Missile site and launch-facility construction monitoring
Excavation scars, silo-lid geometry, and road extensions into empty terrain are legible from commercial and open satellites long before a facility is declared. This page documents the published OSINT methodology used to identify missile garrison construction worldwide.
Sensors
- Maxar WorldView-3: Panchromatic resolution of 0.31 m, multispectral at 1.24 m. Sufficient to resolve silo-lid diameter, construction equipment type, and surface material. Revisit roughly 1 day at mid-latitudes with tasking priority, though cloud and scheduling contention can extend this.
- Maxar WorldView-2: Panchromatic at 0.46 m, eight-band multispectral including coastal blue and near-infrared. Useful for detecting bare-soil exposure through vegetation-index differencing; archive depth to 2009 provides long baselines for change detection.
- Planet SkySat: 0.5 m panchromatic, 1 m multispectral, video-capable. Revisit on tasked targets can reach sub-daily. Shallower archive than Maxar but useful for high-cadence monitoring of active construction phases.
- Sentinel-1 SAR (IW mode): C-band synthetic aperture radar at 5 × 20 m (IW mode, range × azimuth). Cloud-independent and day/night capable. Six-day repeat at mid-latitudes with both satellites active. Coherence-change detection flags ground disturbance even when optical is obscured; backscatter change highlights compacted road surfaces and excavated spoil.
What the ground gives away before the missiles arrive
A missile silo is, for most of its construction life, an engineering project indistinguishable in signature from a large well or a mine shaft. The distinction lies in the pattern. Single excavations are ambiguous. A regular grid of circular pits, each spaced to avoid sympathetic detonation, connected by newly graded roads and served by a cluster of support buildings, is not.
The Middlebury Institute of International Studies documented this methodology in detail when analysing North Korean ICBM infrastructure, and it gained wider public attention in 2021 when analysts using commercial Maxar imagery identified a field of approximately 120 silo-candidate excavations near Yumen in Gansu Province, China, reported by the Federation of American Scientists and subsequently by major press outlets. The analytical steps were not classified. They relied entirely on geometry, pattern, and change over time.
The four signatures analysts look for
Excavation scarring is the first and most visible indicator. Fresh digging exposes subsoil with a spectral signature measurably different from surrounding terrain. In multispectral imagery, the normalised difference vegetation index drops sharply and bare-soil bands brighten. WorldView-3's eight multispectral bands resolve this even in semi-arid terrain where the contrast with natural surface is modest.
Silo-lid geometry is the second. A hardened ICBM silo lid is a large, roughly circular reinforced concrete structure, typically 5 to 15 metres in diameter depending on missile class. At WorldView-3's 0.31 m panchromatic resolution, the lid itself, its hinges, and the blast-deflection apron around it are all resolvable. The Yumen analysis relied partly on this geometric signature to distinguish missile silos from other circular structures such as water tanks.
Road-network extension is the third. Garrisons in previously undeveloped terrain require access. Freshly graded roads appear in SAR backscatter as high-return linear features against low-return natural ground, and in optical imagery as pale linear scars. The rate of extension, direction, and terminus all carry analytical value.
Support-facility patterns are the fourth. Launch-control bunkers, diesel generator buildings, security perimeters, and crew accommodation follow recognisable layouts documented in open military engineering literature and visible in declassified imagery of known Cold War facilities. Their presence alongside excavations substantially raises confidence in the assessment.
Where SAR earns its place in the workflow
Optical imagery is the primary identification tool, but it fails under cloud and cannot be scheduled retrospectively. Sentinel-1 fills both gaps. Its six-day repeat cycle (with both Sentinel-1A and 1B operational, though 1B experienced a failure in 2022 and the constellation has operated on longer revisits since) produces coherence-change maps that detect ground disturbance independently of illumination or weather.
Coherence measures how similar the radar phase is between two passes over the same area. Undisturbed natural surfaces maintain moderate to high coherence. Excavation, compaction, and construction machinery destroy it. A coherence-change product derived from sequential Sentinel-1 IW acquisitions can flag active construction zones at 20-metre posting, providing a persistent alert layer that cues optical tasking. This is not a substitute for the geometric specificity of WorldView imagery; it is a screening tool that prevents analysts from missing activity between optical revisits.
SAR backscatter also responds to the metallic returns from construction equipment and the dielectric contrast of freshly turned earth. These are secondary indicators, but they add temporal resolution to a workflow that might otherwise depend on cloud-free optical windows that arrive weeks apart.
Honest limits of the method
Resolution sets a hard floor on what can be confirmed. At Sentinel-1's 5 m range resolution, you can detect that something changed. You cannot confirm it is a silo. WorldView-3 at 0.31 m can resolve the lid and its geometry, but only if the target has been tasked and the image is cloud-free. Commercial tasking over sensitive locations is sometimes refused, delayed, or degraded by the operator for legal or political reasons.
