There is no business as usual when modelling climate change

We live in a disequilibrium world, where chaos is always close and business as usual an illusion. Model-based assessments are under challenge, says Phil Goodwin. In these circumstances, the roles of scrutiny and challenge are supremely important

New ferry routes, connecting ports in Ireland to ports in Portugal, Spain, France, Netherlands, Belgium and Germany, as a substitute for the ‘land-bridge’ through the UK. There are obvious non-trivial implications for ports, road traffic and services in Wales and England. Image © BBC

Modelling climate change was a key topic at the inaugural Modelling World International 2021, which took place in April. My presentation tackled the challenge of modelling, forecasting and appraisal when there is no ‘business as usual’; an issue that modellers are working hard to get to grips with, and which will be a key topic for the forthcoming Modelling World on 5-6 October.

Embedded in decades of transport modelling is the assumption that the future will be sufficiently like the past that any relationships observed (or thought to be observed), were stable enough to use as a guide to the future, and reliable enough to support decisions, with a few modest caveats. This philosophical starting point underpinned the rationale for the key transport models used to forecast future movement and support major projects. Brexit, Covid19, and climate change all challenge the credibility of that view. 

Brexit and Covid-19 affect transport

Both Brexit and Covid-19 affect economic growth, structure, behaviour and the geography of movement. Therefore they will change the traffic forecasts which have been used for the current multi-billion pound programme of road projects. 

To take one example, within the first couple of months we already saw one effect, a new pattern of freight movement from Ireland to the rest of Europe: new ferry routes, connecting ports in Ireland to ports in Portugal, Spain, France, Netherlands, Belgium and Germany, as a substitute for the ‘land-bridge’ through the UK. There are obvious non-trivial implications for ports, road traffic and services in Wales and England.  

Now such a shift in a well-established pattern of movement from a well-defined set of origins to a well-defined set of destinations, due to a change in the relative cost of different routes, is exactly the sort of thing that the big conventional models are able to deal with. A decent model would have been able to give quite a good prediction of this aspect of a future different from the past. The problem was that no-one wanted to know: not a failing of the models, but a failure of imagination or courage in the way they were used. 

The lessons from Covid-19 are rather different. The near inevitability of some sort of pandemic has been so predictable that, years ahead of the reality, many countries had undertaken detailed simulation studies about what to do. 

The effects on travel choices were uncertain, but it was always obvious that transmission would follow aviation routes. Epidemics become pandemic by international travel, which happens in timescales of hours or days. 

Vaccines can be produced – far quicker now – but still in timescales of months at best. Therefore, control of viruses (or mutant variants) require much more immediate control of international travel. That is an issue of transport planning, and surely modelling has something to contribute. Even before the end of this pandemic we are aware that it will not be the last virus to confront. The possibility of serious interruptions in travel, requiring very rapid control and logistical complications, in effect become part of the definition of business-as-usual, with indefinitely continued uninterrupted growth being a limiting case, not a confident secure future.  

Behaviour changes during lockdown

As far as I know, there were no uses of transport models to address the effects of the type of constraints on movement that would necessarily be involved, or to optimise their application. It would have seemed a waste of time to ask models to forecast the effects of something that was so different from the conditions during which they were built up. (It would still have been very informative to find out how good – or bad – their predictions would have been).

Anyway, the pandemic happened, uninformed by modelling, and we learned, as though it were a surprise, how swiftly, and in how many dimensions, travel behaviour could, and did, adjust. Is that a completely original finding? Not completely. There had already been many examples of disasters which have affected travel patterns very substantially and very swiftly; volcanos, earthquakes, tsunamis, bridge collapses, for example. 

We have also had decades of experience of implementing policies and transport provision which give positive feedback to trends, for example embedding car dependent lifestyles; or in seeking to reverse that positive feedback with initiatives such as the pedestrianisation of town centres. We have already observed these things taking place. 

So why is it surprising that the boundaries of human and political response to crisis, or to positive feedback from policy, are quite different from those which appear when we imagine we are in a state of steady, uninterrupted, growth? 

