Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Journal Pre-Proof: European Journal of Operational Research

Download as pdf or txt
Download as pdf or txt
You are on page 1of 33

Journal Pre-proof

Identifying the market areas of port-centric logistics and hinterland


intermodal transportation

Yann Bouchery , Johan Woxenius , Jan C. Fransoo

PII: S0377-2217(20)30133-8
DOI: https://doi.org/10.1016/j.ejor.2020.02.015
Reference: EOR 16335

To appear in: European Journal of Operational Research

Received date: 8 July 2019


Accepted date: 10 February 2020

Please cite this article as: Yann Bouchery , Johan Woxenius , Jan C. Fransoo , Identifying the mar-
ket areas of port-centric logistics and hinterland intermodal transportation, European Journal of Oper-
ational Research (2020), doi: https://doi.org/10.1016/j.ejor.2020.02.015

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier B.V.


Highlights

 We analyze port-centric logistics and hinterland intermodal transportation.


 We develop new analytical results identifying the optimal market areas.
 Tension between these options is quite likely to happen.

* Corresponding author: Yann Bouchery, yann.bouchery@kedgebs.com, Tel: +33 5 56 84 55 37


Identifying the market areas of port-centric
logistics and hinterland intermodal transportation
Yann BOUCHERY*(1), Johan WOXENIUS (2), Jan C. FRANSOO (3)

(1) The Centre of Excellence in Supply Chain (CESIT), KEDGE Business School
680 cours de la Libération, 33400 Talence, France

(2) Department of Business Administration, University of Gothenburg


Box 610, SE-405 30 Gothenburg, Sweden

(3) Kuehne Logistics University,


Grosser Grasbrook 17, 20457 Hamburg, Germany

December 18, 2019


Abstract: Many port authorities have developed ambitious strategies to foster hinterland intermodal
transportation. In addition, port-centric logistics, that is, the provision of distribution facilities and value-
adding activities in the port area, has expanded in multiple ports. Obviously, such port-centric logistics may
impact the operations in the hinterland substantially and could potentially reduce opportunities for intermodal
transport in the hinterland. We analyze the interaction between port-centric logistics and hinterland intermodal
transportation. We take a logistics service provider’s perspective and we include some key elements in the
model, such as detention fees, extra handling, transport efficiency and empty container repositioning. We
develop new analytical results identifying the optimal market areas of truck-only transportation, port-centric
logistics and hinterland intermodal transportation. Our results show that tension between port-centric logistics
and hinterland intermodal transportation is quite likely to happen in practice. We additionally study the use of
continental containers as a way to reconcile port-centric logistics and hinterland intermodal transportation and
we derive further results. We illustrate our results via an example and we highlight managerial insights.

Keywords: logistics, hinterland intermodal transportation, port-centric operations, market area theory.

2
Introduction
Maritime containers have become the backbone of globalization by providing an entirely standardized
transportation system to connect nonadjacent economic areas, so that container-based trade now represents
more than half of the value of all seaborne trade (UNCTAD, 2015). The vast majority of the distance covered
by a maritime container during its journey occurs at sea. However, the costs related to maritime shipping have
drastically decreased for three main reasons. First, the ships built in 2012–2015 are on average 2.8 times larger
than the ones built 20 years earlier (UNCTAD, 2016). This enables economies of scale and energy efficiency.
Second, alliances (see e.g., Slack, Comtois, & McCalla, 2002; Song & Panayides, 2002; Agarwal & Ergun,
2010) and advances in network design (see e.g., Meng, Wang, Andersson, & Thun, 2013; Zheng, Meng, &
Sun, 2015; Tran & Haasis, 2015) have increased the efficiency of shipping. Third, bunker costs have been
significantly decreased by applying slow-steaming (Finnsgård, Kalantari, Roso, & Woxenius, 2019), but also
as a result of lower oil prices during the past few years. Consequently, the costs of inland-related activities
often exceed the costs of maritime transportation nowadays. For instance, Notteboom and Rodrigue (2005)
state that the proportion of inland costs relative to the total transportation costs of containers ranges from 40%
to 80% and with larger vessels the proportion is likely to be even higher now. We focus on the hinterland of a
deep-sea port, as this is where much of the competitiveness of global supply chains currently lies.
The main advantage of containerization in the hinterland is the ability for containers to be easily
transshipped from one transport mode to another, fostering the development of intermodal transportation.
Hinterland intermodal transportation (HIT) reduces congestion and accidents in the surroundings of deep-sea
ports and is recognized for its superior environmental performance (Roso, 2007; Dekker, Bloemhof, &
Mallidis, 2012; Craig, Blanco, & Sheffi, 2013; Bouchery & Fransoo, 2015). HIT also leads to an increase in
competition between ports from the same region (van Klink & van den Berg, 1998; Fan et al., 2015;
Vermeiren & Macharis, 2016) as distant locations can be served more efficiently. The main container ports
worldwide have developed ambitious strategies to foster high-capacity corridors that enable moving
containers from the port to regional load centers and vice versa (van den Berg, 2015). Accordingly, HIT has
dramatically affected the development of the hinterland in many regions, so that the rise of inland terminals,
dry ports (Roso, Woxenius, & Lumsden, 2009), extended gates (Veenstra, Zuidwijk, & van Asperen, 2012)
and intermodal hubs—referred to as the ―terminalization‖ of the hinterland—is considered to be a new phase
in port development (Rodrigue & Notteboom, 2009). (See Nguyen and Notteboom, 2019, and Witte,
Wiegmans, & Ng, 2019, for recent overviews of the related literature).
Another notable trend in port development relates to the concept of port-centric logistics (PCL). PCL
is defined by Mangan, Malwani, and Fynes (2008) as the provision of distribution facilities and value-adding
activities in the port area. Two main types of port-centric activities can be distinguished. First, shippers can
locate a full-service distribution center in the port surroundings. For instance, Heitz, Dablanc, Olsson,
Sánchez-Diaz, and Woxenius (2018) identifies a trend that warehouses are increasingly located either in close
proximity to the Port of Gothenburg or in cities about 100 kms from the port, whereas few new warehouses
are being built in the area in between. Heitz (2019) finds a similar duality between city-center and peripheral

3
distribution warehouse locations in the Paris metropolitan area. In PCL, the maritime container is generally
transferred from the deep-sea terminal to the distribution center directly by truck due to the short distance.
Although some full-service distribution centers can be found in the immediate vicinity of a port, this
represents only a small share of port-centric activities nowadays (Monios, Notteboom, Wilmsmeier, &
Rodrigue, 2016). There are three main reasons for this limited presence. First, the center of gravity of the
demand is often located farther away in the hinterland and locating distribution centers closer to the demand’s
center of gravity enables the reduction of transportation costs (see e.g. Ambrosino & Scutella, 2005, for a
review of the literature on distribution network design). Second, locating distribution centers in the hinterland
enables shippers to benefit from port competition and limit the risk of being anchored to a single port (Monios
& Wilmsmeier, 2013). Third, shippers usually combine international and local flows in their distribution
centers and PCL is not necessarily efficient for local flows if a port is not close to the center of gravity.
Hence, in this paper we focus on another type of port-centric activities. This latter includes cargo
transloading to semi-trailers or continental containers and vice versa (also referred to as stuffing and
stripping), palletization, cargo consolidation and sorting as well as light transformation and customization
(such as labeling). These activities enable optimizing hinterland transportation without affecting the location
of the shippers’ distribution centers. This type of port-centric activity has expanded in the surroundings of
various ports such as Savannah (Rodrigue, Debrie, Fremont, & Gouvernal, 2010), Los Angeles/Long Beach
(Rodrigue & Notteboom, 2009), Antwerp (Notteboom, 2016) and Felixstowe (Mangan et al., 2008). These
activities in the port induce additional handling and associated costs. However, the advantages of PCL can be
significant for the shippers. For instance, Leachman (2008) states that sorting activities in the Los
Angeles/Long Beach perimeter induces a net savings to the American economy of about $1.1 billion per year
due to inventory reduction obtained from cargo rebalancing between distribution centers, based on updated
information. (See Mangan et al., 2008; Pettit & Beresford, 2009; Monios & Wilmsmeier, 2012; Monios et al.,
2016; and Monios, Bergqvist, & Woxenius, 2018, for further details on port-centric activities.) This article
discusses only port-centric activities that do not affect the location of the shippers’ distribution centers.
PCL and HIT are rarely studied jointly in the literature. Therefore, the assessment of their potential
advantages and drawbacks is made in isolation. In this article, we aim at studying the relationship between
these two concepts. Indeed, many port authorities worldwide are trying to develop the two concepts in
synergy. For instance, the Port of Le Havre opened a new intermodal terminal in 2015 (HAROPA, 2015)
while developing in parallel facilities for PCL (Bolloré, 2015). Likewise, the Port of Gothenburg opened a
new intermodal terminal in 2017 while also investing considerably in warehouses in the port vicinity (Heitz et
al., 2018). Still, there are indications in the academic literature of a possible tension between PCL and HIT. In
Europe, for instance, when containers are unloaded in the vicinity of the port, the cargo is often transloaded
into semi-trailers. Woxenius and Bergqvist (2011) observe that rail is used to move semi-trailers to a lesser
extent than it is used for maritime containers. The same applies to barge transportation. As volume is a key
requirement for efficient use of HIT (Bouchery & Fransoo, 2015), Monios et al. (2016) conclude that PCL
may reduce the value of HIT, that is, that PCL might induce a reverse modal shift. This is in line with Monios
and Wilmsmeier (2012), who state that PCL and HIT are two competing logistics concepts. To the best of our
4
knowledge, the research that studies the relationship between PCL and HIT is mainly conceptual and there is a
need for model-based research on this topic. Such model-based strategic analyses of ocean container shipping
systems have become more common over the past few years following the position paper by Fransoo and Lee
(2013) but so far have not addressed the consequences of port-centric logistics.
In this article, we apply the market area theory (Erlenkotter, 1989) to identify the demand zones that
would be more efficiently served by HIT or PCL. We take a logistics service provider’s (LSP) perspective, as
Notteboom and Rodrigue (2005) state that the development of the hinterland mainly results from logistics
decisions from shippers and their contracted LSPs. Our model includes transport-related costs as well as travel
time (through its related value of time) as these two criteria are among the most relevant to model shippers’
decisions (Larranaga, Arellana, & Senna, 2017). We additionally study the effects of travel time reliability in
Appendix B. We include some key elements in the model, such as detention fees, extra handling, transport
efficiency and empty container repositioning. We analytically derive conditions for the use of direct truck
(DT) shipment, PCL and HIT and we identify the market areas of these options. Our results show that the
tension between PCL and HIT is quite likely to happen in practice. We derive managerial insights from the
results obtained, including an analysis of how different stakeholder groups are affected. We additionally study
the use of continental containers as a way to reconcile PCL and HIT, and we derive additional results.
The rest of the article is organized as follows. Section 2 highlights the related literature and helps at
better positioning our contribution. The model is presented in Section 3 and the main analytical results are
derived in Section 4. Section 5 aims at studying the impact of PCL on direct truck shipment and on HIT.
Moreover, we analyze the use of continental containers as a way to reconcile PCL and HIT. Section 6
provides an illustrative example that allows us to obtain some additional insights. Finally, Section 7 highlights
our main conclusions and some potential future research directions.

