Shipborne radar type classification from spectral signatures
Space-based RF receivers can classify the type of radar fitted to a vessel by measuring pulse repetition frequency, bandwidth, and carrier frequency from LEO, providing an independent check on AIS-declared vessel type.
Sensors
- HawkEye 360 cluster satellites: Three-satellite formation flying in LEO at roughly 575 km altitude; measures time-difference and frequency-difference of arrival across the cluster to geolocate and characterise RF emitters. Frequency coverage spans approximately 144 MHz to 15 GHz, capturing both X-band (9–10 GHz) and S-band (2–4 GHz) marine radars. Revisit over a given ocean area is typically a few hours with the current constellation, improving as additional satellites are added.
- Spire LEMUR-2: Spire's LEMUR-2 satellites carry software-defined radio payloads capable of recording raw IQ data across a configurable frequency range. The constellation exceeds 100 satellites, giving sub-hourly revisit globally. Published use cases include AIS collection and RF monitoring; the SDR architecture allows post-processing for pulse parameter extraction when tasked appropriately.
- Generic space-based ELINT receivers (published literature context): IEEE GRSL and related peer-reviewed literature describes space-based electronic intelligence receivers that extract pulse repetition interval (PRI), pulse width, carrier frequency, and intra-pulse modulation from shipborne radars. Detection sensitivity depends on antenna gain, integration time, and the vessel's radar peak power, which for commercial marine X-band units typically runs 2–25 kW.
What the signal actually carries
Every shipborne radar transmits a characteristic set of parameters that are, in effect, a fingerprint of its design. The three most diagnostic are carrier frequency, pulse repetition frequency (PRF), and intra-pulse modulation. X-band navigation radars operate between 9.2 and 9.5 GHz and use PRFs typically in the 500–3000 Hz range with short pulses of 50–1000 nanoseconds. S-band radars, common on larger vessels requiring longer range, sit at 2.9–3.1 GHz with somewhat lower PRFs. Pulse-compression fishing sonars and scientific echosounders use linear frequency-modulated chirps with bandwidths that can reach tens of kilohertz, a waveform shape that is essentially absent from standard navigation sets.
From LEO, a receiver intercepts these pulses directly. The observable parameters are the carrier frequency (measured via the received centre frequency corrected for Doppler), the PRF (derived from the inter-pulse interval histogram), and the pulse envelope shape (indicating whether compression is in use). Together these three measurements place a radar into a relatively small number of categories: conventional short-pulse X-band, conventional S-band, dual-antenna dual-band, or pulse-compression sonar. That categorisation maps, imperfectly but usefully, onto vessel class and intended activity.
From waveform to vessel class: the processing chain
The signal processing chain has four stages. First, the receiver records raw IQ samples or, in less capable systems, detected power versus time. Second, a pulse detection algorithm identifies individual pulses above the noise floor and measures their arrival times, widths, and amplitudes. Third, a de-interleaving step separates pulses from different emitters when multiple vessels are present in the receiver's field of view, which at LEO altitudes can span hundreds of kilometres. Fourth, the extracted parameter set is compared against a library of known radar models.
De-interleaving is the hardest step. Published methods use PRI histograms, carrier frequency clustering, and, where available, TDOA constraints from multi-satellite formations to assign pulses to individual sources. HawkEye 360's three-satellite cluster geometry is specifically suited to this: the formation baseline allows TDOA and FDOA measurements that both geolocate the emitter and provide an independent frequency reference, reducing ambiguity in the de-interleaving. Even so, when two vessels of similar type are within roughly 20–30 km of each other, pulse streams can be difficult to separate cleanly.
Radar type as a proxy for vessel activity
A vessel fishing with active sonar will almost always have its echosounder running. A tanker on a standard passage will run a single X-band navigation radar, possibly supplemented by S-band in poor visibility. A vessel equipped with a high-PRF, narrow-beam X-band set and simultaneous S-band is more likely a warship or coast-guard vessel than a bulk carrier. These associations are probabilistic, not deterministic, but they are grounded in the economics of marine electronics: operators fit the sensors their activity demands.
The value is in cross-checking AIS declarations. A vessel broadcasting AIS type 'cargo' but emitting a waveform consistent with a fishing sonar suite is worth a second look. Equally, a vessel with no AIS at all but a radar signature matching a small purse-seiner class narrows the search space considerably. This is not a conviction; it is a cue for tasking optical or SAR assets, or for alerting a coast-guard patrol.
