A small share of flights makes most contrail warming. A 2,000 ft change often avoids it.
Contrails that persist in ice-supersaturated air spread into heat-trapping cirrus. This map forecasts where that air sits tonight, which flights will cross it, what an altitude change would cost in fuel, and how rocket launches and satellite re-entries compare on the same scale.
About 2% of flights produced 80% of contrail energy forcing over Japanese airspace (Teoh et al., 2020, Environ. Sci. Technol.).
Flights, launches and re-entries on a single climate scale
Every event is converted to tonnes of CO₂-equivalent over 100 years. Contrail energy forcing from pycontrails is divided by the absolute global warming potential of a tonne of CO₂. Rocket particles get their own term with wide uncertainty. The bars show the central estimate and the plausible range, on a log scale.
The launch and re-entry rows here are illustrative: no published factor yet converts rocket soot, alumina or re-entry metal into CO₂-equivalent. The space ledger below shows what real launches and re-entries actually left in the upper atmosphere, in kilograms, from public catalogues.
Hover or focus a row for its assumptions.
Who made tonight's contrails
Contrail warming here is the contrail model's prediction for each flight, totalled by carrier or aircraft type. Satellite detections of contrails are matched to the aircraft whose wind-advected track lines up with them; "Confirmed" counts the flights with at least one match, so you can see how much of the predicted warming has been observed. "Avoidable" is the warming removed by each flight's best altitude option, after subtracting the extra fuel CO₂.
Telling a contrail from natural cirrus by its shape and its aircraft
Simulated satellite patches in ash false-colour, where thin ice cloud shows dark blue. Each tile is measured from its own pixels: the largest region of thin ice cloud, how straight it is (from the principal axes of its pixels), its width, and how far the nearest advected flight track lies from it. A two-rule classifier then decides: straight means contrail, and a track within 10 km attributes it. There is no trained model yet; the planned one is a segmentation network trained on Google's labelled OpenContrails set (GOES-16). Streaky cirrus and contrails crossing cirrus are the hard cases, and the rule misses some of them.
What launches and re-entering spacecraft leave in the upper atmosphere, and who put it there
Loading space data…
Launches, re-entries and masses come from public catalogues: Launch Library 2, CelesTrak and Jonathan McDowell’s GCAT. What they leave in the air is worked out from published rocket data and emission studies. Every number carries a tag saying how sure it is. Hover over or tab to a tag to see its range and source.
Aircraft stay low. Rockets and re-entries reach layers nothing else does
Each row is a 5-km slice of atmosphere, from the ground to 120 km. Aviation burns vastly more fuel, but almost all of it below 15 km, where rain washes particles out within weeks. Rockets put soot, alumina and chlorine straight into the stratosphere and above, and re-entering spacecraft vaporise in the mesosphere. Pick one substance at a time: they can’t be added together.
What is up there now, and what is about to come down
A sample of the active catalogue, the space stations, and every object on CelesTrak’s list of orbits that are decaying. Triangles mark the pads that launched in the window. Drag the globe, or focus it and use the arrow keys, to turn it.
Loading orbits…
Who launched, and whose hardware came down
Launch providers come from Launch Library 2; owners of re-entering objects come from GCAT. They are separate catalogues, so the two tables are kept apart: SpaceX the launch company and SpaceX’s Starlink division are different rows. Click a column heading to sort.
Built to burn up: safer on the ground, more metal in the sky
Rules now make satellites in low orbit come down quickly. In 2022 the US Federal Communications Commission required the satellites it licenses to leave low orbit within five years of finishing their mission, down from a 25-year guideline. Satellites are also designed to burn up completely on the way down, so no debris reaches the ground.
That is safer for people below, but it puts the most metal into the upper atmosphere. In 2023 NOAA research aircraft found aluminium and other metals from re-entering spacecraft in about 10% of the sulphuric acid particles they sampled in the stratosphere. What that metal does to the ozone layer and the climate is still being studied.
FCC Second Report and Order on orbital debris, IB Docket 18-313, 29 September 2022 (coverage). Murphy et al. (2023), Metals from spacecraft reentry in stratospheric aerosol particles, PNAS 120, e2313374120.
Sources for the space ledger
Out of scope: predicting where re-entries land, modelling ozone or climate effects, and converting space emissions to CO₂-equivalents. The launch and re-entry figures in the contrail map and the one-score chart above are still illustrative placeholders.
Eight public sources
| Source | What it gives us | Feeds |
|---|---|---|
| OpenSky Network | Aircraft positions, altitudes, ICAO24 codes, and an aircraft database mapping each code to type and operator | Flight tracks; carrier and aircraft type for the ledger |
| Contrails.org API | Contrail line segments detected in GOES-East satellite images every 10 minutes | Matching observed contrails to flights, which confirms or contradicts the model's predictions |
| Google Research contrail datasets | Labelled satellite imagery (OpenContrails, GOES-16 ABI) | Evaluating the shape rule now; training a segmentation model later (not trained yet) |
| Copernicus ERA5 | Temperature, humidity and wind on pressure levels at cruise altitude | Ice-supersaturation layer, contrail drift |
| pycontrails | CoCiP contrail model: formation, lifetime, energy forcing | Warming per flight, reroute options, the shared score |
| CelesTrak | Satellite catalogue, decay dates and current orbits | Space ledger: re-entries, the globe, objects about to come down |
| GCAT (J. McDowell) | Masses and owners of catalogued objects (CC-BY) | Space ledger: re-entered mass by owner |
| Launch Library 2 | Launches, vehicles and liftoff masses | Space ledger: propellant and emissions by provider |
- Everything on this page is simulated. The humidity field, flights, carriers, satellite detections and space events are generated to show how the tool behaves.
- Contrail formation assumes the Schmidt–Appleman criterion is met at every level above FL300, which is usually but not always true.
- Energy forcing per km is a rough stand-in for CoCiP output: strong at night, weaker or net-cooling in midday sun, scaled by humidity and aircraft type.
- Flights are real; their contrails are modeled. Tracks and operators come from OpenSky, humidity and wind from ERA5, and contrail formation, lifetime and energy forcing from pycontrails CoCiP.
- AGWPCO₂ over 100 years is taken as 1.48 × 10¹² J per tonne. Contrail efficacy is not applied.
- Rocket and re-entry values are illustrative placeholders. Black carbon and alumina in the stratosphere are not yet well quantified, which is why their ranges span an order of magnitude.
- Reroutes change altitude only across the humid stretch plus 100 km either side, using the semicircular flight-level rule.