Literature review
In this section, we first provide a short overview of the empirical literature on shippers’ preferences of
transport options. Then, we present the main streams of model-based literature related to our work.
Several recent contributions aim at analyzing shippers’ preferences of transport modes. We focus here
on comparisons made for inland transportation where the main modes available are trucks, trains and barges.
As trains and barges rarely enable reaching the shipper’s location, this requires extra trucking for perhaps both
pre- and post-haulage. Therefore, train and barge are generally operated as part of an intermodal solution.
Most research comparing direct shipment by truck with intermodal transportation generally selects a number
of relevant criteria and then conducts stated-choice experiments through surveys. Recent contributions
(Nugroho, Whiteing, & de Jong, 2016; Larranaga et al., 2017; Duan, Tavasszy, & Rezaei, 2019) highlight that
the key decision variables affecting modal choice are transport costs, travel time and reliability of travel time.
These results are confirmed by other studies comparing inland transport options with short-sea shipping (see
Brooks, Puckett, Hensher, & Sammons, 2012; López-Navarro, 2014; Chang & Thai, 2017). Among these
three criteria, transport cost and travel time are well defined, while the reliability of travel time is more

5
difficult to apprehend. Therefore, our model accounts for both transport costs and travel time. We additionally
study the effect of the reliability of travel time as an extension in Appendix B. Note that Shams, Asgari, and
Jin (2017) specifically focus on this latter topic and help us better understand the value of reliability.
There are three main streams of model-based literature related to our work. First, the modal choice
decision is studied from a shipper’s perspective by jointly considering inventory decisions and transport mode
selection. This stream of literature can be traced back to Baumol and Vinod (1970), who included
transportation features in the economic order quantity model. The model they propose was then extended and
refined (see e.g., Liberatore, 1979; McGinninis, 1979; Sheffi, Eskandari, & Koutsopoulos, 1988; McGinnis,
1989; Tyworth, 1991; Tyworth & Zeng, 1998). (See Meixell & Norbis, 2008 for a detailed review.) In those
models, the lead time and lead-time variability that differ from one transportation option to another play a key
role by affecting the inventory levels of the shippers. More recently, Dullaert and Zamparini (2013) assess the
impact of lead-time variability on the traffic mode selection using simulation. They show that the
consequences of lead-time variability reduction depend heavily on the demand distribution structure and on
the targeted service level. This first stream of literature is very relevant and enables getting a good knowledge
of the dynamics between different transportation options from a shipper’s perspective. However, the models
do not take the location of the shippers directly into account and, therefore, the possible tension between PCL
and HIT is not easy to analyze using these models.
Second, modal choice is a key issue in the transportation literature. Modal choice is usually
considered as a tactical decision in transportation models. (See SteadieSeifi, Dellaert, Nuijten, Van Woensel,
& Raoufi, 2014, for a review of the operations research literature on intermodal transportation planning.)
According to SteadieSeifi et al. (2014), the decisions on intermodal transportation at the tactical level relate to
service network design (i.e., the planning of intermodal services, including traffic mode selection) and
network flow planning (i.e., the routing of the flow in the network, and consequently the selection of the
intermodal services used). (See Ypsilantis & Zuidwijk, 2013; Bouchery, Fazi, & Fransoo, 2015; Demir et al.,
2016; van Riessen, Negenborn, & Dekker, 2016; Wang & Meng, 2017, and references therein for recent
developments in this field.) Note that service network design and network flow planning are also studied for
the maritime leg of container transportation. (For this, see Fan, Wilson, & Dahl, 2015; Rothenbächer, Drexl,
& Irnich, 2016; Balakrishnan & Karsten, 2017). Network flow and service network design models generally
include discrete demand, where the objective is to serve a set of consignees considered as sinks in the
network. The models enable the selection of the best intermodal option for each consignee, but, as for
inventory-theoretic models for traffic mode selection, the tension between PCL and HIT is generally not
analyzed. Note that some models approximate the demand as continuous over a demand zone. This modeling
technique can be traced back to Newell (1973) and Daganzo (1991), who strongly contributed to a widespread
application. (See Langevin, Mbaraga, & Campbell, 1996, for a review of the early development in continuous
demand approximation models, and Dasci & Verter, 2001, and Tsao, Mangotra, Lu, & Dong, 2012, for some
more recent applications.) Note that continuous demand approximation is generally handled at the network
design level, since hub location decisions are considered. In our work, we focus on tactical decisions by
assuming that the terminals are already located and we aim at identifying the areas that are more efficiently
6
served by several transportation options, including PCL and HIT. This is a much more common decision
situation than locating new terminals since intermodal terminals are significant, long-lived investments and
transport demand reflects rather stable economic geography patterns. We are not aware of any article that
focuses on tactical decisions for HIT with continuous demand approximation.
The third stream of literature consists of applications of the market area theory (also called spatial
theory). The aim of this technique is to identify the size and topological feature of the demand zone that is
optimally served by a facility. To the best of our knowledge, Erlenkotter (1989) was the first author to connect
the continuous demand approximation models cited above to the market area theory. Erlenkotter (1989) also
offers a review of the early work on market area theory and proposes a typology to characterize general
optimal market area models. The key elements of this typology include the shape of the market, the distance
norm, the facility cost function and the transport cost function. Whereas the market area theory has become
standard in the facility location and hub-and-spoke network design literature (see e.g., Carlsson & Jia, 2013;
Averbakh, Berman, Kalcsics, & Krass, 2015; Sedghi. Shavandi, & Abouee-Mehrizi, 2017), applications to
intermodal transportation are less common. From our knowledge, Nierat (1997) presented the first application,
identifying the shape and size of the market area for an intermodal terminal. Nierat (1997) showed that under
the Euclidean distance norm and for strictly positive origin haulage distances (and therefore costs), the market
area of an intermodal terminal is part of the family of Descartes ovals. The author also compared the
theoretical results obtained to a survey about shippers’ choices for a number of existing terminals. The
theoretical results appeared to be consistent with shippers’ decisions. This conclusion strengthened the validity
of the market area theory in the context of intermodal transportation. The model developed by Nierat (1997)
was then refined by Janic (2008) to assess the impact of using longer trains for intermodal transportation in
Europe. Limbourg and Jourquin (2010) also propose an application to evaluate the market area of intermodal
rail–road container terminals in Europe. Kim and van Wee (2011) apply the model developed by Nierat
(1997) to identify the most influential factors affecting the break-even distance between truck-only and
intermodal transportation by using Monte Carlo simulation. Craig et al. (2013) use the model of Nierat (1997)
with carbon emissions parameters to assess the carbon-optimal market area of intermodal transportation.
Finally, Saeedi, Wiegmans, Behdani, and Zuidwijk (2017) propose a four-step methodology to evaluate the
European intermodal freight market. The methodology embeds the market area theory to identify the
transshipment submarket of each terminal.
Our contribution is threefold. First, this article is, to our knowledge, the first model-based contribution
about PCL. Instead of studying this concept in isolation, we focus on the relationship between PCL and HIT in
order to assess whether these two concepts are competing for the same market. Second, by applying the
market area theory, we develop new analytical results to identify the optimal market areas of truck-only
transportation, PCL and HIT. We identify necessary and sufficient conditions so that HIT is not impacted by
PCL and we show that those conditions do not hold very often in practice. We conclude that the competition
between PCL and HIT is likely to occur. Third, we evaluate if the use of continental containers might allow
for reconciling PCL and HIT. We illustrate our results with some examples and we derive additional insights.

7
Model description
We consider an LSP in charge of serving shippers (both consignees and consignors) located in the hinterland
of a single deep-sea port. Our objective is to assess under which conditions different transportation options
would be most favorable to deliver a given demand zone. Therefore, we express the costs and time for serving
the demand zone with one unit of flow, in accordance with the market area theory. This setting enables us to
derive insights into the geographic area of interest for different transportation options. Modeling transport
costs as linear in the number of containers handled is also in line with current practices as the transport rates
proposed to the shippers are generally expressed ―per container‖. We further discuss in Subsection 6.3 how
volume-dependent costs would affect our results. Let H be the area of operations of the LSP under
consideration. The space is orthonormal and equipped with an x-y coordinate system. The deep-sea port is
located at 0,0  . We assume that H is continuous and delimited by a straight coastline along the y-axis. Let

i  H be a location in H identified through its coordinates ( xi , yi ) . (See e.g. Wang, Meng, & Miao, 2016, for
more details on how to delimit the port’s hinterland.)
The maritime flows under consideration are containerized, which means the cargo reaches and leaves
the deep-sea port for the maritime leg of its transportation in maritime containers. These maritime containers
have standard dimensions, with the two most common types being 20 ft. and 40 ft. containers, of which the
latter is mostly used. We account for both types of containers and for both import and export flows. We
consider that one unit of demand equals one forty-foot equivalent unit (FEU). We consider a merchant
haulage setting, as this setting is particularly relevant for the operations of many large LSPs. Under merchant
haulage, the use of maritime containers in the hinterland is subject to detention fees. Shipping lines often
propose detention-free periods, so shippers or their providers are paying detention costs in case they are not
able to send the container back to the shipping line before the end of the detention-free period F  0
(expressed in days). Let us denote the detention fees by d  0 €/day.FEU (incurred after the detention-free
period). We also account for empty container movements. Let    0;1 be the empty container movement

ratio modeled as the proportion of full container movements. In the case of   0 , no empty movements
occur. In the case of   1 , the movements of empty containers equal the movement of full containers. Note
that  is a measure of how effectively empty containers can be repositioned. It reflects an equilibrium of the
container transport system. In practice,  is quite high. For instance, Lei and Church (2011) highlight that
  0.9 in the hinterland of Los Angeles/Long Beach. Overall, Braekers, Janssens, & Caris (2011) estimate
that    0.66;1 in practice. (See Legros, Bouchery, & Fransoo, 2019, for more details on empty container

matching in the hinterland.)


