Ontwikkeling van een multimodaal goederenvervoermodel voor ...

Vrachtmodel Vlaanderen Intermediate report :generation model and distribution model – P763 1

September 2006

1. GENERATION MODEL

1.1. Purpose and principle

This generation step aims at calculating the global volume of goods that is emitted and attracted by each european zone, differentiated by kind of goods (10 NST categories) :

• The traffic emitted by one zone corresponds to the sum in tons of all the shipments transported from the different sites of production or warehouses located in the regarded zone.

• The traffic attracted by one zone corresponds to the sum in tons of all the shipments transported to the different sites of production or warehouses located in the regarded zone.

The zonal system that is used for the generation model is the base zoning, that is to say the NUTS3 level in Belgium (43 arrondissements).

Figure 1.1 : base zonal system


September 2006

The general principle of the modelisation consists in considering that the emissions and attractions are directly linked to the intrinsic economic characteristics of the regarded zones expressed in socio-economic figures.

The purpose of the generation model is to explain the interdependence between socio-economic data and emitted/attracted volumes with a mathematical function. Obviously, this function is not the same for the productions and for the attractions, and it also differs for each commodity. Consequently, the model must be differentiated by direction (emission/attraction) and by commodity.

1.2. Methodology

Since it is very difficult to obtain socio-economic data for the whole Europe, these equations are only used to calculate the Belgian domestic emissions/attractions (i.e. from Belgian zones to Belgian zones). The other flows are calculated by using annual growth rates that differ between the kind of goods and the kind of traffic.

The following table sums up the methods that have been used to calculate all the emissions and attractions of the model.

Table 1.1: Methods for calculating emissions and attractions

These three methods are described in the following sections.

The model structure enables more than just the pure generation. It actually enables to :

• grow the socio-economic data (population and employment) until the selected forecast year.

• take the evolutions of the productivity and the consumption of households into account.

• correct some localised errors after having modelled the Belgian emissions and attractions, either to correct some figures in the base situation or to take specific forecasted changes into account.

Zone Type of flow Method

Belgium Domestic Linear formulations

Belgium Import and Export Use of average annual growth rates

Abroad Import, Export and Transit

Use of the BIP's growth rates of the

destination countries and use of

attraction elasticities


September 2006

1.3. Domestic emissions and attractions

PRINCIPLEPRINCIPLEPRINCIPLEPRINCIPLE

The modelling of the domestic emissions and attractions is based on linear formulations that take the socio-economic context into account. Thus, the evolution of the emission and attraction figures is directly linked to the evolution of the socio-economic data, at least after the first step of the modelling. Some other parameters like the productivity or the households consumption are included to balance this close link and hereby avoid some misinterpretations. Eventually, a multiplicative pivot is applied to the base emissions and attractions.

EVOLUTION OF THE SOCIOEVOLUTION OF THE SOCIOEVOLUTION OF THE SOCIOEVOLUTION OF THE SOCIO----ECONOMIC DATAECONOMIC DATAECONOMIC DATAECONOMIC DATA

As far as a future situation is concerned, the first step of the forecast calculation in the generation model consists of estimating the evolution of the socio-economic data. This evolution is based on a function using the Average Annual Growth Rate (AAGR) of each variable. The following equations are used :

Evolution of the population :Evolution of the population :Evolution of the population :Evolution of the population :

n

pop

BaseScenario

TPopulationPopulation

+⋅=

1001

Evolution of the employmentEvolution of the employmentEvolution of the employmentEvolution of the employment (per employment category) :(per employment category) :(per employment category) :(per employment category) :

n

empl

BaseScenario

TEmploymentEmployment

+⋅=

1001


September 2006

where :

ScenarioPopulation = Population in scenario situation

ScenarioEmployment = Employment in scenario situation per employment category

BasePopulation = Population in base situation

BaseEmployment = Employment in base situation per employment category

n = Number of years between base situation and scenario situation

popT = AAGR of the national population

emplT = AAGR of the employment per employment category

PRODUCTIVITY AND HOUSEHOLDS CONSUMPTIONPRODUCTIVITY AND HOUSEHOLDS CONSUMPTIONPRODUCTIVITY AND HOUSEHOLDS CONSUMPTIONPRODUCTIVITY AND HOUSEHOLDS CONSUMPTION

