Semi-submersible and narco-submarine route detection
Low-profile vessels used for cocaine transshipment leave near-zero radar cross-section and carry no AIS, but Kelvin wake signatures in high-resolution SAR imagery betray their passage. Detection is probabilistic; this page covers the physics, the published methods, and the honest limits.
Sensors
- Sentinel-1 SAR (C-band, ESA): 5 x 20 m resolution in Interferometric Wide Swath mode, 250 km swath, 6-12 day revisit at the equator. Free archive back to 2014. C-band backscatter detects surface roughness changes in wakes but is susceptible to wave clutter in sea states above Beaufort 4-5.
- ICEYE SAR (X-band): Spotlight mode delivers approximately 0.5 m resolution; Strip mode around 3 m. Revisit to a specific ocean corridor can be reduced to hours through tasking. X-band is more sensitive to fine wake turbulence than C-band but has a narrower swath, typically 5-15 km in Spotlight.
- Capella Space SAR (X-band): Spotlight imagery to approximately 0.5 m, with tasking latency of a few hours. Comparable to ICEYE for wake detection; the two constellations together improve revisit probability on a given corridor.
- Airbus Pléiades Neo (optical, VIS/NIR): 30 cm panchromatic resolution. Useful for confirmation when cloud is absent: a semi-submersible at the surface produces a visible hull silhouette and a turbulent wake in calm conditions. Not useful at night or under cloud, which is a significant operational constraint in the Inter-Tropical Convergence Zone.
What a Kelvin wake gives away
Every vessel moving through water generates a Kelvin wake. The half-angle of that wake is approximately 19.47 degrees regardless of vessel speed, a consequence of the dispersion relation for deep-water gravity waves first described by Lord Kelvin in 1887. What varies with speed and hull form is the intensity of the turbulent wake, the diverging wave arms, and the narrow-V stern wash. For a low-profile vessel (LPV) or semi-submersible running at 6-10 knots with most of its hull below the waterline, the surface disturbance is small but not zero.
In synthetic aperture radar imagery, a wake appears as a bright turbulent centreline (increased backscatter from foam and bubbles) flanked by darker bands where the diverging waves suppress Bragg-scale roughness. The contrast between wake and background depends heavily on sea state. In calm conditions, even a small LPV can produce a detectable signature in high-resolution X-band imagery. In a 2-3 metre swell, that same signature is buried in clutter. This is the fundamental physical limit: space-based wake detection of LPVs is a fair-weather capability, and the Eastern Pacific ITCZ is not reliably fair.
What the published research actually shows
Peer-reviewed work published in Remote Sensing (MDPI) and related journals has demonstrated wake detection for conventional vessels in Sentinel-1 data using Radon-transform and Hough-transform approaches, which convert the linear wake structure into a detectable peak in parameter space. Detection rates for large commercial vessels in moderate sea states run at 70-90 percent in published evaluations. For small, low-freeboard targets the numbers drop considerably. Several studies report that the minimum detectable vessel length in Sentinel-1 IW mode is in the range of 15-20 m under favourable conditions, which covers the larger semi-submersibles (typically 10-25 m) but not the smallest narco-submarines.
ICEYE and Capella Spotlight imagery changes the calculus. At sub-metre resolution, the turbulent centreline of a wake from a vessel as small as 10 m becomes distinguishable from background texture in sea states up to approximately Beaufort 3. Published ICEYE technical documentation confirms sub-metre Spotlight capability. No public study has yet claimed reliable LPV detection statistics from commercial X-band constellations at operational scale, which is an honest gap in the literature rather than a failure of the physics.
How JIATF-South uses space data: the public record
The US Joint Interagency Task Force South, based in Key West, is the publicly documented coordination body for counter-narcotics operations in the Eastern Pacific and Caribbean. Congressional testimony and US Government Accountability Office reports describe a layered detection architecture: wide-area surveillance (including satellite cues and maritime patrol aircraft radar) narrows the search area, and then P-8 Poseidon or P-3 Orion aircraft with high-resolution sensors confirm and track. Satellite SAR is documented as one input to this cueing process, not the terminal sensor.
The GAO has noted in published reports that JIATF-South intercepts only a fraction of estimated cocaine flows, partly because detection resources are finite and partly because LPVs are specifically designed to defeat radar. The publicly stated strategy is probabilistic: raise the cost and risk of each transit rather than guarantee interdiction. Space-based wake detection fits that strategy as a wide-area screening tool that cues scarcer airborne assets, not as a standalone evidence source.
False positives are the operational problem
Ocean clutter produces linear features that mimic wakes. Ship wakes from legitimate vessels decay over hours; a satellite pass may image the residual wake of a fishing boat that has since moved 50 km away. Internal waves, current boundaries, and rain cells all generate linear SAR signatures. Published false-positive rates for automated wake detectors on open ocean imagery range from 20 to over 50 percent depending on sea state and the detection threshold chosen.
