RF activity signature monitoring near nuclear facilities
Space-based RF sensing can characterise emission patterns around nuclear sites, flagging state changes that optical and SAR imagery alone may miss. Ambiguity is real; multi-source corroboration is mandatory.
Sensors
- HawkEye 360 cluster satellites: Three-satellite formation flying in LEO at roughly 575 km altitude. Uses time-difference and frequency-difference of arrival (TDOA/FDOA) to geolocate emitters. Published geolocation accuracy is approximately 1 km CEP for cooperative signals and degrades for low-power or intermittent emitters. Covers VHF and UHF bands. Revisit at a given latitude is irregular, typically multiple passes per day globally.
- Spire LEMUR-2 RF payload: Spire's LEMUR-2 cubesats carry AIS and GNSS-RO payloads as primary instruments, with some units carrying additional RF monitoring capability. The constellation numbers over 100 satellites, giving sub-hourly average revisit globally, though dwell time per pass over a fixed point is short (minutes). Useful for detecting presence and frequency of active emitters rather than precise geolocation.
- Sentinel-1 SAR (corroborating layer): C-band SAR at 5.405 GHz, 6-day repeat at the equator with the two-satellite constellation. Resolves structures down to 5 m in Stripmap mode. Does not detect RF emissions directly, but provides physical site-change context: new antenna arrays, vehicle concentrations, construction activity. Essential for disambiguating RF signature changes.
- Commercial optical constellations (corroborating layer): Platforms such as Maxar WorldView and Planet SkySat provide sub-metre to 3 m resolution imagery for visual confirmation of antenna installations, vehicle presence, and facility expansion. Cloud cover is the primary constraint; persistent cloud over mountainous or coastal nuclear sites can block optical collection for days at a time.
What RF emissions around a nuclear facility actually reveal
A nuclear facility in operation generates a characteristic RF environment. Power plant control systems use licensed VHF and UHF voice communications for shift coordination. Security perimeters run push-to-talk radio networks. Radar systems, where present, emit on predictable frequencies. Telemetry links from instrumentation can appear in monitored bands. None of this is secret in principle; the diagnostic value lies in change.
When a facility shifts from cold standby to active operations, or when an undeclared site begins construction and staffing, the RF environment changes measurably. New emitter types appear. Traffic density on existing frequencies increases, particularly during shift-change windows. Unusual operating hours, such as sustained high-traffic periods at 02:00 local time, can indicate emergency operations or covert activity. The Middlebury Institute of International Studies has published open-source methodology applying exactly this logic to North Korean and Iranian facilities, combining RF inference with commercial imagery.
How space-based RF collection works, and where it struggles
Satellite RF collection in this context relies on passive interception of signals that propagate to LEO. HawkEye 360's formation-flying cluster computes emitter position by measuring the tiny differences in signal arrival time and Doppler shift across the three spacecraft. The published 1 km CEP accuracy figure applies to relatively strong, sustained emitters. A handheld radio used briefly inside a reinforced building may not be detectable at all.
The fundamental constraint is dwell time. A LEO satellite passes over a fixed point in roughly 5 to 10 minutes. An emitter that is silent during those minutes leaves no record. This is not a flaw unique to nuclear monitoring; it applies to every RF collection mission from orbit. The practical mitigation is statistical accumulation: many passes over many days build a baseline, and departures from that baseline become visible even if no single pass captures a complete picture.
Frequency coverage matters too. HawkEye 360 publicly covers VHF and UHF. Military and government communications at nuclear facilities may use frequency-hopping spread-spectrum waveforms, encrypted links, or fibre-connected internal networks that produce no detectable RF at all. Space-based RF collection sees the perimeter and the logistics layer, not the classified core.
Attribution is the hard problem
Detecting an emitter at a location is not the same as attributing that emitter to a specific facility function. A VHF burst near a declared enrichment plant could be a security guard, a maintenance contractor, or a farmer on adjacent land. TDOA/FDOA geolocation at 1 km CEP is sufficient to associate an emitter with a facility footprint but not to assign it to a specific building or function.
IAEA safeguards literature is explicit on this point: RF indicators are considered corroborating evidence, not standalone proof of a safeguards violation. The analytical standard in published OSINT methodology requires at least two independent sensor types showing consistent anomalies before an inference is reported with any confidence. In practice this means RF change plus optical change plus, where available, SAR-detected physical alteration. Any single-source finding should be treated as a cue for further collection, not a conclusion.
Building a baseline and detecting departures from it
The analytical workflow begins with baseline characterisation: systematic collection of RF activity at a site over a period of weeks or months, cataloguing emitter frequencies, typical power levels, daily and weekly traffic rhythms, and any seasonal patterns. HawkEye 360's archive, which extends back to commercial operations beginning around 2019, provides historical data for sites that were of interest before a client engagement begins.
