You are on page 1of 10

IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 11, NO.

6, JUNE 2018 1913

Radio Frequency Environment for Earth-Observing


Passive Microwave Imagers
David W. Draper , Member, IEEE

Abstract—This paper examines the radio frequency interference also share most allocated bands, causing abundant interference
(RFI) environment within microwave imager bands for sensors within the bands. Interestingly, certain types of RFI occur within
that have observed the earth over the last two decades. Since the the ITU allocated bands rather than at other nearby frequencies
microwave imagers have used various bands both within and out-
side of International Telecommunications Union (ITU) allocated outside allocated frequency bands (see Section V).
frequencies for passive satellite earth exploration, this survey pro- For the moment, RFI is concentrated at the lower frequencies
vides valuable insight into band selection and mitigation strategies utilized by remote sensing applications for surface applications.
for future missions. Several conclusions are drawn from this pa- Instruments have observed RFI at the L-band (where passive
per. First, significant land-based RFI exists at the C-Band. The remote sensing and radio astronomy are given exclusive claim
two-band mitigation solution for the C-Band utilized by Advanced
Microwave Scanning Radiometer 2 provides some RFI reduction to the spectrum from 1400 to 1427 MHz) [8]–[10], at the C-Band
of high-level RFI (>10 K), and therefore, may have merit for future (where there is no ITU allocation) [1]–[4], the X-band (where
missions in cases where only two subbands can be feasibly imple- there is a shared allocation from 10.6 to 10.7 GHz) [1]–[2], [11],
mented. At the X-band, ocean reflections from direct broadcast and the K-band (where there is a shared allocation from 18.6 to
and communication satellites (especially around Europe) provide 18.8 GHz and from 21.2 to 21.4) [3], [11]–[12]. Thus far, no
considerable interference above the ITU-allocated 10.6–10.7 GHz
band, whereas very little reflected interference is observed within contamination has been noted for bands given exclusive passive
the allocated band. Good out-of-band rejection is required to avoid remote sensing/radio astronomy access at 23.6 to 24.0 GHz
the reflected RFI above 10.7 GHz. Both inside and outside of the X- and 31.3 to 31.8 GHz. The shared allocated bands at higher
band allocation, significant terrestrial RFI exists. For the K-band, frequencies such as the Ka-band (36.0 to 37.0) and the V-band
the 19.35-GHz band utilized by Special Sensor Microwave Imager (52.6 to 59.3 and 65 to 66 GHz) have also not yet observed any
Sounder and Tropical Rainfall Measuring Mission Microwave Im-
ager avoids reflected RFI from satellites that is present around the significant RFI.
continental United States for satellites observing within the 18.6 to For frequency bands that are dedicated only to science ap-
18.8 GHz allocated band. The analysis suggests that the 19.35-GHz plications, such as the L-band, the radiometer community has
band may be preferable to the 18.7 GHz allocated band for ocean some recourse to have illegal transmitters turned OFF [13]; how-
applications if avoiding RFI is desired. ever, for microwave imagers, which operate at higher frequen-
Index Terms—Algorithms, electromagnetic interference, mi- cies with mostly shared allocations, taking such legal action
crowave radiometry, radio frequency interference (RFI). is much more difficult if not impossible. As such, developing
RFI-mitigating radiometers has been a topic of recent research
I. INTRODUCTION
[14]–[19]. Most past radiometer systems had no on-board RFI
N EVER-INCREASING demand for the radio frequency
A spectrum makes preserving pristine bands for earth re-
mote sensing a nearly insurmountable uphill climb. With the C-,
detection and mitigation capability. At the time of conception
of these radiometers, the frequencies of operation were not used
much by other services, the RFI environment was not as well
X-, and K-band having wide-spread commercial use and higher understood or the technology did not readily exist to mitigate the
band RF components becoming more viable, space-based mi- interference as it does now. A combination of high-speed digital
crowave imagers have seen evolving radio frequency interfer- technology advances along with continued RFI mitigation re-
ence (RFI) within their operating bands over the last several search have provided multiple new options for mitigating (rather
decades [1]–[7]. Although most microwave imager channels than just detecting) RFI [20]–[24]. To date, only the most recent
operate within ranges allocated by the International Telecom- the C-band and L-band missions incorporate RFI mitigation into
munications Union (ITU) for passive earth exploration satel- their design. The Advanced Microwave Scanning Radiometer 2
lite service (EESS), active ground and space-based applications (AMSR2) achieves moderate RFI mitigation with two separate
analog channels in the C-Band [25]. The L-band Soil Moisture
Manuscript received September 15, 2017; revised December 6, 2017; ac- Active Passive (SMAP) radiometer, launched in 2015, is the first
cepted January 16, 2018. Date of publication March 12, 2018; date of current of its kind to incorporate on-board digital subbanding, subsam-
version June 29, 2018. This work was supported in part by the Goddard Space
Flight Center and in part by Precipitation Processing System. pling, polarimetry, and kurtosis processing [26] that are used in
The author is with the Ball Aerospace and Technologies Corp, Boulder, CO ground processing to detect and mitigate RFI.
80301 USA (e-mail: ddraper@ball.com). As the remote sensing community moves forward to fu-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. ture missions, it is learning how to reduce the impact of RFI
Digital Object Identifier 10.1109/JSTARS.2018.2801019 on science measurements. The purpose of this paper is to

