Optimizing Empty Container
Repositioning at Inland Terminals
Using Data Analytics to Make an Informed Decision for a
Seaport in Sweden

Master’s Thesis in Supply Chain Management

Rahul R Kowshik

DEPARTMENT OF TECHNOLOGY MANAGEMENT AND ECONOMICS
DIVISION OF SERVICE MANAGEMENT AND LOGISTICS

CHALMERS UNIVERSITY OF TECHNOLOGY
Gothenburg, Sweden 2023
www.chalmers.se
Report No. E2023-017

www.chalmers.se




Report No. E2023-017

Optimizing Empty Container Repositioning at
Inland Terminals

Using Data Analytics to Make an Informed Decision for a Seaport in
Sweden

Rahul R Kowshik

Department of Technology Management and Economics
Division of Service Management and Logistics
Chalmers University of Technology

Gothenburg, Sweden 2023



Optimizing Empty Container Repositioning at Inland Terminals
Using Data Analytics to Make an Informed Decision for a Seaport in Sweden
Rahul R Kowshik

© Rahul R Kowshik, 2023.

Supervisor: Vendela Santen, Maritime at RISE (formerly SSPA), Sweden
Supervisor: Sara Rogerson, Maritime at RISE (formerly SSPA), Sweden
Supervisor & Examiner: Ivan Sanchez-Diaz, Department of Technology
Management and Economics

Report No. E2023-017
Department of Technology Management and Economics
Division of Service Management and Logistics
Chalmers University of Technology
SE-412 96 Gothenburg
Telephone +46 31 772 1000

Cover: A Seaport with vessels docked for loading and unloading.

Gothenburg, Sweden 2023



Optimizing Empty Container Repositioning at Inland Terminals
Using Data Analytics to Make an Informed Decision for a Seaport in Sweden
Rahul R Kowshik
Department of Technology Management and Economics
Chalmers University of Technology

Abstract
Sweden, as a country, maintains a fair balance between imports and exports.
However, in this post-pandemic era, there has been a rise in both imports and
exports. Consequently, the transportation of goods involves extensive movement of
containers. This thesis focuses on optimizing the existing container transport
network from the perspective of the seaport, shipping lines, inland terminals, and
rail operators. The network and flows are modelled as a mixed integer linear
programming problem using Python and Gurobi. The goal is to meet the demand
for empty containers while reducing turnaround time, container kilometres, the
number of containers, and emissions. The results of the optimization model aid
decision-making in selecting inland terminals for triangulation. The selection
criteria for inland terminals include the number of containers handled, total
emissions, triangle distance factor, and container kilometres. Currently, the flow of
containers does not employ any strategies for utilizing empty containers. However,
through the analysis of the optimization model, the thesis achieves the reuse of
empty import containers by implementing triangulation and storage at inland
terminals. Implementation of these strategies holds significant potential for cost
savings for large shipping lines. In the current container transport system, there
are four legs of transport: two for import flows (from port to importer and back)
and two for export flows (from port to exporter and back). By implementing the
proposed strategies, the number of legs in the transportation system can be
reduced from four to three, resulting in increased efficiency. The results of the
optimization model have demonstrated promising potential for cost savings across
various inland terminals. The implications for different actors involved in the
container transport network are thoroughly investigated in this study. Moreover,
this research addresses the gap between qualitative decision-making and the need
for substantial quantitative analysis. By utilizing the optimization model,
decision-makers now have a valuable tool that provides quantitative insights,
enabling more informed and data-driven decision-making processes.

Keywords: triangulation, optimization, containers, inland terminals, strategies





Acknowledgements
I would like to take this opportunity to express my heartfelt gratitude to all the
individuals who have contributed to the successful completion of this thesis report.
The culmination of my efforts over the last 6 months has resulted in this Thesis
report in front of you.

First and foremost, I am immensely grateful to my examiner and supervisor, Ivan
Sanchez-Diaz, for their unwavering support, guidance, and expertise throughout
this research journey. Their valuable insights, constructive feedback, and constant
encouragement have been invaluable in shaping the direction and quality of this
thesis.

I would also like to extend my deepest appreciation to Vendela Santén and Sara
Rogersson, my supervisors at Maritime at RISE (formerly SSPA), for their
invaluable contributions, mentorship, and practical insights. Their wealth of
industry knowledge and guidance has provided a solid foundation for this research.

I am also thankful to the faculty members of Chalmers University of Technology,
for providing me with a conducive learning environment and for their academic
mentorship. Their knowledge and expertise have been influential in expanding my
understanding of the subject matter.

I am indebted to Emmelie Gustafsson from the Seaport for her external assistance
and expertise. Her willingness to share her insights, engage in discussions, and
provide feedback has been instrumental in enhancing the depth and breadth of this
research.

Special thanks go to Stavros Kontos from Maritime at RISE for his assistance with
Python programming. Furthermore, I would also like to express my gratitude to
the Gurobi Community and Forums for their support and assistance with Gurobi
optimization. Their prompt responses and wealth of knowledge have been
invaluable in resolving technical queries and optimizing the modelling process.

I would like to extend my heartfelt thanks to Lokesh Kumar Kalahasthi from IIT
Delhi for help during the start of the thesis, Fredrik Barthel from Trafikverket, Per
Wide for fruitful discussions, Sönke von Wieding and Joanna Ellis from Maritime
at RISE for insights on the energy and emission analysis all of who generously
shared their time, experiences, and expertise during the course of this research.
Their willingness to impart knowledge and engage in insightful discussions greatly
enriched the content and quality of this thesis.

Stig Staghøj Knudsen and Sharat C S, two individuals who have been of great
support during the difficult times, I dearly appreciate your support and feedback.



I would also like to express my gratitude to my family and friends for their
unwavering support, understanding, and encouragement throughout this
challenging endeavour. Their belief in my abilities and constant motivation
provided the strength and inspiration needed to overcome obstacles and persevere.

In conclusion, I am deeply grateful to everyone who has played a part in this thesis
project. Your contributions have been indispensable, and I am honoured to have
had the opportunity to work with such remarkable individuals.

Rahul R Kowshik, Gothenburg, May 2023



List of Acronyms

Below is the list of acronyms that have been used throughout this thesis listed in
alphabetical order:

CIR Container Imbalance Ratio
ECR Empty Container Repositioning
H&S Hub and Spoke
IT Inland Terminal
IP Integer Programming
LP Linear Programming
MILP Mixed-Integer Linear Programming
MPC Multi-Port Calling
Q Triangle Distance Factor
RBV Resource-Based View
SL Shipping Line
TEU Twenty-foot Equivalent Unit
TIR Triangle Imbalance Ratio
VRIO Value, Rarity, Imitability and Organization

i





Nomenclature

Below is the nomenclature of indices, sets, parameters, and variables used
throughout this thesis.

Indices

i,j Indices for network nodes
k Indices for shipping line operators/container owners
t Index for time step

Sets

N ∈ Nodes Set of all nodes in the network.
IT ⊂ N − {P} Set of all Inland Terminals except Seaport.
SL ∈ Shipping
Line

Set of all shipping lines which own the containers.

Parameters

S_et
(P,j,k) Supply of empty export container at node

D_et
(P,j,k) Demand of empty export container at node

C(i,j) Cost of movement of containers between nodes. Unit in
kilometres.

X_ef t
(i,j,k) ∈ R+ Flow of laden export containers.

X_if t
(i,j,k) ∈ R+ Flow of laden import containers.

δt
(i,j) Binary parameter for having at most one active supply

arc between inland terminals.

iii



Variables

X_eet
(i,j,k) ∈ R+ Flow of empty export containers from node i to node j

for k ∈ SL.
X_iet

(i,j,k) ∈ R+ Flow of empty import containers from node i to node j
for k ∈ SL.

iv



Contents

List of Acronyms i

Nomenclature iii

List of Figures vii

List of Tables ix

1 Introduction 1
1.1 Purpose and Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Literature Review 7
2.1 Key Actors for Inland ECR . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Factors Contributing to ECR . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Analytical Strategies for ECR Reduction . . . . . . . . . . . . . . . . 9
2.4 Research and Management Strategies for ECR Reduction . . . . . . . 11

3 Methodology 13
3.1 Case Study: Seaport . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Inland Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2.1 Sets and Indices . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.3 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.3 Scenario Description . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3.1 Complete Collaboration Scenario . . . . . . . . . . . . . . . . 20
3.3.2 No Collaboration and Alliance Collaboration Scenario . . . . . 20

3.4 Performance Metrics and Indicators . . . . . . . . . . . . . . . . . . . 21
3.4.1 Criteria for Triangle Selection . . . . . . . . . . . . . . . . . . 23

3.5 Energy Consumption and Emissions . . . . . . . . . . . . . . . . . . . 23
3.6 Triangle Selection Process . . . . . . . . . . . . . . . . . . . . . . . . 24

4 Results and Analysis 27
4.1 Case Study Descriptive Analysis . . . . . . . . . . . . . . . . . . . . . 27
4.2 Base Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

v



Contents

4.3 No Collaboration Scenario . . . . . . . . . . . . . . . . . . . . . . . . 32
4.4 Alliance Collaboration Scenario . . . . . . . . . . . . . . . . . . . . . 33
4.5 Complete Collaboration Scenario . . . . . . . . . . . . . . . . . . . . 35
4.6 Savings from Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.7 Inland Terminal Selection based on Re-Utilization . . . . . . . . . . . 37

4.7.1 Asymmetry and Container Imbalance Ratio (CIR) . . . . . . . 37
4.8 Triangle Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5 Discussion 45
5.1 Re-utilization of Containers . . . . . . . . . . . . . . . . . . . . . . . 45

5.1.1 Top 3 Inland Terminals . . . . . . . . . . . . . . . . . . . . . . 46
5.2 Triangulation of Containers . . . . . . . . . . . . . . . . . . . . . . . 46

5.2.1 Containers in Triangulation Flows . . . . . . . . . . . . . . . . 47
5.2.2 Container-kms, Energy & Emissions Savings . . . . . . . . . . 48

5.3 General Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.3.1 Global Implementation and Disruptions . . . . . . . . . . . . 50
5.3.2 Turn Around Time . . . . . . . . . . . . . . . . . . . . . . . . 50
5.3.3 Number of Containers and Utilization . . . . . . . . . . . . . . 51
5.3.4 Trust, Alliances and Collaborations . . . . . . . . . . . . . . . 51
5.3.5 Energy Consumption and Emissions . . . . . . . . . . . . . . . 51

5.4 Implications on Actors . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.4.1 Shipping Line Perspective . . . . . . . . . . . . . . . . . . . . 52
5.4.2 Freight Forwarder Perspective . . . . . . . . . . . . . . . . . . 53
5.4.3 Seaport Perspective . . . . . . . . . . . . . . . . . . . . . . . . 53
5.4.4 Inland Terminal Perspective . . . . . . . . . . . . . . . . . . . 53

6 Conclusion 55
6.1 Academic Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.2 Industrial Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.3 Future Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

References 59

A Appendix 1 I

vi



List of Figures

1.1 Flows without Triangulation . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Flows with Triangulation . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Thesis Flow Chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3.1 Process Flow Chart for Base Case . . . . . . . . . . . . . . . . . . . . 14
3.2 Methodology Flow Chart . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Representation of Flows in the Base Case . . . . . . . . . . . . . . . . 16
3.4 Process Flow Chart with Utilization Strategies . . . . . . . . . . . . . 19
3.5 Representation of Flows in the Base Case . . . . . . . . . . . . . . . . 20
3.6 Representation of Flows in the No Collaboration Scenario and the

