ISKRA · data feed

ISKRA Data Streams

Every satellite, the path its data reaches us by, the product that arrives, and whether it reaches the alerts, only the map, or does not work for us yet.

taken from the code on 15 Sept 2026 · delays measured live, not quoted from documentation

30streams along 11 separate paths

19reach the alerts

6map only, or used as a gate

5arrive, but do not work for us

Path

State

Geostationary Watching without a blink from 36,000 km, but the pixel is 1–3 km and stretched over Bulgaria. Every frame waits for the full-disc scan to finish — hence the identical ~13 minutes along every path.
Satellite	Path	Product	What it gives	Does it reach ISKRA
Meteosat-12	Data Store	FCI Fire Radiative Power `EO:EUM:DAT:0801`	Hotspots at 1 km — the best geostationary resolution we have. every 10 min	alerts Alerts and map.
Meteosat-12	Portugal · LSA SAF	MTG FRP-Pixel	The same hotspots, ready-made, along a different path. every 10 min	alerts Alerts and map.
Meteosat-12	EUMETView · WMS	MTG FRP-Pixel, third route	A third route to the same product. It exists because Portugal went down on 23 July 2026.	alerts Only when the other two go quiet.
Meteosat-12	our dish · `hvs-2`	MTG FRP-PIXEL	The subscription is active, but the transponder has delivered four files in total, the last on 9 Sept.	not working Effectively no feed. We have it along the other three paths.
Meteosat-12	Data Store	FCI Cloud Mask `EO:EUM:DAT:0678`	Where the cloud is — that is, where the thermal sensors are blind.	map The “Cloud cover” layer and its time slider.
Meteosat-10	Portugal · LSA SAF	MSG FRP-PIXEL	Hotspots at 3 km with radiative power in megawatts. every 15 min	alerts Alerts and map.
Meteosat-10	our dish · `E1B-GEO-4`	MPEF Active Fire (FIRC)	A ready fire product with a confidence rating, but no radiative power. 13.5 min after the slot	alerts Wired into the pipeline on 9 Sept 2026.
Meteosat-10	our dish · `E1B-GEO-3`	HRIT L1.5, hourly	The raw imagery for our own detector — but only 24 frames a day. 13.0–13.4 min	not working Accumulates for cross-checking only. The hourly cadence eats the two-hour lead.
Meteosat-10	Data Store	HRSEVIRI L1.5, 15-minute	The same imagery, 96 frames a day instead of 24. The price is 26 GB a day. 13.4 min	not working So far only for calibrating our own detector.
Meteosat-10	our dish · `E1B-GEO-4`	NWC SAF CMA / CTTH	The official cloud mask — more accurate than our own crude 10.8 µm threshold. every 15 min	gate Used as a gate inside the SEVIRI detector.
Meteosat-11 (RSS)	our dish · `E1B-GEO-6`	MPEF Active Fire, 5 min	The fastest fire product we receive. It scans Europe only, which is why it comes every 5 minutes.	alerts ⚠️ Silent since 8 Sept — an announced monthly interruption.
Meteosat-9 (IODC)	Portugal · LSA SAF	MSG FRP-PIXEL IODC	A second angle on the same fire from 45.5° E. Two viewpoints separate a fire from a reflection.	alerts Alerts and map. A witness when suppressing solar-panel glint.