Change detection then operates against that baseline. Statistically significant increases in emitter count, new frequency bands appearing, or traffic patterns inconsistent with declared operational status all generate alerts for human review. The word 'significant' carries real weight here: given the noise in any RF dataset collected from LEO, a single anomalous pass is rarely meaningful. A pattern across ten or more passes over two to three weeks is more defensible.
Satellize structures this kind of multi-source baseline work for government clients. The approach is similar in logic to the crop-estimation programme run for the Kingdom of Tonga, where the value came from systematic time-series analysis rather than any single observation.
Integrating RF with optical and SAR: what each layer adds
Optical imagery at sub-metre resolution can confirm or deny physical changes that would explain an RF signature shift. New antenna masts, additional vehicle compounds, expanded perimeter fencing, or cooling system activity all appear in optical and SAR data. Sentinel-1's 6-day C-band repeat provides a consistent structural record at no cost, suitable for monitoring dozens of sites simultaneously. Commercial tasking fills the gap when higher resolution or shorter revisit is needed.
The combination is more diagnostic than any layer alone. An RF anomaly with no corresponding physical change might indicate a temporary operational shift or a sensor artefact. An RF anomaly coinciding with new construction visible in SAR is a materially stronger indicator. This layered approach is standard in published OSINT methodology and is consistent with IAEA complementary access procedures, which explicitly draw on open-source information to focus inspector attention.
Honest limits buyers should understand before commissioning work
Several constraints are worth stating plainly. Cloud cover does not affect RF collection but does affect the optical corroboration layer, sometimes for extended periods over high-latitude or tropical sites. Frequency-hopping and spread-spectrum communications used by military facilities may be undetectable by commercial LEO sensors. Geolocation accuracy degrades for low-power emitters, short transmissions, and sites at low elevation angles relative to the satellite formation.
Archive depth for commercial RF data is limited compared to optical archives. Landsat optical data goes back to 1972; commercial RF collection from LEO at useful density is a post-2018 capability. For sites with no prior RF baseline, the characterisation period before change detection is meaningful will be months, not days. Buyers should plan collection timelines accordingly and not expect instant attribution from a standing start.
Typical figures
| Geolocation accuracy (TDOA/FDOA, HawkEye 360) | Approximately 1 km CEP for sustained emitters; degrades for low-power or intermittent sources |
| Frequency coverage (HawkEye 360, published) | VHF and UHF bands; specific sub-band coverage not fully disclosed commercially |
| LEO revisit (Spire LEMUR-2 constellation) | Sub-hourly average globally across 100+ satellites; per-point dwell is 5–10 minutes per pass |
| Corroborating SAR resolution (Sentinel-1 Stripmap) | 5 m range, 20 m azimuth; 6-day repeat with two-satellite constellation |
| Corroborating optical resolution (commercial) | 0.3–3 m depending on platform; cloud cover is the primary availability constraint |
| Minimum detectable emitter (indicative) | Sustained VHF/UHF transmissions above approximately a few watts ERP in open terrain; shielded or low-power sources may not be detectable |
| RF archive depth (commercial LEO) | HawkEye 360 commercial operations from approximately 2019; pre-2019 coverage limited |
| Baseline characterisation period | Typically 4–12 weeks of systematic collection before statistically defensible change detection |
| Delivery formats | Emitter activity reports, GIS layers (GeoJSON/KML), time-series charts, alert feeds |
Analytics Satellize can run
| Emitter baseline catalogue | Statistical aggregation of TDOA/FDOA geolocation fixes across multiple passes; frequency and power characterisation | GIS layer of catalogued emitters with frequency, estimated power class, and activity schedule; updated monthly |
| Traffic density time-series | Pass-by-pass emitter count and signal duration aggregation; anomaly detection against rolling baseline using published change-detection statistics | Time-series chart and alert report flagging statistically significant departures from baseline activity levels |
| New emitter type detection | Frequency band and waveform class comparison against baseline catalogue; novel entries flagged for human review | Alert report with emitter geolocation, frequency, first-detected timestamp, and confidence assessment |
| Operating-hours anomaly report | Temporal pattern analysis of emission activity against declared facility operating schedule and historical rhythm | Annotated activity calendar identifying off-schedule high-traffic periods, with pass-level evidence citations |
| Multi-source corroboration assessment | Fusion of RF anomaly flags with Sentinel-1 SAR structural change detection and optical imagery review; consistent with published OSINT multi-source methodology | Integrated site assessment report rating overall confidence level (low/medium/high) with explicit statement of ambiguities and collection gaps |
| Physical site change context layer | Sentinel-1 coherence change detection and optical change analysis for antenna, vehicle, and construction activity | GIS overlay of detected physical changes with timestamps, for overlay against RF anomaly timeline |
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.