Ceasefire and arms-control treaty verification from open imagery
Commercial satellite imagery has become a primary tool for verifying ceasefire lines, weapon-exclusion zones, and arms-control obligations. This page covers the sensors, observables, resolution requirements, and legal standing of open imagery in treaty monitoring.
Sensors
- Maxar WorldView-3: 31 cm panchromatic resolution, enabling identification of equipment class (tank versus armoured personnel carrier versus artillery piece) and counting of individual vehicles. Revisit roughly 1–4.5 days at mid-latitudes depending on tasking priority. The resolution floor for distinguishing a T-72 from a BMP-2 by silhouette and track width sits at approximately 30–50 cm, which WorldView-3 meets.
- Airbus Pléiades Neo: 30 cm native resolution, four-satellite constellation achieving sub-daily revisit over priority areas. Tri-stereo tasking enables digital surface model generation useful for detecting new earthworks, berms, or revetments within exclusion zones. Radiometric quality supports shadow-length analysis for height estimation of parked equipment.
- Planet SkySat: 50 cm resolution, up to 15 frames per day over a single target. Lower throughput per image than WorldView-3 but high temporal density is valuable for confirming whether equipment observed in one pass has moved or been concealed. Used by open-source analysts during the 2020 Nagorno-Karabakh conflict to track armoured column positions.
- Sentinel-1 SAR IW mode: C-band synthetic aperture radar at 5 m resolution in Interferometric Wide swath mode, 250 km swath, 6-day repeat (12-day per orbit direction). Cloud-independent and day/night capable. Detects large metallic objects, vehicle concentrations, and ground disturbance through coherence change. Cannot distinguish equipment class at this resolution but confirms presence or absence of vehicle concentrations in a zone regardless of weather.
What the agreements actually require monitors to see
Arms-control and ceasefire agreements define compliance in physical terms: heavy weapons must remain beyond a specified distance from a line; certain equipment categories must be absent from a buffer zone; troop concentrations must not exceed defined thresholds. The Minsk agreements for eastern Ukraine, for example, specified 15 km and 25 km withdrawal distances for different weapon calibres. The Libya ceasefire of October 2020 prohibited armed vehicles in and around Sirte. The UN Mission in South Sudan has long maintained a mandate to monitor cantonment sites where forces are supposed to be disarmed and assembled. Each of these creates a concrete spatial question: is prohibited equipment present at a known location, or absent as required?
That question is, in principle, answerable from space. The challenge is not conceptual. It is a matter of resolution, revisit cadence, and the willingness to treat imagery as evidence rather than illustration.
The resolution floor for equipment identification
Identifying an object as a tank rather than a truck requires resolving features at roughly the 0.5 m scale or finer. Published NATO STANAG 3769 intelligence imagery interpretation standards define 'identification' of ground vehicles as requiring a ground-sampled distance of approximately 0.3 m. At 1 m resolution, you can confirm a vehicle is present and estimate its rough category. At 30–50 cm, a trained analyst can distinguish a main battle tank from a self-propelled howitzer by hull length, track width, and turret shape. At 15–30 cm, markings and antenna configurations become readable.
WorldView-3 and Pléiades Neo both sit at or below the 30 cm threshold in their native panchromatic bands. That is why they are the sensors of record for equipment-class identification in open-source treaty monitoring. Planet's Dove constellation, at 3–5 m resolution, cannot perform this function. It is useful for detecting the presence of a vehicle concentration, not for classifying what the vehicles are. Sentinel-1 at 5 m IW resolution sits in the same category: a useful alarm layer, not a classification tool.
Documented use by monitoring missions and open-source investigators
The UN Panel of Experts on Libya has cited commercial satellite imagery in multiple annual reports to the Security Council, using it to document the presence of armed vehicles in Tripoli and to track the movement of equipment between factions. The UN Mission in South Sudan used imagery to monitor cantonment sites and confirm whether forces had actually reported for assembly. The OSCE Special Monitoring Mission to Ukraine used both its own UAVs and commercially acquired imagery to document ceasefire violations along the contact line in Donbas before 2022.
Outside formal UN structures, organisations such as Bellingcat, the American Enterprise Institute's Critical Threats Project, and the Middlebury Institute's James Martin Center for Nonproliferation Studies have published imagery-based analyses of treaty compliance. The Nagorno-Karabakh conflict of 2020 produced a particularly dense record: Planet SkySat and Maxar imagery were used within hours of events to confirm or refute claims by both sides about equipment losses and territorial control. This is not a future capability. It has been operational practice for at least a decade.
The honest caveat: imagery analysis identifies what is visible at the moment of acquisition. Concealment under netting, inside buildings, or in forested terrain defeats optical sensors. Sentinel-1 SAR reduces but does not eliminate this problem, since dense forest canopy attenuates C-band returns and purpose-built radar-absorbent materials exist. No single pass proves sustained compliance; a time series does.
Legal and evidential standing
Commercial satellite imagery is not automatically admissible as legal evidence in international tribunals, but it has been accepted as supporting documentation in UN Security Council deliberations, ICC proceedings, and Human Rights Council commissions of inquiry. The key requirements are chain of custody (provenance metadata showing acquisition time, sensor, and processing level), analyst qualification, and corroboration with other sources.
Maxar, Planet, and Airbus all embed acquisition metadata in their imagery products, including UTC timestamp, sensor identifier, and ground-sample distance. This metadata is the starting point for any evidential use. A properly documented analytic report, citing imagery with verified metadata and cross-referenced against ground reporting or signals intelligence where available, has been accepted in international legal contexts. The International Criminal Court's Rules of Procedure and Evidence do not exclude remote sensing data; admissibility turns on reliability and relevance, both of which well-documented satellite analysis can satisfy.