The socio-economic data must be corrected to take into account evolutions that could affect the volumes of goods emitted and attracted. Actually, a diminution of the number of employees in a sector does not necessarily result in a diminution of the corresponding volumes, even on the contrary in some cases. As we cannot get the average annual growth rates of the volumes emitted and attracted by each employment category – which would be optimal for the forecast side of the model – the productivity and the households consumption are introduced. The correction functions are the following :


September 2006

SocioSocioSocioSocio----economic data linked to the populationeconomic data linked to the populationeconomic data linked to the populationeconomic data linked to the population ::::

n

ratio

n

cons

ScenarioScenario

TTPopulationPopVar

+

+⋅=

1001

1001)(

SocioSocioSocioSocio----economic data linked to the employment (per employment category) :economic data linked to the employment (per employment category) :economic data linked to the employment (per employment category) :economic data linked to the employment (per employment category) :

n

ratio

n

prod

ScenarioScenario

TTEmploymentEmplVar

+

+⋅=

1001

1001)(

where :

ScenarioPopulation = Population in scenario situation

ScenarioEmployment = Employment in scenario situation per employment category

ScenarioPopVar )( = Socio-economic variable linked to the population in scenario situation

ScenarioEmplVar )( = Socio-economic variable linked to the population in scenario situation per employment category

n = Number of years between base situation and scenario situation

consT = AAGR of the households consumption

prodT = AAGR of the productivity per employment category

ratioT

= AAGR of the ratio value/ton :

• per employment category for the employment

• global for the population


September 2006

LINEAR FORMULATIONS OF THE EMISSIONS AND OF THE ATTRACTIONSLINEAR FORMULATIONS OF THE EMISSIONS AND OF THE ATTRACTIONSLINEAR FORMULATIONS OF THE EMISSIONS AND OF THE ATTRACTIONSLINEAR FORMULATIONS OF THE EMISSIONS AND OF THE ATTRACTIONS

The socio-economic data that have been calculated in the last step are now used in the generation model.

The generation model results of the estimation of the following generation functions :

Emitted trafficEmitted trafficEmitted trafficEmitted traffic : EM [zone i, NST x, t0] = fx-e[socio-eco (i,k,t0), ße(x,k)]

Attracted trafficAttracted trafficAttracted trafficAttracted traffic : AT [zone i, NST x, t0] = fx-a[socio-eco (i,k,t0), ßa(x,k)]

where :

EM [i, x, t0] (AT [i, x, t0])

= Annual volume in tons emitted (attracted) by the zone i, for the category of goods x, at the instant t0

socio-eco (i, k, t0) = Socio-economic variables (k variables) standing for transport generating activities in the zone i, at the instant t0

ße(x,k) (ßa(x,k))

= Parameter applied to each variable (k variables) in the equation of generation, for the category of goods x, for the emission (attraction)

fx-e[...] (fx-a[...])

= Function of generation of the volume of goods emitted (attracted) per commodity

The equations of emission and attraction are different for each commodity because they obviously do not depend on the same socio-economic variables (e.g. the agricultural products generated by one zone are linked to the agricultural figures of this zone, but the emission of metal products depends on the metal industry of the concerned zone).

The following table shows the socio-economic data that have been used in the model. These are population and employment figures split in different economic activities (NACE typology), available for Belgium on the NUTS3 level (arrondissements) for base year 2004.


September 2006

Table 1.2 : Socio economic base data used in the generation model

The parameters ß(x,k) that are used in the generation function have been calculated by using the least squares method. They fit the observed figures of emission and attraction and correspond to logical links between fields of activities and commodities. They are differentiated by commodities and by direction. These parameters are available in the annex.