Alliance Collaboration Scenario . . . . . . . . . . . . . . . . . . . . . 21
3.7 Triangle Selection Process Flow Chart . . . . . . . . . . . . . . . . . 25

4.1 Share of Import and Export Containers for every Inland Terminal . . 28
4.2 Percentage Share of Shipping Lines . . . . . . . . . . . . . . . . . . . 29
4.3 Source of Empty Containers . . . . . . . . . . . . . . . . . . . . . . . 30
4.4 Base Case - Flows of Empty Export Containers . . . . . . . . . . . . 31
4.5 Base Case - Share of Empty Export Containers . . . . . . . . . . . . 31
4.6 No Collaboration - Flows of Empty Export Containers . . . . . . . . 32
4.7 No Collaboration - Share of Empty Export Containers . . . . . . . . 33
4.8 Alliance Collaboration - Flows of Empty Export Containers . . . . . 34
4.9 Alliance Collaboration - Share of Empty Export Containers . . . . . . 34
4.10 Complete Collaboration - Flows of Empty Export Containers . . . . . 35
4.11 Complete Collaboration - Share of Empty Export Containers . . . . . 36
4.12 Import - Export Asymmetry . . . . . . . . . . . . . . . . . . . . . . . 37
4.13 Container Imbalance Ratio (CIR) . . . . . . . . . . . . . . . . . . . . 37
4.14 Triangle Imbalance Ratio (TIR) for Selected Triangles . . . . . . . . . 40
4.15 Selected Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.1 Container Triangulation Over Time (in Months) . . . . . . . . . . . . 47
5.2 Container Distribution Over Time (in Months) . . . . . . . . . . . . . 49

6.1 Desirability Feasibility Viability . . . . . . . . . . . . . . . . . . . . . 56

vii



List of Figures

viii



List of Tables

3.1 Functions of Different Types of Container Flows . . . . . . . . . . . . 14

4.1 Total Containers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Costs from Optimization Models . . . . . . . . . . . . . . . . . . . . . 27
4.3 Relative Savings of Empty Container Distance Travelled . . . . . . . 28
4.4 Savings and Contribution in Number of Containers in each Scenario . 29
4.5 Cost Break-Up per Container Flow Type . . . . . . . . . . . . . . . . 30
4.6 Cost Break-Up per Container Flow Type . . . . . . . . . . . . . . . . 32
4.7 Cost Break-Up per Container Flow Type . . . . . . . . . . . . . . . . 33
4.8 Cost Break-Up per Container Flow Type . . . . . . . . . . . . . . . . 35
4.9 Energy Consumption and Emission Savings . . . . . . . . . . . . . . 36
4.10 Percentage Re-utilization and Savings due to Re-utilization at Each

Inland Terminal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.11 % Share of Containers for Each Shipping Line in Triangulation . . . . 40
4.12 Comparison of Savings in Triangles . . . . . . . . . . . . . . . . . . . 42
4.13 Comparison of Savings in Triangles (continued) . . . . . . . . . . . . 43

5.1 Re-utilization Savings . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.2 Percentage Triangulation within a Triangle . . . . . . . . . . . . . . . 47
5.3 Savings in Container-kms, Energy and Emissions . . . . . . . . . . . 48

A.1 Origin - Destination Matrix of Inland Terminals in kilometres (km) . I

ix



List of Tables

x



1
Introduction

Global is the new local for the world. Due to disruptions caused by the pandemic
and the invasion of Ukraine, there are some shifts to local production and supply
with increased back-shoring and a focus on sustainability goals. No matter how
global or local, transportation and logistics are ubiquitous in keeping the supply
chain persistent. Global and local supply chains are the backbone of the system
making goods available to consumers and enabling globalization. These chains
refer to the interconnected network of manufacturers, suppliers, distributors, and
retailers transporting goods and services worldwide (Cohen & Mallik, 1997).
Supply chains are made up of individuals, businesses, and organizations connected
to ensure the efficient production, delivery and sales of goods. They allow
companies to source materials, products and services from around the world to
create a competitive advantage in their respective markets. Aided by these chains,
the majority of things present around us have been transported at some point.
There are several modes enabling transport; rail, air, sea and road transport.
Earth is 29% land and 71% water, instantly making sea/maritime transport the
largest and most practical between destinations.

As much as 90% of everything around us has been on a vessel at some point (Faber
et al., 2021). To move such quantities, the maritime sector requires containers to
store goods and commodities during transport. Containers are standard in the
maritime sector and are available in certain standard sizes. The calculation of
volume is by TEUs (twenty-foot equivalent units). The movement of containers in
the chain leads to benefits enabled by shipping but comes with a cost and some
negative impacts. The goal in general is to minimize the cost and negative impacts
while maintaining reliability and quality. Before reaching the origin port and after
reaching the destination port, the containers are moved using road and/or rail
transport. The loaded containers are delivered to customers and stripped. Once
the containers are empty, the containers occupy the same space in the system.
They need to be moved to customers who want to export goods or sent back to the
port into storage or sometimes even shipped empty to fulfil demand in other
locations. The movement of these empty containers is termed empty container
repositioning (ECR) (Song & Carter, 2009). Dealing with several empty containers
and ECRs reduces the percentage utilization of resources and leads to high costs
for some or no value addition in the chain. The estimated incurred losses due to
ECR was $20 billion in 2015 (Sanders et al., 2015). A solution with high potential
to reduce cost and emissions is the use of dry ports/inland terminals (Lättilä et al.,
2013; Roso & Lumsden, 2010). In their article Roso and Lumsden (2010) defines a

1



1. Introduction

dry port as: ’dry port is an inland intermodal terminal directly connected to
seaport(s) with high capacity transport mean(s), where customers can leave/pick
up their standardized units as if directly to a seaport’.

Upon reaching the country of destination, there are numerous operations to be
conducted in the hinterland. The movement of loaded containers cannot be avoided
but can be reduced by having a higher container fill rate and utilization. On the
other hand, the movement of empty containers is necessary and can only be reduced
but not stopped completely. The transport of empty containers creates several new
instances of travel. The cost of logistics is high predominantly and further adds
to commotion, emissions and delays. However, ECR can be reduced drastically
to improve the utilization of the containers in the supply chain resulting in the
reduction of the number of containers in flows and the turn around time mainly
via street turns and triangulation (Tarudin, 2013). Turnaround time for a vessel
is the time a vessel takes from the day of arrival at a port to the day of departure
from the port (Ducruet & Merk, 2013). The same applies to containers as well, the
time difference between the day a container is sent from the port to the day it is
returned to the port. Street turn is defined as the movement of empty containers
from the local assignee directly to the local shipper, reducing extra trips(Jula et
al., 2006). Along similar lines, an import followed by an export using the same
container is called triangulation (Reinhardt et al., 2012). An illustration of flows
without and with triangulation is shown in Fig.1.1,1.2 based on (Feng & Moreno
Sanchez Briseno, 2021).

Figure 1.1: Flows without Triangulation

Maritime movements globally are responsible for 3% of all anthropogenic CO2
emissions (Faber et al., 2021), 15% of nitric oxides (NOx) and 4-9% of sulphur
oxides (SOx) (Eyring et al., 2010). Focusing on Europe, where the maritime sector
is a major contributor to the economy of a country, emissions produced are lowest
at 13,5% compared to 71% and 14,4% from road transport and aviation
respectively (EEA & EMSA, 2021). The long-term goal set forth by the European
Union is to reduce emissions from the transport sector by 60% by 2050 (Santén &
Rogerson, 2018).

2



1. Introduction

Figure 1.2: Flows with Triangulation

Sweden imported goods approximately worth SEK 173,2 billion and exported
worth SEK 172,1 billion in 2022 (Statistics, 2022). The value resembles a 50-50
share from a monetary value perspective and does not represent the same for the
number of containers. Sweden handled nearly 1,65 million TEUs in 2022 (Placek,
2022). Sweden imports several of its commodities. This in turn translates to
having great waterways and maritime control across the country. Handling such
quantities is no small feat which leads to substantial emissions from all operations.
The transport of the containers between the port and inland terminals is mainly
by rail. The use of rail between the Seaport and inland terminals reduces
emissions as the source of energy for rail is from renewable sources (NTM, 2023).
Rail is a greener alternative to road and air (EEA, 2021). The use of rail benefits
the Seaport in handling larger volumes in a greener way.

Based on requests from shipping lines, along with the self-interest of the Seaport
have led to the inception of this project led by Maritime at RISE (previously
SSPA) in collaboration with Chalmers University of Technology and other
organizations. A first study proposed a simulation to show the feasibility of
implementing street turns in Eskilstuna, Sweden (Hao & Ibrahim, 2023). The
study focused on the clients connected to the dry port in Eskilstuna. Similar
studies have been conducted in other countries and be highly beneficial (Song &
Carter, 2009; Meng & Wang, 2011; Furió et al., 2013; Kolar et al., 2018; Kuzmicz
& Pesch, 2019; Gusah et al., 2019). This study will focus on implementing
triangulation as a solution for ECR reduction from the Seaport’s perspective and
the inland terminals it is connected to. Thus, shedding light on the possibility of
potential savings for all actors with the bigger and more complex model
established across the country to optimize ECR.

This master thesis proposes a quantitative study to assess potential benefits and
disclose the potential improvements, consequently reducing lead times and costs,
traffic, energy consumption and emissions in Sweden’s hinterland transport sector.
The study from the Seaport’s perspective will give an idea of potential savings for

3



1. Introduction

other ports globally.

1.1 Purpose and Scope

The concept of re-utilization and triangulation is an initiative with significant
potential for reducing empty container repositioning (ECR) and has already been
implemented in the hinterlands of several other countries. A data analysis
approach can provide valuable insights for decision-making regarding the feasibility
of implementing these strategies. In Sweden, a case study is conducted from the
perspective of a seaport, with a specific focus on container flows and interactions
between the seaport and inland terminals. The objective of this thesis is to assess
the demand for empty containers while minimizing container movements,
presenting a supply and demand problem. Achieving economies of scale would
result in greater benefits. The most suitable triangulation scenario would involve
terminals that handle a large volume of containers and are located near each other,
relative to the seaport. Another important parameter is the comparison of total
container numbers before and after implementation, considering different models.
Terminals meeting these conditions will be identified as suitable candidates for
triangulation. Regarding re-utilization, inland terminals with substantial import
activity will have greater potential for re-utilizing containers. This can be
measured using the container imbalance factor (CIR).

The goals of the thesis are: to identify the best triangle that can be formed between
inland terminals and to analyse the feasibility, cost savings, energy consumption and
emissions reductions, and implications of implementing it. The best inland terminals
where re-utilization is highest resulting in substantial savings. The current network
will be compared with the optimized model based on time, and distance travelled
by the empty containers to reduce the turnaround time and the total number of
containers in the network and distance travelled.

1.2 Research Questions

To meet the scope of the project and fulfil the requirements, the following research
questions have been formulated:

1. What cost savings can be achieved through the practice of triangulation at
inland terminals? Additionally, which specific inland terminals are considered
optimal choices for implementing triangulation?

