The radiation model in geography is an advanced spatial interaction model for human mobility, commuting, migration, transport geography and origin–destination flows. It asks whether movement between places can be predicted from population and opportunities around origins and destinations—without beginning by fitting the familiar distance-decay parameter used by gravity models.
In human geography, spatial interaction modelling usually begins with a tension between attraction and friction: larger destinations attract more movement, while distance, travel time and cost suppress it. The radiation model approaches the same mobility problem from another direction. Instead of saying that travellers respond directly to a calibrated distance penalty, it imagines people searching through opportunities and accepting destinations according to how attractive those opportunities are relative to the alternatives encountered before them.
This makes the radiation model relevant to geography, commuting patterns, migration models, mobility modelling, transport planning, urban systems, accessibility, intervening opportunities, gravity models, population distribution and complex spatial networks. It also gives us a useful theoretical test. If much of observed distance decay can emerge because more opportunities are encountered as the search radius expands, then some apparently distance-driven behaviour may actually be opportunity-driven behaviour expressed through space.
Gravity asks how strongly distance suppresses interaction. Radiation asks whether the geography of opportunities can generate the suppression without putting distance decay into the model by hand.
Quick Read: The Radiation Mechanism
ORIGIN POPULATION → SEARCH FOR OPPORTUNITY → INTERVENING OPPORTUNITIES → DESTINATION OPPORTUNITIES → COMPETITION AMONG CHOICES → EXPECTED ORIGIN–DESTINATION FLOW
The radiation model belongs to the same broad family of questions as Gravity Models and Intervening Opportunities, but it owns a narrower and more advanced reader job: understanding a parameter-light opportunity-selection model in which expected flows arise from the relative opportunity masses of origins, destinations and the surrounding search field.
1. Why Geography Needed Another Mobility Model
Gravity models are extraordinarily useful. They compress a great deal of geographic behaviour into a simple proposition: interaction tends to increase with the importance of origins and destinations and decrease with separation. The problem is not that this is wrong. The problem is that the friction function usually has to be specified and calibrated.
How quickly should interaction decline with kilometres? With minutes? With monetary cost? Should the decay be exponential, a power function, or something else? Does one fitted parameter transfer from one city to another? Does a commuting coefficient estimated before a new rail line remain valid after the network changes?
The radiation model became interesting because it offered a different theoretical route. Perhaps we can predict a substantial part of mobility by modelling competition among opportunities rather than fitting a distance-deterrence curve directly.
2. The Core Intuition: Search, Compare, Accept
Imagine a worker at origin i looking for a job. Jobs differ in attractiveness. The worker does not necessarily care about kilometres as an abstract quantity. The worker cares whether a sufficiently attractive opportunity is available.
Nearby opportunities are encountered before farther ones under a spatial search. If one of them is better than the worker’s current opportunity threshold, the worker can accept it and stop searching. A farther destination receives the worker only when its opportunities outperform the relevant alternatives encountered closer to the origin.
This makes mobility an order-statistics problem over opportunities distributed in space. Distance matters because it orders the search field and determines how many opportunities intervene, not necessarily because a traveller carries a universal psychological dislike of kilometres.
3. The Canonical Variables
In the classic radiation formulation, three quantities are central:
- mᵢ: the population or opportunity mass associated with origin i.
- nⱼ: the population or opportunity mass associated with destination j.
- sᵢⱼ: the population or opportunity mass inside the circle centred on i whose radius reaches j, excluding the origin and destination masses themselves under the standard definition.
These quantities allow the model to represent a destination not only by what it contains, but by what lies between it and the origin. A destination surrounded by many intervening opportunities faces stronger competition than an otherwise similar destination reached after passing through few alternatives.
4. The Familiar Radiation Expression
A common form of the radiation model expresses the expected flow from i to j as a fraction of the total trips originating at i:
Tᵢⱼ = Tᵢ × [mᵢ nⱼ / ((mᵢ + sᵢⱼ)(mᵢ + nⱼ + sᵢⱼ))]
The notation varies across presentations and later model variants. The important conceptual structure is more valuable for most readers than memorising symbols. Destination opportunity nⱼ increases potential flow. Intervening opportunity mass sᵢⱼ suppresses it. Origin mass enters both the production of travellers and the comparison structure.
The model is often described as parameter-free in its original idealised form because it does not require the empirically fitted distance-decay exponent characteristic of many gravity models. That does not mean it is assumption-free, data-free or universally calibration-free in practical applications.
5. Parameter-Free Does Not Mean Free of Assumptions
This distinction is essential. A formula can contain no fitted distance coefficient and still depend on strong assumptions about how opportunities are distributed, how travellers select them, how population represents opportunity, how the search field is ordered and how total origin trips are specified.
“Parameter-free” is therefore a statement about one part of the mathematical architecture, not a declaration that the model contains no modelling choices.
6. Why Distance Decay Can Emerge Without a Distance Parameter
As we move farther from an origin, the area enclosed by the search radius usually increases. More people, jobs, services or other opportunities fall inside that area. The farther destination must therefore compete with a larger pool of intervening alternatives.
