Accessibility to freight transport networks in Belgium: a geographical approach

Tijdschrift voor Economische en Sociale Geografie – 2003, Vol. 94, No. 4, pp. 424–438.© 2003 by the Royal Dutch Geographical Society KNAGPublished by Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden MA 02148, USA

ACCESSIBILITY TO FREIGHT TRANSPORT NETWORKS IN BELGIUM: A GEOGRAPHICAL APPROACH

ISABELLE THOMAS*,**,***, JEAN-PIERRE HERMIA*,

THIERRY VANELSLANDER**** & ANN VERHETSEL****

*Department of Geography, Université Catholique de Louvain, Louvain-la-Neuve, Belgium. E-mail: [email protected]; [email protected] **Centre of Operational Research and Econometrics, Université Catholique de Louvain, Louvain-la-Neuve, Belgium ***National Fund for Scientific Research, Brussels ****Department of Transport and Regional Economics, Universiteit Antwerpen, Belgium. E-mail: [email protected]; [email protected]

Received: March 2002; revised July 2002

ABSTRACTThis paper focuses on the role of transportation infrastructure in freight accessibility from asustainable and multi-modal perspective. Several accessibility measures are applied separately yetsimultaneously to water, rail and road transportation systems. It appears that the transportationsystem is structuring the Belgian space geographically and topographically (inertia of space dueto transportation infrastructure), but that economic activities are less associated with transporta-tion systems than with the population. This means that either economic activities are footloose,or that there is a major discrepancy between transportation infrastructure and economic locations.Land-use planners should be aware of this when making decisions about the mobility of the future.

Key words: Freight transport, Belgium, accessibility, externalities

INTRODUCTION

The main question in this paper is, to whatextent freight transportation infrastructureaffects space in terms of accessibility in Belgium,and to what extent accessibility measures canbe operational tools for transport and land-useplanning decision-making. The study of theinterrelationships between land-use planningand transportation is not new in urban andregional economics or in geography (see e.g.Anas et al. 1998; Banister 1995; Banister &Berechman 2001; Berechman 1994; Berechmanet al. 1996; Giuliano 1989; Rietveld 1994;Thomas 2002; Wegener 1998). Today, the press-ing nature of the environmental debate hasrenewed interest in this field of research. The

relationship between transportation invest-ment and regional economic growth is a verycomplex problem, not easily summarised byreferring to one regional economic theory oranother. One of the reasons this problem is socomplex is that transportation infrastructurehas both spatial and economic properties. Onthe one hand, transportation infrastructure hasnetwork properties, in that it has the extra-ordinary ability to shift market areas and to affectcommunication channels (Rietveld 1989). Onthe other, it provides input into the productionof private and public sector goods. Therefore,it affects the socio-economic landscape in waysno single spatial model like von Thünen, Weberor Hoover can fully anticipate (Rephann &Isserman 1994).

ACCESSIBILITY TO FREIGHT TRANSPORT NETWORKS IN BELGIUM 425

© 2003 by the Royal Dutch Geographical Society KNAG

Hence, this paper analyses transportationnetworks both topologically (the characteris-tics of the infrastructure) and in relation totransport-generating activities (consumption andproduction of goods). It simply tests to whatextent the geographical space is heterogeneousin terms of accessibility and to what extent trans-portation networks correspond to the distribu-tion of population and/or economic activities.A final question is the sensitivity of the acces-sibility measures to modal changes. Will moresustainable solutions for freight transport affectthe present spatial pattern of accessibility withina country like Belgium?

This paper presents the results of a broaderresearch programme, undertaken at the requestof the Belgian government (Beuthe & Meersman2001), which attempts to gain insight into thecomplex relationships between freight trans-port, infrastructure, external effects and econ-omic growth in Belgium. On the one hand, agroup of economists analysed the relationshipsbetween modal choice and prices (Meersman& Van de Voorde 1999), and worked on anevaluative analysis of infrastructure projectsand transport policies (Beuthe et al. 2001). Onthe other, geographers and planners focused onthe question of the optimal location of inter-modal platforms (Arnold & Thomas 1999;Arnold et al. 2000), and on the measurement ofspatial accessibility. This paper deals with thislatter aspect.

The format of the paper is as follows. Thesecond section reviews the literature aboutaccessibility and its links with consumptionand production. The third section presentsthe methodology used in this application. Thefourth section presents the results pertainingto topological and economic accessibility. Thefinal section concludes the paper and suggestssome land-use planning rules for the future.

