Meeting 2 of the Committee on a Survey of the
Active Sensing Uses of the Radio Spectrum
November 8, 2013
Jet Propulsion Laboratory
The National Academies of Sciences, Engineering, and Medicine’s Committee on a Survey of the Active Sensing Uses of the Radio Spectrum convened a 1-day workshop of experts in the area of radio frequency interference (RFI) and active users from the radio frequency spectrum scientific community on November 8, 2013, at the Jet Propulsion Laboratory (JPL) in Pasadena, California. The purpose of the workshop was to obtain information from experts working in the area of active remote sensing of Earth’s environment on the problems being encountered in the use of radars for such purposes. Radio-frequency interference has been increasingly observed in data recorded by several airborne and spaceborne radar remote sensing systems. The Academies established the Committee on a Survey of the Active Sensing Uses of the Radio Spectrum to study this problem.
The workshop was organized by committee chair Fawwaz Ulaby and committee member Mike Spencer. Attendees included the following:
NOTE: The notes on which this summary is based were taken by committee member Albin J. Gasiewski, University of Colorado. Acronyms not defined in the text are defined in Appendix D.
Fawwaz Ulaby, University of Michigan, workshop co-organizer and committee chair
Mike Spencer, JPL, workshop co-organizer and committee member
Mark Davis, Independent Consultant
Steve Durden, JPL
Albin Gasiewski, University of Colorado, committee member
Leif Harcke, JPL
Scott Hensley, JPL
Bryan Huneycutt, JPL
Jim Reis, Fugro EarthData
Frank Sanders, National Telecommunications Industry Association (NTIA)
Dean Sangiorgi, Canadian Space Agency (CSA) (by telecon)
Masanobu Shimada, Japan Aerospace Exploration Agency (JAXA)
Pierluigi Silvestrin, European Space Agency (ESA) (by telecon)
The workshop was opened by Fawwaz Ulaby, who discussed the context and defined the charge for the workshop. The workshop purpose stemmed from the following charges to the committee:
- Describe the science that is currently being conducted using the radio spectrum for transmission and measurement of these active signals and identify the spectrum requirements necessary to conduct this research;
- Identify the anticipated future spectrum requirements necessary to continue to conduct and expand this research for the next 10-20 years, taking into account trends in overall active use of the spectrum;
- Discuss the value to the nation of accommodating the active scientific use of the spectrum, recognizing the need to balance the needs of multiple communities;
- Assess the active science communities’ current and anticipated future access to the spectrum required for research; and
- Recommend strategies to accommodate the continued active use of the spectrum for scientific purposes in order to maintain the needed science capabilities identified above.
RFI has been identified by the committee as an important issue. Even though several research institutions and space agencies have reported RFI observations, and in some cases techniques were developed to remove or filter out the interfering signals, there is a need to generate a comprehensive report to document the
RFI issue across all radar-frequency allocated bands from the P-band through the Ka-band as part of this study.
Accordingly, the purpose of the workshop was to answer the following questions for all radar bands between the P-band and the millimeter-wave part of the spectrum:
- What is the extent of the RFI problem? Where does it occur geographically, and who is the interfering source? Conversely, are remote sensing radars interfering with other users?
- How does the RFI impact the quality of the information extracted from the radar data?
- What mitigation techniques, if any, are used to deal with the RFI problem?
- Projecting 20 years into the future, what future missions are likely to be negatively impacted by the RFI issue?
Prof. Ulaby reminded the attendees of the timeline of the study, which should be completed by approximately the middle of 2014.