Three options are available for delivering a container from origin to destination, namely direct
shipment by truck (T), the use of HIT (H) via an inland terminal and the transloading of the cargo from a
maritime container into a semi-trailer and vice versa based on port-centric operations (P). We assume that the
same transport option is selected for all movements from and to a given demand zone. In what follows, we
model the total costs for each transport option. These total costs are the combination of direct costs with time-
8
related costs obtained by multiplying the time spent for shipping the cargo by the value of time   0
€/FEU.day. (See Appendix A for a detailed discussion of each transport option.) We focus here on
highlighting the expressions that drive our analysis. (Note that we additionally account for the reliability of
travel time as an extension in Appendix B.)
The first option involves direct truck transportation. The average costs per FEU of shipping to and
from i  H by using direct shipment by truck are expressed as follows:

CiT   P,i C T  DiT . (1)

where:
 P ,i = the distance from the port to i (expressed in km),

C T = truck transportation costs (expressed in €/FEU.km),


DiT = detention costs (expressed in €/FEU).

Moreover, DiT can be expressed via the following formula:

DiT  (1   ) max(0;  P,iT T  U  F )d   max(0;2 P,iT T  U  F )d , (2)

where:
 = average empty container movement ratio,
 P ,i = the distance from the port to i (expressed in km),

T T = truck transportation time (expressed in day/km),


U = loading/unloading time at the shipper (expressed in day),
F = detention-free period (expressed in day),
d = detention fees (expressed in €/day.FEU).

The second option involves HIT. We assume that n  ℕ (where ℕ refers to the set of natural numbers) inland
terminals are available in the hinterland for intermodal operations. We consider that terminal locations are
predetermined in line with tactical decisions made by LSPs. Indeed, large LSPs often purchase the terminal-
to-terminal service from railway companies or intermodal service providers, who offer services between a set
of existing intermodal terminals. For all j  1,...,n, terminal j is located at  x j , 0   H , with 0  x j . This is in

line with practice as inland terminals are often located along a corridor that can be modeled as a line. Note
also that our results can easily be extended to the case where terminals are located all over the hinterland. In
what follows, we limit ourselves to terminals located along the x-axis for reasons of clarity. If n  1 we
assume, without loss of generality, that for all j  1,...,n  1 , x j 1  x j . Throughout this article, we focus on

train transportation because this is the most developed HIT solution worldwide. However, the model is also
valid for road or barge HIT systems. The average costs per FEU of shipping to and from i by using HIT via an
inland terminal j are expressed as follows:

Ci j  C j   j ,iCT  Di j , (3)

9
where:
C j = train transportation costs from the port to terminal j (expressed in €/FEU),

 j ,i = the distance from inland terminal j to i (expressed in km),

Di j = detention costs (expressed in €/FEU).


Similar to Expression (2), we can express the average detention costs for shipping to and from i by hinterland
intermodal transportation as follows:

Di j  (1   ) max(0; x jT H  T j   j ,iT T  U  F )d 

 max(0;2 x jT H  2T j  2 j ,iT T  U  F )d , (4)

where:
T H = train transportation time (expressed in day/km),

T j = transfer time at terminal j (expressed in day).

The third option involves port-centric operations. This involves transloading the containers into road vehicles
(and vice versa) at the deep-sea port. We consider that the cargo loaded into one FEU is on average reloaded
into   (0; ) semi-trailers. The average costs per FEU of shipping to/from i via PCL are then expressed as
follows:

CiP  C 0   P,i  C P , (5)

where:
C 0 = the costs of port-centric operations (expressed in €/FEU),
 = average transloading ratio,
C P = semi-trailer transportation costs (expressed in €/FEU.km).

Throughout the article, when i refers to a specific point in the hinterland, we replace it in the notation by its
T
coordinates; that is, Cxi , yi refers to CiT for i  ( xi , yi ) .

Market areas of the transportation options

In the remainder of the article, we assume that the distance between two points ( xk , yk )  H and

( xl , yl )  H is evaluated by the Manhattan (or rectilinear or taxicab) distance:

 k ,l  xk  xl  yk  yl . (6)

The Manhattan distance is often selected to evaluate distances in the literature. The Manhattan distance
enables clear and meaningful insights of the market areas of the different logistics options we consider.
Bouchery and Fransoo (2015) use both the Euclidean and the Manhattan distance for an intermodal hub
10
location problem. They show that the results obtained are quite similar in both cases. Equation (6) enables us
to express  P, i and  j ,i for all j  1,...,n as follows:

 P,i  xi  yi , (7)

 j ,i  x j  xi  yi , for all j  1,...,n. (8)

In this section, we aim at separating the hinterland region into different zones. Each zone is optimally served
via one of the three options considered. We analytically identify below the best option to serve a given
demand zone. We start by stating some additional assumptions and by highlighting some preliminary results
for detention costs. First, we assume that for all j  1,...,n, x jT T  x jT H  T j , namely that hinterland

intermodal transportation is not faster than direct truck. Indeed, transfer time at the inland terminal and the
departure frequency limit the perceived speed of hinterland intermodal transportation. This assumption
implies that detention costs for HIT are equal to or higher than detention costs for direct truck as stated in
Lemma 1.

Lemma 1.
DiT  Di j for all i  H .
(All proofs appear in Appendix C.)
We further assume that U  F ; that is, the loading/unloading time at the shipper is strictly less than the

detention-free period. We can then express DiT as follows:

0 if  P,i  d1
DiT   2 P ,iT T  U  F  d if d1   P,i  d2 , (9)

 (1   ) P ,i T T  U  F  d if  P ,i  d 2
where:
F U F U
d1  T
, d2  .
2T TT
Note that Expression (9) implies that DiT is increasing and convex in  P ,i (as 2  1   ). The following

lemma highlights the variation of Di j and Di j  DiT .

Lemma 2.
Let j  1,...,n. Then:

1) Di j is increasing and convex in  j ,i .

2) Dxi ,0  Dxi ,0 is decreasing in xi for xi  0; x j  and increasing in xi for xi  x j .


j T

3) Dxi , yi  Dxi , yi is increasing in yi .


j T

The results of Lemma 2 will be used in the following subsections to identify the market areas of each
transportation option.
11
Direct truck versus hinterland intermodal transportation
We aim at characterizing demand zones for which the cost of HIT is strictly less than the cost of direct truck
shipment. We start by identifying the demand zones for which Ci j  CiT when xi  x j .

Theorem 1:

Let j  1,..., n . If C  x j C  (1   )( x jT  T j  x jT )d  0 , then Ci j  CiT for all xi  x j and for all yi .


j T H T

Else, there exists BT , j such that Ci j  CiT for all ( xi , yi )  H such that xi  x j and  i , j  B .
T, j

Theorem 1 highlights a possible tension between transportation costs and detention fees in the hinterland for
locations farther away from the inland terminal. Indeed, intermodal transportation would outperform direct
trucking for such locations unless the time penalty of HIT ( x jT T  x jT H  T j ) leads to excessive detention

fees. The condition for such a tension to occur is C  x j C  (1   )( x jT  T j  x jT )d  0 . Note that


j T H T

shipping lines, which control detention fees, would be reluctant to create conditions that deter a larger use of
HIT. We conclude that C j  x j C T  (1   )( x jT H  T j  x jT T )d  0 in many practical cases. We focus on such a

setting in what follows. Theorem 1 thus implies that Ci j  CiT for all xi  x j and for all yi ; that is, that HIT

shows higher performance than direct trucking for locations farther away from the inland terminal.

We focus next on the behavior of Cxji ,0  CxTi ,0 when xi  x j . We start with a lemma.

Lemma 3.
C j  x j C T  (1   )( x jT H  T j  x jT T )d  0  Cxjj ,0  CxTj ,0

Theorem 2 enables an understanding of the behavior of Cxi ,0  Cxi ,0 when xi  x j .


j T

Theorem 2:

There exists a unique 0  B


T, j
 x j such that CxjT  j ,0  CxTT  j ,0 for all j  1,..., n . Moreover,

x jCT  C j
B T, j
 .
2C T

Theorem 2 implies that for locations along the x-axis with xi  x j , direct trucking is more effective up to a
T, j j T
certain threshold distance B . Note that C xT  j , yi and C xT  j , yi can be different for some yi , as Lemma 2

12
states that Dxi , yi  Dxi , yi is increasing in yi . This implies that the border between direct truck and HIT is not a
j T

straight line if there exists yi such that DBjT , j , y  DBTT , j , y  DBjT , j ,0  DBTT , j ,0 . However, approximating this border
i i

by a straight line is acceptable in practice as long as TTd remains small. Indeed,


DBjT , j , y  DBTT , j , y  DBjT , j ,0  DBTT , j ,0  2 yi T T d . Therefore, if we take yi  300 km, if a truck runs on average
i i

1,000 km/day and if detention fees are €60/day, this leads DBjT , j , y  DBTT , j , y  DBjT , j ,0  DBTT , j ,0  36 . If we further
i i

assume that C T  €2.5/FEU.km, the maximum difference in x-coordinates for the border between direct truck
and HIT is 14.4 km for a difference in y-coordinates of 300 km, thus equaling less than 5%. Therefore, we
approximate the border between direct truck and HIT as a straight line in what follows.
Theorem 1 and Theorem 2 enable identifying the locations for which direct truck and HIT outperform
each other. Indeed, if we exclude PCL, the results above show that H can be optimally split into two market
areas. The first one, referred to as the DTnoP market area, is served by direct truck shipments and includes the

locations for which xi  BT , j with j  1 (as Cxji ,0  CxTi ,0 is strictly decreasing in xi for xi  0; x j  and the

border between direct truck and HIT is assumed to be a straight line). The second one, referred to as the
HITnoP market area, is served by HIT (via various inland terminals) and includes the locations for which
xi  BT , j , with j  1 , that is:


DTnoP  ( xi , yi )  H 0  xi  BT ,1 ,  (10)


HITnoP  ( xi , yi )  H xi  BT ,1 , (11)

These results are in line with the literature stating that HIT requires a certain distance to break even.

