Scholarly article on topic 'Urban freight establishment and tour based surveys for policy oriented modelling'

Urban freight establishment and tour based surveys for policy oriented modelling Academic research paper on "Social and economic geography"

CC BY-NC-ND
0
0
Share paper
OECD Field of science
Keywords
{"Urban freight" / "establishment and tour based surveys" / "policy oriented modelling" / "urban policy" / "decision-making support"}

Abstract of research paper on Social and economic geography, author of scientific article — Christian Ambrosini, Daniele Patier, Jean-Louis Routhier

Abstract Commonly collected data about urban freight does not provide appropriate input to urban freight transport policy-oriented models. According to us, a policy-oriented model (Ambrosini et al., 2008) is a data-adjusted model, oriented towards operational policy analysis. The aim is to simulate freight distribution within urban areas for evaluation, control and design of the urban freight transport system. Policy-oriented models help public decision making about the main stakes associated with an urban sustainable development. These models have to simulate the present situation properly enough to provide a good quality forecast to satisfy both local urban and global environmental issues. Data collected in order to calibrate those models have to fit into the methods carried out for their working. So we propose first doing a critical investigation of different possible methodological approaches for urban goods movement surveying. Secondly we present some recent approaches staying more close to reality, stressing on the necessary quality of data to improve the reliability of the results.

Academic research paper on topic "Urban freight establishment and tour based surveys for policy oriented modelling"

Available online at www.sciencedirect.com

ScienceDirect

Procedia Social and Behavioral Sciences 2 (2010) 6013-6026

The Sixth International Conference on City Logistics

Urban freight establishment and tour based surveys for policy

oriented modelling

Christian Ambrosinia*, Daniele Patiera, Jean-Louis Routhiera

aLaboratoire d'Economie des Transports, ISH, 14 Av. Berthelot, 69363 Lyon cedex 07, France

Abstract

Commonly collected data about urban freight does not provide appropriate input to urban freight transport policy-oriented models. According to us, a policy-oriented model (Ambrosini et al., 2008) is a data-adjusted model, oriented towards operational policy analysis. The aim is to simulate freight distribution within urban areas for evaluation, control and design of the urban freight transport system. Policy-oriented models help public decision making about the main stakes associated with an urban sustainable development. These models have to simulate the present situation properly enough to provide a good quality forecast to satisfy both local urban and global environmental issues. Data collected in order to calibrate those models have to fit into the methods carried out for their working. So we propose first doing a critical investigation of different possible methodological approaches for urban goods movement surveying. Secondly we present some recent approaches staying more close to reality, stressing on the necessary quality of data to improve the reliability of the results. © 2010 Elsevier Ltd. All rights reserved

Keywords: Urban freight; establishment and tour based surveys; policy oriented modelling; urban policy; decision-making support

1. Introduction

It is important to keep in mind the main issues and objectives of local authorities, noting that some of those objectives are also related to national or international concerns (e.g. pollution and greenhouse gas). The framework of the urban freight transport system has substantially evolved recently. Since about the middle of the 90's, there has been a quick growth of car traffic in cities. At that time, the top priorities have changed. Until then, delivering goods in city centres was not a major problem, because of relatively low congestion. Local authorities have to cope with substantially higher levels of traffic congestion since then. Other issues have appeared simultaneously regarding the city residents, firms and stores, confronted with new requirements about the quality of the urban life, involving

* Corresponding author. Tel.: +33-(0)4-72-72-65-26; fax: +33-(0)4-72-72-64-48. E-mail address: christian.ambrosini@let.ish-lyon.cnrs.fr.

1877-0428 © 2010 Published by Elsevier Ltd. doi:10.1016/j.sbspro.2010.04.015

attempts to reduce local environmental impacts and global nuisance (air and noise pollution as well as greenhouse gas emissions). All this has led to:

• control traffic flows in the city centres (speed limits or restricted parking policies),

• the management of a worrying lack of space for shipment consolidation and storage facilities, linked to important changes in logistic chains (hence emergence of issues associated with the location of logistic platforms in the cities), and

• the will for revitalising city centres (preservation of local retail trade stores to support economic activity).

Such objectives that involve pick-up/delivery operations have to be explicitly taken into account regarding a new road sharing between car and truck traffic, as well as a new urban development, including more urban goods movement concerns (delivery parking areas, urban distribution centres, types of vehicles, appropriate routes and specific delivery time windows).

2. Objectives of Urban Freight Data Collection and Methodologies to Get Appropriate Data

Urban freight data collection takes place in the framework briefly described above. Many reasons underlie this gathering (specific projects to produce national or local estimates concerning vehicle flows, regulation or else environmental impacts; public authorities monitoring and performance measurement; freight transport modelling and forecasting; requirements into line with EU directives; commercial monitoring; legal requirement for licensing and safety controls, etc.). We underline that the variety of motives and purposes to collect information make international comparisons difficult. Besides, it is urgent to harmonise definitions and units of observation to perform appropriate comparative diagnoses.

In the same way, there is a large variety of methods (especially different types of surveys) for collecting urban freight data. Information can be provided by different means, including amongst others:

• traffic counts,

• roadside surveys,

• interviews with freight transport company managers,

• interviews with establishment managers (consignees/ shippers),

• on consolidation platforms or on roadside interviews with drivers,

• questionnaires sent to freight transport company managers, or drivers, consignees or shippers,

• accompanied trips with vehicle drivers,

• parking and loading activity surveys,

• data collection using on-board new technology, including use of satellite tracking data containing goods vehicle activity (GPS), etc.