Camouflage and deception are real. Netting, spoil redistribution, and construction sequencing designed to mimic civilian activity can delay or confuse assessment. The Yumen analysis was possible partly because the construction pace outran the concealment effort. A slower, more deliberate programme could present a harder problem.
Ambiguity is not always resolvable from imagery alone. Circular excavations occur in mining, geothermal drilling, and water infrastructure. Pattern, scale, and context reduce ambiguity but rarely eliminate it. Honest assessments carry confidence levels, not binary conclusions. The published OSINT work on both North Korean and Chinese sites was explicit on this point.
Running a monitoring programme in practice
A practical monitoring programme for a defined area of interest combines three layers. First, a Sentinel-1 coherence-change baseline updated on each six-day pass, covering the full area of interest regardless of cloud. Second, automated change-detection alerts that cue WorldView or SkySat tasking when coherence loss exceeds a threshold. Third, analyst review of optical imagery against a feature checklist: excavation geometry, road extension, support-facility footprint, and spoil-heap volume estimated from shadow length.
Archive depth matters for establishing baselines. Sentinel-1 data is freely available from the Copernicus Data Space from 2014. WorldView-2 archive extends to 2009. Establishing a pre-disturbance baseline, and confirming that a site was genuinely undeveloped before a construction event, is often as analytically important as characterising the construction itself.
Satellize structures analytics of this type as periodic monitoring reports with GIS-layer deliverables, drawing on open Sentinel data and adding commercial tasking on client licence where the geometry demands sub-metre confirmation. Clients with treaty-monitoring or national-assessment mandates can request alert feeds rather than periodic reports.
What the public record already shows
The 2021 reporting on the Gansu silo field, and the earlier Middlebury work on North Korean sites including the Yongjo-ri and Hoejung-ni garrisons, demonstrated that commercial satellite imagery has permanently changed the information environment around strategic weapons programmes. Facilities that would previously have remained obscure for years are now identified within months of breaking ground.
This is not a claim about what any particular government knows through classified means. It is a claim about what is knowable from openly available data by analysts with no special access. The analytical methods are published, the imagery is purchasable, and the archive is deep enough to reconstruct construction timelines retrospectively. For governments and organisations with legitimate monitoring interests, the question is no longer whether this is possible. It is whether they have a systematic programme in place to do it.
Typical figures
| Best optical resolution (panchromatic) | 0.31 m (Maxar WorldView-3) |
| Multispectral resolution | 1.24 m (WorldView-3); 1 m (Planet SkySat) |
| SAR ground resolution (Sentinel-1 IW) | 5 m range × 20 m azimuth |
| Optical revisit (tasked, mid-latitude) | Sub-daily (SkySat); ~1 day (WorldView-3 with priority); subject to cloud and scheduling |
| SAR revisit (Sentinel-1, single satellite) | 12 days; 6 days with both satellites active (1B status subject to operational recovery) |
| Minimum detectable excavation diameter (optical) | ~1 m at WorldView-3 resolution; confident characterisation from ~3 m |
| Spectral bands used | Pan, coastal blue, blue, green, yellow, red, red-edge, NIR1, NIR2 (WorldView-3); C-band VV/VH (Sentinel-1) |
| Archive depth | Sentinel-1 from 2014 (Copernicus Data Space); WorldView-2 from 2009; Landsat from 1972 for coarse baseline |
| Latency (open SAR) | Sentinel-1 products typically available within 1 hour of acquisition via Copernicus Data Space |
| Delivery formats | GeoTIFF change layers, vector feature files (GeoJSON/KML), PDF assessment reports, alert feeds |
Analytics Satellize can run
| Coherence-change screening layer | Sentinel-1 interferometric coherence differencing between sequential IW acquisitions; thresholded to flag ground-disturbance events | GeoTIFF or GeoJSON alert layer, updated per SAR pass, covering defined area of interest |
| Excavation geometry characterisation | Sub-metre optical feature extraction from WorldView-3 or SkySat; pit diameter, spacing, and orientation measured against known silo-class dimensions from open literature | Feature annotation layer with confidence classification and dimensional measurements |
| Road-network extension mapping | Multi-temporal optical and SAR backscatter change detection; linear-feature extraction on newly exposed road surfaces | Vector road-extension layer with timestamps and direction-of-advance annotation |
| Support-facility footprint assessment | Object-based image analysis of optical imagery against published template patterns for launch-control, security, and accommodation structures | Annotated image report with facility-type classifications and confidence levels |
| Construction-timeline reconstruction | Archive trawl across Sentinel-1 and available commercial optical scenes to establish pre-disturbance baseline and sequence construction phases | Chronological PDF report with dated imagery panels and activity timeline |
| Spoil-heap volume estimation | Shadow-length photogrammetry on high-resolution optical imagery; cross-checked against excavation diameter to estimate depth and volume removed | Tabular volume estimates per pit with uncertainty ranges, included in assessment report |
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.