POP Population

A0 Agriculture

B0 Agricultural and food industry

C1 Clothing industry

C2 Printing industry

C3 Pharmaceutical industry

C4 Household equipment

D0 Automotive industry

E1 Shipbuilding, aircraft industry, railway equipment industry

E2 Mechanical industry

E3 Electric and electronic equipment

F1 Minerals

F2 Textile industry

F3 Wood and paper industry

F4 Chemical industry

F5 Metal industry

F6 Electric and electronic industry

G1 Production of combustible and fuels

G2 Water, gas, electricity

H0 Construction industry

J2 Wholesale

K0 Transport

R0 Other service industry

S0 Administration

POPULATION 2004

EMPLOYMENT 2004


September 2006

USE OF THE MULTIPLICATIVE PIVOTUSE OF THE MULTIPLICATIVE PIVOTUSE OF THE MULTIPLICATIVE PIVOTUSE OF THE MULTIPLICATIVE PIVOT

Once these steps have been achieved, a pivot is applied on the base emissions and attractions, that is to say on the observed emissions and attractions for base year 2004, which leads to :

• ln current situation (2004), the pivot results in the observed emissions and attractions.

• In forecast situation (scenario), the pivot multiplies the observed emissions and attractions by the evolutions calculated between the current synthetic emissions/attractions and the forecast synthetic emissions/attractions of the corresponding scenario. This can be accounted for by the following scheme :

Figure 1.2 : principle of the multiplicative pivot

Base matrices Base matrices Base matrices Base matrices (i.e. observed)(i.e. observed)(i.e. observed)(i.e. observed)

Matrices in scenario Matrices in scenario Matrices in scenario Matrices in scenario situtation, ressitutation, ressitutation, ressitutation, resulting from ulting from ulting from ulting from the multiplicative pivotthe multiplicative pivotthe multiplicative pivotthe multiplicative pivot

Synthetic matrices in Synthetic matrices in Synthetic matrices in Synthetic matrices in scenario situationscenario situationscenario situationscenario situation

+ x% Current synthetic Current synthetic Current synthetic Current synthetic matricesmatricesmatricesmatrices

+ x%


September 2006

1.4. Importations and exportations of the Belgian zones

The importations and the exportations of the Belgian zones in scenario situation have been calculated by introducing annual growth rates which are based on the following parameters :

• Average Annual Growth Rate of the importations or exportations per NST category of goods.

• Average Annual Growth Rate of the ratio value/ton of the importations or exportations per NST category of goods.

1.5. Emissions and attractions of the foreign zones

The generation step does not take into account the emissions and attractions between foreign zones (i.e. transit). These flows are only calculated in the distribution model, which directly follows the generation model.

However, the growth of the attraction of the foreign zones is taken into account in the generation model, so as to limit the flows proportionally to the incoming capacity of their destination. The annual growth rates of the BIP of the principal destination countries are used in conjunction with elasticities on the attractions of the latter.

The emissions of the foreign zones do not need to be restricted of even increased as the distribution model bases the concerned flows on the attractions of Belgium.

1.6. Correction of punctual anomalies

As it is based on the socio-economic data, the generation model assumes that these data can result in a realistic transport activity.

Even if this assumption is very often true, it may happen that some anomalies appear in some zones. In this precise case, the transport activity cannot be completely accounted for by socio-economic data. The following reasons can for instance lead to this problem :

• Concentration of industries with high productivity rates (storage centers, logistic activities...)

• Unique big structures in few populated zones / zones with low employment figures (metal factory, productive industry...)

• Implantation of new resident areas or enlargment of cities / industries

These anomalies can be directly compensated by integrating additionnal volumes to the original emissions or attractions of the model (not only in Belgium but also abroad), either for the base case (2004) or in scenario situation (one separate file for each scenario).


September 2006

1.7. Calibration

The following graphics give the results of the calibration of the attractions (figure 1.3) and of the emissions (figure 1.4). One can clearly see that the results are satisfying, even before the application of the multiplicative pivot.

