Drone base and UAV operating infrastructure detection
Tactical UAV operating bases occupy a fraction of the ground a manned airfield needs, but sub-metre optical and SAR imagery reveals the shelters, launch rails, antenna arrays and generator pads that betray them. Open-source analysis of Shahed-series sites has already proved the method works at scale.
Sensors
- Maxar WorldView-3: 31 cm panchromatic resolution, 13.1 km swath. The current commercial ceiling for optical detail on small structures. Multispectral bands at 1.24 m support material and camouflage-net discrimination. Revisit approximately 1 day at mid-latitudes with off-nadir tasking.
- Airbus Pléiades Neo: 30 cm panchromatic, 70 cm multispectral, 14 km swath. Native stereo and tri-stereo collection enables 3D height extraction of shelters and masts without ground control. Revisit up to twice daily over a given point using the two-satellite constellation.
- Planet SkySat: 50 cm collect resolution, 6.6 km swath. Less detail than WorldView-3 or Pléiades Neo but a larger constellation allows more frequent revisit windows for monitoring known sites. Video mode captures vehicle movement at candidate locations.
- Capella Space Spotlight SAR: Synthetic-aperture radar at X-band (9.65 GHz), achieving approximately 50 cm resolution in Spotlight mode. Penetrates cloud and collects at night, which matters when optical access is denied. Coherent change detection between repeat passes flags ground disturbance and new surface features at centimetre-scale displacement sensitivity.
- Thermal infrared (Landsat 9 TIRS, ECOSTRESS): Landsat 9 TIRS provides 100 m thermal resolution with 16-day revisit. ECOSTRESS on the ISS reaches approximately 70 m at irregular revisit. Neither resolves a generator pad directly, but a persistent thermal anomaly co-located with a sub-metre optical signature strengthens identification confidence considerably.
Why a twelve-metre shelter is the tell
A Shahed-136 loitering munition has a wingspan of 2.5 metres and a body length of 3.5 metres. Its operating infrastructure is correspondingly compact: a fibreglass or canvas shelter roughly 10 to 15 metres long, a launch rail angled between 15 and 30 degrees, a ground-control station with one or two dish antennas, and a generator or vehicle-mounted power unit. The entire active pad can sit inside a single WorldView-3 image tile with room to spare. Open-source investigators, drawing on publicly available Maxar and Planet imagery, have geolocated Shahed launch sites in Iran and staging areas in Russia with sufficient confidence to publish coordinates. The methodology is not classified.
The challenge is not resolution in isolation. It is distinguishing a UAV shelter from a farm outbuilding, a comms relay hut or a temporary construction site. That requires reading the ensemble: shelter orientation relative to prevailing wind (launch rails need a clear run), proximity to a hardstand or vehicle track, the presence of antenna masts with specific dish geometries, and thermal signatures consistent with continuous generator operation. Each element alone is ambiguous. Together they are not.
What 30 cm resolution actually shows, and what it does not
At 31 cm (WorldView-3 panchromatic), an analyst can resolve the ridge line of a prefabricated shelter, the shadow cast by a dish antenna, the wheel ruts of a vehicle that positioned a launch rail, and the scorch or compaction mark left by repeated launches. Camouflage netting appears as a textured low-contrast patch with a characteristic shadow edge. These are real, measurable features in the public imagery record.
What 30 cm optical cannot do: see through cloud, operate at night without a thermal complement, or resolve internal equipment through a closed shelter roof. It also cannot confirm a site is active rather than abandoned. A single collect is a snapshot. Confidence in operational status requires temporal depth, ideally a sequence of collects over days or weeks showing vehicle presence, antenna repositioning or launch-rail angle changes. Archive imagery from Planet's daily constellation is often the fastest route to that temporal baseline.
SAR fills the gaps optical cannot
Capella Space Spotlight SAR at approximately 50 cm resolution produces a radar backscatter image that is independent of cloud cover and solar illumination. Metal structures, including dish antennas and launch rail frames, produce strong coherent returns that stand out against soil or vegetation backgrounds. A fibreglass shelter has lower backscatter but still produces a detectable double-bounce return at its base where the wall meets the ground.
Coherent change detection is the more powerful SAR technique for this use case. By comparing two co-registered complex SAR images of the same site taken days apart, analysts can identify areas where surface coherence has been lost, indicating ground disturbance, new construction or vehicle movement. Published studies using Sentinel-1 (5 m resolution, 12-day repeat) have demonstrated coherence-loss detection of earthworks at sub-metre displacement. Capella's 50 cm Spotlight mode extends that sensitivity to smaller features. The honest limit: SAR geometry produces layover and shadow artefacts in hilly terrain that can mask or mimic features, requiring careful geometric correction before interpretation.
Thermal signatures: supporting evidence, not a primary discriminant
A diesel generator running continuously to power ground-control electronics and communications equipment produces a heat signature detectable in thermal infrared imagery. At Landsat 9 TIRS resolution of 100 m, that signature is a diffuse warm pixel, not a resolved object. Its value is as corroborating evidence when an optical or SAR collect has already flagged a candidate site. A warm anomaly that persists across multiple Landsat passes, co-located with a compact compound identified in sub-metre optical imagery, raises confidence materially.