CONCEPT AND MEASUREMENT OF ACCESSIBILITY

Accessibility is usually defined as the ease withwhich activities can be reached from a certainplace and with a certain system of transport(Morris et al. 1979). Since the seminal works(by Stewart 1947; Harris 1954 for example),this topic has recently attracted renewed atten-tion among economists, geographers and regional

planners. Undoubtedly, this has to do with themobility problems faced in most urbanisedand industrialised areas, problems caused bymassive urban congestion, rising transportationcosts, emerging external costs, growing freightflows, that is phenomena affecting sustainabilityin a negative sense. Nowadays, ‘the concept ofaccessibility is revisited’ (Martellato & Nijkamp1999). Most useful academic publications onaccessibility, date from the early 1970s onwards(see e.g. Ingram 1971; Vickerman 1974; Pirie1978; Morris et al. 1979). Note that as early as1959, Hansen had published a paper entitled‘How accessibility shapes land use’, in whichsimple potential measures were used. In spiteof this early start at the beginning of quantita-tive analysis in geography, accessibility remainsa challenge for spatial data analysis today (Foth-eringham et al. 2000, p. 245).

A review of related literature (Vanelslander& Verhetsel 2001; Hermia & Thomas 2001)reveals that at least as far as freight transport isconcerned, little has been written and is knownabout accessibility. Most publications deal withpassenger transport and are often limited tourban areas. Furthermore, examples of inte-grating demand for and supply of transporta-tion capacity or multimodal approaches areseldom found in literature. Finally, in most cases(at least for passengers), accessibility measuresare restricted to the use of distance (or a measuredirectly derived from it). Hence, in this paper,the authors intend to begin by investigatingwhether the methods used for passengers areapplicable to freight transport, and then to con-sider different input into the accessibility meas-ures, including distance, direct costs and timecongestion, as well as total population and econ-omic activity as weighting factors.

As the literature directly linked to the trans-port of goods is quite limited in terms of volumeas well as of originality, scientific publicationsthat treat non-specific indicators will be reliedupon. These indicators will then be adjustedto the characteristics of the transportation ofgoods. There are six aspects of accessibility thatlead to different definitions and hence to dif-ferent techniques by which accessibility shouldbe measured (Hilbers & Verroen 1993). First ofall, perspective is important: of the viewpoints of(i) the individual, (ii) the activity, (iii) the trans-portation system and (iv) the Government and

426 ISABELLE THOMAS ET AL.


the individual. The nature of the activity forwhich the authors want to calculate accessibilityis what determines the suitable techniques.Here, transportation generating socio-economicactivity (industrial activity and commerce) isrelevant. As motives for mobility, the authors areespecially interested in freight transportation.The interest groups in the research are the indus-trial and commercial sectors, or at least theirtransportation-generating subgroups. The trans-portation modes considered in this paper in-clude roads, railways and waterways, which arethe only relevant modes at this scale of analysisin Belgium. Finally, the levels of scale are of primeimportance in the results; there will depend ondata availability, which will be considered at alater stage.

The measurements considered by Hilbersand Verroen (1993) are based on disclosurecharacteristics, position in the network, actualaccessibility, potential accessibility, revealedpreference and activity patterns. Possible indi-cators are used in a sequence ranging fromsupply-oriented to demand-based. Althoughnowadays, indicators based on revealed prefer-ence and activity chains seem to contribute tothe progress in modelling passenger transport,no such data are available for freight trans-port. Consequently, the potential accessibilitymeasures, and more specifically the gravity-typemeasures, appeared to be the most suitable.For the gravity measure, two components areneeded. First, a denominator has to include ameasure of the affect of friction on distancetravelled. It can be expressed in different costunits such as, distance, travel time or general-ised cost of transport. The numerator has toinclude a resistance factor or an attractivenessmeasure; possible variables are population as areflection of the location of consumption andeconomic activity generating transport, reflect-ing the location of production.

METHODOLOGY

Types of accessibility measures – In this paper,general accessibility measures (Morris et al. 1979;Gutierrez et al. 1998; Pooler 1994) representingthe degree of interconnection between a par-ticular reference location i and all or a set ofother locations j in the area are studied. Thesemeasures take the following general form:

(1)

where: Ai is the accessibility of a place i wj is the weight representing the attractivenessof location j dij is a measure of separation between i and j,andf(dij) is an impedance function, generally afunction of distance, travel time or cost.