2. What are the potential cost benefits associated with container re-utilization
at individual inland terminals?

3. What are the benefits of implementing the strategies for optimizing the
utilization of empty containers, and how does it impact seaports, shipping
lines, and inland terminals?

4



1. Introduction

1.3 Limitations
The study is based on the data for import and export containers between 1/1/2022
to 30/6/2022. The data is provided by the Seaport which is confidential and thus,
actual data will be masked.

The potential solutions presented in this report are just a few of the alternatives
which meet the requirements of the project to a great extent. The terminals dealing
with certain container quantities are considered and others are removed from the
study. Some shipping lines did not have a high activity level with the Seaport and
thus were neglected from the study. Different types of containers are not considered
in the analysis. 90% of containers are 40ft. eq containers based on information and
statistics from the Seaport. Thus all containers are considered to be the same and
not differentiated further. The model does not consider the potential of storage at
inland terminals and how this can change the triangles and utilization of empty
containers. We do not have details on the actual costs of container handling, and
train schedule. The study does not cover the potential savings from a monetary
aspect.

1.4 Thesis Outline

Figure 1.3: Thesis Flow Chart

This thesis is divided into six chapters. The introduction feeds into Chapter 2,
containing the background information about empty container repositioning in the
hinterland, triangulation and strategies developed for ECR reduction.

Chapter 3 presents the method of examining the Base Case to investigate
improved models and feasibility, including the data description.

In Chapter 4, the results are presented for different test models with descriptive
and prescriptive models. The results are analysed for the practicality of the
solutions.

Chapter 5 leads to the discussion of the interaction between various actors
involved and the implications of the proposed solutions on the actors.

5



1. Introduction

Finally, chapter 6 concludes the discussions, in addition to the general conclusions
drawn from the study and the future scope and suggestions.

6



2
Literature Review

2.1 Key Actors for Inland ECR

The movement of loaded containers is a value addition but the movement of empty
containers is an added expense with little to no value addition as discussed above.
Several actors are working and interacting with containers in the supply chain.
The shipping line is the owner of the containers and controls the long-distance
segments of global freight distribution. The carrier (including freight forwarders
and transport providers) enables the transport of containers to ensure it reaches
their destination (Rodrigue, 2020). Different carriers use different modes of
transport as per their core competencies, geographical market locations and area
of focus. Carriers could be using rail, inland waterways, trucks and air transport.
The importers and exporters are customers in this case and are generating the
demand for containers (Rodrigue, 2020).

Shipping lines are responsible for owning more than 50% containers globally and
are also tasked with ensuring availability at different locations as per demands.
They are owners of vessels, operating them on trade routes (Tegbrant &
Karlander, 2023). The focus is on high volumes and high vessel fill rates for each
trip. The formation of strategic alliances is fueled by such goals with other
shipping lines having common goals (Rodrigue, 2020). This also leads to
coopetition, a situation when alliances are formed but everyone focuses on
optimizing their profits (Lin, Huang, & Ng, 2017). The shipping lines have several
customers and to ensure a high level of service, 80% of transport functions have
been outsourced to freight forwarders (Balci, Caliskan, & Yuen, 2019). The
underlying reason could be a lack of competency in inland logistics (Lun, Lai, &
Cheng, 2010). To improve this shortcoming, shipping lines have started to focus on
door-to-door deliveries (Tegbrant & Karlander, 2023). They also are sceptical
about implementing new strategies without enough existing data. This feeds into
analysing the desirability, viability and feasibility framework (Konietzko et al.,
2020; Fitzsimmons & Douglas, 2011).

Inland carriers’ different modes of transport focus on using the established network
for enabling transport for the shipping lines and attracting more shipping lines,
importers and exporters (Walter & Poist, 2003). Inland terminals are established
by private operators with support from governing bodies (Wilmsmeier, Monios, &
Lambert, 2011). Freight forwarders and transport providers enable transport from

7



2. Literature Review

inland terminals to importers and exporters. They have the resources and
knowledge of local geographical layout enabling themselves to attract customers
(Lai, Xue, & Hu, 2019). Furthermore, several services are also provided by freight
forwarders to ensure customer retention and attract more customers. Customs
clearance, warehousing and repackaging are some common ones. Collaboration
with importers and exporters also ensures that they are up-to-date with different
demands (Tegbrant & Karlander, 2023; Saeed, 2013).

Playing the role of end consumers, importers and exporters are driving the change
in the industry (Rodrigue, 2010). The profits for the other actors in the chain are
directly linked to the customers, inadvertently putting customer satisfaction as the
priority which feeds into customer retention and long-term relations and profits
(Raza, Woxenius, Vural, & Lind, 2023). However, their role in ECR is
unintentional, but has a big factor in leading to an increase or decrease in ECR
(Johansson & Williscroft, 2022).

The extra costs incurred due to ECR is generally borne by the carrier and the
shipping lines. On average 20-30% of the total containers on vessels are empty
(Kuzmicz & Pesch, 2019). Approximately 50% of containers are transported empty
in the hinterland (Lee & Meng, 2014). That is a significant number of containers
that are being underutilized. The customers return the container as soon as they
are stripped or loaded to save costs (Jula et al., 2006). Back in 2009, the estimated
cost of ECR was around $15 billion which accounts for almost 27% of the total
operating costs (Song & Carter, 2009). When sent back, the containers must be
stored and/or transported empty, reducing percentage utilization. In 2009, the
potential of optimizing ECR by 10% would have resulted in 30-50% profitability for
the industry (Theofanis & Boile, 2009).

2.2 Factors Contributing to ECR
Trade imbalance is defined as the ratio of net trade by gross trade, where exports
do not match imports and vice versa(Greenaway & Milner, 1981). This can be
extended to imbalances between continents and imbalances within a country. The
major contributor to ECR is trade imbalance on a global scale (Song & Carter,
2009). The difference in the share of imports and exports is also a contributing
factor for ECR in a country or location. There are several other factors affecting
ECR like dynamic behaviour, uncertainty, technological aspects, withholding
information/lack of communication, actor’s operational strategies and planning
(Song & Carter, 2009; Basarici & Satir, 2019).

• Dynamic behaviour refers to the change in the demands at different
geographical locations at any given time. Though the actual demand can be
predicted to a larger extent using historical data. The demand for empty
containers and the arrival of loaded containers will also vary due to
imbalances (Basarici & Satir, 2019). Storing and accumulating empty
containers in advance might help reduce the dynamism of the network.

8



2. Literature Review

Holiday seasons like Christmas and Chinese New Year, trigger greater
demand at a certain period of the year in a particular location. (Greenaway
& Milner, 1981; Song & Carter, 2009).

• Uncertainty in aspects of customer demands, consolidation, handling,
customs, discharges, disruptions caused due to weather, condition of
containers, delays due to human errors, and construction work at port
leading to a change in schedules for vessels will lead to deviation in the
movements of containers (Song & Carter, 2009; Borg, 2021; Basarici & Satir,
2019). According to Song and Carter (2009), demand uncertainty is the most
common phenomenon. For example, the delays and disruptions due to
Evergreen blocking the Suez Canal combined with the sudden surge in
demand in the later stage of the pandemic are still witnessed today by the
industry (Ramos et al., 2021; Kuźmicz, 2022).

• Technological aspects related to the types of containers that are used.
Containers are varying in dimensions and functionality. Furthermore, the
capacity and infrastructure of ports and terminals also play a role in
increasing/reducing ECR (Song & Carter, 2009).

• Withholding information/lack of communication arises due to reasons related
to trust in the network and the advantage an actor possesses aided by the
information. Information is viewed as a resource and relies on the
resource-based view (RBV) and the value, rarity, imitability and
organization (VRIO) framework actors try to withhold information (Franc &
Van der Horst, 2010). Forecasting and data analytics combined with open
communications can go a long way in reducing turnaround time and the
number of containers in the network (Heilig, Stahlbock, & Voß, 2020).

• Actors’ operation strategies and planning are tools for them to reduce costs and
are highly based on actual empty container movements. Though these are also
sometimes the reason for specific movements (Song & Carter, 2009; Basarici
& Satir, 2019). For example, a rail operator might have specific requirements
for consolidation and a minimum number of containers to run the trains. This
is planned to aid them to save costs, however, might increase/decrease the
turnaround time (Ülkü, 2012).

2.3 Analytical Strategies for ECR Reduction
The inefficiencies in the hinterland due to ECR can be categorized as a generic
supply and demand problem. Such problems are modelled and solved as transport
problems having constraints such as multiple destinations, time periods, quantities
and storage. The transportation problem is a popular type of linear program that
has been used in numerous fields (Chanas & Kuchta, 1996). It is modelled as a
network of nodes and paths that can be used to simulate flows of goods to form a
virtual network similar to that of the actual one (Patrick, 2022). Based on the

9



2. Literature Review

results of the model, the finer details related to specific destinations, commodities,
and periods can be focused on. The dynamic nature of supply and demand can be
modelled through this method (Chanas & Kuchta, 1996). The transport problems
fall under the category of linear programming and have parameters, constraints,
variables and an objective function depending on the case (Chanas & Kuchta,
1996; Patrick, 2022). The complex versions of the transport problems consider
multi-echelon, multi-product, multi-modal and multi-period and may be
deterministic or stochastic depending on the requirements of the scenario (De et
al., 2020). For more examples kindly refer to Table 2 in Kuzmicz and Pesch (2019)
and Table 3.2 in Hao and Ibrahim (2023).

Increasing global demand resulting in higher movements calls for improvements in
efficiency and utilization of available resources. One of the aspects to improve is to
reduce ECR movements and reduce strain on hinterland supply chains and
networks. Different methods ranging from qualitative studies to fold-able
containers to data analytics have been implemented and tested at various levels
and by different actors in the supply chain. The involvement of dry ports plays a
vital role in ECR movements as they act as a centre for carrying out several
terminal functions like consolidation, transhipment, storage, cross-docking, sorting,
etc (Roso & Lumsden, 2010; Hao & Ibrahim, 2023). In 2019, the number of dry
ports was 230, 400 and 100 in Europe, the USA and Asia respectively (Luo &
Chang, 2019). The number of dry ports within Sweden in 2022 was 12
(Khaslavskaya & Roso, 2022).

Numerous studies have been conducted to check the potential for reducing ECR
within Europe. While some focused on the hinterland and some on the global
scale, they analysed different methods. Furió et al. (2013) studies the optimization
of ECR using street turns in Valencia, Spain. Their study focuses on the cost of
operations with and without street turns. The focus was from a single-port
perspective. The integer programming (IP) program considers the time which
increases the complexity due to its dynamism (Furió et al., 2013). The scope of
improvement in large ports using street turns was also studied in Central and
Eastern Europe hinterland. The study was a qualitative study discussing the
strategies of carriers and their unwillingness to revise them. The results also reveal
the global orientation of ECR instead of local or regional orientation (Kolar et al.,
2018).

Building on the study, Kuzmicz and Pesch (2019) approaches ECR problems with
a focus on Eurasian intermodal transportation. The area of focus is the Chinese
One Belt One Road initiative. The model is a mixed integer linear program with
stochastic demand and uncertain future demands. Their approach falls under the
category of dynamic programming (Kuzmicz & Pesch, 2019). Moving further away
from Europe, Gusah et al. (2019) carried out a systematic analysis for empty
container analysis in Melbourne, Australia. The study is categorized as
agent-based modelling with simulation. In the United States, at Los Angeles and
Long Beach, Jula et al. (2006) studied the dynamic use of empty containers based

10



2. Literature Review

on a two-phase optimization problem. Eventually concluding that the reduction in
ECR will save costs, but also reduce congestion, noise and emissions in the local
area around the port and boost the economy as a result (Jula et al., 2006). In New
York and New Jersey, a study on the location of inland terminals/depots wrt. the
port shows that the location and demand around there plays a major role in ECR
(Boile et al., 2008).