PERSPECTIVES FROM THE EUROPEAN SPACE AGENCY
The first presentation was by Pierluigi Silvestrin of ESA, assisted by Elena Daganzo, and presented via teleconference. He described the ESA Earth Observation Programme, which included radar altimeters (e.g., Cryosat-2, Sentinel-3 series), a cloud profiling radar on EarthCare (2018), and synthetic aperture radars on the Sentinel-1 series, and scatterometers on the MetOp operational series of satellites. Silvestrin reminded the attendees of the operation of the Ku- and C-band radars on EnviSat, which operated for 10 years. The Sentinel series is meant to establish a long-term, continuous monitoring capability. A set of interference criteria was presented based on ITU RS 1166. The issue of RFI is an increasing problem for both passive and active ESA sensors. Management of the spectrum by the national authorities is more reactive than pro-active, and pro-active management is urged. Silvestrin proceeded to describe the ESA active microwave sensor bands of interest ranging from 0.435 GHz (P-band synthetic aperture radar (SAR) for surface biomass) to 238 GHz (multimillimeter wave cloud radar). RFI cases for ERS-1/2 SAR were rare (10 cases in 20 years), and also in ENVISAT/ASAR. RFI to ASCAT scatterometer on MetOp were also very low, but increasing after ~mid-2012. Occurrences are in the Middle East and (less so) the Far East and North America. He also reported about RFI to the (passive) SMOS synthetic aperture radiometer, which has seen more than 500 cases. More than 50 percent of these cases have been resolved, but high RFI is still seen in the Middle East, China, and Japan.
Examples of RFI cases in ESA C-band SAR (ERS-2 SAR) and EnviSat (5250-5350 MHz, burst RFI) were presented. It seemed that these cases were possibly due to a military radar. Random position and duration suggests radar pulses, possibly a countermeasure radar. Silvestrin noted that the problems may not be seen in the Level 1 data, and processing of RFI at the Level-0 data is needed to identify RFI. RFI mitigation on the Level-0 data in the time domain was able to be used to remove this RFI based on statistics. However, the mitigation algorithm was not implemented in the ground segment as there was no guarantee of the accuracy of the algorithm.
Silvestrin further described the ESA/European Union Sentinel series’ radio band needs from 5250 to 5570 MHz. He indicated that reallocation of local area network allocations by WRC-2003 within this band prompted reassignment of the Sentinel C-band from 5250-5350 MHz to 5350-5470 MHz. In preparation for WRC-2015, he noted with concern that discussions are occurring to expand radio local area network (RLAN) allocations to 5350-5470 MHz, which would hasten the likelihood of SAR interference to Sentinel. The tolerable interference to the Sentinel SAR would be exceeded by 30 dB, and even a single outdoor wireless RLAN would be detectable. He noted that the Sentinel design stemmed from 10 years of effort, while the request to use the new subband for RLAN arose only over the previous year. A relevant question in this regard is this: Will future C-band SARs have the capability to frequency hop from one sub-band to another to avoid RFI? The requisite technology may become available in the future, but it is also likely that the entire C-band range may become a source of RFI, thereby making the frequency hopping capability irrelevant. Also, the C-band RFI sources are anticipated to be nonstationary, and difficult to predict and sense around. How wide a bandwidth would be ultimately needed by C-band SARs? At least 90-100 MHz is needed for data continuity (e.g., for surface topographical change detection using SAR interferometry, InSAR), and higher-resolution design would demand even wider bandwidths. He noted that the original C-band allocation was 320 MHz, and so current systems are operating at reduced bandwidths from even this specification.
The potential for interference between ESA active remote sensing SARS systems operating at 420-450 MHz and other services, such as Space Object Tracking Radars (SOTRs) at the P-band, was also discussed. Silvestrin noted the existence of a proposed approach for sharing which is being discussed with the U.S. Department of Defense, but no reciprocal coordination by the United States has been offered. As a result, secondary mission objectives (e.g., ice sounding) are strongly impacted. Missions using the X-, Ku-, and Ka-bands are also being studied, but no RFI concerns have yet been identified.
PERSPECTIVES FROM THE CANADIAN SPACE AGENCY
Dean Sangiorgi of CSA discussed the needs and current RFI situation of the CSA and, specifically, the use of the C-band. Radarsat-1 used a carrier centered at 5300 MHz, and the Radarsat-2 Constellation Mission (RCM) will use 5405 MHz center frequency. A bandwidth of up to 100 MHz is used. Launch of three satellites is due in 2018. A two-part study with MacDonald Dettwiler Associates (MDA) was undertaken to assess current and past RFI and to simulate RFI. The study used Ottawa as the primary scene with 85 scenes between 1996 and 2012. A consistent backscatter increase was seen in 4 of 9 scenes between 2001-2002; all of these were covered by water and close to urban centers. These were likely a result of RFI increases. Overall an average increase of 3 dB was noted over the 15-year period, although there was no definite way to identify what was the source of the increase.