Mainstream modelling frameworks have not been built around that understanding, but this was a choice. The observations were there to be seen, but the choice made was that this is not the core business of models. 

Climate change – what if decarbonisation fails? 

‘Business-as-usual’ is also the casualty of what we know about the most important environmental issue of our time. We face two crucial scenarios which affect the future of travel, but neither of them are taken seriously in transport modelling. The first is that we completely ignore in mainstream transport modelling is the implications of failure, at a global level. Suppose human societies fail to change the production of CO2 and other atmospheric drivers sufficiently to halt warming. We are advised by the climate models that this will mean, among other effects, a rise in sea levels drastically affecting coastal populations. It will also convert many currently hot locations into places that are simply unliveable. 

Taken together, these two consequences directly affect some of the wealthiest, most pleasant and desirable locations in the world, and also some of the poorest and most challenging. 

Both would mean mass population movements, maybe in the order of a billion migrants, starting in our own (possibly shortened) lifetimes, and including many of us. It is inconceivable that current patterns of international movement (and national movement in countries whose geography stretches across equatorial, temperate and cold regions), could then continue or recover steady growth along past trends. 

Now that outcome is of course not in the hands of any one country. But if that is our future (whether we have done as much to avoid it as we can, or not), then the idea of our present infrastructure and transport management needs being usefully informed by business-as-usual traffic growth is utter nonsense. If planning is useful to cope with that scenario, it will be about flexible and resilient transport and living arrangements, not infrastructure designed to cope with business as usual traffic growth, and last for a century. 

The engineering model will not resemble the norms of civil engineering we have inherited, but its predecessor, military engineering: temporary encampments, pontoon bridges, emergency logistics, the management of huge population movements – traffic movements – under pressure. 

There could be a vitally important role for contingency modelling to plan for such huge changes in our own population movements and changes in economic structure. I do understand a resistance to modelling of this sort, because it might be seen as preparing people to accept such preparations and have undesirable effects on social order. But ignoring it comes close to assuming that the disruption would not happen. 

And what if decarbonisation succeeds?

Refusing to model failure seems unwise. But refusing to plan for success is mystifying. What happens to transport if, at a global level, we are able to reduce carbon output sufficiently to avoid calamity, forced evacuations and migrations?  

I’d ask you to agree that among the many conditions for such a success would be rapid progress in reducing the volume and distance of travel substantially. There would be a near disappearance of fossil fuel vehicles, except possibly for museums or ritual purposes. We would need swiftly to have achieved a reduction in the total volume and distance travelled by vehicles – including electric ones – whether cars or planes. This would then also help to reduce the associated carbon emissions from vehicle manufacture and road construction. 

Now I know there are some who say this is not necessary, and electrification will allow business-as-usual growth. I would assert that it is already clear that this trajectory is insufficient, and many Governments also assert that. It is commonplace in official statements that we must use our cars less. We must favour walking and cycling and buses and metro systems. We must replace short distance flights by better trains. This is no longer a minority or crank idea, it is the very widespread received wisdom (in policy statements, anyway). 

So why is there such resistance to making the ‘success’ scenario a central case for project appraisal, and the agenda for systematic, appropriate modelling of what projects and policies would most swiftly bring this success about?  

What we mostly model is neither the success nor failure of stopping global climate change, but a strange future where the climate is not different enough from the present to have any impact on travel patterns, and the travel patterns have not changed enough to have any effect on climate, and the rest of the world is not different enough to have a material effect on the patterns of trade which affect either our imports or our exports. And that such a state can continue for 60 years or maybe 100. This is the least probable scenario, and the classic inclusion of a few minor sensitivity tests of plus or minus 10% here and there does not remotely address what would be happening. We have two radically different futures in front of us – both entirely possible – but we choose to model a third, utterly unconvincing one. 

So how to promote healthy scrutiny and challenge? Does it require independent arbitration? Separation of client and promoter? Open access to the models themselves? Or statutory funding for challenge?