The data collection technique is influenced by the type of data to be collected and the use of it; e.g. a data collection providing a brief look at a situation differs from a survey feeding a freight model.

3. Common Surveys

Recent reviews (Patier and Routhier, 2008) confirm that the regular national approaches are often similar but most of the time not specifically urban-oriented. So it is important to expound positive and negative sides, in the light of the urban freight modelling objectives. Usual approaches generally consist of:

• Automatic counting by means of magnetic loops: most of the countries make ample use of it allowing traffic counts easily and inexpensively. Heavy goods vehicles (> two axles) can be distinguished from cars and light goods vehicles. Counts are made throughout diverse time periods. This method is useful in case of traffic monitoring and forecasting, as well as to calibrate models for traffic generation and O/D matrices, or to get rough estimates.

• Roadside surveys ("cordon" surveys): vehicles are stopped alongside the curb to interview the drivers about their trip, getting information about origin, final destination and purpose of the trip; to get the number of possible intermediate stops, their location and purposes, etc. The management of such surveys is not easy, in that stops may quickly lead to traffic jams or road safety problems. As it is difficult to ask a lot of questions to the truck drivers, it is impossible to link accurately the industry activity with the purpose of trips or the exact kind of the freight carried. So information will not be much appropriate for thorough studies. Moreover, internal traffic inside a zone cannot be captured. Nevertheless, such surveys are useful to calculate the number of different types of vehicles using a road section or running in a traffic zone; so rough estimates of the through and the inbound/outbound traffic are possible. This technique permits validation and calibration of the results of specific surveys and models.

• Vehicle-based surveys: consist of picking a sample from a vehicles register. In most of the countries, these LGV and HGV surveys, are carried out by national authorities. In France, two types can be considered, compulsory freight road transport business vehicles (more than 3.5 tons) surveys, carried out yearly by the national French Transport Observation and Statistical Service (SOeS). They permit an estimation of the traffic performance of HGV, tons and types of goods carried, number of vehicles moving from one region to another. It is the main way to calibrate interurban freight traffic models (commodity flow models). As well, periodic surveys on LGV (<3.5 tons) fulfil the same function as that for HGV.

• Shipper surveys: in France, INRETS (French National Institute for Transport and Safety Research) carried out a "Shipper" survey in 1988 and the "Echo" (Envois, CHargeurs-Operateurs de transport) survey in 2006. Meanwhile, a test survey had been conducted in 2003 (Rizet et al., 2004). One of the main features of these surveys is to keep track of the shipments all along the transport chain (from the shipper to the consumer). So supply and demand of goods can be thoroughly analysed. The main stake is to identify the real logistic determinants of the transport chain to understand the behaviour of all the stakeholders. Information has been collected directly from shippers, carriers, customers, forwarders and providers. In the Echo survey, a representative sample of 2,935 establishments was picked among 78,000. 9,742 dispatches among 10,462 have been tracked from the shippers to the final destination. The establishment sample was stratified according to the activity and the number of employees.

These surveys have fed many studies on freight demand analysis and modelling: traffic generation, spatial distribution, transport organisation, etc. A logit modal choice model enables the modal split for each segment to be estimated and to predict the new split in case of demand changes. So it is possible to analyse the evolution of transport chains between 1988 and 2005, as well as to compare shipments according to different sizes of urban areas (Dufour J.G., 1995). But it is impossible to analyse accurately the last miles travelled according to the rounds, because dispatches from retailers are not considered. Weightings are used in order to extrapolate the results from the samples, but not precisely fitted for different urban areas. In the end, it is not easy to use these results at an urban level.

In conclusion, among the approaches described above, none of them are really designed to encompass the global specificity of urban freight transport. Most of these surveys consider only one type of vehicle. Usually, a complete description is made only at the starting point of the transport chain (cf. shipper surveys). For those reasons among others, the observation units used (vehicles or shippers) do not allow an exhaustive description of the variety of the urban goods transport flows. Moreover European freight experts have recently identified appreciable range of types of data to be collected (Browne and Allan, 2006), including:

• data about light goods vehicle activity,

• data about the supply chain considered as a whole,

• data about urban freight and logistics infrastructure,

• data about loading and unloading operations and infrastructure for goods vehicles,

• geographical data about goods vehicle trips in the urban areas,

• data about trips carried out by the consumers for shopping purposes, and

• speed and route data of the goods vehicles; data for non-road modes.

So those current approaches are not sufficient to provide a significant increase in the efficiency of the models. So these methods have to be compared with some more recent approaches: establishment and tour-based surveys.

4. Establishment Based and Tour Based Surveys

4.1. Tokyo metropolitan freight surveys

Although not exactly conducted in the present perspective, but considering the extent and the wealth of it, it is interesting to take a look at the last exhaustive Tokyo Metropolitan Freight Survey (TMFS), conducted in 2003 by the Transport Planning Commission of the Tokyo Metropolitan Region (TMR). It consisted of three surveys (Hyodo et al., 2007; Shimizu et al., 2007):

• a questionnaire survey to logistics establishments (30,000 respondents from about 120,000 questionnaires mailed out) about daily commodity flows and logistics activities of their facilities,

• a questionnaire survey to large size truck drivers about their routes, and

• a local delivery survey of loading/unloading activities in five Central Business Districts (CBD).

In 2005, the Commission analysed the survey results and established several transportation policies as well as city and regional planning policies. Those were classified into four fields:

• support location of regional logistics facilities,

• control location of local delivery facilities,

• improve road network for large trucks, and

• promote effective (un)loading activities in the CBD.

