Hyperion
Where the water
is likely to move.
A multi-region urban flood intelligence system for compound flooding, combining rainfall, tides, rivers, terrain, urban structure, and historical events.
Hyperion
in one place.
Hyperion is a multi-region urban flood intelligence system for compound flooding.
It combines rainfall, tides, rivers, terrain, urban structure, historical events, and operational signals to study where water is likely to move, what is driving the event, and where action may be required.
Its first operational phase focuses on Florida, Big Bend, and Louisiana. Recife is planned as a later transfer phase, after the shared technical core is stable.
The reason
it was started.
Urban flooding is rarely caused by one variable.
Rainfall, tide, river stage, elevation, drainage, roads, buildings, and prior ground conditions interact differently across neighborhoods and events.
What stands
in the way.
Cities and operators face fragmented environmental data, rainfall-only warnings, coarse forecasts, false precision, long-term scores that do not answer immediate questions, and models that perform retrospectively without proving operational value.
How the work
is approached.
Hyperion builds a shared multi-region architecture with regional adapters.
Its approach includes common data layers, susceptibility modelling, event modelling, compound nowcasting over approximately 0–6 hours, calibrated scores, visible uncertainty, event-specific validation, operational alerts, and transfer testing across regions with different data quality.
What it can
contribute.
Hyperion gives Genesis contact with a physical world in which several causal systems interact at once.
It can contribute spatial relations, temporal dynamics, multi-cause events, incomplete observations, uncertainty under real consequences, transfer between environments, and evidence that the same schema may require different regional interpretations.
The state
today.
The modelling stack runs as a data and operations pipeline across the first-phase regions. Uncertainty is shown rather than hidden.
Hyperion works over an approximately 0–6 hour nowcasting window. It does not promise deterministic address-level prediction at any horizon.
Canonical repository: cesaremcasa/Hyperion (renamed from Brasil, 2026-07-20). Product shell Home → Access → Internal lives in that repo.
- 01Type — Environmental intelligence
- 02Status — Active research