Although the impedance may take many forms,the negative exponential form has been usedmore often than other forms and is also themost closely tied to travel behaviour (Handy &Niemeier 1997). There is a long history behindthe use of these gravity-type measures.

In addition to general accessibility, a relativeaccessibility index is also computed:

(2)

where: n is the number of nodes included in thestudy area.

This is a rewriting of the preceding gravity-type index, closer to the spatial interaction for-mulation when considering several places in thespace. Despite the differences between generaland relative accessibility, mainly due to an alter-nate use of the mass of the origin area or node,we will see that the resulting spatial patterns arethe same.

Three different interpretations of theseformulas were used. First, topological accessibility,which is defined as the index obtained by exclud-ing the weights associated to the nodes (all wjvalues are set to 1). In doing so, the authorsconsidered the accessibility linked to the char-acteristics of the network only. Topologicalaccessibility reflects the degree of connectivitybetween locations (Ingram 1971); it is expressedin terms of the presence/absence of a trans-portation link, or the physical distance or traveltime or cost between locations. In this case, dis-tance in kilometres and transportation costswere considered. Second, what is here called thegeographical accessibility was computed by intro-ducing the population as a weight that reflectsthe attractiveness of a node, the populationbeing considered as a proxy for the amount of

A w f di j ijj

( )= ×∑

A

f d w

n wi

ij jj

jj

( )

=×∑

∑



consumption that requires distributive freighttransport. A variant of this geographical acces-sibility was also calculated by replacing the dis-tance in kilometres by a time-distance variableas impedance. This time distance was con-structed by using the average distance for eachtype of road or transportation mode, and byintroducing road congestion. Third, economicaccessibility was obtained by inserting as a weightin the numerator, a variable reflecting theimportance of the economic activity (wj). Econ-omic accessibility was computed both withdistance and with transportation costs as animpedance function.

All computations were made for the threemodal networks separately (roads, railways andwaterways); in addition, multimodal accessibilitywas also computed for both geographical andeconomic accessibility, integrating the threemodes in a sustainable transportation context.

Study area and data collection – The analysis wasrestricted to one country (Belgium), and todomestic transport only. Domestic transportrepresents 59% of the total tonnage and 48%of total tonne-kilometres on the Belgian road,rail and waterway networks. This percentage isthe highest for road and the lowest for rail.Undoubtedly, it would be interesting to con-duct the same kind of research at an interna-tional scale; between Belgium and neighbouringcountries, or even at a European scale.

The main data source is a graph (not shown)representing the transportation networks, withnodes and edges that can be weighted. Thefirst concern was to get digitised networks ofroad, rail and inland waterways large enough toreflect the local disparities and small enough toallow computation of indices, for statisticalanalyses and for mapping. NODUS data wereused (Jourquin 1995; Beuthe et al. 2001), thoughthe network was simplified. The characteristicsof the arcs (roads, railway line and waterways)include the maximum speed allowed, the num-ber of lanes, the mean amount of traffic theysupport, the possible presence of congestion,and an estimation of the time distance sep-arating each pair of neighbouring nodal points.Distance data and direct cost data are suppliedtogether with the network. The number ofNODUS nodes and edges were reduced in orderto obtain manageable data sets. Justification

for these selections can be found in Hermiaand Thomas (2001). This resulted in threetopological networks:

• a road network (1,068 nodes, 1,609 edges)• a rail network (222 nodes, 281 edges) and • a waterway network (146 nodes, 166 edges).

As already mentioned, characteristics of thenodes are population and economic activityvariables. Population is expressed by the totalnumber of inhabitants residing in each node;hence, population is considered as a proxy forconsumption of goods. Economic activity isharder to measure; the cadastral surface occu-pied by industrial, distributive and service activ-ities is used as a proxy in this case. This surfaceroughly estimates the production of freighttransport. Since economic data are availableonly at the commune level, the number ofnodes for calculating economic accessibility willequal 589, which corresponds to the 589 Bel-gian communities. Geographical accessibility iscomputed using 1,068 nodes because popula-tion data are available on a less aggregated level.

ANALYSIS AND RESULTS

Topological accessibility – The impedance wasfirst simply set equal to the distance expressedin kilometres. The resulting variations in thetopological accessibility indices are mapped(Figures 1 to 3). The darkest areas have thehighest values, which means that distances to allother nodes in the overall network are the low-est. As expected, the most accessible areas forroad and rail in Belgium are centred on Brus-sels; waterway accessibility is – in conformingwith the inland waterways system – highest inthe northern part of the country. For rail andwater, some areas are blank on the maps; this isdue to the lack of links to the networks considered.