Direct truck versus port-centric logistics


In this subsection, we will characterize demand zones for which port-centric logistics is strictly more costly
than direct truck shipment. By combining Expression (1) and Expression (5), we obtain:

CiP  CiT   P,i ( C P  C T )  C 0  DiT , (12)

Expression (12) enables us to derive the following theorem:

Theorem 3.
If  C P  CT  (1   )T T d , CiP  CiT for all  P ,i .

Else, there exists a unique distance B P ,T such that CiP  CiT when  P,i  B . Moreover, CiP  CiT for all
P ,T

 P,i  B P,T and that CiP  CiT for all  P,i  B P,T .

In what follows, we assume that  C P  CT  (1   )T T d , as otherwise, CiP  CiT for all i  H and PCL is

not viable. In case  C P  CT  (1   )T T d , PCL is less costly than direct truck transportation for locations

farther away than B P ,T kilometers from the port. This implies that PCL requires a certain distance to break
13
even as the fixed costs of port-centric operations need to be compensated by the reduction in transportation
costs. Note also that this is a first sign of potential tension between PCL and HIT as they both will be
competitive over direct trucking for locations distant enough from the port. Theorems 1, 2 and 3 imply that the
direct truck (DT) market area (i.e., the demand zone for which direct truck shipment is optimally used) can be
obtained as:


DT  ( xi , yi )  H 0  xi  BT ,1 , xi  yi  B P ,T .  (13)

Hinterland intermodal transportation versus port-centric logistics


We aim here to characterize locations for which PCL is cheaper than HIT. Let j  1,..., n . We focus first on

locations for which xi  x j . By combining Expression (3) and Expression (5) and by taking  P,i  x j   j ,i ,

we obtain:

CiP  Ci j  C 0  C j  x j  C P   j ,i ( C P  C T )  Di j , (14)

Expression (14) enables us to derive the following theorem:

Theorem 4.

Let j  1,..., n and let xi  x j . Assume that C  C  x j  C  Dx j ,0  0 . There exists a unique BP , j such
0 j P j

that CiP  Ci j when  j ,i  B . Moreover, CiP  Ci j for all  j ,i  B and CiP  Ci j for all  j ,i  B .
P, j P, j P, j

Note that if C 0  C j  x j  C P  Dxjj ,0  0 , the terminal will be more efficiently served by PCL than by HIT,

which is unreasonable for the viability of the terminal. Therefore, we assume that

C 0  C j  x j  C P  Dxjj ,0  0 in what follows. Theorem 4 implies that in case xi  x j , PCL performs better

than HIT for locations farther away than BP , j kilometers from the inland terminal. This highlights a second

sign of possible tension between HIT and PCL for distant locations. HIT is often considered as efficient for
long distance transportation (Kim and van Wee, 2011), but our model shows that a different pattern may occur
if PCL is considered. We now focus on situations for which xi  x j and we derive the following theorem:

Theorem 5.

If C  C  x j C  D0,0  0 , Cxi ,0  Cxi ,0 for all xi  x j .


0 j T j P j

Else, there exists a unique BP , j such that CB P , j ,0  CB P , j ,0 . Cxi ,0  Cxi ,0 for all xi  BP, j and Cxi ,0  Cxi ,0 for
P j P j P j
 

all xi  BP , j . Moreover,  C P  CT  0  CiP  Ci j for all xi  BP, j .

Theorem 5 first implies that if PCL is more expensive than HIT for serving the port
( C 0  C j  x j C T  Dxjj ,0  CxPj ,0  Cxjj ,0  0 ), HIT will be more competitive than PCL for all i  H such that

xi  x j and yi  0 . The fact that HIT might be more competitive than PCL in the port area is quite
14
unreasonable. Therefore, we focus on the situations for which C  C  x j C  D0,0  0 . In such a case,
0 j T j

Theorem 5 implies that there exists a threshold BP  j for HIT to become competitive against PCL along the x-

axis (for xi  x j ). Note also that  C P  CT is likely to be negative in practice. Indeed, truck transport costs

are often comparable for containers and semi-trailers and   1 in many practical applications. We

consequently assume that  C P  CT  0 in what follows. In such a case, Theorem 5 additionally implies that

 
CiP  Ci j for all xi  BP, j . Let P, j  i  H xi  x j , Ci j  CiP . Theorem 5 implies that  P , j is piecewise

 
linear and convex. Note also that in case Di j  0 for all i  H xi  x j , P, j  i  H yi  Sxi  L j 
with S   C  C T and x jCT  C j  C 0
 
P
Lj  . We further define HITB P, j  i  H xi  x j , Ci j  CiP .
C  C
T P
C  C
T P

Market area split between consecutive inland terminals


In this subsection, we aim at identifying the split in market areas between the various inland terminals. We
further define the train transportation costs via terminal j for all j  1,...,n as follows (see Appendix A for a
more in-depth discussion):

C j  x j  T H  (1   ) Z H   (1   ) K j   T j , (15)

where:

T H = train transport time (expressed in day/km),


Z H = average train transportation costs for containers (expressed in €/FEU.km),
K j = transfer costs at terminal j (expressed in €/FEU),

T j = transfer time at terminal j (expressed in day).

We assume for the remainder of this article that inland terminals have similar operating conditions, namely
that in case n  2 , Tj 1  Tj and K j 1  K j for all j 1,..., n  1 . Furthermore, we limit ourselves to

n  2 through this subsection. We start by stating the following lemma:

Lemma 4.

 x j  k  x j 
Ci j  Ci j k for all j 1,..., n  1 , for all k  1,..., n  j and for all i  ( xi , yi )  H xi  .
 2 

The following theorem enables characterization of the borders between two consecutive inland terminals j and
j+1 for all xi  x j 1 :

15
Theorem 6.

For all j 1,..., n  1 and all k  1,..., n  j :

If Cxjjkk,0  Cxjjk ,0 , then Ci j k  Ci j for all i  H .

j k j k
Else, then there exists a unique B j , j  k such that Cxi ,0  Cxi ,0 for all xi  B j , j k and Cxi ,0  Cxi ,0 for all
j j

x j k  x j
B j , j k  xi  x j k . Moreover,  B j , j k  x j k .
2
j 1
The following lemma shows that Cx j1 ,0  Cx j1 ,0 based on the assumptions previously made:
j

Lemma 5.

C j  x j CT  (1   )( x j (T H  T T )  T j )d  0  Cxjjkk,0  Cxjjk ,0 for all j 1,..., n  1 .

Note that by applying Lemma 5 and as we assume that C j  x j CT  (1   )( x j (T H  T T )  T j )d  0 , there exists


j 1 j 1 j , j 1
a unique B j , j 1 such that Cxi ,0  Cxi ,0 for all xi  B j , j 1 and Cxi ,0  Cxi ,0 for all B  xi  x j 1 , according
j j

j j 1
to Theorem 6. Note that CB j , j1 , y and CB j , j1 , y can be different for some yi due to the potential delay in
i i

growth between Di j and Di j 1 . This implies that the border between consecutive inland terminals is not a

straight line if there exists yi such that DBj j ,1j1 , y  DBj j , j1 , y  DBj j ,1j1 ,0  DBj j , j1 ,0 . However, approximating this
i i

border by a straight line is acceptable in practice for the same reasons that led us to approximate the border
between direct truck and HIT as a straight line. Therefore, we approximate the border between consecutive
inland terminals as a straight line in what follows.

Theorem 7.

 
For all j 1,..., n  1 , if C T  (1   ) Z H  d (T H  T T )   T H then Cxi ,0  Cxi ,0 for all x j 1  xi . Else,
j 1 j

there exists a unique Bj , j 1 such that Cxj ,01  Cxj ,0 for all x j 1  xi  Bj , j 1 and Cxj ,01  Cxj ,0 for all Bj , j 1  xi .
i i i i

Note that C  x j C  (1   )( x j (T
j T H
 T T )  T j )d  0  C T  (1   )  Z H  d (T H  T T )    T H and
j 1
therefore, Cxi ,0  Cxi ,0 for all x j 1  xi , by applying Theorem 7. We conclude that Ci j  Ci j 1 for all
j

xi  B j , j 1 and that Ci j  Ci j 1 for all xi  B j , j 1 . We additionally show that two nonconsecutive inland
terminals have nonadjacent market areas.

Corollary 1.

For all j 1,..., n  1 and for all k  1,..., n  j , Cxi ,0  Cxi ,0  Cxi ,0  Cxi ,0 .
j j k j 1 j

16
Based on the results presented through this section, we can define the market areas of each transportation
option as follows:


HITB1  HITB P,1  i  H BT ,1  xi  x1 ,  (16)

 
HITB j  HITB P, j  i  H B j 1, j  xi  x j , for all j  2,..., n , (17)

 
HITF j  i  H x j  xi  B j , j 1 , xi  yi  min( BT , j , BP, j ) , for all j  1,...,n  1 , (18)


HITF n  i  H xn  xi , xi  yi  min( BT ,n , BP,n ) ,  (19)

HIT j  HITB j  HITF j , for all j  1,...,n, (20)


n
HIT  HIT j , (21)
j 1

PCL  H \ ( DT  HIT ) , (22)


with DT defined in (13). These expressions allow for an analytical characterization of the market areas of each
transportation option in case PCL is available.
In the next section, we investigate the effect of PCL on direct truck shipments and HIT. To do so, we
compare DTnoP with the DT market areas and HITnoP with the HIT market areas. Then, we assess whether

the use of continental containers might help to reconcile HIT and PCL.

