Professional Documents
Culture Documents
Management of Marine Logistics in The Case of Emergency or Disaster
Management of Marine Logistics in The Case of Emergency or Disaster
3, 2016 251
Biographical notes: Heng Wang received his BSc in Systems Innovation from
The University of Tokyo in 2015. He is currently an MSc student at The
University of Tokyo. His research interests include marine logistics.
Kenji Tanaka received his BSc in Marine Engineering from The University of
Tokyo in 1998, and PhD in Systems Innovation from The University of Tokyo
in 2009. He is currently an Associate Professor at The University of Tokyo. His
research interests include logistics and energy management.
1 Introduction
Disasters have caused serious damage to society many times in the past. For example,
300,000 people died from the Sumatra earthquake in 2004. 200,000 people died from the
Haiti earthquake in 2010. Today, even though we have taken emergency measures
against disasters, they still continue to cause serious damage. Therefore, it is important to
improve measures of recovery from disasters.
One of the problems after an earthquake is provision of goods for evacuees. When the
Kobe earthquake occurred in 1995, evacuees were short of water and portable toilets. On
the other hand, when Niigata earthquake occurred in 2007, excessive clothing was send
to the damaged area. Then, warehouses around damaged area were full of clothing, and
other goods could not be put in warehouses. Judging from these facts, it is important to
provide proper kind and amount of goods for evacuees.
In past disasters, most of the goods were transported by land logistics. On the other
hand, marine logistics have some advantages in emergency transportation: It is believed
that marine logistics suffer less damage than land logistics when a disaster occurs. If it
occurs close to coast line, we can provide goods for evacuees more efficiently and
effectively by using marine logistics.
In this paper, we propose a marine logistics system in case of emergency disaster,
including an estimation of the demand on goods for the damaged area and an
optimisation of the assignment of ships to transportation routes. Thanks to this system,
we can estimate how much goods and how many ships should be prepared for future
disasters.
2 Literature review
Holguín-Veras et al. (2014) summarised interviews and field work research after the
Tohoku disasters in Japan. Kaynak and Tuğer (2014) focused on discovering the clusters
for collaboration within disaster coordination operations. Bastos et al. (2014) identified a
set of logistics processes and actions which are related to the success of disaster response
operations. Kuse and Yano (2011) proposed logistics method in detail. However, these
studies only proposed qualitative methods. Though qualitative methods are useful to
determine the policy of plan for disaster relief, we cannot determine the detail of plan for
disaster relief, such as how many ships are required. Therefore, we should also consider
quantitative method to determine detail of plan for disaster relief.
There are other studies about resilience of logistics. Fizzanty et al. (2013) developed a
framework for evaluating and improving failed supply chains. Carvalho et al. (2012)
established links between supply chain disturbances and corresponding strategies of
resilience to counter them. Vlachos et al. (2012) developed robust supply chains by
reducing vulnerability and improving resilience. However, these studies did not mention
the disaster.
Some studies proposed optimisation of logistics in case of disaster by quantitative
method. Namimatsu et al. (1997) optimised logistics by vehicle routing problem. Senda
and Suzuki (2013) simulated logistics in case of disaster in Nagoya, Japan. Lin et al.
(2011) proposed a logistics model for delivery of prioritised items in disaster relief
operations. However, these studies only proposed land logistics in case of disaster. We
can provide goods for evacuees more efficiently and effectively by using marine logistics
than using land logistics, because marine logistics suffer less damage than land logistics
when a disaster occurs. Therefore, we should also consider marine logistics when a
disaster occurs.
Management of marine logistics in the case of emergency or disaster 253
Only few studies about marine logistics analyse the case of disasters by quantitative
methods. Majima et al. (2013) developed a marine transportation simulator for
emergency disaster. However, this system did not include an optimisation algorithm.
Therefore, this paper develops marine logistics system for emergency disaster control,
including an optimisation algorithm.
3 Model
First, we select the available origin ports, destination ports and ships for marine logistics.
Origin port is the port in a non-damaged area, and destination port is the port in the
damaged area. Then, we calculate the demand of goods in the damaged area based on its
population and the amount and type of goods required per person.
Next, we determine the assignment of ships to transportation routes based on these
conditions by assuming that ships transport goods by a shuttle service between origin port
and destination port. We optimise marine logistics by determination of the assignment of
ships to transportation routes. This is done using linear programming. From this
optimisation, required conditions to supply demand on goods of evacuees can be
calculated. For example, how many ships are required, how many quays are required for
each port, or how to assign ships to transportation routes. Then, we carry out marine
logistics simulation based on calculated conditions.
