TIMSBHOPAL-01

Thermal Intelligence & Monitoring System

BHOPAL-01 · Madhya Pradesh, India

Thermal Intelligence &
Monitoring System

Satellites see every fire over Bhopal, but cannot tell a brick kiln from burning crop residue. TIMS fuses NASA FIRMS thermal anomalies with OpenStreetMap industrial geometry and PostGIS spatial analysis to detect anomalies, assess their risk, and identify what is exposed nearby.

375 m
VIIRS pixel resolution
7217
Boundary polygon vertices
4326
WGS 84 spatial reference
1/3/5 km
Impact zone radii

Pilot envelope 77.3,23.1,77.55,23.35
Marker position derived from the configured bounding box

Processing chain

Detect → Understand → Assess Risk → Identify Impact → Support Response. Each stage is implemented and observable; where a capability is not wired up, the system says so rather than showing a plausible placeholder.

  1. 01

    Detect

    NASA FIRMS VIIRS (375 m) and MODIS (1 km) thermal anomalies are pulled on a 30-minute schedule and de-duplicated by acquisition identity.

  2. 02

    Geofence

    A detection counts as a Bhopal incident only when PostGIS ST_Within places it inside the real 7217-vertex district polygon — not merely inside the request bounding box.

  3. 03

    Understand

    Distance to the nearest industrial site and a recurrence count over a 1 km radius are computed in the database on the WGS84 spheroid, never approximated in JavaScript.

  4. 04

    Classify

    An XGBoost model returns a hedged class and a confidence. When it is unreachable a transparent rule engine takes over, and the record says which path ran.

  5. 05

    Assess

    A weighted 0–100 triage score with its drivers listed in full: radiative power, confidence, persistence, industrial proximity, populated proximity.

  6. 06

    Support

    Concentric 1/3/5 km exposure counts, nearest hospital and fire station, and a live alert pushed over Socket.io for anything HIGH or above.

What this system does not claim

A monitoring instrument is only useful if its limits are as legible as its readings.

  • It does not predict fires

    The risk score is a transparent weighted triage figure whose drivers are listed on every incident. It is not a validated probability.

  • It does not confirm causes

    Every class is hedged — "possible industrial fire". Co-location with a mapped factory is suggestive, not probative, and there is no ground truth to check against.

  • It does not estimate affected population

    No licensed gridded population layer is loaded, so the interface reports "population estimate unavailable" instead of producing a number.

  • It does not dispatch

    Acknowledging an alert records an analyst’s name for audit. It notifies no external agency and is not an official emergency alert.