The Belgian road network is quite dense andcontains a large and intensely used motorwaynetwork (Thomas & Verhetsel 1999). The infra-structure is, however, denser in the northernpart of the country due to higher populationdensity, which can be explained historically.The relative homogeneity of the road networkis especially apparent when compared withwaterway and railway networks. Topologicalaccessibility indices show that the regionbetween Brussels and Antwerp is the most



accessible (Figure 1). Nodes in the southern partof the country are the least accessible; theseare preceded in the ranking by nodes in WestFlanders (Western part of the country). Thisis mainly due to the topology of the network,to the history of the country, but also to bordereffects that were not taken into account.

The railway network has a long tradition andis dense in Belgium (van der Herten et al. 2001).The network has a polycentric structure, with afirst-order central function for Brussels. Theinfrastructure density is quite close to thepopulation demography: while much denser inFlanders than in Wallonia, the branching ismuch better in Hainaut (south-west Wallonia)and rather sparce in Campine (north-eastFlanders), although this is for historic reasonsonly (economic development, history of settle-ments). Topological accessibility indices showthat the zone north of Brussels contains thenodes that are most accessible (Figure 2).

The waterway network is quite well developed inBelgium. It is adapted to the natural topographyof the country. In addition to waterways that arenaturally navigable or adapted for this purpose,Belgium also boasts a large number of canals.Topological accessibility indices computed forwaterways show that nodal points are the mostaccessible in Flanders, and more specifically ina triangle with the Eastern border correspond-ing to the Antwerp-Brussels axis (Figure 3).

In a second step, kilometres were replaced bytransportation costs as a measure of impedance.It is indeed generally accepted that freighttransport behaviour is inspired by considera-tions over transportation costs rather than bydistance. Transportation costs include; costs forvehicle operations, handling costs, commodityinventory costs, fixed costs for vehicles andworkers during (un-)loading and transshippingoperations, labour costs for goods handling at thepoint of origin and of destination, congestion

Figure 1. Topological accessibility by road.



costs at certain points of the network and theopportunity cost of the capital tied to thegoods. This generalised cost remains incom-plete because information on relative safety,reliability and other qualitative attributes werenot available and were therefore not includedin the analysis (Beuthe et al. 2001). But usingthese cost calculations as impedance broughtus much closer to the economic considerationsof freight transport decision-makers than usingthe distance in kilometres. This time-consumingtask did not, however, reveal any major differ-ences between topological accessibility computedwith the transportation costs and accessibilitycomputed with distances. The correlation coef-ficients (n = 589) between the accessibility indi-cators were high: for the road network, r isequal to 0.957, for the rail network to 0.971, andfor waterways to 0.971. By way of conclusion, wecan state that the metric characteristics of thenetworks determine topological accessibility to

a considerable extent. Introducing a measure torepresent economic impedance demands a lotof extra work, but only results in minor andlocal differences compared with topologicalimpedance. While the introduction of real costsis very important for the economic evaluationof infrastructure projects, it is less important forthe geographical understanding of the trans-portation structures, especially in a densely-inhabited country like Belgium.

Geographical accessibility – Geographical acces-sibility was computed using the total number ofinhabitants as a weight (wj) in the gravity-typeformula. In a first set of computations, distancewas used as a measure of impedance (frictionof distance). A second form of geographicalaccessibility was then computed by replacingdistance expressed in kilometres by transporta-tion time as impedance (while population wasmaintained as wj). Transportation time was

Figure 2. Topological accessibility by rail.



obtained by combining data about every arc onthe average speed, given the type of road anda proxy for the risk of congestion on that arc.Expected theoretical congestion measures wereobtained by combining a theoretical functionand average traffic measures (Hermia & Thomas2001). Figure 4 represents the spatial vari-ation of expected congestion. Although due toa lack of data, there is no way to test to whatextent this measure estimates the observedcongestion, its geographical pattern is quiterealistic. The only objective in this case was tosee to what extent the addition of such a vari-able would modify accessibility structure.

Mean road speed conditions and automobilecongestion only slightly modify the accessibilityhierarchy established with distance: correlationcoefficients are significant, positive and quitehigh (see Table 1). This means that reinforcingnodal weights increases the nodal position of allnodes of the Brussels’ agglomeration; the same

is true for some nodes situated between Brusselsand Gent. To the contrary, nodes in the Leuvenand Waremme areas fall back in the same hier-archy. In the same way, increasing the relative

Figure 3. Topological accessibility by waterway.

