|
|
Scenario-wise analysis of
transport and logistics systems
with a SMILE
dr. ir. L.A. Tavasszy *
ir.M.J.M. van der Vlist *
prof. drs.C.J. Ruijgrok * ***
ir.J. van der Rest **
With the dynamics of global logistics processes in mind, private and public decisionmakers regularly require strategic information on the expected development of freight flows. The article reports on the development and the first applications of a new strategic model for freight transport and logistics in and around the Netherlands. The need for such a model has emanated from discussions on freight transport modelling that were started in the early nineties. In these discussions it was concluded that a model for policy analysis on the long term, with a low level of detail but with a wide scope was missing. Modelling with a wide scope would involve a treatment of not only sectoral and spatial exchanges together with transport processes but in addition also the logistics processes underlying freight transport.
This implies that the relations between transport and the economy have to be well understood, including the appropriate feedbacks between markets for goods and services at different levels. The development of decision support systems should aid this understanding, in particular when strategic management or public policy analysis requires a systematic sketch of the impacts of certain changes in the system. Here, the term "logistic and transport system" should be clarified. It denotes the economic activities underlying the complete supply chain, including production, sales and sourcing, inventory and transport. Note also that the use of "system" implies that the information about these activities should concern structural issues (about the elements of the system like infrastructure and regions), functional issues (related to economic choice behaviour) and issues related to the dynamics of the markets concerned. The policy issues are treated in more detail in the following section.
The paper describes the background and the intermediate results of a project aimed at the development of a new decision support system (DSS) for public and private decision makers in the transport and logistics sector. The model is being constructed as a joint effort of the Transport Research Centre of the Ministry of Transport and the research organisations NEI (Netherlands Economic Institute) and TNO Inro. The name of the model is SMILE (Strategic Model for Integrated Logistic Evaluations). We discuss the applicability of this model for exploring the possible effects of transport policy and changes in the logistic organisation of firms and through these on the corresponding freight flows. Also, the design of the information system is explained, covering the specification of the underlying models, the graphical interface by means of which scenarios in SMILE can be prepared, and the databases.
What will be the specific information needs of policy analysts and decision makers in transport and logistics? We will try to formulate an answer by exploring in short the main topics in policy analysis in this area. Before we do this, however, a number of starting points should be identified which characterise the scope of the information provided (and not so much its contents). These relate to the type of decision maker who is in need of information and the stage of decision making which is supported. From these two premises we can give a definition of a DSS for the case of policy analysis in logistics and transport.
Note that information needs differ for the private and public sector in terms of the uncertainties involved, the instruments available and the performance measures. The instruments available to the private sector will primarily focus on direct (des)investment, whereas governments can also resort to fiscal and regulatory policy. Typical for the needs of the public sector is the heterogeneity of the goods and the stakeholders studied, and the system’s complexity and uncertainty with regard to changes in the (economic) environment (see e.g. Patton ans Sawicki, 1986). In this paper, we focus on information needs for public policy analysis.
The flow of information through the decision making process can be looked at in a number of stages, which are:
Depending on the stage of decision-making in which supporting information is given, the nature of the information will be different.
Typically, in practice, a DSS will fulfil a role in all three stages of the decision making process by complementing the information available. In the strict sense a DSS will play its main role in the third stage and will be normative in nature, i.e. provide a solution to the problem. However, the view of a DSS as a management information system in the wider sense implies a focus on complex problems, with a wider scope and, usually, a lower level of detail.Some argue that the impact of DSS on organisations could be increased if, beside analytical support information, conceptual activities to support lawyers, policy advisors and consultants were provided (Taylor and Weaver, 1991). As within the area of policy DSS usually are not used within a production or managerial decision making environment, they should therefore even "move away from the monitoring of trends, the modelling of probable outcomes, the quantitative assessment of alternative options".
This point of view is clearly not being taken in this paper. It may serve to illustrate, however, the range of alternative requirements that information systems, and DSS in particular, can be confronted with. The type of DSS that we will propose focuses on an analytical processing of quantitative information related to freight flows, to be produced within a fairly complex set of assumptions (regarding hundreds of variables) that concern the economy in general and the transport and logistic systems in particular. Therefore, we will follow Ginzberg and Stohr's (1982) definition of DSS as "...a computer-based information system used by decision makers to support their decision making activities in situations where it is not possible or not desirable to have an automated system perform the entire decision process". As a comment on this definition we note that in practice the users will usually be an agent or staff member supporting the decision maker. Figure 1 illustrates the different types of information that are required to fulfill this task; the shading indicates the type of information system we will be discussing in this paper.