Impact of port-centric logistics and use of continental containers


In this section, we start by studying the impact of PCL on direct truck shipment and on HIT. We show that the
tension between PCL and HIT is very likely to happen in practice. Therefore, we assess whether the use of
continental containers might help to reconcile the two practices. To do so, we assume that H is a rectangle of
width xmax and height 2ymax . Moreover, the deep-sea port is located in the middle of the rectangle’s height.

Impact of port-centric logistics


The idea we focus on in this subsection consists of identifying whether HIT is affected by PCL. We start by
identifying conditions such that direct truck shipment is not impacted by PCL.

Lemma 6.

DTnoP  DT if and only if B P,T  BT ,1  ymax .

Lemma 6 explains that the DT market area is not affected by PCL in case the area of interest for PCL
corresponds to the area for which HIT performs better than direct truck shipment. If the condition of Lemma 6
holds, the market area of PCL is either empty, or PCL is in pure competition with HIT by capturing part of the
HITnoP market area. This is a sign of possible tension between HIT and PCL.

17
We now identify conditions such that HIT is not affected by PCL. For all j  1,..., n , let

P, j
xmax  xi i   P , j  i  H yi  ymax  .

Theorem 8.
P, j
xmax  BT ,1
P, j
xmax  B j 1, j , j  2; n
HITnoP  HIT if and only if .
xn  BP ,n  xmax  ymax
x j  BP , j  B j , j 1  ymax , j  1; n  1

Theorem 8 states that HIT is not affected by PCL if the catchment area of each inland terminal is not affected
by PCL. If the conditions of Theorem 8 hold, HIT and PCL can be implemented in synergy. In such a case,
either the PCL market area is empty, or this market area is taken solely from the DTnoP market area. In this

setting, HIT is used to serve distant locations, and PCL enables improvement of operations in the surroundings
of the port. Indeed, if there are fewer semi-trailer movements than that of trucks transporting containers,
which is quite likely to happen in practice, the number of trucks needed in the hinterland is reduced if PCL is
implemented instead of direct truck shipment and, therefore, congestion in the port surroundings can be
reduced. Many port authorities and actors of the hinterland supply chain target this setting when implementing
PCL. However, Corollary 2 highlights a critical feature.

Corollary 2.
0
Conditions of Theorem 8 hold only for small values of xmax and ymax for large values of C .

We show in Corollary 2 that the conditions of Theorem 8 hold only for small hinterlands with large port-
centric operations costs. In such a case, the condition of Lemma 6 is likely to hold. Therefore, PCL is not
likely to be used. This implies that if PCL is a viable option in the hinterland of a port, this practice will
compete with HIT in terms of market area. We conclude that the tension between PCL and HIT is very likely
to happen in practice.
We further attempt to identify the conditions under which the market area of PCL is mainly taken
from the HITnoP market area (i.e., more than 50% of the market area of PCL comes from the HITnoP market

area). Some of these sufficient conditions are stated in Theorem 9. Note that the conditions stated in Theorem
9 are not mutually exclusive. Moreover, even in cases where none of the conditions stated in Theorem 9 hold,
the combined market area of PCL can be predominantly taken from the HITnoP market area.

Theorem 9.

18
More than 50% of the market area of PCL comes from the HITnoP market area if any one of the following

conditions holds:

xmax  xn  BP,n  BT ,1  B P,T  0 and BP,n  B P,T ,

or xmax  xn  BP,n  BT ,1 ,

or 2 B
T ,1
 BP,1 .

Define 1
HITBnoP  
 ( xi , yi )  H BT ,1  xi  x1 and
n
HITFnoP  ( xi , yi )  H xi  xn  . Theorem 9

describes two extreme cases, namely when the PCL area taken from DTnoP is smaller than the area taken from
n
HITFnoP (in the first two cases) and when the PCL area taken from DTnoP is smaller than the area taken from
1
HITBnoP (in the third case). The first two cases hold if xn is not large enough (compared to xmax ). In this

case, HIT is penalized by a long haulage distance by truck to reach locations far away in the hinterland, and
1
PCL becomes more favorable. In the third case, the area taken from HITBnoP is larger than DTnoP . This

happens when PCL is efficient for quite short distances (i.e., if BP,1 is large) and when DTnoP is small.

Reconciling PCL with HIT by using continental containers


Using a maritime container in the hinterland seems straightforward and simple. However, several
disadvantages might appear. First, the regulations differ from one country to another and therefore the
dimensions of the maritime containers are not always optimal for hinterland transportation. For instance, the
maximal size of an intermodal unit in North America is 53 feet. Therefore, the contents of three maritime
containers (40 ft. each) can often be transferred into two North American continental containers (Rodrigue &
Notteboom, 2009). Second, the shipping lines that own (or lease) the maritime containers are reluctant to let
their containers be used in the hinterland as this is associated with perceived opportunity costs. Therefore, they
charge detention fees that reduce the efficiency of hinterland operations. Third, empty container repositioning
in the hinterland will lead to considerable costs.
The reason for using continental containers is to combine the advantages of an intermodal unit without
the drawbacks of a maritime container used in the hinterland. In this subsection, we aim at assessing whether
the use of continental containers might help in reconciling PCL and HIT. This idea is quite straightforward as
the use of continental containers induces port-centric operations (i.e., transloading from maritime containers to
continental containers and vice versa) as well as HIT. In what follows, we model the costs associated with the
use of these continental containers. Note that even if we refer to a continental container for clarity reasons, the
modeling applies to any type of continental intermodal unit. Moreover, we assume that a continental container
is used only for intermodal movements, as there is no difference with PCL from a modeling perspective in
case of truck-only movements. (See Appendix A for more details of the modeling of continental container

19
transportation.) The average cost per FEU of shipping to and from i via a continental container through an
inland terminal j  1,....,n is expressed as follows:

CiCj  C 0'   'C j '   j ,i  'C P ' . (23)

In what follows, we limit our analysis to the special case of C  C ,  '   , and C P '  C P . In such a
0' 0

setting, the continental container can be used for intermodal operations, in contrast to the road-only
transportation unit considered previously, but doing so will induce higher truck transportation costs. The
average cost per FEU of shipping to and from i via a continental container through an inland terminal
j  1,....,n can be rewritten as:

CiCj  C 0   C j '   j ,i  C P ' , (24)

We start by proving that the use of continental containers via inland terminal j  1,....,n will be viable only

C j'
if  CP .
xj

Lemma 7.
C j'
For all i  H ,  C P  Ci j  CiP for all j  1,..., n .
C

xj

C j'
Based on Lemma 7, we assume that  C P in the remainder of the article; that is, there is an advantage in
xj
using train instead of truck to move continental containers to terminal j. Otherwise, the analysis made in

C j'
Section 4 can be directly transposed in case continental containers are available. In case  C P , using HIT
xj
enables reducing transportation costs to the terminal but induces larger costs for the final leg of transportation
by truck (as C P '  C P ). There is then a trade-off in using continental containers by HIT for locations farther
away from the terminal as compared to truck-only transportation. In Theorem 10, we state a sufficient
condition such that the use of continental container helps to reconcile PCL and HIT.

Theorem 10.

xmax  ymax  xn  BP ,n


If x C P  C n' , there exists i  H such that CiP  Cin and CiCn  CiP .
B P ,n
  n P'
C CP

The locations i  H such that CiP  Cin and CiCn  CiP identified by Theorem 10 have the following

features:
i) They belong to the hinterland.
ii) They are far enough from terminal n so as to make PCL cheaper than HIT.
ii) They are not too far away from terminal n so as to make HIT valuable for moving continental containers.
20
Illustrative example and insights
In this section, we apply the results obtained above to an example. We derive insights and we include a brief
analysis of how different stakeholder groups are affected.

An illustrative example
We consider the data from Table 1. These data have been obtained from discussions with professionals in a
Western European context. Note first that the hypotheses we made through the article are consistent with these

data; that are, x jT T  x jT H  T j for all j  1,..., n , C  x j C  (1   )( x jT  T j  x jT )d  0 for all


j T H T

j  1,..., n , C P  CT  0  (1   )T T d , C 0  C j  x j C P  Dxjj ,0  0 for all j  1,...,n,

C j'
C 0  C j  x j C T  D0,0
j
 0 for all j  1,..., n , C P '  C P and  C P for all j  1,..., n .
xj

x max 1,000 km C 2' 2'


C760 €/FEU €/FEU
760 d 60d €/FEU.day60 €/FEU.day
y max 300 km C
Z
Tft Tft
€/FEU.km
Z2.87 2.87
C €/FEU.km  100 €/FEU.day100 €/FEU.day

x1 250 km C P P
€/FEU.km
C1.83 2.02 €/FEU.km T T
0.00139T T
day/km
0.00139 day/km
x2 600 km C P' P'
C2.05 €/FEU.km
2.25 €/FEU.km T H 0.00104T H day/km
0.00104 day/km
1
C 612 €/FEU α α0.95 0.95 Z H
0.9Z H
€/FEU.km0.9 €/FEU.km

C 2
1,232 €/FEU  1.1 1.1 T1 0.5T 1 days 0.5 days
0
C 250 €/FEU F F 2 2
days days T2 0.5T 2 days 0.5 days
1'
C 392 €/FEU U U0.2 0.2
days days K1 50 €/FEU
K2 50 €/FEU
 C P  CT
S T  5.71
Table 1: Data of the illustrative example. C  CP

The results from Subsection 4.1 lead to BT ,1  232 km. Consequently, we find that

  
DTnoP  ( xi , yi ) 0  xi  232, yi  300 and HITnoP  ( xi , yi ) 232  xi  1000, yi  300 . Moreover, 
the results from Subsection 4.4 lead us to B1,2  534 km. We refer to Figure 1 for an illustration. By applying
the results of Subsection 4.2, we learn that B P ,T  292 km; that is, PCL is competitive against direct trucking
only over distances longer than 292 km. Indeed, trailers are much more competitive than direct trucking of
maritime containers (  C P  2.02€/FEU.km, i.e., 30% lower than C T ). However, port-centric operations are

costly ( C 0  250€/FEU). Therefore, PCL requires 292 km to break even. This enables us to define


DT  ( xi , yi ) 0  xi  232, xi  yi  292 . 
The results of Subsection 4.3 lead to BP,1  106 km, BP,1  221 km., BP,2  200 km and

BP,2  560 km. Note that the border between HIT and PCL for xi  x j is well-approximated by

 C P  CT

P, j  i  H yi  Sxi  L j  with S 
CT   C P
 5.71 , L1 
x1C T  C j  C 0
CT   C P
 1262 and L2  3159 .

Figure 2 illustrates the results when excluding the continental container option.

21
Figure 1: The market areas of DT and HIT (excluding PCL and continental containers).

PCL

PCL

Figure 2: The market areas of DT and HIT (excluding continental containers).

The comparison between Figure 1 and Figure 2 clearly illustrates the tension between HIT and PCL.
We can notice at first that the DTnoP market area is not strongly impacted by PCL, in opposition to the

HITnoP market area. We can also easily observe that more than 50% of the market area of PCL comes from

the HITnoTR market area (even if the conditions of Theorem 9 are not met). In the example, HIT is at best a

slightly better performer than DT. For instance, reaching inland terminal 1 by using HIT is just 8% cheaper
than using direct trucking (when excluding detention costs) and the cost advantage decreases as we move
farther away from the terminal. On the other hand, the performance of PCL relative to DT increases as we
move farther away into the hinterland. This explains why the market area of HIT is quite small in Figure 2.
We conclude that all decision makers involved in the management of containerized cargo in the hinterland

22
should make themselves more aware of the interaction between HIT and PCL as the tension between these
options is likely.
The results when including the possibility of using continental containers appear in Figure 3. Note
that we distinguish between the area served by an inland terminal using maritime containers and continental
containers in Figure 2. For inland terminal 1, both options are used ( HIT1 vs. C1 ), while for inland terminal 2,

only continental containers are used. We can observe that the use of continental containers helps extend the
market area of HIT for faraway locations (in accordance with Theorem 10). Indeed, the cost advantage of
reaching inland terminal 2 with a continental container is large as compared to PCL (1086€/FEU as compared
to 1460€/FEU) and the difference between C P and C P ' is not large enough to compensate for this difference.
As the impact of using continental containers is large in the example studied, we conclude that continental
containers have a real potential in reconciling PCL and HIT.

PCL PCL

PCL PCL

Figure 3: The market areas of DT and HIT (including continental containers).

Practices in some ports


We can draw some additional insights from the results we have obtained by comparing the settings of several
types of hinterland following the recommendations by McCalla (1999), who compares developments at two
Canadian ports in Halifax, Nova Scotia, and Vancouver, British Columbia, to show that the ports’ responses to
changes in global flows is highly contextual. Also, Ducruet and Lee (2006) identify local and regional factors
that in combination affect ports’ adaptation to globalization. First, consider the hinterland of the ports of Los
Angeles/Long Beach. We mentioned previously that 53 ft. containers are allowed in the US. This implies that

 is likely to be small (smaller than 1). As Ci is increasing in  , this is the first sign that continental
Cj

containers might be of interest. Moreover, the hinterland of the ports of Los Angeles/Long Beach is large with
inland terminals quite distant from the ports (to allow for east-west transfer). Therefore, the conditions of
Theorem 10 are very likely to hold. Our model leads us to conclude that continental containers are likely to be
used heavily. This is in line with the data, as in 2009, about 45% of the containers imported through Los
Angeles/Long Beach were transloaded into domestic containers (Rodrigue & Notteboom, 2009).

23
Second, consider the hinterland of the Port of Le Havre. This one not being very large, there are a
limited number of inland terminals, and rail transportation costs are high. Therefore, the market area of HIT is
small. If we assume that the truck transportation cost data of the example above are relevant for a French
setting and that the transloading costs are €250/FEU, our model leads us to conclude that (1) HIT will be
limited to locations close to the inland terminals, (2) locations 200 km or less from the port will be served by
direct truck, and (3) PCL will be used to serve those shippers located more than 292 km from the port (and out
of the market area of an inland terminal). As  is quite large and rail transportation costs are high, the use of
continental containers will not enable a reconciliation between PCL and HIT. We observe that this setting is
consistent with the current situation in the hinterland of Le Havre.
Third, consider the hinterland of the Port of Rotterdam. The hinterland is of medium size, rail
transportation costs are low, there is a very high density of inland terminals and  is quite large. Therefore,
direct truck is used for locations close to the port and HIT is used for more distant locations. Our model leads
us to conclude that PCL is not very likely to be developed. Note that we exclude short-sea shipping from our
analysis; that is, we do not consider market areas overseas. In case of short-sea shipping, the port is a natural
transloading point. This is in line with the practice in the Port of Rotterdam where the development of port-
centric operations mainly targets short-sea market areas and central warehouses for distribution throughout
Europe.
Fourth, the largest customer of the Port of Gothenburg, Stora Enso, uses an extreme form of
continental container, the Stora Enso Cargo Unit (SECU), to move forest products by train from their four
Swedish mills situated 160-450 km from the port of Gothenburg. Each SECU can carry almost 80 metric tons
and each train up to 3000 metric tons, utilizing the full capacity of the rail infrastructure. This is required for
managing large volumes on congested tracks but also for rail to compete with the generously sized trucks that
are up to 25.25 m long and with a capacity of up to 64 metric tons allowed in Sweden. Most SECUs leave
Gothenburg for continental Europe on RoRo ships, but, according to Hansen (2018), an annual amount of
30,000 FEUs of goods destined for overseas markets arriving in SECUs is transloaded into maritime
containers within or in the vicinity of the port. This shows that there is room for specialized PCL solutions for
certain products over both short and medium distances if the flows are large and concentrated. In addition,
Gothenburg’s general intermodal terminal, mostly used for domestic flows, was moved into the port area in
2018, improving the competitiveness of transloading in combination with hinterland transport by rail.

Effects of volume and value of time


We first discuss here how volume might affect our results. Indeed, our analysis assumes that all costs are
linear for whatever the number of containers. This is in line with the tariffs faced by LSPs, but we additionally
discuss here how volume might create economies of scale that would impact our results. Note that costs are
differently affected by economies of scale. For instance, truck transport costs are quite insensitive to the
volume effect.
H 0
Conversely, the costs that are the most sensitive to volume in our model are K j , Z and C . Indeed,

K j and C 0 includes the fixed costs to operate the inland terminal or the facility and therefore, more volume
24
leads to a decrease in this cost. Similarly, train transportation costs depend on the average load factor of the
T, j
train, that is, on the total volume handled. If more volume is handled by the LSP, our results show that B
and B P ,T would decrease. This will result in a reduction of the DT market area. Therefore, volumes favor both
HIT and PCL over direct truck and it is in line with the results of Crainic, Dell’Olmo, Ricciardi, & Sgalambro
(2015), who designed a train system serving three ports and two dry ports in Italy based on aggregating
volumes and adhering to time restrictions. Note that volume might also affect travel time by creating
congestion (mainly affecting truck transport) or by enabling more frequent intermodal services. These two
features tend to favor HIT. (See Bouchery and Fransoo, 2015, for an in-depth analysis of the effect of volume
on direct truck vs. HIT). We would also like to stress that volume is double-edged. On one hand, volume
favors both HIT and PCL over direct truck leading to what Hesse and Rodrigue (2004) identify as a
concentration of distribution centers at either gateway ports or inland hubs as trade volumes increase. On the
other hand, our results show that tension between HIT and PCL is likely to happen in practice. In such a case,
H 0
the volume is shared between both options, and, therefore, K j , Z and C might increase. This could

subsequently favor direct shipment by truck. We conclude that port authorities should carefully assess the
cannibalization risks between PCL and HIT before investing in the development of both options. This is in
line with Notteboom and Rodrigue (2005), who advise port authorities to do a thorough analysis of the truly
port-related activities and what is better done in the hinterland. Also McCalla (1999) states that ports must
respond to global trade changes but acknowledges the difficulty of ―who or what is to define these
responses?‖ (p. 253). Note also that the use of continental containers or semi-trailers is a promising solution to
make use of HIT and PCL without compromising volumes for either option. Many ports in northeast Europe
combine deep-sea container and short-sea semi-trailer flows. Transloading containerized goods into semi-
trailers can then make up sufficient volumes for HIT services, as investigated by Woxenius and Bergqvist
(2011).
We now discuss the impact of the increased value of time. Some shippers and some cargo are more
sensitive to time than others (Finnsgård et al., 2019) and may well require more urgent delivery. This
translates in our model to an increase in  . We can easily show that the fastest option for all i  H is direct
truck. Indeed, as compared to PCL, DT enables saving the time necessary for port-centric operations.
Moreover, the total distance traveled when using HIT is greater or equal to the total distance traveled by truck.
As we assume that hinterland intermodal transportation is not faster than direct truck (i.e., x jT T  x jT H  T j

for all j  1,...,n), we conclude that HIT is slower than direct truck. We conclude that an increase in  leads
to an increase in the DT market area. This is in line with practice where direct truck is used in case of urgent
deliveries, at least over moderate distances.

Effects on other stakeholders


Our analytical results focus on an LSP, as LSPs’ decisions are pivotal in the development of the
hinterland. We complement our results with a brief analysis of how the different options affect other
stakeholders and their incentives for promoting them.