Figure 1.3 : Comparison between estimated and observed attractions (in million tons)

Figure 1.4 : Comparison between estimated and observed emissions (in million tons)

y = 1.3172x + 815672

R2 = 0.6966

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100

Observed attraction volumes

Esti

mate

d a

ttra

cti

on

vo

lum

es

y = 1.1301x + 2E+06

R2 = 0.6208

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100

Observed production volumes

Esti

mate

d p

rod

ucti

on

vo

lum

es


September 2006

2. DISTRIBUTION MODEL


The distribution step aims at modeling all the flows between the different zones of the study, and not only the emission and the attraction of each zone. Actually, these data that have been calculated thanks to the generation model will now be used to get the origin-destination (O-D) flows on the NUTS3 level (43 arrondissements in Belgium).

A gravity model is used to distribute these flows on the whole zoning. The results are provided for each kind of goods by O-D matrices in tons and for the sum of all modes (the mode differentiation follows in the modal split model). More information about the gravity model is given further.

Eventually, the transit flows are also taken into account in this step, even if they are not calculated with a gravity model.

2.2. Distribution of the different kinds of traffic

The domestic flows and the international flows are distributed in a different way :

• The domestic flows are distributed thanks to a gravity model. This type of model uses an exponential resistance function depending on the squared distance in our case. It can also depend on the times of transport, on the costs or on the generalised costs.

• The import and export flows are elaborated in the same way with a gravity model that distributes the emitted and attracted volumes of Belgium to/from the foreign zones.

• The transit flows are calculated with growth rates applied to the current observed flows.

As a result, the distribution model provides matrices of flows per kind of goods.

2.3. Modeling process

The modeling of the distribution flows takes place in the following steps :

• distribution of the domestic flows

• distribution of the exportations

• distribution of the importations

• growth of the transit flows thanks to growth rates

• use of a pivot on the base matrices (observed situation 2004)


September 2006

DISTRIBUTION OF THE DOMESTIC FLOWS, THE EXPORTATIONS AND THE IMPORTATIONSDISTRIBUTION OF THE DOMESTIC FLOWS, THE EXPORTATIONS AND THE IMPORTATIONSDISTRIBUTION OF THE DOMESTIC FLOWS, THE EXPORTATIONS AND THE IMPORTATIONSDISTRIBUTION OF THE DOMESTIC FLOWS, THE EXPORTATIONS AND THE IMPORTATIONS

For each kind of flow (domestic, import or export), the distribution is calculated with a gravity model corresponding to the following formulation :

).F(D.A.E.T ijxj,xi,xj,xi,xij, βα=

where :

x = NST category of goods

xij,T = Flow of goods, in tons, from the zone i to the zone j

xj,xi, , βα = Coefficients enabling to equilibrate the emissions/attractions in the

end of the process

xi,E = Emissions in tons of the zone i *

xj,A = Attractions in tons of the zone j *

)F( = Resistance function

ijD = Distance on the road between the zones i and j *

(* calculated by the model)

The proportionality coefficients αi,NST et βj,NST enable to respect the balance of the attractions

and the emissions in the end of the iterative process.

The resistance function is an exponential formula which looks like following :

).D100

.(1exp(-p)F(D ijyx

xij )γ

+=

where :


) F( = Resistance function


September 2006

ijD = Distance on the road between the zones i and j *

xp = Fitting parameter per NST category of goods

xγ = Annual growth rate applied to the fitting parameter per NST category of goods

y = Number of years between base situation and scenario situation

(* calculated by the model)

The parameter xγ enables to model an evolution of the average length of the goods flows.

However, it is possible to keep it equal to zero. In this case, one considers that the distribution of the volumes in base situation and in scenario situation goes the same way.

Once all the kinds of flows have been distributed, the model provides matrices of flows that include the domestic exchanges as well as the importations and the exportations for each NST category of goods.

GROWTH OF THE TRANSIT FLOWSGROWTH OF THE TRANSIT FLOWSGROWTH OF THE TRANSIT FLOWSGROWTH OF THE TRANSIT FLOWS

The transit flows in scenario situation are calculated thanks to annual growth rates which are estimated from :

• Growth prospects of the European BIP for the periods 2004-2015 and 2015-2025.