Table 1. Pearson correlation coefficients between accessibilityvalues obtained with different measures of input (n = 1,433,significance level < 0.001).

wj = 1, f(dij) = km

wj = 1, f(dij) = time

wj = population,f(dij) = km

wj = 1, f(dij) = time

0.98 – –

wj = population, f(dij) = km

0.97 0.96 –

wj = population, f(dij) = time

0.96 0.98 0.97



spread between mean speed of different road types(from motorways to provincial roads) leavesus with a different hierarchy. In this case,nodes in peripheral agglomeration areas andclose to motorways in particular, saw their posi-tion improved, while nodes situated at a con-siderable distance from a motorway axis, woundup in a relatively lower position, just like thenodes in the heart of the urban agglomerations.These hierarchical changes become clear fromthe moment that congestion is taken intoaccount: Brussels then moves from the 12th tothe 24th rank in terms of accessibility in a roadnetwork comprising nearly 250 nodes in differ-ent classes.

The transportation and accessibility informa-tion mentioned in the preceding sections wereactually included in a geographical informationsystem. This enabled us to build an integratedtransportation system, in which the transporta-tion nodes and edges associated with the threemodes were included. This new graph consistsof 1,433 nodes. Each node is characterised byan accessibility measure for each transportationsystem. A weight is then added in order to take

the relative importance of each transportationsystem into account. α represents the share ofroad freight transport, β that of rail and γ thatof waterways. The new indices are standardisedin order to vary between 0 and 1; the value of1 means a very accessible node. Current marketshares for freight lead to an integrated index ofwhich spatial variation is mapped in Figure 5(α = 88%, β = 10% and γ = 2%). Given the largeproportion of freight transported by road,Figure 5 is unsurprisingly dominated by roadaccessibility and centred on Brussels; hence itis comparable to Figure 1.

For a sustainable freight transportationsystem, a better balanced solution should beadopted: α should decrease in favour of β (rail)and/or γ (waterways). However, perspectivesare more pessimistic, leading to a continuedincrease in α. Consequently, several simulationswere run letting α vary from 33% to 90%, βfrom 5 to 33% and γ from 2 to 33% respectively.Figure 6 represents one of the many simulationscorresponding to the most sustainable solution:a reduction in road and an increase in rail andwaterways. On average and on topological basis

Figure 4. Estimated road congestion.



only (no economic weight), we can concludethat globally, whatever the market shares of thedifferent transportation modes, spatial struc-ture does not change (the maps look alike); theconcentric structure of the country persists.This can be explained by the historical spreadand growth of the population and the fact thatroads represent the most important transpor-tation mode and hence dominates the spatialstructure. Most simulated solutions are highlyand significantly correlated (r varies between0.624 and 0.999). In Figure 7, an example of anaccurate comparison of two simulated solutionsis given, showing that even if maps are globallysimilar, local differences can be observed. Moreparticularly, if water and rail gain more import-ance now, then peripheral regions gain acces-sibility. The most accessible locations currentlylose accessibility; this is in fact not acceptablebecause these central regions are the mostcongested. This means that an overall Belgianpolicy of favouring freight transport over waterand rail and of penalising road transport would

worsen the situation in centrally congestedareas. Without an accompanying spatial policy,this would lead to a further devolution of econ-omic activities towards peripheral locations.This is not acceptable because of the high(public) development costs related to peripherallocations (Fujita & Thisse 2002). Furthermore,negative agglomeration effects in the currentproduction and employment areas wouldincrease and threaten economic growth in thecentral areas with the greatest concentrationsof economic activities. In other words, transpor-tation policies favouring water and rail trans-port should be accompanied by a stringentspatial policy.

Economic accessibility – The most relevantvariable representing transport-generatingpower would be the real commodity flows enter-ing and/or leaving each node. Unfortunately,spatially-detailed freight traffic data are notavailable. A proxy variable was used in thiscase: surface occupied by transport-generating

Figure 5. Geographical multimodal accessibility. Present situation (88% road – 10% rail – 2% waterway).



industries or services (these are official land-use data provided by the cadastral survey).Hence, Table 2 shows that while the correlationcoefficients computed between topological andeconomic accessibility values are significant,they are, in this case, quite low. This is particu-larly true of the road network. This meansthat the road network historically followedthe same spatial organisation as that of popu-lation (Table 1), but that the spread of economicactivities since 1960 has been controlled by‘other’ locational factors, confirming the fact

that economic activities are nowadays said to befootloose from a transportation point of viewdue to the extensive supply of roads. This canalso mean, to the contrary, that the present trans-portation network did not adapt to economiclocations, and that an enhancement of the trans-portation system might lead to a better corre-lation between economic and transportationneeds. Moreover, it has to be kept in mind thatonly the relation between the Belgian domestictransportation network and the location ofeconomic activities has been considered. The

Figure 6. Geographical multimodal accessibility. Hypothetical sustainable situation (80% road – 15% rail – 5% waterway).