254 H. Wang and K. Tanaka
With the results of the marine logistics simulation, we can evaluate whether we can
supply demand on goods of evacuees under these conditions. We determine the required
resources such as how many ships are required and how to assign ships to transportation
routes.
Kind of goods Required amount (per person*day) Target age Type of goods
A MA1 Division 1 α
B MB2, MB3 Division 2 α
Division 3
C MC2 Division 2 β
… … … …
‘Kind of goods’ means the kind of goods which may be required for evacuees in case of
disaster, such as water and food. ‘Required amount’ means the amount of goods per
person and per day. We define Mij as the weight of goods i required for age division j per
person and per day. ‘Target of age division’ describes the age of people of who require
the kind of goods. Finally, we divide goods into two types according to their occurrence.
Demand of type α goods occur regularly, such as water and food. On the other hand,
demand of type β goods only occur after disasters, such as blankets and toilets.
Management of marine logistics in the case of emergency or disaster 255
∑ ∑ ∑
nship norigin ndestination
f = aijk xijk (1)
i =1 j =1 k =1
Next, we define the constraints. To determine the best assignment, we define three
constraints.
• one ship is only assigned to one transportation route
• αj to βj ships are assigned to transportation routes which are related to origin port j
• γk to δk ships are assigned to transportation routes which are related to destination
port k.
The first constraint means that one ship is never assigned to several transportation routes.
The second constraint means that there is proper number of ships to be assigned to each
origin port. This is because the number of quays is different at every origin port. αj and βj
are determined by the number of quays which can be used for transporting goods in case
of disaster at origin port j. The third constraint means that there is proper number of ships
to be assigned to each destination port. This is because the demand of goods is different
in every destination port. γk and δk are determined by the shortage of goods at destination
port k. Therefore, the best assignment is not the assignment which simply maximises the
transportation amount of goods but the assignment which is matched with the demand of
destination ports. These three constraints are expressed by equations (2) to (4).
∑ ∑
norigin ndestination
xijk ≤ 1 (2)
j =1 k =1
∑ ∑
nskip ndestination
αj ≤ xijk ≤ β j (3)
i =1 k =1
∑ ∑
nskip norigin
γk ≤ xijk ≤ δk (4)
i =1 j =1
256 H. Wang and K. Tanaka
First, the simulator reads the required data. Required data includes port data, ship data,
and demand data which we determined in Sections 3.3 and 3.4. Next, we determine
whether the simulator should use the reassigning algorithm. The reassigning algorithm is
the optimisation algorithm to reduce number of required ships for transporting goods.
The reassigning algorithm is explained in detail in Section 3.5.2.
Then, the simulator determines the first assignment of ships to transportation routes
by means of linear programming. If we determine not to use reassigning algorithm, the
simulator carries out simulations based on this assignment until the end of the simulation
period. If we determine to use reassigning algorithm, the simulator carries out simulations
Management of marine logistics in the case of emergency or disaster 257
based on this assignment until the time when the simulator judges to change the
assignment. Then, the simulator determines another assignment of ships to the
transportation routes based on demand and supply of every destination port at that time.
Finally, the simulator carries out another simulation based on the newer assignment until
it judges to change assignment again.
The simulator repeats this process until the end of the simulation period.
When the simulation ends, we obtain the data of the output result. The result of the
simulation is an assignment of ships to routes, the supply of goods at every destination
port, and the stock of goods at every origin port.
After we obtain the data from the simulator, we change the number of ships and
simulator carries out simulations again. Through this loop, we obtain results of all
combinations of number of ships and reassigning patterns.
⎛
∑ ∑ Lij ⎞
ndestination _ port ngoods
⎜ i =1 j =1 ⎟
F = ⎜1 − ⎟ × 100 (5)
∑ ∑
ndestination _ port ngoods
⎜ Dij ⎟
⎝ i =1 j =1 ⎠
F fill rate
ndestination number of destination ports
ngoods number of kind of goods
Lij shortage of goods i at destination port j
Dij demand of goods i at destination port j.
According to this definition, fill rate equals to 100% if there is no shortage of goods at
destination ports. Figure 3 shows the relationship between the fill rate and the number of
required ships. In this case, the fill rate is equal to 100% if we prepare 16 ships for
transportation.
