Intermodal container flows in a port system network

Share Embed

Descrição do Produto


Int. J. Production Economics 97 (2005) 75–88

Intermodal container flows in a port system network: Analysis of possible growths via simulation models Francesco Parola, Anna Sciomachen Department of Economics and Quantitative Methods, University of Genoa, Via Vivaldi 5, 16126 Genoa, Italy Received 2 August 2003; accepted 17 June 2004 Available online 8 October 2004

Abstract We present a discrete event simulation modelling approach related to the logistic chain as a whole in the northwestern Italian port system. We analyse the potentiality of the system by giving particular attention to the land transport and the modal split re-equilibrium with the aim of evaluating a possible future growth of the container flows. Some simulation models are analysed for highlighting both features and problems of the logistic activities of the intermodal network. In particular, the first experiment is performed according to the present configuration for validating the model itself and setting the parameters; successive models are developed for evaluating possible different scenarios of the land infrastructures in a 2012 vision. r 2004 Elsevier B.V. All rights reserved. Keywords: Logistics; Simulation; Performance analysis; Intermodality; Maritime container terminal

1. Introduction Containerisation has increasingly facilitated the transportation of goods since 1970s and it is still spreading all over the world. On the basis of the present trends, it is expected that the containerisation ratio will be over 70% of all general cargo at the end of this decade. In order to Corresponding author. DIEM, Universita` di Genoa, Via

Vivaldi 2, 12126 Genoa, Italy. Tel.: +39-10-2095484; fax: +3910-2095243. E-mail address: [email protected] (A. Sciomachen).

improve the benefits of the economies of scale, the size of containerships has greatly increased, and today the greatest vessels have a capacity of more than 8,000 twenty foot equivalent units (TEUs). Consequently, the containerisation requirements in the ports had to be taken into account to meet national and international demands. This noticeable raise in the standardisation of goods has permitted the introduction of intermodal transportation system, such that containerised goods can travel by rail, truck or sea. This represents a very important change for ports to increase their traffic, but, on the other side, can lead to congestion problems, with a subsequent

0925-5273/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpe.2004.06.051


F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

relapsed on port competitiveness and negative impact for local economies. Companies managing maritime terminals have the need of optimising all the operations involved in the container flow in order to achieve the maximal global productivity, that is expressed in terms of some opportune performance indices, such as turnaround time, throughput or hourly containers handled (Thomas, 1989). In fact, an increase in ship tonnage may lead to multimodal transport operators over sizing, with subsequent pressures on Port Authorities and especially on terminal operators for lower port charges and tariffs. Moreover, the handling and stacking phases require more space as a result of the increase of the ships’ size. Port competitiveness would therefore be more than ever depend on the quality of services provided by the terminal operators, which are required to improve flexibility in management and cost reduction through an adequate set of facilities and an efficiently yard organisation for handling and moving containers (see e.g. Kia et al., 2002; Kim et al., 2000, 2003; Yun et al., 1999). A container terminal layout is usually based on various requirements for container storage and transfer between ship and land feeder modes. When planning a new container terminal, quite a number of decisive factors have to be considered, which again will lead to the type of handling system that in turn requires appropriate and special types of equipment. Discrete event simulation models are more and more frequently used for capturing the synchronisation processes between the handling resources and the arrivals and departures of vessels, trains and trucks, since the complex resource allocation rules and the variety of stochastic processes involved make almost impossible the use of analytical approaches. Many cases, in fact, have been recently approached by using simulation tools for the performance evaluation of maritime terminals, with respect to either their productivity (see e.g. Kia et al., 2002; Legato and Mazza, 2001; Merkuryev et al., 1998; Shabayek and Yeung, 2002; Silberholz et al., 1991; Yun et al., 1999) or their intermodal connections (see e.g. Benacchio et al., 1998; Fabbri et al., 2001; Gambardella et al., 1998).