At the request of the Commission, Hyodo et al. focused on logistics facility location choice, for measuring impact of road construction, because the control of location of logistics facilities is considered crucial in the city and the regional planning in the TMR. So a model was developed to quantitatively estimate location choice probabilities for regional logistics facilities, to help decision-makers (improvements in logistics location efficiency near expressway interchanges). Moreover, Hyodo et al. focused also on heavy trucks routes choice modelling from the recent TMFS, because of a specific problem set by new heavier and taller containers movement on the Tokyo road network.

From the TMFS, Shimizu et al. developed a study of delivery distribution in the Tokyo central business district, to help improving the delivery system. But this fruitful source of information is not based on the implementation of a model. Nevertheless several proposals could have been made to reduce the amount of delivery vehicle traffic, to secure parking for on-street delivery vehicles, to improve the transport of goods between delivery vehicles and their final destination. A report combines the results of the different Tokyo surveys and concludes by summarising the various policies to implement, given that those surveys had been conducted as part of the TMR's transportation plan. Notably, these measures aim to separate delivery traffic from other types of traffic, using control and efficient management of space, time and other demand measures.

4.2. Canadian business surveys

Recent Canadian business surveys have been conducted in Calgary in 2000 (about 3,000 business establishments) and Edmonton in 2001 and 2002 (about 4,500) to improve the knowledge of goods movements generation, distribution patterns, influence of employment size and type on commodities shipments, relations between vehicle movement and commodity movement (pattern of delivery for a given commodity flow, use of depots, all the trips needed to accommodate a commodity flow, the types of vehicles used). Information was collected on the full range of commodities being transported by business establishments, together with description of vehicle movements arising with a transportation activity. An establishment-based process was implemented, analogous to the household-based approach used in personal travel surveys. It was possible to distinguish goods and services in all the vehicle movements (Hunt et al., 2005). For this purpose, three basic categories of data have been collected:

• Establishment data: 25 categories of employment based on industrial sectors, relating to the nature of production process performed at establishments and to the space occupied. Establishment forms concerned location, category, employment and total movements of goods and services produced at the establishment and leaving it by category. Depots were concerned too and establishment depot forms consisted of the same points, with questions in addition about goods and services entering the depot.

• Shipper behaviour data: including location of destinations for goods and services on the survey day, broken down by size, type and value of shipments to each destination. Goods shipment forms concerned quantities values, routings and final destination for goods leaving the establishment. These were distributed to vehicle drivers taking goods from the establishment on the collection day. Goods depot forms concerned quantities, origins and destinations, vehicles used for goods entering and leaving the depot.

• Vehicle movement data: type, location and timing of stops, size, type and value of shipment exchange at stops, vehicle configuration. Goods vehicle forms concerned description of the pattern of all stops made by the vehicle departing the establishment carrying goods produced at the establishment. Service vehicle forms consisted of the description of the pattern of stops made by vehicles carrying goods and services provided from the establishment.

Establishment forms were completed by the initial contact person, shipment forms by the shipper at the establishment, vehicle forms by the drivers. Such a survey consists in providing information about only the shipments of production leaving the establishment. This way to proceed was chosen to lighten the task of the respondent, in order to reduce the potential amount of information to provide.

Methodologically, the establishment sample was developed from data supplied by Alberta Treasury, including a business name, address and industry classification. In Calgary, the lack of telephone numbers made it necessary to spend extra time. As for the French register, a large number of duplicate and out of business establishments complicated the process. Establishments were contacted three times: first by mail, followed by a phone call to get information about the employment of the establishment, nature of it (shipping goods/services or depot), acceptance of the survey. Second, a formal letter was sent by mail or fax, followed by a new call to send the contact person the appropriate forms. Third, data collection on the survey day was carried out according to three levels of assistance:

• minimal (the person participating called before and after the survey day and then the establishment is later visited in order to review the completed form),

• partial (a visit made at the beginning to provide initial guidance as the survey got underway. A phone call is given after the survey day, followed by a visit as above), and

• full (someone working for the survey remains on site for the entire survey day).

Concerning sample expansion and weighting, the individual scaling factor was determined according to three variables: establishment size (number of employees), standard industry category and geographical location. To check the results of the expansion process and potential respondent bias, the expanded set of vehicle movements were loaded to the node-and-link modelled network in the regional travel model. Assigned truck flows were found to match closely the observed values from vehicle classification counts. Some results are not far from those obtained with the French urban goods surveys (share of movements and of light vehicles) with light vehicles being used in about two-thirds of all commercial stops on a typical weekday. About a third of all commercial stops concern service deliveries. Note these service deliveries are often ignored in most of studies concerning freight movement or commodity flow.

Data collected has been used to develop a novel form of tour-based micro-simulation of commercial vehicle movements which has proved very successful in practice. It has been integrated with a conventional aggregate equilibrium model of household movements, and has the potential to work with emerging tour-based / activity-based models of household travel. This approach brings new avenues of research for urban commercial movement modelling. As Hunt et al. (2005) wrote, "it is important to underline that the development of this model required the data got in this survey, and surveys of the sort described here would be required as part of developing this sort of model elsewhere". The only drawback is the cost: in Calgary the total amount was $600,000 and $800,000 in Edmonton.