258 H. Wang and K. Tanaka
Figure 3 Relationship between fill rate and number of ships (see online version for colours)
However, if we only evaluate the fill rate at the end of the simulation period, the output
like Figure 4 is evaluated to be fine result. In this case, serious shortage of goods
occurred in the middle of the simulation period.
Figure 4 Serious shortage of goods occurred during simulation period (see online version
for colours)
Therefore, we calculate the fill rate in the middle of simulation to evaluate the output.
Figure 5 shows the relationship between the fill rate and the number of ships n days after
the disaster. According to this figure, if we prepare 19 ships for transportation, fill rate
equals to 100% seven days after the disaster; however fill rate equals to only 80% three
days after the disaster. We have to prepare 34 ships to achieve a 100% fill rate three days
after the disaster.
Management of marine logistics in the case of emergency or disaster 259
Figure 5 Relationship between fill rate and number of ships n days after the disaster (see online
version for colours)
Figure 6 shows the number of required ships to achieve 100% fill rate n days after the
disaster. This figure visualises the trade-off between the number of required ships and the
fill rate in the middle of simulation.
Figure 6 Number of required ships to achieve a 100% fill rate n days after the disaster (see online
version for colours)
Next, we determine the required kind and amount of stock of goods. To carry out the
assignment the simulator came up with, shortage of goods should not occur at the origin
port until the end of simulation period. Therefore, we have to store enough goods to
cover the required kind and amount at each origin port. In our model, the required goods
at each origin port are defined as the kind and amount of goods which are exported from
the origin port during the simulation period.
260 H. Wang and K. Tanaka
4 Case study
Figure 7 Expected distribution of seismic intensity in times of Tonankai earthquake (see online
version for colours)
Figure 8 The region which is damaged most seriously in each scenario (see online version
for colours)
Management of marine logistics in the case of emergency or disaster 261
When we estimate the required resources for marine transportation, we have to estimate
several scenarios of disaster. This is because the degree of damage is different for each
scenario, meaning that required resources are different, too. In this case study, we
expected two scenarios of earthquake: The red X marks in Figure 8 indicate the regions
which are most seriously damaged in each scenario. The blue boxes indicate the origin
ports and green boxes indicate the destination ports.
We assumed that all origin ports can be used in scenario 1. On the other hand, origin
port 2 cannot be used in scenario 2 because of the earthquake damage. In addition, we
estimated the number of evacuees around the destination ports for each scenario as shown
in Table 2.
Table 2 Number of evacuees for each scenario
Next, we calculate demand on goods. Table 4 shows the kind and amount of required
goods for evacuees. In this table, goods for children include milk powder and baby’s
diapers, and goods for elderly people include food for the elderly people and adapted
diapers.
Table 4 List of required goods
Then, we estimate how many evacuees will arise at each region based on the scenarios.
From this expected evacuees and the list of goods, we calculate the demand of goods for
each region.
Figure 9 The state immediately after the earthquake (see online version for colours)
Management of marine logistics in the case of emergency or disaster 263
Figure 10 The state four days after the earthquake (see online version for colours)
Figure 11 The state seven days after the earthquake (see online version for colours)
After determining ports and ships conditions as well as the demand of goods, we carry
out the marine logistics simulation. Simulation period is set to seven days. We calculate
the minimum number of required ships to cover the demand in this period. Figure 9 to
Figure 11 show the state of demand, supply and transportation routes obtained by the
simulation of scenario 1.
Figure 12 Difference of required ships with and without the reassigning algorithm in scenario 1
(see online version for colours)
Then, we consider the trade-off between the number of required ships and the fill rate in
the middle of the simulation. Figure 13 shows the number of required ships for 100% fill
rate n days after disaster in scenario 1. According to this figure, if we prepare 24 ships for
transportation, the fill rate equals to 100% seven days after disaster. However, we have to
prepare 38 ships to achieve 100% fill rate three days after disaster.
Figure 13 Number of required ships for 100% fill rate in scenario 1 (see online version
for colours)
Next, we determine the kind and amount of goods to stock. Figure 14 shows the required
stock of goods at each origin port when we use 24 ships under the reassigning algorithm.
Management of marine logistics in the case of emergency or disaster 265
Figure 14 Required stock of goods at each origin port in scenario 1 (see online version
for colours)
4.3.2 Scenario 2
Figure 15 shows the relationship between the number of ships and fill rate of goods in
scenario 2. We will see that 40 ships are required for a 100% fill rate when we do not use
the reassigning algorithm. On the other hand, only 24 ships are required for a 100% fill
rate when we use the reassigning algorithm. Similarly to scenario 1, this result suggests
that the reassigning algorithm decrease the number of required ships for transportation.
