Modelling Tomorrow's World: what needs to change?

Hugh Neffendorf is the founder and Managing Director of Katalysis Limited

Hugh Neffendorf is the founder and Managing Director of Katalysis Limited

'Times are changing,' says Hugh Neffendorf, founder, Katalysis Limited, and one of the coordinators for the Transport Systems Catapult-organised Modelling Tomorrow's World. Our models need to change too, he says. 'We need to ask whether many of the models we use today are fit for purpose in terms of modelling for a very different future...' 

The models we use today, says Neffendorf, in conversation with Modelling World programme director Juliana O'Rourke, tend to be based on how people behave today, or are projections from trends or from cross-sectional analysis of people's current behaviour.

But the disruptions that the transport sector is facing, and the dramatic changes in society, technology and information, mean that such a basis no longer holds true. 'We must think about what is changing, and how we get to where we need to be. Developed countries are not escaping from pressing economic constraints, nor from environmental challenges.

These need to be taken into account, and we do not really model them today.' Modelling in the face of unprecedented change is one of the key aims of the Modelling Tomorrow's World event, he adds. 

'We still need to make the best possible predictions because, even in a world of uncertainty and risk, governments want to plan, and businesses need to plan. The problem is that it is getting increasingly hard to do it because the world is not stable. Especially in terms of what is going to change.' The challenge is that we have to move to new ways of modelling while still being able to assess schemes, policies and investment decisions in a consistent or realistic manner. We can't just throw everything away. We need a transition from current WebTAG-type models, towards future-sensitive models that can be trusted to give us a reliable sense of priorities. Currently, models have at their basis the idea that people of a certain type behave in a certain way, he suggests. But when you face them with choices that they don't have today, then they will behave differently. One of the main reasons for doing this now is to ask: how do we model the future if we don't know what it's going to be like? 

Issues for Modelling Tomorrow's World 

Types of changes

•  Demographics
•  Land use changes
•  Economy
•  Environment
•  Attitudes, expectations and behaviour
•  Lifestyles and travel habits – age groups differ
•  Transport delivery
•  Information to travellers
•  Data sources

Models that are being challenged

•   Car ownership
•   Trip rate categories
•   Values of time
•   Mode choice – very different types of modes and inter-modality
•   Generalised cost – much more complex; more factors
•   Route choice – no longer a matter of choice; directed or information en-route

Many types of model needs

•   Operational
•   Near term forecasts
•   Data blending and enrichment
•   Transport planning – schemes and policies under new assumptions
•   Uncertainty and risk
•   Information in the loop
•   What if?
•   LUTI – new issues to be considered

Model solutions

•   Activities – family dynamics; age groups
•   Simulations – large scale, city level. Person choices, not just vehicle and pedestrian agents. Synthetic populations. Tools and methods from other domains
•   Scenario models – many sensitivity tests
•   Intermodal, whole journey models
•   Faster, bigger models (array computing)
•   Systems of systems – lessons from other domains

What to do?

•   International view
•   Systems view
•   Research programmes
•   Improved guidance
•   Clients should request new insights and methods
•   Live trials – real time/near term
•   Keep spreading the message – More events like Modelling Tomorrow's World

Speakers at Modelling Tomorrow's World will be contemplating what 'tomorrow' in developed countries might be like, based on what we know today. We do know much of what is changing, says Neffendorf, but not necessarily its impact. Firstly, demographics, as we have ageing populations, more people living in single occupancy or shared homes, ethnic concentrations, and dramatically pronounced social groups. We shop online. Our behaviours, expectations and our lifestyles, are continually being modified by new opportunities. For example, says Neffendorf, increasing numbers of professionals, particularly younger ones, have meetings by text or online. They no longer need to always be in a traditional meeting environment. We need to think about segments of the population differently, for example we are going to have to start thinking about age-based travel segments. It's no longer one household making trips. Adults might engage in one set of behaviours, but teenagers often have completely different outlooks and activity patterns. 

For example, says Neffendorf, one colleague told me that his daughter and her friends, who live in the USA, are moving to an area better served by Uber. They are moving their homes because they no longer want to own a car. Traditionally, people may have sought homes close to railway stations or metros. In future, people will also be seeking out concentrations of Mobility as a Service (MaaS) operations as well as public transport offerings. People will clearly do more online shopping; younger people will do more home working and 'network' working. They will want to live relative to their activities, but these may no longer be dictated by work places. Place will not be as stable and static as in the past. Changes in society are one aspect of the need to evolve our models; another key issue is transport supply and the delivery of transport. We're looking at a huge increase in the number of travel choices, and more dynamic associated information availability, but how do we model this? 