A more general view about multimodal transport is presented in Rizzoli et al. (2002), where the authors use a simulation model for representing the flow among and within inland intermodal terminals, focusing their analysis on the rail/road connections. In this work, we consider a set of maritime terminals together with the interconnections between them and land infrastructures. In particular, we are involved with the intermodal chain of the Italian ports of the Ligurian sea, and use the simulation software environment Witness (2000) for analysing the flow within the chain, focusing the attention on the prospect of empowering its connections, especially the railway network. More specifically, the main aim of the present work is to study how to face the impact of the sea traffic growth on the land infrastructures and highlight the degree of saturation of the railway lines and the level of congestion of the truck gates. The efficiency of the logistics chain as a whole is related to the good co-ordination of its different links and also some apparently small variables play an important role in its definition. Therefore, although the aim of the present simulation experiments is to study the behaviour of macro variables in the proposed scenarios, such as rail traffic share, the simulation model that will be discussed in details in Section 2 has been realised on the basis of single micro-systems (e.g. berth management and marshalling yard) composing it. In fact, different from the simulation models that we found in literature, that are mainly devoted at analysing only one link of the transport chain, our study has been carried out by the motivation of studying the macro environment without neglecting the least variables. In this work, we use simulation as a tool for supporting some strategic decisions involving too many different variables, that is not possible to manage otherwise. In particular, decisions about a huge amount of monetary investment, a possible development of infrastructure connections that, in turns, needs financial, environmental impact and project design approvals, cannot be undertaken without any realistic and technical evaluation based on the kind of behavioural results that simulation methods are able to grant.

ARTICLE IN PRESS F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

turn of different maritime container terminals, represented in our simulation model as single node, for which both the ship and the road & rail arrivals are considered. Given the characteristics of the port system network for which we have to analyse the main bottlenecks, we had the need of using a modular structure simulation approach. In particular, in the classical modelling cycle required for the organisation of any simulation study we first focus our attention on the main logistics network by considering all nodes as a ‘‘black boxes’’ that have been developed successively. More explicitly, in the two macro-areas of Genoa and La Spezia the main container terminals are modelled by using a modular structure that is able to efficiently represent both the logistic activities and the interconnection phases between each mode of transport. Note that another advantage of this modular approach is that it allows us a quicker and more rational analysis; in particular, within each ‘‘module’’ parameters and rules could be modified any time if necessary (see Section 3) without changing the functional behaviour of the global model. The modules that are explicitly represented in our model are those related to the input/output flow of containers in the system, that is the ship berthing, the truck gate and the rail yard (one for

Simulation methodological discussions are given along with all the paper. The paper is organised as follows. The main characteristics of the terminals under consideration and some information about their connection axes towards the hinterland are given in Section 2, together with details about our simulation methodology and the development of the model. The simulation experiments are presented in Section 3, where we focus the attention on the alternative scenarios and the parameters set for the input/ output analysis; such parameters affect the decision about the performance indices of interest. Section 3 also gives the main simulation results with respect to possible development in the direction of an enlargement of the traffic throughout the railway modality. Finally, some conclusions are given in Section 4.

2. The integrated port system and the simulation model The system we are involved with consists of two main components, namely the ports of Genoa (in the following considered as split into two ports, namely ‘‘old port’’ and ‘‘new port’’) and La Spezia, that are both located in Liguria county (Italy) (see Fig. 1). Each component consists in




Switzerland motorways present railway lines




“third” rail line




La Spezia

Ligurian Sea Fig. 1. Ports of Genoa and La Spezia (Italy).


F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88 r ai l y a r d m odule (arrivals)

truck gate module

r a i l y ar d m o du l e (departures)

YARD (main buffer)

ship berthing modules (variable number)

Fig. 2. The simulation modules within each maritime container terminal.

arrivals and one for departures). The interconnections among such modules are depicted in Fig. 2. Note that each module has to manage the import/export container flow from/to the yard, that represents the main buffer of a container terminal; each module is hence able to process the different means of transport entering into the system, namely train, vessel and truck, and to manage loading/unloading operations of containers for both sea and road & rail connections. We consider a peak month of maritime traffic, such as July, to obtain good validation results and successively to highlight the degree of saturation of the connection axes between couples of nodes belonging to the network. Appropriated goodness-of-fit tests, such as chisquare and Kolmogorov and Smirnov tests (see, e.g., Law and Kelton, 1999) have been used in order to evaluate the distributions extrapolated from the real world implemented in the model. Obviously, in all modules crane-working times are set in accordance with the real performances of the different handling facilities. We use the attribute ‘‘size’’ for distinguishing between 20 and 40 ft containers. In the following subsections we describe the main features and the implemented rules of each one of the above modules. 2.1. The ship berthing module This module is related to the arrivals and departures of vessels and their loading/unloading

