Airfield runway repair and reconstruction monitoring
Military runway repair, extension, and crater infill leave measurable spectral and radar signatures. Open Sentinel data flags the change; sub-metre commercial imagery confirms the extent.
Sensors
- Sentinel-2 MSI (SWIR bands B11 and B12): 10 m visible, 20 m SWIR resolution; 5-day revisit at mid-latitudes with two satellites. SWIR bands at 1610 nm and 2190 nm are the primary open-data change signal: fresh asphalt absorbs strongly in SWIR, aged bitumen reflects more as oxidation and aggregate exposure increase. Free archive from 2015.
- Sentinel-1 SAR IW mode: C-band (5.4 GHz), 5 × 20 m ground-range resolution in IW mode, 6-day repeat. Fresh tarmac and disturbed aggregate both shift backscatter relative to smooth aged pavement. Cloud-independent, day/night capable. Coherence differencing between pre- and post-repair acquisitions highlights surface change even when optical is obscured.
- Planet SkySat: Sub-metre panchromatic (0.5 m), 1.0 m multispectral. Tasked on demand; latency typically same-day to 24 hours. Confirms repair geometry, measures patch dimensions, and resolves crater-fill quality. Limited SWIR coverage, so used in combination with Sentinel-2 for spectral context.
- Maxar WorldView-3: 0.31 m panchromatic, 1.24 m multispectral, with 8 SWIR bands at 3.7 m. The SWIR capability at sub-5 m resolution makes WorldView-3 the highest-fidelity sensor for distinguishing fresh asphalt from aggregate fill within individual repair patches. Tasked commercially; not open-access.
What fresh asphalt gives away in shortwave infrared
Bituminous asphalt has a distinctive SWIR signature when freshly laid. New pavement absorbs strongly across the 1400–2500 nm range because the binder is dense, dark, and relatively smooth. As a surface ages, ultraviolet exposure oxidises the binder, aggregate becomes exposed, and the surface brightens in SWIR. The reflectance difference between a fresh repair patch and the surrounding aged runway can reach 0.05–0.10 in normalised reflectance units in Sentinel-2 Band 11, a shift large enough to detect at 20 m resolution even on a runway only 30 m wide.
This is not a subtle effect. Published analysis of Syrian and Libyan airfields used exactly this contrast to identify runway resurfacing activity and crater repair after airstrikes, demonstrating that the method works on standard open-data archives without commercial tasking. The signal fades over months as the new surface weathers, which means the timing of a Sentinel-2 acquisition relative to the repair matters. A patch laid in week one may be undetectable in SWIR by month six, so change-detection workflows need to compare against a baseline from before the repair, not against a contemporaneous scene from a different airfield.
SAR backscatter as a cloud-independent corroborator
Optical SWIR is the primary detection method, but it fails under cloud. Sentinel-1 C-band SAR provides an independent signal. Disturbed aggregate and freshly compacted fill scatter C-band energy differently from smooth aged tarmac: the surface roughness change shifts backscatter by 1–3 dB in published airfield studies, which is detectable in a simple pre/post ratio image.
Coherence interferometry adds a second layer. Sentinel-1 repeat-pass coherence over stable, smooth surfaces such as runways is typically high (coherence values above 0.6 over 6-day pairs). Construction activity, crater fill, or resurfacing breaks that coherence sharply. A coherence-loss map over a 6-day window isolates exactly where the surface changed, independent of cloud cover or time of day. The method cannot distinguish runway repair from, say, a fuel spill or parked equipment without corroborating optical evidence, so the two sensor streams are always used together.
Crater repair has a different signature from routine resurfacing
Routine runway resurfacing tends to produce large, geometrically regular patches aligned with the runway centreline. Crater repair after a strike produces irregular, smaller patches at discrete points, often accompanied by disturbed soil or aggregate outside the runway edge where fill material was stockpiled. In sub-metre optical imagery these two scenarios are visually distinguishable. In 20 m Sentinel-2 data they are not always separable, which is an honest limit of the open-data baseline.
The practical workflow is therefore two-stage. Sentinel-2 SWIR flags any change in runway spectral character within days of the repair. Commercial tasking of SkySat or WorldView-3 then provides the geometry and extent needed to classify the repair type and estimate the time to operational readiness. A runway with a single crater-fill patch of 15 m diameter is not operationally equivalent to one that has been fully resurfaced, and the sub-metre scene is what makes that distinction possible.
Extension and threshold displacement: the longer game
Runway extension is a slower process than repair but strategically more significant. Adding 500 m to a runway changes the aircraft types it can support. Extension activity is detectable in Sentinel-2 imagery through progressive soil disturbance and grading at the threshold, followed by the appearance of new pavement with a strong SWIR absorption signature. The timeline from ground-breaking to paving at a major airfield is typically months, giving multiple acquisition opportunities.