A generalized approach using quantitative methods along with mathematical
modelling provides insights for better decision-making to implement strategies and
construct policies for ECR reduction. An approach of flow balancing considering
multiple shipping lines but without considering the dynamic characteristics of
container shipping is studied by Song and Carter (2009). The relation between
internal coordination and external container sharing is modelled as a mathematical
problem in a macro environment (Song & Carter, 2009). Moving ahead, Meng and
Wang (2011), examined a hub and spoke (H&S) system combined with a
multi-port-calling (MPC) shipping line(port calling sequence). The study focused
only on a single shipping organization and their operations. The optimization
model used a CPLEX solver and the result reflected large cost savings achieved by
combining H&S and MPC(Meng & Wang, 2011).

2.4 Research and Management Strategies for
ECR Reduction

The examples above are from an analytical perspective to check the potentiality of
methods to reduce ECR. However, quantitative checks need to be based on some
strategies and theories which are a more qualitative approach. Some methods that
have been analysed are based on policy-making, decision-making, collaborations,
communication, openness to information sharing and trust.

Braekers, Janssens, and Caris (2011) mention the different levels of
decision-making leading to empty container movements. The strategic level deals
with the network design while the tactical level deals with the services in the
network design. Further down in the value stream, the operational level is linked
to the container allocations and routing problems (Braekers et al., 2011). While
the different levels are a demarcation of different functions, factors like trust, risks
and commitment also play a role in shaping the supply chains.

"Trust is most elusive but also the most sought after" (anonymous) (Sahay, 2003).
For collaborations between actors having similar market value and size, trust and
risks are two factors that are most commonly a barrier forcing organizations to act
in an individualistic manner. Vilko, Ritala, and Hallikas (2019) has developed a
model to show how the size of an organization, visibility and control play their
part in an alliance/collaboration. Similar models developed by Kwon and Suh
(2005); Ireland and Webb (2007); Dekker, Sakaguchi, and Kawai (2013) portray
the detailed path to study trust and risk in large multi-modal supply chains and

11



2. Literature Review

organizations. Strategic alliances have been analysed and several barriers have
been identified that show that alliances may not be the best solution always
(Panayides & Wiedmer, 2011). The contracts and policies are only a part of the
alliances but the operational synergies also play an important part (Pierre & du
Tertre, 2000).

Container leasing merged with internet-based systems will drastically reduce ECR
if planned and executed properly (Chang et al., 2008; Braekers et al., 2011).
Container leasing focuses on the shipping lines leasing out containers instead of
owning them permanently. Containers could be leased for long-term or short-term
depending on the needs of the customers (Hao & Ibrahim, 2023; Braekers et al.,
2011). Internet-based systems using continuous tracking and information sharing
make it easier for all actors to know their containers. Theofanis and Boile (2009)
mentions that the success of these methods relies heavily on the willingness of
firms to share information, which eventually returns to the basic need for trust.

Kuzmicz and Pesch (2019) elaborate on the possibility of container substitution
and sharing. For satisfying the demand for 2 empty 20-foot containers, a single
40-foot container could be used instead. The substitution depends on several
factors and might not work in multi-commodity scenarios for instance. If the
capacity, destination and other factors overlap then container substitution can
have a beneficial impact (Chang et al., 2008; Kuzmicz & Pesch, 2019). Increasing
the chances of substitution container sharing will also increase the savings.
However, sharing of containers has its own set of problems which are related to
container types, location mismatch, import-export time mismatch, etc (Chang et
al., 2008; Hellekant & Rudal, 2021).

Foldable containers are one way to overcome barriers and save space. Six folded
containers stacked upon one another will occupy the same space as that of one
container (Konings & Thijs, 2001). The upfront purchasing cost of foldable
containers is high, maintenance is high and lifespan is lower compared to generic
containers. Research shows even after considering these factors, usage of fold-able
containers can bring about 50-60% savings in cost(Goh & Lee, 2016). Another
research estimates a savings of 75% in storage space with the use of foldable
containers (Moon, Do Ngoc, & Konings, 2013).

Connectainers is a new design for a 20-foot container which can be joined with
another 20-foot container to form a 40-foot container (Connectainer, 2018). The
company manufacturing the connectainers also claim the containers to be the same
dimensions as the standard containers and are watertight and resistant.
Furthermore, they are multi-modal transport compatible. The connecting and
disconnecting of such containers can take 30 minutes by skilled workers
(Connectainer, 2018; Kuzmicz & Pesch, 2019).

12



3
Methodology

3.1 Case Study: Seaport

The Seaport plays a vital role in the imports and exports in Sweden and
Scandinavia, as mentioned in Section.1. The data received from the Seaport
contains the details of individual container movements. For modelling, we use the
data related to the origin, destination, line operator, and date. The data is
further divided into sub-categories. The type of container is based on import or
export type. The owner of the container is a different shipping line. The different
rail operators handling each container and the unique container ID.

The total supply and demand are the same in all cases and the inland terminals
are the same as well. The total demand for empty export is ≈ 27000 containers.
The total supply of laden import containers is ≈ 28000 containers.

The current network under operation functions with flows originating from the
Seaport and terminating at the Seaport. The import containers are loaded and
bought into Sweden from other countries. These are then loaded onto trains and
sent to inland terminals for importers to collect and strip them. Once the
containers are stripped they are sent back to the port empty from the inland
terminals. Upon demand from exporters, empty containers are shipped from the
port to the exporters, then filled with goods and sent back to the port for export.
This movement of empty containers results in ECR (Song & Carter, 2009). The
port stores these empty containers unless exporters request for them or shipping
lines take them back to be used in another location, which also falls under ECR.
At present, the inland terminals do not store empty containers. The rail operators
currently shuttle between the Seaport and the inland terminals. There are no
trains scheduled by the shipping lines to run between the inland terminals. Thus,
triangulation is not possible in the current network. However, the rail network is
present and other trains are operating on those routes. The turn-around time at
present is around 5 days for importers and 7 days for exporters.

However, it can vary depending on the type of contract that a customer has with
the shipping line. There are a few instances when a container is reused instead of
having to send a new container with a new ID to meet the demand. However, such
instances are very rare and are not pre-planned as a solution to reduce ECR but
rather a coincidence resulting in some savings. The container flows in the network

13



3. Methodology

are distinguished by the container activity, the state of the container and the
purpose of the container. There are four types of container flows:

1. Laden Import Containers
2. Empty Import Containers
3. Empty Export Containers
4. Laden Export Containers

Figure 3.1: Process Flow Chart for Base Case

Container Flow Function
Type Based on Base Case Utilization Strategies

Activity and State Re-utilize ECR
Laden Import Supply Goods Supply Goods Supply Goods
Empty Import Backflow Re-utilize Triangulate
Laden Export Goods to Port Goods to Port Goods to Port

Empty Export Supply of Supplementary Supplementary
Empty Containers Containers Containers

Table 3.1: Functions of Different Types of Container Flows

The laden import containers carry the import goods to the receiver. Backflow can
be defined as the return flow of empty import containers from inland terminals to
the Seaport. The empty import containers in the base case contribute entirely to
the backflow. If the strategies are implemented, the empty import containers can
either be re-utilized at the inland terminal or transported to other inland terminals

14



3. Methodology

to meet the demand for empty export containers(triangulation). If both strategies
are not possible, it is then allocated to backflow. The laden export containers are
responsible for the transport of goods from the shippers to the Seaport. The empty
export containers are supplemented by the Seaport in the event of a shortage when
the demand is not fulfilled by the implementation of utilization strategies.

Figure 3.2: Methodology Flow Chart

The approach towards optimization of the network through triangulation is
promoted by the existing literature and simulations carried out by others. The
quantitative analysis using operational research tools will further reinforce the
findings. This will help broaden the transferability, improve quality and help with
validity. In this case study, from the perspective of the Seaport, the focus is on
flows originating and terminating at the Seaport. With similar ports across
Sweden, a similar model could be used to optimize flows from the perspectives of
other ports.

The case study of the Seaport is a transportation problem which can be further
categorized as a multi-period, single-commodity supply and demand problem. A
transport problem deals with minimizing the cost function or maximising the
profit for the entire network as modelled where the objective function is decided by
the shareholders. The approach used is to model it as a mixed-integer linear
program (MILP). The mathematical approach ensures that the quantitative half of
the analysis supports the qualitative half.

The Base Case represents the current modus operandi. The data on current
operations are used to calculate the total cost, which will be used for comparison
with the cost from the proposed models. The Seaport is the only node in the
network supplying containers. There is no triangulation and re-utilization of
containers.

3.2 Inland Network Model
The mathematical model used for optimization for all scenarios is elaborated
below. The total cost of operations is given by Eq 3.1.

Min Z =
∑
t∈T

∑
k∈SL

∑
(i,j)∈N

(X_eet
i,j,k + X_iet

i,j,k) · Ci,j+
∑
t∈T

∑
k∈SL

∑
(i,j)∈N

(X_if t
i,j,k + X_ef t

i,j,k) · Ci,j

Z =
∑
t∈T

∑
k∈SL

∑
(i,j)∈N

(X_et
i,j,k · Ci,j) +

∑
t∈T

∑
k∈SL

∑
(i,j)∈N

(X_f t
i,j,k · Ci,j)

(3.1)

15



3. Methodology

Figure 3.3: Representation of Flows in the Base Case

Subject to:

∑
ij

δ(ij) <= 1

if δ(ij) = 1 =⇒ Xt
(ij) ≥ 0

if δ(ij) = 0 =⇒ Xt
(ij) = 0

(3.2)

∑
k∈SL

S_et
(P,j,k) +

∑
k∈SL

∑
i∈IT

(X_et
(ijk) − X_et

(jik)) +
∑

k∈SL

X_et
(P,j,k) =

∑
k∈SL

D_et
(P,j,k)

∀i, j ∈ IT , if i ̸= j, k ∈ SL, t ∈ T

(3.3)
∑

k∈SL

X_et
(ijk) ≤

∑
k∈SL

S_et
(P,i,k) ∀i, j ∈ IT , if i ̸= j, k ∈ SL, t ∈ T (3.4)

∑
k∈SL

X_f
(t+1)
(j,P,k) =

∑
k∈SL

D_et
(i,j,k) ∀i, j ∈ IT, k ∈ SL, t ∈ T (3.5)

The notations used in the mathematical model are elaborated below:

3.2.1 Sets and Indices
N ∈ Nodes = {Seaport, Terminal 1, Terminal 2, Terminal 3, Terminal 4, Terminal
5, Terminal 6, Terminal 7, Terminal 8, Terminal 9, Terminal 10, Terminal 11} - All
nodes in the network.

IT ⊂ N − {P} Inland Terminals = All nodes except Seaport.

SL ∈ Shipping Line = {SL1, SL2 ... SL10} - All different owners of containers.
Note: For the Alliance Collaboration scenario, certain shipping lines in an alliance
are modelled differently. For example SL1=SL2=SL5.