Where the method breaks down
Honest accounting of the limits matters here. Several commercial marine radar manufacturers produce X-band sets with nearly identical PRF and pulse-width ranges, meaning the classification can place a radar in the right family without identifying the specific model. Multi-radar vessels, which carry two or three independent sets for redundancy or regulatory compliance, produce overlapping pulse streams that stress de-interleaving algorithms even with formation-satellite geometry.
Vessels can also simply switch their radars off. A fishing vessel in well-known grounds may navigate by GPS alone and run no radar at all, producing no RF signature. Radar silence is itself informative in some contexts, but it eliminates the classification signal entirely. Atmospheric ducting can occasionally propagate X-band signals far beyond the geometric horizon, causing a distant vessel's radar to appear at an unexpected location in the receiver's data; this is rare but not negligible in tropical maritime environments. Finally, the public radar-model libraries used for matching are incomplete: newer integrated bridge systems from Asian manufacturers are under-represented in open literature.
What a classified radar signature delivers operationally
For a maritime authority running a vessel monitoring programme, radar-type classification adds a layer that AIS alone cannot provide. A fleet of vessels whose AIS declares 'fishing' but whose RF signatures are consistent with navigation-only X-band sets may in fact be transiting, not fishing, which matters for quota monitoring. Conversely, a vessel whose AIS is off but whose radar signature matches the pulse-compression sonar profile of a mid-water trawler is a plausible illegal fishing candidate in a protected zone.
The output is typically a per-vessel record combining geolocation (from TDOA/FDOA), radar type classification with a confidence score, and timestamp, delivered as a structured data feed or as an alert layer over a maritime picture. Satellize can integrate this RF classification layer with optical and SAR tasking pipelines, a workflow relevant to any maritime-domain-awareness programme that needs to move from signal detection to vessel identification. The analytical approach mirrors the signal-parameter extraction methods published in IEEE Geoscience and Remote Sensing Letters, adapted for the specific constellation geometries available under client licence.
Typical figures
| Frequency coverage (HawkEye 360) | Approximately 144 MHz to 15 GHz, covering S-band (2.9–3.1 GHz) and X-band (9.2–9.5 GHz) marine radars |
| Geolocation accuracy (TDOA/FDOA, published) | Typically 1–5 km CEP from three-satellite cluster geometry at LEO altitudes; degrades with poor satellite geometry or low SNR |
| Revisit (HawkEye 360 constellation) | Several hours per ocean area with current constellation; improves as further cluster pairs are launched |
| Revisit (Spire LEMUR-2, 100+ satellites) | Sub-hourly globally for AIS; RF monitoring revisit depends on tasking configuration |
| Minimum detectable radar peak power | Commercial marine X-band sets typically 2–25 kW; detection sensitivity depends on receiver gain and integration time; low-power sets at range margins may be missed |
| PRF measurement range | 500–3000 Hz for conventional X-band; lower for S-band; pulse-compression sets identified by chirp bandwidth rather than PRF alone |
| Classification output confidence | High for single-radar vessels with well-characterised waveforms; reduced for multi-radar vessels or when de-interleaving is ambiguous |
| Latency (detection to delivery) | Dependent on downlink schedule and processing pipeline; near-real-time possible with direct-to-cloud downlink architectures |
| Deliverable formats | Structured JSON or CSV event feed, GeoJSON alert layer, PDF vessel-record report |
Analytics Satellize can run
| Radar type classification per vessel contact | PRI histogram analysis and carrier frequency clustering against published marine radar parameter libraries (IEEE GRSL methods) | Per-contact record: geolocation, radar family, confidence score, timestamp; delivered as structured data feed |
| AIS-versus-RF vessel type discrepancy flag | Cross-referencing AIS-declared vessel type against RF-derived radar class; statistical thresholding on mismatch probability | Alert layer in maritime picture; daily digest report of high-confidence discrepancies |
| Dark vessel radar activity detection | RF contact detection in absence of AIS transmission; geolocation via TDOA/FDOA from multi-satellite formation | GeoJSON alert with estimated position, radar type, and confidence interval; triggers optional SAR or optical tasking |
| Pulse-compression sonar activity mapping | Intra-pulse chirp detection distinguishing fishing sonar from navigation radar; spatial clustering to identify active fishing grounds | Heatmap layer of sonar-active contacts over defined ocean area; periodic summary report |
| Multi-radar vessel flagging | De-interleaving of simultaneous pulse streams from a single geolocation; presence of dual-band or high-PRF sets flagged as potential non-commercial vessel indicator | Annotated vessel record with radar suite description; included in maritime-domain-awareness feed |
| Historical RF activity timeline per zone | Archive replay of collected RF contacts over a defined maritime zone; trend analysis on radar-type distribution over time | Time-series chart and CSV export; supports pattern-of-life analysis for enforcement planning |
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.