Sensors
- HawkEye 360 cluster satellites: Three-satellite clusters flying in formation measure time-difference-of-arrival (TDOA) and frequency-difference-of-arrival (FDOA) simultaneously. Published geolocation accuracy from a single pass is approximately 10–20 km CEP for non-cooperative emitters in the 100 MHz–6 GHz range. Revisit over a fixed point varies by latitude but is typically several times per day across the full constellation.
- Spire LEMUR-2: LEMUR-2 carries a software-defined radio payload (STRATOS) capable of recording raw IQ data across configurable bands including the GSM-R uplink (876–880 MHz) and downlink (921–925 MHz). The constellation of over 100 satellites provides high revisit; individual pass duration over a ground segment is short (roughly 5–10 minutes), limiting dwell time per observation.
- Generic LEO TDOA/FDOA formation: Any multi-satellite cluster with precisely synchronised clocks and shared ephemeris can perform TDOA. Positional accuracy improves with wider cluster baseline and higher signal-to-noise ratio. Single-pass 10–20 km uncertainty is the honest published figure; repeated passes with FDOA fusion can narrow this, but sub-kilometre accuracy from LEO alone remains an active research problem, not a routine product.
- Ground-truth GSM-R monitoring receivers (reference layer): The European Railway Agency (ERA) and national infrastructure managers deploy fixed GSM-R monitoring units at selected mast sites. These provide high-accuracy frequency and power data but cover only instrumented locations. Space-based data is most useful as a cuing layer that directs these ground assets.
Why GSM-R interference is a safety problem, not just a nuisance
GSM-R (Global System for Mobile Communications, Railway) operates in a narrow paired band: 876–880 MHz uplink, 921–925 MHz downlink. It carries cab-to-signaller voice, European Train Control System (ETCS) data, and shunting radio. Degraded signal quality on any of these channels can trigger emergency brake applications or prevent drivers from receiving movement authorities. The International Union of Railways (UIC) has documented cases where adjacent LTE 800 MHz base stations, operating legally in the 791–821 MHz band, produced intermodulation products that fell inside the GSM-R passband.
The interference mechanism is not exotic. LTE signals at high power, mixed in a non-linear receiver front-end with a nearby GSM-R carrier, can generate third-order intermodulation products at frequencies that sit directly on GSM-R channels. Illegal transmitters and poorly suppressed industrial equipment add further sources. The problem is that a 1,000 km rail corridor may have dozens of interference hotspots, and finding them with a drive-test vehicle takes weeks.
What a satellite pass actually measures
A LEO RF monitoring satellite does not receive the GSM-R signal as a railway device would. It receives the upward-leaking energy from transmitters on the ground, including both legitimate GSM-R base stations and any co-channel or adjacent-channel interferers. By comparing signal power, frequency offset, and arrival timing across a cluster of satellites flying in formation, the system can estimate where on the ground a transmitter is located.
The physics sets hard limits. TDOA from a two- or three-satellite cluster at 500–600 km altitude, with a baseline of tens to low hundreds of kilometres between cluster members, produces a position hyperbola. Intersecting two hyperbolas from one pass gives a fix, but the geometry from LEO is shallow, and timing errors translate directly into positional uncertainty. Published figures from HawkEye 360 and the broader academic literature consistently cite 10–20 km CEP as a realistic single-pass result for emitters in this frequency range. That is enough to identify which section of a rail corridor is affected and which LTE cell tower or industrial site is the likely source. It is not enough to identify a specific mast without follow-up.
FDOA (frequency-difference-of-arrival) adds a second observable: the Doppler shift of the signal differs between satellites moving at slightly different velocities relative to the emitter. FDOA tightens the position estimate and helps discriminate between a stationary interferer and a moving one, which matters when trying to separate a fixed LTE base station from, say, an interfering transmitter on a freight vehicle.
National corridor survey in hours, not weeks
A ground survey team with a calibrated receiver and a GPS logger can measure interference with high precision, but they can only be in one place at a time. A national rail network of 10,000 km takes weeks to survey systematically, and intermittent interferers may not be active during the survey window.
A constellation with multiple daily passes over a given corridor can observe the entire network repeatedly. An interferer that is only active during peak LTE traffic hours (say, morning and evening commutes) will be caught if the satellite passes during those windows. The resulting dataset is a time-stamped map of detected emissions along the corridor, with approximate geolocations and signal characteristics. This is a fundamentally different product from a ground survey: lower positional precision, but national coverage and temporal depth. The two methods are complementary. Space-based survey identifies candidate hotspots; ground teams verify and resolve them.
Honest limits: what the data cannot tell you
The 10–20 km positional uncertainty is not a failure of the technology; it is a consequence of geometry and clock physics at LEO altitudes. Buyers should not expect a satellite pass to identify a specific LTE base station by location alone. The output is a corridor segment, perhaps 15–30 km long, that warrants ground investigation.
