ght Modelling Research

Table of contents

• Report on Workshop for freight modelling research
  held on 7^th July 2006, Room H4, Great Minster House at 10am
• Report written by John Bates, Independent Consultant, 30 August 2006
• Background
• List of Attendees
• Session 1 (1000-1100)
• Session 2 (1100-1330)
• Discussion Session (1400-1630)
• Freight modelling workshop - Recommendations from the discussion session on the National Freight Transport Model (NFTM)

Report on Workshop for freight modelling research
held on 7^th July 2006, Room H4, Great Minster House at 10am

Report written by John Bates, Independent Consultant, 30 August 2006

Background

Freight forecasts have been part of the Department's National Road Traffic Forecasts (NRTF) for more than 30 years. Up till the 1997 NRTF this had been carried out at a very aggregate level, but, following some work by ITS Leeds, the 1997 Forecasts used a more disaggregate approach based on 15 commodity groups. At the same time, MDS Transmodal (MDS-T) had developed a model of Cross-Channel lorry freight, and the two organisations worked together in a subsequent EU Project to generalise the model to all modes, whereupon it became known as GBFM. Further development of GBFM by MDS-T has since been commissioned by, among others, the SSRA.

Since then, the Department has made use of GBFM from time to time in the context of its policy testing. It has also carried out a major program of forecasting enhancements on the personal travel side, which have built on the tools used in NRTF and are now part of the National Transport Model.

In the latter part of 2001 the Department, recognising that a sound understanding of freight movement is essential for the creation of realistic freight models used for policy, strategy and scheme appraisal, at national, regional and local levels, embarked on a project to enhance its freight modelling capability, and commissioned a large consortium led by WSP to carry out a review of needs and possibilities. The outcome of this (in June 2002) was a recommendation for a major work programme involving data collection, research and an approach to modelling at the National level.

Given the ambitious nature of these recommendations, an additional Short List of "Initial Projects" was also prepared (November 2002), in order to provide solid foundations for the further development of the [putative]NFTM (National Freight Transport Model).

2 key new projects were recommended:

A regional pilot for Economic/Logistic Models

An audit of the current GBFM modelling approach

In addition, recommendations were made in connection with the following related issues:

A London Pilot of local agent-based microsimulation

Data needs

Technical advice and co-ordination of the NFTM development

Since then, progress has been made on a number of fronts. As a first step in developing an improved understanding of the current state of research, DfT decided to hold a workshop in order to bring together a number of streams of research and practice which are of potential relevance to the problem. The Workshop was organised by ITEA Division of DfT, who invited a number of colleagues from within DfT, as well as a group of academics and consultants. The list of attendees is given in the next Section.

The Agenda for the day is given as Annex A. The morning sessions set out the Department's context and considered current practice, both in the UK and elsewhere. The afternoon was given over to a discussion session facilitated by Ian Williams of WSP. At the end of this, a large number of issues were recommended for further consideration, and participants were invited to send in comments and priorities after the event.

This Report reviews both what took place during the meeting and the subsequent comments.

List of Attendees

Session 1 (1000-1100)

The first presentation was made by MS, with some of the material contributed by Russell Harris (who was unable to attend). A recent development is the delivery of Version 5 of GBFM, which forms part of the updated (2006) version of the National Transport Model (NTM). This could be taken as a reference point for the DfT's current freight capability, and hence suggest what the future requirements might be.

The base matrices in GBFM are assembled from a combination of data sources: CSRGT, Rail data, UK trade data, Maritime Statistics. 60 commodities are distinguished, and international and domestic are given separately. The O-D pattern has been disaggregated to around 2700 postcode sectors using gravity model techniques, and the highway network is now the same as that used for the rest of NTM.

In terms of Demand responses to changes in generalised cost, the model's capability can be represented as follows:

For non-cost drivers, the model's capability can be represented as follows:

Port capacity is not modelled, though handling improvements could be modelled through the maritime network.

A number of research issues were identified:

Base matrix

Generalised cost function

Responses to cost change

Drivers of change over time

LGVs - boundary of freight system

The last is a particularly difficult and important issue, given the high level of growth in this sector. Since the main part of the Base Matrix data is derived from CSRGT, freight carried by LGVs will be more or less excluded from the model.

