Synthetic population
The synthetic population describes who travels. It is the foundation the rest of the demand chain is built on, and if it is wrong, everything downstream inherits the error.
Why the population has to be synthesised
A complete individual-level representation of the real population is generally unavailable to researchers and planners, particularly at fine geographical scales. The reason is privacy and data-access restrictions (Garrido et al., 2020; Lomax et al., 2022; Wu et al., 2019).
Nobody can hand a modeller a file listing every real resident of a city with their age, income and employment status, and nobody should.
Instead, synthetic population generation is used to construct a statistically representative population from available aggregate and sample data.
The objective, stated precisely
The objective is not to reproduce identifiable real individuals. It is to generate a population whose statistical and spatial characteristics closely resemble those of the observed population.
This distinction is what makes the method both ethically and legally workable. The synthetic people are not anonymised real people; they are invented people who collectively behave like the real ones.
What each synthetic individual carries
Each synthetic individual or household is assigned characteristics such as:
| Attribute | Example use in demand modelling |
|---|---|
| Age | Trip rates and purpose mix vary sharply by age |
| Sex | Influences travel patterns and mode choice |
| Household composition | Determines joint and escort trips |
| Income | Affects mode choice and car availability |
| Occupation | Shapes work-trip destination and timing |
| Employment status | Determines whether a work trip exists at all |
| Car availability | A primary control on mode choice |
These are derived from sources such as censuses, household surveys, administrative datasets and other mobility datasets (Moeckel et al., 2003; Gearda et al., 2013; Yaméogo et al., 2020).
What makes a synthetic population credible
The extent to which the synthetic population reproduces observed demographic and socioeconomic distributions is important for the credibility of subsequent transport simulations (Garrido et al., 2020; Zhou et al., 2022).
This is a validation obligation, not a formality
Because every later result rests on this layer, a synthetic population that misrepresents, for example, the spatial distribution of car ownership will produce mode shares that look plausible in aggregate while being wrong in exactly the neighbourhoods an equity analysis cares about.
Checking the synthetic population against observed distributions is therefore part of the modelling work, not a preliminary to it.
The tools
Several tools can generate or project a synthetic population. Each takes a different approach, and they are not mutually exclusive: a workflow may use one to establish a base population and another to project it forward.
| Tool | Approach | Documented on |
|---|---|---|
| NTEM | Trip-end forecasting for Great Britain, from DfT | NTEM |
| SILO | Land-use and demographic microsimulation over time | SILO |
| SPENSER | Spatial and dynamic microsimulation under scenarios | SPENSER |
| Python codes | Not yet specified | Other tools |
| Eqasim | Not yet specified | Other tools |
What happens next
A synthetic population on its own describes people, not travel. It becomes demand only once each individual is given a daily schedule.
That step is activity plans, and combining the two produces the synthetic travel demand that the transport model simulates.