Multi-pass confirmation is the standard mitigation. If a wake signature appears in a Sentinel-1 pass and a tasked ICEYE pass two hours later shows a consistent heading and position consistent with a 7-knot vessel, confidence rises substantially. Without that second pass, a single wake detection is an alert, not a finding. Any operational system that presents single-pass detections as confirmed LPV tracks is overstating what the data supports.
Practical detection architecture for an enforcement client
A workable space-based screening system for a known trafficking corridor combines three elements. First, systematic Sentinel-1 coverage of the corridor at its 6-12 day free revisit, processed with an automated wake detector to flag candidate tracks. Second, commercial tasking of ICEYE or Capella on flagged areas within the next available pass window, typically 4-12 hours depending on orbital geometry and tasking queue. Third, optical confirmation with Pléiades Neo if cloud permits and if the estimated vessel position at the time of the optical pass is within the sensor's tasking range.
The output is a probability-ranked list of candidate tracks, not a confirmed vessel list. Feeding that list into a maritime operations centre for cross-referencing with AIS gaps, known departure ports, and historical route data is where the detection value compounds. Satellize runs exactly this kind of multi-source correlation workflow; the underlying method is the same class of analysis used in the Tonga crop-estimation programme, applied to a very different domain. If you are building or procuring a maritime surveillance capability, the architecture question to resolve first is not which satellite to use but what decision the satellite data needs to support and at what latency.
Honest limits, stated plainly
Cloud cover across the Eastern Pacific ITCZ averages above 70 percent, which eliminates optical confirmation for most passes. SAR penetrates cloud but cannot image through heavy rain cells, which also cause significant clutter. Revisit on a specific 100 km corridor segment with free Sentinel-1 data is roughly twice per week at best; commercial tasking raises that to once or twice per day at cost. A semi-submersible transiting at 7 knots covers approximately 168 nautical miles per day, meaning a 12-hour detection gap translates to an 84-nautical-mile position uncertainty ellipse.
Sub-metre SAR is not yet proven at operational scale for LPV detection; the physics supports it, the published case studies do not yet close the loop on false-positive rates in real operational conditions. Any vendor claiming reliable, automated, space-only LPV detection at acceptable false-positive rates should be asked for the validation dataset.
Typical figures
| SAR spatial resolution (Sentinel-1 IW) | 5 x 20 m (range x azimuth) |
| SAR spatial resolution (ICEYE/Capella Spotlight) | ~0.5 m |
| Optical resolution (Pléiades Neo) | 30 cm panchromatic |
| Sentinel-1 revisit (equatorial corridor) | 6-12 days per pass direction; combined ascending/descending ~3-6 days |
| Commercial SAR tasking revisit | 4-12 hours to first available pass on a given target area (ICEYE/Capella) |
| Minimum detectable vessel (Sentinel-1 IW, favourable conditions) | ~15-20 m length; smaller vessels require sub-metre SAR |
| Operational sea-state limit for wake detection | Approximately Beaufort 3-4; performance degrades above Beaufort 4 |
| Sentinel-1 archive depth | 2014 to present (free via Copernicus Data Space) |
| SAR frequency bands in use | C-band (5.4 GHz, Sentinel-1); X-band (~9.6 GHz, ICEYE, Capella) |
| Delivery formats | GeoTIFF, GeoJSON track candidates, PDF alert report, API feed |
Analytics Satellize can run
| Automated wake candidate detection | Radon/Hough transform on calibrated SAR backscatter; linear-feature extraction against local clutter baseline | GeoJSON layer of candidate wake centrelines with heading, estimated speed, and confidence score; refreshed per satellite pass |
| Multi-pass track correlation | Kinematic consistency check: candidate positions from successive passes tested against plausible speed/heading envelope for LPV class | Ranked track list with probability score and supporting imagery chips; delivered as PDF alert or API push within 2 hours of second pass |
| Corridor baseline and anomaly flagging | Historical Sentinel-1 archive processed to establish expected wake density per corridor segment; new passes compared against baseline | Monthly corridor activity report with anomaly dates and locations; GIS layer of elevated-activity zones |
| AIS dark-vessel cross-reference | Candidate wake positions cross-referenced against AIS vessel positions at pass time; residual candidates (no AIS match within 5 km) flagged as dark | Filtered alert list of AIS-dark wake candidates; integrated into maritime operations centre feed |
| Optical confirmation tasking recommendation | Forward-projection of estimated vessel position at next Pléiades Neo pass time; cloud-cover probability from NWP model used to score confirmation likelihood | Tasking recommendation with predicted position box and cloud-risk score; issued within 30 minutes of SAR alert |
| Sea-state suitability forecast | Integration of ECMWF wave-height and wind-speed forecasts with published detection-limit curves to predict detection probability windows over a corridor | 72-hour detection-probability map per corridor segment; updated daily |
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.