Space launch facility launch-readiness and vehicle rollout monitoring
Pre-launch ground operations at orbital and sub-orbital complexes follow a documented sequence visible in commercial satellite imagery. Transporter-erector movement, propellant tanker presence and flame-trench flooding provide days of warning before ignition.
Sensors
- Planet SuperDove: 3 m resolution, daily global revisit across the full constellation of roughly 200 satellites. Captures rollout timing and vehicle-erection events within a 24-hour window; insufficient resolution to discriminate vehicle class but reliable for detecting pad occupancy changes.
- Maxar WorldView-3: 0.31 m panchromatic resolution, enabling shadow-length measurement to estimate vehicle height and fairing diameter. Tasked commercially; latency from order to collect typically one to three days depending on cloud and tasking priority.
- Airbus Pléiades Neo: 0.30 m panchromatic, 1.2 m multispectral. Stereo collection in a single pass supports three-dimensional height estimation of erected vehicles. Revisit to a specific site roughly once per day with both satellites combined.
- Sentinel-1 SAR (C-band): 5 m by 20 m resolution in Interferometric Wide Swath mode, six-day repeat at mid-latitudes, all-weather day-and-night collection. Detects large metallic objects such as transporter-erectors through backscatter change; cloud cover that defeats optical sensors during propellant loading is not an obstacle.
The pre-launch sequence as an observable timeline
Every orbital launch programme, regardless of the operator's desire for secrecy, must execute a ground-operations sequence that is physically constrained. Vehicle assembly or transport from a horizontal integration facility to the pad takes hours to days. The transporter-erector, a large metallic structure weighing hundreds of tonnes, produces a distinctive radar return and an unambiguous optical signature. Erection to vertical is followed by umbilical connection, propellant loading and range-safety checks. That sequence, documented in open technical literature for programmes including Soyuz at Plesetsk, Long March at Jiuquan and the Hwasong-derived vehicles at Sohae, spans roughly two to ten days from rollout to ignition depending on propellant type.
Cryogenic propellants such as liquid oxygen impose the tightest timelines: a vehicle cannot sit fuelled for more than a few hours before boil-off forces a scrub or a launch. Storable hypergolic propellants allow longer hold times, which compresses the warning window but does not eliminate it. Propellant tanker vehicles are themselves visible in imagery; their presence at a pad that was empty the previous day is a meaningful indicator even before the rocket appears.
What a shadow tells you about a rocket
At WorldView-3 and Pléiades Neo resolutions, a shadow cast by an erected vehicle across a known-geometry pad can be measured to within a few metres. Combined with the solar elevation angle at the time of collection, shadow length gives vehicle height. Fairing width is measurable directly in the image. These two figures, cross-referenced against the known dimensions of the candidate vehicle families operating from a given facility, allow class-level identification without any classified source.
This is not a novel technique. Analysts at commercial firms and think-tanks including the James Martin Center for Nonproliferation Studies have published vehicle identification work using exactly this method on Sohae and other facilities. The honest limit is that vehicles of similar height and diameter, such as the various Long March 2 variants, can be difficult to discriminate on geometry alone. Payload fairing shape adds discriminating information when the image geometry allows it to be resolved.
SAR when the clouds arrive on cue
Coastal launch sites, including Sohae on the Yellow Sea and Jiuquan in the Gobi, can experience cloud cover during periods of meteorological activity that conveniently coincides with launch preparation. Sentinel-1's C-band radar penetrates cloud and collects at night. A transporter-erector in transit produces a strong double-bounce return from its corner geometry; an empty pad produces a lower, more uniform backscatter. Change detection between consecutive Sentinel-1 passes, six days apart at mid-latitudes, will flag pad occupancy changes reliably for objects larger than roughly 10 metres.
The limitation is resolution. At 5 m by 20 m in Interferometric Wide Swath mode, Sentinel-1 can confirm that a large metallic object is present but cannot measure it precisely enough for vehicle-class identification. Commercial SAR constellations, including Capella Space and ICEYE, offer sub-metre resolution and more frequent revisit, but those are tasked assets with associated cost and scheduling constraints. For persistent monitoring of known sites, combining Sentinel-1 as the always-on tripwire with tasked high-resolution optical or commercial SAR as the confirmation layer is the practical architecture.
Flame-trench flooding and the final hours
Several launch facilities flood the flame trench beneath the pad with water immediately before ignition to suppress acoustic energy and protect the vehicle base. At Plesetsk and at NASA's Kennedy Space Center, this deluge system is visible in high-resolution imagery as a water pool or wet concrete signature at the base of the launch structure. The appearance of this feature in an image collected within hours of a scheduled window narrows the warning to the final launch day.