operations. In order to simplify its complexity, this module represents a berth, that consists of a group of quay cranes, which perform the loading and unloading operations of vessels; consequently, both ship-to-shore cranes and harbour mobile cranes are considered as elements belonging to the module. In our simulation model, we can have a variable number of such modules depending on the length of the quay of the terminal under consideration. Each terminal has its disposal roads, represented by a buffer in the model, that receives the part ‘‘ship’’ entering in the system according to an exponentially distributed interarrival time (see the screenshot visualisation of this module depicted in Fig. 3). Containers are handled from the ship to the yard and vice versa. A quay crane is represented as split into two machines, for the import and export operations, respectively, that are considered as a unique component in the final report (see Sections 3 and 4). As soon as a ship arrives at the berth it communicates two attributes, that is the number of charging and discharging operations that have to be performed. At first, the available cranes begin the unloading operations and, only after the last lift in import, they start the loading operations. The most interesting rule used in this module is referred to the quay operations. In fact, it is necessary to have a correct sequence of ship arrivals, loading and unloading operations and departures. When a vessel arrives at the roads, one of the ship berthing modules calls the ship and puts it into a buffer depending on its associated variable state, whose value could be either ‘‘free’’ or ‘‘busy’’. In this way, when the berth is engaged, that is to say the buffer is full, the other ships in the roads have to wait until one of the other module becomes free. In practice, the rule used in the software environment Witness for governing the quay operations is the following: IF NPARTS (ship_buffer)=0 PULL from roads ELSE Wait ENDIF

ARTICLE IN PRESS F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88


unloading machine

ship departure

loading machine

loading crane

ship buffer

discharging buffer

unloading crane


ship Fig. 3. The ‘‘ship berthing’’ module.

After the berthing process, the quay cranes begin their unloading operations. We use a variable denoted ‘‘import_container’’ for counting the number of import moves, its initial value is the attribute related to the ship. Using the simple rule import_container ¼ import_container  1 we decrease the value of the variable at each discharging move of the quay crane. In this way, when the value goes to zero the element ‘‘export_crane’’, that is represented as a machine, calls the first container in the yard and begins the export operations. Analogously, we use a variable denoted ‘‘export_container’’ for counting the containers loaded on the vessel. When its value is equal to zero, the ship is ready to leave the terminal, since this implies that all loading operations are completed. We consider each day of the week as split into some time slots characterised by different arrival profiles. In those cases in which inter-arrival times were available, ship arrivals are represented by a negative exponential distribution; otherwise, a particular function of the software Witness has been used for simulating in a random order the arrivals of a certain number of vessels in different

time intervals. This function is reported in Fig. 4, where column ‘‘Time’’ shows the cumulative time expressed in hours, column ‘‘Length’’ shows the length of each time interval and column ‘‘Volume’’ gives the number of elements entering in the system in the corresponding time period; note that, for instance, the truck gate is not usually open during the night and thus the number of arrivals is zero. The same time function is also used for road & rail connections. Unfortunately, we did not have at our disposal any time series about the working times of the cranes, the ‘‘call size’’ of the vessels and the number of wagons of block trains; therefore, we had to exploit the information obtained from interviews made to terminal operators. In most cases, we use a normal distribution with a lower and an upper bound (T-Normal), related, for instance, to the minimum and maximum number of handled containers. 2.2. The truck gate module Another important phase of the intermodal chain is represented by the gate, at which hundreds of trucks arrive every day.


F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

Fig. 4. The window of the ‘‘arrival profile’’ function.

The truck gate module consists of some lanes where the ‘‘parts’’, representing the trucks, are processed both in import and export (see e.g. Rizzoli et al., 2002). In the model, we create an active part denoted ‘‘truck_in’’, that has different daily inter-arrival times. This part arrives at a buffer, namely a yard near the terminal, and joins the queue in a lane of the gate. Each lane, represented by a machine, processes the part ‘‘truck_in’’ and pulls the containers charged on the trailer to the stacking yard. The truck has to wait in a buffer for the availability of some import containers, that are stored in the yard. The time that each truck has to spend in the buffer, that is in the terminal, is defined by a T-Normal distribution, extrapolated by information of the stevedores, as well as the data concerning the truck arrivals during the day. Note that in some cases we have to distinguish between rush and non-rush hours. Both for the import and export cycles we assume that each trailer has either one or two containers, for a maximum length of 40 ft. In particular, we create an integer variable called ‘‘exit_rules’’ which

can assume the following values: exit_rules 8 20; > > < ¼ 40; > > : 2020;

if the truck has to charge one 20 ft container if the truck has to charge one 40 ft container if the truck has to charge two 20 ft containers:

In this way, the machine that has to manage the exit of trucks can load the trailer in accordance with the value of the above variable. 2.3. The import–export flows via rail The rail modality has a great importance in the model as both the European Community and the Italian Government aim at re-equilibrating the present modal split unbalance. In our analysis we decided to focus our attention only on block trains, called intermodal quality system (IQS), as they represent a large part of the volumes transported via rail. As far as the rail interconnections are concerned, we use two different kinds of modules, differentiating between inward and outward containers.