1939-1404 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
1914 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 11, NO. 6, JUNE 2018

TABLE I
CENTER FREQUENCIES (GHZ) OF RECENT MICROWAVE IMAGERS ADDRESSED
IN THIS PAPER

Instrument C-band X-band K-band K-band∗ Ka-band V, W mm-wave

TMI 10.65 19.35 21.3 37.0 85.5


GMI 10.65 18.7 23.8 36.64 89 166
183.31 ± 3
183.31 ± 7
WindSat 6.8 10.7 18.7 23.8 37.0
AMSR-E 6.925 10.65 18.7 23.8 36.5 89
AMSR2 6.925, 7.3 10.65 18.7 23.8 36.5 89
SSMI 19.35 22.235 37.0 85.5
SSMIS 19.35 22.235 37.0 50.3– 150
63.28, 183.31 ± 1
91.65 183.31±3
183.31±7


Water vapor channels.
Fig. 1. Frequency bands utilized by microwave imagers. The pink bands See Fig. 1 for Bandwidths of 6 to 24 GHz Imaging Channels.
identify ITU-allocated dedicated (d) passive bands for remote sensing, radio
astronomy, and space research. The gray bands are ITU-allocated bands that are
shared with active ground or space-borne transmitters. The map of the RFI environment for the C-band through
the Ku-band is shown in Fig. 2. It shows the maximum RFI
provide guidance for future missions by offering context into index in Kelvins over a 1 month period. Much of the land-
the current and past RFI environment that has been measured based RFI segments nicely by geopolitical boundary. For the
by recent microwave imagers from the C-band to K-band. The C-Band, the two AMSR2 channels exhibit sufficiently different
measurement passbands for current and past sensors in the C- RFI corrupted regions that significant RFI mitigation is possible.
to K-band range are given in Fig. 1. The instruments of inter- The RFI observed around the United States (18.7 GHz)
est include the Special Sensor Microwave/Imager (SSM/I) and and Europe (10.7 GHz) originates from geosynchronous di-
its successor Special Sensor Microwave Imager Sounder (SS- rect broadcast satellite transmission reflecting off the ocean sur-
MIS), which have flown on multiple Defense Meteorological face. In the current RFI environment, the ocean-reflected RFI
Satellite Program (DMSP) satellites, the Advanced Microwave is avoided by remaining within the ITU-allocated the X-band
Scanning Radiometer – Earth Observing System (AMSR-E) shared allocation at 10.65 GHz as GMI does, and using a higher
aboard the Aqua satellite and is successor AMSR2 on the Global band (19.35 GHz) outside of the ITU allocation with less inter-
Change Observation Mission-Water developed by the Japanese ference as TMI does.
Aerospace Exploration Agency, the Tropical Rainfall Measur- Section II presents the method for RFI detection and mag-
ing Mission Microwave Imager (TMI) and the follow-on Global nitude estimation. Using the method, Section III provides a
Precipitation Measurement Microwave Imager (GMI) flown by multiyear trend analysis. Section IV analyzes the dual-band
the National Aeronautics and Space Administration, and Wind- mitigation approach from AMSR2. Section V compares RFI re-
Sat instrument flown on the Coriolis spacecraft by the Naval flected from the ocean within various instruments with different
Research Laboratory. bands. Section VI provides conclusions.
We detect RFI using a relatively simple generalized spectral
index method, where each channel of interest is compared to a
linear combination of other channels and their squares, which II. METHOD
represent a best fit to the channel of interest depending upon Where on-board mitigation techniques are not available, var-
surface type [27]. This type of detection scheme is advantageous ious methods have been utilized over the last two decades to
because it operates on the measured data itself, without needing characterize and flag RFI in ground algorithms. Fundamentally,
external information. It is limited by a lower sensitivity in cases the methods identify unnatural variation of the data that is not
of atypical surface conditions or natural phenomenon such as explained by common geophysical phenomena. The various RFI
convective storms, melting snow, or salt flats. It also has lower detection methods used in postprocessing are given in Table II.
sensitivity along coast lines or inland waterways where both
land and water fall within the same measurement.
A. Past RFI Detection Methods
Center frequencies for channels of each of the imagers are
provided in Table I. We examine the past RFI environment over The simplest methods are the brightness temperature (Tb)
various areas of the world and how the environment has changed threshold and spectral difference methods. The Tb threshold
over time. We look at the spectral variation of the RFI insofar as detects RFI when its effect increases the Tb some observed
it can be measured with the current and recent past radiometers natural limit such as 320 K [28]. It is the simplest method
in operation. We find that in various developed countries, the RFI of detection, and therefore, has the lowest sensitivity, and only
has slowly decreased over the last 1.5 decades, but is increasing works well over land where the natural Tb is high, and therefore,
in developing areas. not far from the Tb threshold.

