Transport modelling

Transport for West Midlands for West Midlands Analytics Forum

West Midlands Data Insight: Transport Modelling – an introduction

Welcome to the transport modelling part of the Data Insight website – so if you didn’t want transport modelling, time to select reverse now… or maybe it’s worth spending a little time here to find out if you do in fact want to use transport modelling?

  • Question: Given the need to build a lot more houses so that our young people have an opportunity to find a home of their own when they grow up, how would their access to different types of ‘life activities’ – work, education, shopping, leisure, healthcare, seeing their family and friends – be affected by putting those houses inside or alongside the existing conurbation, in towns peripheral to the conurbation, or in a new town?  And how would those different spatial strategies affect the performance of the transport network – e.g. road congestion, crowding on public transport, space for cycling?   
  • Question: We want to reduce the overcrowding of trains on an existing railway line – what would be more effective: longer trains, more frequent trains, or improving the parallel bus service so that some of the short-distance trips transfer from train to bus? And what would be the effect of each on congestion on the roads around the railway line?
  • Question: If we introduce a pedestrian crossing phase at an existing traffic signal junction, how much benefit would that provide to pedestrians and how much extra delay would it cause to people in vehicles and to freight movement?

If you are interested in the answers to these or similar questions, then transport modelling could help you decide what’s best – for your stakeholders, e.g. residents, industry, and the providers of services in both the public and private sectors, for the local environment and global eco-systems, e.g. climate change, and for obtaining good value from public spending [or for private business and/or third-sector organisations]. 

A note on formatting: we’ll be using bold for section headings and underline for links to other sections of this website and hyperlinks to other websites; so that leaves us using italics for emphasis, and single apostrophes to highlight jargon and/or technical use of ‘common words’ where we use them for the first time.  If this formatting doesn’t work well on your browser, please tell us and we’ll reconsider it; which gives us the first chance to say “We want your feedback!” – use the feedback form to provide comments.

What is ‘transport modelling’?  

Before we dive into transport modelling, let’s examine what we mean by ‘transport’ and what we mean by ‘modelling’.

Transport:   Movement of people and of goods between land-uses, e.g. from home to work, from wholesale warehouse to retail outlet, for the purposes of ‘gain’; whether that be economic, social, spiritual, recreational, or some other purpose that the person travelling or the organisation consigning the goods considers gainful.  It does not cover travel for the purpose of travel itself or for exercise, e.g. a sight-seeing tour bus, a leisure cycle ride, walking the dog, or going jogging.

Modelling:  Any simplified representation of actuality; thus: a photograph is a simplified version of a view; a documentary book is a simplified version of any episode in history, for example; and a house built out of snap-together plastic blocks no matter how elaborate is a simplified version of an actual house.  A mathematical model is a simplified version of reality expressed numerically and through formulae and algorithms that turn input data into output data that forms information.   From now on when we refer to ‘model[s]’ and ‘modelling’ we mean mathematical models unless we state otherwise.

What is transport modelling for?

Transport modelling is undertaken to help transport decision makers – and related decision makers, e.g. land-use planners, industrial investors, public health officers – to better understand the current, the future and sometimes the past transport systems, and to make better decisions about what to do.  Modelling does not tell us what to do: that is for leaders to do in conjunction with their stakeholders; modelling provides information, not advice or recommendations. 

We – the whole interleaved set of transport authorities, industries and regulators – use transport models for three reasons:

