeeps, ships, trains and trucks.
The disaster and environment modelling is implemented in
a highly flexible way. Humanitarian organizations may use this
simulation environment as a cockpit in order to simulate
multiple scenarios and supply chain configurations easily.
Excel sheets can be used to quantify the damage of the
disasters by inserting the location and amount of beneficiaries.
GIS regions can be utilized to shape the impact of the disaster
on the GIS map. Here, it is assumed that the transportation into
these areas is difficult directly after a disaster has occurred due
to for example impassable roads. This is regarded in the
simulation by a deceleration of the delivery on the last mile.
The user can deactivate this effect interactively during the
simulation runs. Similar, the disaster activation takes place.
Buttons are available for activating multiple disasters.
Additionally, new cross docking points, storage centers or FDC
can be created during a simulation run. Logistics centers are
neglected because the setup of this agent type is a long term
process with significant impact on the supply chain structure
which would not fit into the underlying time horizon and
scope. Furthermore, it is possible to activate and deactivate
certain agents in course of a simulation run. This applies to
cross docking points and FDC. For instance, it might be
possible that an airport is shortly not available or an FDC needs
to be shut down due to varying reasons (e.g., infrastructural
problems). This functionality was not implemented for the
other stationary agents because firstly they are in most of the
cases not located in the destinations where the disaster
occurred and secondly significant events (e.g. an extensive fire
in a storage center) would need to transpire in order to cause a
breakdown over a longer period of time which happens only
very rarely. An overview of the cockpit’s features is provided
in Fig. 6.
Fig. 6: Cockpit features of the simulation environment
Fig. 6 shows on the left side the different buttons, which
can be activated during the simulation as well as a chart for the
total number of victims. By double-clicking on an agent,
specified windows open concerning the selected agent (here:
storage center Castrovillari). Time stack charts encompass
information concerning the inventory levels and the backlogs.
The GIS map, located in the center of the cockpit, depicts the
flow of goods between the distinct agents and on the right side
modifications can be executed concerning the inventory levels.
The implemented collaboration mechanism concentrates on
the ordering processes between logistics centers, cross docking
points and storage centers (see Fig. 7).
Fig. 7: Collaboration mechanism between two
organizations
The collaboration mechanism works as follows: Besides an
individual backlog for an organization, a collaborative backlog
for the affected organizations is created at the beginning of the

simulation run (cf. step 1, Fig. 7). It contains all the different
orders for the associated elements (e.g. logistics centers).
Collaborating organizations can work on an aggregate backlog
to ensure a more efficient and effective delivery of the goods.
A reservation system with messages is implemented to ensure
that the organizations are not working on the same order at the
same time. After an organization has selected an order, a
reservation message is directly send to the other organizations
to inform them that the order is in progress (cf. step 2, Fig. 7).
When the delivery process is finished, the order is deleted from
the collaborative backlog and the reservation is deleted from
the reservation system.
Different performance measures have been implemented in
order to assess the performance of the supply chain
configuration during and after the simulation run. Therefore,
on the one side performance measures are used to allow for an
analysis of the performance during the simulation which is
done in form of multiple time stack charts. Additionally, output
data is constantly inserted in an Excel sheet in order to enable a
multi-dimensional analysis after the simulation run. These
results can ideally be used to compare multiple supply chain
configurations. After the core functionalities of the simulation
environment are presented, it is necessary to check whether the
requirement catalogue was implemented and to which extent.
D. Exemplary Simulation Study Results
To demonstrate an exemplary usage of the developed
simulation environment and the connected benefits for
humanitarian organizations, a simulation study was conducted
which will be briefly outlined. This step also facilitates the
validation of the simulation environment’s design. For
conducting the simulation study, different sources have been
regarded (Banks 2013; Kelton 2010; Law 2007; VDI 2000). As
illustrated previously, the goal is to analyze the performance
and costs of different collaboration modes between
humanitarian organizations during disaster relief operations (no
interacting or horizontal collaboration). As described above, an
elaborated performance measurement system was integrated in
the simulation environment but due to the focus in a
presentation of the designed artifact, we limit the evaluation
results on two performance indicators being the amound of
saved lives and total costs.
The following artificial scenario is considered in the
simulation study. Three independent humanitarian
organizations A, B and C are regarded with supply chains
distributed in Europe. Scenario 1 assesses the case that no
interactions occur between all three organizations and scenario
2 considers a horizontal collaboration between the
organizations. Their help is needed due to a tsunami on the
southern west coast of Sicily, in particular Palermo, Messina
and Pollina. A tsunami was selected because it may be
expected that a tsunami would in most of the cases lead to
destroyed inventories as well as trading centers resulting in the
immediate need of externals. Especially, the damage around
Palermo is significant causing problems during the distribution
of goods. This is highlighted by the GIS area on the map.
The simulation study provided two explicit results. First, it
can be concluded that collaboration can have an impact on the
efficiency of humanitarian operations.
The communication mechanism between the three
organizations led to a substantial cost reduction for the total
operation as well for the organization A and C. Furthermore, it
was possible to save almost 4000 people more than in
scenario 1. Through a more efficient delivery and steady
communication within the supply chain, it was feasible to
realize more constant inventory levels over the duration of the
simulation study. This was observable by the corresponding
inventory levels and lower order fulfillment times for the
distinct logistics centers. Consequently, it was feasible to
realize better results in the FDC concerning the ratio of saved
people. Tab. 10 summarize the key results.
IV. CONCLUSION AND OUTLOOK
From a practical point of view, the constructed simulation
environment encompasses a multitude of analysis techniques
and solutions for diverse business needs in the humanitarian
sector. Features like the implemented cockpit view with a
multitude of modification possibilities, provide a generic
answer for handling the uniqueness of disasters and the
dynamic environment of humanitarian logistics during a
simulation run. By assessing multiple disaster and
configuration runs in course of a simulation, it is not necessary
to construct simulation models and conduct related studies for
each disaster or configuration setting singularly. This feature
makes the utilization of this environment not only interesting
for humanitarian organizations but also for cities or
international institutions concerning their crisis management.
They could use this simulation environment to conduct
resilience and responsiveness analysis.
However, Firstly, restrictions exist concerning the usage of
the simulation environment. In particular, there are certain
aspects and problems in the selected four application areas:
bottleneck analysis, collaboration, route planning and
prepositioning of facilities and supplies which cannot be
modelled or solved by using the simulation environment and
simulation in general. Besides, due to the focus on the design
of the environment only an excerpt of the evaluation results
Category\
Scenario
Scenario 1 –
No collab.
Scenario 2 –
Collab.
Results
Victims
(No.
Killed)
13,200 9,300 3900 more
people have
been saved
than in
scenario 1
Ratio of
saved
people
46,34 % 62,2 % Almost 20 %
improvemen
t
Total
Costs (in
€)
3.51E+11 2.24E+11 36 % of the
total costs
have been
saved

were presented and discussed here, which might lead to a
biased and overdrawn picture.
As outlined previously, simulation results cannot produce a
perfect replication of the reality. These models can only
capture the complexity of the real system to a certain degree. It
is even feasible that multiple simulation runs of the same
model result into distinct results denying clear statements
concerning the behavior of the system. Hence, decisions should
not be based solely on the results of a simulation study.
Especially, in these complex simulation environments, which
are necessary for modelling humanitarian operations in a
certain level of detail, it is sometimes hardly possible to
identify logical errors since the simulation runs properly
considering verification aspects.
The evaluation of the simulation environment revealed
several limitations which need to be addressed in the future.
Here, especially a more detailed modelling of suppliers would
enable a more particularized analysis of the interactions
between an external supplier and the humanitarian
organizations. Additional collaboration mechanisms could
provide a more diversified view on the performance of the
collaboration between the organizations. Besides, the
incorporation of system dynamic components in the behavior
of a beneficiary could result into an overall more realistic
consideration of beneficiaries within the simulation
environment. Last but not least, in the future works a more
detailed presentation of the simulation study, covering all
measured Key Performance Indicators, need to be given in
order to share a more comprehensive evaluation picture of the
simulation environment.
REFERENCES
[1] Kunz, N., and Reiner, G. 2012 “A meta-analysis of humanitarian
logistics research,” in: Journal of Humanitarian Logistics and Supply
Chain Management, Vol. 2(2), pp. 116–147.
[2] Thomas, A. S., and Kopczak, L. R. 2005. From Logistics to Supply
Chain Management: The path forward in the humanitarian sector,
http://www.fritzinstitute.org/pdfs/whitepaper/fromlogisticsto.pdf,
downloaded on 2 June 2015.
[3] Blecken, A. 2010. Humanitarian Logistics: Modelling Supply Chain
Processes of Humanitarian Organisations, Bern et al.: Haupt Verlag.
[4] McGuire, G. A. 2006. Development of a Supply Chain Management
Framework for Health Care Goods Provided as Humanitarian assistance
in Complex Political Emergencies, PhD Thesis, Wirtschaftsuniversität
Wien.
[5] Tomasini, R., and Van Wassenhove, L. 2009. Humanitarian Logistics,
Basingstoke: Palgrave MacMillan.
[6] Schulz, S. F. 2008. Disaster Relief Logistics: Benefits of and
Impediments to Horizontal Cooperation between Humanitarian
Organizations, dissertation at the Technische Universität Berlin.
[7] Benini, A. 2012. “A computer simulation of needs assessments in
disasters. The impact of sample size, logistical difficulty and
measurement error,” Washington DC, USA.
[8] VDI. 2000. VDI 3633 - Simulation von Logistik-, Materialfluss- und
Produktionssystemen Grundlagen, VDI-Guidelines, Verein Deutscher
Ingenieure, Düsseldorf.
