Open data feeds
The useful question is not what data exists. It is which data changes an answer.
Plenty of open feeds could be bolted to a globe and would look impressive. Very few of them would move a footprint. This page ranks candidates by whether they touch the physics that actually determines where something lands — and the top of that list is not imagery or news, it is the atmosphere the model flies through.
Implemented now
Live atmosphere
—Feed status
—Levels
—Peak departure
—Top of profile
Requesting current pressure-level data…
Live comparison
Today's atmosphere against the 1976 standard
Density from pressure-level temperature and geopotential height through the ideal gas law, compared against USSA-76 at the same altitude. Drag scales linearly with density, so a departure of a few percent is a few percent on the drag term throughout the densest part of the flight.
Candidate feeds
Ranked by whether they move a footprint
Every feed listed is open access and keyless or near-keyless. The limitation column is the part usually left off a capability slide.
| Feed | Access | What it provides | What it would buy | Limitation | Value |
|---|---|---|---|---|---|
| NOAA GFS / ECMWF pressure levels via Open-Meteo |
Live No key, CC BY 4.0 |
Temperature and geopotential height on 19 pressure levels, surface to ~24 km, refreshed hourly. | Replaces a 1976 long-term average with today's actual density where almost all the drag happens. Density enters drag linearly, so a few percent departure moves time of flight and footprint size directly. Implemented — see the live comparison below. | Stops around 24 km. A glide vehicle spends most of its flight above that, so the profile blends back to USSA-76 at the top. | high |
| NOAA SWPC space weather services.swpc.noaa.gov/json |
Live No key |
F10.7 solar radio flux and Ap geomagnetic index, including a 45-day forecast. | These two numbers drive thermospheric density in NRLMSISE-00. Above 86 km this build currently uses a crude exponential extension that ignores solar activity entirely — the weakest part of the atmosphere model, in the regime where a glide vehicle enters. | Only useful paired with an NRLMSISE-00 implementation, which is a meaningful port rather than a fetch. | high |
| NASA / USGS elevation (SRTM, 3DEP, Copernicus DEM) opentopodata.org, USGS |
Live Mostly keyless |
Terrain elevation at the impact point. | Trajectories here terminate at sea level. Real impact happens at terrain altitude, which changes time of flight and shifts the footprint. On a 2 km plateau the vehicle arrives measurably earlier than modelled. | Sampling a footprint's worth of terrain is many point queries; a tiled DEM subset is the sane approach. | high |
| Winds aloft via the same pressure-level feed |
Live No key |
Wind components at each pressure level. | Currently unmodelled. Wind is a first-order cross-range error source in the terminal phase and is the most common reason a well-calibrated ballistic prediction still misses. | Requires adding a wind term to the equations of motion — a real change to the propagator, not a data swap. | high |
| EGM2008 gravity model NGA, public coefficients |
Static Public |
Earth's actual gravity field including oblateness. | The model uses spherical inverse-square gravity. The J2 oblateness term alone produces a systematic, direction-dependent error over long trajectories. | Low-order coefficients are cheap; the full model is large and unnecessary here. | medium |
| CelesTrak / Space-Track element sets celestrak.org |
Live Keyless (CelesTrak) |
Orbital elements for tracked objects. | Not for the warning problem, but the obvious bridge to space-domain surfaces — cueing, sensor-tasking geometry, and reentry prediction for decaying objects, which is the same propagator with a different entry state. | Different problem class. Worth building beside this, not inside it. | medium |
| NASA EONET / FIRMS eonet.gsfc.nasa.gov, firms.modaps.eosdis.nasa.gov |
Live Keyless / free key |
Natural event boundaries and thermal anomaly detections. | Not physics. Useful as a demonstrated-detection layer for the same globe, and as a genuinely live data source to sit beside modelled ones so the provenance chips have something real to distinguish themselves from. | Adds credibility to the provenance argument more than to the model. | low |
| ERA5 reanalysis Copernicus / Open-Meteo archive |
Historical No key via Open-Meteo |
Hourly reconstructed global atmospheric conditions from 1940. | The single most valuable item on this list for the readiness argument. It would let the calibration corpus be re-run against real historical atmospheres rather than a synthetic one — moving the calibration harness from a synthetic corpus toward a real environment. | Still not truth data about a vehicle. It closes the atmosphere half of validation, not the vehicle half. | high |
| NASA POWER power.larc.nasa.gov |
Live Keyless |
Long-term meteorological and solar climatology. | Useful for seasonal and regional atmosphere climatology — how much a given theatre departs from standard across a year, which bounds how much error the standard atmosphere assumption is worth. | Climatology, not conditions. Bounds the error rather than removing it. | medium |
| IGRA radiosonde archive NOAA NCEI |
Historical Keyless |
Actual balloon-measured temperature, humidity and wind profiles. | Measured rather than modelled. The right way to validate that the density profile derived from pressure-level data is itself correct — a verification step for the feed, not just for the solver. | Sparse in space and time, and rarely where you want it. | medium |
The ordering is the point. Imagery and news feeds are the easiest to add and the least useful here — they make a globe look alive without touching the equations. Atmosphere, winds, terrain and gravity are unglamorous and each one directly changes where the footprint lands. A capability that adds the first group and not the second is decorating, not improving.
And none of them close validation. Real atmosphere, real terrain and real winds would make the environment real. They say nothing about whether the vehicle model is right, because no open feed carries measured trajectories of the vehicles in question. That gap stays exactly where it was.