Fire: how HazardNet classifies and scores it
Vegetation and burn risk from heat, wind and drying.
How the model arrives at this class
Classified from the same drivers. There is no active-fire input and no burnt-area input, so this class is a dryness-and-wind index wearing a hazard name, and the page says so.
Uncalibrated softmax; the confidence is the classifier's own certainty across the eight classes.
Advisory levels follow the published alert policy and the escalation boundaries of the responsible authorities; nothing here overrides them.
The independent physics cross-check
The three terms are heat, wind and drying: the fuel-moisture chain, not fire occurrence.
Drivers: maximum temperature (T_max, °C) · maximum wind (W, km/h) · evapotranspiration over the horizon (ET, mm).
- Expression: 0.4*clip((T_max-25)/15) + 0.3*clip((W-5)/20) + 0.3*clip(ET/6)
- That expression is executed against the pipeline’s own implementation by the content tests in CI: this page cannot silently describe a formula the pipeline no longer runs.
What the current run says about Fire
In the run this deployment ships (prediction date 2026-11-02), 2 district trajectories are classified as Fire across the continuous 15-day window (Days 1 to 15).
Highest model severity: Narail (0.30, Days 1 to 15), Sunamganj (0.30, Days 1 to 15).
These are classifications from the current run, not a forecast of impact. Advisory levels follow the published escalation gate alongside official Bangladesh authorities.
Season and geography
Typical season: March–May, before the monsoon; also the December–January dry spell. This describes when the hazard is climatologically plausible, not when this run flags it.
What this class cannot tell you
- No fuel load, no land cover, no ignition source, no thermal-anomaly feed (MODIS/VIIRS active fire is not ingested).
- Rice-straw burning and brick-kiln seasonality are invisible to a district-aggregated weather index.
- The class is retained because the model's vocabulary has eight classes and this is one of them, not because the pipeline measures fire.
- Confidence is published as an uncalibrated model score; no calibration accuracy is claimed.
Inputs behind it
Published assessments and official bulletins are the record; this site reports what they name.
- published forecast records
Questions and answers
Is a high Fire severity a prediction of damage?
No. The severity index is the model's continuous score for how strongly the drivers resemble this class, on a 0.00–1.00 scale. It is not a probability, not a percentage, and it does not model exposure. The physics cross-check is a second, independent estimate; where the two diverge, both numbers are shown rather than averaged.
Why is the confidence not a probability of Fire?
The published confidence is the model's own confidence over its eight classes, labelled uncalibrated_model_confidence in the data. It says how certain the classifier is about the class it picked, not how often that class actually materialises. No calibration has been fitted yet, which is also why nothing above the WATCH level can be published automatically.
Which districts does this page cover?
The district pages cover all 64 districts. This page summarises the districts the current run covers: a run can be partial, and the coverage stamp on the snapshot says exactly how partial, which the status page reports.
HazardNet is not an official warning service. Official warnings come from BMD and FFWC. In an emergency call 999.
Loading the interactive HazardNet application…