16



3. Methodology

t ∈ T = {02/01/2022 .... 30/06/2022} - Range of dates.

i, j ∈ N : Indices of nodes (N).

k ∈ SL: Index of the shipping line.

3.2.2 Parameters
Parameters are numerical constants which help define the attributes of a system.
It is used to set up the scenarios for the model.

S_et
(P,j,k): Supply of empty export container at node Seaport ∈ N , i ∈ IT and

k ∈ SL.

D_et
(P,j,k): Demand of empty export container at node P ∈ N , i ∈ IT and k ∈ SL.

C(i,j): Cost of movement of containers between nodes (i,j) ∈ N . Unit in kilometres.
Note: Notice that the cost from node i to node j is the same as the cost from
node j to node i i.e. C(i,j) = C(j,i).

X_ef t
(i,j,k) ∈ R+: Flow of laden export containers from node i to node j for

k ∈ SL.

X_if t
(i,j,k) ∈ R+: Flow of laden import containers from node i to node j for

k ∈ SL.

δt
(i,j): Parameter for having at most one active supply arc between inland terminals

i, j ∈ IT for t ∈ T .

3.2.3 Variables
All variables are non-negative integers.

X_eet
(i,j,k) ∈ R+: Flow of empty export containers from node i to node j for

k ∈ SL.

X_iet
(i,j,k) ∈ R+: Flow of empty import containers from node i to node j for

k ∈ SL.

The objective function, Eq 3.1 includes two parts. The first part deals with the
total cost of empty container movements. The second with the total cost of laden
container movement. The constraints are the conditions that delimit the feasible
set for the decision variables. The constraints are explained further below
elaborating on their function in the model.

17



3. Methodology

Binary Flow Constraint: Eq 3.2 is modelled to restrict incoming flows into an
inland terminal. This ensures that an inland terminal can receive empty containers
from only one other inland terminal. It is structured so, to create a one-to-one
relationship between inland terminals to facilitate triangulation.

Flow Constraint: Eq 3.3 is responsible for ensuring that the demand is met for
every time instance at each terminal. The equation includes all types of possible
container providers, i.e., flows from the Seaport, containers available for
re-utilization and through triangulation.

Flow Conservation Flow: The flow conservation for each inland terminal is
denoted by Eq 3.4. The flow conservation focuses on the containers that are
received through the supply flow and are the only ones that can be used for
re-utilization or triangulation.

Laden Container Flows: Eq 3.5 represents the flow of laden containers which are
constant throughout all the scenarios. The laden import containers are sent from
the Seaport to the inland terminals and the laden export are sent from the inland
terminals to the Seaport.

3.3 Scenario Description

Building further on the Base Case and incorporating the utilization strategies,
some assumptions are necessary and are listed below. These assumptions are used
for modelling the different scenarios.

1. The Seaport has infinite supply and storage capacity. It is always available
to support any inland terminal when there is a demand which is not fulfilled
by either of the utilization strategies.

2. All containers are considered to be the same size i.e., 40ft. containers (2
TEUs).

3. Trains are available and running between inland terminals.

4. The laden import containers are stripped in a day and available the next day
for use at the inland terminal. They are termed as "supply of empty export
containers" during modelling.

5. The empty export containers are filled with goods in a day and available the
next day for return transport to the Seaport.

18



3. Methodology

Figure 3.4: Process Flow Chart with Utilization Strategies

19



3. Methodology

Based on the data and assumptions, three different scenarios are formulated and
analysed. The costs for transporting all laden containers will remain the same in
all cases. Hence, the objective functions in the proposed models do not have a
term focusing on the flows of laden containers. Rather, we focus just on the
variable associated with empty containers.

1. Complete Collaboration

2. No Collaboration

3. Alliance Collaboration

3.3.1 Complete Collaboration Scenario
In this scenario, the model is a triangulation model with an assumption that the
containers are common for all shipping lines. This allows any container to be used
by any customer at any time by surpassing the need for container matching. The
remaining containers at the end of each day are returned to the Seaport. The idea
is to reduce the constraints on the model and thus resulting in lower costs. The SL
set does not apply to this scenario.

Figure 3.5: Representation of Flows in the Base Case

3.3.2 No Collaboration and Alliance Collaboration Scenario
The containers are differentiated based on the shipping line and thus container
matching is a constraint on the system. Simply put, a container owned by SL1
cannot be used by a customer of SL2. The SL set is applicable here and each shipping
line is individually used. This scenario is also constrained by other flow constraints.
The Alliance Collaboration scenario is similar to the No Collaboration scenario
but the number of shipping lines involved is reduced as shipping line alliances are
formed, under which the SLs in an alliance can use each other’s containers. The

20



3. Methodology

SL set is based on the alliances. The alliances are based on the existing alliances in
the maritime industry. The three major alliances operate almost 80% of the flows
(Zheng & Luo, 2021).

Figure 3.6: Representation of Flows in the No Collaboration Scenario and the
Alliance Collaboration Scenario

3.4 Performance Metrics and Indicators
These metrics are measured as an output from the optimization model. They help
with comparisons and investigate further into the scenarios. The indicators are
also dependent on the data related to distances, quantities and flows.

• Container*km: Container-kilometer is directly proportional to the number
of containers as it is a product of the distance and the number of containers
in a link. This is also the output obtained as a result of the optimization
model. As the result is simply a value calculated by multiplying a constant
with a variable. The constant is the distance between two nodes and the
variable is the number of containers in flow between the said two nodes.
Hence, the reduction in the number of containers and the reduction in
container kilometres means the same. The percentage savings will be the
same for both values. However, the monetary savings would significantly
vary.

• Energy Consumption: Energy consumption is another criterion that acts
as a deciding factor for triangle selection. Higher the savings in energy
consumption the better the triangle. The procedure for energy calculations is
explained further in this chapter.

• Triangle Distance Factor, Q: The triangle distance factor is the ratio of
the distance between two inland terminals to the sum of the distances between

21



3. Methodology

the Seaport and each inland terminal. The energy and emission savings are
inversely proportional to the factor. The lower the factor, the higher the
potential for savings and vice versa. This is true only if decision-making is
based on the triangle distance factor. Since the number of containers directly
affects the energy values, the proportionality does not hold always.

Savings α
1
Q

Q is formulated as:
Q = dt

dp1 + dp2
(3.6)

Where,
dt = Distance between inland terminals, in kilometres.
dp1 = Distance between the Seaport and inland terminal 1, in kilometres.
dp2 = Distance between the Seaport and inland terminal 2, in kilometres.

• Container Imbalance Ratio (CIR): CIR is used as an indicator to depict
the imbalance in import and export containers at each terminal. It measures
the proportion of containers imported versus exported. An inland terminal
will have more imports as compared to exports if CIR is positive. If the CIR
is negative, there are more exports as compared to imports. A high positive
magnitude will indicate significantly higher imports compared to exports. A
low negative magnitude will indicate higher exports compared to imports. Eq
3.7 is the formula for CIR.

CIR = (Total Import Containers − Total Export Containers)
(Total Import Containers + Total Export Containers) (3.7)

• Triangle Imbalance Ratio (TIR): When triangles are selected, a TIR value
is calculated for each triangle to check the imbalance. TIR is based on CIR but
calculated for two terminals involved in the triangle. The import and export
containers are aggregated before using them in Eq 3.7. The TIR will provide
insight into whether the triangle is import-dominated or export-dominated.
If a triangle is import dominated there are higher potential for savings. If a
triangle is export-dominated the potential savings are reduced. This relation
might not hold true if there is a significant imbalance between the individual
terminals within a triangle. For example, if terminal A has very high exports
and terminal B has low imports, the potential for savings is reduced. In
an opposite scenario where terminal A has low exports and terminal B has
high imports, the savings are higher as more containers are available for re-
utilization and triangulation. If TIR is zero, it may not provide any significant
insights about how the triangle might perform but shows that there is no
imbalance. In such a case, other parameters and indicators will have to be
studied to analyse the triangle. TIR is one of the many indicators supporting
decision-making.

22



3. Methodology

3.4.1 Criteria for Triangle Selection
The triangles are selected based on criteria put together with the results of the
optimization model. The performance metrics and indicators are dependent on the
number of containers that are transported in each flow. The container*km, energy
consumption, emissions, CIR and TIR are also directly proportional to the number
of containers. However, the triangle distance factor is calculated based on the
distances between two nodes in the network and is not affected by the number of
containers.

• Number of Containers in Flows: The commodity in transport is a
shipping container. Each additional container in the network requires space
and adds a cost. An important factor to consider while selecting triangles
post-optimization is the reduction in the number of containers to satisfy the
same demand under different conditions. The goal is also to reduce the
number of containers that are added to the flow from the Seaport. More
containers that can be re-utilized or triangulated reduce the number of new
containers entering the system to fulfil the required demand.

3.5 Energy Consumption and Emissions
The use of rail as transport saves time, emissions and costs. To transport
containers across the country, rail is the best choice, given the flexibility and access
to a wider already existing rail network at a significantly reduced cost and lower
emissions. Trains help increase consolidation as well. The following method is used
to calculate the energy consumption and the emissions due to train movements to
satisfy the demand in the case study.

The trains in Sweden are electrified and run on electricity generated from a mix of
hydro and wind energy. The average train used by the Seaport in Sweden is 630m
long and carries 40 containers, as per information from NTM (2023). The energy
consumption for an electric train using electricity generated from a mix of wind
and hydro is estimated at 0.08 kWh/tonne-km from a well-to-wheel scenario. The
estimation includes a correction factor for topography as well. It is calculated
based on the total weight the train pulls when loaded and empty, the fill rate and
the extra distance travels when empty. Similarly, the value for emissions is
estimated at 0.9 g/tonne-km of CO2e (NTM, 2023). Based on assumptions, all
containers are 40-foot containers and the average weights for empty and loaded
containers are selected from Table 1 in Yildiz (2019). For calculating the total
energy consumption and emissions for all types of flows, it is necessary to include
the weights of empty and loaded containers. When focusing only on empty
container flows, 3.75 tonnes is considered for calculations (Yildiz, 2019).

The calculation procedure for energy consumption and emissions is presented
below:

23



3. Methodology

D = Distance travelled by containers, in container-km.
Ec = Total energy consumption, in kWh.
q = Energy consumed for 1 tonne-km, in kWh/tonne-km.
Em = Total emissions, in tonnes CO2e.
e = Emission generated for 1 tonne-km, in g/tonne-km.

W = Weight of a 40-foot container.
Wt = 3.75 tonnes - Average tare weight of a 40-foot container.
Wp = 26.7 tonnes - Average net weight (payload + container weight).
Wm = 30.4 tonnes - Average maximum weight when fully loaded.

Ect = Energy consumption of an empty container.
Ecp = Energy consumption of a loaded container at an average weight.
Ecm = Energy consumption of a fully loaded container at maximum weight.
Emt = Emissions generated from the transport of an empty container.
Emp = Emissions generated from the transport of a loaded container.
Emm = Emissions generated from the transport of a fully loaded container.