The second presentation was made by AMcF, specifically in the context of the treatment of freight within NTM. There are three main requirements:

• long term forecasts of HGV and LGV traffic growth, to reflect the impact on congestion on other road users
• assess the impact of freight policies on emissions and other indicators
• allow for growth sensitivity to exogenous variables (eg GDP, fuel prices)

GBFM provides HGV forecasts at a certain level of spatial detail, distinguishing between rigid and articulated vehicles. LGV forecasts are derived from a simple time series model (as a function of GDP and fuel prices), applied globally. When the latent growth forecasts are input to FORGE, some of the growth may be re-allocated to other areas or suppressed, though the elasticities used to do this are of dubious provenance.

The short term aims for NTM were to test the new GBFM model, in particular by validating the base matrices to build up confidence about the reliability of its geographical detail. There are some inconsistencies between CSRGT and the traffic statistics which are relevant here. In addition, the model's responses to key policy and forecasting assumptions need to be critically assessed.

Longer term aims involve the assessment of possible land-use changes and logistics effects, as well as improving the values of time and reliability.

The third presentation was made by DB (Logistics division). The aims of policy in this area are to facilitate the distribution of goods subject to minimising environmental impacts. This includes movement by water. These aims are generally in line with EU policy.

A key modelling issue is considered to be the relationship between freight and the economy. There was also a general concern about ensuring security of the supply chain.

In this connection BT observed that the increase in EU traffic meant that a growing share of total HGV traffic was associated with the ports. The development of the supply chain was partly influenced by international labour cost trends. He noted that maritime modelling was based on average costs. RC also observed that there was a difference in port choice depending on whether we are talking about EU or non-EU trade.

The fourth presentation was made by CAS, and concerned appraisal. He considered that the current guidance in respect of freight was minimal and inadequate: it involved factoring the Base Matrix on the basis of the 1997 NRTF predictions for growth. In the short term, a quick re-review of the methodology was required, with the intention of WebTAG update next year, but research was needed in the longer term.

The fifth presentation was by SW, and related to freight values of time. The DfT is in the process of reviewing the current guidance in this area, based on the "cost savings approach". The current guidance takes into account the cost of driver's working time (based on marginal productivity concepts - as with other "working time" values) and the impact on vehicle operating costs.

As part of the review, they will be considering whether other costs should be considered (eg costs borne by other players in the supply chain, reliability effects, logistics considerations such as "just-in-time"), as well as reviewing the accuracy of the data on which current cost assumptions are based.

The next step is to carry out or commission a literature review, and based on that a decision will be taken as to whether primary research in this area is required.

CO then made some comments about the CSRGT, which is a rolling survey of about 15,000 HGVs per year, completing a one-week diary. Apparently a quality review had been carried out in 2004, but though he had requested more information about this, he had nothing to report so far.

ASF asked whether the published figures in TSGB included foreign hauliers (apparently there had been a survey of 2-3000 hauliers in 2003): according to CO, the answer was no. CO also mentioned an initial survey in connection with LGVs, which was a planned development across the EU, courtesy of Eurostat.

AFR mentioned that there would be some interest in marrying operator/fleet type information with CSRGT. SN expanded on the point about discrepancies between CSRGT and the traffic census data - the latter showing a strong growth trend, while the CSRGT results over time were more or less constant. He wondered whether the difference could be due to an increasing presence of foreign vehicles.

HFG wondered whether any use was made of ETIS [European Transport policy Information System], in connection with EU data and modelling. According to SN, this is used. HFG made a more general point about the need for good economic data, including input/output information.

Session 2 (1100-1330)

The first presentation was made by INW, and related to the recently completed (2005) EUNET2 project, which was referred to in the "Background" section as the "regional pilot for Economic/Logistic Models". This project had built on a previous EU-funded model of the Trans-Pennine corridor, and related to two regions within GB (though the rest of GB is also included, at coarser levels of detail).

The specific objectives of the study were to carry out a number of research and implementation tasks, and to set up an operational pilot model to be:

• a means of forecasting future levels of goods transport demand as a function of economic transactions and freight logistics
• a means of forecasting origin-destination (O-D) movements of freight in a form that takes due account of:
• • changes in the costs and other supply characteristics of transport
  • any structural changes in the national and local economies, and
  • the influence of logistics decisions on transport demand.