4.3. French UGM surveys

Establishment-driver surveys have been conducted first in France from 1995 - 1997, on behalf of the Ministry of Transport and ADEME (French public agency for the environment and the energy). At that time neither suitable statistics were available nor systems to gather information from the various economic stakeholders involved, nor analytical methods, nor methodological tools, nor monitoring of experiments on urban logistics. The first urgent task was to carry out comprehensive surveys in a number of different fields. No research bodies were available to assist the decision-making of the local authorities, and from the outset, it was quite clear that the human and financial resources required exceeded the capacities of any single conurbation. Moreover there was a common need for a framework which could be used by anyone, not just for a specific city. So the Laboratoire d'Economie des Transports (LET) has built and analysed three large original surveys to provide useful quantitative elements for an appropriate demand forecasting and vehicle flow generation in the French towns Marseilles (1.1 M inh.), Bordeaux (750,000 inh.) and Dijon (240,000 inh.) (Ambrosini et al., 1997 and 1999). In each city the same method has been implemented, based on an establishment-based survey coupled with a tour-based (drivers) survey. They are called Urban Goods Movement (UGM) surveys (Patier and Routhier, 1998).

Several methodological precautions have been taken to build a really operational urban goods movement survey, because relevant data were required to reflect real situations. Indeed, the number of vehicles involved in deliveries and pick-ups depends on the type of goods carried, packaging, frequency of deliveries, type of delivery vehicles used, but such data does not exist in usual statistics. Generally usual known data are: population, employment, number of establishments with the type of their activity. In order to shed light on the relationship between the activity and the flows of vehicles they caused, it is necessary to survey trip generators, sending/receiving goods. Only generators know precisely how many vehicles stop every day in their establishment, the size of the latter, the management of the delivery trips, and the constraints of management regarding the type of goods.

Relevant data is needed: in the urban areas, the knowledge of the freight flows is not reduced to knowing the routes of delivery vehicles, because about 75 % of pick-ups/deliveries are carried out in more or less complex rounds. So it is advisable to re-examine the traditional 4 steps gravity approach (generation of tons, O/D goods distribution, modal choice and network assignment) usually used in the long haul modelling. An urban goods transport origin-destination matrix has no meaning in terms of transport: one ton (or m3), from one zone i to one zone j, can be carried in a single payload in direct trip between i and j, as well as hundred small parcel deliveries, some of them being delivered straightforward and some other delivered in complex rounds with light goods vehicles. As the current modelling approaches are not efficient at the urban scale, so the underlying hypotheses do not fit.

After having pointed this fact out, what would be the most quantifiable observation unit? To observe the different ways of organisation of a goods vehicle on-street, several statistical units may be considered: monitoring of a street section during a defined period allows one to know on one hand parking locations and parking times and on the other hand the movement of goods vehicles going through this section. Making a survey of the routes of goods vehicles permits a thorough description of their stops. Through the knowledge of the shippers' activity, all the pickups/deliveries can be registered. But each of these observation units show some drawbacks: rules of sampling concerning street sections are difficult to implement; routes may not be precisely located considering land use characteristics; shipper surveys do not provide easily the characteristics of vehicle routes.

The "movement of vehicle", in the sense of deliveries or pick-ups achieved at each establishment, has been chosen as the statistical unit because:

• a movement of vehicle has to be considered according to the road occupancy (running and double park time), what ensure a quantified measurement of the objective,

• the movements have to be described through an establishment survey, to get the useful characteristics of each movement (especially the space occupied by the vehicles have to be precisely measured), and

• in order to quantify the on-street occupancy as a vehicle runs from a depot to a delivery zone or between two delivery stops, it is necessary to complete the establishment survey with a delivery-man survey.

So a self-managed questionnaire has to be filled in by the drivers who have visited the surveyed establishments. Thereby, it is possible to piece together the route including the movement initially surveyed.

To focus the objective, we have to assume that the main issue about goods movement is road occupancy by the goods vehicles which are in competition with individual cars. Thus congestion and accessibility have to be transformed into road occupancy. If the size of the sample is sufficient, such a survey may give an accurate picture of the urban goods vehicles traffic. This traffic is considered in a general sense of road occupancy both for running and on-street parking time. This choice allows us to circumvent the difficulties inherent in identifying the origin/destination flows which are one of the priority aims of the approaches usually encountered (Bonnafous, 2000).

4.3.1. Case study

The first major survey was launched in Bordeaux with an intense participation of the city authorities. Three joint surveys have been implemented: - an establishment survey, carried out from establishments shipping/receiving goods (industrial, commercial, tertiary activities), a driver survey, concerning drivers delivering/picking-up goods to an establishment (own account, third party), and a carriers survey, from the most frequently cited transport companies.

4.3.1.1. Survey method

Carrying out such a survey depends on the commitment of the body responsible for the management of an urban transport system e.g. the Urban Community of Bordeaux has been much involved, especially at the time of the choice of the urban area to survey (about the density of the economic activity and the availability of local databases). In the end, the area was accepted in accordance with the individual household surveys area, in order to have available data about car and public transport road occupancy. So the survey area was divided into zones similar to household trips survey zones, in order to ensure a correspondence between population indicators and movements of goods. The establishment survey was carried out using a questionnaire, providing data through interviews of the establishment manager, about establishment's activity, fleet of vehicles, storage capacity, parking facilities and surroundings of the establishment.

A log, kept by the person responsible for the logistics, provided data on all incoming/outgoing movements of goods for a week. The log consisted of a set of data sheets. Each of them contained information relating to one movement in the course of the week, including pick up/delivery data (location, type of vehicle, time for delivering), carrier's name, frequency of pick-ups/deliveries and data on the product (type, packaging, weight, origin and destination).

