On 13 August 2026, a day before the Hautes Fagnes, a forest fire broke out in the Hürtgenwald in the district of Düren that became the largest in North Rhine-Westphalia since official record-keeping began in 1991. This fire is instructive for one reason: our geostationary satellite was not first here. VIIRS NOAA-21 saw it at 13:45, MTG only two hours later.
The fire broke out on the afternoon of Thursday 13 August in woodland between Großhau and Gey. Authorities initially reported about 25 hectares; within a few hours that was around 300. Because of the rapid spread and a fire front rolling towards Gey, the village was evacuated during the night of 13 to 14 August: about 1,800 residents.
There were two incidents in the Hürtgenwald that day. First a small patch of woodland between Großhau and the Wehebach reservoir, brought under control at around 16:00. Then the fire at Gey. Our earliest detection at 13:45 lies further north than the main cluster and probably belongs to that first, smaller fire.
The geostationary strip only starts at 15:50 here, while VIIRS had already seen the fire at 13:45. An MTG pixel is about 2 kilometres; a ground fire under canopy covering 25 hectares does not fill that pixel with enough heat to cross the detection threshold. Only when the fire spread to 300 hectares and reached the canopy did the cell light up.
Eight satellite sources saw this fire. Measured from 13:45, the moment of the first detection. Note the first two rows: acquisition and delivery do not coincide.
| Source | Acquired | Available | After first detection |
|---|---|---|---|
| VIIRS NOAA-21polar | 13:45 | 16:09 | 144 min |
| MTG FRP-PIXELgeostationary | 15:50 | 16:09 | 144 min |
| MODISpolar | 16:29 | 17:34 | 230 min |
| MSG FRP-PIXELgeostationary | 17:00 | 17:34 | 230 min |
| Sentinel-3Apolar | 22:55 | 23:52 | +608 min |
| VIIRS Suomi-NPPpolar | 02:35 | 13:56 | +1,451 min |
| VIIRS NOAA-20polar | 02:52 | 14:17 | +1,473 min |
| Sentinel-3Bpolar | 12:08 | 14:17 | +1,473 min |
Sentinel-3 is included for completeness. That source is valuable for mapping a burnt area, not for the first warning.
VIIRS NOAA-21 acquired the fire at 13:45 but only delivered at 16:09; MTG acquired it at 15:50 and delivered in the same minute. So the polar satellite saw it earlier, the geostationary one delivered faster, and by coincidence they arrived in the same quarter of an hour. Without that coincidence the first warning would have been two hours later.
Detections per hour, local time. The dotted line marks the highest power measured: 408 MW during the night of 14 August, when the fire at Gey flared up again and the village was evacuated.
Geostationary is not always faster. For a small, cool ground fire under trees, a 2-kilometre pixel loses to a 375-metre polar measurement. Selling one source as the solution sells half the truth: it is the combination that matters, and which of the two sees something first depends on the kind of fire.
Our own area calculation came out at nearly 3,000 hectares for this fire, while the authorities reported about 300. That is a factor of ten.
The enclosing shape we compute around a detection cloud is an upper bound, not a measured perimeter: the further apart the detections, the larger the shape. For an elongated fire front in woodland that adds up sharply. We use that value internally to rank fires, not to report area.
We marked the fire as ended as soon as detections stopped. In reality it was contained but not extinguished at that point, with flames still visible in two places.