Andrea Serna-Castaño, from the University of Bath, discusses the advantages of using synthetic populations to evaluate the impacts of a changing climate and the interventions put in place to mitigate them.
What if we could test climate policies before implementing them, seeing exactly who benefits and who gets left behind? Synthetic populations are detailed, privacy-safe simulations of real communities that could make this possible, giving cash-strapped councils the evidence they need to protect vulnerable residents from climate impacts.
Climate change is not a distant problem. Heatwaves are hitting city centres harder, floods are threatening market towns, and air pollution is harming vulnerable populations. While councils control over 80% of the UK’s emissions through their decisions about transport, housing, and land use, they are trying to tackle these problems with shrinking budgets, limited data, and too few staff. The difficulty lies in knowing where and who is most affected, and which policies can best protect those at risk.
Virtual models of society
This is where synthetic populations come in. They are simulated datasets at the individual level that look like the real populations. The synthetic data mirrors the statistical characteristics of real communities. They are complete datasets showing age, education, household types, car ownership, and more, but without including any real people’s information.
We build these by combining survey data like Understanding Society with Census information and other public datasets. The result is a privacy-safe virtual model of society that lets us test policy scenarios before they’re implemented.
For example, a council may need to choose between two air quality interventions: expanding green spaces and restricting high-polluting vehicles. Using a synthetic population, we can estimate not just the overall impact of each intervention on air pollution, but crucially, who benefits most and who might be left behind.
While building such models may not be the fastest approach initially, they provide greater precision in understanding the trade-offs between policy alternatives and the populations affected. They also allow councils to integrate medium and long-term perspectives into decision making, supporting more informed and equitable policy choices.
Approaches to policy modelling
Policy models can either look backwards to evaluate what has already happened or look forwards to explore potential outcomes.
Backward-looking models use methods such as causal inference to identify cause-and-effect relationships after interventions have taken place. Forward-looking models ask “what if” questions before policies are implemented.
Microsimulation makes it possible to test alternative policy options and assess long-term outcomes by simulating how individuals might be affected under new conditions. Predictive models, such as machine learning, can complement microsimulation by improving parameter estimates or predicting probabilities. The synthetic population provides the structural foundation – a virtual society on which those probabilities can be applied.
Why this matters now
The links between the environment and health are well established. The World Health Organization estimates that almost a quarter of all deaths globally are linked to preventable environmental risks. But these risks are not shared equally. The wealthiest 10% of people contribute over six times more to global warming than the average person, while the health burden often falls on low-income communities, older adults, and those living with chronic conditions.
Without clear, locally grounded numbers, short-term economic pressures usually win out over longer-term health and environmental benefits. Synthetic populations with microsimulations can change this equation by providing the quantitative evidence councils need to make fairer, more sustainable decisions. An example of this is the work done by SIPHER, which tests policy interventions that affect household income and the impact this has on health outcomes.
What we are doing at Bath
At the University of Bath, our team is developing a high-resolution synthetic population for England as part of the Local Health and Global Profits research consortium. We are working directly with local authorities to evaluate policies and outcomes that matter to them.
Synthetic populations can’t replace real data, but they can open doors that were previously closed by confidentiality barriers. They allow us to ask new questions, test policy alternatives, and place the needs and experiences of different groups at the heart of climate discussions. In the face of limited resources, this can give local authorities the evidence they need to act decisively, equitably, and effectively.
This collaboration shows what happens when research meets real-world decision-making. By combining the UK Data Service‘s open data resources with synthetic methods, we can move from broad generalisations to tailored local insights.
From national data to local insights
Previous synthetic populations have supported models of travel demand, changes in household income, and food expenditure. What we’re doing differently is applying this approach specifically to climate change and health.
By linking local environmental exposures with population characteristics, we can identify health vulnerabilities that might otherwise stay hidden. For instance, we can explore how air quality policies affect people with asthma living in high-deprivation areas, or how urban heat policies protect elderly residents living alone.
Building on existing tools
We are not starting from scratch. Excellent resources already exist: Friends of the Earth’s database of local climate actions, the Climate Just tool identifying disadvantaged areas, and the Local Climate Adaptation Tool linking climate and health data.
These tools show where vulnerabilities exist and which actions are possible. Our synthetic population approach adds something new: estimates the extent to which different policy options might affect health, wellbeing, and the economy in specific local contexts over time.
The challenges ahead
Building and running these models takes time and technical expertise. Synthetic populations require careful design, validation, and significant computing power. Their outputs can be complex to interpret without specialist support.
Another challenge is how to maintain and update a synthetic population beyond the life of a research project. Previous initiatives have demonstrated what is technically possible, but updates often stop once funding ends. A potential solution would be to establish an official, regularly updated synthetic population that is refreshed after each Census and maintained as a shared resource for both academia and government.
The bigger picture
Synthetic populations cannot replace real data, but they can open doors that were previously closed by confidentiality barriers. They let us ask new questions, test policy alternatives, and put people at the centre of the climate conversation.
For councils facing impossible choices with limited resources, this kind of innovation provides the evidence needed to act decisively, equitably, and effectively.
About the author
Andrea Serna-Castaño is an applied environmental and health economist at the University of Bath, working within the Local Health and Global Profits (LHGP) consortium.
Her research focuses on using data and modelling to support evidence-based policy on health, wellbeing, and sustainability, including work on the links between climate change and population health.
Other stories you’ll find interesting
Comment or question about this blog post?
Please email us!