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
DRAPER: RADIO FREQUENCY ENVIRONMENT FOR EARTH-OBSERVING PASSIVE MICROWAVE IMAGERS 1915

Fig. 2. RFI observed by microwave imagers from the C-band to K-band. Each color represents the maximum RFI index over a month period at a several bands
observed by the radiometers (See Fig. 1 for a description of bands used by each sensor). RFI at 6.9 GHz (green) is most notable in the continental United States,
Brazil, India, Middle East, and Japan. 7.3 GHz RFI (yellow) is pronounced in Eastern Europe, Russia, Turkey, Africa, Indonesia, and South-East Asia, 10.65 GHz
RFI (within ITU shared band, red) is observed at high levels in Great Britain, Italy, Turkey, the Nile Delta, Japan, and Mexico (although somewhat obscured by
other frequencies in Mexico). The wider WindSat 10.7 GHz is also seen in the same locations as well as around the continent of Europe as direct broadcast satellites
reflect off the ocean. The RFI at 18.7 GHz (also within ITU shared band, navy) appears most dramatically around the continental United States and Hawaii from
direct broadcast satellites. The 19.35 GHz Band (light blue) shows up in Mexico and Johannesburg South Africa. Depending upon availability of data, the data
may have been acquired anywhere between late 2014 and 2017.

TABLE II can be approximated by comparing a channel of interest to a


VARIOUS RFI CHARACTERIZATION METHODS
linear combination of other channels and their squares. The
model difference algorithm is similar to the method used in the
Method Surface Type Sensitivity∗ Reference paper, presented in the next section.
Tb Threshold Land Low e.g., [28] The principal component analysis decomposes the multispec-
Spectral difference Land Low to Moderate [1]–[3] tral Tb data or RFI indices into a set of vectors that form a
Model difference Ocean Moderate to High [2]
Principal component Land, Ocean Moderate to High [2], [29],
new orthonormal data space. The decomposition forms a natu-
Polarization index Land (second, Low to High [30], [31] ral basis set for the data, where the first basis vectors span the
third, fourth) space of highest variance. Unnatural emissions express them-
Ocean (third,
fourth)
selves as large magnitudes in the lowest ranked components. [2],
Chi-squared probability Ocean Land, Moderate to High [32] [29] [29], [30].
Probability Distribution Ocean Moderate to High Other methods include polarization index method and chi-
Generalized RFI Index Land, Ocean Moderate to High This work
squared probability method. The polarization index involves