Polar Passing at 800 km and seeing down to 375 m, but only on an overpass — two or three times a day over one place. They see the small fire and miss the moment.
Satellite	Path	Product	What it gives	Does it reach ISKRA
Suomi NPP	NASA FIRMS	`VIIRS_SNPP_NRT`	Hotspots at 375 m — the finest we receive ready-made. ~2 h	alerts Alerts and map.
NOAA-20	NASA FIRMS	`VIIRS_NOAA20_NRT`	The same instrument on another platform — another overpass time. ~2 h	alerts Alerts and map.
NOAA-21	NASA FIRMS	`VIIRS_NOAA21_NRT`	A third VIIRS, a third overpass time. ~2 h	alerts Alerts and map.
NOAA-20 / -21 Suomi NPP	Met service · FTP	`VIIRS-n20 / n21 / npp`	The same hotspots, already filtered for Bulgaria and three times faster than FIRMS. ~60 min	alerts Switched on 31 Aug, when the granule feed stopped.
NOAA-21 and others	our dish · `hvs-1`	EARS-VIIRS `SVMC` + `SVDNBC`	Raw M-band channels at 750 m and the night lights. Eight times faster than the met service route. 7.4 min	not working 73 granules sit unread. Our own detector finds 73% of the ready-made product.
Terra / Aqua	NASA FIRMS	`MODIS_NRT`	Hotspots at 1 km from the oldest instruments still flying. ~2 h	alerts Alerts and map.
Sentinel-3A / 3B	Data Store	SLSTR FRP `EO:EUM:DAT:0417`	Hotspots at 1 km with radiative power, two satellites in opposite phases.	alerts Alerts and map.
Sentinel-3A / 3B	Met service · FTP	`bg_FRP_MWIR1km`	Exactly the same product, but measured 75 minutes earlier than along our own path.	alerts Alerts and map. Not a duplicate — a faster route.
Sentinel-3A / 3B	Met service · FTP	`bg_FRP_SWIR500m`	500 m, a night-time product for weak fires. We do not get it anywhere else.	alerts Our only source for this product.
Sentinel-3C	—	to come	A third SLSTR. Launch 15 Sept 2026 at 01:21 UTC, then months of commissioning.	not yet The labels are already prepared in three places, so it cannot arrive mislabelled.
MetOp-B / -C	our dish · `E1B-EPS-10`	AVHRR L1B	Raw channels — the hotspots are our own detection, not a ready product. ~70 min	alerts Alerts and map. Our own detector.
FY-3D	our dish · `E1H-RDS-2`	EARS-MERSI 1 km	A Chinese instrument reaching Europe over EUMETCast. The hotspots are our own detection.	map The Bulgarian and the European map.
GCOM-C (Shikisai)	JAXA G-Portal · `sftp /nrt`	SGLI L1B 250 m	A raw Japanese instrument. The hotspots are our own detection using their own algorithm; checked against their answer — we find 100% of theirs. Night only. ~90 min	alerts Alerts and both maps. Their ready-made product arrives 34 h later and is no use for an alarm.
MetOp-SG A1	Data Store	METimage VII L1B `EO:EUM:DAT:0464`	Five fire channels at 500 m. Over the village of Anton one fire gave three separate pixels.	map, trial Its own layer, off by default. The product is still in commissioning.
Landsat-8 / -9	Microsoft Planetary Computer	L1 / L2, 30 m	One pixel is 0.09 ha. The hotspots are our own detection from shortwave reflectance.	map Map only — it arrives over 3 hours late, and that is the cut-off for an alert.
Sentinel-2A / 2B	AWS · Element84	L2A, 20 m	It sees the burn scar, not the heat. A 0.04 ha pixel — half a decare leaves a trace.	verification Settles weak single detections after the event.

Not from the sky The two sources with the most accurate location. No sensor tells you which way the front is moving.
Source	Path	Service	What it gives	Does it reach ISKRA
People on the ground	the site and the bot	manual pin, photo, call for help	The exact location no satellite gives, plus the direction of the front and what is needed.	alerts Alerts and map. A human report is drawn as a circle.
Amateur stations	our own key	submitted detections	Third-party receivers feeding the common stream under their own label.	alerts Alerts and map, with a “Radio amateurs” filter.

The three things that arrive and do not work for us

EARS-VIIRS from our dish — 7.4 minutes against 60 through the met service, and 73 granules sit unread.
The 15-minute SEVIRI — we have it through Data Store. Over the Ljubuša fire our own threshold saw the fire at 12:30, the ready product at 14:30. The hourly HRIT simply has no 12:30 frame.
MTG over hvs-2 — the subscription is active, the transponder delivers four files.

The same fire arrives along two or three paths on purpose: MTG along three, VIIRS along two, MSG FRP along two. Losing one source does not leave us blind. We run five detectors of our own — MetOp AVHRR, FY-3D MERSI, Landsat SWIR, SEVIRI IR_039 and METimage — but only the first two reach the alerts.