A removable driver-questionnaire was attached to each movement data sheet. Driver surveys were self-managed: questionnaires were given to the drivers in the establishment in which they were making a delivery/pickup and then were returned filled in by post. The questionnaire described the round exhaustively (number of stops in the city, type and weight class of the vehicles used, type of handling equipment used, distance covered and type of each establishment delivered). The route was drawn on a city map and information was added about the number, the location, the schedule and the parking time for each stop.

The carriers' surveys were conducted by a transport specialist during a face to face interview. The main points surveyed were: company's activity (express, consignment, (inter)national, fleet, number of employees), organisation of the transport chains, frequency of the deliveries, fleet of vehicles allocated to deliveries in the city, trucks movements, etc., number of pickups/deliveries, number of daily rounds (per time periods), type of the vehicles used. Concerning the organisation of the activity: location of most frequently used terminals, main logistics chains, number of rounds, number and type of vehicles involved, etc.

4.3.1.2. Sampling

The French exhaustive establishment register (SIRENE), includes a lot of invaluable information on each establishment (industry, number of employees, nature of the premises, etc.). In Bordeaux, the sample was picked from 38,507 among 40,466 establishments of the urban area. Public services (eg. schools, local authorities, post offices and hospitals) were covered by separate studies and not included in the survey. In the conurbation, the latter activities generate few goods movements but however account for 25% of the employment. The survey had to cover a large number of establishments to obtain acceptable results about the generation of movements in each class of activity.

So 1,500 establishments were surveyed. On average, each establishment provided five data sheets, concerning the different movements carried out by the same carrier in one establishment (6,600 data sheets in total). Dummy data

sheets (on average one per establishment), were made for regular movements not carried out during the survey week. A precise description of the movements for a week was given (nature and path of the goods of each shipment, packaging, weight) according to 8,300 product lines.

Results show that a majority of movements are involved in a single shipment. The sample of the establishments was picked from this database. A stratified sampling method was used so as to ensure an adequate representation of the different categories of establishments according to activity and size with regard to the generation of movements. An a priori partition (37 strata) was built, according to the industry class, the number of employees and a comprehensive geographical coverage of the conurbation.

A total of 6,000 driver questionnaires were handed out by the establishments taking part in the survey, with 903 valid questionnaires returned by the drivers (17%). It was possible to link them to the surveyed establishments. Owing to them, 69 of most frequently used carriers could have been surveyed. In Marseilles and Dijon, establishment surveys have been partly realised by means of Computer Assisted Telephone Interviews (CATI), for 2,000 establishments in Marseilles and 1,000 in Dijon. But this method was rather unwieldy and replaced by a usual phone call accompanied by a questionnaire by post. Concerning the driver surveys, they have been realised in a different way compared with Bordeaux, to add some points: instead of giving the questionnaire to the driver when they (un)loaded at the establishment concerned, the survey took place in the company in charge of the deliveries. For direct trips and simple rounds, delivery persons were interviewed at the carriers/forwarders platforms or at the consolidation centres. In the case of complex rounds, on-board surveys were realised by a pollster. The number of valid questionnaires was 800 in Marseilles and 500 in Dijon. The total amount of the database for the three cities is 4,500 establishments and 2,200 drivers' questionnaires.

4.3.1.3. Weighting and sample expansion

The expansion of the sample results allowed extrapolation of the observed data to the whole urban area, but this expansion required several weightings: concerning the generation of movements, the number of movements is mainly sensitive to the type of industry and the size of the establishments. So it was assumed that the fluctuations linked to other characteristics cancel each other out when the movements were aggregated on large groups of establishments.

Weighting of the establishments: this weight ws is specific to each stratum s, permits to build the sample and is determined by the ratio rs (number of establishments of the stratum in the conurbation)/ (number of establishments of the same stratum in the sample):

Weighting of the movements: in the log, some movements have been left out or poorly captured. So it was necessary to change each weighting at the establishment level to make possible the extension of the results from each data sheet of the movements. Results showed that the number of movements obtained by the sum of the frequencies of the movements in the data sheets were 20% lower than the actual amount described by the establishments interviewed. To provide a correction, the scaling factor ce of the characteristics for each movement of e is such as:

Ce = mSe / Zfi (2)

Where :

mse is the number of movements of an establishment e in the stratum s during a week;

Zfi is the number of movements obtained by the sum of the frequencies (number of movements realised each week) fi of data sheet movements of the establishment e.

Each stratum is built so that the distribution of the establishments of the sample according to the number of generated movements is statistically representative of the distribution of the establishments in the conurbation. Some

big generators (automotive industry and warehouses) therefore, have to be surveyed separately. For each stratum s, each movement i of e in the sample has a weight of mvsei movements in the whole study area, such as:

mVsei = Ws * Ce.*fi (3)

The total number of movements Ms of the vehicles generated by the establishments of the stratum s is:

Ms = Zmse* Ws= Zmvsei (4)

eCs eCs, i

Weighting of driver trips : two bias have been found in the driver sample: first, for each management mode, there is an imbalance between the number of movements generated by the different types of activities of the establishments delivered by the drivers and the number of movements generated by the whole movements delivery slips described in the establishment sample; second, drivers who belong to the surveyed firm answered more frequently than third party drivers (the rate of questionnaires sent back was different according to the closeness of the driver and the establishment delivered). In order to correct those biases, the following weighting of a tour r-tma of the type (m,a) was used (weight in weighted tour number):

r-tm,a = Nmvtm,a/ nmvr^a and Nmvtm,a = Zmvm,e (5)

Where:

a is the activity branch (industry, craftsman, wholesale, retail, large store, tertiary services, warehouses; agriculture); m is the management mode, in five classes (own account realised or not by the establishment as consignee, own account realised or not by the establishment as forwarder, third party);

Nmvtm,a is the weighted number of the movements of type m calculated in the urban area for the establishments of type a;

nmvrma is the total number of stops of the (m,a) drivers trips in the driver sample.