Figure 1 Classification of DSS for logistics systems analysis
| type of information provided | descriptive | prescriptive |
| Conceptual qualitative | e.g. reasoning models | e.g. diagnosis systems |
| analytical quantitative | e.g. SMILE | e.g. routing/ scheduling optimization |
Decision makers have to take in account a large number of factors influencing their decision and a large number of factors influenced by the decisions. This is true for a decisions in several fields, but especially in the field of traffic and transport, as transport is dependent on economical developments (growth and spatial organisation of economic activities). Two key issues can be distinguished when it comes to decision support using descriptive information:
Essential for a model is the notion that developments in freight flow demand are the result of changes in economic structures that create a demand and a supply of goods in specific geographic regions and form the basis for transport flows between regions.
The general aim for the SMILE model is to get a better view on future developments in freight flows that use Dutch infrastructure. Therefore we take two kinds of freight flows into account, each of which we treat differently.
The first category of flows is a direct result of the production structure within the Netherlands and the logistic organisation of flows that exists between these production units. The second category holds the same but these have the extra choice possibility of not using Dutch infrastructure and therefore have to be treated differently. The logistical organisation is defined as the way the goods flows are controlled, both as necessary activities to make production activities possible as well as all the activities necessary to fulfil the demand by customers.
Figure 2 Global modal structure of SMILE
In general, a distinction can be made between product logistics and transport logistics. Product logistics has to do with the control of good flows from basic products, via inventories and intermediate production activities till the physical distribution of final products to the customers. This is visualised in figure 1. This figure shows that the whole logistic process consists of several repetitious activities involving the basic activities Production, Inventory and Transport. Transport logistics involves the optimisation of the organisation of freight flows so that the utilisation of transport equipment is optimised, considering costs and quality elements such as reliability and speed.
Figure 3 Relationship between PIT and Product Logistics
With SMILE, some first steps have been taken to resolve the shortcomings in knowledge about the relationship between production, inventory and transportation by a three level chain modelling approach. The first level concerns the linkage between manufacturing activities within product chains, the second concerns the modelling of stockholding within distribution chains and the third relates to the movement and transshipment of goods within (multimodal) transport chains. Below we explain how these three levels have been operationalised in the model.
Production, Sales and Sourcing
At the first level, Make/Use Tables create the opportunity to set up a production function for each sector (if the sector produces only one commodity group for each commodity). In contrast to the prevailing models of sectoral exchanges of goods, in the SMILE model the conventional Input/Output modelling approach has been abandoned. The available Make/Use tables provide a detailed insight into the production factors connected to the activity of each sector, including the commodities that are produced and consumed. By linking production and consumption, product chains can be built that, when combined further, result in production networks.
One of the main features of the SMILE model is the application of these production functions in the assessment of the effects of a change in final demand for one product group on all the others, through the production network. This new approach allows to trace e.g. the effects of a 20% replacement of steel by composites in the car manufacturing industry on the volumes of freight flows. Moreover, the Make and Use Tables are very helpful in establishing the location pattern of both production and consumption.
After having determined the volume and nature production and consumption at different locations, the spatial distribution of flows between these locations is calculated. This spatial distribution results from the sales and sourcing processes at each location. As trade theory tells us it is a result of comparative price differences and the resistance of geographical, organisational and institutional differences that have to be bridged. This part of the model provides the user with the trade relations in connection with the Netherlands.
Inventory
The main function of the second level is to link trade relations to transport relations by considering warehousing services. Here, distribution chains are described by a logistic choice model. Several configuration options for distribution chains are investigated (see figure 2), which are characterised by the number and location of distribution centres.
Figure 4 Optional distribution structures
In the choice model, total logistic costs are calculated that account for handling, inventory and transport costs for homogeneous product categories denoted as logistical families. A survey on product characteristics has been undertaken to support the model with real-life data; its results clearly show the relevance of distribution chains for modelling freight transport demand, as for almost half of the 300 products investigated, the same type of distribution centre (continental, national or both) was in use.