Even if the traveller never applies a direct distance penalty, the probability of reaching a distant destination can decline because the probability of finding an acceptable nearer opportunity rises as the search field expands.
This is the theoretical beauty of the model. A familiar macroscopic pattern—fewer long-distance trips—can emerge from microscopic opportunity competition rather than being inserted as a friction curve at the outset.
7. Radiation Is Not Gravity
Gravity Models own the formal size–separation framework in which interaction is usually a function of origin mass, destination mass and a fitted or specified friction of distance, time or cost.
The radiation model owns a different mechanism. It predicts interaction from opportunity masses and intervening opportunity structure. Physical distance matters through the ordering and accumulation of opportunities, not necessarily through an explicit deterrence exponent.
8. Radiation Is Not the General Intervening-Opportunities Article
Intervening Opportunities owns the broad behavioural and geographic idea that nearer acceptable alternatives can absorb movement before a farther destination is reached.
The radiation model is one formal mathematical implementation of opportunity competition. It does not own every intervening-opportunity model, nor should every satisficing or search process be called radiation.
9. Radiation Is Not Flow Mapping
Flow Mapping owns the visual representation of origin–destination movement. Radiation models estimate the flows that might later be drawn as lines, arrows or matrices.
10. Radiation Is Not Accessibility
Accessibility owns the broader question of which opportunities can realistically be reached. Radiation models predict how interaction may distribute among opportunities. Accessibility can be an input to mobility behaviour without being identical to the mobility model.
11. Population Is a Proxy, Not the Opportunity Itself
The original model often uses population as a proxy for the number of opportunities. This can be reasonable at some scales because more populous places often contain more jobs, services and social contacts. But the proxy can fail.
A residential town can contain many people and relatively few jobs. A business district can contain enormous employment opportunity and relatively few residents. A specialist hospital can contain an opportunity unavailable anywhere else despite a small surrounding population.
Advanced applications therefore ask whether population, employment, vacancies, service capacity, floor area, school places or another variable better represents the opportunity field relevant to the trip purpose.
12. Trip Purpose Changes the Opportunity Field
The same city contains several overlapping geographies of opportunity. A commuter sees jobs. A patient sees appropriate healthcare. A shopper sees products. A student sees schools or study spaces. A freight operator sees customers, warehouses and ports.
One population surface cannot represent all these opportunity systems equally well. A radiation model should therefore be interpreted relative to a defined movement purpose.
13. The Circle Is a Model Choice
The classic formulation counts intervening opportunities inside a circle around the origin with radius equal to the origin–destination distance. That is elegant in unconstrained planar space. Real cities are networks.
A river, expressway, border or rail interchange can make a geometrically close opportunity behaviourally distant. Two places equally far in kilometres can differ radically in travel time. The search field may therefore be better defined by network distance, generalised cost or travel time than by Euclidean radius.
This connects directly to Network Distance and Cost Distance.
14. Search Space Can Be Anisotropic
People do not search equally in every direction. Rail corridors, highways, coastlines, mountain valleys and administrative borders channel movement. A circular search field assumes isotropy: that direction does not matter. That can be badly wrong.
A more realistic opportunity field may stretch along fast transport corridors and contract across barriers. The model’s conceptual logic survives, but the geometry used to count intervening opportunities needs adaptation.
15. A Singapore Thought Experiment: Punggol to Employment
Imagine a worker living in Punggol. Several employment clusters lie across Singapore. A simple geometric model orders them by straight-line distance. A transport-aware model orders them by door-to-door travel time. A job-aware model then counts suitable vacancies encountered within those increasing travel-time contours.
A nearby employment cluster with many suitable vacancies can absorb a large share of demand before the worker considers a farther central location. But if nearby jobs do not match the worker’s occupation, the nominal intervening opportunity mass overstates real competition.
This demonstrates why advanced mobility modelling must distinguish population mass, employment mass and relevant opportunity mass.
16. A Singapore Thought Experiment: Healthcare
For routine care, nearby clinics can act as strong intervening opportunities. For a rare specialist procedure, most nearby facilities are not relevant opportunities at all. A specialist centre farther away may receive flows that a population-based radiation model underpredicts.
The lesson is not that radiation modelling fails. It is that opportunity must be defined at the level of the need.
17. A Singapore Thought Experiment: Schools
School choice demonstrates another complication: eligibility and institutional rules. A geographically nearby school can be irrelevant if the student is not eligible for the programme or if places are unavailable. An opportunity surface should therefore represent accessible capacity, not merely buildings on a map.
18. A Singapore Thought Experiment: Retail
For ordinary groceries, many nearby stores can intervene strongly, producing short trips. For a rare luxury product, the opportunity field becomes sparse. People may travel much farther because nearer retail locations do not contain the required opportunity.
The same person therefore exhibits different apparent distance decay for different trip purposes because the opportunity landscape changes.