Illicit river-route vessel activity detection
Small vessels on drug-transshipment rivers run without AIS and often at night. SAR imaging and VIIRS night-light analysis reveal their presence anyway, using radar cross-section and thermal-light anomalies on open water reaches.
Sensors
- Sentinel-1 SAR (C-band, ESA): 5 by 20 metre resolution in Interferometric Wide Swath mode; 250 km swath; 6-day repeat at the equator, improving at higher latitudes. Penetrates cloud and operates day and night. Detects vessel-scale radar cross-sections on open river reaches; small fibreglass or wooden boats typically register between 1 and 10 square metres RCS, which is near the detection floor in IW mode but resolvable with CFAR processing.
- Capella Space SAR (X-band, commercial): Spotlight mode delivers sub-0.5 metre resolution, sufficient to resolve hull geometry and estimate vessel length. On-demand tasking with latency of a few hours. X-band is more sensitive to small metallic targets than C-band, improving detection of outboard-motor housings and fuel drums on open decks.
- VIIRS Day/Night Band (NOAA/NASA, Suomi NPP and NOAA-20): Panchromatic low-light channel, 750 metre nominal resolution. Detects radiance from vessel lights, lanterns and engine-exhaust glow at night. Published work using VIIRS DNB has identified fishing-vessel clusters on the Mekong; the same detection logic applies to motorised cargo canoes on the Amazon and its tributaries. Single nightly overpass; cloud cover is a hard limit.
- Planet SkySat (optical, commercial): 50 cm resolution in collect mode; tasked on demand. Useful for daytime confirmation of vessel presence and wake pattern analysis on wide river reaches. No cloud penetration; limited utility under Amazonian convective cloud. Best used as a follow-up sensor after SAR or VIIRS cues a location of interest.
- Sentinel-2 MSI (optical, ESA): 10 metre resolution in visible and near-infrared bands; 5-day revisit with both satellites. Useful for mapping open-water channel geometry, identifying cleared riverbank staging areas and detecting turbidity plumes from propeller wash on shallow reaches. Not a vessel-detection sensor in its own right at this scale, but essential for contextual mapping.
Why rivers are harder than open sea
Maritime dark-vessel detection on the open ocean benefits from wide, unobstructed radar swaths and a relatively simple background. Rivers are the opposite. Narrow channels, braided meanders, overhanging canopy, and radar shadow from forested banks all reduce the effective detection area in any single SAR pass. A dugout canoe with a 40-horsepower outboard motor presents a radar cross-section that, in C-band IW mode, sits close to the noise floor. Detection requires careful constant false alarm rate (CFAR) processing tuned to the local clutter statistics of the water surface, not the generic ocean thresholds used in maritime AIS-correlation workflows.
The Mekong and Amazon basins have both been the subject of published research using Sentinel-1 and VIIRS DNB precisely because the enforcement problem is acute and the sensor physics are tractable. The Amazon's várzea floodplain creates seasonal variation in channel width and surface roughness that must be accounted for in any persistent monitoring scheme. Dry-season low water concentrates traffic on fewer navigable channels, which paradoxically makes detection easier: the population of vessels is funnelled into a smaller search area.
What a floating radar cross-section gives away
SAR backscatter from a vessel depends on its material, geometry and orientation relative to the sensor look angle. Metal fuel drums, outboard-motor casings and corrugated roofing used as makeshift cargo covers all produce strong corner-reflector returns. A wooden hull with no metal fittings is genuinely difficult at C-band; Capella's X-band spotlight mode, with sub-0.5 metre resolution, can resolve individual structural features that create dihedral reflections.
Wake detection is a secondary method. A moving vessel produces a Kelvin wake and a turbulent wake; both alter the local normalised radar cross-section of the water surface. Synthetic aperture radar can image these wakes even when the vessel itself is below the detection threshold, provided the river reach is wide enough and the vessel speed is above roughly 3 to 5 knots. On narrow tributaries under 100 metres wide, the wake is often clipped by the bank return, making this method unreliable. Honest assessment: SAR vessel detection on rivers narrower than about 150 metres is difficult and produces significant false-negative rates with current open-constellation imagery.
Night light as a proxy for fleet activity
VIIRS DNB does not resolve individual vessels at 750 metres. What it does resolve is the aggregate radiance of a cluster: several boats moored together at a transshipment point, or a convoy moving through a dark river reach, can produce a detectable radiance anomaly against the near-zero background of an unlit forest river. The VIIRS Nightfire product, developed at the Colorado School of Mines and distributed through the Earth Observation Group, was designed for gas flare detection but uses the same DNB and M-band data that inform vessel-light analysis.
The practical workflow is to establish a radiance baseline for each river segment across cloud-free nights over a reference period, then flag exceedances. A single anomalous night is noise. A pattern of anomalies on the same segment on multiple nights, correlated with known seasonal trafficking calendars, is a signal worth acting on. Cloud is the dominant operational limit: the Amazon basin averages fewer than five cloud-free VIIRS overpasses per month over many reaches during the wet season. This is not a system that provides nightly coverage; it provides probabilistic accumulation of evidence.