What future for modelling?

The problem in that it is not clear that models, whose history is the assessment of projects to provide infrastructure for permanent growth in movement, are capable of providing good answers to either scenario.  

The first reason is obvious. A model that only includes vehicle travel cannot be used for optimise sustainable transport policies. This is surely self-evident in towns, but I’d say it affects the strategic network also, because any behavioural model has to allow for redistribution of the matching of origins and destinations,  not only mode choice, and certainly not only route choice. 

Now surely one of the robust findings, from half a century of modelling, is that if costs increase over the whole network, redistribution will shorten average journey distances, and vice versa if costs reduce. What that means is that the mechanisms of changing mode choice interact with those of destination choice. Policies which make local movement easier and more attractive – by local public transport, walking and cycling, and indeed land-use planning – must also affect the amount of long distance travel. 

We cannot implicitly assume that the journey length distribution is a stable modelling parameter, or we would substantially underestimate the potential. We would get an answer, but it would be wrong. 

There are some in-built features of the best established models which tend to underestimate the dimensions and speed of behavioural change when needed, and therefore contribute less helpfully to sustainability.

The modelling concepts of equilibrium, and steady uninterrupted long-term growth, do not cope with asymmetrical relationships; for example with price or income, or cohort effects of aging. We do not have an approach to transport pricing which copes with an electricity market paid for at different frequencies, different units, changes prices from minute to minute, offering opportunities for hedging and buying in advance or smart charging in two directions. We have not taken seriously the crucial importance of timing and speed of change and sequence of policy interventions. 

We live in a disequilibrium world, where chaos is always close and business as usual is an illusion. There is no ‘usual’. Therefore, I would say we need dynamically specified models which can accommodate imperfectly reversible relationships, discontinuity, and path dependance, and give outputs of an evolving, uncertain pattern in explicit time, not an equilibrium one.

The modelling industry – including many LTT readers (and writers) – is an ecosystem of social relationships between clients, contractors, vested interests and users. The main clients, public or private, are often the main funder and promoter of specific projects. The modellers’ own business plan needs contracts for the sort of model they are good at. 

All these parties are co-dependent, and sometimes complicit. When things go wrong – like the Australian experience of modelled forecasts of toll revenues which led to investors losing money – there is a row about whose fault it was, but mostly the errors are not apparent until years or decades later, and by then there is no one accepting responsibility, or for that matter interested in whether error existed. Models are complex, very rarely transparent, and almost never fully open access. They are run by a sort of priesthood with lengthy training, and dependent on their patrons. 

Models under challenge?

In these circumstances, the roles of scrutiny and challenge are supremely important. 

What we are seeing now is that model-based assessments are under challenge in circumstances outside professional conferences and official reports. I’m involved in one at the moment as an expert witness in our High Court, suggesting that official figures underestimate the carbon consequences of road projects. The action was initiated by a campaigning group, and recently reported in The Guardian, in an article then retweeted by, among others, Greta Thunberg to her global followers. These are not contexts that most of us are familiar with.  

In my experience, the main active challenge often comes from political groups which are out of power, or have recently acquired it, and environmental campaigning groups which are well informed, but poor. Statutory restrictions and procedures do not always make scrutiny easy. 

So how to promote healthy scrutiny and challenge? Does it require independent arbitration? Separation of client and promoter? Open access to the models themselves? Or statutory funding for challenge? Could we have a consulting industry of skilled modellers who would be prepared both to provide the models for a notion of ‘business as usual’, and also to critique them? Is there a market niche for professional modellers to make a commercially viable living from representing the challengers? Would you want to? Would you dare? The social and market framework supporting modelling will change. There is a role for models, and modellers – many roles – but they will not be the same as those we have grown up with.  

Phil Goodwin is Emeritus Professor of Transport Policy at UCL and UWE, and Senior Fellow of the Foundation for Integrated Transport. philinelh@yahoo.com 

A recording of the conference presentation is available, with additional graphics: https://tinyurl.com/j6ymuzju