Table 2. Pearson correlation coefficients between topological and economic accessibility (n = 589) (significant at the0.01 level).

wj = economic activity surfacef(dij) = km

wj = economic activity surfacef(dij) = cost

road wj = 1 0.163 0.161rail wj = 1 0.513 0.523water wj = 1 0.491 0.496



location of recent economic activities is moreoriented towards international transport gates(the ports of Antwerp, Gent and Zeebrugge andthe airport of Brussels). As a consequence, thecorrelation between the location of economicactivities and rail and water accessibility meas-ures is indeed much higher than in the case ofroad accessibility.

An interesting comparison can be drawnbetween the spatial patterns of topological andeconomic accessibility (Figure 8). Some localcommunities (nodes) are topologically and econ-omically quite accessible; this means that theireconomic activity is proportional to their rela-tive location on the network. This observationis certainly true for certain places in the area ofBrussels. The same proportion is observed forsome communities with low topological valuesand low economic accessibility; this is true formany communities beneath the Sambre-et-Meuse axis. More interesting are communitieswith low topological values, but relatively high

economic accessibility, such as certain communi-ties in West Flanders. This means that in spiteof their peripheral position on the transpor-tation network, these places attract a relativelylarge number of economic activities. On thecontrary, there are communities with hightopological values, but rather low economicaccessibility indices (for example, certain com-munities in the province of Liège). These areasattract relatively few economic activities giventheir relative location in the freight transporta-tion network. The fact that economic activitiesare largely represented (high economic acces-sibility) in an area with poor network connec-tions (low topological accessibility) and viceversa, means that other location-related factorsplay a more important role than their mereposition in the network.

The data on economic accessibility for thedifferent freight transport networks (road, rail,water) are highly correlated due to the hugeimpact on the surface for economic activities as

Figure 7. Geographical multimodal accessibility. Differences between present situation and hypothetical sustainablesituation. (Value > 1: present accessibility > sustainable accessibility).



impedance. To get a synthetic image of multi-modal economic accessibility, a cluster analysiswas applied (hierarchical ward grouping proce-dure). In this analysis, four variables were usedas discriminant criteria: three variables measur-ing the topological accessibility associated toeach transportation mode (road, rail, water)and a variable that reflects the amount of econ-omic activity. In doing so, we obtained groupsof communities with the same position in thethree freight transportation networks and withthe same characteristics for economic activities.

Four clusters of municipalities were obtainedand mapped (Figure 9).

1. Cluster 1 consists of communities withrelatively low scores on all variables exceptfor topological accessibility by road, whichis very high. This cluster consists mainly ofcommunities in traditionally rich agrarianregions that are now integrated in periurbanor suburban areas (especially around Brussels).

2. Cluster 2 contains communities with rela-tively positive scores on all variables, espe-cially on topological accessibility by rail. They

correspond to traditional industrial areas,rooted in the 19th century; most largeurban regions are included in this cluster.These communities are the best suited forsustainable policies that promote the use ofrail for freight transport.

3. Cluster 3 groups communities with low rateson all variables; this means that the overallaccessibility is very low and subsequentlythat economic activities are absent. Theregion corresponding to the area south of theWalloon industrial axis, to the central part ofWest Flanders and to many Flemish commu-nities along the border with the Netherlandsare included in this cluster. From a Belgianperspective, these are deprived areas interms of freight transportation and economicactivities, but are often part of the mostbeautiful natural areas of the country withlow population densities.

4. Cluster 4 communities are characterised byaverage values on all variables, except foraccessibility by water which is very high. Thespatial pattern reflects the stretches of naturalwaterways and canals that are not well served

Figure 8. Road accessibility weighted by surface economic activities.



Figure 9. Cluster analysis by surface economic activities and accessibility (road, rail, waterway).



by rail or road. Policies stimulating freighttransport by inland waterways can thereforehave effects in these municipalities, but un-fortunately infrastructures for intermodalitywith rail are lacking.