25
Municipalities are often involved in the governance of port authorities and the port-city interface is
attracting a lot of attention (Monios et al., 2018). Conveying maritime containers directly to the hinterland by
rail relieves city infrastructure from the presence of quite a few trucks and their associated external effects,
whereas PCL creates jobs in the port city, albeit to a decreasing extent due to warehouse automation.
Container terminal operators often prefer HIT as it improves the port’s competitiveness in the hinterland,
facilitates increased throughput with constant land use and offers efficient terminal operations (Roso et al.,
2009). As an example, Brühl (2018) states that the interface with trucks at APM Terminals in Gothenburg
requires four to five times more straddle-carrier resources than transferring to on-dock rail.
Browne and Woxenius (2019) argue that a stage of reintegration of the interests of the city and the
port can be added to Hoyle’s (1989) five stages of port development. Hoyle’s division into stages is used by
Ducruet and Lee (2006) in their analysis of port–city relationships, but they also criticize it as it offers no
universal development path to follow. Nevertheless, Browne and Woxenius (2019) also identify potential
tensions due to the significant land use requirements for new distribution property near ports and a
complicated mix between public and private interests. The case of Gothenburg illustrates these somewhat
unaligned stakeholder interests as the port authority focuses on PCL and extends its landlord role by investing
in warehouses (Heitz et al., 2018), while the container terminal operator APM Terminals builds its growth
strategy firmly on HIT (Kristensen, 2019).
Shipping lines might lose control of the transport chain with PCL, as the maritime container fleet they
supply is used only port to port. The transport chain is divided at the PCL warehouse and it is not likely that
the shipping line is engaged in the hinterland transport beyond that warehouse. A shipping line serving a
single port in a port region, however, would prefer a PCL structure since it raises the barrier for the shipper
and LSP of using an adjacent port. As the port’s gates are often congested, trucking companies often prefer
driving longer distances once they have loaded a container, thus lowering the risk exposure related to waiting
times. However, they are surprisingly often in favor of HIT as they can avoid congestion in the port city as
well as at the gates, as discussed by Roso et al. (2009). On the other hand, PCL might enable some companies
to avoid picking up containers at deep-sea terminals and to focus instead on semi-trailer transportation from
the port-centric facility to the shippers’ locations. Moreover, there are examples of dedicated high-capacity
shuttles between a port and PCL warehouses, as in Gothenburg where 3 FEU trucks are used, as the port area
has private roads with generous legislation.

Conclusions
This article explores the impact of hinterland intermodal transportation and port-centric logistics on the
efficiency of inland operations for international containerized cargo flow. This topic has attracted extensive
attention in the maritime logistics and transport geography literature but predominantly with qualitative
approaches and as case studies. To our knowledge, this is the first model-based research contribution on PCL.
Instead of studying this concept in isolation, we focus on the relationship between PCL and HIT in order to
assess whether these two concepts are competing for the same market area. Our analytical results enable the

26
identification of the optimal market areas for each option. We also identify necessary and sufficient conditions
such that HIT is not impacted by PCL and we show that those conditions are not likely to occur in practice.
Based on additional results, we conclude that tensions between PCL and HIT are quite likely to occur, as PCL
tends to reduce the volume available for HIT. We additionally analyze how continental containers might help
reconcile the two concepts. We show a high potential for continental containers if they can carry a lot of
cargo, if the hinterland is large and if the inland terminals are located far away from the port. We draw further
insights by proposing an illustrative example and by analyzing how our results can be interpreted in the case
of different ports in Europe and in North America. The explanatory power of our model appears quite high
and may help managers identify the key elements influencing the development of each option in a given
setting.
We hope that this article will pave the way for future research. Some potential avenues of interest are
detailed below. First, we focus only on the economic impacts of the different options considered. Other
aspects, such as environmental impact or employment, could be additionally considered. Second, recall that as
we take an LSP perspective, we consider a tactical model with fixed locations for the inland terminals. This
approach might be complemented by a strategic study of where to locate inland terminals. In such a setting,
accounting for flow-dependent economies of scale for HIT would be particularly relevant. Third, our study
focuses solely on the LSP perspective. We additionally draw some preliminary insights of how the different
options affect the other stakeholders. This shows that shipping lines and deep-sea terminal operators might be
quite reluctant to accept the development of PCL if the concept is not implemented cautiously. A multiple
actor analysis might lead to further findings. This might also complement our results on the potential use of
continental containers. In Europe for instance, continental containers are not used extensively as part of
intercontinental flows. We highlight some causes for this lack of efficiency, but additional reasons could be
sought in the different stakeholders’ incentives for using them. Overall, we consider that many new and
interesting models could be derived from the development of the research on the inland transportation for
containerized cargo.

Acknowledgments
Part of this work was supported by research grants from the Normandie Region under the FLUIDE project.

27
References
Agarwal, R., & Ergun, Ö. (2010). Network design and allocation mechanisms for carrier alliances in liner
shipping. Operations Research, 58(6), 1726–1742.
Ambrosino, D., & Scutella, M. G. (2005). Distribution network design: New problems and related models.
European Journal of Operational Research, 165(3), 610–624.
Averbakh, I., Berman, O., Kalcsics, J., & Krass, D. (2015). Structural properties of Voronoi Diagrams in
facility location problems with continuous demand. Operations Research, 63(2), 394–411.
Balakrishnan, A., & Karsten, C. V. (2017). Container shipping service selection and cargo routing with
transshipment limits. European Journal of Operational Research, 263(2), 652–663.
Baumol, W. J., & Vinod, H. D. (1970). An inventory theoretic model of freight transport demand.
Management Science, 16(7), 413–421.
Bolloré, (2015). Bolloré Logistics announces the launch of its project to build a 36,000m2 logistics hub at the
port of Le Havre. Accessed Sept. 5, 2017 http://www.bollore-transport-logistics.com/en/media/press-
releases/bollore-logistics-confirms-its-position-as-leader-and-announces-the-launch-of-its-project-to-
build-a-new-36000m2-logistics-hub-at-the-port-of-le-havre.html.
Bouchery, Y., & Fransoo, J. (2015). Cost, carbon emissions and modal shift in intermodal network design
decisions. International Journal of Production Economics, 164, 388-399.
Bouchery, Y., Fazi, S., & Fransoo, J. C. (2015). Hinterland transportation in container supply chains. In
Handbook of Ocean Container Transport Logistics (pp. 497–520). Springer International Publishing.
Braekers, K., Janssens, G. K., & Caris, A. (2011). Challenges in managing empty container movements at
multiple planning levels. Transport Reviews, 31(6), 681–708.
Brooks, M. R., Puckett, S. M., Hensher, D. A., & Sammons, A. (2012). Understanding mode choice decisions:
A study of Australian freight shippers. Maritime Economics & Logistics, 14(3), 274–299.
Browne, M. & Woxenius, J. (2019). Port cities and urban logistics, In: M. Browne, S. Behrends, J. Woxenius,
G. Giuliano, & J. Holguin-Veras (Eds.) Urban logistics – Management, policy and innovation in a rapidly
changing environment (pp. 124–137). London: Kogan Page.
Brühl, M. (2018). Chief Commercial Officer, APM Terminals Gothenburg, interview 2018-04-23.
Carlsson, J. G., & Jia, F. (2013). Euclidean hub-and-spoke networks. Operations Research, 61(6), 1360–1382.
Chang, C. H., & Thai, V. V. (2017). Shippers’ choice behaviour in choosing transport mode: The case of
South East Asia (SEA) Region. Asian Journal of Shipping and Logistics, 33(4), 199–210.
Craig, A. J., Blanco, E. E., & Sheffi, Y. (2013). Estimating the CO2 intensity of intermodal freight
transportation. Transportation Research Part D: Transport and Environment, 22, 49–53.
Crainic, T. G., Dell’Olmo, P., Ricciardi, N. & Sgalambro, A. (2015). Modeling dry-port-based freight
distribution planning. Transportation Research Part C: Emerging Technologies, 55, 518–534.
Daganzo, C. F. (1991). Logistics systems analysis: Lecture notes in economics and mathematical systems.
Berlin: Springer-Verlag.
Dasci, A., & Verter, V. (2001). A continuous model for production–distribution system design. European
Journal of Operational Research, 129(2), 287–298.
Dekker, R., Bloemhof, J., & Mallidis, I. (2012). Operations research for green logistics–An overview of
aspects, issues, contributions and challenges. European Journal of Operational Research, 219(3), 671–
679.
Demir, E., Burgholzer, W., Hrušovský, M., Arıkan, E., Jammernegg, W., & Van Woensel, T. (2016). A green
intermodal service network design problem with travel time uncertainty. Transportation Research Part B:
Methodological, 93, 789–807.

28
Duan, L., Tavasszy, L. A., & Rezaei, J. (2019). Freight service network design with heterogeneous
preferences for transport time and reliability. Transportation Research Part E: Logistics and
Transportation Review, 124, 1–12.
Ducruet, C. & Lee, S.-W. (2006). Frontline soldiers of globalisation: Port–city evolution and regional
competition, GeoJournal, 67(2), 107–122.
Dullaert, W., & Zamparini, L. (2013). The impact of lead time reliability in freight transport: A logistics
assessment of transport economics findings. Transportation Research Part E: Logistics and
Transportation Review, 49(1), 190–200.
Erlenkotter, D. (1989). The general optimal market area model. Annals of Operations Research, 18(1), 43–70.
Fan, L., Wilson, W. W., & Dahl, B. (2015). Risk analysis in port competition for containerized imports.
European Journal of Operational Research, 245(3), 743–753.
Finnsgård, C., Kalantari, J., Roso, V. & Woxenius, J. (2019). The shipper’s perspective on slow steaming –
Study of six Swedish companies, Transport Policy. Published online 2019-10-19.
Fransoo, J. C., & Lee, C. Y. (2013). The critical role of ocean container transport in global supply chain
performance. Production and Operations Management, 22(2), 253–268.
Hansen, K. (2018). Senior Vice President, Stora Enso Logistics, interview 2018-05-08.
HAROPA, (2015). Multimodal Terminal. Accessed Sept. 5, 2017. http://www.haropa-
solutions.com/en/logistique-marchandise/infrastructures-portuaires/terminal-logistique-multimodal-
haropa.
Hoyle, B. S. (1989). The port-City interface: Trends, problems and examples, Geoforum, 20(4), 429–435.
Heitz, A. (2019) Planning urban freight and logistics: Duality in the logistics real estate market, the case of
the Paris Metropolitan area. World Conference on Transport Research (WCTR) 2019, Mumbai, 26-31
May, pp. 1–20.
Heitz, A., Dablanc, L., Olsson, J., Sánchez-Díaz, I., & Woxenius, J. (2018). Spatial patterns of logistics
facilities in Gothenburg, Sweden. Journal of Transport Geography, Available online 2018-03-21.
Hesse, M. & Rodrigue, J.-P. (2004). The transport geography of logistics and freight distribution. Journal of
Transport Geography, 12(3), 171–184.
Janic, M. (2008). An assessment of the performance of the European long intermodal freight trains (LIFTS).
Transportation Research Part A: Policy and Practice, 42(10), 1326–1339.
Kim, N. S., & van Wee, B. (2011). The relative importance of factors that influence the break-even distance of
intermodal freight transport systems. Journal of Transport Geography, 19(4), 859–875.
Kristensen, H. (2019). General Manager, APM Terminals Gothenburg, Speech at Port of Gothenburg Day,
Gothenburg, 3 October.
Langevin, A., Mbaraga, P., & Campbell, J. F. (1996). Continuous approximation models in freight
distribution: An overview. Transportation Research Part B: Methodological, 30(3), 163–188.
Larranaga, A. M., Arellana, J., & Senna, L. A. (2017). Encouraging intermodality: A stated preference
analysis of freight mode choice in Rio Grande do Sul. Transportation Research Part A: Policy and
Practice, 102, 202–211.
Leachman, R. C. (2008). Port and modal allocation of waterborne containerized imports from Asia to the
United States. Transportation Research Part E: Logistics and Transportation Review, 44(2), 313–331.
Legros, B., Bouchery, Y., & Fransoo, J. (2019). A time‐based policy for empty container management by
consignees. Production and Operations Management, 28(6), 1503–1527.
Lei, T. L., & Church, R. L. (2011). Locating short-term empty-container storage facilities to support port
operations: A user optimal approach. Transportation Research Part E: Logistics and Transportation
Review, 47(5), 738–754.
Liberatore, M. J. (1979). A model of freight transport selection. Transportation Journal, 92–100.