Low Sensitivity = 20 K+. examining the polarization characteristics in either second, third,
Moderate Sensitivity = 10 to 20 K. or fourth Stokes for anthropomorphic signatures. The second
High Sensitivity < 10 K. Stokes can be relatively good for identifying linearly polarized
RFI over dry land as the RFI has a clear v-to-h signature. For
The spectral difference method compares the channel of in- fully polarimetric instruments such as WindSat, the third and
terest to the next highest frequency of the same polarization, and fourth Stokes channels are very useful in detecting RFI, since
flags RFI when the two channels differ by more than a threshold the polarimetric signature of the earth for these channels is
[1]–[3]. The Tb threshold and spectral difference methods are typically less than 5 K [31].
useful over land, but do not work as well over ocean where the The chi-squared probability method examines the residual
Tb is lower and spectral difference variability is much greater. chi-squared function in the retrieval algorithm and flags areas as
More sophisticated methods such as the model difference RFI where the “goodness of fit” reflected by high chi-squared
and principal component analysis methods are most useful over values suggest RFI [32]. Furthermore, research continues into
ocean [2]. The model difference method compares a channel of even more sophisticated methods, but this listing suffices for the
interest to the result of a radiative transfer model. This method purposes of this paper.

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
1916 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 11, NO. 6, JUNE 2018

TABLE III
GMI GENERALIZED RFI INDEX COEFFICIENTS USED IN (2)

LAND OCEAN SEAICE

coefficient 10.65 V 10.65 H 18.7 V 18.7 H 10.65 V 10.65 H 18.7 V 18.7 H 10.65 V 10.65 H 18.7 V 18.7 H

a 10 V 1 0 0.2678 0.8294 1 0 1.0743 0.2818 1 0 0.2562 0.1469


a 10 H 0 1 −0.3393 −0.9613 0 1 −0.1198 0.1751 0 1 0.0387 −0.2926
a 18 V −5.1728 −5.1287 1 0 −4.3637 −1.2731 1 0 −4.5888 3.7739 1 0
a 18 H −0.1691 −0.6539 0 1 1.0042 –0.2804 0 1 −2.3296 −2.5049 0 1
a 23 V 7.2909 7.2682 −1.9501 −1.5338 0.1149 −0.2418 0.1759 0.2402 10.5330 0.1952 −2.4211 −1.2576
a 36 V −3.4811 −4.1260 1.5762 2.4969 1.8648 −0.7479 −0.7066 1.5860 −5.2574 −0.9190 1.1131 1.1546
a 36 H 0.8877 1.7169 −0.0866 −1.3592 −0.5954 −0.0604 0.0204 −0.7539 3.9066 1.8001 −0.5157 −1.3764
a 89 V 1.5663 2.5001 −0.7921 −1.2043 −0.4187 0.0790 1.2913 1.5226 2.1925 0.9800 −1.0718 −1.0059
a 89 H −1.3052 −2.1763 0.4082 0.9501 0.0666 0.1374 −0.3753 −0.5219 −3.6059 −1.5765 1.2233 0.9447
b10 V 0 0 −0.001049 −0.001245 0 0 −0.004329 −0.000281 0 0 −0.001286 0.000094
b10 H 0 0 0.000688 0.001164 0 0 0.000415 −0.003704 0 0 0.000080 −0.000284
b18 V 0.005787 0.008734 0 0 0.004994 −0.000287 0 0 0.006185 –0.011295 0 0
b18 H 0.000569 −0.001983 0 0 −0.001200 −0.000084 0 0 0.005120 0.002463 0 0
b23 V −0.012336 −0.012041 0.002321 0.001326 0.000691 0.002084 −0.001268 −0.002088 −0.020891 0.003943 0.003344 0.000754
b36 V 0.007578 0.007470 −0.003043 −0.003140 −0.001569 0.003922 0.001136 −0.002348 0.011734 0.000860 −0.001932 −0.000332
b36 H −0.002089 −0.002182 0.000058 0.000911 0.000352 −0.000752 −0.000329 0.000339 −0.008653 −0.002662 0.000775 0.001033
b89 V −0.003400 −0.005376 0.001572 0.001926 0.000648 −0.000189 −0.002658 −0.003353 −0.004599 −0.001890 0.002052 0.001585
b89 H 0.002800 0.004666 −0.000685 −0.001325 −0.000123 −0.000368 0.001054 0.001605 0.007966 0.003407 −0.002396 −0.001494
a o −85.56 −53.96 −12.32 −30.56 130.31 138.29 –240.74 −358.70 −263.26 −356.19 54.78 85.61