If nbsm,a is the observed number of stops of a round for a driver of type (m,a), the weight (in number of movements) of this driver is:

r-mvm,a = nbsm,a * r-tm,a (6)

The survey period was an ordinary week. Results showed that the frequency of supplying an establishment is not daily but weekly. In the city, many stakeholders with many different logistic organisations work together, so it is necessary to capture any day of a week. Moreover, it is impossible to calibrate data from only one day without knowing the weight of this day in the global week. It is important to have information on seasonally changes, so questionnaires included issues about the weight of the surveyed week regarding the others weeks of the current month, and the proportion of this month in the year. Finally, concerning the budget to be expected, such surveys are expensive (around 500,000 €). Funding of the French surveys was shared between the Ministry of Transport, the ADEME and the city authorities.

4.3.1.4. Some results in Bordeaux

• number of deliveries/pick-ups per week: 270,000,

• average number of movements per employee: 1.1,

• ratio deliveries to pickups: 61% / 3%,

• part of own account deliveries/pickups: 56 %,

• delivery/pickups carried out with vehicles <3.5 tons: 52%,

• ratio of movements realised in rounds: 75%,

• ratio of rounds among all vehicles involved: 25%,

• share of retail trade (number of deliveries/pickups): 33%,

• part of the small establishments (< 5 employees) in vehicle movements: 50%.

Goods vehicle generation: for each type of activity, the number of deliveries and pick-ups carried out by trucks is calculable. Average size of the rounds is 13 delivery points (19 for third party, 11 for consignor-own account, 5 for consignee own account).

Road occupancy: in the centre of Bordeaux, the duration of road occupancy by the double parked delivery vehicles is twice that of those running. In order to compare the road occupancy of all the stakeholders in the city (residents, firms, services and urban managers), the veh.km car unit (vehicle km in car unit) is proposed as the statistical unit, as follows: a light delivery commercial vehicle (van or LGV<3.5 tons) is equivalent to 1.5 cars; a truck (rigid) to 2 cars; an articulated lorry to 2.5 cars. In this way, the components of urban goods transport are: deliveries/pick-ups (all establishments) 40%; purchasing trips (private cars) 50%; urban management (building sites, networks maintenance, waste collection) 10%. According to the town, the share of the urban goods traffic in the total traffic is from 9% to 15% of the trips, 13% to 20% of the veh.km and 15% to 25% of the veh.km car units.

One of the main contributions of such surveys consists of the improved knowledge of the urban management rules (links between activities, operation mode, management mode, type of vehicles used, distances covered, number of served establishments, running time, parking location and parking time). The number of deliveries/pickups, operation structure, management mode and organisational mode are linked with the type of activity. The main relationships revealed are:

• management mode and organisational mode (own account carries out mostly direct trips and third party transport mostly rounds);

• type of vehicle and management mode (own account uses more light commercial vehicles and third party transport uses rather heavy trucks);

• distance covered and management mode and organisational modes (rounds of third party transport are longer than rounds of own account);

• distance covered between two stops and size of the round (the longer the round trip is, the shorter the distance between stops is);

• stops duration and rounds size (the more stops on the round, the shorter the delivery time).

A very important result is that the same relationships have been observed in the three surveyed towns. Thus the number of deliveries caused by each type of activity and, inside each class of activity, the number of deliveries realised by types of vehicle; the share of the own account and the share of the different types of rounds. For each class of activity, it shows that the overall economic and logistics structure is depends on size and geography of the towns. Thanks to such a result, it has been proved it is possible to transfer the knowledge from those surveys to other towns on one hand and to build a general model applicable to French and even European towns on the other hand. Those relationships are the core of the Freturb model (see next part).

The uniqueness and the contribution of the French establishment-driver surveys consist of:

• taking into account the exhaustiveness of the urban goods transport,

• coupling the survey methodology and objectives (diagnosis, knowledge of the demand, link between generation and development of flows and simulation of policy-oriented scenarios), and

• weighting which permits a calibration method for planning and modelling (generation of deliveries and pickups, distance covered according to each activity, each management mode and each organisational mode).

An update of the 1995-97 French surveys is ongoing. The new surveys (2010-11) will be implemented, using the above establishment-driver methodology. In order to get a better knowledge of warehousing practices, data collection has been greatly improved. The results are eagerly expected in order to know the recent changes in urban

logistic chains and the impacts of new supply practices (e-commerce, home deliveries, low motorised and friendly environmental vehicle deliveries). In the case of embarked surveys, the use of GPS will be generalised (the use of light and passive GPS devices by the pollsters will improve the accuracy about leg lengths, speeds, stop locations and duration, all along a tour). Five cities, including Bordeaux (from 100,000 to 10 M inh., to check the irrelevance of town size to UGM) are expected. This new wave of surveys will permit future French UGM surveys to be standardised.

5. Simulation and Diagnoses to Helping Decision Making

It is important to examine the capability for doing appropriate diagnoses from the surveys described above. Decision makers need to know, as precisely as possible, the share of each activity in the generation of traffic flows, to understand and control the functioning of the urban freight transport system and its impact on the environment. Establishment-driver surveys permit such a diagnosis to be realised, thanks to various indicators.

