Validating Impact-Based Forecasts for Anticipatory Action

Anticipatory actions (AA) are pre-emptive measures taken to reduce disaster impacts based on forecasted hazards. Trigger models using impact-based forecasts (IBF) play a central role in deciding when these measures are activated, by combining hazard forecasts with information on exposure, vulnerability, and past impacts. The resulting impact predictions can cover a range of outputs such as number of people or infrastructures affected, crop losses, or area that is expected to be inundated. Loss predictions are typically given as percentages, such as of buildings destroyed within a given administrative unit, with some impact assessment models also considering crop phenology or specific building vulnerability information. This diversity in approaches, in itself linked to the often limited data availability for the loss estimation, creates further challenges for validation.

While the work on designing these trigger models is growing, systematic methods to validate their impact forecasts remain relatively under-explored. There are open questions on how well forecasted impacts align with observed damage, how performance of IBF varies across events, and how repeated events or different damaging mechanisms (e.g., wind, storm surge, landslide) influence the relationship between hazard and impact. Through this topic, these questions can be explored using one or more existing AA trigger models for floods or tropical cyclones.