[9] Besiou, M., Stapleton, O., and Van Wassenhove, L. N. 2011. “System
dynamics for humanitarian operations,” in: Journal of Humanitarian
Logistics and Supply Chain Management, Vol. 1(1), pp. 78-103.
[10] Star, S. L. (1991) Invisible work and silenced dialogues in knowledge
representation. In: Women, Work and Computerization. I. Eriksson, B.
Kitchenham, and K. Tijdens, Eds. North Holland, Amsterdam, 1991, 81-
92.
[11] Mays, R..; Walton, R.; Lemos, M.; Haselkorn, M. 2014. “Valuing what
works: success factors in disaster preparedness”, The Global Disaster
Preparedness Center, American Red Cross.
[12] Borshchev, A. 2013. The Big Book of Simulation Modeling,
Multimethod Modeling with Anylogic 6, Chicago: AnyLogic North
America.
[13] Leiras, A., de Brito Jr, I., Peres, E. Q., Bertazzo, T. R., and Yoshizaki,
H. T. Y. 2014. “Literature review of humanitarian logistics research:
trends and challenges,” in: Journal of Humanitarian Logistics and
Supply Chain Management, Vol. 4(1), pp. 95–130.
[14] Özdamar, L. 2011. “Planning helicopter logistics in disaster relief,” in:
OR Spectrum, Vol. 33(3), pp. 655-672.
[15] Trautsamwieser, A., Gronalt, M., and Hirsch, P. 2011. “Securing home
health care in times of natural disasters,” in: OR Spectrum, Vol. 33(3),
pp. 787-813.
[16] Hu, Z.-H. 2011. “A container multimodal transportation scheduling
approach based on immune affinity model for emergency relief,” in:
Expert Systems with Applications, Vol. 38(3), pp. 2632-2639.
[17] Cooper, H. M. 1988. “Organizing knowledge syntheses: A taxonomy of
literature reviews,” in: Knowledge in Society, Vol. 1, pp. 104-126.
[18] Das, R., and Hanaoka, R. 2014. “An agent-based model for resource
allocation during relief distribution,” in: Journal of Humanitarian
Logistics and Supply Chain Management, Vol. 4(2), pp. 265–285.
[19] Xanthopoulos, A. S., and Koulouriotis, D. E. 2013. “A Multi-agent
Based Framework for Vehicle Routing in Relief Delivery Systems,” in:
Zeimpekis, V., Ichoua, S., and Minis, I. (eds.), Humanitarian and relief
logistics - Research issue, case studies and future trends, New York:
Springer-Verlag, pp. 167-182.
[20] OCHA. 2015. Logistics Cluster, http://www.logcluster.org/logistics-
cluster, accessed on 25 August 2015.
[21] WFP. 2016. Logistics Cluster, https://www.wfp.org/logistics/cluster,
accessed on 13 February 2016.
[22] Pettit, S., Roh, S., Harris, I., and Beresford, A. 2015. “The pre-
positioning of warehouses at regional and local levels for a humanitarian
relief organisation,” in: International Journal of Production Economics,
http://www.sciencedirect.com/science/article/pii/S0925527315000274,
downloaded on 17 August 2015.
[23] Van Wassenhove, L. N., and Pedraza Martinez, A. J. 2012. “Using OR
to adapt supply chain management best practices to humanitarian
logistics,” in: International Transactions in Operational Research, Vol.
19, pp. 307–322.
[24] Widera, A., Dietrich, H.-A., Hellingrath, B., & Becker, J. (2013).
Understanding Humanitarian Supply Chains — Developing an
Integrated Process Analysis Toolkit. In Proceedings of the International
Conference on Information Systems for Crisis Response and
Management (ISCRAM) 2013, Baden-Baden, Germany, 210–219.
[25] Detzer, S., Gruczik, G., Widera, A., Nitschke, A. (2016). Assessment of
Logistics and Traffic Management Tool Suites for Crisis Management.
In Proceedings of the European Transport Conference, Barcelona.
[26] Widera, A.; Lechtenberg, S., Gurczik, G.; Bähr, S.; Hellingrath, B.
(2017) Integrated Logistics and Transport Planning in Disaster Relief
Operations. Proceedings of the International Conference on Information
Systems for Crisis Response and Management (ISCRAM) 2017, Albi,
France, 752–764.
[27] Wilde, T., and Hess, T. 2007. “Forschungsmethoden der
Wirtschaftsinformatik: Eine empirische Untersuchung,” in:
Wirtschaftsinformatik, Vol. 49(4), pp. 280–287.
[28] Grigoryev, I. 2012. AnyLogic 6 in Three Days, A quick course in
simulation modeling, Chicago: Anylogic North America.
[29] M. Young, The Technical Writer’s Handbook. Mill Valley, CA:
University Science, 1989.