Planet's daily collect is the right tool here. A morning image showing a dry trench and an afternoon image showing a flooded one, combined with a vehicle already vertical, constitutes a high-confidence imminent-launch indicator. The practical constraint is tasking latency and cloud. Planet's constellation provides the temporal density; cloud cover over the site on the critical day can eliminate the optical cue entirely. That is why multi-sensor fusion, not reliance on any single system, is the only credible monitoring architecture.
Archive depth and the baseline problem
Detecting change requires knowing what normal looks like. Planet's archive extends to 2016 for most sites; Maxar's commercial archive goes back further for selected high-priority locations. Sentinel-1 data is freely available from 2014. A site like Sohae has been imaged hundreds of times across these archives, providing a well-characterised baseline for pad occupancy, support-area vehicle density and road-traffic patterns.
Less-documented facilities present a harder problem. A new pad at a facility with sparse archive coverage may lack the baseline needed to make change detection reliable. In those cases, the first months of monitoring are partly spent establishing the baseline rather than generating alerts. Buyers should expect a calibration period of four to eight weeks for any facility not already covered by an existing monitoring programme. Satellize's analytical workflow for this use case draws on the same open-constellation archive approach used across its other national-security programmes, including the Tonga crop-estimation work, adapted for the specific object classes and change signatures of launch infrastructure.
Honest limits of open-source launch monitoring
Several things cannot be determined from commercial imagery alone. Payload identity is almost never recoverable from optical data; a fairing is a fairing. Orbital parameters are not knowable before launch. Sub-orbital versus orbital intent is sometimes inferrable from vehicle class but not always. Underground or indoor vehicle assembly, as practised at some facilities, is invisible until the vehicle emerges. And a government that chooses to conduct a short-notice launch from a mobile platform or a previously unused pad will defeat a monitoring programme built around fixed known sites.
The value of persistent monitoring is not omniscience. It is the systematic compression of surprise. A programme that provides 48 to 72 hours of warning before a launch at a known facility, with a low false-alarm rate, is operationally useful even if it cannot name the payload or predict the trajectory. That is the realistic performance envelope, and it is worth stating plainly.
Typical figures
| Best available optical resolution | 0.30 m panchromatic (WorldView-3, Pléiades Neo) |
| Typical daily-revisit optical resolution | 3 m (Planet SuperDove constellation, ~200 satellites) |
| SAR resolution (Sentinel-1 IW mode) | 5 m range × 20 m azimuth; all-weather, day/night |
| Sentinel-1 revisit at mid-latitudes | 6 days (single satellite); 3 days with both Sentinel-1A and 1B operational |
| Tasked high-resolution collect latency | 1 to 3 days from order (WorldView-3, Pléiades Neo); subject to cloud and tasking queue |
| Minimum detectable object (SAR change) | Approximately 10 m metallic structure producing corner-reflector return in Sentinel-1 |
| Vehicle height estimation accuracy | ±2 to 5 m from shadow-length method at sub-0.5 m resolution, depending on solar elevation angle |
| Archive depth | Planet from 2016; Sentinel-1 from 2014 (free); Maxar commercial archive from early 2000s for priority sites |
| Spectral bands used | Panchromatic (0.45–0.80 µm) for resolution; C-band SAR (5.405 GHz) for all-weather change detection |
| Delivery formats | Georeferenced GeoTIFF, annotated PDF report, GIS vector overlay, alert notification |
Analytics Satellize can run
| Pad-occupancy change alert | Pixel-level change detection on Planet daily collect against site-specific baseline; threshold set on pad-area brightness and object presence | Automated alert within 24 hours of a new collect showing occupancy change; GeoTIFF with annotated bounding box |
| Vehicle-class identification report | Shadow-length and fairing-geometry measurement on sub-0.5 m tasked imagery, cross-referenced against published vehicle dimension tables | Analyst report with candidate vehicle list, confidence rating and annotated image extract |
| Support-vehicle concentration score | Object detection on high-resolution optical imagery to count and classify vehicles in defined support-area polygons; compared against historical baseline counts | Time-series chart of vehicle counts per zone; threshold-breach alert when count exceeds two standard deviations above baseline |
| SAR pad-occupancy layer | Sentinel-1 backscatter change detection between consecutive passes; double-bounce signature flagging for large metallic objects on pad | GIS polygon layer updated on each Sentinel-1 pass; binary occupancy status per pad |
| Flame-trench flood detection | Spectral water-index analysis (NDWI or equivalent) applied to multispectral Planet or Pléiades collect at pad base polygon | Binary wet/dry flag per collect, timestamped; integrated into launch-readiness composite score |
| Launch-readiness composite score | Fusion of pad occupancy, vehicle presence, support-vehicle count and flame-trench status into a weighted indicator score, updated daily | Daily scored dashboard entry per monitored facility; escalation report when score crosses defined threshold |
| Historical launch-cycle baseline report | Retrospective analysis of archive imagery to characterise typical pre-launch observable sequence and timing at a named facility | Written baseline report with annotated image timeline; used to calibrate alert thresholds for ongoing monitoring |
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.