Following the survey, new groups of products and market conditions were identified in which it is reasonable to assume that within these groups logistic choice behaviour is homogeneous. At present logistic families are distinguished using the following product and market characteristics (see also the figure below):
The authors would argue that in aggregate freight transport system modelling, this issue has not received the attention it deserves, particularly if one looks at the ongoing trend of "reconfiguring European logistics systems". The existing aggregate modelling approaches that combine production and network modelling do not represent this second level. Neglecting this level may adversely affect the accuracy of modelled freight flows as well as the policy sensitivity of the model.
Transportation
At the third level, a multi-modal network for 6 modes of transport is available in SMILE. It is a strategic network which means that only the network structure is modelled. Not all alternative links between regions are visible to the user, but only one representing all alternatives.
Figure 5 Multimodal network
The underlying network however (that is used for calculations) is modelled in analogy with the physical network. The figure above illustrates this network type for three modes. The optimal route in SMILE will be sought for every commodity type, and for every element in the logistics chain. The route choice disutility has a mode abstract (see Baumol and Vinod, 1966), weighted cost formulation where a combination is made of the physical distribution costs and time spent during transportation.
The weights which are used are specific per goods type and reflect the opportunity costs of the resource "time" within the entire logistics process. Comparing this model with existing approaches in physical distribution (see e.g. Goss, 1991, Higginson, 1993; Mc.Ginnis, 1989; Jourquin, 1995, Vieira, 1992) this formulation is new within this complex setting (using empirical values for 50 logistical families, reflecting total logistic costs including estimated random preferences).
Dynamics
Do we wish to give a forecast for one point in time or do we need information on the growth patterns of system variables? Insight in growth patterns gives insight in the existence of saturation levels and equilibrium mechanisms. Although two scenarios may produce similar results for a specific forecast year, they may show a completely different development in the years before and after the forecast year. The understanding of these dynamics is at least as important as the exact prediction of the levels in the horizon year under study.
Figure 6 Simulative modelling approach
The above stages of production, distribution and transportation form three steps in the calculation process that are repeated in a cyclical simulation process, as illustrated in the figure above. Each simulation cycle represents a time span of one year; a complete simulation results in time series data related to transport flows and costs. Naturally, the intermediate output concerning economic activities concerning the exchange of goods and the logistics service markets are obtained in a similar way and stored in the SMILE database.
The above examples give an idea of the structure of the SMILE model. In this section we will treat in more detail the possibilities for interaction between the DSS and the user. In order to use the model, its should provide sufficient possibilities to specify and analyse different scenarios, related to socio-economics as well as transport and logistics. Therefore, attention has been devoted to modules which smooth the input/ output interaction between user and the system (a scenario module and a presentation module) a database management system and the graphical interface. The figure below shows the relations between the model system and the user. The transport and economy modules form the logistics model and have been discussed in the previous sections; below we concentrate on the other facilities of SMILE.
Figure 7 Relations between SMILE and the user
Database
The database was implemented in Microsoft Access and can be accessed in three different ways: firstly, the processing of data for the calculations; secondly, through the graphical interface and thirdly, data can be accessed by means of a number of commercial database packages. The relational database contains data on:
Scenario design
By means of the graphical user interface, the DSS assists the user with designing his/her own scenarios for simulations up to 40 years ahead and shows the impacts of policy measures on freight flows and the environment. Scenarios can be stored as a complete database of all variables used by the model, i.e. including a complete specification of the modelled world of the user. In order to assist the user in his steps with specifying the transport system for a 40 year period, reference scenario’s were built. When specifying a scenario the user is guided in choosing variables by means of selection screens and input screens, where the baseline or reference scenario’s chosen are shown (figure 9, see the horizontal line).
These reference scenarios are derived from general economic scenario's as have been developed by the Dutch Central Planning Agency (CPB). They represent the economic environment in which the freight demand model is further specified. The user is assumed to define certain policy options or exogenous interventions in the system that create specific changes in the cost structures and choice options available. These exogenous changes are checked upon internal consistency and then used to create a scenario, if available, using outcomes of earlier scenario calculations as reference.
Figure 8 Specification of yearly changes in variables
A (geo)graphical interface
SMILE operates within a Windows 95 or Windows NT environment. Beside the standard Microsoft interface, specific applications were developed to aid the user in specifying the input to the calculations. In order to picture regions and networks throughout the world, a geographical interface is required. The main development concerns an application that compares to a Geographic Information System (GIS) environment. In fact, the main practicality of a GIS, an integrated treatment of geographical database with processing and visualisation facilities is available in SMILE, not in the least due to the relational structure of the database. Figures 10 and 11 give a view on the features that will be provided by these applications.
