Part 1: Where is Cycling in Transport Modelling?

Back to Introduction

Large scale modelling of transport networks is rooted in a tradition that focuses on vehicular movement; originally many such models focused only on car trips, with a ‘mode choice’ element coming later. At a micro level, for example in terms of junction modelling, a key aim has been to get as many motor vehicles as possible through the junction, hence in the UK the multiplying of motor traffic lanes at junctions and the related disappearance of cycle lanes there.

I have been asking around about how transport models deal with cycling, to get a picture of where we’re at and what we want models for – rather than just including cycling in models for the sake of it, I was usefully reminded that we needed to ask what the purpose would be. Some people I’ve spoken to don’t think models should deal with cycling. When cycling rates are very low, any change can be drowned out in statistical noise. And cycling is locally variable. One person told me ‘I am suspicious of cycling modelling because propensity to cycle depends on so many local conditions.’

This reluctance to model cycling is understandable. The age of transport modelling is also the age of cycling’s decline.


And with cycling low and/or declining in many places, even when promoting cycling became part of transport policy, the data and techniques needed to deal with it in modelling travel behaviour and traffic flow have remained underdeveloped. However, given the substantial yet still very localised rises in cycle commuting in some urban areas, alongside some evidence (albeit contested) that ‘Peak Car’ might be on the horizon, the need for models that help us understand cycling is growing urgent. Leaving out cycling risks keeping the emphasis on cars, by not exploring how we can develop models that can deal with cycling’s ‘local’ factors (which, I think, do also apply with respect to other modes, in different ways).

Part of the hesitation over modelling cycling seems to stem from a feeling that cycling has some kind of an irrational or at least subjective element absent from other modes (except perhaps walking). This interests me as a sociologist. What we define as ‘rational’ and ‘irrational’ is often a good guide to what is seen as normal and privileged, I think. I’ve written a paper – currently under peer review – looking at the concept of ‘utility cycling’ as partly an attempt to make cycling respectable as a rational mode choice. Yet this rationality is potentially illusory: the fact that we can define car trips as ‘rational’ is somewhat strange given the focus of so much car advertising, which aims at generating purchases and trips that stem from the heart, not the head. (Leaving aside the collective irrationality of mass car travel).

Many modellers are in fact sceptical of the assumption that travel behaviour is rational, and one session at the practitioner conference Modelling World 2012 was themed on this topic. There are two issues. Firstly, equations involving a range of variables may predict driving behaviour, but that does not necessarily mean the correct causes of this behaviour have been identified, merely that behaviour has been simulated using the chosen variables. Secondly, even if we have got the variables right for now, what happens if tastes, preferences, habits, and so on change? What happens if the future does not (or given environmental pressures, cannot) replicate the past? What if future people are or must be different to ourselves? There is evidence that in London at least, young people are less interested in the car than in the past. They want their Oyster card and their smart phone, but a car just seems like something Mum and Dad wanted.

So I would be sceptical of the idea that cycling is uniquely cultural, irrational, subjective and so on (choose your own adjective, with positive or negative connotations, depending on your discipline and outlook…) It appears so, because models have been developed based around the idea that car travel (and to a lesser extent public transport) represents a rational choice, leading to a neglect of data and techniques that could accommodate cycling. However, it’s true that cycling implies the need at least to extend modelled rationality as it currently exists. Cyclists often don’t choose the quickest route (assuming they know what it is!), as it may be unpleasant and/or unsafe. Such calculations are not assumed to enter the brain of the car traveller, who has been shielded from thinking about the risk they run (or to which they expose others), and who is more alienated from the surrounding environment than is the cyclist. But decisions about whether and where to cycle are not inherently un-modellable; they just pose different questions to the ones we’ve been answering.

Forward to Part 2: Time, Money, and Routes

Bookmark the permalink.

3 Responses to Part 1: Where is Cycling in Transport Modelling?

  1. Pingback: Cycling and transport modelling - Modelling on the Move

  2. Adam J says:

    You raise some excellent points:
    -If we want to grow the number of trips by bicycle, will we have to include bicycles more centrally and strategically in modelling, rather than perpetuating the status quo? It can be done:

    -Rationality: I’d argue that riding a bike is not such a rational option when you have to share space with vehicles much larger, faster and harder than the bicyclist, hence the oversupply lycra bullets and other extreme sports enthusiasts masquerading as bicyclists on our streets. If the only safe place to ride is the footpath and an old law bans that, it’s perfectly rational not to ride.

    -What happens when tastes change? I’m reminded of the fact that a handful of years ago no-one had a smartphone, nor ever dreamed they’d need one; now we can’t imagine how we’d live without one. Sometimes the person producing a product has to preempt future needs and use framing/marketing/PR to help the people understand that the new product can improve their lives.

  3. Rosemary Sharples says:

    I’m a bit concerned that this discussion makes no mention of the work that has been done on modelling of cycling. My MSc dissertation from the early 1990s found various reasons that cyclists could not be included in a standard transport model:
    Rosemary Sharples

    Modelling cyclists in SATURN

    Traffic Engineering + Control, October 1993: v. 34, issue 10,
    pp 472-475


    This paper reports the results of a study which attempted to include cyclists with motorized vehicles in a SATURN transport model. The network modelled was a stretch of the Wilmslow Road corridor in Manchester, England. This major arterial road runs eight kilometres south from the city centre. It has the largest number of bus routes and the heaviest flow of buses in the county and the largest volume of cyclists in north-west England. The network was considered both with and without proposed bus and cycle facilities.

    Network data and trip matrices were supplied by the Greater Manchester Transportation Unit from existing models. The figures for cyclists came from screenline counts, and were added into existing trip matrices. It was necessary to estimate values for the traffic characteristics of cyclists, including average speeds, PCU ratings, saturation flows, extent of illegal manoeuvres, speed dispersion, platoon dispersion, sharing of lanes, interaction with other vehicles, value of time and gap acceptance characteristics, because of lack of information. As a linear network, knowledge of cyclists’ route choice was not required. However, the differences between cyclists’ origins and destinations (university halls of residence and buildings) and those of motorized traffic led to congestion. Limitations in the software included restrictions on the number of entry lanes to a node, lack of provision for extra classes of vehicle to have distinctly different average speeds and outputs which do not take into consideration different types of vehicles. The steps taken to overcome these problems are described.

    Several conclusions can be drawn from this study. Bicycles are not little cars. Cyclists are not just motorists in another type of vehicle, albeit one without an engine. Software designed for motor vehicles is not suitable for modelling cycles. Much research remains to be done on cycle traffic characteristics.

    Keywords: cycling; modelling; cycle speeds; traffic characteristics

    Howard Boyd was also working on cycle modelling at the same time and he made use of the Dutch Quo Vadis/Bicycle model for Trafford Borough Council. I don’t know the history of Quo Vadi/Bicycle but presumably the Dutch have done some work on modelling cycling.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.