• BIP elasticities on the attractions for each NST category.

USE OF THE PIVOTUSE OF THE PIVOTUSE OF THE PIVOTUSE OF THE PIVOT

This step consists in applying a multiplicative pivot to the base matrices. The process is quite similar to the one of the generation model :

• In base situation, the base matrices can directly be used.

• In forecast situation (scenarii), we apply to the base matrices the evolutions calculated between the current synthetic matrices and the forecast synthetic matrices.

2.4. Calibration of the distribution model

The model has been calibrated through a comparison between the results of the model and the data of the base matrices on the following points :

• Difference between the desired and the modeled attractions for each zone.


September 2006

• Distribution of the flows in distance classes.

ANALYSIS OF THE CONVERGENCE (RMSE)ANALYSIS OF THE CONVERGENCE (RMSE)ANALYSIS OF THE CONVERGENCE (RMSE)ANALYSIS OF THE CONVERGENCE (RMSE)

The analysis of the convergence of the iterative process does also enable to validate the calibration. The method consists in calculating the value of the Root Mean Square Error (RMSE) for each iteration until the best result has been reached.

1

)(2

arg

−

−=∑

i

FFRMSE

etti

i

where :

RMSE = Root Mean Square Error

i = Number of iterations that have been run

iF = Modeled flow at iteration i

ettF arg = Observed flow

The RMSE enables to analyse how the attractions differ between the model and the base matrices, as the distribution process is based on variations of the attractions. Actually, the emissions are considered as basis and the attractions are modified at each iteration to distribute the flows while keeping a symetrical matrix.

In our case, the distribution results in very good values of RMSE in so far as it is equal to zero for each NST and for each kind of flow (domestic, import and export). That means that the matrices converge well and that the final attractions correspond to the observed ones.

DISTRIBUTION OF THE FLOWS IN DISTANCE CLASSESDISTRIBUTION OF THE FLOWS IN DISTANCE CLASSESDISTRIBUTION OF THE FLOWS IN DISTANCE CLASSESDISTRIBUTION OF THE FLOWS IN DISTANCE CLASSES

The values given by the RMSE enabled to test the convergence of the model and the quality of the modeled emissions and attractions in comparison to the observed ones. The second phase of the calibration does now consist in testing the distribution in terms of separate flows, and no more in terms of emissions and attractions.

The comparison of the observed and modeled flows by distance classes is a good indicator to test if the flows have been correctly distributed and correspond to the reality. The following graphic shows this comparison :


September 2006

Figure 2.1: Comparison between observed and estimated flows per distance classes, before calibration of the intra-zonal traffic

The difference between the observed and modeled figures is quite low, except for the distance class 0-10 kilometers, which can be accounted for by the fact that the intra-zonal traffic has been increased by a factor of 2.5 in the observed matrices. If the modeled figures are multiplied by the same factor, the difference gets much lower (figure 2.2).

-100

0

100

200

300

400

500

10 100 1000

distance

millio

n t

on

ne

s

Synthetic Base Polynom (base) Polynom (synthetic)


September 2006

Figure 2.2: Comparison between observed and estimated flows per distance classes, after calibration of the intra-zonal traffic

-100

0

100

200

300

400

500

10 100 1000

distance

millio

n t

on

ne

s

Synthetic Base Polynom (base) Polynom (synthetic)


September 2006

3. MODE CHOICE MODEL


The mode choice model aims at estimating the modal distribution of the flow matrices that result from the distribution model. It enables to split these total matrices into four single-mode matrices for each of the following modes of transport:

• Road

• Combined Transport (TC)

• Conventional railway

• Inland waterway

Moreover, the mode choice is applied to each NST category of goods on the aggregated zoning (NUTS3 in Belgium) so as to keep a global approach of the modal split. It requires utility functions that are actually linear combinations of transport times and prices. These parameters – also called levels of service – are calculated on a more precise level so as to take advantage of the precision of the multimodal network. The latter is thus scanned on the coarse level for each mode in order to get the distances and the times of each relation of the matrix. This finally enables to calculate the mode choice after re-aggregating the coarse results to the NUTS3 level.