B. Generalized RFI Index Method to be flagged and missed detections may occur more frequently.
Given the spatial distribution of the X- and K-band RFI, there are
Most of the methods discussed in the previous section rely
on the correlation between channels to estimate significant de- few actual earth locations where RFI occurs simultaneously in
both bands (Mexico and Japan being notable exceptions). More
viations from typical conditions. This correlation is used to
sophisticated methods may be developed in the future to ignore
derive a generalized RFI detection method for both land and
ocean. The generalized method expands the idea of the model other channels in the bracketed portion of (1) where known RFI
exists and use a different fit in those cases.
difference method to land scenes. A set of coefficients is de-
The generalized RFI index of (1) may be written in an even
termined such that the magnitude of RFI is estimated as the
difference between the channel of interest and a best fit repre- simpler form, as a linear combination of all channels,
sentation of that channel using channels of different center fre-  
quencies and their squares. The generalized RFI index ΔTb[i] is ΔTb[i] = ao [i] + a j [i]Tb[j] + b j [i]Tb2 [j] (2)
j
expressed as
where j is summed over all channels, aj [i] = 1 for the channel of
ΔTb[i] = Tb[i]
⎧ ⎫ interest, aj [i] = 0 for channels with the same center frequency
⎨  ⎬ of the channels of interest, ao [i] = −ao [i], and aj [i] = −aj [i]
 
− a0 [i] + aj [i]Tb[j] + bj [i]Tb2 [j] for all channels (j) with a different center frequency than the
⎩ ⎭ channel of interest. Likewise, bj [i] = 0 for the channel of in-
j ∈{f c(j )= f c(i)}
(1) terest and all channels with the same center frequency, and
bj [i] = −bj [i] for all other channels.
where i represents the channel index to the channel of interest A separate set of coefficients is determined for land, ocean,
and j is the channel index to all other channels with different and sea ice. The generalized RFI index coefficients derived for
center frequency (f c). The ao [i] are constant terms, the aj [i] GMI are shown in Table III. The generalized RFI index method
are linear coefficients and bj [i] are quadratic coefficients. has moderate sensitivity for native data and high sensitivity
We do not use the opposite polarization of the channel of for temporally averaged data, for most earth conditions. It can,
interest in the fit, but use all other channels of both polarizations however, produce false alarms along coast lines, near lakes and
up to 89 GHz. The exception is the C-Band channels of AMSR2, islands, over certain odd terrains, mis-flagged sea ice, or melt-
where the C-band channels are not used in the best-fit represen- ing snow. It can also be highly dependent upon geometry, and
tation, i.e., bracketed portion of (1), due to the high RFI. With may produce anomalous results if the attitude or altitude of the
this exception, other channels with different center frequencies spacecraft changes. For this analysis, stormy seas, sea-to-ice
that are contaminated by RFI at the X-band and K-band are still interfaces, and certain other geographical features have been
included in the best-fit representation to keep the algorithm sim- masked to avoid known sources of false alarms [27].
ple. The other channels that contain RFI corrupt the RFI index The generalized RFI index method was initially introduced
of the channel of interest, typically creating negative RFI in- for operational flagging of GMI data. The distribution of RFI
dices. In such cases, RFI in the channel of interest is less likely index values for RFI-free data from the GMI instrument’s 10.65

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
DRAPER: RADIO FREQUENCY ENVIRONMENT FOR EARTH-OBSERVING PASSIVE MICROWAVE IMAGERS 1917

Fig. 3. Distribution of generalized RFI indices over land and ocean for RFI-
free data from GMI. For GMI, RFI thresholds were selected to be 6 sigma or
more to reduce the probability of false alarm in operational data.

and 18.7 GHz channels are given in Fig. 3. The distributions pro-
vide insight into the sensitivity of the method. A set of thresholds
are applied to the data on the native grid to flag each measure-
ment cell as containing RFI. A cell is flagged as suspect if the
generalized RFI index exceeds the threshold
Fig. 4. RFI measured for the 6.9 GHz V-pol channels of AMSR-E and
ΔTb[i, type] > Tbthreshold [i, type] (3) AMSR2. Top: Maximum measured RFI Index over 1 month period (October
2017). Middle: Percentage of various regions with RFI detected with a magni-
where i is the channel index and type is the surface type (ocean, tude over 10 K persistently detected more than 10% of the time, as a function of
land, seaice). time. Bottom: Average RFI index for all measurement points within the region
that have been detected with a magnitude over 10 K more than 10% of the time.
To avoid false alarms on GMI, the thresholds used in na-
tive gridded data for processing by the Precipitation Processing
System at Goddard Space Flight Center have been set relatively index based on a monthly average Tb may be set much lower
high, at 6-sigma or more relative to the RFI-free areas (see (2 K for 18 GHz and 5 K for 10 GHz over land). Temporally
Fig. 3 for distribution of the RFI over RFI-free areas). For GMI averaged data with lower RFI thresholds have been used for
10.65 GHz channels, the thresholds are set to 15 K for ocean and identifying a database of regions where GMI is persistently
20 K for land in the operational algorithm because of the desire corrupted by RFI [27].
to reduce false alarms. For 18.7 GHz channels, the thresholds
are set to 10 K for both ocean and land. Given the distributions
in Fig. 3, it is evident that thresholds could be tightened if the III. MULTIYEAR RFI TREND
application required it, and higher level of false alarms were Over the last several decades, the radiometer missions have
tolerable. For example, the 10 GHz threshold could be reduced provided a rich database of Tbs from which we evaluate the
to 10 K over ocean and 18 GHz thresholds could be reduced to changing RFI environment. Four sets of imagers carry a nearly
5 K over the ocean and still be greater than 6-sigma of RFI-free continuous Tb record over the last 15 to 30 years:
areas. 1) The SSM/I sensors, followed by the SSMIS sensors on the
The land and ocean thresholds apply to 80% of the earth. DMSP satellites have operated nearly continuously from
Thresholds are set higher for coastlines, sea ice, and high- 1987 through the present.
latitude regions where significant snow persists. In these re- 2) The AMSR-E and AMSR2 sensors provide a nearly con-
gions, which are about 15% of the earth, the method has lower tinuous dataset from 2002 through the present.
sensitivity, and only high RFI can be differentiated using the 3) The WindSat instrument flown on the Coriolis spacecraft
generalized RFI index. Stormy areas over the ocean (less than by NRL has operated from 2003 through the present.
5% of the data, identified by high 36 GHz horizontal polar- 4) The TMI has operated from 1997 through 2015, followed
ization) are ignored by the flagging due to the highly variable by the GMI from 2014 through the present.
nature of the spectrum. For each sensor, the thresholds have For multiyear trending, we concentrate on the AMSR-E and
been customized to meet the sensitivity of that sensor. AMSR2 satellites since these contain most of the bands of inter-
The generalized RFI index method has very good sensitivity est and provide full global coverage. SSM/I and SSMIS have a
when used on temporally averaged data. Thresholds for RFI much longer data record, but do not have the C-Band or X-Band

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
1918 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 11, NO. 6, JUNE 2018

Fig. 5. RFI measured for the 10.65 GHz V-pol channels of AMSR-E and Fig. 6. RFI measured for the 18.7 GHz V-pol channels of AMSR-E and
AMSR2. Top: Maximum measured RFI Index over 1 month period (October AMSR2. Top: Maximum measured RFI Index over 1 month period (October
2017). Middle: Percentage of various regions with RFI detected with a magni- 2017). Middle: Percentage of various regions with RFI detected with a magni-
tude over 15 K persistently detected more than 10% of the time, as a function of tude over 10 K persistently detected more than 10% of the time, as a function of
time. Bottom: Average RFI index for all measurement points within the region time. Bottom: Average RFI index for all measurement points within the region
that have been detected with a magnitude over 15 K more than 10% of the time. that have been detected with a magnitude over 8 K more than 10% of the time.

