Pontoon and military bridging construction monitoring
Military pontoon bridges produce distinct physical signatures detectable by high-revisit optical and SAR sensors. This page explains the physics, the sensors, and the honest limits of remote detection.
Sensors
- Planet Dove (PlanetScope): 3–4 m native resolution, daily global revisit. In practice a single pass over a river valley may come once per day per orbital plane; cloud is the dominant gap. Sufficient to resolve individual pontoon sections and vehicle tracks at bridgeheads.
- Sentinel-1 SAR (C-band, ESA): IW mode delivers 10 m ground range resolution with 6-day repeat at mid-latitudes (12 days for a single satellite). Penetrates cloud and operates at night. Coherence change between two passes reveals new surface structures on water and disturbed soil on banks. The 2022 Dnipro monitoring relied heavily on this dataset.
- ICEYE SAR (X-band): Stripmap mode at roughly 3 m resolution; spotlight mode can reach sub-metre. Tasking allows sub-daily revisit over a nominated area. X-band is more sensitive than C-band to small metallic structures such as pontoon floats, but also more affected by surface roughness from wind on water.
- Maxar WorldView-3: 0.31 m panchromatic resolution, 16-band multispectral including SWIR. Resolves individual pontoon sections and vehicle types. Revisit over a fixed point is roughly 1–4 days depending on latitude and tasking priority. Cloud and tasking cost are the constraints.
What a pontoon bridge looks like from 500 km up
A military pontoon bridge is, from a sensor's point of view, a narrow linear feature of high radar backscatter or elevated reflectance crossing a surface that normally returns almost nothing. Water at low wind speeds is a near-perfect specular reflector in both optical and radar bands: it bounces energy away from the sensor. Place a row of metal pontoon floats and a decking superstructure across it and you introduce double-bounce radar returns and a sudden bright line in optical imagery. The contrast ratio between open water and a bridging set is large enough to be detectable even at Sentinel-1's 10 m resolution.
The secondary signature is often more diagnostic than the bridge itself. Vehicle tracks on soft riverbank soil are disturbed ground: lower coherence in SAR repeat-pass pairs, and spectrally distinct from undisturbed vegetation in multispectral imagery. A bridgehead that has seen dozens of armoured vehicles will show track networks spreading hundreds of metres inland. Those tracks persist after the bridge is removed, which matters for post-event reconstruction.
Cloud over river valleys: the problem that will not go away
River valleys concentrate moisture. The Dnipro, the Dniester, the Euphrates, the Irrawaddy all sit in terrain that generates persistent low cloud and fog, particularly in autumn and winter. A daily optical constellation like Planet Dove provides excellent temporal coverage in theory; in practice, a river crossing attempt timed to coincide with overcast conditions can remain invisible to optical sensors for 24 to 72 hours. That is long enough to establish a bridgehead.
SAR does not care about cloud. Sentinel-1's C-band signal passes through cloud cover with negligible attenuation. The trade-off is resolution and revisit: a 6-day repeat cycle on a single Sentinel-1 satellite means a bridge assembled and dismantled within that window could be missed entirely in coherence analysis, though the backscatter anomaly may still appear in individual acquisitions. ICEYE and other commercial SAR operators fill the gap with tasked collections, but tasking requires knowing roughly where to look. Fusing optical and SAR into a single alert pipeline, with SAR as the primary source and optical as confirmation, is the operationally sound approach.
The 2022 Dnipro crossings and what analysts actually saw
The attempted Russian river crossings near Bilohorivka in May 2022 became a widely cited demonstration of open-source satellite monitoring. Planet imagery showed pontoon sections being assembled and subsequently destroyed within a single day in some cases. Sentinel-1 coherence change maps, computed by comparing phase information between repeat passes, showed loss of coherence at the water-bank boundary consistent with new surface structures and vehicle disturbance. The combination of sources allowed analysts to reconstruct the timing and location of multiple crossing attempts from open data alone.
The episode established a practical benchmark: a pontoon bridge of battalion-tactical-group scale, using standard Russian PMP or PP-91 bridging equipment, produces a signature detectable at 10 m SAR resolution within one repeat cycle. It also demonstrated the limits. Several attempted crossings that occurred under cloud were reconstructed only retrospectively from bank disturbance, not from direct bridge observation. Real-time warning requires SAR tasking, not just archive analysis.
Honest limits of the method
Detection is not identification. A SAR bright-line anomaly across a river could be a fish weir, a low-head dam, floating debris, or a civilian pontoon ferry. Disambiguation requires either sufficient resolution to resolve the structural form (sub-3 m SAR or optical) or corroborating context: vehicle tracks, nearby force concentrations, upstream logistics activity. Single-source detections should be flagged as candidates, not confirmed bridges.
