The live hazard map

What you are looking at

The interactive map at the top of this page colours each district by the highest-severity hazard of its current outlook, where the latest model run covers it; districts outside that run keep their static baseline colour, and the district card in the dashboard states which of the two is shown. Selecting a district opens its forecast card: hazard type, severity index, confidence bin, the crops and land cover at risk, and the lead time in days.

Every number on the map comes from the most recent forecast snapshot the platform holds. The snapshot's prediction date is printed next to the map — if that date is more than a few days old, treat the outlook as stale and read the banner that explains why.

How to read the numbers honestly

• Severity index (0.00–1.00) is the model's continuous damage-potential score. It is not a probability and not a percentage of damage.
• Confidence (Certain ≥ 0.85, Probable 0.70–0.85, Uncertain < 0.70) is the model's own certainty in the hazard class it selected. It is not a measure of whether the forecast is correct, and it is not derived from the physics cross-check — the two tracks are computed separately and can disagree, and each district card shows both side by side so the divergence is visible.
• Physics severity is an independent, formula-based estimate (for example, a vegetation-health index for drought or an excess-heat factor for heat waves). Model and physics severity diverging is normal and is shown rather than hidden.
• Lead time is the forecast horizon — 7 or 15 days. Both are driven by the same deterministic weather window (Open-Meteo serves at most about 16 days), so the 15-day outlook is the more uncertain of the two and is a planning aid, not a storm-specific prediction. A longer 10/20/30-day horizon set is described in ADR 0005 but is not implemented in the pipeline, so it is not published.

HazardNet is decision support, not an official warning service. For evacuation, cyclone signals, flood bulletins and emergency instructions, follow the Bangladesh Meteorological Department (BMD), the Flood Forecasting and Warning Centre (FFWC) and your local administration.

Where the data comes from

Sentinel-1 radar and Sentinel-2 optical imagery (ESA Copernicus, accessed through Google Earth Engine), ERA5-Land reanalysis (ECMWF Copernicus), Open-Meteo deterministic forecasts, HDX COD-AB administrative boundaries and OpenStreetMap exposure data. Full provenance, cadence and licences are on the data sources page.