A forecast that says "150 mm of rain expected" means little to most people. A forecast that says "low-lying houses along the river are likely to flood tonight; move valuables upstairs" gets action.
That's the shift impact-based forecasting makes: from what the weather will be to what the weather will do.
The idea
The World Meteorological Organization has promoted impact-based forecasting and warning services for years. The model combines:
- Hazard forecast: how likely and how severe
- Exposure: who and what is in the way
- Vulnerability: how badly they'll be affected
That's the same structure as climate risk modelling, applied in real time.
The risk matrix
Many services use a matrix of likelihood × impact:
| Minimal impact | Minor | Significant | Severe | |
|---|---|---|---|---|
| High likelihood | Yellow | Amber | Red | Red |
| Medium | Green | Yellow | Amber | Red |
| Low | Green | Green | Yellow | Amber |
A low-likelihood but severe event can warrant a stronger warning than a likely but minor one. The UK Met Office's warning system follows this kind of approach.
Anticipatory action
Impact-based forecasts unlock anticipatory action: agreeing in advance what to do when a forecast crosses a threshold, and pre-arranging the money.
Example structure:
- Trigger: "Forecast probability of severe flooding > X% at Y days lead time."
- Action: distribute cash to pre-registered households, move livestock, reinforce shelters.
- Funding: pre-approved, released automatically when the trigger hits.
The Red Cross Red Crescent's forecast-based financing and various UN anticipatory action pilots work this way.
What this means for engineers
The systems behind impact-based forecasting need:
- Hazard data pipelines: forecasts ingested and processed automatically (flood forecasting).
- Exposure databases: buildings, roads, population, assets, kept current.
- Vulnerability information: housing type, livelihoods, past impacts.
- Trigger logic: transparent, versioned, auditable rules.
- Messaging: templates per audience and language, sent across SMS, radio, and apps.
- Feedback: logging what happened after each event to refine thresholds.
Getting the messages right
- Specific places and actions: "Avoid the bridge at X after 6pm."
- Plain language, local language.
- Consistent colours and terms across agencies.
- Clear end of warning: people need to know when it's safe.
Challenges
- Vulnerability data is often missing or outdated.
- Trigger design balances false alarms against missed events, and money is wasted either way.
- Institutional coordination: met services, disaster agencies, and responders must agree.