Inland waterway barge traffic monitoring from SAR and optical
Most inland barges are invisible to AIS networks. SAR and optical satellites can detect and count them, but low freeboard, river clutter and sun-glint geometry impose real limits that any serious monitoring programme must account for.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground range resolution in Interferometric Wide Swath mode, 250 km swath, 6-day repeat at the equator (effectively 1-3 days over Europe with both satellites). C-band backscatter detects vessels as bright point targets against darker water, but low-freeboard barges at wind speeds above roughly 5-7 m/s can fall below the clutter threshold.
- Planet SuperDove (optical, Planet Labs): 3 m native resolution, 8 spectral bands (coastal blue through NIR), near-daily revisit over most river corridors. Useful for counting barges and estimating convoy length in clear conditions, but sun-glint on river surfaces at low solar angles can saturate pixels over the vessel footprint, and 3 m is marginal for freeboard measurement.
- Spire Global spaceborne AIS: Constellation of over 100 LEO cubesats receiving VHF AIS messages globally. Provides ground-truth for vessels that do carry transponders, and exposes coverage gaps where inland fleets are legally exempt from AIS carriage. Useful as a complement to imagery, not a replacement.
- Sentinel-2 MSI (optical, ESA): 10 m resolution in visible and NIR bands, 5-day revisit with both satellites. Coarser than Planet for individual vessel detection but free, archived since 2015, and useful for river-width and water-level context that affects barge routing and draft.
Why inland fleets are largely invisible to conventional tracking
Maritime AIS carriage is mandatory for ocean-going vessels above 300 gross tonnes and all passenger ships. Inland waterway regulations are fragmented. On the Rhine, CESNI's IS EURS standard has progressively extended AIS to inland vessels since 2015, but enforcement varies and smaller push-tow operators on the Mississippi, Yangtze and Paraná remain outside any mandatory reporting regime. A loaded hopper barge on the Mississippi can carry 1,500 tonnes of grain with no electronic signature whatsoever.
The consequence for trade-flow analysis is significant. Commodity flows on inland waterways, particularly coal, grain and aggregates, are often inferred from port declarations rather than direct observation. Those declarations lag by weeks. Satellite detection offers a near-real-time alternative, with caveats.
What a flat hull gives away on radar, and when it does not
SAR vessel detection relies on the contrast between a vessel's metallic superstructure, which returns strong backscatter, and the relatively smooth water around it. Ocean-going ships present high superstructures and substantial radar cross-sections. A loaded grain barge is a different problem: freeboard can be as little as 30-50 cm when fully laden, and the flat steel deck presents a weak double-bounce return compared with a ship's hull-superstructure corner reflectors.
At wind speeds above roughly 5-7 m/s, Beaufort 3-4, C-band backscatter from wave capillary action on a river surface can match or exceed the return from a low-freeboard barge. Published studies using Sentinel-1 over the Rhine have demonstrated detection probabilities above 80% for convoys longer than 100 m in calm conditions, dropping sharply in moderate wind. Single dumb barges under 60 m are frequently missed entirely. This is not a processing failure; it is a physical limit of C-band at this target class.
X-band SAR (3 cm wavelength, used by some commercial constellations) improves contrast slightly but is not freely available at Sentinel-1's revisit rate or swath width. The honest position is that SAR alone cannot guarantee detection of all laden barges on a busy river in anything other than near-calm conditions.
Optical detection: sun-glint, shadows and the 5-metre resolution floor
Planet SuperDove imagery at 3 m resolution can resolve individual barges in a convoy, which typically range from 60 m to 185 m in length and 9 m to 23 m in beam. Counting barges and measuring convoy length is feasible in good light. Identifying vessel type from hull geometry is possible for the larger classes.
Two optical problems recur on river corridors. First, sun-glint: at low solar elevation angles, specular reflection from the river surface saturates pixels in the visible bands, obscuring vessels near the centre of the glint patch. River orientation relative to the sun azimuth determines which reaches are affected on any given overpass, and it changes daily. Second, cloud cover over river valleys in temperate and tropical zones is persistent; the Rhine and Yangtze both average more than 60% cloud cover in winter months.
Draft estimation from freeboard measurement deserves specific attention because it is commercially attractive: a deeper draft implies a heavier load, which implies commodity volume. At 3 m resolution, the freeboard of a standard European barge (roughly 1.5-2.5 m above waterline when laden) subtends fewer than two pixels. Reliable freeboard measurement from optical imagery requires sub-metre resolution and very precise water-level reference data. Published literature consistently places the reliable threshold at 5 m resolution or better, and even then uncertainty is large enough to make tonnage estimates speculative rather than operational.