ARTICLE IN PRESS F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

The module of export via rail represents the receiving of wagons from the freight station and the unloading operations of containers to the yard. It is not characterised by difficult rules, as they are very similar to those that have been already described for the previous modules. Wagons, arriving from the marshalling yard external to the terminal, are moved by a locomotive to the operating tracks, where the discharging operations take place. In this module there is a group of machines that unload containers and move them on the yard. As in the case of the truck gate, in this module it is necessary to create an integer variable, denoted ‘‘teus_unloading’’, that defines the number and the size of the containers discharged from each wagon. The possible values assumed by this variable are the following: teus_unloading 8 20; > > > > > > 40; > > < ¼ 2040; > > > > 2020; > > > > : 202020;

if the wagon has one 20 ft container ifthe wagon has one 40 ft container if the wagon has one 20 and one 40 ft containers if the wagon has two 20 ft containers if the wagon has three 20 ft containers:

This variable follows an integer distribution, that is built in accordance with the information that have been provided to us by the Cargo Division of the Italian Railways. The module of import trains represents the loading of wagons at the terminal, the delivery of containers to the freight station and the composition of trains before their departure. This module presents a further difficulty in comparison with the previous one, due to the necessity of reproducing the cycle of departure. In fact, for the export trains there is not any problem as it is sufficient to create the part ‘‘export_train’’ with an inter-arrival time on the basis of the official schedule of the Cargo Division. The train paths are reproduced in the model by using some conveyors; trains, running along the railway lines, arrive at one of the marshalling yard external to the port where the wagons are split and transported to each container terminal. On the contrary, in the import cycle, it is necessary to reproduce the overall process from


the loading of the containers on the wagons to the train departure from the marshalling yard. Consequently, we create a ‘‘virtual part’’, with the same inter-arrival time as the import trains, thus getting a signal of the departure. This virtual part allows us to modify the value of an integer variable denoted ‘‘go’’ used in the following statement: IF go=1 PULL from marshalling_yard ENDIF ELSE Wait ENDIF This rule is used for a machine which plays the role of moving the wagons in the marshalling yard where, after possible arrivals of trucks coming from other terminals, the block train will start. Before the departure of the train from the freight station, we label this part with the following two attributes that are very useful with respect to the technical limits of the railway lines:


‘‘n_wagons’’, which shows the number of wagon of the train; ‘‘tonn’’, which gives the weight of the transported freight.

The value of such attribute is obtained through a relation existing among the elements that is represented in a correlation matrix. The loading operation of each wagon is managed in this module, thanks to rules very similar to those used in the truck gate module for the departure of the lorries.

3. Model validation and scenario analysis For validation purposes, we have used the data concerning the present scenario and compared our resulting indices with those extrapolated from the real world. All the simulation runs have been performed within the experimental framework of the software environment Witness. In particular, we fixed the time unit of the simulation clock to 1 h


F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

and made a warm-up period equal to one month (i.e. 31 days), corresponding to 744 time units. Then, we carried out five consecutive replications of length equal to 1 month, changing the random number for each replication. In each run we computed the data of interest (e.g. rail and road flows in TEUs, number of leaving/arriving trains, etc.) and the reporting interval has been fixed to the end of each replication, asking for an output file in a CSV type. Note that the resulting data about the performance indices of interest are hence given by the average value of all the monthly indices computed within each run. In particular, in order to be able to update any of the above performance indices we used an integer variable acting in the following way. Let us consider, for instance, the rail yard module (departures) (see Section 2); in this module there is a machine aiming at generating the block trains leaving the terminal. We created a counter able to point out the number of trains leaving the terminal from the beginning of the simulation run. In the window ‘‘actions on finish’’ of this machine we used the following iterative string: train_departures ¼ train_departures þ 1: Thus, when a train leaves, the value of the integer variable ‘‘train_departures’’ increases by a unit. The same way has been used for obtaining the value of the other indicators, such as ‘‘TEUs transported via rail’’, ‘‘TEUs transported via road’’, and so on. An example of how we use such counters in the Witness framework is reported in Fig. 5, where the case of one terminal in the port of Genoa is considered. It is worth mentioning that our analysis is mainly focused on the performance indices related to the modal split, such as the total number of daily trucks and trains and the volume in TEUs handled by land transport. Note that we consider the modal split as given by the rail and road traffic shares; they are given (in percentage), respectively, by t ¼ rail traffic share TEUs transported via rail ¼ ; l



wagons out

rail import (TEUs)

wagons in

rail export (TEUs)

trucks out

road import (TEUs)

trucks in

road export (TEUs)



sea import (TEUs) sea export (TEUs)