Total energy consumption is denoted by:

Ec =D · q · (Wt + Wp + Wm)
Ec =Ect + Ecp + Ecm

(3.8)

Total emission is denoted by:

Em =D · e · (Wt + Wp + Wm)
Em =Emt + Emp + Emm

(3.9)

3.6 Triangle Selection Process
The selection of triangles is based on analysing the results in three steps. The
first step deals with investigating the output from the optimization model. The
total output file is exported to Excel and the number of containers in each flow
is calculated. The top 10 triangles based only on the number of containers are
shortlisted as potential candidates. This gives an overview of all potential triangles
that can be implemented in the entire network based only on the criterion. Upon
selection of all potential triangulation candidates, the optimization model is executed
again in the second step. However, this time the inland terminals in a triangle are
isolated and optimized as a smaller model consisting of only the Seaport and the two
inland terminals. This presents the maximum savings possible in a triangle without
any interference or overlap with other triangles. It also eliminates the chances of
cannibalization of containers. When individual triangles are analysed, the focus is
only the flows within the triangle and the rest of the network is not considered at
all for analysis. The savings are calculated within the triangle and compared to the
Base Case. Moving further, the performance metrics and indicators for each triangle
are calculated. The third step is a deep dive into performance metrics and indicators

24



3. Methodology

Figure 3.7: Triangle Selection Process Flow Chart

for each triangle to select the top 3 triangles that are most beneficial to implement
from a holistic view. However, this method is only a quantitative approach helping
prioritize the best triangles in the network. Moving ahead the stakeholders will
have the option of ranking the different criteria as per their requirements and then
comparing the triangles to finalize on one or a few that can be implemented.

25



3. Methodology

26



4
Results and Analysis

4.1 Case Study Descriptive Analysis
The results from the different scenarios are analyzed and compared with each
other to have a better understanding of the optimization models. The triangles
and terminals are selected which meet the requirements of the utilization
strategies. The total number of containers are same in all scenarios. Table.4.1
shows the total containers in each category.

Container Type Quantity
Import ≈28000
Export ≈27000
Total ≈56000

Table 4.1: Total Containers

As seen in Table 4.1. there are 51% import containers and 48% export containers.
These are distributed over the first half of 2022 i.e., January to June. To segregate
the flow, it is divided into four types. The focus is on the reduction of containers
and the percentage re-utilization without creating fresh/new flows originating from
the Seaport to satisfy demands. There are four types of container flows:

1. Laden Import Containers
2. Empty Import Containers
3. Empty Export Containers
4. Laden Export Containers

Container Flows (million cont-km) Total
Scenario Laden Empty (million

Import Export Import Export cont-km)
Base Case 7.03 10.5 7.03 10.5 35.06

No Collaboration 7.03 10.5 4,53 8.39 30.45
Alliance Collaboration 7.03 10.5 3.71 7.6 28.84

Complete Collaboration 7.03 10.5 2.5 6.35 26.38

Table 4.2: Costs from Optimization Models

The costs of transporting laden import and laden export containers remain the same
in all scenarios as these flows are invariable. These flows are value-addition flows

27



4. Results and Analysis

and are based on the orders placed by customers (importers and exporters). The
actual savings will be measured across the empty import and empty export container
flows and the re-utilization of empty containers. The figures visualizing flows are
differentiated based on colours and the thickness of the lines represents the quantity
transported. Thick lines indicate higher quantities and thin lines represent lower
quantities The visuals of flows are divided based on distances into three sub-figures
to avoid overlapping of lines. To keep the names confidential the shipping lines are
masked. The total costs in each scenario decrease as there are fewer constraints
allowing the relaxation of flows in the network, as seen in Table 4.2.

Figure 4.1: Share of Import and Export Containers for every Inland Terminal

The constraints are mainly associated with the shipping lines which translates to
container-sharing restrictions. The relative percentage savings compared to the Base
Case for empty containers are calculated and presented in Table 4.3.

Scenario

Empty Import Empty Export
Distance

Savings
Distance

Savings(million (million
cont-km) cont-km)

Base Case 7.03 - 10.5 -
No Collaboration 4.53 36% 8.39 20%

Alliance Collaboration 3.71 47% 7.6 28%
Complete Collaboration 2.5 64% 6.35 40%

Table 4.3: Relative Savings of Empty Container Distance Travelled

28



4. Results and Analysis

Figure 4.2: Percentage Share of Shipping Lines

The saving in distance is a parameter that indicates the possibility of savings by
changing the constraints of the network and the flows in the network. Another vital
parameter for shipping lines is the savings in the number of containers required to
satisfy the demand for empty export containers. In the Base Case, there are ≈
27000 unique containers that are used to meet the total demand. However, with
re-utilization and triangulation, the number of unique containers required reduces
drastically as the empty import containers are re-utilized at the inland terminal or
transported through the triangle to satisfy the demand at the other terminal. The
number of containers originating from different sources is mentioned in Table 4.4.

Scenario Source of Empty Container and Contribution
Port Re-utilized Triangulation

Base Case 0 - -
No Collaboration 62% 20.4% 17.5%

Alliance Collaboration 49.3% 28.8% 21.8%
Complete Collaboration 32.6% 40.7% 25.6%

Table 4.4: Savings and Contribution in Number of Containers in each Scenario

The re-utilized containers belong to the laden import containers which are stripped
and then available at the inland terminal to be used again as an empty export
container. If there is no demand it is transported back to the Seaport or for
triangulation. With the Seaport as the source, fresh containers are requested to
fulfil the demand. If an inland terminal receives containers from another inland

29



4. Results and Analysis

terminal instead of the Seaport, triangulation is possible in that instance and this
contributes to the share of re-utilized empty containers.

(a) No Collaboration (b) Alliance Collaboration

(c) Complete Collaboration

Figure 4.3: Source of Empty Containers

4.2 Base Case
Base Case represents the current cost incurred by the Seaport to transport all
containers. The transport includes flows from the Seaport to inland terminals and
vice versa. The table represents the break-up of the costs for different flows. The
total cost is a result of Eq 3.1.

Container Type Cost (million cont-km.)
Laden Import 7.03
Empty Import 7.03
Empty Export 10.5
Laden Export 10.5

Total 35.06

Table 4.5: Cost Break-Up per Container Flow Type

The cost calculated is for the total containers mentioned in Table4.1. This scenario
is only a representation of the current network that is in use. There are no provisions

30



4. Results and Analysis

for flows between inland terminals and for reusing empty import containers. This
results in the same costs for flows to and from the Seaport.

Figure 4.4: Base Case - Flows of Empty Export Containers

Figure 4.5: Base Case - Share of Empty Export Containers

31



4. Results and Analysis

4.3 No Collaboration Scenario

Similar to the Base Case with some variations to incorporate triangulation.
Containers are distinguished based on ownership. There is a reduction in the total
number of containers used as the total cost reduces due to triangulation between
inland terminals. The total cost is a result of Eq 3.1.

Container Type Cost (million cont-km.)
Laden Import 7.03
Empty Import 4.53
Empty Export 8.39
Laden Export 10.5

Total 30.45

Table 4.6: Cost Break-Up per Container Flow Type

Compared to the base case only 62% of the containers need to be requested from the
Seaport since 20.4% can be solved through re-utilization within the same terminal
and 17.5% through triangulation from another terminal. This amounts to a 38%
reduction in fresh containers from the Port required to satisfy the demand.

Figure 4.6: No Collaboration - Flows of Empty Export Containers

32



4. Results and Analysis

Figure 4.7: No Collaboration - Share of Empty Export Containers

4.4 Alliance Collaboration Scenario

The alliances existing in the maritime industry do not allow container sharing but
focus on vessel fill rate, wherein the partners in the alliance can load their containers
on partner vessels. The alliance in our modelling is an extended version of the
same but also extends to container sharing along with vessel sharing. Theoretically,
this reduces the number of constraints and results in a lower cost which is seen
mathematically as well. The total cost is a result of Eq 3.1.

Container Type Cost (million cont-km.)
Laden Import 7.03
Empty Import 3.71
Empty Export 7.6
Laden Export 10.5

Total 28.84

Table 4.7: Cost Break-Up per Container Flow Type

Compared to the base case only 49.3% of the containers need to be requested from
the Seaport since 28.8% can be solved through re-utilization within the same
terminal and 21.8% through triangulation from another terminal. This amounts to
a 50.6% reduction in fresh containers from the Port required to satisfy the demand.

33



4. Results and Analysis

Figure 4.8: Alliance Collaboration - Flows of Empty Export Containers

Figure 4.9: Alliance Collaboration - Share of Empty Export Containers

34



4. Results and Analysis

4.5 Complete Collaboration Scenario

Modelled as an ideal scenario where all shipping lines willingly let go of container
ownership and matching. There is no constraint for containers to belong to the
same shipping line which enables complete freedom to be transported anywhere in
the network. This also reduces the total number of containers. As expected from
an ideal scenario, the solution from this scenario is the lowest and the best out of
others. The total cost is a result of Eq 3.1.

Container Type Cost (million cont-km.)
Laden Import 7.03
Empty Import 2.5
Empty Export 6.35
Laden Export 10.5

Total 26.38

Table 4.8: Cost Break-Up per Container Flow Type

Compared to the base case only 32.6% of the containers need to be requested from
the Seaport since 40.7% can be solved through re-utilization within the same
terminal and 25.6% through triangulation from another terminal. This amounts to
a 66.3% reduction in fresh containers from the Port required to satisfy the demand.

Figure 4.10: Complete Collaboration - Flows of Empty Export Containers

35



4. Results and Analysis

Figure 4.11: Complete Collaboration - Share of Empty Export Containers

4.6 Savings from Scenarios
The savings in container-kms are shown in Table 4.2. The savings in energy and
emissions are substantial from the network perspective, shown in the table below.
The energy and emission savings are low as the mode of transport is rail and the
emissions from rail transport are lowest due to the source of the electricity, as
mentioned in Section.3.5.

Scenario
Cont-km Energy Emissions Energy &

(in Savings (million (ton. Emission
millions) (%) kWh) CO2e) Savings (%)

Base 35.06 - 42.7 480 409.65 -Case
No 30.45 13% 41.32 464 850.9 3.2%Collaboration

Alliance 28.84 18% 40.84 459 417.15 4.4%Collaboration
Complete 26.38 25% 40.1 451 114.65 6.1%Collaboration

Table 4.9: Energy Consumption and Emission Savings

36



4. Results and Analysis

4.7 Inland Terminal Selection based on
Re-Utilization

4.7.1 Asymmetry and Container Imbalance Ratio (CIR)
To get the individual values for which feed into CIR, each inland terminal’s total
import and export was calculated and has been plotted as a butterfly chart in
Fig.4.12.

Figure 4.12: Import - Export Asymmetry

Figure 4.13: Container Imbalance Ratio (CIR)

While Fig.4.13 depicts the imbalance ratio for each inland terminal. There are
5 terminals that are import dominant and 6 terminals that are export dominant.
Terminal H, Terminal J and Terminal Q are entirely export-dominant terminals with
no incoming flow of import containers. Terminals with only export containers have
a CIR of -1.0. The CIR is directly dependent on the number of containers. It acts as
the first step in identifying the top inland terminals which have high imports, which
eventually result in high % re-utilization. The re-utilization strategy is analyzed
only for the No Collaboration scenario.

