Maritime sensing is an observability problem: AIS, EO, SAR and RF
In brief
- AIS provides reported identity and movement; optical imagery, radar and radio-frequency sensing provide different checks on vessel presence and activity.
- Observations must be matched in time and space, with uncertainty and shared data sources accounted for.
- A useful system explains what was observed, which vessel it may belong to and what remains unresolved.
A radar image shows a vessel where the tracking map shows none. That could be a vessel without a matching AIS report, an incorrectly located detection, or a mismatch between the image time and the tracking data.
Adding an optical image may help identify the object. It may also introduce another problem: the image was acquired an hour later, after the vessel moved.
This is an observability problem: which properties of the maritime environment can we determine from the measurements available? Sensors provide evidence about position, identity and activity, but those are separate questions. Establishing that a vessel is present does not establish its name or the purpose of its voyage.
I would model the vessel and the observation process together: what may be happening at sea, and what each sensor could have recorded under those conditions.
This note explains the main data sources, the limits of combining them and a proposed approach to monitoring smaller vessels. The design proposals are distinct from the capabilities documented in the references.
What each data source measures
Earth observation (EO) includes several forms of remote sensing. Here, optical imagery means visible and near-infrared images, while synthetic aperture radar (SAR) is discussed separately. Radio-frequency (RF) sensing refers here to detecting and locating transmissions; AIS itself also uses radio.
| Source | What it measures or reports | What it can help establish | Main interpretation limit |
|---|---|---|---|
| Automatic Identification System (AIS) | broadcast identity, position, course and speed | reported vessel identity and movement history | messages may be missing or inaccurate |
| Optical imagery | light reflected from the scene | visible objects, vessel configuration and local context | cloud, illumination and image resolution limit interpretation |
| SAR | radar energy reflected from the surface | possible vessels and surface features, including through cloud and at night | a radar target does not carry a vessel identity |
| RF sensing | detectable radio emissions | the presence and possible location of a transmitter | the emitter must be associated with a vessel |
| Onboard sensors | local position, motion or equipment measurements | conditions experienced by an equipped vessel | access depends on installation, calibration and data delivery |
These sources can complement one another, but none supplies the complete operational account.
AIS: a detailed report with a reception history
AIS broadcasts structured vessel information over maritime radio channels. Shore stations and satellites can receive those messages. The resulting track depends on both the vessel’s broadcasts and the receiving system (US Coast Guard AIS overview).
Equipment classes also differ. Class A and Class B devices have different reporting capabilities; Class B includes carrier-sense and self-organising variants. An analysis should account for the installed equipment rather than assume that all vessels produce equally frequent observations (US Coast Guard types of AIS).
I would check three things before interpreting an AIS track:
- Reception: was the feed likely to receive reports in that area and period?
- Identity: do the identifier and vessel details remain consistent across reports and external records?
- Movement: are the timestamps, positions and implied speeds physically plausible?
A gap in a data feed does not establish that a vessel stopped transmitting. A continuous track does not establish that every reported position is correct. These are different failure modes and need separate checks.
Dataset coverage matters beyond individual alerts. If the observation system captures some vessel classes or regions more reliably than others, counts and activity estimates may reflect that selection. Comparisons should report coverage alongside the behaviour being measured.
Optical imagery: useful detail when the scene is visible
Optical images can help distinguish vessel shapes, configurations and surrounding activity. The European Maritime Safety Agency (EMSA) describes applications including vessel detection, rendezvous inspection and monitoring fixed fishing assets. What can be interpreted depends on resolution and viewing conditions; cloud, poor weather and darkness restrict visible-light observations (EMSA maritime surveillance service).
A clear image may help test whether an AIS identity is compatible with the vessel’s apparent size or type. It usually cannot establish the vessel’s purpose from appearance alone. Two vessels alongside one another, for example, provide evidence of a meeting; the image must contain sufficient detail to support a claim about what was exchanged.
For a proposed monitoring system, I would record usable image coverage separately from nominal coverage. An acquisition obscured by cloud should not count as an opportunity in which the detector confidently found no vessel.
SAR: imaging through cloud is not continuous tracking
SAR transmits radar pulses and forms an image from the returning signals. It can operate at night and through cloud. Vessels and platforms can appear as bright targets against the sea surface, while surface conditions influence the radar response (EMSA SAR explanation).
A radar detection is a candidate object, not an automatic identity assignment. Its interpretation should account for the sensor configuration, surrounding structures and conditions at acquisition. A missed target may reflect detection limits rather than absence.
The distinction between weather capability and temporal coverage is important. A satellite image covers a particular place at a particular time. The next usable acquisition may occur after the vessel has moved well beyond that area. “All-weather” should therefore never be used as a substitute for a measured revisit interval or demonstrated detection performance.