Sangiorgi also discussed the impact of World Radiocommunication Conference (WRC)-2015 RLAN reallocations on the 5305-5470 MHz band. A joint task group has been formed to discuss this issue, and the MDA study considered additive Gaussian noise to study the potential impact. A scene from May 2010 of the Victoria, British Columbia, airport was used as a baseline. Small detectable noise features are seen at an interference-to-noise level (I/N) of 0 dB in Radarsat-2 images. The additive interference is visible in areas of low backscatter such as runways. Eventually, runways and highways become undetectable at levels of ~6 dB, and at ~12 dB the image becomes useless. Such interference impacts the use of the data for change detection, specifically with respect to archival imagery observed without interference. An experiment using actual RLANs to study the impact empirically is being considered. A graph of number of RLAN connections required to exceed an interference threshold was presented. Approximately 1,050 RLAN connections with an equivalent (or effective) isotropically radiated power (EIRP) between 25 mW and 1 W over the 225 km2 can be tolerated within a 100 MHz bandwidth. A –6 dB I/N threshold was concurred as a rough threshold for observing an effect on InSAR topography. The conclusion of the study is that sharing between RLANs and Radarsat Constellation Mission (RCM) measurements is not practical. Potential use of RLANS in 5350-5450 MHz is thus a growing concern.
CSA foresees continued use of C-band SAR sensors in future missions to ensure data continuity beyond 2025. RCM launches are scheduled to begin in 2018.
PERSPECTIVES FROM THE JAPAN AEROSPACE EXPLORATION AGENCY
Masanobu Shimada of JAXA provided an assessment of RFI concerns to the JAXA L-band SAR missions, including ERS-1/SAR (1992-1998), ALOS/PALSAR (2006-2011), and ALOS-2 (2013-). A bandwidth of 28 MHz is used by PALSAR, and 85 MHz is used by ALOS-2. He also discussed the airborne Pi-SAR radars.
RFI has been recognized as the noise source for the JAXA L-band SAR imaging problem. The contamination bandwidth as become wider as time has progressed. A possible correction is zero padding the contaminated frequency components, resulting in a contamination ratio that has been increasing in time. The RFI for examples shown was believed to be due to ground radars operating over Japan. Both JERS-1 (15 MHz bandwidth about 1.275 GHz center frequency) and PALSAR (28 MHz bandwidth over 1.270 GHz center frequency) were both shown to be affected. Shimada noted that the RFI depends on observation angle. Notch filters are applied to the spectral data to produce images that have RFIs of ~3 MHz in width removed. Significant image improvement over Hawaii and Alaska were shown after RFI mitigation, including changes in overall backscatter baseline. RFI over Korea was shown to be quite large and not removed by notch filtering. The notch filter is applied automatically; however, it also removes some noticeable geophysical features.
Global maps of RFI observed by JERS1-SAR and PALSAR from 1992-2011 using the notch filtering methods were shown. Up to a few percent bandwidth contamination was observed from 1992-1998, increasing to up to 10 percent in 2011. Contamination is noted in Western Europe, Middle East, China, Korea, and a radar line across northern Canada. Contamination also occurs over water. Sidelobe coupling of RFI from over land (for water observations) was also noted. Major increases in contaminated areas are observed from JERS-1 to PALSAR, although differences in radar bandwidth and power level would warrant a major reprocessing effort to ensure consistency between the two data sets.
FREQUENCY ALLOCATION CHALLENGES FROM UWB RADARS
Frequency Allocation Challenges from ultra wideband (UWB) radars were discussed by Mark Davis, an independent consultant and representative of the Institute of Electrical and Electronics Engineers (IEEE) Aerospace and Electronic Systems Society (AESS). He noted that advanced radar techniques are demanding more bandwidth and frequencies, but wideband communications for government and personal/business use are being allocated more bandwidth. The economics for frequency allocation are presenting a serious impediment for radar frequency allocation. AESS has the charter for maintaining the IEEE radar definitions and standards. A subpanel on radar waveform diversity has been established to explore the impact of spectrum allocations. IEEE standard 1675 defines UWB as being either “500 MHz bandwidth or a percentage bandwidth greater than 25 percent” of the carrier frequency. Two examples of airborne UWB radars: FOPEN ATD (100 percent and 120 percent bandwidth for the UHF band and the VHF band, respectively), and GeoSAR (50 percent bandwidth for the UHF band and 2 percent bandwidth for the X-band) were discussed. Both systems experienced a 3-year frequency approval process.
Davis discussed the UWB intercept power measurement concept. The NTIA Part 15 UHF Frequency Avoidance Table shows very few open bands within UHF, with –70 to –90 dBm sensitivity thresholds. All UWB airborne systems must use a very conservative approach to notching across the UHF band in order to get permission to operate. The range from 960 to ~1470 MHz is particularly important for GPS and other critical applications. However, the NTIA requirements cause significant loss of bandwidth due to excessive notching. Linear frequency-modulated notching or frequency jump bursts are used for notching. Significant loss of resolution occurs for current notching requirements. Bridging notched frequencies can improve resolution, as can use of cognitive radar techniques. The cognitive radar approach senses and adapts to the radio frequency environment, and uses adaptive transmit waveforms to achieve greater average bandwidth.
The conclusion is that compliance standards are conservative and inflexible, and that new technologies need to be considered. The international radar community needs to adapt and develop new technologies analogous to cognitive radio. Systems demonstrating cognitive radio include those developed by the U.S. Army and the Defense Advanced Research Projects Agency, but only now are these technologies being studied for use in radar.
NATIONAL TELECOMMUNICATIONS INDUSTRY ASSOCIATION PERSPECTIVE
Frank Sanders of the NTIA provided the issue of RFI between active L-band long-range search radars and satellites. Terrestrial radar usually transmit upwards at angles of ~0.5 to 20 degrees above the horizon and, thus, can scan across many satellites. The high sensitivity of satellite and radar receivers suggests that RFI may potentially occur in both directions. Propagation distances are generally in the range of 440 to 900 km. Dr. Sanders presented typical solid-state air-search radar EMC characteristics. These radars operate between 1200-1400 MHz, and have very gain antennas (30-40 dBi). Scan rates are ~1 rotation per second. Unintentional out-of-band (OOB) emissions may only be 40 dB down at 1400 MHz due to a non-conservative OOB spectral emission mask. Sanders noted that the P(Y) channel frequency of GPS receivers at 1227.60 MHz are not affected by long-range search radars since they integrate over the brief search radar pulse. Peak EIRP at 1425 MHz and 440 km distance would be approximately -83 dBm/MHz into a 0 dBi satellite receiver (see NTIA report TR-06-444, available at http://www.its.bldrdoc.gov/publications/2481.aspx). A transmit power of ~250 W might be possible without exceeding a typical –6 dB I/N limit.
Interference to long radars from satellite can also occur, as shown in a plan position indicator (PPI) long-range radar image with ~5 degree blanked azimuthal sector. Interference as part of NASA SMAP design studies were performed showing
reductions of the probability of detection (Pd) at interference levels of ~20 dB I/N for three long-range radar types. SMAP has a 1-MHz transmit bandwidth, which is comparable to the long-range radars (a few MHz).
In summary, RFI can occur in both directions, and the U.S. recommendations on OOB rolloff provides some protection. Satellite transmitters can provide RFI into long-range radars, potentially causing loss of tracked aircraft. Occurrence of loss of tracking is more frequent that 1 in 106, but difficult to further specify.
NASA ACTIVE REMOTE SENSING SPECTRUM OVERVIEW
Bryan Huneycutt of JPL provided as overview of NASA JPL satellite experiment, beginning with Seasat in 1978. Sensor types are SARs (e.g., Seasat, AIR-MOSS), altimeters (e.g., JASON, SWOT), scatterometers (e.g., QuikSCAT, Aquarius), precipitation radars (e.g., TRMM TMI or GPM GMI), and cloud radars (e.g., CloudSat). Applications impact viewing geometry and required bands, but requirements include frequencies from the P-band (0.432-0.438 GHz) to the millimeter wave spectrum (237.9-238.0 GHz). This is frequency range of over 500 (or ~9 octaves). Several current and planned missions use bands that do not necessarily have primary allocations for EESS active remote sensing. The EESS-Active allocated bands and their current uses were discussed, specifically radiolocation. Interference criteria for missions range from –3 to –10 dB I/N, and data availability criteria are between 95 and 99 percent.
