Inland Waterway and River Vessel Traffic Monitoring
On the Amazon, Congo, Mekong and Yangtze, most vessels carry no AIS. Very-high-resolution optical and SAR satellites can still count, classify and track them, turning vessel traffic into a proxy for commodity flows, illegal logging supply chains and humanitarian access.
Sensors
- Maxar WorldView Legion: 30 cm native panchromatic resolution, revisit up to 15 times per day over a target site in mid-latitudes, somewhat less at equatorial latitudes. Resolves vessel type, deck configuration and visible waterline, making load-state estimation feasible on vessels longer than roughly 20 metres.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, four-satellite constellation with same-day revisit over priority sites. Tri-stereo tasking allows 3-D reconstruction of vessel superstructures, which aids classification of high-sided barges versus flat-deck timber carriers.
- ESA Sentinel-1 SAR (C-band): IW mode delivers 10 m ground range resolution across a 250 km swath, with a 6-day repeat at the equator (12-day per satellite). Penetrates cloud and rain, which is decisive during tropical wet seasons. River bends and steep banks cause layover and shadow artefacts that can obscure vessels entirely; this is a hard limit, not a processing problem.
- Planet SuperDove: 3 m multispectral resolution, 8 spectral bands, near-daily global revisit. Resolves vessels longer than roughly 15 metres and provides spectral context for distinguishing vessel hull colours and deck cargo type, though it cannot match the classification detail of 30 cm systems.
Why rivers are harder than oceans
Open-ocean AIS coverage, even with its well-documented gaps, gives analysts a starting framework. On the Amazon, the Congo or the upper Mekong, that framework barely exists. The International Maritime Organisation mandates AIS only for vessels over 300 gross tonnes on international voyages; the timber barge, the artisanal fishing pirogue and the fuel-laden river tanker operating domestically fall outside that requirement entirely. Surveys of Amazonian river traffic suggest the vast majority of commercial movements are AIS-dark by default, not by intent.
The physical environment compounds the problem. Tropical river systems spend months under near-continuous cloud cover. The Amazon basin averages cloud cover exceeding 80 percent on any given day across much of the year. Optical satellites, however capable, simply cannot see through it. SAR is the only consistent sensor during those periods, and it introduces its own geometry problems that any honest analyst must account for.
What a floating roof gives away
Very-high-resolution optical imagery at 30 cm can resolve vessel class, approximate length, deck cargo configuration and, in some cases, load state from visible freeboard. A laden timber barge sits low and shows a narrow strip of hull above the waterline; the same barge running empty rides high enough that the hull colour band is clearly visible. Pléiades Neo's multispectral bands add discrimination between fresh-cut timber (bright reflectance in near-infrared), bagged agricultural produce and containerised cargo. These are not certainties, they are probabilistic classifications that improve with analyst ground-truth and repeated observation of the same routes.
Planet SuperDove's near-daily cadence at 3 m resolution fills the temporal gap between expensive VHR tasking events. It cannot confirm vessel type reliably for anything under about 15 metres, and it will miss small pirogues entirely. The practical workflow pairs SuperDove for daily counting and route mapping with periodic Pléiades Neo or WorldView Legion tasking for classification confirmation at key chokepoints such as river confluences, logging concession boundaries or border crossings.
SAR in the wet season: what it sees and what it hides
Sentinel-1 C-band SAR detects vessels as bright radar cross-section returns against the relatively low-backscatter surface of calm river water. On straight reaches with calm water, even a 20-metre wooden vessel can produce a detectable return above background noise. Published detection studies on rivers using Sentinel-1 report minimum detectable vessel lengths in the range of 15 to 25 metres under calm-water conditions, though this degrades significantly when wind roughens the surface or when vessels are moored against vegetated banks whose backscatter masks the hull return.
River bends are where SAR fails most visibly. The side-looking geometry of SAR means that steep riverbanks on the inside of a bend can shadow the water surface behind them, creating a zone of no return that looks like open water but contains no usable information. On the outside of the same bend, layover can compress the bank and adjacent water into a single bright feature, burying any vessel return within it. These gaps are predictable from the SAR acquisition geometry and the river's digital elevation model, so analysts can map the blind zones explicitly rather than treating the absence of a detection as the absence of a vessel. Optical data, when cloud permits, must fill those gaps.
Vessel traffic as a commodity signal
The analytic value of inland waterway monitoring is rarely the individual vessel. It is the pattern. A sustained increase in barge traffic on a particular Amazonian tributary during the dry season, correlated with a known logging concession boundary upstream, is a materially different signal from seasonal agricultural transport. Time-series vessel counts at fixed transects, derived from a combination of SAR detections and VHR optical snapshots, can be compared against legal export records, concession maps and deforestation alerts from systems such as GLAD at the University of Maryland.
