Observations
What the evidence shows
What we see
Optical satellite imagery of the airfield is available before the reported recapture and not after it. The 4 October pass is 56-60% cloud. Radar covered the same ground six hours after the event.
What it is consistent with
The optical check could not be made, and that is different from the check coming back empty. October is the tail of the rainy season over the Tigray highlands, so this will recur: optical imagery of this region in this season is unreliable on any given date. The gap this exposed has since been closed — the radar pass overhead at the time has now been processed, and is recorded here as s1-airfield-2026-10-03. At 10 m, optical would in any case have shown large changes on the apron or a burn scar, not damage to a runway or an individual structure.
What we see
The runway reads on all three passes as a single connected low-backscatter feature. Its footprint, outlined on the 9 September pass alone, measures 3,706 m along its axis on a bearing of 22 degrees, which corresponds to the published 02/20 designation. Inside that fixed outline the brightest pixel is -16.0 dB on 9 September, -11.5 dB on 21 September and -11.5 dB on 3 October; no pixel reaches -10 dB on any pass. Mean backscatter inside it rises 0.68 dB between 21 September and 3 October, against a scene-wide median rise of 1.14 dB over the same interval.
What it is consistent with
A smooth paved surface reflects radar away from the sensor and reads dark. Cratering, rubble, large vehicles or aircraft standing on the surface are bright at this wavelength and would raise returns within the outline. The figures are consistent with a runway surface carrying no obstruction at a scale of about 20 m or larger on 3 October, and the rise inside the outline is smaller than the rise across the scene as a whole. This says nothing about who held the airfield, and radar cannot be made to say it.
What we see
Across the event interval, 1.2% of the 6 km box changed by more than 3.7 dB. Across a control interval of the same length between two quiet dates on the same track, the figure is 0.8%. The difference between them is 0.3 percentage points of area.
What it is consistent with
The event interval produced no more change than a quiet interval of the same length over the same ground. Radar amplitude at this resolution registers large, bright or geometric change; it is not interferometric coherence and is far less sensitive to structural damage. A null result here is weak evidence that nothing happened, and no evidence at all about control.