  • Anticipating the Future: To understand what travel problems will emerge in the future if we don’t intervene over-and-above the current transport provision and existing committed interventions, e.g. buses could become less punctual due to more private cars on the road; trains could get more crowded in the future and rail’s travel mode share growth become constrained; or residents’ range of employment opportunities within acceptable commute time could reduce as journeys slow down due to increased road congestion. 
  • Deciding on Interventions: To predict and thus appraise – meaning assess the value of – what the effects of possible interventions in the transport system would be, e.g. if we adjust traffic signal timings in near real-time does that improve bus service punctuality; how many extra train carriages would be needed to keep pace with future demand growth; and how could both the transport system and land-use pattern be changed to maintain a good range of employment opportunities for future residents.
  • Understanding the Past – to enlighten debate about alternative futures: For hypothetical appraisal of alternative past choices to facilitate public policy debate – e.g. what would obesity rates be like now if we had started promoting cycling to school fifty years year – or private dispute resolution – e.g. around hypothetical development for land that is subject to compulsory purchase order; there are rare applications, but modelling is the only way to answer “What if we had done …. in the past?” questions that we face in the present.  This type of discourse – comparing the actual now to other ‘nows’ that might have been can be more engaging for lay people than speculating about different possible futures that could come to pass based on decisions that we make now.  

There follows some further exploration of the scope of transport modelling, which you could skip past and “just get stuck in” by going to our how-to guidance for programme directors [coming in the future], handbook for project managers [currently being scoped] and manual for modelling practitioners [for PRISM 5: published in part, other parts being drafted; for other models: to follow]; but transport modelling is not flat-pack furniture so we advise that you don’t skip what follows below.

Types of transport problems, interventions, and models – from ‘strategic’ to ‘engineering’.

We are using ‘strategic’ to refer to transport interventions that affect how people live their lives, e.g. substantial land-use change such as urban area extensions or the changing an existing business park into a mixed-used employment, residential and leisure park; and changes in the transport system that are so transformational that they greatly affect where people choose to live, work and do other activities; and/or influence people’s daily and weekly activity patterns and not just the way they make particular journeys within their activity pattern.  [NB:  In this context we are not using ‘strategic’ as meaning large or long-distance.]

We are using ‘tactical’ to refer to transport interventions that affect the mode that people use and the route that people take to make a particular journey; and changes in trip destination that do not represent a change in the pattern of living, e.g. going to a different edge-of-town retail park than we used previously to buy white goods.  Thus, in tactical modelling the chain of trips between different land-uses – e.g. home -> work -> gym -> home remains similar even if the modes of travel, the activity locations, and the trip routes change.

‘Operational’ modelling is about how traffic units – e.g. vehicles on the highway network, passengers at a train station – move through a small part of that network but not about the routeing of that trip other than at the microscopic level, e.g. whether to make three left-turns instead of a right-turn, whether to use the north-end or south-end footbridge at a major station.  In operational modelling the trip routeing is fixed at the wider network level.

‘Engineering’ modelling is about how the passing traffic interacts with the transport system infrastructure, e.g. heavy goods vehicles wearing the road surface, pedestrians deflecting the deck of a footbridge as they pass over it [do you remember the Millennium Bridge debacle – see].  In engineering modelling the volumes and movement path of traffic units is fixed, e.g. the number of persons passing over the north-end footbridge at a major railway station; it is the infrastructure’s behaviours, not that travel and trip making activity, that is being investigated.


We have summarised these aspects of transport modelling below; let us know your thoughts on these definitions of modelling scope.


What are the limitations of transport modelling?  

Firstly, modelling does not give you the answer to your transport question; it provides you with information that can help you or another person in authority make a decision.  

To understand the limitations of transport modelling it is useful to think about the three stages of modelling:

  • In the middle we have ‘transport prediction’, or sometimes this part is called transport modelling but we consider that the term ‘transport modelling’ better encompasses all three stages. Predictive capabilities in a strategic and tactical transport model could include:
    • The extent to which car drivers re-route off a main road when they meet an unexpected delay or just stick on the main road accepting the delay.
    • How the combinations of travel time and expense influence whether someone takes public transport or uses their car for a journey.
    • How changes in travel time and expense influence someone’s job search area, there being for each individual only so much travel ‘cost’ they will incur in commuting.
    • How the overall nature of mobility across an area – how quick and how cheap journeys are between the whole range of origins and destinations – influences future land-use patterns in that area.
  • At the start we have the forecasting of model inputs. For example, take a model where the transport prediction functions use formulae that link an input, e.g. the number of working age people living in an area, to an output, e.g. the number of travel-to-work commute trips made in the morning peak period: we could forecast the number of commute trips in the future if we have a forecast of the future number of working age people living in an area.  Similarly, if our model can predict how increases in train service frequency leads to car drivers switching to train travel, we need to have a forecast of the future train service to calculate a forecast of future shift of travel from car to train.
  • When we have made a transport prediction, e.g. the shift of travel from car to train, we usually want to put a value on that, e.g. what is the benefit to those who switch from car to train travel, and what is the benefit to remaining road users who can use the roads with less congestion delay; and that means we need to have a forecast of the value of such improvement, e.g. how much each minute of journey time saving is worth to car business car traveller who gains when some car commuters have switched to train.


From this outline of the stages of modelling we can see that to undertake transport modelling we need:

  1. Predictive functions that link non-transport inputs to travel demand, e.g. people and volume of travel, and link travel conditions to trip making, e.g. times and expense by mode of travel to the split of travel into trips by each mode.
  2. Forecasts of the future inputs, e.g. future year population by age, future year transport network conditions and
  3. Forecasts of the values the society will attach to the benefits of having access to activities, e.g. being able reach a wider range of job options, and the characteristics of making trips, e.g. how much time the take.

Each of these three stages of modelling is prone to short-comings, whether they are:

  • statistical ‘error’, which simply means the situation when the size of our data sample means that there is a range of uncertainty about inferences about the whole population, not that we have made a mistake;
  • statistical ‘bias’, which simply means that our data sample is not a good representation of the cross-section of the population, e.g. contains too many older people and not enough younger people, not that we have some sort of bigoted attitude;
  • we make a methodological mistake, e.g. our model allows for trip-makers to choose between drive-by-car and public transport but does not allow them to choose between drive-by-car and take-a-taxi because when we developed our model in the past we didn’t anticipate how big a part of the transport mix taxis would become in the ‘future’, i.e. now; or
  • human operator error, i.e mistakes in common parlance, e.g. when coding a proposed train service into the public transport network, a train that is scheduled to depart at 12:20 is coded as departing at 12:02.

In addition to these short-comings in model set-up, there uncertainty about the future – which potential housing developments will be built, what train timetable changes be introduced, what the fuel-type mix of private cars will be in the future – that will affect predictions of travel behaviour and trip making.  This uncertainty should be managed, not ignored.  

Well-developed models will provide you with better information than just guessing, but all models are will have short-comings because they are a simplified version of reality, not reality itself, and contain short-comings and uncertainties.   To quote Box and Draper, 1987:

“... all models are approximations. Essentially, all models are wrong, but some are useful. However, the approximate nature of the model must always be borne in mind....” [1].

It is the for the developer of the model to be transparent about the data sources that they used to develop the model, about the mathematical functions within the model, about the uncertainty about forecasts of the model inputs; and for the user of the model to be questioning about how useful the model is in the situation they are applying it to.   Top tips: form a clear, ‘whole-system’ overview before you engage in modelling; challenge non-disclosure by anyone and everyone involved in the modelling; and be realistic regarding uncertainty about the future.  You will find more about systems thinking, scenario testing, realism testing and sensitivity testing in the programme directors and project managers manual [n.b.: not yet drafted – in the meantime, use the contacts in the ‘How to get more help section.

Models, sub-models and embedded models

Even the simplest of transport models typically comprise several sub-models, some of which are embedded and thus you might overlook them.  For example, an operational model such as of vehicle queuing at a simple side-road priority junction comprises: a description of the ‘use’ on the junction, i.e. (i) types and numbers of vehicles making each turning move at the junction; (ii) the physical configuration of the roads, e.g. at what angle and gradient the side road approaches the main road, and how many lanes at the side-road give way line; and (iii) a statistical distribution of main-road traffic gaps that are acceptable to side-road vehicle drivers, e.g. 10% of car drivers require a gap of 5 seconds or longer in the main road traffic in order to pull out.  Model (iii) is often embedded in the modelling software or is held in default input data and not specified in each application of the software; and thus many users forget that this sub-model is a specifiable input to the overall model. 