Fig. 5. Example of changing parameters in the validation phase.

r ¼ road traffic share TEUs transported via road ; ¼ l


where l represents the land traffic (TEUs) given by l ¼ total containers handled by sea  total transhipped containers:


Presently, the considered ports, excluding transhipment and bulk cargo, show different intermodal ratios. In particular, in Genoa according to Eq. (1) we have t=31%; this is a good percentage if we consider that in the 1980s the same value was about 10% less than the present one; however, in La Spezia, we have t ffi 40%. The rapid growth in the maritime transport and the unsustainability of present road traffic flows impose a radical change in the current policies of land traffic management. In fact, the congestion both in the motorway network and in the urban areas has persuaded Port Authorities to plan great investments aimed at reducing the traffic share via road given in Eq. (2), thus achieving r ffi 50%. We started our steady-state analysis for studying future scenarios in accordance with three different hypotheses, that are described in the following subsections. The simulation results of each scenario have been then compared with those of the simulation runs of the present configuration, that is year 2002. The three different operative scenarios synthesized in Table 1 are obtained in our model by

ARTICLE IN PRESS F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88


Table 1 Our hypothesis for future scenarios 2012 Scenarios First best

Second best

New railway line and doubling of the Pontremolese line

New railway line

Worst option

Only restructuring of the main railway lines New rail leg in the Genoa hub Technological updating of the existing railway lines Reorganisation of the main freight stations Opening 24 hours a day of the truck gate in the main container terminals

3.1. The ‘‘first best’’ scenario: 50% rail versus 50% road transportation modality

O ld li ne

“Pontremolese” Pontremolese” line

Tyrrhenian line


new rail leg

La Spezia

Fig. 6. The structure of the present railway network in the considered area.

The corresponding first simulation experiment has been hence performed by taking the following assumptions


We begin our analysis by considering the scenario, denoted ‘‘first best’’ option, that certainly represents the most favourable configuration. This scenario has been made in accordance with the aims outlined by the National Transport and Logistic Plan in 2001 and the Port Planning Schemes of Genoa and La Spezia.

“Succursale” line

New line

ine ”l da va “O

modifying the parameters (e.g. inter-arrival times, train length, etc.) in the truck gate and rail modules (see Fig. 2). Note that the parameters of the ship-berthing module have been not modified, since we assume that the maritime flows are constant in all scenarios. For instance, it is possible to modify the number of daily trains from/to the terminals in the different time bands of the day by changing the value of the parameters in column ‘‘Volume’’ of the ‘‘arrival profile’’ window of the part representing the block train (see Fig. 4). Column ‘‘Volume’’ shows the number of trains leaving the terminals in a defined time slot; by looking at the 7th row we have 10 block trains leaving the terminal in accordance to an exponential distribution (with a random number stream equal to 315) in a time interval of 8 hours (column ‘‘Length’’). This interval begins at the 42nd hour of the simulation run (column ‘‘Time’’). Obviously, it is also possible to modify the length of the intervals by changing the parameters in column ‘‘Length’’.


realisation of the ‘‘New line’’ and doubling of the ‘‘Pontremolese’’ line (see Fig. 6). Note that this strategy would permit to use trains up to 650 m in length (equal to 32–33 wagons) and over 1.200 tonnes in weight with only one locomotor; technological updating of the main existing railway lines; reorganisation of the main freight stations by allowing track length up to 650 m and rationalisation of the arrival and departure operations of block trains 7 days a week; opening 24 hours/day of the gate in the two main terminals of Genoa and in the main terminal of La Spezia; doubling the motorway to/from Genoa.


F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

In this scenario, we have carried out the same monthly runs as before with the aim of getting some information related to both the short-term and the intermediate one. The main figure is certainly represented by the radical change in the value of l given by Eq. (3) that is, respectively, l ffi 267;000 TEUs in Genoa and l ffi 137;000 TEUs in La Spezia; in this case we have an almost equal distribution between road and rail modality, that is t ¼ 51:9% and r ¼ 48:1% in Genoa, while t ¼ 50:9% and r ¼ 49:1% in La Spezia. In this proposed configuration the main container terminals, namely three in Genoa and one in La Spezia, handle a volume via rail greater than the present land traffic as a whole. The results of this experiment show that the average number of daily block trains is about 63 in Genoa and 30 in La Spezia; note that in the present configuration (year 2002) we have, respectively, only 30 and 18 block trains. Moreover, we have a better distribution of the traffic during the week, thanks to the unexploited time slots of the passenger trains on Sundays and holidays. Table 2 shows the key-role played by the ‘‘Third Apennine line’’ (see Fig. 6), above all as regards the northbound traffic, that is about 73% of the overall flows; on the contrary, southbound trains, that do not have problems of ruling gradient, also use both the Old Line, that is about 5% of the overall traffic, and the ‘‘Succursale’’ one (36%).