According to the authors, the study addresses two key areas of innovation in freight demand modelling:

• First, it connects the pattern of freight demand to the underlying spatial pattern of economic transactions (i.e. production, consumption, and trade) of the UK, thus providing a transparent interrelationship between the growth of freight and the patterns of economic activity. This makes it possible to explore opportunities for de-coupling freight growth from economic growth.
• Secondly, it represents the main logistic stages of freight transport explicitly1. By representing the main logistic stages explicitly, the model is better able to capture the choices between road and rail, and between the use of different sizes of lorries that are typical of specific logistic stages.

The main innovative features of EUNET2.0 are the representation of logistic movements, and the integration of this representation in a Spatial Input-Output (SIO) Model. These features make it possible to forecast changes in freight demand that result from future transport and logistics policy options, whilst taking into account of the growth and change in the spatial economy.

The model deals with 31 commodity categories and estimates the production and consumption in each zone. These are used to generate O-D matrices of trade (in value terms) which are then converted to tonnes, and allocated between mode and vehicle sizes. The options available are strongly dependent on the commodity type. The overall structure of the model is illustrated in Annex B (taken from the Final Report).

RC asked whether data was available on distribution centres. INW replied that it could be gleaned from floorspace information from the Valuation Office, and from Employment data.

The second presentation was made by SN, who described the essentials of the update of GBFM from V4 to V5. Domestic and International flows had been clearly separated, the Base matrices had been improved (the 1991 ODIT survey was no longer being used). He noted that GBFM underpredicts HGV veh-kilometres in comparison with the traffic census.

He suggested a number of areas where research was needed. Floorspace data needed to be analysed by SIC. More spatial detail was needed from CSRGT with a view to improving the base matrices. And shipment size data and the treatment of empty running remained problem areas.

The third presentation was made by ASF, and dealt with three distinct topics. The first was a brief presentation of a strategic freight model developed by ITS called LEFT (the current version is LEFT3). This has no geography but allows for 9 different classes of distance, as well as 7 commodity types and 2 consignment sizes. Given this, it aims to predict broad modal shares between road and rail, and can give some insight into the impact on emissions.

The second item was a description of the "Leeds Adaptive Stated Preference" tool LASP, which provides the current valuations of time and reliability etc used in GBFM. The presentation concluded with the third item - a discussion of the key elements in freight VTTS (value of travel time savings), and noted that "each occasion some time category is unpacked from VTTS, the value of what remains as VTTS will change."

The fourth presentation was by GdJ, who gave an overview of current practice outside the UK. In summary, most models made use of the same 4 "steps" as in passenger models (ie, production and attraction, distribution, modal split and assignment), but encountered problems because of the diversity of decision-makers and shipments, the additional logistics choices (shipment size, distribution centres, scheduling), and limited data, particularly that of a disaggregate nature.

He suggested a typology of (inter)national freight transport models along the following lines (a key distinction being the explicit representation of logistics):

Aggregate, 4 steps (or less):

• France, Austria, Brenner, EU (NEAC, SCENES, EXPEDITE).

Partly disaggregate:

• Italy, Germany, Great Belt, Fehmarn Belt, EU (Eufranet).

Aggregate with logistics choices:

• Netherlands (SMILE+), EU (SLAM), UK (EUNET), Norway + Sweden (prototype).

Partly disaggregate with logistics choices:

• Portland, US and Norway + Sweden (planned).

He then described in more detail the approach which was being currently taken for the development of freight models in Sweden and Norway, of which the core element involves a choice model which operates as a micro-simulation model at the level of individual sender-receiver relations, and produces shipment size and transport chain for each firm-to-firm flow, based on a full logistics and transport cost specification.

A key element was the availability of data for calibration. Sweden has the great advantage of the Commodity Flow survey, and he reported on the successful use of this data in a small project at ITS Leeds to explain mode choice and shipment size.

Discussion Session (1400-1630)

The basis for the afternoon session is described below. However, it should be noted that there was so much to talk about and so many different points of view that the provisional agenda was not really followed! Hence, after setting out the intended agenda, what follows is a general record of the various issues which were raised.

"The afternoon session is devoted to a discussion on the way ahead. This is not aimed at making any commitments but it is about planning and prioritising the forthcoming research work on freight modelling in the next 2-3 year's time. It is anticipated that the workshop will cover the following questions: