Sensors
- Sentinel-1 (ESA, C-band, 5.405 GHz): Interferometric Wide Swath mode delivers 250 km swath at 5 x 20 m resolution (range x azimuth) with a 6-day repeat at the equator from two satellites. ESA has publicly documented systematic RFI contamination in IW imagery over the Middle East, South Asia and East Asia, attributable to ground surveillance radars and communications systems operating near 5.4 GHz.
- ALOS-2 PALSAR-2 (JAXA, L-band, 1.2578 GHz): L-band is susceptible to a different interference population: aeronautical and maritime navigation radars, and certain military systems. Stripmap mode achieves 3 m resolution; ScanSAR reaches 100 m over a 350 km swath. L-band RFI tends to produce narrower spectral spikes than C-band contamination, making frequency-domain filtering somewhat cleaner.
- RADARSAT-2 (MDA, C-band, 5.405 GHz): Commercial C-band SAR with programmable polarisation (single, dual, quad). Spotlight mode reaches 1 m resolution, useful for confirming emitter location hypotheses at fine spatial scale. 24-day repeat but commercial tasking allows off-track acquisition to shorten revisit.
- Sentinel-1 NRT processing (ESA Copernicus): Level-1 SLC products are available within 1 hour of acquisition over priority areas and within 3 hours globally, enabling near-real-time RFI monitoring rather than retrospective analysis only.
What a bright streak in a radar image actually means
A SAR system illuminates the ground with a precisely timed pulse train and records the backscattered echo. When a ground-based transmitter radiates energy in or near the SAR operating band, the receiver cannot distinguish that signal from genuine ground return. The result is a coherent artefact: a bright streak aligned with the range direction (for a continuous-wave or narrowband emitter) or a diffuse elevation in the noise floor (for a wideband source). Range-ambiguity ghosts appear when the interfering signal arrives at a delay that the processor interprets as a second echo from a different range.
These artefacts are not random. Their geometry encodes the emitter's frequency offset from the SAR carrier, its pulse repetition interval if it is pulsed, and, through triangulation across multiple passes or multiple look angles, an approximate ground position. ESA's own quality-control publications flag persistent RFI sites in Sentinel-1 imagery over Iran, Saudi Arabia, India, China and parts of Southeast Asia, most of them consistent with air-surveillance or weather radars operating near 5.4 GHz.
The suppression problem and why it is not fully solved
RFI mitigation in SAR falls into three broad algorithm families. Frequency-domain notch filtering excises the contaminated sub-band before image formation; it is fast and works well against narrowband continuous-wave emitters, but it discards genuine backscatter at the notched frequencies and degrades radiometric fidelity. Time-frequency methods, including short-time Fourier transforms and wavelet decompositions, localise the interference in both dimensions and can preserve more of the signal, at the cost of higher computational load. Sub-space or parametric approaches model the interfering signal explicitly and subtract it; these work best when the emitter's waveform is stable across the synthetic aperture.
None of these methods is clean when the interfering transmitter is wideband, pulsed at a rate close to the SAR PRF, or mobile. Residual contamination after filtering can still corrupt coherent change detection and interferometric phase, which are far more sensitive than amplitude imagery. Analysts should treat post-suppression interferograms over known RFI zones with explicit scepticism, particularly for deformation studies where a spurious phase ramp of a few millimetres per year is indistinguishable from genuine ground motion without independent validation.
Reading the interference pattern as an emitter signature
Suppression and characterisation are two sides of the same analysis. Before filtering, the raw Single Look Complex (SLC) data carries the emitter's spectral fingerprint. A narrowband emitter at a fixed offset from the SAR carrier produces a sinusoidal modulation in the azimuth direction whose spatial frequency maps directly to the frequency offset. A pulsed emitter produces a comb of harmonics. Comparing that comb across multiple Sentinel-1 passes over the same area can confirm whether the emitter is the same physical system or has been retuned.
Spatial persistence is equally informative. An emitter that appears in every descending pass over a fixed coordinate but never in ascending passes is almost certainly ground-based and directional, its beam geometry favouring one satellite look angle. An emitter that drifts between passes, or that correlates with known shipping lanes, suggests a shipborne radar. This kind of multi-temporal, multi-geometry stacking is not routinely done by national mapping agencies; it requires deliberate analytic effort against an archive that now extends back to Sentinel-1A's launch in April 2014.
What the archive depth makes possible
Ten years of open Sentinel-1 SLC data is an underused resource for spectrum regulators and defence planners. A persistent emitter that has been contaminating imagery since 2015 leaves a decade-long record of its operating pattern: when it is active, how its frequency has drifted, and whether its intensity has changed. That is behavioural intelligence derived entirely from publicly available imagery, with no requirement for dedicated signal-intelligence collection.