ECOSTRESS on the ISS offers slightly finer resolution (approximately 70 m) but irregular revisit tied to ISS orbital precession. Neither sensor is a substitute for sub-metre optical in primary detection. Thermal is most useful for assessing whether a known site remains active after camouflage or structural changes have obscured the optical signature.
Building a detection workflow: from tip to confident identification
Practical detection follows a tiered logic. A medium-resolution daily optical collect (Planet Dove at 3 to 5 m) or a Sentinel-1 SAR pass flags a candidate area through anomaly detection or change from a baseline image. That tip triggers a tasking request for a sub-metre collect, WorldView-3 or Pléiades Neo, over the specific coordinates. The high-resolution collect is then analysed against a feature checklist: shelter geometry, antenna presence, launch-rail shadow, hardstand, thermal corroboration. If three or more features are confirmed, the site is classified as a probable UAV operating location.
Archive depth matters. WorldView-3 has been collecting since 2014; Pléiades since 2012. For a site that has been active for months, archive imagery often shows the construction sequence, which is itself analytically valuable. Planet's daily archive from 2016 onwards provides the temporal density needed to track vehicle presence patterns.
Satellize runs this tiered workflow for defence and national-security clients, combining open-constellation monitoring with commercial tasking on client licence. The same analytical logic that underpins the Tonga crop-estimation programme, systematic feature extraction against a changing baseline, applies directly to compound-level change detection in security contexts.
Honest limits and the ambiguity problem
No remote-sensing method produces a confirmed identification without some residual ambiguity. A well-camouflaged site under a continuous cloud deck in a SAR-shadow zone will degrade even a well-resourced collection plan. Deliberate deception, dummy shelters, repositioned antennas and decoy generator pads, is a known countermeasure that open-source analysis has documented at several locations. The methodology described here is resistant to casual concealment but not to a disciplined denial and deception programme.
Resolution is also not the only variable. A 30 cm image collected at a steep off-nadir angle introduces geometric distortion that complicates feature measurement. Analysts should prefer nadir or near-nadir collects for precise shelter-dimension estimation. Finally, the legal and policy framework governing commercial imagery tasking over active conflict zones varies by operator and jurisdiction. Clients must account for those constraints in collection planning.
Typical figures
| Minimum optical resolution for confident identification | 30 cm panchromatic (WorldView-3, Pléiades Neo) |
| SAR resolution (Spotlight mode) | ~50 cm (Capella Space X-band) |
| Thermal resolution | 100 m (Landsat 9 TIRS); ~70 m (ECOSTRESS) |
| Optical revisit at tasked priority | ~1 day (WorldView-3 off-nadir); up to 2× daily (Pléiades Neo 2-satellite) |
| SAR revisit | Irregular on-demand (Capella); 6–12 days (Sentinel-1, 5 m IW mode) |
| Minimum detectable structure | Shelter ~10 m length at 30 cm; antenna mast ~1 m diameter dish at 31 cm |
| Archive depth | WorldView-3 from 2014; Planet daily from 2016; Sentinel-1 from 2014 |
| Cloud penetration | SAR only; optical blocked by cloud cover |
| Coherent change detection sensitivity | Sub-metre surface displacement (Capella Spotlight repeat pairs) |
| Delivery formats | Georeferenced GeoTIFF, annotated PDF report, GIS vector overlays (GeoJSON/KML), change-alert feed |
Analytics Satellize can run
| Candidate site detection | Anomaly detection on medium-resolution daily optical (Planet Dove 3–5 m) or Sentinel-1 SAR coherence baseline | Coordinate list of flagged candidate locations with confidence tier, delivered as GeoJSON alert feed |
| High-resolution feature confirmation | Structured analyst checklist (shelter geometry, antenna shadow, rail orientation, hardstand presence) applied to sub-metre optical collect | Annotated image report with feature labels and site classification (probable / possible / negative) |
| Temporal activity assessment | Multi-date optical stack analysis for vehicle presence, antenna repositioning and launch-rail angle change across Planet or archive WorldView imagery | Activity timeline chart and summary report indicating periods of probable operational use |
| SAR coherent change detection | Complex interferometric coherence comparison between Capella Spotlight repeat passes to detect ground disturbance and new surface features | Coherence-loss map (GeoTIFF) overlaid on optical basemap, with change polygons classified by magnitude |
| Thermal corroboration layer | Landsat 9 TIRS or ECOSTRESS anomaly extraction co-registered to candidate site coordinates | Thermal time-series chart per site with anomaly flags, delivered as supplementary GIS layer |
| Network mapping of dispersed sites | Spatial clustering of confirmed and probable sites to infer supply-line geometry and command relationships | Network diagram and georeferenced site-relationship map in PDF and GeoJSON |
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.