If you run a station copying polar-orbiting weather satellites and you're within reach of the Balkans, one extra pass a night isn't noise, it's data — sometimes the difference between catching a fire at 2am and hearing about it from a farmer at sunrise. Reply or DM.
What we need, technically: raw HRPT/AHRPT downlink from NOAA/MetOp AVHRR (L-band, ~1698–1707 MHz) or raw CADU frames before any vendor processing — SatDump output works fine.
We do our own calibration and detection on it.
Coverage gap is 41–44N, 22–29E, night passes especially.
Proof it works: Michael in Athens sent us a raw MetOp AVHRR capture that caught a farm fire burning overnight in southern Bulgaria — missed by our geostationary feed and by FIRMS, decoded straight from the CADU stream, nothing else involved.
What would help most: station operators willing to share raw captures over
41–44 N, 22–29 E — especially night passes.
We run our own detection on raw AVHRR 3.7 and 10.8 µm. The processing we can do. The
coverage we cannot.
Map: iskra.gamera.bg/map/
What it cannot do, since that matters more:
Two of our main geostationary products are still demonstrational — the crews are told
which. A small fire inside a village falls under the pixel; we missed one on 13 Aug
while catching a 550-decare forest fire 40 km away.
Measured latency, sensing to alert:
geostationary 15–25 min
MetOp AVHRR 58–113 min (global stream)
VIIRS via FIRMS 114–130 min
The polar sensors see the small and the night fires, and they reach us an hour and a
half late. That is the number that decides everything else.
Where it stands today:
· 20 channels — 14 volunteer formations, the forest agency, a park guard service
· 814 alerts for 301 separate fires
· 12 satellites in the stream
· burned area measured on Sentinel-2 for 68 fires, about 3000 ha