RF sensing: evidence about an emitter
RF sensing can help locate a source of radio transmissions. Its usefulness depends on which signals a system can receive and how accurately it can estimate their origin.
One example is AIS position validation. Spire describes using Doppler shift and timing information from received AIS signals to estimate the transmitter’s location separately from the coordinates encoded in the message. This is a provider-described capability, rather than a guarantee of performance in every setting (Spire AIS Position Validation).
A disagreement between the two locations warrants investigation. It does not, by itself, distinguish deliberate manipulation from a timing, equipment or processing error.
The physical source of the evidence also matters. An AIS position and an RF location estimate derived from that same transmission contain different information, but they are not wholly independent observations. Treating them as independent can overstate confidence. Conversely, a system that observes no relevant emission cannot conclude that no vessel is present.
Coverage and latency are different quantities
Orbit affects where and when a satellite can observe. Low Earth orbit (LEO) supports many imaging missions, but an individual satellite moves relative to the surface. Medium Earth orbit (MEO) is used by navigation systems, while geostationary satellites remain above the same equatorial longitude and support broad regional communications and weather observation (ESA types of orbits).
For maritime monitoring, the operational questions are more specific than the orbit label:
| Quantity | Question to ask |
|---|---|
| Acquisition opportunity | When could the sensor observe the area? |
| Usable coverage | Was an observation actually acquired and suitable for the task? |
| Revisit | How long until another usable observation of the area? |
| Delivery latency | How long after acquisition did the data reach the analyst? |
| Alert latency | How long after acquisition was the result processed and available? |
A quickly delivered image can still be an infrequent snapshot. A constellation with frequent passes can still leave gaps if acquisitions are not scheduled or usable. I would evaluate these quantities for the intended region and service rather than infer them from satellite count alone.
Combining observations requires a time-aware match
Consider an illustrative sequence: an AIS report at 10:01, an RF detection at 10:05, a SAR image at 10:14 and an optical image at 10:37. These measurements do not describe the same instant.
At 12 knots, a vessel travels about 4.8 km in 13 minutes. Matching the SAR target directly to the 10:01 AIS point could therefore select the wrong vessel. A motion estimate must first account for the elapsed time, with a wider range of plausible positions when manoeuvres or reporting gaps make prediction uncertain.
I would assess each association using the observation time, possible vessel positions, location error, estimated size and alternative matches. In crowded water, the output may remain ambiguous. The system should allow a detection to stay unmatched instead of assigning the nearest available identity.
Combining sources also requires checking how they can fail together. Two products may reuse the same AIS feed, or two detections may come from the same image. These should not be counted as separate confirmations. Shared gaps and processing dependencies belong in the evidence record.
A successful combination can strengthen a specific conclusion: a physical object is present, its dimensions fit a candidate vessel, and its position is compatible with that vessel’s recent track. Intent or cargo activity may still require operational records or additional observations.
Small-vessel monitoring adds a communication problem
Remote sensing can observe a vessel without equipment installed onboard. A local monitoring device offers different information: position history, motion and, where sensors are installed, engine or equipment condition. Getting those measurements ashore introduces power, storage and connectivity constraints.
For a proposed small-vessel device, I would start with satellite navigation, local storage and only the sensors needed for a defined task. Near-shore cellular service could carry bulk records; a supported satellite service could carry selected offshore messages. Suitability would need to be checked against coverage, power demand, antenna placement and operating cost.
I would begin with an explicit transmission policy:
| Situation | Proposed device behaviour |
|---|---|
| Routine offshore operation | store detailed measurements locally and send scheduled position and status summaries |
| Priority event | attempt prompt transmission of the event, time, position and essential context; record acknowledgement or failure |
| Limited power or poor connection | preserve priority data, report degraded operation when possible and retry within defined limits |
| Suitable bulk connection | upload buffered records and check for missing or duplicated data |
Local processing can reduce data volume, but it also introduces the risk of suppressing an important event. Safety-related rules need explicit delivery targets, fallback behaviour and testing under link and power failures. An optimisation objective cannot guarantee that critical information is never lost.
Adaptive transmission could later balance energy, message cost and delay. It should first be compared with a simple fixed policy, using measured delivery reliability and record completeness.
References
- US Coast Guard Navigation Center. Automatic Identification System (AIS) Overview.
- US Coast Guard Navigation Center. Types of Automatic Identification Systems.
- European Maritime Safety Agency. Copernicus Maritime Surveillance Service. Service information sheet.
- European Maritime Safety Agency. How does SAR detection work?
- Spire. AIS Position Validation. Provider description of RF-based position validation.
- European Space Agency. Types of orbits.