This requires the use of the following input data:

• The current base matrices, per mode

• The scenario matrices resulting from the distribution model, sum of all modes

• The levels of service matrices (times and prices) for the current base situation

• The levels of service matrices for the considered scenario situation

As a result, the mode choice model provides matrices of goods per mode and commodity type.

3.2. Structure of the mode choice

The structure that has been selected to calculate the mode choice is non-hierarchic, which means that the choice is directly made between the following four modes:

• Road


September 2006

• Combined Transport

• Conventional railway

• Inland waterway

The following figure shows that structure.

Figure 3.1: Structure of the mode choice

The asset of this structure is the interactivity of the mode choice in so far as the model takes each parameter into account to define the modal split. Let us still mention that the different networks are connected together (multimodal network), which is for instance useful for the combined transport that uses both the road and the rail.

3.3. Modeling of the mode choice in Cube

CCCCALCULUS OF THE UTILIALCULUS OF THE UTILIALCULUS OF THE UTILIALCULUS OF THE UTILITIESTIESTIESTIES

The utilities for each mode are calculated as follows:

xm,ij,xxm,ij,xxm,ij, .Price.TimeU βα +=

MatrixMatrixMatrixMatrix all modes, all modes, all modes, all modes,

10 10 10 10 NSTNSTNSTNST

RoadRoadRoadRoad

10 NST10 NST10 NST10 NST

Combined transportCombined transportCombined transportCombined transport


Conventional Conventional Conventional Conventional railwayrailwayrailwayrailway


Inland waterwayInland waterwayInland waterwayInland waterway



September 2006

where:


m = Transport mode

xm,ij,U = Utility between zone i and zone j, for mode m

xα = Parameter associated with the time

xm,ij,Time = Time in minutes between zone i and zone j, for mode m

xβ = Parameter associated with the price

xm,ij,Price = Price in euros between zone i and zone j, for mode m

The times and the prices result from the levels of service matrices which have been calculated with network attributes and estimated figures based on statistics. The distances do of course also play a role in the calculation. The following table shows the different components of times, costs and prices that are used for each mode.


September 2006

Figure 3.2: Components of times, costs and prices

RoadRoadRoadRoad

Time = Driving time + resting time

Cost = Fuel and maintenance cost + driver costs + material cost

Price = α.(Cost road) β

Combined transportCombined transportCombined transportCombined transport

Time = Road-haul driving time for the pre and post carriage + rail-haul driving time

Cost = Road cost + hourly cost + reservation fees + energy cost + fix cost

Price = Road price + Combined transport price

ConventionConventionConventionConventional railwayal railwayal railwayal railway

Time = Driving time

Cost = Hourly cost + reservation fees + energy cost + fix cost

Price = Rail price

Inland waterwayInland waterwayInland waterwayInland waterway

Time = Road-haul driving time for the pre and post carriage + water-haul driving time

Price = Road price + water price + handling price

The time and price parameters that have been used for the calculation of the utilities result from our experience and have already been used in several models in which they proved to deliver good results. These parameters are the following.


September 2006

Figure 3.3: Parameters of the mode choice

NST commodity NST commodity NST commodity NST commodity groupgroupgroupgroup xα xβ

0 -0,09447 -0,014703

1 -0,09447 -0,014703

2 -0,03784 -0,011703

3 -0,03784 -0,011703

4 -0,03784 -0,011703

5 -0,03784 -0,011703

6 -0,04222 -0,006980

7 -0,04222 -0,006980

8 -0,04222 -0,006980

9 -0,155 -0,009057

CCCCALCULUS OF THE MODALALCULUS OF THE MODALALCULUS OF THE MODALALCULUS OF THE MODAL DISTRIBUTION DISTRIBUTION DISTRIBUTION DISTRIBUTION