CONCLUSION

‘Sustainable mobility’ has been adopted as anoverall objective for European transport policyand similar intentions have been expressed inother parts of the world. Defining sustainablemobility for passengers and goods is, however,a difficult task that requires a multitude ofapproaches. In this paper, the focus is limitedto freight transportation and accessibility. Thedifferent transportation modes and differentaccessibility measures are compared, leadingto the conclusion that accessibility analysis is astrong tool for integrating spatial and networkcharacteristics.

With the example of Belgium, the authorsshow that transportation infrastructure stronglyshapes geographical space and that this struc-ture is highly dependent upon the history of thecountry (high spatial inertia). This is true forthe topology of the area studied, which is almostequal to the spatial distribution of population,however the spatial pattern of economic acces-sibility differs. Indeed, sensitivity analyses haveshown that changing the measures of imped-ance does not generate large differences in thespatial structure of accessibility. In the samesense, weighting the nodes by population doesnot lead to a different spatial structure than thatproduced by the topological structure: the dis-tribution of population and freight transpor-tation system fit spatially. However, weightingthe nodes by importance of economic activitydoes matter: the transportation system and econ-omic land-use do not fit spatially. This results inthe emergence of favoured and penalised areasin terms of accessibility

While these conclusions are valid nationally,differences have been observed locally. Thisdemonstrates that applying global policiesabout sustainable mobility of goods will gener-ate local differences that will need controlling.

In short, the empirical results obtained in thispaper reveal that while transportation and econ-omic investments may seem complementary,they are both embedded in their political

context and do not always converge. On the onehand, spatial patterns of accessibility seem fairlyinert, in that only stringent spatial policiescould possibly change the overall accessibility inthe long run. On the other hand, short-termeconomic policies (e.g. road pricing, trafficmanagement, regulations) could have impor-tant local consequences. The locational prob-lems associated with economic activities andwith the mobility of commodities are questionsthat require joint treatment by policy-makers.Unfortunately, governmental agencies havethus far, shown only a limited ability to coordin-ate their transportation and economic activitiesand to design complementary policies. Thiscould prove a challenge for the future ofregional economic development.

Acknowledgements

The authors are indebted to the Belgian FederalResearch Department, O.S.T.C. (contract no. MD/10/003) for financing this research project. Theywould also like to thank B. Jourquin for providingdata and the anonymous referees for their usefulcomments on a draft version of this paper.

REFERENCES

Anas, A., R. Arnott & K. Small (1998), UrbanSpatial Structure. Journal of Economic Literature 36,pp. 1426–1464.

Arnold, P. & I. Thomas (1999), Localisation desCentres de Transbordement dans un SystèmeMulti-réseaux: Essai de Formalisation. L’EspaceGéographique 28, pp. 193–204.

Arnold, P., H. Marchand, D. Peeters & Thomas I.(2000), Pour une Localisation Optimale desCentres de Transbordement Intermodaux EntreRéseaux de Transport: Formulation et Extensions.The Canadian Geographer/Le Géographe Canadien 45,pp. 427–436.

Banister, D. ed. (1995), Transport and Urban Develop-ment. London: Spon.

Banister, D. & Y. Berechman (2001), TransportInvestment and the Promotion of Economic Growth.Journal of Transport Geography 9, pp. 209–218.

Berechman, J. (1994), Urban and Regional Impactsof Transportation Investment: a critical assess-ment and proposed methodology. TransportationResearch A 28, pp. 351–362.



Berechman, J., H. Kohno, K. Button & P. Nijkampeds. (1996), Transport and Land Use. Modern Classicsin Regional Science: 2. Cheltenham: Edward Elgar.

Beuthe, M. & H. Meersman (2001), Transport andMobility: 10 Years of Research. Brussels: OSTC, FederalOffice for Scientific,Technical and Cultural Affairs.

Beuthe, M., B. Jourquin, J. Geerts, A. Koul & C.Ndjang Ha (2001), Freight Transport DemandElasticities: A Geographic Multimodal Trans-portation Network Analysis. Transportation ResearchE 37, pp. 253–266.

Fotheringham, A., C. Brunsdon & M. Charlton(2000), Quantitative Geography, Perspectives on SpatialData Analysis. London: Sage.

Fujita, M. & J. Thisse (2002), Economics of Agglomera-tion. Cities, Industrial Location, and Regional Growth.Cambridge: Cambridge University Press.

Giuliano, G. (1989), New Directions for Under-standing Transportation and Land Use. Environ-ment and Planning A 21, pp. 145–159.