JPL RADAR RFI EXPERIENCE WITH ATMOSPHERIC RADARS
Steve Durden of NASA JPL provided a synopsis of airborne and satellite atmospheric radar missions. The ARMAR radar operated at 13.4 GHz, and later (APR-2) at 13.4, and 35.6 GHz. The Airborne Cloud Radar (ACR) at 94.9 GHz was developed in the 1990s was a forerunner of CloudSat. This allocation was rooted in historical continuity. Getting a temporary assignment has been straightforward. Bandwidths required are ~100 MHz. The CloudSat allocation was originally at 78 GHz, but moved to 94 GHz to improve sensitivity to small cloud particles and take advantage of the availability of extended interaction klystron amplifier technology. Potential damaging interference from CloudSat to radioastronomy facilities (e.g., Haystack) remains a concern, but has apparently never occurred since most radio astronomy facilities do not observe at zenith. In summary, there are no major interference problems occurring with cloud radars, and the frequency allocations and bandwidths currently allocated are adequate for current and future missions. However, the possibility that automobile collision avoidance radars are being built to operate at the W-band raises some concerns about the future viability of this band.
JPL RADAR RFI EXPERIENCE WITH P-BAND RADARS
Leif Harcke of NASA JPL provided a synopsis of AirMOSS P-band radar RFI experience. AirMOSS is capable of operating from 280 to 450 MHz with an instantaneous bandwidth of 80 MHz. Only a 20 MHz chirped operating bandwidth without notching has currently been approved. The CleanRFI code has been used for RFI removal. The algorithm was illustrated using 430-450 MHz P-band data from AIRSAR in 1998 where significant RFI was successfully removed. Narrowband amateur radio communication sources at 450 MHz were cited as likely sources. The Point Target Simulator (PTS) for simulating interference was described. It was noted that the interference was often so time dynamic (for a number of reasons, including frequency hopping, multipath fading, sidelobe reception) that derived least mean square filter coefficients could not readily be reused from pulse to pulse. Changes of 2-3 dB in backscattering cross section were noted upon filtering. Data obtained using 32,000 fast Fourier transform samples over 80 MHz reveal many interference sources that are 20-30 dB above the radar receiver noise floor. Most of these sources are narrowband and are ubiquitous over all AirMOSS calibration sites. The strongest RFI was observed in the United States from 406-420, 450-470, and 440-450 MHz. The 420-440 MHz spectrum has been observed to be the quietest. Some wideband sources are present from time to time and at certain locations (Duke Experimental Forest).
JPL PERSPECTIVES ON L-BAND RADAR INTERFERENCE
Mike Spencer of NASA JPL provided a discussion of the impact of RFI on L-band active remote sensing. The science applications include soil moisture, salinity, biomass measurement, disaster management, and polar studies. Sensors include Aquarius (~1260 MHz) and SMAP (~1230 MHz), and NI-SAR and UAVSAR (~1280 MHz). A number of radionavigation and radar systems operate over the bands of these instruments. Aquarius observations provided a global map of radar RFI in dBm. The interference is globally conspicuous. Over North America the strongest sources are known emitters, most of which are long-range radars. The current interference environment seems mostly manageable from a science perspective. However, there are worrisome trends, including lower power/longer pulses and higher pulse repetition frequency Common Air Route Surveillance Radars, which are harder to remove. Also worrisome is the general increase of the urban noise floor, which was illustrated by PALSAR measurements over Hong Kong.
The SMAP frequency approval process required changing from low-power long pulses (200 W/40 usec) to higher-power shorter pulses (500 W/15 usec). The process required 11 months and extensive testing against Federal Aviation Administration (FAA) radars in Oklahoma City, compatibility testing against GPS systems,
and an extensive dynamics analysis. A short duty cycle is preferred to minimize interference to long-range radars. Some operational restrictions on SMAP may yet be imposed. SMAP is scheduled to be launched in January 2015.1 It is noted that Aquarius has been operating for 2 years without incurring any interference to ground-based radars.
In summary, the status quo is acceptable, but there are concerns about the long-term future. There are also ill-defined spectral acceptability criteria that hinder the approval process, thus carrying operational risk deep into the project life cycle.