Canopy as cover: the partial solution
Traffickers use overhanging forest canopy deliberately. A vessel moored under a 20-metre forest overhang is invisible to optical sensors and largely invisible to SAR, which cannot penetrate a closed forest canopy at C or X-band. L-band SAR (as used by ALOS-2 or the forthcoming NISAR mission) offers some canopy penetration, but not enough to reliably detect a vessel-scale target beneath a dense tropical forest.
The practical detection window is the open-water reach between canopy sections. Vessels must transit these exposed stretches. Persistent monitoring of the open reaches, combined with mapping of the canopy-covered sections as probable staging areas, creates a detection logic based on transit events rather than static presence. This is a meaningful constraint on what the system can claim: it detects movement, not concealment. A vessel that never moves from under canopy during a sensor overpass will not appear in the data.
Building an evidence layer, not an arrest warrant
Satellite data on its own does not identify a vessel, its crew or its cargo. What it produces is a spatial and temporal pattern of anomalous activity: a reach with statistically elevated vessel detections, night-light anomalies on specific dates, turbidity signatures suggesting frequent propeller disturbance of shallow sediment, and cleared riverbank areas that may serve as loading points. Each of these is a layer in a cumulative evidence picture.
Published enforcement programmes in the Amazon, including those supported by Brazilian federal agencies, have used exactly this layered approach: SAR-derived vessel counts on major tributaries, VIIRS anomaly calendars, and optical imagery of bank infrastructure. The satellite record does not replace human intelligence or in-situ interdiction capacity. It tells enforcement planners where to concentrate finite patrol resources, and it provides a documented historical record that is admissible as contextual evidence in some jurisdictions.
Satellize's analytics pipeline can run this kind of persistent river-reach monitoring on open Sentinel-1 and VIIRS data, with commercial SAR tasking added for targeted high-resolution confirmation.
Operational limits worth stating plainly
Sentinel-1's 6-day revisit means a vessel can complete a transshipment run in the gap between passes. Commercial SAR constellations reduce this, but tasking costs scale with frequency. VIIRS provides nightly coverage in principle but cloud cover in tropical basins removes most of it in practice. No current open-access constellation provides the combination of fine resolution, cloud penetration and sub-daily revisit that would make river-route interdiction a near-real-time capability. What is achievable is pattern-of-life analysis over weeks and months, which is genuinely useful for strategic enforcement planning even if it cannot support tactical interception on a given night.
Typical figures
| SAR spatial resolution (Sentinel-1 IW) | 5 × 20 m (range × azimuth) |
| SAR spatial resolution (Capella spotlight) | < 0.5 m |
| Sentinel-1 revisit (equatorial) | 6 days (single satellite); 12-day with one satellite |
| VIIRS DNB resolution | 750 m nominal |
| VIIRS overpass frequency | Once nightly per satellite; cloud cover is the dominant gap |
| Minimum detectable vessel (SAR, open water) | ~5 m hull length with metal fittings in C-band IW; smaller targets resolvable at X-band spotlight |
| Minimum detectable river width (SAR vessel detection) | ~150 m for reliable open-water detection; narrower reaches subject to bank clutter |
| Cloud penetration | SAR: full; VIIRS DNB and optical: none |
| Archive depth (Sentinel-1) | From 2014; freely accessible via Copernicus Data Space |
| VIIRS archive depth | From 2012 (Suomi NPP); NOAA-20 from 2018 |
Analytics Satellize can run
| Vessel density map by river reach | CFAR (constant false alarm rate) target detection on Sentinel-1 IW SAR, adapted to river-surface clutter statistics | GIS polygon layer with vessel-count time series per reach segment, updated per available pass |
| Night-light anomaly calendar | VIIRS DNB radiance baseline modelling; exceedance flagging per river segment per cloud-free overpass | Monthly anomaly report with flagged dates, reach coordinates and radiance magnitude |
| High-resolution vessel confirmation | Capella Space X-band spotlight tasking triggered by SAR or VIIRS cue; hull-geometry estimation | Tasked image with annotated vessel detections and estimated length class, delivered within agreed latency window |
| Riverbank staging-area identification | Sentinel-2 multispectral change detection on cleared bank areas; turbidity plume mapping in NIR band | GIS point layer of candidate staging locations with optical image chips and change date |
| Pattern-of-life activity score | Multi-sensor fusion of SAR vessel counts, VIIRS anomalies and optical change detections; temporal clustering to identify persistent vs. episodic activity | Ranked reach-priority list for enforcement resource allocation, updated monthly |
| Seasonal channel navigability model | Sentinel-1 water-surface extent mapping across wet and dry seasons; channel-width time series per reach | Seasonal navigability GIS layer informing patrol planning and detection-window scheduling |
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.