As MaaS puts pressure on traditional public transport to survive, public transport will inevitably shrink as a result. Local authorities are going to have to decide which aspects of MaaS they subsidise – instead of buying socially necessary public transport they'll be buying socially necessary MaaS. And they will need information coupled with models to determine which are the most progressive investments in terms of society. 

Impacts on land use

Transport changes will also have inevitable impacts on land use, comments Neffendorf. As MaaS and, potentially, autonomous vehicles come on stream we will need to build fewer roads, and we won't see so many facilities along the roads we have. We may see less 'traditional' parking, but will need more space for MaaS vehicles on the kerbside. Our urban spaces could look very different. 

The other big change in transport is that we now have access to vast sources of data, he continues. As we engage with these, we are learning whether they are useful, reliable, or biased and misleading. Is big data really the modeller's Holy Grail? We will soon be dealing with real-time data – from mobile phones and other sources, and no longer confined to survey-captured data. Having access to real-time mobile phone and passenger travel data means more than the fact that we no longer need to undertake so many surveys: it also allows for more timely and higher-resolution data to help improve our models. Some people say the data is the model; I don't necessarily think that this simple statement is right but there is no doubt that the data, coupled with suitable forecasting methods is a great foundation for future models. 

Another disruptive element in the mix is that is passengers are being fed information as they're on the move, instead of pre-planning their journeys. Passenger transport users with smartphones are being advised of their travelling options whilst on their journeys. Many drivers don't question their SatNavs; don't question the information they are being given. If models that work out which routes people are going to take are no longer based on what goes on in people's heads, but are based on Sat Nav data, then we need to know which proportion of the population uses Sat Navs, and how implicitly they trust the information given. 

From pre-planning to decisions on the move

Currently, our models don't work in terms of people getting information while they are travelling, they work in terms of pre-planned travel. This is yet another aspect of modelling that is going to break down. Many other issues on which models are based are also being challenged, for example car ownership trends, the value of time, and modal choices. Many people no longer wish to own a car. Current thinking on trip rates is based on households of a certain structure. Once we start breaking these structures down, into individual types of people such as the elderly and young people, then the old equations simply don't fit anymore.

Mode choice modelling faces special challenges, says Neffendorf. Today we can say, here is your choice of modes – bus, train, car, walk, cycle – and we can calculate them. But MaaS and ride-sharing service offerings are much harder to build into the network in terms of generalised costs. 'We currently describe things in terms of time and money, but in future will be faced with much shorter-term decisions as new choices – and rejection of choices – become instantly available. Introducing new types of transport and, ultimately, autonomous vehicles as mode choices is much more complicated than today's static supply networks. I already mentioned route choice, with many people simply following instructions. And although dynamic algorithms can be built into models, this is not the way it's done today.'

The modelling that we'll talk about at Modelling Tomorrow's World is not only strategic transport planning modelling, done by consultants for cities, he adds. It's also modelling for transport operators. Nor will it be just about 25 years in the future, but also about what is happening now, and what's going to happen in an hour's time. We will be talking about near-time forecasting which we can do when we have real-time data mixed with models, even built into solid-state devices.

One of the big things we are faced with in a changing world, and also the hardest thing to build in to modelling today, Neffendorf muses, is uncertainty. 'We can talk about new mobility services but we don't yet know their impacts. We are moving away from cause and effect to cause and uncertain effect. This happens in many walks of life, for example the financial world and meteorology, but transport modelling hasn't really embraced uncertainty because it's difficult. It requires many scenarios which need to be weighted. Typically, a 'traditional' transport planning modeller from an engineering background will cope with the idea of high, medium and low estimates, but uncertainty as a statistical quantity that you can actually put weights on is much harder to work with. However, I believe that we have to rise to the challenge, and that uncertainty will need to feature in modelling guidance in the future.'

Dealing with uncertainty

Politicians and decision-makers struggle with uncertainty, as they prefer a straightforward recommendation from an expert, rather than a range of options and possibilities. But however inconvenient, it's wrong of us as professionals to ignore this. Many existing modelling initiatives, for example the modelling of HS2, will be hugely challenged by changes in society and in technology, and I'm not at all sure that they have a way of taking this into account. Are they re-running the models with autonomous vehicles in the mix, for example? Many current decisions have been taken on far too simplistic a basis, in my opinion.

We must remember that we will need a transition period. We need to know that the decisions we've taken are robust, even if something unexpected should happen. There is scope for moving to new types of modelling, some of which may be familiar, but not yet in common usage. 

At Modelling Tomorrow's World we will hear about modelling family activities as opposed to modelling household groups. We need to be more realistic and segment the family into its needs and behaviours over a day. Such an approach is much more widely used in the in the United States, but recently authorities in the UK, for example Transport for London, have undertaken interesting and relevant exercises that involve identifying neighbourhood characteristics as part of the definition of their population base. Something we don't know, and won't know until we experiment, is how much better the models become because of this segmentation in prediction. They should be better because we will be breaking things down into behavioural units that matter, but whether the output will be significantly different, we don't know. 