5.1. Indicators

Common long distance transport indicators (tons, tons kilometres, energy intensity or empty running) do not seem relevant at the urban level. Energy intensity (fuel consumption per ton-kilometre) is relevant to measure the impact of freight transport between two regions. But to carry goods inside a town, types of vehicles and frequencies of their use are variable and the unit of measurement depends on the type of the vehicles used, so measures and calculations are difficult since the flows are totally disaggregated. This difficulty is the same concerning empty running of which calculation needs knowledge of the load graphs for each round carried out in the town. Following indicators have been worked out from the establishment-driver surveys:

• the number of loading/unloading, from the number of deliveries/pick-ups per week and per employee for each activity, this can consist in a fast appraisal of the generation of deliveries/pick-ups (for each industry) in a town, without any survey,

• the distance covered for loading/unloading, from the number of kilometres covered for one-pick-up or delivery in a zone, per vehicle and per activity, allows to the contribution of the moving vehicles to road congestion to be estimated;

• greenhouse gas and pollution (according to zone, vehicle, activity and management), from g pollutant per km, g CO2 per km and litres of fuel per km, can give a measure of the impact of urban goods movement on energy consumption, local and global nuisance and greenhouse gases,

• the loading/unloading density, from the number of deliveries and pick-ups per km2 inside a zone, gives the importance of the goods flows inside the zone. This calculation can be done for each branch of industry,

• the loading/unloading time (spent in on-street double parking for delivery/pick-up's in a zone, per vehicle and per activity permits to know the contribution of each branch of industry to road congestion due to on-street double parking deliveries,

• the average length of the first leg (from a platform to the delivery area), measures the impact of the platform's location on goods delivery in the catchment's area,

• the average distance travelled per pick-up/delivery, measures the contribution of one pick-up/ delivery to the whole urban traffic (per type of involved vehicle). The total distance travelled on the network in the urban area by type of heavy truck or light goods vehicle (<3.5 t) used, from the total amount of vehicle km per week, measures the contribution of the whole industry activity to the total traffic,

• the average time per delivery (per activity, per vehicle, etc.) gives the time taken for deliveries in a tour, on a segment, for an industry activity, etc., and

• the average speed per round (including and excluding stops) measures the performance of the rounds according to the organisational mode and the type of vehicle. The average weight per kilometre, per tour, per activity and type of vehicle, measures the performance of the rounds according to the organisational mode and the type of vehicle.

The description above shows the wealth of the indicators which can be calculated by the means of an urban goods "establishment-driver" survey. Each indicator is useful to make a comprehensive diagnosis of the urban goods transport system and helps to measure the involvement of the main components of the urban goods movement: types of activity, management method, organisational mode (vehicles size and rounds size). Consequently, such an approach is a very promising tool in order to improve completely the knowledge of the relationships between activity and goods transport at the urban scale, including the main components of sustainability (Ségalou et al., 2006).

5.2. Capability for modelling

The comparison of the results of the three French freight transport surveys show a great similarity in terms of diagnostics, e.g. the number of deliveries generated per employee in various activities is quite the same in the three towns. Other indicators are also quite similar: share of the different sizes of vehicle (<3.5 tons, rigid heavy vehicles and articulated vehicles); share of the own account and the third party; different sizes of rounds. These results show that, in the French urban areas, economic organisation is more important than geographic topography. Consequently, it is possible to link together into a model the main relationships between economy and transport system by the surveys.

It has to be underlined that the policy-oriented model Freturb has been developed (Routhier J.L. et al., 2001; Routhier and Toilier, 2007) on the basis of an establishment-driver surveys, designed and carried out in accordance with the structure of the model. On the basis of an establishment data register and some geographical zones data, the model simulates the number of vehicle movement generated in each zone, the impact of the economic activity on road occupancy (traffic and parking duration) and several indicators for diagnoses (e.g. average stop delivering times for each size of vehicle). Recent developments of the model allow the calculation of the distribution of the flows. It takes into account the round organisation of vehicles routes. Moreover, an environmental impact module is now available.

The main advantage of this model is that no localised specific survey is necessary for calibration. It only needs available data in a city: establishment registers, geographical and network data (zoning, road network, speed average on links). At the present time, a software package is used by more than 25 French cities. It permits the simulation of the effects of several control levers on urban logistics (outcome of the activity location for firms and households). New regulations as time windows for delivering zones can be taken into account (new ways of supplying as home deliveries and e-commerce); urban logistics facilities and organisation (urban distribution centres, spatial cooperative systems, etc.).

Table 1 summarises the key aspects of the different types of surveys:

Table 1 Key aspects of the different types of surveys

Types of survey Roadside Vehicle-based Shipper Establishment and tour-based

Survey cost inexpensive medium expensive Expensive

Accuracy of estimates rough good Good very good

Explanatory power of the related models Traffic management Regional O/D Commodity flow Urban freight movement generation

Forecast Network planning Infrastructure planning Supply chains evolution Town planning Urban logistics

Simulation and decision-making aid Traffic models calibration Routing and scheduling Distribution channel optimisation Logistic behaviour <=> traffic flow

6. Conclusion

Here we aim to show that specific methods of data collection are required to build models finding appropriate answers to the expectations and needs of decision-makers (about town planning, regulation, forecasts, etc.). It seems

necessary to combine establishment and driver surveys to get an optimal efficient modelling. Methodologically, some points are very important:

• the choice of the most relevant observation unit: the movement (one vehicle, one delivery and/or pick-up),

• selecting the sample from a reliable and comprehensive establishments register (if exists), including address, detailed activity code (NACE) and number of employees, and

• the representativeness of economic activities depends on good stratification.

