Sherpur — current hazard outlook

What the model this deployment ships currently says about Sherpur, horizon by horizon, with the independent physics cross-check beside it.

Current outlook from the model run

The run this deployment ships (prediction date 2026-09-16) carries one row for Sherpur. Coverage of that run: produced 25 units at the 7-day horizon and 49 units at the 15-day horizon, covering 60 of 64 districts (coverage status: partial). A district can appear at one horizon and not the other.

  • 15-day outlook (target date 2026-10-01): Flash Flood, severity 1.00, confidence Certain (≥ 0.85); physics cross-check 0.72, divergence 0.28.

How to read these numbers

  • Severity index (0.00–1.00) is the model's damage-potential score for the class it selected — not a probability and not a share of damage.
  • Confidence is the model's own certainty in that class (softmax), published as uncalibrated; it is not a measure of whether the forecast is correct.
  • The physics severity is an independent formula-based estimate over the same drivers. Divergence between the two tracks is normal and is shown rather than hidden.
  • Horizons are 7 and 15 days. The 15-day outlook is the less certain of the two and is a planning aid, not a storm-specific prediction.
  • Nothing above the WATCH level can be published automatically while no calibration map exists, and nothing at all can be published while the pipeline stamps no model version — the alert engine records that as `publication_blocked` rather than issuing the alert.

Static baseline shipped with the app

Sherpur is in Mymensingh division. The application's district table carries a static baseline for it: risk Moderate, baseline severity 0.59, baseline hazard Flash Flood, elevation 23 m, main crops Aman Paddy & Vegetables.

The baseline is a static entry shipped with the app — it is what the map colours a district with when the current run does not cover it, and it is not a model output. The outlook section above is the model output, and it exists only for the districts and horizons the current run covers.

Recorded hazard history

This deployment has no event archive loaded, so this page states no historical count for Sherpur. The HazardNet model card quotes 2,931 historical events (2000–2025) for the training prior, and that figure is reported rather than verified in this repository — the loader reports the drift against it instead of asserting it. Until the archive is loaded, this section stays empty rather than filling itself with an approximation.

Ground truth for this district

Ground-truth reports (district, hazard, horizon, date, what was observed) are the only route by which this project can publish skill metrics. The contact page reaches the maintainers.

Questions and answers

What does the severity number for Sherpur mean?

It is the model's continuous score for how strongly the forecast drivers resemble the class it selected, on a 0.00–1.00 scale. It is not a probability, not a percentage of damage, and it does not account for how many people or how much cropland is exposed. The physics cross-check is a separate, formula-based estimate over the same drivers and is shown next to it.

Why does Sherpur sometimes have no outlook?

A run can be partial: the pipeline reports a coverage stamp naming how many districts it produced rows for, and a district outside that run keeps its static baseline colour on the map and shows no model row here. The status page reports the coverage of the run this deployment ships. Nothing is interpolated to fill the gap.

How current is this page?

Every page here is generated at build time from the artifacts the deployment ships. The forecast section states the prediction date it came from (2026-09-16), and the /status page states the age of each source against the SLO it is measured against.

Content reviewed 2026-09-16. HazardNet is decision support, not an official warning service — see the methodology for scope and limitations.

Loading the interactive HazardNet application…