Simulation impacts

Simulation will have a big impact on models in future. Improbable will be speaking at the event about its big simulator, which deals with individual simulation agents (units) that represent decision processes during an active day. Improbable is also building agent persistency into the model, meaning that the same person (unit) can be tracked for a complete time period or day. 

We will need more 'quick look' scenario models and 'what if' scenarios. I think we'll see a lot more of this type of model. Some of these models and simulations will be very hungry for computer time so we will also see more arrays of computers, and Improbable and others are talking about running simulations in the cloud. Consultants are already buying arrays of computers in order to manage multiple model runs much faster, so enabling better and more rapid testing. 

Another thing that's going to happen comes under the heading of 'system of systems'. I think of this as a military expression but it really means that things are interconnected. Transport, environment, health, the economy. At the moment each has its own models but we need to bring them together. We can no longer say, 'let's model transport and have constraints on environmental issues'. As previously mentioned, developed countries are facing increasingly complex and pressing economic constraints, not to mention serious environmental challenges. 

There are types of models 'out there' that look promising, and need to be worked on, says Neffendorf. It's not easy to do this because it's tough for consultants to explain to clients that they wish to use a new method, or one that may be less well-known. Clients don't say: I would like you to adapt a conventional model to study how MaaS will change mode choice. Nobody requests such experiments as part of the standard study on Leicester or Woking; instead such initiatives are treated as add-on research projects. Bringing new thinking into mainstream planning is a key challenge, but one we have to tackle. We'll need a lot more experimentation and research programmes than we have today, and it's not at all clear who is going to pay for this, says Neffendorf.

Pathways to change

There are people who get it, who realise how important this is, and will be communicating this to their clients. We will also see organisations like the Transport Systems Catapult (TSC) doing more facilitation, he adds.  TSC is not in the business of mainstream transport studies, so is in a good position to work between potential funders such as Innovate UK, DfT or Highways England, and encourage them to move away from the mainstream. And to experiment more. Much more. 

There are several UK local authorities that are beginning to move in the right direction, says Neffendorf. Oxford's Llewelyn Morgan will be speaking at the event; Oxford is alert to the fact that they are being faced with a whole set of new challenges that they can neither predict nor model today. Bearing this in mind, they are beginning to develop new data sources, tools and processes to bring new thinking into their modelling and data acquisition approaches. This is to help them to plan better, but they are also alert to the need to bring new data and methods into standard national guidance for scheme and policy appraisal.

Transport for London, Cambridge, Exeter and Manchester also have units which are beginning to think outside the box. Some, for example Cambridge, have funding which is embedded within their Smart City Deal. Exeter has funding from venture capitalists who, via Exeter City Futures, are helping the city to tackle key challenges like transport and health together, as a community. Change and innovation will also come from other directions, for example transport operators beginning to invest in near-term modelling in order to better understand future impacts. The TSC University Partner Programme is also expected to help deliver new insight to modelling approaches for intelligent mobility.

However, it remains true that many transport operators need to move further and faster in considering how their services may need to change, says Eifion Jenkins, Programme Director, Transport Systems Catapult. Technology is adding an extra layer onto public transport services, so putting further pressure on asset providers – technology such as apps hold the key to interaction with customers, and therefore to profit, he notes.

This event is about how we might begin to model inevitable new horizons, including developments such as MaaS, explains Neffendorf. Software providers like PTV are already beginning to build MaaS 'modules' into software, for example ride-sharing, as this increasingly becomes part of the modal choice. The models must explain the choice set and the growing complexity of intermodal options, and agent-based modelling is one way of doing this.

We simply can't keep modelling the way we do now. It will increasingly become derided, to the extent that we won't be able to convince decision makers. Models are breaking, and it is essential that the message gets across that we need to experiment with techniques that are more sensitive to apparent changes. This event is designed to do just that. As we understand which issues people want to study in more depth, TSC will organise further workshops and seminars. We want to engage with the modelling community, and this event is the starting point. 

Modelling Tomorrow's World is on 23 November 2016 in Milton Keynes.

View details online

Hugh Neffendorf is the founder and Managing Director of Katalysis Limited. Prior to 1997, Hugh was Managing Director of MVA Systematica, a prominent consultancy in software and information management. He is also Visiting Professor at the University of Southampton (Geography), an Ordnance Survey Accredited Consultant, a retained consultant for the Office for National Statistics and was a member of the previous UK Government's Advisory Panel on Public Sector Information. He is advisor to the Transport Systems Catapult on modelling, data and analysis for intelligent mobility.