29
Limbourg, S., & Jourquin, B. (2010). Market area of intermodal rail‐road container terminals embedded in a
hub‐and‐spoke network. Papers in Regional Science, 89(1), 135–154.
López-Navarro, M. Á. (2014). Environmental factors and intermodal freight transportation: Analysis of the
decision bases in the case of Spanish motorways of the sea. Sustainability, 6(3), 1544–1566.
Mangan, J., Lalwani, C., & Fynes, B. (2008). Port-centric logistics. International Journal of Logistics
Management, 19(1), 29–41.
McCalla, R. J. (1999). Global change, local pain: Intermodal seaport terminals and their service areas. Journal
of Transport Geography, 7(4), 247–254.
McGinnis, M. A. (1979). Shipper attitudes toward freight transportation choice: A factor analytic study.
International Journal of Physical Distribution & Materials Management, 10(1), 25–34.
McGinnis, M. A. (1989). A comparative evaluation of freight transportation choice models. Transportation
Journal, 36–46.
Meng, Q., Wang, S., Andersson, H., & Thun, K. (2013). Containership routing and scheduling in liner
shipping: Overview and future research directions. Transportation Science, 48(2), 265–280.
Meixell, M. J., & Norbis, M. (2008). A review of the transportation mode choice and carrier selection
literature. International Journal of Logistics Management, 19(2), 183–211.
Monios, J., & Wilmsmeier, G. (2012). Port-centric logistics, dry ports and offshore logistics hubs: Strategies
to overcome double peripherality? Maritime Policy & Management, 39(2), 207–226.
Monios, J., & Wilmsmeier, G. (2013). The role of intermodal transport in port regionalisation. Transport
Policy, 30, 161–172.
Monios, J., Notteboom, T., Wilmsmeier, G., & Rodrigue, J.-P. (2016). Competition and complementarity
between seaports and hinterland locations for attracting distribution activities. Port Economics Discussion
Report 04/2016, Chios, Greece.
Monios, J., Bergqvist, R., Woxenius, J. (2018) Port-centric cities: The role of freight distribution in defining
the port-city relationship. Journal of Transport Geography, 66, 53–64.
Newell, G. F. (1973). Scheduling, location, transportation, and continuum mechanics: Some simple
approximations to optimization problems. SIAM Journal on Applied Mathematics, 25(3), 346–360.
Nguyen, L. C., & Notteboom, T. (2019). The relations between dry port characteristics and regional port-
hinterland settings: Findings for a global sample of dry ports. Maritime Policy & Management, 46(1), 24–
42.
Nierat, P. (1997). Market area of rail-truck terminals: Pertinence of the spatial theory. Transportation
Research Part A: Policy and Practice, 31(2), 109–127.
Notteboom, T. (2016). The adaptive capacity of container ports in an era of mega vessels: The case of
upstream seaports Antwerp and Hamburg. Journal of Transport Geography, 54, 295–309.
Notteboom, T. E., & Rodrigue, J.-P. (2005). Port regionalization: Towards a new phase in port development.
Maritime Policy & Management, 32(3), 297–313.

Nugroho, M. T., Whiteing, A., & de Jong, G. (2016). Port and inland mode choice from the exporters' and
forwarders' perspectives: Case study—Java, Indonesia. Research in Transportation Business &
Management, 19, 73–82.
Pettit, S. J., & Beresford, A. K. C. (2009). Port development: from gateways to logistics hubs. Maritime
Policy & Management, 36(3), 253–267.

Rodrigue, J.-P., & Notteboom, T. (2009). The terminalization of supply chains: Reassessing the role of
terminals in port/hinterland logistical relationships. Maritime Policy & Management, 36(2), 165–183.
Rodrigue, J.-P., Debrie, J., Fremont, A., & Gouvernal, E. (2010). Functions and actors of inland ports:
European and North American dynamics. Journal of Transport Geography, 18(4), 519–529.

30
Roso, V. (2007). Evaluation of the dry port concept from an environmental perspective: A note.
Transportation Research Part D: Transport and Environment, 12(7), 523–527.
Roso, V., Woxenius, J., & Lumsden, K. (2009). The dry port concept: Connecting container seaports with the
hinterland. Journal of Transport Geography, 17(5), 338–345.
Rothenbächer, A. K., Drexl, M., & Irnich, S. (2016). Branch-and-price-and-cut for a service network design
and hub location problem. European Journal of Operational Research, 255(3), 935–947.
Saeedi, H., Wiegmans, B., Behdani, B., & Zuidwijk, R. (2017). European intermodal freight transport
network: Market structure analysis. Journal of Transport Geography, 60, 141–154.
Sedghi, N., Shavandi, H., & Abouee-Mehrizi, H. (2017). Joint pricing and location decisions in a
heterogeneous market. European Journal of Operational Research, 261(1), 234–246.
Shams, K., Asgari, H., & Jin, X. (2017). Valuation of travel time reliability in freight transportation: A review
and meta-analysis of stated preference studies. Transportation Research Part A: Policy and Practice, 102,
228–243.
Sheffi, Y., Eskandari, B., & Koutsopoulos, H. N. (1988). Transportation mode choice based on total logistics
costs. Journal of Business Logistics, 9(2), 137.
Slack, B., Comtois, C., & McCalla, R. (2002). Strategic alliances in the container shipping industry: A global
perspective. Maritime Policy & Management, 29(1), 65–76.
Song, D. W., & Panayides, P. M. (2002). A conceptual application of cooperative game theory to liner
shipping strategic alliances. Maritime Policy & Management, 29(3), 285–301.
SteadieSeifi, M., Dellaert, N. P., Nuijten, W., Van Woensel, T., & Raoufi, R. (2014). Multimodal freight
transportation planning: A literature review. European Journal of Operational Research, 233(1), 1–15.
Tran, N. K., & Haasis, H. D. (2015). Literature survey of network optimization in container liner shipping.
Flexible Services and Manufacturing Journal, 27(2-3), 139–179.
Tsao, Y. C., Mangotra, D., Lu, J. C., & Dong, M. (2012). A continuous approximation approach for the
integrated facility-inventory allocation problem. European Journal of Operational Research, 222(2), 216–
228.
Tyworth, J. E. (1991). The inventory theoretic approach in transportation selection models: A critical review.
Logistics and Transportation Review, 27(4), 299.
Tyworth, J. E., & Zeng, A. Z. (1998). Estimating the effects of carrier transit-time performance on logistics
cost and service. Transportation Research Part A: Policy and Practice, 32(2), 89–97.
UNCTAD (2015). Review of maritime transport. New York and Geneva: United Nations Conference on Trade
and Development.
UNCTAD (2016). Review of maritime transport. New York and Geneva: United Nations Conference on Trade
and Development.
van den Berg, R. (2015). Strategies and new business models in intermodal hinterland transport. PhD thesis,
Eindhoven University of Technology.
van Klink, H. A., & van den Berg, G. C. (1998). Gateways and intermodalism. Journal of Transport
Geography, 6(1), 1–9.
van Riessen, B., Negenborn, R. R., & Dekker, R. (2016). Real-time container transport planning with decision
trees based on offline obtained optimal solutions. Decision Support Systems, 89, 1–16.
Veenstra, A., Zuidwijk, R., & van Asperen, E. (2012). The extended gate concept for container terminals:
Expanding the notion of dry ports. Maritime Economics & Logistics, 14(1), 14–32.
Vermeiren, T., & Macharis, C. (2016). Intermodal land transportation systems and port choice: An analysis of
stated choices among shippers in the Rhine–Scheldt delta. Maritime Policy & Management, 43(8), 992–
1004.
Wang, X., Meng, Q., & Miao, L. (2016). Delimiting port hinterlands based on intermodal network flows:
Model and algorithm. Transportation Research Part E: Logistics and Transportation Review, 88, 32–51.
31
Wang, X., & Meng, Q. (2017). Discrete intermodal freight transportation network design with route choice
behavior of intermodal operators. Transportation Research Part B: Methodological, 95, 76–104.
Witte, P., Wiegmans, B., & Ng, A. K. (2019). A critical review on the evolution and development of inland
port research. Journal of Transport Geography, 74, 53–61.
Woxenius, J., & Bergqvist, R. (2011). Comparing maritime containers and semi-trailers in the context of
hinterland transport by rail. Journal of Transport Geography, 19(4), 680–688.
Ypsilantis, P., & Zuidwijk, R. (2013). Joint design and pricing of intermodal port-hinterland network services:
Considering economies of scale and service time constraints. ERIM working paper nr. ERS-2013-011-LIS,
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2297698.
Zheng, J., Meng, Q., & Sun, Z. (2015). Liner hub-and-spoke shipping network design. Transportation
Research Part E: Logistics and Transportation Review, 75, 32–48.

32

You might also like