Once the utilities have been calculated by using the levels of service and the parameters exposed above, it becomes possible to calculate the modal parts of each type of goods. The following formula is used:


September 2006

)exp()exp(

PP'xm,Uij,

xm,ij,U'xm,ij,xm,ij, =

where:

xm,ij,P' = Modal part of mode m in scenario situation

xm,ij,P = Modal part of mode m in current base situation

xm,ij,U' = Utility of mode m in scenario situation

xm,Uij, = Utility of mode m in current base situation

The last step of the calculus of the modal parts consists in weighting the latter so as to get their sum equal to one. The application of these modal parts on the distribution model’s output matrix eventually results in matrices per mode and type of goods for each considered scenario.


September 2006

4. FINE DISTRIBUTION MODEL


This fine distribution step aims at disaggregating the zonal level from the NUTS3 zoning (43 zones) to the coarse zoning (518 zones) in Belgium, so as to continue on this more precise level in the next steps. It is a very compact sub-model compared to the previous ones, which can be accounted for by the nature of the input data.

Actually, as we have got these input data on the coarse level (they have been aggregated before their integration in the model), we would lose information during their disaggregation. Therefore the use of socio-economic data has been avoided to prefer a fine distribution keylist that splits the zonal system according to observed weights of production and attraction for each zone. Each mode requires a specific keylist, whose scenario dependant figures can easily be modified by the user.

4.2. Result

The fine distribution model results in 4 matrices of tonnes – one for each mode – detailed in 10 categories of goods on the coarse zoning level.


September 2006

5. TLN MODEL


GGGGLOBAL PURPOSELOBAL PURPOSELOBAL PURPOSELOBAL PURPOSE

The matrices that result from the fine distribution model do only link the different zones of the zonal system. They stay thus quite general in their degree of precision – even if the coarse zoning is already fine enough – in so far as specific concentrations of traffic cannot be directly taken into account by the matrices. However, such a possibility would be particularly interesting for the road traffic whose concentration can strongly depend on logistic activities. The purpose of the TLN (Transport Logistic Nodes) model is actually to integrate these logistic centers into the road matrix so as to stick as close as possible to the reality of the flows and enable better results in the assignments of the next steps.

DDDDEFINITION OF A EFINITION OF A EFINITION OF A EFINITION OF A TLNTLNTLNTLN

A TLN “zone” can be described as a transfer point where loads change the means of transport (and not necessarily the mode!) simultaneously with a re-consolidation of the shipment. This has to be clearly separated from the case of an intermodal terminal, where loads change the mode by using standardised load units (containers, swab bodies or trailers) without re-consolidation of the shipment. TLN also has to be clearly separated from production activities where commodities entering the production site (input) are not necessarily the same that leave the production site (output).

A TLN might be one “single” node (one warehouse or distribution centre) – necessarily of a certain importance worth being considered as “stand-alone” – or an area where logistic activities are concentrated.

RRRROLE OF THE OLE OF THE OLE OF THE OLE OF THE TLNTLNTLNTLNSSSS

Eventually, the TLNs are used to split the road relations into direct and indirect flows, the latter going via the TLNs.

5.2. Structure of the TLN model

OOOOVERVIEWVERVIEWVERVIEWVERVIEW

The following figure shows the integration of the TLN model in the modeling chain and especially the splitting between direct and indirect flows.


September 2006

Figure 5.1: overview of the TLN model

SSSSTRUCTURE OF THE MODETRUCTURE OF THE MODETRUCTURE OF THE MODETRUCTURE OF THE MODELLLL

First, the TLNs must be integrated in the network so as to be taken into account in the matrix. Their localization will be important to determine their catchment area before redistributing the flows of the road matrix.

Each TLN has got a few attributes that enable the splitting of the flows at the end of the TLN model. These attributes are following:

• Capacity: sets the maximal volumes that can transit via the TLN

• Maximal short-haul distance: sets the catchment area for the short-haul side

• Minimal long-haul distance: sets the catchment area for the long-haul side

• Number of the TLN: enables to define the TLN

The following figure shows both catchment areas and the direct/indirect trips.