Narrow rivers below roughly 80–100 m width create ambiguity at Sentinel-1 resolution because the bridge occupies fewer than ten pixels across its span. ICEYE spotlight mode or WorldView-3 resolves this, but at cost and with cloud risk. Dismantling a bridge takes roughly the same time as assembling one: a crossing completed and the bridge removed before the next SAR pass leaves only the bank disturbance as evidence. Finally, inflatable or low-profile bridging equipment has a smaller radar cross-section than steel pontoon sets and may fall below detection thresholds in C-band imagery at 10 m resolution.
Building a monitoring pipeline for a contested river line
A practical pipeline for a government or command authority monitoring a specific river line combines three layers. First, a persistent SAR coherence baseline: Sentinel-1 pairs computed on every available repeat cycle establish what normal looks like at every point along the river. Deviations above a threshold trigger a candidate alert. Second, tasked high-resolution SAR from ICEYE or a comparable operator provides confirmation imagery within hours of a candidate alert, independent of cloud. Third, optical collection from Planet or a tasked WorldView-3 pass provides the visual confirmation and vehicle-count data that SAR alone cannot supply.
Archive depth matters. Sentinel-1 data runs back to 2014 for many regions, and Planet's archive extends to 2016 in useful density. A pre-conflict baseline built from years of imagery makes anomaly detection far more reliable than a baseline constructed after hostilities begin. Satellize builds and maintains these baselines as part of its analytics work; the same coherence-change methodology that underpins the Tonga crop-estimation programme applies, at a different physical scale, to detecting surface-structure change at water boundaries. Analysts wanting to trial the pipeline on a specific river reach can request a scoped baseline assessment.
Typical figures
| Best optical resolution (WorldView-3) | 0.31 m panchromatic |
| Optical revisit (Planet Dove) | Daily global; single-point cloud-free revisit variable, typically 1–3 days |
| SAR resolution (Sentinel-1 IW) | 10 m ground range |
| SAR resolution (ICEYE spotlight) | Sub-1 m achievable; 3 m in standard stripmap |
| Sentinel-1 repeat cycle | 6 days (two-satellite constellation); 12 days single satellite |
| ICEYE tasking latency | Typically under 24 hours from tasking request to delivery |
| Minimum detectable bridge width (Sentinel-1) | Approximately 10 m structure on rivers wider than ~80 m; narrower rivers require higher-resolution SAR |
| Archive depth | Sentinel-1 from 2014; Planet from 2016; Landsat from 1972 for historical baseline |
| Spectral bands (WorldView-3) | 8 VNIR bands, 8 SWIR bands, panchromatic |
| SAR frequency (Sentinel-1 / ICEYE) | C-band 5.4 GHz / X-band 9.65 GHz |
Analytics Satellize can run
| SAR coherence change alert | Interferometric coherence differencing between Sentinel-1 repeat-pass pairs; coherence loss at water-bank boundary flagged against pre-conflict baseline | Automated alert with candidate location, confidence score, and coherence difference image; delivered as GeoTIFF and JSON feed |
| Bridge presence confirmation layer | Backscatter anomaly detection: bright linear feature across water body in single-pass SAR or optical imagery, cross-checked against baseline water-surface reflectance | GIS polygon layer with timestamp, sensor, and classification (confirmed / candidate / false positive); updated per new acquisition |
| Bridgehead track-network mapping | Multispectral change detection using normalised difference indices on Planet or WorldView imagery; disturbed soil classified against pre-event vegetation baseline | Vector track-network layer with estimated vehicle-pass intensity; included in weekly river-line monitoring report |
| Crossing event timeline reconstruction | Multi-source temporal stack analysis combining all available SAR and optical acquisitions; events sequenced by first-detection and last-detection timestamps | PDF event timeline with annotated imagery and sensor provenance, suitable for intelligence briefing |
| River-line persistent monitoring dashboard | Automated ingestion of Sentinel-1 open-access acquisitions plus tasked commercial SAR; anomaly scoring per river segment updated on each new pass | Web dashboard with per-segment status, alert history, and raw imagery access; API available for integration into existing C2 systems |
| Historical baseline report | Multi-year Sentinel-1 and Landsat archive analysis to establish seasonal coherence norms and identify pre-existing infrastructure at all river crossings in an area of interest | Static PDF baseline report with annotated maps; delivered before operational monitoring begins |
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.