High-resolution demographic projections using SPENSER
SPENSER (Synthetic Population Estimation and Scenario Projection Model) is a set of open-source tools for population estimation and projection, funded by the Alan Turing institute. It is a partnership between academic and private sector researchers that introduces a novel data-driven modelling framework to run complex simulation models.
Like a digital twin, SPENSER replicates the structure and behaviour of the real population in terms of demographic, socioeconomic and health characteristics, along with detailed time use data to provide high-resolution geographical and sub-population projections essential for the planning and delivery of services and urban infrastructure developments. SPENSER reproduces data on households and the constituent population across the whole of Great Britain.
“The driving force behind this project is to produce the high-resolution, special scale, geographical and detailed sub-population projections that everyone needs, for resource planning, land use planning, making sure the bins are collected and police are in the right place” explains Dr. Nik Lomax, lead researcher and Alan Turing Fellow. “However, what we currently have access to, generally from official sources at local authority level, is based on high-level scenarios, which doesn’t allow you to tweak a lot of parameters.”
What the local authority data also lacks is a feedback loop, being a population-led set of projections that do not take into account what happens if such things as employment, housing, or the economic outlook suddenly change for a small area. “With SPENSER, we are trying to build models accurately representing human growth and human behaviour, so we can produce useful data to scale,” says Dr. Lomax. By generating an individual level, synthetic population that can be simulated through time, SPENSER makes it possible to produce projections under different, user-defined scenarios.
One of the most interesting outputs of the past year has involved using data from SPENSER in a Covid-19 disease model, in a case study demonstration for Devon. Work to understand the complex links between individuals’ daily activities and Covid-19 transmission in the UK has demonstrated the applicability of SPENSER which provided the high-resolution demographic dataset. This work, building on the existing SEIR model of infectious disease transmission, brought epidemiological modelling, urban analytics, special analysis and data integration together. The project modelled disease transmission within communities and provided the ability to assess different scenarios, including what would have happened if lockdown had begun a week earlier.
Individual ‘agents’ estimated by the model were fed into the Covid SEIR model, which assigned behaviours and exposures to Covid-19. It then worked out the probability of contracting the disease. Data from third-party sources was used, with shops, schools, and hospitals included as destination locations. SPENSER’s flexibility – you can change inputs into it and look at the effects of different interventions – meant the team could show, with the Devon case study, that an earlier lockdown might have resulted in a lower peak in daily infections, almost halved (47% fewer infections overall).
To understand the effectiveness of government policies, detailed data reflecting the everyday lives of the British population is required. This Covid-19 example presents a multitude of opportunities for future scenario development, including exploring the effects of alternative lockdown scenarios across different sub-populations, and the ramifications of the vaccination rollout. SPENSER’s high-resolution demographic forecasting makes a range of application areas possible, and accessible, from physical infrastructure planning to health and social care spending. Changing the parameters means researchers and planners can run ‘what if’ scenarios and inform planning decisions with an evidence base.
SPENSER’s dynamic micro simulations can open up high-resolution data for local authorities, national governments, emergency services, and land use planners interested in railways and roads, all of whom are then able to make better decisions having assessed probable impacts. The Covid-19 example is a tangible piece of evidence, but SPENSER has also fed a transport model developed by UCL; and a land use model developed in Newcastle, looking at environmental and land use impacts of changing populations at small area scale. “It’s useful for any project needing high-resolution data as an input,” says Dr. Lomax. SPENSER is also open-source, and updated regularly.
More experimental work is also being done with SPENSER, around building dynamic models. “I’m interested in how we better project the population,” says Dr. Lomax. “The data from this project, rather than the methods is important in many applications– the data are intended as input to other models which provide insight into a wide range of systems. If you are going to build a digital twin of anything, for example of an urban environment, you need a digital representation of the humans who are going to engage and interact with that space, and SPENSER provides the human input to the digital twin.”
Dr. Lomax explains that, in considering all the infrastructure projects that either do or don’t go ahead, better decisions can be made if you possess good population demand estimates. “Take, for example, the Northern Powerhouse Rail project. SPENSER can enable decisions to be made, based on extrapolated data, on where the trains should stop, what capacity will be needed on the line.”
Dr. Nik Lomax