Thanks to the doubling of the Pontremolese line and the possibility of using trains 650 m in length, in La Spezia all the rail traffic from/to the north can transit through this line, thus saving important time slots in the Tyrrhenian line and in the other lines of the Genoa railway junction. With a comparison between these resulting flows and the current ones (see Table 2), it clearly comes out how, thanks to the planned investments, the relations with the hinterland are only managed by following a north/south direction, without any interference between the systems of Genoa and La Spezia. Furthermore, in this simulation experiment we have observed the increasing importance reached by the main marshalling yards, which show a noticeable traffic growth. As regards the evolution of the road traffic, we have to make a distinction between the ports of Genoa and La Spezia; in fact, our simulation results show that maritime flows are doubled but, thanks to the modal re-equilibrium, the volumes via road have shown an increase of only about 50% apart from the gate of the old port of Genoa, where the number of daily trucks is twice than the present one. Actually, owing to the opening 24 hours a day of the gate of the main terminals, the growth of the road volumes in the rush hours is limited to 20–30%, since the daily traffic is split into a greater amount of working hours.

Table 2 Cargo trains in the first best scenario Direction

Weekly trains

Monthly trains


La Spezia



La Spezia


Ovada line Succursale line New line


3 53 162

— — —


18 234 676

— — —


Old line Succursale New line


9 74 129

— — —


41 333 562

— — —


Tyrrhenian line

Eastbound Westbound

5 12

2 4

3 8

30 37

14 10


Northbound Southbound

89 96

89 96

403 440

403 440

— —

16 27 — —

ARTICLE IN PRESS F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

3.2. The ‘‘second best’’ scenario An alternative scenario, which presents a more difficult situation of the system as a whole, is characterised by the missed doubling of the Pontremolese line, even if we can in any case suppose a maximum train length equal to 500 m. This new scenario shows heavy effects both on port activities and on the rail traffic along the Tyrrhenian line and in the Genoa hub. For a better comprehension of the performance indices of this scenario, we give the following definitions: Q ¼ maximum capacity ¼ maximum number of trains which can use the railway line every day;


f ¼ free time slots ¼ maximum capacity  average daily trains ¼ additional number of trains which could use the railway line every day:

Table 3 Capacity of the different stretches and real traffic flows (real data—2002) Q

Railway lines



Passenger Freight Total Old Line

Stretch 1 125 Stretch 2 125

46 46

21 21

67 67

58 58

Succursale line

Stretch 1 230 123 Stretch 2 230 147

90 101

213 248

17 18

‘‘Ovada’’ line

Stretch 1 Stretch 2

22 8

5 7

27 15

53 45

Tyrrehenian line Stretch 1 220 164 Stretch 2 220 125 Stretch 3 220 108

53 53 53

217 178 161

3 42 59

80 60


n ¼ average daily trains ¼ average number of trains by using the railway line every day;



This scenario certainly presents a less number of free time slots Eq. (6) from and to the Po Valley than the previous one, moreover, the length of the trains concerning the port of La Spezia is limited to 500 m. Consequently, like it is happening now, a share of the traffic from and to La Spezia has been shifted along the coastal line up to the achievement of the complete saturation of the Tyrrhenian line. By analysing the results concerning this experiment we can focus our attention on the number of daily cargo trains along the Tyrrhenian line and on the modal split. In this scenario La Spezia maintains its favourable modal ratio but, due to the shorter length of the trains, a greater number of time slots during the day is necessary. However, in this case, the La Spezia system is not independent as in the previous scenario (first best) and hence it has to be supported (for about 50% of the total traffic) by the Genoa hub as far as the trains from/to the North of Italy are concerned.

As we see in Table 3 referred to the real data of the present traffic flows (year 2002), the Tyrrhenian line can be divided into three stretches with different characteristics and degrees of saturation. In fact, the sections nearest to Genoa, that is ‘‘stretches 1 and 2’’, show a higher level of congestion due to the presence of both the regional service for passengers and, in particular, the metropolitan traffic from/to the suburbs of the city. In Table 3 we present some indices that are derived according to Eqs. (4)–(6). In fact, in this scenario in order to compensate the missed realisation of the doubling of the Pontremolese, we assume an alternative and cheaper investment, namely the building of a new rail leg (see Fig. 6) that could double ‘‘stretch 1’’, thus separating the passing traffic from the metropolitan one. Note that in this scenario ‘‘stretch 1’’ is the actual bottleneck of the overall system. In any case, the proposed investment is already planned by the Italian Railways. In conclusion, we can say that this 2012 vision, although achieving a modal equilibrium, shows a quite unstable situation in the long-term as the La Spezia hub is not self-sufficient for its growing traffic. On the other side, in this scenario the ports of Genoa and the management of theirs traffic flows do not present particular problems.