channels. The TMI and GMI also have a longer data record, but Starting first with the 6.9 GHz RFI in Fig. 4, we find that the
do not have the C-band channels and only cover to +/−38◦ RFI environment has been decreasing gradually over the period
latitude during the TMI era, which limits the extent over which of interest in Japan and the United States. In India, this band
the RFI can be analyzed. Temporal trending of GMI-era data is has shown increased RFI from 2002 through 2010 followed by
also presented in [33]. a decrease in RFI afterward. Other areas such as Brazil and
Figs. 4–6 show the RFI trended from 2002 through 2017 Indonesia have demonstrated a steady increase in RFI extent.
based on the AMSR-E and AMSR2 instruments for three bands: For the 10.65 GHz band, there has been consistent RFI over
6.9 GHz, 10.65 GHz, and 18.7 GHz. We examine the vertical Europe from 2002 through 2017. For countries such as Mexico
polarization channels because they are less sensitive to atmo- and Turkey, the X-band RFI has consistently increased, while
spheric effects and varying surface types that cause a larger Japan has exhibited a gradual decline in RFI.
natural variation in the RFI index. As for 18.7 GHz, In Brazil, RFI was significant around pop-
In each block of Figs. 4–6, the top figure shows the maximum ulated places such as Rio de Janeiro, but has declined in Brazil
RFI index detected over the last month of data collection. For to where it is almost nonexistent at the present at the level of the
regions of the world with significant RFI, we show in the middle algorithm thresholds. Other places such as Belarus and Libya
plot a time series containing the extent of RFI that exceeds show consistently increased RFI. For the United States, much of
the algorithm thresholds (applied to Tb measurements on the the Ku-band RFI stems from reflected RFI from geosynchronous
native grid) more than 10% of the time. Such criteria map out satellites from lakes and frozen areas. This RFI begins in 2007
a smaller area than regions that exhibit any RFI whatsoever, and has steadily increased in amplitude as more direct broad-
but provides the extent of larger-magnitude persistent RFI in cast satellites have been placed in orbit. More will be mentioned
the region. The bottom plot evaluates the average RFI index about reflected RFI from the ocean in Section V.
over the persistently detected region (exceeding thresholds more
than 10% of the time) within each country of interest. Other
IV. DUAL-BAND RFI MITIGATION
criteria of extent and magnitude could have been selected, but
these provide sufficient fidelity to examine longer term trends of The AMSR2 instrument affords a unique assessment of
the RFI. a dual-band RFI mitigation technique. The AMSR2 instru-

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
DRAPER: RADIO FREQUENCY ENVIRONMENT FOR EARTH-OBSERVING PASSIVE MICROWAVE IMAGERS 1919

Fig. 7. RFI Mitigation effectiveness for AMSR2 the C-Band Channels. Top: Fig. 8. 10.65–10.7 being in the protected band, GMI (top) does not exhibit
Maximum RFI index over a month for the 6.925 GHz V-pol channel. Middle: reflected RFI around Europe but does pick up substantial RFI over land. AMSR2
Maximum RFI index over a month for the 7.3 GHz V-pol channel. Bottom: For (middle) is also in the protected band, but does see reflected RFI. WindSat
each measurement over the month, the minimum RFI index of the two frequency (Bottom) whose bandpass is in the nonprotected satellite transmission band
bands is chosen, and the maximum result over the month is shown. AMSR2 exhibits significant interference. GMI and AMSR2 data are taken from October
data taken from October 2017. 2017 and WindSat data are taken from September 2014.