September 2006

Figure 5.2: TLN catchment areas and trips

Then the flows are redistributed via the TLN if its attributes fit the considered relations. The distance matrix and the capacity of the TLN play an important role at this step as they enable to select and weight the relations that will be split into direct and TLN trips.

The following figures illustrate the process with a simple example.

Figure 5.3: TLN attributes

Capacity Short-haul limit Long-haul limit

TLN1 100 t 2.5 km 6.5 km

TLN2 300 t 1.5 km 7.5 km

Long-haul side

Short-haul side

TLN

Direct trip

Trip via TLN

Zone A

Zone B

Long-haul side

Short-haul side

TLN

Direct trip

Trip via TLN

Zone A

Zone B


September 2006

Figure 5.3: Distance matrix including both TLN

Figure 5.4: Tonnes matrix before the TLN model

Tonnes 1 2 3 4 5 6 7 8 9

1 100 130 160 190 220 250 280 310 340

2 120 150 180 210 240 270 300 330 360

3 140 170 200 230 260 290 320 350 380

4 160 190 220 250 280 310 340 370 400

5 180 210 240 270 300 330 360 390 420

6 200 230 260 290 320 350 380 410 440

7 220 250 280 310 340 370 400 430 460

8 240 270 300 330 360 390 420 450 480

9 260 290 320 350 380 410 440 470 500

Distances 1 2 3 4 5 6 7 8 9 TLN1 TLN2

1 0 1 2 3 4 5 6 7 8 9 8

2 1 0 1 2 3 4 5 6 7 8 7

3 2 1 0 1 2 3 4 5 6 7 6

4 3 2 1 0 1 2 3 4 5 6 5

5 4 3 2 1 0 1 2 3 4 5 4

6 5 4 3 2 1 0 1 2 3 4 3

7 6 5 4 3 2 1 0 1 2 3 2

8 7 6 5 4 3 2 1 0 1 2 1

9 8 7 6 5 4 3 2 1 0 1 0.5

TLN1 9 8 7 6 5 4 3 2 1 0 1

TLN2 8 7 6 5 4 3 2 1 0.5 1 0

Legend: Relations of TLN1 corresponding to the distance attributes of TLN1

Relations of TLN2 corresponding to the distance attributes of TLN2

Relations with possible TLN trip via TLN 2

Relations with possible TLN trip via TLN 1


September 2006

Figure 5.5: Tonnes matrix including the TLN and the indirect trips

One can easily find back the capacities of both TLN (100 t for TLN1 and 300 t for TLN2) by summing the line and/or the column of each TLN.

5.3. Result

The TLN model results in a road matrix on the coarse zoning level including all the TLNs which have been added to the network and the indirect trips via these TLNs according to the attributes that have been given to them.

Let us still mention that the functionalities of this model have been lead by the available data about TLN in Flanders. One could have wished more possibilities but the amount of information in this precise field is unfortunately quite limited.

Tonnes 1 2 3 4 5 6 7 8 9 TLN1 TLN2

1 100 130 160 190 220 250 280 224.25 245.95 10.24 169.56

2 120 150 180 210 240 270 300 319.61 348.66 21.73 0

3 140 170 200 230 260 290 320 338.98 368.03 22.99 0

4 160 190 220 250 280 310 340 370 400 0 0

5 180 210 240 270 300 330 360 390 420 0 0

6 200 230 260 290 320 350 380 410 440 0 0

7 220 250 280 310 340 370 400 430 460 0 0

8 173.61 261.5 290.55 330 360 390 420 450 480 21.73 62.61

9 188.08 280.87 309.92 350 380 410 440 470 500 23.31 67.83

TLN1 7.87 17.64 19.53 0 0 0 0 26.3 28.66 0 0

TLN2 130.43 0 0 0 0 0 0 80.87 88.7 0 0

Ontwikkeling van een multimodaal goederenvervoermodel voor ...

Documents

Transcript of Ontwikkeling van een multimodaal goederenvervoermodel voor ...