Figure 15 Difference of the required ships with and without the reassigning algorithm in
scenario 2 (see online version for colours)
266 H. Wang and K. Tanaka
Then, we consider the trade-off between the number of required ships and the fill rate in
the middle of simulation. Figure 16 shows the number of required ships for a 100% fill
rate n days after disaster in scenario 2. According to this figure, if we prepare 24 ships for
transportation, fill rate equals to 100% seven days after disaster. However, we have to
prepare 42 ships to achieve a 100% fill rate three days after disaster.
Figure 16 Number of required ships for 100% fill rate in scenario 2 (see online version
for colours)
Next, we determine the kind and amount of goods to stock. Figure 17 shows the required
stock of goods at each origin port when we use 24 ships under the reassigning algorithm.
Figure 17 Required stock of goods at each origin port in scenario 2 (see online version
for colours)
Management of marine logistics in the case of emergency or disaster 267
Figure 18 Required kind and amount of stock of goods (see online version for colours)
5 Conclusions
We designed a marine logistics system for disaster or emergency situations. This system
includes a reassigning algorithm, which decreases the number of required ships for
transporting goods. Compared with previous approaches, this model determines
quantitative results using optimisation. By using this model, we can calculate required
resources for supply and demand of goods for evacuees.
We applied this system to estimate the required resources for marine logistics in case
of Tonankai earthquake. The simulation results show that we have to prepare 24 ships
and 329,000 tons of goods.
However, we only considered the transportation between origin port and destination
port in this paper. In future research, the transportation between destination port and
evacuees should also be considered.
References
Bastos, M.A.G., Campos, V.B.G. and Bandeira, R.A.M. (2014) ‘Logistic processes in a
post-disaster relief operation’, Procedia – Social and Behavioral Sciences, Vol. 111,
pp.1175–1184.
Carvalho, H., Maleki, M. and Cruz-Machado, V. (2012) ‘The links between supply chain
disturbances and resilience strategies’, Int. J. Agile Systems and Management, Vol. 5, No. 3,
pp.203–234.
268 H. Wang and K. Tanaka
Fizzanty, T., Russell, I. and Collins, R. (2013) ‘Learning from failed supply chains: the application
of complex adaptive systems and a modified SAFE framework in evaluating proposed system
improvements’, Int. J. Agile Systems and Management, Vol. 6, No. 3, pp.232–258.
Holguín-Veras, J., Taniguchi, E., Jaller, M., Aros-Vera, F., Ferreira, F. and Thompson, R.G. (2014)
‘The Tohoku disasters: chief lessons concerning the post disaster humanitarian logistics
response and policy implications’, Transportation Research Part A: Policy and Practice,
Vol. 69, pp.86–104.
Kaynak, R. and Tuğer, A.T. (2014) ‘Coordination and collaboration functions of disaster
coordination centers for humanitarian logistics’, Procedia – Social and Behavioral Sciences
Vol. 109, pp.432–437.
Kuse, H. and Yano, Y. (2011) ‘Logistics planning for disaster prevention’, City Planning Review,
Vol. 60, No. 3, pp.87–90.
Lin, Y., Batta, R., Rogerson, P.A., Blatt, A. and Flanigan, M. (2011) ‘A logistics model for
emergency supply of critical items in the aftermath of a disaster’, Socio-Economic Planning
Sciences, Vol. 45, No. 4, pp.132–145.
Majima, T., Watanabe, D., Takadama, K. and Katsuhara, M. (2013) ‘A development of
transportation simulator for goods in disasters’, SICE Journal of Control, Measurement, and
System Integration, Vol. 6, No. 2, pp.131–136.
Namimatsu, T., Tamura, H., Sengoku, M., Shinoda, S. and Abe, T. (1997) ‘Analysis of algorithms
using the set division on a delivery problem’, IEICE Technical Report, Vol. 97, No. 403,
pp.41–48.
Senda, Y. and Suzuki, A. (2013) The 2013 Spring National Conference of Operations Research
Society of Japan Abstracts, pp.98–99.
Vlachos, D., Iakovou, E., Papapanagiotou, K. and Partsch, D. (2012) ‘Building robust supply
chains by reducing vulnerability and improving resilience’, Int. J. Agile Systems and
Management, Vol. 5, No. 1, pp.59–81.