F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

3.3. Another possible scenario: The modal unbalance The last scenario under consideration is certainly not the most desirable one since it will probably induce the strangling of the port activities which could not have the infrastructures required for land transport of goods. Moreover, the empowering of some Alpine Passes (see Fig. 1), such as the Gothard and the Simplon ones, could transform the chance of expanding the hinterland in the treat of losing further traffic shares in favour of the ‘‘northern range’’ ports. With the same hypothesis of the first scenario (see Table 1), that is the technological updating of the main railway lines, the empowering of the freight stations and a better management of the rolling-stock, we now assume the lack of a new railway line in both hubs. In accordance with the above-mentioned investments, we suppose a maximum train length equal to 550 m, except for the Tyrrhenian and the Pontremolese lines, where it is 500 m. Under these conditions, the simulation runs have demonstrated how it is not possible to achieve a modal equilibrium, in fact, in this case t=34.3% and r=65.7% in Genoa, while t=40.5% and r=59.5% in La Spezia. It is worth mentioning that the high level saturation of some stretches does not permit further increase in the number of daily cargo trains from and to the ports. Another arising problem is represented by the necessity of using two locomotors beyond a certain tonnage, due to the high-ruling gradient of the Succursale and the Pontremolese lines; this issue requires the availability of a great amount of locomotors and time slots along the abovementioned railway lines. Due to the restriction of Q, in this scenario it is not possible to increase too much the cargo trains along the Succursale line, while thanks to the technological updating of the Ovada Line, we have at our disposal a good amount of additional time slots during the day. Obviously, for the southbound traffic we should exploit at maximum the so-called ‘‘Old Line’’ which, for its ruling gradient, is not useful for the northbound one.

The La Spezia system shows a different situation since n (see Eq. (5)) is not decreased with respect to the previous scenario (i.e. the second best scenario), even if the reduced length does not allow to move the same amount of wagons. In order to maintain a certain degree of independence between the systems of Genoa and La Spezia we decide to keep under 20% the quota of trains from/to La Spezia, passing through the Genoa hub. Clearly, this aim is achievable only in the hypothesis of the realisation of the investments already considered in the previous scenario and, in particular, of the new rail leg between the eastern outskirts and the centre of the town (see Fig. 6). Finally, we can briefly analyse the situation of the road traffic, which by considering the inadequacy of the rail infrastructures, has to sustain the greatest part of the cargo volumes. In this scenario, the number of daily trucks increases proportionally with respect to the growth of the sea traffic, note that in the old port of Genoa this raise is particularly high, more than 170%. Although there is the availability of 24 hours gates for main container terminals, this choice is not sufficient to face the traffic volumes, in particular, during rush hours, and to ensure a sustainable growth in the long-term.

4. Conclusions In Section 3 we have presented three different scenarios in which we have analysed the impact on land infrastructures by assuming a constant sea traffic growth in the time period 2002–2012. In conclusion, we want to give a comparison between the main results obtained by performing our simulation experiments concerning the best (scenario 1) and the worst (scenario 3) case configurations. The key output parameters chosen for the evaluation of the different scenarios are


the rail traffic share, that is value t given by Eq. (1), reported in Fig. 7 the percentage increase in road traffic during the period 2002–2012 shown in Fig. 8.

ARTICLE IN PRESS F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88

achieve a modal equilibrium between road and rail traffic, especially in the new port of Genoa, as it is more clearly outlined in Fig. 7. By looking at Fig. 7, we can also see that the port of La Spezia seems to show a better situation than the Genoa ones; in fact, it has a quota of rail traffic over 40%. Obviously, the modal unbalance results in a greater road traffic congestion, as we can see in Fig. 8, where a growth in the sea flow implies an almost proportional increase in the road one in the next 10 years. In particular, the old port of Genoa

We can note that on the basis of the value of t emerges that the first scenario is the best one (see Fig. 7). In particular, Table 4 shows the number of daily trains in each considered port and the corresponding maximum length. We can see that in the best option the supply of trains is able to face the additional volumes of the demand, showing a good modal equilibrium. On the contrary, due to the reduced train length and the lack of new infrastructures, the number of daily trains in the last scenario is not sufficient to Genoa - new port


Genoa - old port

port of La Spezia

100% 90% 80%





53.7% 62.2%







60% 50% 40% 30%


50.0% 30.2%






10% 0%

simulation- first best worst scenario 2002

simulation- first best worst scenario 2002 rail share

simulation- first best worst scenario 2002

road share

Fig. 7. The modal split in the ‘‘first best’’ scenario and in the last one.