Gutierrez, J., A. Monzon & J.M. Pinero (1998),Accessibility, Network Efficiency, and TransportInfrastructure Planning. Environment and PlanningA 30, pp. 1337–1350.

Handy, S. & D. Niemeier (1997), Measuring Acces-sibility: An Exploration of Issues and Alternatives.Environment and Planning A 29, pp. 1175–1194.

Hansen, W. (1959), How Accessibility Shapes LandUse. Journal of the American Institute of Planners 25,pp. 73–76.

Harris, C. (1954), The Market as a Factor in theLocation of Industry in the United States, Annals ofthe Association of American Geographers 44, pp. 315–348.

Hermia, J.-P. & I. Thomas (2001), Evaluation Spatialedes Réseaux de Transport de Marchandises en Belgique.Bruxelles: Rapport final adressé aux ServicesFédéraux des Affaires Scientifiques, Techniques etCulturelles (SSTC).

Hilbers, H. & E. Verroen (1993), Het Beoordelen vande Bereikbaarheid van Lokaties – Definiëring, Maatstaven,Toepassing en Beleidsimplicaties. Delft: TNO Inro.

Ingram, D. (1971), The Concept of Accessibility: ASearch for an Operational Form. Regional Studies 5,pp. 1001–1107.

Jourquin, B. (1995), Un Outil d’analyse Économiquedes Transports de Marchandises sur des RéseauxMulti-modaux et Multi-produits: le Réseau Virtuel,Concepts, Méthodes et Applications. Mons: FacultésUniversitaires Catholiques, PhD Thesis.

Martellato, D. & P. Nijkamp (1999), The Concept ofAccessibility Revisited. In: A. Reggiani ed., Accessibility,Trade and Locational Behaviour. Aldershot: Ashgate.

Meersman, H. & E. Van De Voorde (1999), IsFreight Transport Growth Inevitable? In: EuropeanConference of Ministers of Transport, ed.,Which Changes for Transport in the Next Century?pp. 23–48. Paris: OECD.

Morris, J., P. Dumble & M. Wigan (1979),Accessibility Indicators for Transport Planning.Transportation Research A 13, pp. 91–109.

Pirie, G. (1978), Measuring Accessibility: A Reviewand Proposal. Environment and Planning A 11,pp. 299–312.

Pooler, J.A. (1994), The Use of Spatial Separationin the Measurement of Transportation Accessi-bility. Transportation research A 29, pp. 421–427.

Rephann, T. & A. Isserman (1994), New Highwaysas Economic Development Tools: An EvaluationUsing Quasi-experimental Matching Methods.Regional Science and Urban Economics 24, pp. 723–751.

Rietveld, P. (1989), Infrastructure and RegionalDevelopment. A Survey of Multiregional Econ-omic Models. The Annals of Regional Science 23,pp. 255–274.

Rietveld, P. (1994), Spatial Economic Impacts ofTransport Infrastructure Supply. TransportationResearch A 28, pp. 329–341.

Stewart, J. (1947), Empirical Mathematical RulesConcerning the Distribution and Equilibrium of thePopulation. Geographical Review 37, pp. 461–485.

Thomas, I. (2002), The Shape of the TransportationNetwork and the Optimal Location of Human Activities.A Numerical Geography Approach. Cheltenham:Edward Elgar.

Thomas, I. & A. Verhetsel (1999), Transport Routieret Mobilité Durable: un État de la Question pourla Belgique. Hommes et Terre du Nord 3, pp. 198–203.

Van Der Herten, B., M. Van Meerten & G. Verbeurgt(2001), Sporen in Belgie: 175 jaar spoorwegen, 75 jaarNMBS. Leuven: Universitaire Pers Leuven.

Vanelslander, T. & A. Verhetsel (2001), IntegraleEvaluatie van de Band Tussen Ruimelijk-economischeStructuur en Infrastructuur voor Goederenvervoer.Antwerp: UFSIA, Onderzoeksrapport 2001–2015.

Vickerman, R. (1974), Accessibility, Attraction andPotential: A Review of Some Concepts and TheirUse in Determining Mobility. Environment andPlanning A 6, pp. 675–691.

Wegener, M. (1998), Applied Models of Urban LandUse, Transport and Environment: State of the Artand Future Developments. In: L. Lundqvist,L. Mattson & T. Kim, eds., Network Infrastructureand the Urban Environment, pp. 245–267. Berlin:Springer Verlag.

Accessibility to freight transport networks in Belgium: a geographical approach

Documents