Data coverage & limitations
A trustworthy evidence engine should make the blind spots visible before the user sees a score.
| Source | Measures | Coverage | Time | Useful for | Important limit |
|---|---|---|---|---|---|
| Open-Meteo geocoding | Place search | Global | Current | Place → coordinates/timezone | Place names can be ambiguous |
| Open-Meteo weather | Cloud/visibility/precipitation | Global | Recent + historical model/reanalysis | Selected hour atmosphere | Model estimate, not a rooftop sensor |
| ADSB.lol | Aircraft positions | Receiver-dependent | Live/near-live | Current nearby aircraft | Not complete; not used for old sightings |
| NASA/JPL CNEOS | Energetic fireballs | Global sensor reporting | Historical | Large atmospheric event match | Not every visual meteor; some locations missing |
| NOAA Kp | Geomagnetic disturbance | Planetary | Current | Broad aurora context | Not local visibility |
| NOAA OVATION | Auroral grid | High-latitude/global grid | Current forecast | Local current aurora context | Not historical in this tool |
| Launch Library 2 | Launch schedules/pads | Global launch sites | Past/upcoming | Launch time + pad context | Schedule is not trajectory visibility |
| CelesTrak GP/OMM | Orbital elements | Tracked catalog | Current elements | Catalog freshness/input to propagation | Catalog presence is not exact pass |
Why “not available” is a real result
If an aircraft feed is live-only, it cannot truthfully answer what flew over the observer yesterday. If OVATION is the latest model, it cannot be silently applied to last month. If CNEOS contains only energetic sensor-detected events, absence does not rule out an ordinary meteor. Sky Event Radar exposes these boundaries instead of converting them into zeros.
Why multiple sources are still worth combining
Each dataset is incomplete in a different way. The witness description contributes movement and appearance; weather constrains visibility; aircraft feeds can provide concrete moving objects; fireball data can support unusually energetic meteors; launch schedules add human-spaceflight context; aurora models add local geomagnetic context. Agreement between independent layers is more informative than simply adding more rows from one source.
Cache and request discipline
Fast-changing sources are cached briefly while slower orbital catalogs are cached longer. CelesTrak GP/OMM checks are cached for at least two hours rather than polled repeatedly. This reduces redundant provider load and prevents freshness from becoming a reason to abuse an upstream service.