A stratified sample and a post-stratification improve noticeably the accuracy of the coefficients (generation of deliveries and pick-ups). The extension of the results is possible if the size of the sample is large. The coupling of the establishment and driver surveys thanks to the movement as statistical units makes it possible to formalise the relationship between the various industry activities and the transport sector. The combination of the establishment diary of movements and vehicle travel surveys show an increase in efficiency to answer some important issues (to allow unbiased sample expansion; to explain transport conditions of the establishment logistics functioning; to link street occupancy with parked or running vehicles and to feed a policy-oriented model)

The data collection method: face to face is the more efficient technique, but it is expensive. To reduce costs, new methods can be investigated, as Internet surveys (filling questionnaires in on the Internet, with a direct control by the researcher). The choice of the city can also be linked to the involvement of the local authorities (urban community, Chamber of commerce and industry). Support from the department responsible for road maintenance and issuing building permits may be useful (as in the case of Bordeaux). All these bodies bring invaluable knowledge about the specific characteristics of the city or possible technical problems (for helping the surveys management, for instance). Good selection of consultants for the implementation of the survey is also essential.

The quality of the results depends of the possibility to react quickly, when faced with any problem. Moreover, the consulting has to be familiar with the research approach, not only from the marketing point of view. In the same way, recruiting pollsters must be very careful, because surveys are long and complex. The pollster has to be adaptable, reactive, and must have well understood by the different stakeholders. Good training and tests are necessary, since the cost of such surveys is high. Financing has to be well estimated (and include possible unforeseen factors). Unlike the household travel surveys, urban goods surveys are one-off surveys. It appears difficult to carry out such surveys in a periodic way, due to their high cost, the lack of involvement of the stakeholders and the funding of authorities. Two main results of those surveys have notably been obtained:

• explanation of the goods vehicles flows generation expanded into the whole urban area, thanks to appropriate weighting, and

• a database for the Freturb model's calibration in order to help local decision makers (transport master plans of French cities).

As a possible extension, it would be good to combine local store surveys with customers' surveys. This would allow connections between the stores supplying and household purchasing behaviour to be measured. A tool for modelling the interaction between business traffic and individual traffic should also be very useful and fruitful.

References

Ambrosini, C., Meimbresse, B., Routhier, J. L., & Sonntag, H. (2008). Urban freight policy-oriented modeling in Europe. In E. Taniguchi, & R.

G. Thompson (Eds.), Innovations in City Logistics. Nova Publishers, 4th quarter 2008. Ambrosini, C., Patier, D., Routhier, J. L., & Tranchand, P. (1997, 1999). Transports de marchandises en ville : résultats des enquêtes

quantitatives réalisées à Bordeaux, Marseille et Dijon. LET, DRAST-DTT, ADEME, Lyon, 233 p. Bonnafous, A. (2000). Les marchandises en ville : le problème méthodologique de l'appréhension statistique. Proceedings of the 13th Jacques Cartier Conference, Montreal.

Browne, M., Allan, J. (2006). Best practice in data collection, modelling approaches and application fields for urban transport models I. Theme:

Urban freight data collection — synthesis report. Best urban Freight Solutions II - EU-FP6 Co-ordination Action, 119p. Dufour, J. G. (1995). Exploitation de l'enquête "Chargeurs" de l'INRETS, en termes de marchandises en ville. CETUR. Hunt, J. D., Stefan, K. J., Broownlee A. T. (2005). An establishment-based survey of urban commercial vehicle movements in Alberta: Survey design, implementation and results. TRB annual meeting, CD-Rom.

Hyodo, T., Kuse, H., Hagino, Y., Takebayashi, H., & Endo, K. (2007). Modeling logistics location choice and truck route choice behavior by Tokyo metropolitan region freight survey. In E. Taniguchi, & R. G. Thompson (Eds.), City Logistics V (pp. 221-234). Kyoto: Institute for City Logistics.

Patier, D., & Routhier, J. L. (1998). Livraisons de marchandises en ville. Transport, Environment, Circulation, 1997-12, 8-14.

Patier, D., & Routhier, J. L. (2008). How to improve the capture of urban goods movement data? Resource paper, 8th International Conference on Survey Methods in Transport, Annecy, France.

Rizet, C., & Guilbault, M. (2004). Tracking along the transport chain with the shipper survey. ITSC Costa Rica conference proceedings. Elsevier.

Routhier, J. L., & Aubert, P. L. (1999). Freturb, un modèle de simulation des transports de marchandises en ville. Proceedings of the 8th World Conference on Transport Research, 1, 531-544. Elsevier, Amsterdam.

Routhier, J. L., Henriot, F., & Patier, D. (2009). Préparation des nouvelles enquêtes marchandises en ville. LET, Ademe, February.

Routhier, J. L., & Toilier F., (2007). FRETURB V3, a policy oriented software of modeling urban goods movement. 11th WCTR, Berkeley, June.

Ségalou, E., Routhier, J. L. (LET), Albergel, A. (Aria Technology), & De Rham, C. (System Consult) (2006). Méthodologie pour un bilan environnemental physique du transport de marchandises en ville. ADEME, éd. Connaître pour agir, 88 p.

Shimizu, M., Hyodo, T., Takebayashi, H., Kuse, H., & Hagino, Y. (2007). Study of delivery distribution in the central area by Tokyo

metropolitan freight survey. In E. Taniguchi, and R. G. Thompson (Eds.), City Logistics V (pp. 393-407). Kyoto: Institute for City Logistics.