One structural limitation: commercial imagery providers are private companies subject to the laws of their home jurisdictions. Shutter control provisions in US law (under the Land Remote Sensing Policy Act) allow the US government to restrict collection over certain areas in defined circumstances. Buyers relying on a single provider for treaty monitoring should be aware that access is not unconditional.
Building a monitoring architecture that holds up
Effective treaty monitoring from imagery requires three layers working together. First, a baseline: a complete, dated optical survey of the zone of interest before the agreement enters force, establishing the position of equipment, earthworks, and installations. Second, a change-detection cadence: regular revisits at a frequency matched to the likely pace of violation. For heavy-weapon withdrawal, weekly revisits may suffice; for active ceasefire lines, daily or sub-daily tasking is more appropriate. Third, an alert layer: a coarser, more frequent input (Sentinel-1 SAR, Planet Dove) that flags anomalies for follow-up tasking of high-resolution sensors.
Sentinel-1's 6-day repeat and free data access make it a practical foundation for the alert layer across large areas. High-resolution commercial tasking is then applied only where the alert layer signals a change, which controls cost without sacrificing coverage. This architecture is well established in the published literature on arms-control verification and has been described in detail in studies published in journals including Remote Sensing (MDPI).
Satellize structures analytics engagements around this layered approach, applying open-constellation data as the persistent monitoring base and adding commercial tasking on client licence where resolution requirements demand it. Clients receive a versioned, auditable record rather than ad hoc imagery pulls.
What imagery cannot do, and why that matters for buyers
Cloud cover is a persistent problem in some theatres. South Sudan's wet season, which runs roughly April to October, can produce sustained cloud cover that blocks optical collection for days at a time. SAR partially compensates, but at the resolution cost described above. Buyers should specify cloud-persistence statistics for their area of interest before designing a monitoring architecture.
Imagery answers presence-or-absence questions about physical objects. It does not answer questions about intent, command authority, or the identity of the actors controlling equipment. A tank in a buffer zone is a verifiable fact from a 30 cm image; whether it was placed there by a named faction under a named commander requires corroboration. Conflating the two is the most common analytical error in open-source treaty monitoring, and it has produced embarrassing corrections in the public record. The imagery is the evidence. The attribution is the analysis. They should be presented separately.
Typical figures
| Best available optical resolution | 30 cm (WorldView-3 panchromatic, Pléiades Neo panchromatic) |
| Equipment-class identification threshold | Approximately 30–50 cm GSD per NATO STANAG 3769 'identification' level |
| SAR resolution (Sentinel-1 IW) | 5 m range × 20 m azimuth; 250 km swath; C-band 5.405 GHz |
| Revisit cadence (high-resolution optical) | Sub-daily to 4.5 days depending on sensor and tasking priority; SkySat up to 15 passes per day over a point |
| Revisit cadence (Sentinel-1 SAR) | 6 days per orbit direction; 12-day exact repeat; free and open access |
| Minimum detectable object (optical) | Objects larger than approximately 2–3 times the GSD are reliably detectable; individual vehicles visible at 50 cm GSD |
| Archive depth | Maxar archive from 1999; Planet from 2014; Sentinel-1 from 2014; Pléiades from 2011 |
| Imagery latency (commercial tasking) | Typically 2–24 hours from acquisition to delivery for priority tasking; standard delivery 24–72 hours |
| Evidential metadata | UTC acquisition timestamp, sensor ID, GSD, orbit parameters embedded in product metadata by major providers |
| Cloud limitation | Optical sensors blocked by cloud; SAR unaffected. Persistent cloud cover in tropical or monsoon theatres can interrupt optical collection for days to weeks |
Analytics Satellize can run
| Buffer-zone occupation baseline | Object-based image analysis (OBIA) applied to high-resolution optical imagery to classify and count vehicles, structures, and earthworks within a defined exclusion polygon at agreement entry-into-force date | Georeferenced GIS layer with equipment counts, structure footprints, and annotated imagery report, versioned and timestamped for evidential use |
| Periodic compliance change-detection report | Bi-temporal or multi-temporal image differencing on Sentinel-1 SAR coherence and optical band composites to flag new objects, disturbed ground, or disappeared features within monitored zones | Weekly or fortnightly PDF report with flagged change polygons, confidence ratings, and recommended follow-up tasking areas |
| Equipment-class identification assessment | Analyst-led visual interpretation of 30 cm optical imagery against published vehicle silhouette libraries, cross-referenced with shadow-length and track-width measurements | Annotated image extract with equipment-type assessments, confidence levels, and explicit statement of ambiguities where resolution or viewing geometry limits certainty |
| Withdrawal-distance compliance measurement | Georeferenced vehicle position extraction from high-resolution imagery, with standoff distance calculated against the agreed ceasefire line geometry using GIS buffer analysis | GIS layer showing measured standoff distances per detected vehicle or equipment cluster, with pass/fail assessment against treaty-specified thresholds |
| SAR-based alert for large vehicle concentrations | Sentinel-1 IW backscatter anomaly detection using thresholded change in sigma-naught values consistent with metallic object concentrations, applied across wide-area exclusion zones | Automated alert feed (GeoJSON) identifying candidate anomaly locations for follow-up optical tasking; delivered within 12 hours of Sentinel-1 pass |
| Archival timeline reconstruction | Retrospective multi-date imagery stack from commercial and open archives to reconstruct equipment movements or fortification construction prior to a specified event date | Annotated timeline document with dated imagery extracts, suitable for inclusion in a commission of inquiry submission or UN panel 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.