SAR is particularly useful here because graded soil and compacted sub-base have distinct backscatter signatures from both the original runway and the surrounding terrain. A time-series of Sentinel-1 backscatter over the threshold area will show the progression from natural ground to disturbed soil to compacted base to finished pavement, each stage distinguishable if the baseline archive is long enough. Sentinel-1 data is available from 2014, providing over a decade of retrospective analysis at no data cost.
What this method cannot do
Resolution is the primary constraint. Sentinel-2 at 20 m SWIR cannot resolve individual repair patches narrower than roughly 40 m, and a single-lane taxiway repair may be invisible in the open-data baseline entirely. Revisit is five days under clear skies, which means a rapid repair completed between passes may only be detected after the fact, not in real time.
SAR coherence analysis requires at least two acquisitions separated by the satellite's repeat period, which is 6 days for Sentinel-1. A repair completed and cured within that window may not show coherence loss at all if the new surface is sufficiently smooth. Commercial SAR constellations with shorter revisit exist but are not open-access. Cloud is not a problem for SAR but is a genuine problem for the optical confirmation step: sustained cloud cover over an active conflict airfield can delay sub-metre confirmation by days to weeks. Finally, the method detects surface change, not intent. A repaired runway is a repaired runway; inferring operational status requires integration with other intelligence streams.
Putting it into practice
A practical monitoring programme for a set of airfields of interest combines a persistent Sentinel-2 SWIR change-detection layer, refreshed every five days from open archive, with triggered commercial tasking when a change alert exceeds a defined threshold. The Sentinel-1 coherence layer runs in parallel as a cloud-independent corroborator. Sub-metre tasking is reserved for confirmed alerts, keeping commercial data costs proportionate to actual activity.
Satellize runs this kind of layered open-plus-commercial architecture for analytics clients. The same spectral-change and coherence methods that underpin the Tonga crop-estimation programme apply directly to surface-change detection on hard infrastructure, with the obvious difference that the change signal of interest is bitumen rather than vegetation. Clients receive a GIS-ready alert layer, a change-classification report, and, where commercial imagery has been acquired, a georeferenced patch-extent polygon with estimated repair area in square metres. Analysts wanting to interrogate the underlying Sentinel data directly can request access to the full time-series stack.
Typical figures
| Open-data spatial resolution (SWIR) | 20 m (Sentinel-2 B11/B12) |
| Commercial optical resolution | 0.31 m pan / 1.24 m multispectral (WorldView-3); 0.5 m pan (SkySat) |
| SAR resolution (Sentinel-1 IW) | 5 × 20 m (range × azimuth, ground range) |
| Open-data revisit | 5 days (Sentinel-2, two satellites, mid-latitudes); 6 days (Sentinel-1) |
| Commercial tasking latency | Same-day to 48 hours depending on constellation and cloud |
| Key spectral bands | SWIR 1610 nm and 2190 nm (Sentinel-2 B11/B12); WorldView-3 SWIR 8-band 1195–2365 nm at 3.7 m |
| Minimum detectable repair patch (open data) | Approximately 40 m × 40 m in Sentinel-2 SWIR; smaller patches require sub-metre optical |
| Archive depth | Sentinel-2 from 2015; Sentinel-1 from 2014; WorldView archive varies by site |
| Cloud limitation | Optical SWIR blocked by cloud; SAR coherence unaffected but requires 6-day pair |
| Delivery formats | GeoTIFF change rasters, GeoJSON alert polygons, PDF assessment reports |
Analytics Satellize can run
| SWIR change-detection alert | Normalised difference of Sentinel-2 B11 reflectance between baseline and current acquisition; threshold exceedance flagged per runway segment | Automated GeoJSON alert layer, refreshed every 5 days per site |
| SAR coherence-loss map | Sentinel-1 IW repeat-pass interferometric coherence differencing (6-day pairs); coherence drop below 0.4 flagged as surface-change indicator | GeoTIFF coherence-loss layer overlaid on runway footprint |
| Repair-type classification | Patch geometry and spectral profile from sub-metre optical (SkySat or WorldView-3) combined with SWIR context; rule-based classifier distinguishing crater fill, partial resurfacing, and full resurfacing | Classified polygon layer with repair-type attribute and estimated area in m² |
| Runway extension progress report | Multi-temporal Sentinel-1 backscatter time-series over threshold areas combined with Sentinel-2 NDVI suppression and SWIR increase; stage-of-construction classification | Monthly PDF progress report with annotated imagery and estimated extension length |
| Operational-readiness change indicator | Composite score derived from repair extent, patch age (days since SWIR anomaly onset), and surface-smoothness proxy from SAR backscatter standard deviation | Per-airfield status table updated with each new acquisition cycle |
| Historical repair timeline | Retrospective SWIR and SAR stack analysis over Sentinel archive (2014/2015 to present); repair events dated and mapped | GIS layer of dated repair events with associated imagery chips; Excel timeline export |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.