Figure 9 Interface to region selections
Figure 10 Interface to transport network selections
Recent research in the Netherlands has indicated that existing models - although fairly sophisticated in the sense of the detailed description of the transport system - lack policy sensitivity in terms of the ability to be used for calculating the impacts of alternative policy measures (Tavasszy, 1994). In order to support policy analysis, one should be able however, to intervene at certain points in the model that reflect real world elements like infrastructure, regulations, or services.
One of the first applications of SMILE initiated in the year 1998 concerns preparatory work for a new Transport Structure Plan of the Netherlands, in assignment of the Dutch Ministry of Transport, Waterways and Public Works. SMILE is being applied to calculate expected freight flows within, and connected to the Netherlands, until the year 2030. These flows are one of the key outcomes of three background scenarios for the socio-economic development of the Netherlands.
Policy issues
The policy questions involved range from issues like: ‘what are the effects of economic growth on transport and infrastructure needs’ to what influence has the ‘Channel Tunnel for Dutch transport’. The information needs differ between private companies and public policy makers. Private companies are interested in questions like where should I plan my warehouse activities, what is the optimal distribution structure, how will globalization influence my business. What ports should I use to import my basic materials? These questions tend to have a limited time horizon (mostly not longer than five years) and ask for very direct answers. Public policy makers often use a longer horizon and are interested in effects on society. In fact the policy question can be divided in levels. A few examples of policy questions are listed below:
It should be noted that in policy making a shift can be observed from point forecasts in year t (the volume of road transport in 2010 is 800 mln. ton) to a more scenario like approach. The main elements are:
Policy issues and transport modelling
From the above mentioned policy questions a large number can be quantified and answered with models like SMILE. Key issues which require innovations in modelling include:
Figure 11 Example of output on regional logistics activities from SMILE
Specific handles for policy instruments in SMILE
The scenario and import/export modules allow changes to be made in several variables related to prices, capacities, service performance levels or the users' preferences in the system. A listing of the policy variables is given below:
These variables can be changed per year and per commodity group (at the product or logistical family level), where applicable.
The output of SMILE relates to
The conclusions of this paper are as follows:
References
Ginzberg, M.J. and E.A.Stohr, Decision Support Systems: Issues and perspectives, in: J.Ginzberg, W. Reitman and E.A.Stohr (eds.), Decision Support Systems, North Holland, 1982
Goss (1991), The application of the concept of generalized transport costs to freight, in: Smith, H.G. (ed.), Proceedings of the European Transport Colloquium, European Transport Colloquium Foundation, Delft.
Jourquin, B.A.M. (1995), Le réseau virtuel,concept, méthodes et applications [The virtual network: concept, methods and applications], Ph.D. Thesis, Facultés Universitaires Catholiques de Mons, Mons, Belgium.
McGinnis, M.A. (1989), A comparative evaluation of freight transportation choice models, Transportation Journal, Winter, pp.36-46.
NEI, TNO/INRO, Rotterdam Transport Centre (1992), De rol van het goederenvervoer in de economie [The role of freight transport in the economy], Nederlands Economisch Instituut, Rotterdam.
Patton, C.V. and D.S. Sawicki (1986), Basic methods of policy analysis and planning, Prentice-Hall, Englewood Cliffs.
Quandt, R.W, W.J. Baumol (1966), The demand for abstract transport modes, Journal of Regional Science 22, pp.159-177.
Tavasszy, L.A., Characteristics and capabilities of the Dutch freight transport system models, RAND, Santa Monica, 1994
Taylor, A.D. and J.C. Weaver, Decision support for people who don't make decisions, in: Sol, H.G., J.Vecseny, Environments for supporting decision processes, Elsevier, North-Holland, 1991
Timmermans, H., A. Borgers (1985), Spatial choice models: fundamentals, trends and prospects, Eindhoven University of Technology, Eindhoven.
Vieira, L.F.M. (1992), The value of service in freight transportation, Ph.D. thesis, Massachusetts Institute of Technology, Boston, Mass.
UFSIA, De Nederlandse goederenvervoermodellen, SESO, University of Antwerpen, 1991
Published in the Urban Mobility Professional