Methodology and limits
How SkyWatch normalizes observations, manages freshness, avoids duplicates, and communicates uncertainty.
Open the live mapNormalization and identity
Incoming records use different names, identifiers, coordinate formats, and update intervals. SkyWatch converts them into common entity contracts before rendering. Aircraft are matched primarily by stable transponder or provider identifiers, with conservative aliases used only where the data supports them.
Records from more than one source are merged only when identity evidence is strong enough. When it is not, SkyWatch prefers separate markers over a confident-looking but incorrect merge.
Freshness before appearance
Each live layer has its own expected update cycle. The interface reports the source, last update, and cache age where available. A successful HTTP response containing old data is not treated as a fresh observation.
Network-first behavior is used for emergency information. Cached material can preserve context during an outage, but stale state must remain visible and cannot silently become a live claim.
Alerts and UAV event trails
Repeated copies of the same warning are deduplicated into one event. Recent alerts remain prominent for 30 minutes and become visually subdued for contextual history up to 12 hours. A formal all-clear closes the applicable event.
UAV trails connect the sequence of alert locations within one incident. An all-clear ends that incident so the next incursion starts a separate trail rather than being joined to an older route.
Thermal and environmental signals
Thermal detections are filtered for persistent industrial heat, low-confidence pixels, isolated weak observations, and known environmental noise. Even retained detections are leads for investigation, not attribution.
GNSS interference, network disruptions, radiation, seismic activity, clouds, and precipitation have different spatial resolutions. Their boundaries should be read as measured or modeled areas, not precise lines on the ground.