Transport planner Tom van Vuren believes that in post COVID-19 era transport modelling will be more important than ever.
COVID-19 has had a significant impact on transport, with far fewer people flying or travelling by public transport. There is huge uncertainty over future travel patterns and what policies will be introduced to help the sector deal with the ongoing effects of the pandemic. This unpredictability raises questions about the value of transport modelling. Should we be modelling at all or do we need a new approach?
Even before COVID-19 changes in travel patterns raised questions over whether it was sensible to continue to use the fixed travel behaviour assumptions calibrated in our models. Emerging modes, such as micro-mobility, shared demand options and connected and autonomous vehicles, could render existing modelling methods, built up over decades and supported by significant investments in data, software and human expertise, ineffective.
Activity- and agent-based models promise change – but do they solve the problem or are they an expensive, unnecessary and risky distraction?
At Mott MacDonald, we are convinced of the continued need for and value of modelling. Decisions on investing in transport infrastructure rely on access to the best available evidence and models remain a key part of the process.
In these uncertain times, we have been working with our clients to make best use of their existing analytical assets to answer questions they face now as well as those that are emerging.
How have we done this without breaking the bank at a time when budgets are under pressure, and how can we build on this experience? Four key points emerge:
- The need for a sound starting point remains – Our base year models are as up–to–date as possible. 2020 has been so turbulent that it is not sensible to use it as the basis for forecasting. Instead, we have used passively collected data, such as mobile phone traces that are routinely collected and stored by network owners, to estimate a 2019 base year rather than do a wholesale re-estimation. Historic speed and travel time data are widely available commercially, while many traffic survey companies also store their past data – a possible goldmine for model estimation when data collection is not possible.
- Embed all uncertainty in the forecasting, and not in the base year representation – Forecast uncertain behaviour, but do not include in the models. It may be easier, for example, to model the impacts of homeworking on the number of commute trips within the model itself, rather than the actual amount of homeworking people choose to do. The latter can be tested by assuming different levels of response. Public transport capacity assumptions depend on not just what is safe, but also how operators emerge from their current financial difficulties. In the absence of reliable insights, we can make a judgement call. Ideally, these ‘what-if scenarios’ will be informed by new, and focused research. This will improve our understanding and generate plausible parameters, but without immediately including them in our models. Adding complexity to the models risks making the ‘black box’ even more opaque and inaccessible.
- Population segmentation – Not everybody has had the opportunity to work from home during the pandemic and capacity restrictions on public transport have affected some population groups and job types more than others, while access to a private car remains unevenly spread. Greater segmentation in our modelled population is an efficient way to improve detail in analyses – particularly the outputs. Whether it is data obtained commercially from organisations such as Experian or Census-derived data from the Office for National Statistics, our analytical teams have significantly increased segmentation to reflect different travel behaviour responses to COVID-19. We have done this in such a way that the bulk of strategic models – eg PRISM in the West Midlands, LCRTM in the Liverpool City Region, the South East Wales Transport Model – remain unchanged.
- Greater detail – Consider how to increase the network and zoning granularity and spatial detail so that active modes can be represented much better in the modelling. It is possible to link network models of differing levels of geographical detail – we already do this for highway and public transport networks. These refined spatial models could also reflect the use of slow modes as access and egress modes for public transport, increasing the visibility and value of such trips in the overall travel mix.
Innovating for the future
At the same time as incorporating changes to reflect the realities of the pandemic, we continue to invest in innovation. In the West Midlands, we have implemented a pilot multi-agent activity-based model as an alternative to the demand forecasting component in the PRISM strategic model. Its success illustrates that traditional four-stage models can be incrementally enhanced, retaining the best elements and replacing only the parts that would benefit from improvement. We call them hybrid models and they provide a path to the future, releasing the embedded value in modelling assets and ensuring service continuity to their owners.
A similar benefit was realised by introducing a middle-level mesoscopic element in the highway supply-side representation of PRISM. This innovation permitted modelling at the most appropriate level of detail where it matters, while also making the interchange of data between the large-scale, aggregate strategic model and the more refined microsimulation easier.
For us, it is all about data! Future model scenario results are ‘synthesised’ data and should be treated no differently from today’s real-life observations from, for example, sensors. It is the reason we are also starting to use Moata, our cloud-based geospatial data analytics and visualisation system, in transport modelling. Our Moata-based digital twins for the Malaysian city of Iskandar will enable visualisation and feed real-time information provision into the local Smart Integrated Mobility Management System.
We can also bring our modelling of human behaviour into our transport demand modelling. Given the impact of COVID-19, perceptions of health risks in the post-pandemic era may mean that many people are unwilling to travel during peak times of congestion and overcrowding. Evidence based findings of such human behaviour can be added to models and adjusted as pandemic alert levels change.