Project 01 · Climate & Aviation AI

STRATUM

Forecasts the aircraft contrails that warm the climate over India and gives live altitude advice to avoid them. Then it proves the forecasts mean something: every prediction is checked against what the Meteosat-9 satellite actually saw.

Scroll to watch how it works
Featured · STRATUM

Scene 1 of 5

Planes leave trails.

Most vanish in seconds. In cold, humid air, some last for hours.

Scene 2 of 5

Lasting contrails warm the planet.

At night they trap heat. Together they can warm the climate as much as aviation's CO₂.

Scene 3 of 5

STRATUM forecasts where they'll form.

Weather data and contrail physics flag the humid layers before planes reach them.

Scene 4 of 5

Then it checks against the satellite.

A learned detector scans Meteosat-9 images, and each contrail is matched to the flight that made it.

Scene 5 of 5

And finds the altitude that avoids them.

About 2,000 ft lower or higher, live, for flights over India.

Open STRATUM live
FORECAST: HUMID LAYER, LASTING CONTRAILSMETEOSAT-9Detected · matched to its flight−2,000 ft: contrail avoided

The details

Under the hood.

STRATUM forecasts the aircraft contrails that warm the climate over India and gives live altitude advice to avoid them. Then it proves the forecasts mean something: every prediction is checked against what the Meteosat-9 satellite actually saw, with pre-registered hypotheses and a chance-match test that most contrail work skips. Everything runs live in your browser, on free, open data. Team of four from Indus International School, Bengaluru: Shrugal Vinod (team lead), Aditya Harsoda, Ishaan Kupnur, Abeer Shah. Supervisor: Dr. Varsha Pant.

  • Satellite-checked: Meteosat-9 contrails matched to the flights that made them
  • Learned detector at a pixel Dice of 0.64 (a single expert reaches about 0.72)
  • Conrad Challenge 2026–27: Innovation Stage, team lead
Contrail Forecast Maps
Satellite Contrail Detector
Flight-to-Contrail Matching
Chance-Match Audit
Live Altitude Advice
3D Sky Replay

Built with

PythonPyTorchpycontrailsTypeScriptPreactThree.jsMapLibreCloudflare WorkersGitHub Actions

Next project

MatchMind AI

Football Intelligence