Genoa - new port

Genoa - old port

port of La Spezia

200.0% 172.2% 180.0% 160.0% 135.2% 140.0% 120.0%






100.0% 80.0% 52.5%



40.0% 20.0% 0.0% sea traffic growth

first best

worst scenario

sea traffic growth

first best

worst scenario

sea traffic growth

first best

worst scenario

Fig. 8. The road traffic growth in the ‘‘first best’’ scenario and in the last one compared with the increase in the sea traffic (2002–2012).

ARTICLE IN PRESS F. Parola, A. Sciomachen / Int. J. Production Economics 97 (2005) 75–88


Table 4 Number of daily trains in the ‘‘first best’’ scenario and in the last one

Genoa (new port) Genoa (old port) La Spezia



Maximum train length

Simulation—2002 First best Worst scenario Simulation—2002 First best Worst scenario Simulation—2002 First best Worst scenario

11 27 21 19 36 26 18 30 29

550 mt 650 mt 550 mt 550 mt 650 mt 550 mt 440 mt 650 mt 500 mt

shows the most critical situation and hence in this case the realisation of new rail infrastructures seems to be the only strategic decision for an efficient transport policy aimed at improving the port competitiveness.

Acknowledgements We wish to thank the anonymous referees for their suggestions for improving the paper. References Benacchio, M., Carnevali, C., Genova, R., Mazzucchelli, M., Sciomachen, A., 1998. The port as intermodal node: Models for yard planning and management. In: Sciutto, G., Brebbia, C.A. (Eds.), Maritime Engineering and Ports. WIT Press, Boston, pp. 155–164. Fabbri, E., Parola, F., Sciomachen, A., 2001. Intermodal connections of a maritime container terminal: Analysis of

long terms plans. In: Olivella Puig, J. (Ed.), Maritime Transport, UPC, pp. 735–746. Gambardella, L.M., Rizzoli, A.E., Zaffalon, M., 1998. Simulation and planning intermodal container terminal. Simulation 71, 107–116. Kia, M., Shayan, E., Ghotb, F., 2002. Investigation of port capacity under a new approach by computer simulation. Computers & Industrial Engineering 42 (2–4), 533–540. Kim, K.H., Park, Y.M., Ryu, K.R., 2000. Deriving decision rules to locate export containers in containers yards. European Journal of Operational Research 124 (1), 89–101. Kim, K.H., Lee, K.M., Hwang, H., 2003. Sequencing delivery and receiving operations for yard cranes in port container terminals. International Journal of Production Economics 84 (3), 283–292. Law, A.M., Kelton, W.D., 1999. Simulation modelling and analysis. McGraw Hill, New York. Legato, P., Mazza, R.M., 2001. Berth planning and resources optimisation at a container terminal via discrete event simulation. European Journal of Operational Research 133, 537–547. Merkuryev, Y., Tolujew, J., Blumel, E., Novitsky, L., Ginters, E., Viktorova, E., Merkuryeva, G., Pronins, J., 1998. A modelling and simulation for managing the Riga Harbour Container Terminal. Simulation 71, 84–95. Rizzoli, A., Fornara, N., Gambardella, L.M., 2002. A simulation tool for combined rail/road transport in intermodal terminals. Mathematics and Computers in Simulation 59 (1–3), 57–71. Shabayek, A.A., Yeung, W.W., 2002. A simulation model for the Kwai Chung container terminals in Hong Kong. European Journal of Operational Research 140 (1), 1–11. Silberholz, M.B., Golden, B.L., Baker, E.K., 1991. Using simulation to study the impact of work rules on productivity at maritime container terminals. Computers and Operations Research 18 (5), 433–452. Thomas, B.J., 1989. Management of port maintenance: A review of current problems and practices. H.M.S.O, London. Witness, 2000. User guide. Lanner Group. Yun, W.Y., Park, Y.M., Ryu, K.R., 1999. A simulation model for container-terminal operation analysis using an object oriented approach. International Journal of Production Economics 59 (1–3), 221–230.

Lihat lebih banyak...


Copyright © 2017 DADOSPDF Inc.