The guidance below is focused on strategic and tactical transport modelling, for which software with embedded sub-models are less often available off-the-shelf than is the case with operational and engineering models; but it is easy for users of well-established strategic and tactical models to forget the sub-models that have been embedded within them; for example: at one end, how the future population will segment into different family life-cycle and wealth groups; through to at the other end how observant of traffic lights car drivers will be.  In both cases it is easy to extrapolate the historically observed characteristics and behaviours of the transport system into the future; but such extrapolation should be an active, well-informed choice of the modeller, not a passively adopted default position.

Conclusion:  Transport modelling can be used to help identify likely future transport problems that could emerge; to decide how to intervene in the transport system to make life better; and to inform debate and resolve disputes about transport choices – past, present and future.  To be effective it should be applied:

  • at the appropriate level – from strategic and tactical issues through operational and engineering problems;
  • conscientiously, planned and executed with all parties being aware of assumptions made both explicitly and implicitly;
  • pro-actively anticipated future challenges and responsively adapted to emerging realities; and
  • with effective management and technical assurance.  

Transport modelling is only as good as those who choose to use it, those who manage it, those who undertake it, and those who use the model outputs in advice and decision.  It requires effective team-work and, not super-human .  This guidance is intended to help you be a good transport modeller; whether you are is up to you.


Our Transport Modelling Guidance:  So now we have given a bit of food for thought, your choices are to go forward to:

… or to leave now and go back to the Data Insight homepage.   If you are confused or dissatisfied or both, please leave comments on the feedback form – thank you.

Other Sources:   We will never claim to be the definitive source of guidance on transport modelling.  We have identified some other sources that you might useful be we do not recommend, endorse or in any way approve of the content provided by these other sources; it is for you to critically appraise the validity of their content to your modelling situation.

  • ‘Transport Analysis Guidance – contains guidance on the conduct of transport studies - set objectives and identify problems, develop potential solutions, create a transport model for the appraisal of the alternative solutions, and how to conduct an appraisal which meets the Department for Transport’s requirements; see NB: Do not stop here – WebTAG is not the be-all and end-all of transport modelling and your thinking you should not be constrained by sticking to Department for Transport’s guidance which we devised for its own purposes which are possibly not your purposes.
  • ‘Travel Forecasting Resource’ – a Transportation Research Board [USA] research project designed to examine the state-of-the-practice in metropolitan travel forecasting; see com/definition/mathematical-model.html
  • ‘Transport Knowledge Hub – Forecasting’ – a community of practitioners, modelling specialists and policy advisers that provides a forum for information sharing and collaboration on modelling/forecasting techniques, methodologies and related issues across the transport sector; see
  • ‘Australian Transport Assessment and Planning Guidelines: T1 Travel Demand Modelling’ –provides guidance for developing and applying strategic highway and public transport models when appraising major transport initiatives; focus in this volume is on the modelling of person-based travel demand across road and public transport networks; see
  • ‘Activity-Based Travel Demand Models: A Primer (2014)’ - This publication is a guide for practitioners that describes activity-based travel demand model concepts and the practical considerations associated with implementing them. Activity-based travel demand models portray how people plan and schedule their daily travel. This type of model more closely replicates actual traveler decisions than traditional travel demand models and thus may provide better forecasts of future travel patterns.; see

Are you aware of any other non-commercial sources of guidance?  If yes, please let us know using our feedback form and we will consider adding them to those listed above.

Thank you for reading and wishing you all success in your transport modelling.


Keith Homer

Transport Data Framework Manager, Transport for West Midlands. (11 June 2019)



  1. Box, G. and Draper, N, 1987: ‘Empirical Model-Building and Response Surfaces’, John Wiley & Sons, New Jersey; via Wikipedia, accessed 24th March 2019.