I start this post with a confession to make. While I claim to be an expert in many areas of transport, one area that I have always struggled with is transport modelling. Honestly, ask me anything about models outside of the basic 4 stage model and be prepared for a blank expression. But the art of being a professional is not necessarily to understand all of the technical details of each area of your craft, but to understand its importance and relevance.
Transport modelling is both very important and very relevant. The very fabric of our cities, towns, streets have been determined by computer programmes like ARCADY, LINSIG, SATURN, and VISSIM. In all of this logic and mathematical equations based on vehicle flows and cost-time calculations, many different highway users have lost out. Cyclists are one of the biggest ones.
But there has been some movement recently to challenge this orthodoxy, to see how the profile of cycling can be pushed further up in the modelling world, led by the tireless work of Rachel Aldred. Just over a week ago I went back to my old stomping ground of the University of Westminster to attend Modelling on the Move 6 – Cycling and Transport Modelling. After a bit of pre-seminar swotting, and chats with colleagues far more knowledgeable about this stuff than me, I entered the seminar interested in the answers to 3 questions.
Should cycling be modelled at all?
In an ideal world, transport models are just one factor in a much wider decision-making process that ultimately leads to the delivery of transport schemes. If a model says that a junction will fall over in vehicle capacity terms, do the other benefits (health, societal, economic etc.) of a particular scheme outweigh the model and result in a scheme being built?
This raises a fundamental question of whether or not we should seek to represent cycling in our transport models at all. Considering much wider benefits of cycling are very well-documented, is it worth the time, effort, and expenditure to completely redesign transport models for them? Transport for London certainly seem to think so. So much so, they are actually ahead of the Dutch in doing this.
One of the common areas that most of the attendees seemed to agree on was that transport models – rightly or wrongly – have a significant influence on transport decisions. Roger Geffen put this point across very well, whereby the new roads policy seems to be almost built on making the forecasts of the National Transport Model come true.
A matter of interest to me was there were two presentations during the day that did not even mention traditional transport models – those of Herbert Tiemens and Paul Schepers, both transport friends from the Netherlands. I am not sure what role traffic modelling plays in the redesign of streets of junctions – I am sure that it plays some role. But the feeling from the presentations, and the research that I have done, seems to indicate that model results are very much a secondary consideration.
Perhaps this reflects the relative importance that professionals from different countries and contexts give to model results. In the UK, the model result is almost sacrosanct – we cannot distress the model, annoy it, make it fall over. We don’t even want to test to see if model results are correct. In this context, modelling cycling is needed not because it is the right thing to do to come to a transport decision, but because out decision-making philosophy dictates it.
Do our current models take account of cyclists adequately?
It is safe to say that almost everyone agreed that current models do not take account of cyclists adequately. And a common theme ran through this – adequacy of data.
As explained helpfully by John Parkin, in a typical junction model cyclists are assigned a value of 0.2 PCU, or Passenger Car Units. This figure is based upon quite simplistic research into the matter – assuming that cyclists have the same gap acceptance, relationships with other users, and headways as car drivers.
More research is required into virtually all aspects of cyclist behavior and interactions with other vehicles at junctions and across networks. Is there tighter gap acceptance and does this vary with different vehilces? What impact does the faster acceleration time of cyclists have on capacity? How does cycle-friendly junction design impact on behavior? Some significant and expensive research needs to be undertaken into all of this if we wish to model cyclists properly.
Examples given by Transport for London and Tim Gent for Cambridgeshire represent a decent first step on the road to better cyclist representation in transport models. It shows modellers fundamentally rethinking how their models are structured and undertaking some research on cyclist behavior – even if there is a critical lack of good data validation in both examples. That and cyclists are simply being retrofitted to what are still traffic models at heart.
A point made often in the seminar was how it was impossible to model cycling owing to the variability of current cyclists themselves, and their decision-making processes. In my view this is a side issue. As John Parkin touched on, micro-simulation modelling is increasingly introducing this variability into modelling of highway networks. Also, a model by its very nature is a simplification of a vast number of decision-making processes based on observed behavior. Nobody can say that every driver’s responses are reflected in many transport models at the individual level, but we do understand what the average driver, and drivers as a collective, are likely to do because we have good quality data to support that.
How can our models be improved?
I have already touched on the improvements to transport modelling being undertaken in London and Cambridgeshire, and the issues that they have faced. I wish here to concentrate on another matter: modelling the unknown.
As I have already touched on before, current transport models are based upon projecting forward behavior that has been observed, and thus they are only as good as the data that we have at the time. For cycling, as Nicholas Sanderson points out, any data collected on current cyclists now may become obsolete soon if cycling becomes more normalised.
In this sense, transport models are always playing catch-up to observed trends. Perhaps modellers could make much more use of stated preference data (for all its flaws) in future demand forecasts, but even this is an uncertain science.
Or perhaps, just perhaps, we could simply not rely on a computer programme on a laptop to help us make a transport decision? Nah, that’ll never catch on.