The practical limit is spatial resolution. Sentinel-1 IW mode at 5 x 20 m can localise a persistent emitter to within a few hundred metres using multi-pass triangulation, but it cannot resolve individual antenna structures. RADARSAT-2 Spotlight at 1 m, tasked on a candidate site identified from Sentinel-1 analysis, can narrow that further. For precise geolocation below 50 m, dedicated RF geolocation from orbit is the appropriate tool, covered separately in this library.
Practical applications and honest scope
Spectrum regulators use RFI maps to prioritise enforcement action and to negotiate coordination agreements under ITU Radio Regulations Article 15. Defence and intelligence customers use persistent-emitter catalogues to track radar network changes without tasking dedicated collection assets. SAR data providers use the same analysis to flag contaminated scenes before delivering them to downstream users, avoiding the reputational cost of supplying imagery that looks broken.
The method does not work in all conditions. Sentinel-1 TOPS mode processing introduces azimuth phase jumps at burst boundaries that can mimic or mask certain interference signatures. Very low-power emitters, below roughly 10 dB above the thermal noise floor of the SLC product, will not produce detectable artefacts. And the analysis is inherently retrospective: Sentinel-1's 6-day revisit means a transient emitter active for less than a day may never be captured. Satellize runs this class of persistent-emitter stacking analysis on open Sentinel-1 archives for government clients, applying the same processing pipeline it uses in its Tonga crop-estimation programme to manage large multi-temporal data volumes efficiently.
Getting from raw SLC to a usable emitter catalogue
The processing chain starts with ESA's Sentinel Application Platform (SNAP), which handles SLC ingestion, burst assembly and Doppler centroid estimation. RFI characterisation then branches off before the final multi-look step, preserving the full complex spectrum. Frequency-domain analysis of each range line identifies anomalous spectral peaks; these are catalogued by sub-swath, burst, and acquisition date. Spatial clustering across passes produces candidate emitter locations with confidence ellipses derived from the geometry of the contributing acquisitions.
Output formats that are actually useful to clients differ from what remote-sensing papers typically publish. A GIS layer of candidate emitter polygons with temporal activity flags is actionable. A per-scene contamination severity score, expressed as the fraction of range lines exceeding a spectral anomaly threshold, integrates into existing SAR data quality workflows. Both are straightforward to automate once the spectral analysis pipeline is established.
Typical figures
| Primary sensor frequency (Sentinel-1) | C-band, 5.405 GHz centre frequency |
| Spatial resolution (Sentinel-1 IW mode) | 5 m range x 20 m azimuth (SLC); 10 m x 10 m after multi-look |
| Spatial resolution (RADARSAT-2 Spotlight) | ~1 m (single polarisation) |
| Revisit (Sentinel-1, two-satellite constellation) | 6 days at equator; shorter at higher latitudes |
| NRT product latency (Sentinel-1 priority zones) | Level-1 SLC available within ~1 hour of acquisition |
| Archive depth (Sentinel-1) | From April 2014 (Sentinel-1A launch); ~10 years of open SLC data |
| Emitter localisation accuracy (multi-pass Sentinel-1) | Typically 100–500 m, depending on number of passes and look-angle diversity |
| Minimum detectable interference level | Approximately 10 dB above SLC thermal noise floor; weaker emitters not reliably detected |
| Coverage per Sentinel-1 IW pass | 250 km swath width |
| Delivery formats | GeoTIFF contamination maps, GeoJSON emitter candidate polygons, CSV temporal activity logs, SNAP-compatible metadata |
Analytics Satellize can run
| Per-scene RFI severity score | Spectral anomaly detection on range-line FFT; fraction of lines exceeding adaptive threshold | Scene-level quality flag integrated into SAR data delivery metadata or dashboard |
| Persistent emitter candidate catalogue | Multi-temporal spectral peak clustering across Sentinel-1 SLC archive; spatial intersection of look-angle lines | GeoJSON layer of emitter candidate polygons with activity timeline and confidence ellipses |
| Emitter behavioural timeline | Time-series of spectral peak frequency, intensity and duty cycle extracted from SLC archive per candidate site | CSV or PDF report showing operating pattern, frequency drift and intensity trend over selected date range |
| RFI-suppressed image product | Frequency-domain notch filtering or sub-space subtraction applied before multi-look, with residual contamination flagged | GeoTIFF amplitude image with suppression applied and per-pixel quality mask indicating affected range bins |
| Interferometric phase quality assessment | Coherence estimation and phase-ramp analysis on post-suppression interferograms over known RFI zones | Report quantifying residual phase error attributable to RFI, with recommendation on usability for deformation analysis |
| Regulatory coordination dossier | Aggregation of persistent emitter records with ITU band allocation context and cross-pass confirmation evidence | Structured PDF dossier suitable for submission to national spectrum regulator or ITU coordination process |
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.