Humanitarian applications follow a similar logic. River access is often the only supply route into remote communities during flood seasons. Vessel traffic monitoring at key river junctions can serve as an early indicator of supply-chain disruption before ground reports arrive. The Congo River basin, where road infrastructure is extremely limited, is a case where satellite-derived vessel counts have genuine operational value for aid coordination, though the 6-day Sentinel-1 revisit and persistent cloud cover mean the picture is always somewhat lagged.
Honest limits of the current sensor stack
No current satellite system reliably detects vessels shorter than 10 to 15 metres on inland waterways. That excludes a large share of artisanal and subsistence traffic, which may be precisely what matters in some humanitarian or fisheries contexts. Night-time optical coverage is limited to moonlit scenes or thermal infrared, which at the resolution of available public systems does not discriminate vessel types on rivers. SAR at night works fine geometrically but inherits all the layover and shadow problems already described.
Archive depth is a genuine asset. Sentinel-1 data is publicly available back to 2014, and Landsat archives extend to the 1970s for coarser-resolution context. Building a multi-year baseline of vessel traffic patterns is feasible and often reveals seasonal rhythms that make anomalies statistically detectable. Satellize applies this kind of time-series approach in its analytics work, including the crop-estimation programme for the Kingdom of Tonga, where the principle of building a statistical baseline before declaring an anomaly is the same, even if the sensor stack differs.
Building a monitoring programme that holds up
A credible inland waterway monitoring programme starts with route definition: which tributaries, which chokepoints, which seasonal windows matter most. SAR coverage via Sentinel-1 should be the backbone, tasked systematically at every available pass over priority reaches. Commercial VHR tasking from Pléiades Neo or WorldView Legion should be reserved for classification events at chokepoints, not blanket coverage, because the cost of saturating a river system with 30 cm imagery is prohibitive and unnecessary.
Change detection is more useful than absolute counts in most client contexts. A buyer interested in illegal logging supply chains does not need to know that 47 barges transited a confluence on a given day; they need to know that traffic is 60 percent above the three-year seasonal average and concentrated in a reach adjacent to a newly flagged deforestation polygon. That is a specific, actionable intelligence product. The sensor data is the raw material; the analytic framework built on top of it is where the actual insight lives.
Typical figures
| Best optical resolution (VHR) | 30 cm panchromatic (WorldView Legion, Pléiades Neo) |
| Medium-resolution optical | 3 m multispectral (Planet SuperDove, near-daily revisit) |
| SAR resolution (Sentinel-1 IW mode) | 10 m ground range, 250 km swath |
| SAR revisit (Sentinel-1, single satellite) | 12 days at equator; 6 days with both satellites |
| VHR optical revisit | Up to 15 passes/day (WorldView Legion over mid-latitude sites); same-day (Pléiades Neo constellation) |
| Minimum detectable vessel length (SAR, calm water) | Approximately 15 to 25 m (published range; degrades near vegetated banks) |
| Minimum detectable vessel length (VHR optical) | Approximately 8 to 10 m at 30 cm resolution under good illumination |
| Cloud penetration | SAR only; optical sensors blocked during tropical wet season (cloud cover >80% on many basin days) |
| Sentinel-1 archive depth | From 2014 to present (publicly accessible) |
| Minimum river width for reliable SAR detection | Approximately 80 m (narrower channels risk bank-return contamination) |
Analytics Satellize can run
| Vessel count time series at fixed transects | Automated SAR ship detection (CFAR thresholding on sigma-nought) combined with optical cross-validation at cloud-free epochs | Monthly GIS layer with transect-level vessel counts and seasonal anomaly flags |
| Vessel classification by type and load state | VHR optical object detection and morphological classification; freeboard estimation from shadow length and multispectral hull-band analysis | Per-vessel attribute table (estimated type, length, load state) appended to detection shapefile |
| SAR blind-zone map for a target river reach | Geometric simulation of layover and shadow from SAR acquisition parameters and river DEM; output flags detection-unreliable pixels | Raster mask delivered alongside each SAR-derived detection layer, with recommended optical tasking windows to fill gaps |
| Traffic anomaly alert for suspected illegal logging corridors | Three-year seasonal baseline from Sentinel-1 archive; z-score anomaly detection at weekly cadence; spatial join to concession and deforestation alert polygons (GLAD alerts) | Weekly alert report with anomaly score, affected reach coordinates and supporting imagery thumbnails |
| Humanitarian access corridor status | Binary passability classification per river segment based on vessel presence and absence over rolling 14-day SAR composite | Bi-weekly corridor status map in GeoJSON, compatible with common humanitarian GIS platforms |
| Multi-year vessel traffic trend analysis | Long-term time-series regression on SAR detection counts normalised for acquisition geometry and seasonal water level variation | Annual trend report with confidence intervals and breakpoint identification for significant traffic-pattern changes |
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.