Fusing SAR, optical and AIS to build a traffic index
No single sensor resolves the problem. The practical approach used in published hydrological and trade-flow studies is to treat the three data streams as complementary evidence. Spire AIS data identifies vessels that do carry transponders, anchoring the count and providing speed and heading vectors. SAR detections, run through a CFAR (constant false alarm rate) detector tuned to the expected radar cross-section range of inland vessels, add candidates in the AIS gaps. Optical imagery, when cloud-free, provides the highest-confidence count and allows convoy geometry to be measured.
The output is a traffic index rather than a precise vessel census. On a well-observed corridor like the Rhine between Rotterdam and Duisburg, where Sentinel-1 passes are frequent and AIS coverage is relatively good, a daily barge-passage count with perhaps 15-25% uncertainty is achievable. On the upper Mississippi or the Yangtze above Wuhan, where AIS penetration is lower and cloud cover is higher, weekly aggregation is more honest than daily claims.
Satellize applies this fusion approach on open Sentinel and commercial Planet tasking, with Spire AIS as the transponder layer, calibrating outputs against lock-passage records where those are publicly available.
Practical limits buyers should price into any programme
Detection probability for single laden barges in moderate wind: 40-60% with Sentinel-1 C-band alone. Convoy detection (three or more barges) in the same conditions: closer to 75-85%, because the aggregated radar cross-section is larger and easier to separate from clutter.
Cloud persistence is the dominant availability constraint for optical data, not satellite revisit. A 3-day cloud event over the Rhine in November produces a gap that SAR partially fills but cannot fully substitute for, because SAR cannot confirm vessel type or measure freeboard.
Archive depth is an asset. Sentinel-1 data runs back to 2014 for European corridors, Sentinel-2 to 2015. Planet's archive over major waterways extends to roughly 2016-2017 depending on corridor. Multi-year archives allow seasonal traffic pattern analysis and anomaly detection, which is often more valuable to a commodity trader or infrastructure planner than a single-day count.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 10 m ground range, 20 m azimuth (published ESA specification) |
| Optical spatial resolution (Planet SuperDove) | 3 m native; orthorectified to ~3.7 m in standard delivery |
| SAR revisit (Sentinel-1, European corridors) | 1-3 days with both satellites; 6 days at equator |
| Optical revisit (Planet SuperDove) | Near-daily over most river corridors; cloud-limited availability |
| Minimum detectable convoy length (SAR, calm conditions) | ~60-80 m (single barge marginal; two-barge convoy more reliable) |
| Reliable freeboard measurement threshold | Sub-5 m resolution required; 3 m optical insufficient for operational tonnage estimates |
| SAR frequency / optical bands | C-band 5.4 GHz (Sentinel-1); coastal blue to NIR 8 bands (SuperDove) |
| Archive depth | Sentinel-1 from 2014, Sentinel-2 from 2015, Planet from ~2016-2017 on major corridors |
| AIS complement (Spire Global) | VHF 161.975 / 162.025 MHz; global LEO reception, latency typically under 90 minutes |
| Typical traffic-index uncertainty | 15-25% on well-observed European corridors; higher on Yangtze and Mississippi upper reaches |
Analytics Satellize can run
| Daily barge-passage count by river segment | CFAR SAR vessel detection on Sentinel-1 IW scenes, fused with Spire AIS transponder layer to avoid double-counting | GIS polyline layer with passage counts per segment per overpass, aggregated to daily and weekly CSV |
| Convoy geometry measurement | Object detection on Planet SuperDove 3 m imagery; convoy length and barge count extracted from cloud-free overpasses | Per-convoy attribute table (length, barge count, position, timestamp) delivered as GeoJSON |
| Traffic anomaly alert | Statistical deviation from rolling 30-day baseline passage count; triggers when count falls or rises beyond two standard deviations | Email or webhook alert with supporting SAR and optical thumbnails |
| Seasonal traffic index (multi-year) | Time-series aggregation over Sentinel-1 archive from 2014; normalised for cloud and wind-speed conditions using ERA5 reanalysis wind data | Annual report with monthly traffic index charts per corridor, exportable to Excel |
| AIS coverage gap mapping | Comparison of SAR-detected vessel positions with Spire AIS message density; gaps indicate non-reporting fleet segments | Heatmap layer showing AIS dark zones overlaid on river network, updated quarterly |
| Water-level and routing context layer | Sentinel-2 water extent mapping combined with published gauge records (e.g. German BfG, USACE) to flag low-water routing constraints | Weekly GIS layer showing navigable width and gauge-correlated low-water alerts per reach |
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.