JPL PERSPECTIVES ON SCIENCE DEGRADATION METRICS TO TRACK THE IMPACT OF AN EVOLVING RFI ENVIRONMENT
Scott Hensley of NASA JPL provided a discussion of metrics to track the impact of an evolving RFI environment and observations of increasing RFI over the past decades of active science studies. A goal is that NASA develop a standard way of setting RFI criteria. NASA operates a variety of sensors over a wide range of frequencies that cannot be shifted in frequency due to the physics of the measurement. Hensley’s fundamental recommendation is for NASA to fund development of metrics addressing the impact of RFI on active sensor science. Questions such metrics need to quantitatively answer are the following: (1) What is the impact of RFI to science? (2) When is RFI so great that precludes useful science? and (3) What is the benefit of this science to society? Anecdotal evidence is insufficient to track these impacts.
Hensley next outlined the qualities of a good metric. It needs to be relatively easy to compute, measure degradation in way that is easy to explain, accurately reflect the loss of science to changes in the RFI environment, be able to be tracked over a mission lifetime to inform future missions, and be meaningfully averaged both spatially and temporally. With radars, every aspect of the received signal is used (amplitude, frequency, time delay, Doppler shift, and phase). The impact to science depends on the character of the RFI. The full characteristics of the RFI signals should be characterized empirically. He suggested a table that would begin a means of archiving RFI.
Hensley considered the example of SMAP, which offers a simple metric for quantifying the impact of RFI on the error standard deviation (STD) of the backscattering cross section. This error STD can be directly related to soil moisture error. In radar interferometry the signal-to-noise ratio (SNR) correlation is also a straightforward function of the SNR and RFI signal-to-interference ratio (SIR). Carrying these degradation ratios back to the accuracy of the science product depends on the differential relationship between backscattered signal and geophysi-
1 As of January 24, 2015.
cal quantity of interest. The example of RFI impact on topography and deformation measurements was suggested.
RFI IN THE P-BAND AND THE X-BAND SEEN BY THE GEOSAR SYSTEM
Jim Reis of Fugro EarthData discussed a user’s perspective on a commercial SAR system, including the licensing issues associated with allocations. The GeoSAR is a wideband, interferometric, fully polarimetric P- and X-band system that has been in continuous commercial use since 2003. It uses a Gulfstream II aircraft to provide a 20 m interferometric baseline. Reis provided a number of examples of RFI occurring at various locations, including Papua New Guinea, Columbia, Central Alaska, Maryland, and Barrow Alaska. Notch filters over the ~160 MHz GeoSAR bandwidth (from 260 to 440 MHz) are required to in some regions to remove some authorized and unauthorized RFI. Both cases of narrowband and broadband RFI are observed, some of which is persistent over at least 2 years.
Certification of the GeoSAR system to Stage 4 through NTIA and Federal Communications Commission (FCC) was a lengthy process. Since it was a system that had routinely licensed over the last 8 years, it was considered for allocation. An early RFI study led to a requirement for mandatory notching of all users within a 20 nmi zone from nadir to prevent communications disruption. However, the Air Force Spectrum Management Office (AFSMO) imposed an umbrella standard requiring matching of all users up to 500 nmi from the acquisition site. Overall, the final notching requirements became more strict. Also, an active U.S. government contract was required at the time of allocation. The new process for allocation is through the FCC, which submits the request for a 30-day suspension with affected government agencies (NTIA, AFSMO, FAA, etc.), followed by a special temporary authorization. Such slow and uncertain allocation processes inhibits the regular operation of the system required to maintain operational readiness.
DISCUSSION AND SUMMARY
A brief workshop discussion session was held after the last presentation. It was requested by Prof. Ulaby that the attendees share with the organizer the three most important observations and recommendations from the workshop.
General observations from discussions held during the workshop presentations and afterwards during the discussion session follow:
- The approval process for transmit allocations is too cumbersome, lengthy, and ill-advised. The U.S. Interagency Radio Advisory Committee (IRAC) operates by consensus of its members and, thus, provides numerous opportunities to table or veto applications. The allocation for GeoSAR radar allocations is ineffective and encourages only voluntary self-compliance by the applicant.
- The development of a common set of standardized metrics connecting RFI levels to science product accuracies would be useful in assessing the costs of observing with increasing levels of RFI.
- The associated problems of adaptive notching and how notch variation on a pulse-to-pule basis affects SAR resolution and processing are open areas that need engineering investigation.