37



4. Results and Analysis

In
la

nd
T

er
m

in
al

C
IR

%
R

e-
ut

i.
B

as
e

C
as

e
N

o
C

ol
la

bo
ra

ti
on

R
eu

se
C

on
tr

ib
ut

io
n

&
%

Sa
vi

ng
s

T
ot

al
T

ot
al

T
ot

al
C

on
t-

km
E

ne
rg

y
E

m
is

si
on

C
on

t-
km

E
ne

rg
y

E
m

is
si

on
Sa

vi
ng

s
Sa

vi
ng

s
Sa

vi
ng

s
(i

n
10

00
s)

(1
00

0
kW

h)
(t

on
.

C
O

2e
)

(i
n

10
00

s)
(1

00
0

kW
h)

(k
g.

C
O

2e
)

Te
rm

in
al

A
0.

46
53

%
33

4
10

0
1.

13
17

8
53

60
2

Te
rm

in
al

B
0.

25
48

%
14

73
44

2
4.

97
70

8
21

2
23

90
Te

rm
in

al
C

0.
48

42
%

48
3

14
3

1.
63

20
4

61
68

9
Te

rm
in

al
D

-0
.5

6
2%

26
23

78
7

8.
85

60
18

20
3

Te
rm

in
al

E
-0

.7
8

5%
14

84
44

5
5.

01
78

24
26

5
Te

rm
in

al
F

0.
48

49
%

24
1

72
0.

81
11

7
35

39
6

Te
rm

in
al

G
0.

22
17

%
39

8
11

9
1.

34
68

20
23

0
Te

rm
in

al
H

-1
.0

-
11

87
35

6
4.

01
-

-
-

Te
rm

in
al

I
-0

.9
1%

19
59

58
8

6.
61

15
5

52
Te

rm
in

al
J

-1
.0

-
14

6
45

0.
5

-
-

-
Te

rm
in

al
Q

-1
.0

-
23

6
71

0.
8

-
-

-

T
ab

le
4.

10
:

Pe
rc

en
ta

ge
R

e-
ut

ili
za

tio
n

an
d

Sa
vi

ng
s

du
e

to
R

e-
ut

ili
za

tio
n

at
Ea

ch
In

la
nd

Te
rm

in
al

38



4. Results and Analysis

4.8 Triangle Selection

In the base case, we looked at the whole sending system between many inland
terminals. Here it is investigated if flows would be established only between
specific terminals. The selection of triangles is based on the criteria mentioned in
Section.3.4.1. The top 9 links with the maximum containers transported are
selected for the triangles as quantities lower than those are unfavourable for
triangulation. Quantities over 250 containers for the entire duration of 6 months
are considered as high. The most frequent flows that result in the transport of
significant quantities over the time period are selected. From Fig.4.6, the inland
terminals that have the highest shares from Fig.4.7 are considered. The flows
originating from the Seaport (the Seaport) are disregarded during the selection of
triangles.
The top nine triangles selected for triangulation are:

• Seaport - Terminal A - Terminal B
• Seaport - Terminal A - Terminal E
• Seaport - Terminal A - Terminal F
• Seaport - Terminal A - Terminal I
• Seaport - Terminal B - Terminal I
• Seaport - Terminal B - Terminal H
• Seaport - Terminal B - Terminal D
• Seaport - Terminal F - Terminal D
• Seaport - Terminal C - Terminal G

The comparisons between the triangles and the different scenarios are presented in
a tabular format below. The percentage savings provide insight into the potential
triangles where the new changes can be implemented. Savings only from
triangulation are shown in the table and not from re-utilization. A deep dive into
the data based on shipping lines will help with decision-making related to
individual shipping lines and optimization within their operations. If a shipping
line has a large market in a country, it can implement triangulation in an
individualistic manner. Individual triangles are visualized on the map in Fig4.15.
The thickness of the lines indicates the quantity. The names of nodes are denoted
by the starting letters of their names. The key for the table is presented below:

P - Seaport, TA - Terminal A, TB - Terminal B, TC - Terminal C, TD - Terminal
D, TE - Terminal E, TF - Terminal F, TG - Terminal G and TI - Terminal I.

Q is the triangle distance factor. TIR is the triangle imbalance ratio.

Column ’Containers’ shows the total number of empty containers that are
transported between the three nodes in a triangle. Energy consumption and
emissions are calculated as shown in Section.3.5.

39



4. Results and Analysis

Triangles Shipping Lines Container Share
SL 1 SL 2 SL 3 SL 4 SL 5 SL 6 SL 7 SL 8 SL 9 SL 10 SL 11

P - TC - TG - 1% 5% 1% 4% - 55% 31% 3% - -
P - TA - TF 1% 4% 2% 4% 24% 1% 18% 39% 6% 1% -
P - TB - TH - - - 2% - - 98% - - - -
P - TB - TD - 1% - 3% 4% - 85% 5% 2% - -
P - TF - TD - - - 5% - 4% 62% 25% 3% - -
P - TA - TB - 5% 3% - 25% - 30% 28% 1% 8% -
P - TA - TE - 16% - - 16% - 12% 53% 3% - -
P - TB - TI - - 1% 23% 6% - 2% 63% 5% - -
P - TA - TI - 2% - 1% 19% - - 77% 1% - -

Table 4.11: % Share of Containers for Each Shipping Line in Triangulation

Table 4.11 represents each shipping line and its share of containers transported
within the triangle. It is evident that there are certain shipping lines stand out with
their quantities. The approach moving forward for the shipping lines should be to
realize the potential for savings within the suggested triangles and further work on
implementing the strategies. The imbalance in triangles is shown in Fig.4.14. It
presents the overall imbalance when both the inland terminals are considered.

Figure 4.14: Triangle Imbalance Ratio (TIR) for Selected Triangles

40



4. Results and Analysis

F
ig

ur
e

4.
15

:
Se

le
ct

ed
Tr

ia
ng

le
s

41



4. Results and Analysis

T
ri

an
gl

es
Q

T
IR

B
as

e
C

as
e

N
o

C
ol

la
bo

ra
ti

on
Sa

vi
ng

s
C

on
ta

in
er

En
er

gy
Em

iss
io

ns
C

on
ta

in
er

En
er

gy
Em

iss
io

ns
En

er
gy

&
C

on
ta

in
er

A
bs

ol
ut

eA
bs

ol
ut

e
A

bs
ol

ut
e

-k
m

(1
00

0
(t

on
ne

s
-k

m
(1

00
0

(t
on

ne
s

Em
iss

io
ns

Sa
vi

ng
s

co
nt

-k
m

En
er

gy
Em

iss
io

ns
(in

10
00

s)
kW

h)
C

O
2e

)
(in

10
00

s)
kW

h)
C

O
2e

)
Sa

vi
ng

s
Sa

vi
ng

s
Sa

vi
ng

s
Sa

vi
ng

s
P

-T
C

-T
G

0.
09

0.
38

88
0

26
4

2.
97

37
0

11
1

1.
25

58
%

31
%

51
0

15
3

1.
72

P
-T

A
-T

F
0.

23
0.

47
57

4
17

2
1.

94
21

8
65

0.
74

62
%

52
%

35
6

10
6

1.
2

P
-T

B
-T

H
0.

24
0.

01
3

2,
66

0
79

8
8.

98
1,

72
0

51
6

5.
81

35
%

30
%

93
9

28
1

3.
17

P
-T

B
-T

D
0.

25
-0

.0
8

4,
09

6
1,

22
8

13
.8

3
2,

84
5

85
3

9.
6

31
%

21
%

1,
25

0
37

5
4.

22
P

-T
F

-T
D

0.
33

-0
.1

2
2,

86
4

85
9

9.
67

2,
15

3
64

6
7.

27
25

%
11

%
71

0
21

3
2.

4
P

-T
A

-T
B

0.
53

0.
35

1,
80

6
54

1
6.

1
81

0
24

3
2.

73
55

%
50

%
99

6
29

8
3.

36
P

-T
A

-T
E

0.
57

-0
.0

2
1,

81
7

54
5

6.
13

1,
49

0
44

7
5.

03
18

%
21

%
32

7
98

1.
1

P
-T

B
-T

I
0.

62
0.

1
3,

42
0

1,
02

6
11

.5
5

2,
24

7
67

4
7.

59
34

%
35

%
1,

17
3

35
2

3.
96

P
-T

A
-T

I
0.

84
0.

29
2,

29
2

68
7

7.
74

1,
79

7
53

9
6.

07
22

%
36

%
49

4
14

8
1.

67
T

ri
an

gl
es

Q
T

IR
B

as
e

C
as

e
A

lli
an

ce
C

ol
la

bo
ra

ti
on

Sa
vi

ng
s

P
-T

C
-T

G
0.

09
0.

38
88

0
26

4
2.

97
22

4
67

0.
76

75
%

56
%

65
6

19
6

2.
22

P
-T

A
-T

F
0.

23
0.

47
57

4
17

2
1.

94
40

12
0.

14
93

%
66

%
53

3
16

0
1.

8
P

-T
B

-T
H

0.
24

0.
01

3
2,

66
0

79
8

8.
98

1,
36

1
40

8
4.

6
49

%
37

%
1,

29
8

38
9

4.
38

P
-T

B
-T

D
0.

25
-0

.0
8

4,
09

6
1,

22
8

13
.8

3
2,

20
8

66
2

7.
45

46
%

28
%

1,
88

8
56

6
6.

37
P

-T
F

-T
D

0.
33

-0
.1

2
2,

86
4

85
9

9.
67

1,
73

0
51

9
5.

84
40

%
17

%
1,

13
4

34
0

3.
83

P
-T

A
-T

B
0.

53
0.

35
1,

80
6

54
1

6.
1

58
0

17
4

1.
96

68
%

63
%

1,
22

5
36

7
4.

14
P

-T
A

-T
E

0.
57

-0
.0

2
1,

81
7

54
5

6.
13

1,
41

6
42

4
4.

78
22

%
26

%
40

1
12

0
1.

35
P

-T
B

-T
I

0.
62

0.
1

3,
42

0
1,

02
6

11
.5

5
1,

89
3

56
8

6.
39

45
%

44
%

1,
52

7
45

8
5.

16
P

-T
A

-T
I

0.
84

0.
29

2,
29

2
68

7
7.

74
1,

57
5

47
2

5.
32

31
%

45
%

71
6

21
5

2.
42

T
ab

le
4.

12
:

C
om

pa
ris

on
of

Sa
vi

ng
s

in
Tr

ia
ng

le
s

42



4. Results and Analysis

T
ri

an
gl

es
Q

T
IR

B
as

e
C

as
e

C
om

pl
et

e
C

ol
la

bo
ra

ti
on

Sa
vi

ng
s

C
on

ta
in

er
En

er
gy

Em
iss

io
ns

C
on

ta
in

er
En

er
gy

Em
iss

io
ns

En
er

gy
&

C
on

ta
in

er
A

bs
ol

ut
eA

bs
ol

ut
e

A
bs

ol
ut

e
-k

m
(1

00
0

(t
on

ne
s

-k
m

(1
00

0
(t

on
ne

s
Em

iss
io

ns
Sa

vi
ng

s
co

nt
-k

m
En

er
gy

Em
iss

io
ns

(in
10

00
s)

kW
h)

C
O

2e
)

(in
10

00
s)

kW
h)

C
O

2e
)

Sa
vi

ng
s

Sa
vi

ng
s

Sa
vi

ng
s

Sa
vi

ng
s

P
-T

C
-T

G
0.

09
0.

38
88

0
26

4
2.

97
89

26
.7

8
0.