Cloud cover and weather are irrelevant for RF monitoring, which is an advantage over optical and SAR approaches. But urban RF environments are dense. A satellite pass over a major city will see dozens of signals in adjacent bands simultaneously, and separating a weak GSM-R interferer from a crowded spectrum requires careful signal processing and, often, multiple passes to build a statistical picture. Weak interferers below the satellite's noise floor will not be detected at all. Published detection thresholds depend on antenna gain, integration time, and the specific satellite design; they are not uniform across all LEO RF payloads.
Archive depth varies by operator. HawkEye 360 has been collecting data since 2018 and maintains a historical archive. Spire's STRATOS payload coverage depends on tasking history. For a retrospective investigation of when interference began on a specific corridor, archive availability should be confirmed before committing to an analysis approach.
From raw detections to an actionable interference map
Raw TDOA/FDOA fixes are points with uncertainty ellipses. Turning these into an actionable interference map requires several processing steps: clustering fixes by frequency and time to separate distinct emitters, correlating candidate locations with known LTE cell tower databases (publicly available from national regulators in most European countries), and flagging detections that fall outside expected GSM-R operational parameters.
The ERA has published technical guidance on GSM-R interference thresholds, including acceptable levels of adjacent-channel interference from LTE. These thresholds can be encoded as detection rules applied to the satellite-derived signal power estimates, producing a tiered output: confirmed exceedance, probable exceedance, and monitor. Satellize applies this kind of threshold-referenced classification in its analytics work, drawing on published ERA methodology to ensure outputs are interpretable by railway safety engineers rather than only by RF specialists.
A practical deliverable is a GIS layer updated on each constellation pass, showing corridor segments colour-coded by interference risk, with attached metadata on detected frequency, estimated power, and the satellite pass geometry that produced each fix. This can feed directly into a network operations dashboard or be exported for review by the national spectrum regulator.
Integration with regulatory and safety workflows
Railway spectrum management sits at the intersection of two regulatory worlds: the national communications regulator (who licences LTE operators) and the railway safety authority (who certifies signalling systems). Space-based interference data is most useful when it is formatted to speak to both. A fix that correlates spatially with a specific LTE operator's cell plan is evidence for a regulatory conversation. A fix that correlates temporally with a reported ETCS communication failure is evidence for a safety investigation.
The European Railway Agency's published interference studies, and the UIC's ongoing work on GSM-R protection, provide the methodological baseline for this kind of cross-domain analysis. National infrastructure managers in countries planning migration from GSM-R to Future Railway Mobile Communication System (FRMCS) will face a transition period during which both systems coexist. Monitoring the spectrum during that transition, at a national scale, is exactly the kind of problem where LEO RF surveillance has a practical advantage over any ground-based approach.
Typical figures
| Frequency coverage | GSM-R uplink 876–880 MHz, downlink 921–925 MHz; adjacent LTE 800 band 791–821 MHz / 832–862 MHz |
| Positional accuracy (single pass, TDOA) | 10–20 km CEP typical from LEO cluster; improves with multi-pass fusion and FDOA combination |
| Revisit rate | HawkEye 360: multiple passes per day at mid-latitudes; Spire LEMUR-2: high revisit, pass duration 5–10 min per ground segment |
| Detection latency | Hours to same-day for processed fix delivery; near-real-time tasking available on some commercial platforms |
| Minimum detectable signal | Payload-dependent; weak or intermittent emitters below satellite noise floor will not be detected; confirm thresholds with operator for specific use case |
| Coverage | Global; full national rail corridor observable within a single day's passes at European latitudes |
| Archive depth | HawkEye 360 archive from 2018; Spire STRATOS coverage depends on historical tasking; verify availability before retrospective analysis |
| Delivery formats | GeoJSON / KML interference fix layers, CSV event logs, dashboard-compatible feeds; ERA-threshold-referenced risk tiers |
Analytics Satellize can run
| Corridor interference heat map | TDOA/FDOA fix clustering aggregated across multiple passes; spatial binning along rail alignment | GIS layer (GeoJSON) updated per constellation pass, colour-coded by interference frequency and estimated severity |
| Candidate interferer identification | Spatial correlation of satellite fixes with national LTE cell tower registry and industrial facility locations | Ranked candidate list with location, operator attribution where possible, and confidence score per candidate |
| Temporal activity profile | Time-series analysis of detections per corridor segment across archive; peak-hour versus off-peak comparison | PDF report showing interference activity patterns by time of day and day of week, to guide ground survey scheduling |
| ERA-threshold exceedance alerts | Signal power estimates compared against ERA published adjacent-channel interference thresholds for GSM-R protection | Automated alert feed flagging segments where estimated interference exceeds ERA guidance, formatted for safety engineering review |
| Ground survey prioritisation layer | Bayesian ranking of corridor segments by detection frequency, estimated power, and proximity to known ETCS failure reports | Ordered survey schedule with segment coordinates and recommended measurement parameters for ground teams |
| FRMCS transition monitoring baseline | Repeated spectrum occupancy snapshots in GSM-R and adjacent bands before and during transition period | Quarterly spectrum occupancy report for regulatory submission, with trend analysis across the transition window |
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.