3
90

%
78

%
79

1
23

7
2.

67
P

-T
A

-T
F

0.
23

0.
47

57
4

17
2

1.
94

33
5

10
0

1.
13

42
%

84
%

23
9

71
0.

81
P

-T
B

-T
H

0.
24

0.
01

3
2,

66
0

79
8

8.
98

1,
01

0
30

3
3.

41
62

%
48

%
1,

64
9

49
4

5.
57

P
-T

B
-T

D
0.

25
-0

.0
8

4,
09

6
1,

22
8

13
.8

3
1,

29
2

38
7

4.
36

68
%

45
%

2,
80

4
84

1
9.

46
P

-T
F

-T
D

0.
33

-0
.1

2
2,

86
4

85
9

9.
67

1,
17

5
35

2
3.

97
59

%
34

%
1,

68
8

50
6

5.
7

P
-T

A
-T

B
0.

53
0.

35
1,

80
6

54
1

6.
1

58
3

17
5

1.
97

68
%

80
%

1,
22

3
36

6
4.

13
P

-T
A

-T
E

0.
57

-0
.0

2
1,

81
7

54
5

6.
13

1,
31

5
39

4
4.

44
28

%
33

%
50

2
15

0
1.

69
P

-T
B

-T
I

0.
62

0.
1

3,
42

0
1,

02
6

11
.5

5
1,

36
0

40
8

4.
59

60
%

57
%

2,
06

0
61

8
6.

95
P

-T
A

-T
I

0.
84

0.
29

2,
29

2
68

7
7.

74
80

7
24

2
2.

72
65

%
57

%
1,

48
5

44
5

5.
01

T
ab

le
4.

13
:

C
om

pa
ris

on
of

Sa
vi

ng
s

in
Tr

ia
ng

le
s

(c
on

tin
ue

d)

43



4. Results and Analysis

44



5
Discussion

The results will be discussed and the final triangles and the top inland terminals
with re-utilization will be decided in this chapter. The Complete Collaboration
provides a theoretical maximum for each strategy. However, its feasibility is
limited based on the discussion with the different stakeholders and the Seaport.
Thus, this section will focus on the No Collaboration scenario and its potential
benefits which are more likely to be realized. The No Collaboration scenario is
closest to the current conditions. The Alliance Collaboration scenario is under the
assumption that the shipping lines that are in alliance now for improving vessel fill
rate would also agree to share containers within the alliance.

The comparison between the selected triangles with data about the different
indicators and performance metrics is discussed below. The utilization of
containers has significantly increased across all scenarios. The No Collaboration
scenario has a 36.2% utilization. The Alliance Collaboration scenario increases
utilization to 48.3% and further to 64.1% with Complete Collaboration. The
reduction in the number of containers leads to fewer lifts and reduced container
tracking data to handle. The container request is reduced by 37.9%, 50.6% and
66.3% for the No Collaboration, Alliance Collaboration and Complete
Collaboration scenarios respectively. From Table.4.12, the analysis will focus on
the No Collaboration scenario. This will be the most intuitive for decision-making
as the No Collaboration scenario is the most realistic representation of the existing
conditions. The optimization and analysis have led to an interesting crossroads
where two ECR reduction solutions are possible. One is triangulation which still
involves ECR but with high savings as compared to Base Case operations. The
other solution is to re-utilize. Re-utilizing available empty import containers at the
same inland terminal looks highly promising and has the potential for high
savings. When reading ahead, please remember that the percentages are only
relative. A higher percentage does not mean that the inland terminal or the
triangle is better. The actual number of containers each terminal deals with is
sensitive data and thus cannot be disclosed here. However, they are taken into
consideration during the selection process.

5.1 Re-utilization of Containers
A surprising outcome of optimizing container flows showed the possibility of reusing
containers at the same inland terminal that received them from the Seaport as a

45



5. Discussion

laden import container. This results in a reduction in empty container requests
sent from inland terminals to the Seaport. The re-utilization is also discussed in
Tegbrant and Karlander (2023). The container imbalance ratio (CIR) is calculated
directly based on the number of containers. In this case, a high positive ratio has
greater potential for re-utilization.

5.1.1 Top 3 Inland Terminals
Based on the No Collaboration scenario the highest re-utilization of containers
without any additional ECR of any kind, the top three inland terminals are listed
below in the lowest to highest % re-utilization quantity order.

• Terminal B: 48%
• Terminal F: 49%
• Terminal A: 53%

The re-utilization approach is more alluring as it requires little to no investments
from the actors involved. It is instead a greater focus on optimizing the existing
transport and operations. The empty import containers present at an inland
terminal before being returned to the Seaport are considered to satisfy a part of
the demand in this approach. There is no repositioning of the containers except for
moving them around within the inland terminal as per convenience. There will be
instances when the containers cannot be matched for re-utilization. The next best
option is to triangulate the containers with another inland terminal. This will
reduce the distance a new container travels from the Seaport and reduce the time
to satisfy the demand at the inland terminal. In the Base Case, all containers were
requested from the Seaport. Table 5.1 represents the savings achieved due to
re-utilization. There are other inland terminals in Table 4.10 with higher savings
in container-km and energy, however, the criteria for selection is the number of
containers re-utilized and not based on other performance indicators of the inland
terminal itself.

Inland Terminal Cont-Km Savings Energy Savings Emission Savings
(in 1000s) (1000 kWh) (kg. CO2e)

Terminal B 708 212 2390
Terminal F 117 35 396
Terminal A 178 53 602

Table 5.1: Re-utilization Savings

5.2 Triangulation of Containers
The selection of the top 3 triangles is dependent on numerous factors as discussed
above. However, the results of our study and information from the Seaport have
led us to fixate on two most important factors for the decision makers; the total
number of containers in triangulation flows and the total container-kms, the energy

46



5. Discussion

& emissions savings. We discuss the two different perspectives when each factor is
given priority individually.

5.2.1 Containers in Triangulation Flows
Focusing mainly on the containers in triangulation flows from Table 4.12,4.13, the
top three triangles are:

1. Seaport - Terminal B - Terminal D
2. Seaport - Terminal A - Terminal E
3. Seaport - Terminal C - Terminal G

Based entirely on the containers re-positioned through triangulation, the three
triangles have significant quantities. The table below represents the total %
contribution from triangulation. The number of containers triangulated in each
triangle is ≈ 1000.

Triangle Q TIR % Containers Contributed
Triangulation Re-utilization From Port

Seaport - TB - TD 0.25 -0.08 12% 21% 67%
Seaport - TA - TE 0.57 -0.02 13% 21% 66%
Seaport - TC - TG 0.09 0.38 23% 31% 46%

Table 5.2: Percentage Triangulation within a Triangle

Figure 5.1: Container Triangulation Over Time (in Months)

47



5. Discussion

Triangles 1 and 2 are the triangles having the highest total demand for empty
containers. Though they have a negative TIR showing that they are slightly export
dominant, the 2 triangles have the highest number of containers in triangulation
flows. In contrast, Triangle 3 has the lowest demand for empty containers while
having the 3rd highest number of triangulated containers. The underlying reason
is the lowest value of Q. Due to the short distance between the inland terminals,
though having low demand, the possibility of triangulation is quite high. This is
also supported by the high TIR, showing that Triangle 3 is import-dominant and
has several containers which can be triangulated easily without having to request
fresh ones from the Seaport. Fresh containers from the Seaport would travel 3-4
times more distance to satisfy the demand. The direction of flows in the triangles
will aid with fixing the direction of trains as seen in Fig.5.1. In Triangle 1, there
are 93.8% of container flows from TB to TD and 6.2% of container flows from TD
to TB. In Triangle 2, 100% of container flows are from TA to TE. In Triangle 3,
64.3% of containers flow from TC to TG and 35.7% of containers flow from TG to
TC. Considering the information on the direction of flows and as all triangles have
different terminals, all three triangles can be implemented simultaneously. There
will be no cannibalization of containers. Furthermore, Fig.5.1 shows the distribution
of containers over time for each month. Such distribution charts will help accurately
plan train timetables and the number of trains required within each triangle.

5.2.2 Container-kms, Energy & Emissions Savings
Giving priority to container-kms, energy and emissions savings, from Table4.12,4.13,
the top three triangles are:

1. Seaport - Terminal B - Terminal D
2. Seaport - Terminal B - Terminal I
3. Seaport - Terminal A - Terminal B

Based only on container-kms the three triangles have the absolute maximum savings
in distance travelled. This also translates to savings in energy consumption and
emissions. This approach also reduces the time factor linked while transporting
over longer distances. Table 5.3 represents the absolute values and % savings.

Triangles

Absolute Absolute Absolute %
Cont-kms Energy Energy Savings
Savings (1000 kWh) Savings

(1000 kms) (ton.CO2e)
Seaport - TB - TD 1251 375 4.22 31%
Seaport - TB - TI 1174 352 3.96 34%
Seaport - TA - TB 996 299 3.36 55%

Table 5.3: Savings in Container-kms, Energy and Emissions

Terminal B is repeated in all three selections. With this approach, the possibility
of implementing triangles 1 and 2 simultaneously will lead to the cannibalization of
containers. The flows will try to satisfy demand from the same pool of containers

48



5. Discussion

which will result in lower savings. In Triangle 1, there are 93.8% of container flows
from TB to TD and 6.2% of container flows from TD to TB. In Triangle 2, 100% of
container flows are from TB to TI. In Triangle 3, 100% of container flows are from
TA to TB. Having the details of direction and quantity inflows, it can be concluded
that Triangles 1 and 3 can be implemented together as there are no conflicts. It
should be noted that the selection is entirely based on factors and has no relation
to the quantities of containers it handles.

Figure 5.2: Container Distribution Over Time (in Months)

5.3 General Discussions
The above selections are entirely based on factors and criteria. This answers both
research questions. However, implementing such changes is not limited to a
quantitative approach. A mix of policies, collaborations between shipping lines,
technology, strategic planning and operational excellence can make the
implementation smoother and more efficient. As the models have been named on
the interaction of different actors in the supply chain, trust and openness in the
system is the sine qua non for the maritime industry. In a list of numerous factors,
the inertia around data sharing can be one of the top concerns hindering the
potential for savings. A previous thesis, Hao and Ibrahim (2023) discuss the
economic and environmental effects of ECR reduction from the perspective of an
inland terminal and its flows to customers. This focuses on the last segment of the
value chain. A more detailed analysis from a recent study focusing on the last
segment, using data analytics and a take on urban logistical problems (Castrellon,

49



5. Discussion

Sanchez-Diaz, & Kalahasthi, 2023). Tegbrant and Karlander (2023) and Hellekant
and Rudal (2021) discuss the improvements in transport efficiency and different
ECR reduction strategies that can be implemented across the board by one or
several actors involved in the supply chain. Hagelin and Knutsson (2020)
investigate the drivers and barriers in ECR reduction.

5.3.1 Global Implementation and Disruptions
The effects of disruptions in the supply chain were not considered during the
modelling of the scenarios. The time period of the data in this study is at a state
when the pandemic was over but the invasion of Ukraine had started. However,
the data does not reflect these disruptions in any manner and neither have we
added any constraints to recreate disruptions in the model. The disruptions affect
the flows in the waterways and indirectly affect the hinterland flows. The
optimization is focused on the hinterland flows and thus disruptions on a global
scale are not considered. Similarly, the strategies are applicable in all re