ment measures neighboring the C-Band channels at 6.925 and shows substantially less RFI providing a cleaner product over the
7.3 GHz. To implement dual-band RFI mitigation, the geophys- United States. For most of the rest of the world, the 6.925 GHz
ical retrievals would simply use the channel in the algorithm channel is much cleaner than 7.3. Using the dual-frequency mit-
(either 6.925 or 7.3) with lowest RFI index. This technique igation technique, the number of flagged (exceeding 10 K RFI
mitigates RFI in regions where only one of the bands is con- index) measurements over a 1 month period dropped from 0.2%
taminated. Where both bands are contaminated, the RFI would to 0.02% over the earth, a 90% decrease. The analysis suggests
remain at the lower of the two levels. that, although imperfect, a two-band solution removes a substan-
Such a method may be useful for future missions if there is a tial portion of the moderate to high RFI and could be used for
requirement for low-bandwidth downlink, or where power con- future missions where data downlink or digital processing com-
straints necessitate lower end Field Programmable Gate Arrays plexity are limited. This technique would also remove a portion
in digital implementations. We evaluate the effectiveness of us- of the lower-level RFI (< 10 K). Furthermore, analysis would
ing these two neighboring the C-band channels by examining need to be performed to quantify the mitigation effectiveness
the minimum RFI index of the two channels over the world map. for lower-level RFI.
The RFI index has been produced on the native grid for
AMSR2. For each measurement, the RFI index is computed
and the minimum RFI index for the two bands is selected. The V. RFI REFLECTED FROM THE OCEAN
result is shown in Fig. 7. The top figure shows the maximum Of the bands utilized by microwave imagers, two of the bands
V-pol RFI over a one-month period for 6.925 GHz. The mid- currently also share utility with direct broadcast downlink fre-
dle plot is the maximum V-pol RFI for 7.3 GHz. The Bottom quencies and have observed significant interference. The in-
shows the result of selecting the minimum RFI index of the two terfering broadcast services are allocated 10.7–12.75 GHz and
bands and plotting the maximum of the result over the 1-month 17.3–21.2 GHz, both of which are either near or encompass
period. remote sensing bands.
As shown in Fig. 7, the 6.925 GHz channel exhibits heavy Fig. 8 shows a comparison of the 10.65 GHz channels on
RFI over the continental United States, while the 7.3 channel GMI and AMSR-2 to the 10.7 GHz WindSat band around

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
1920 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 11, NO. 6, JUNE 2018

Fig. 9. 18–19 GHZ RFI around the continental United States for GMI (top)
and TMI (Bottom). Even though in the protected band, GMI picks up substantial Fig. 10. Top: 18 GHz V-pol RFI index mapped through earth reflection to
RFI from geosynchronous satellites. TMI, at a different center frequency avoids the distance of geosynchronous orbit. Middle: Zoomed between −3◦ and +3◦
those transmitters. GMI and TMI data are taken from October 2014. latitude, −120◦ and −80◦ longitude. Bottom: Average RFI index from −1◦ to
+1◦ latitude for longitudes −120◦ to −80◦ . Also listed in the bottom plot are the
satellites with 18 GHz transmitters at the position of the interference sources.

Europe. All sensors show heavy land-based RFI in Europe.


As for reflected RFI off the ocean, GMI carefully maintains its The Direct broadcast satellites over the continental United States
passband within the ITU allocation for passive earth observation utilize the 18.3 through 18.8 GHz downlink band, which directly
between 10.6 and 10.7 GHz. No contamination from reflected interferes with the GMI, WindSat and AMSR2 band. The re-
geosynchronous satellite signals over the ocean around Europe flections are most noticeable over water, but are also noticed
is observed by GMI in this band [34]. over land as well, likely reflecting off lakes and frozen areas.
Moderate amounts of reflected RFI are observed around the The 18.7 GMI band is compared to the 19.35 TMI band around
European continent for AMSR2 and heavy RFI is observed for the continental United States in Fig. 9. The TMI RFI index does
WindSat, which has a much broader bandwidth that ventures not exhibit the same reflections, suggesting that the 19.35 GHz
into the heavily utilized direct broadcast range. It is interesting band would be better suited for ocean applications in future
that AMSR2 claims to utilize the same band as GMI, yet still is sensors.
corrupted by reflected RFI around Europe. To identify the interfering direct broadcast satellites, the GMI
The AMSR2 sensitivity to 10.65 GHz RFI may be due 18.7 GHz RFI index is projected to the location on the geosyn-
to inadequate out-of-band filtering. GMI does not have suffi- chronous sphere where the satellites would need to be located
ciently different orbital geometry from AMSR2 and WindSat such that the transmitted signal would reflect off the earth /ocean
to attribute GMI’s absence of reflected RFI around continental surface and be received by the main beam of the GMI antenna.
Europe to geometry. All three sensors view the earth at similar The results of projecting the RFI index through reflection off
earth incidence angle (55.1° for AMSR2, 52.8° for GMI, and the earth onto the geosynchronous sphere are shown in Fig. 10.
49.9° for WindSat), and due to the periodic 180° yaw maneu- Most of the RFI sources concentrate on a small number of points
vers performed by GMI, the GMI views the earth at all possible around zero latitude and longitude from −80 to −120 degrees.
azimuth angles. The only other explanation is poor out-of-band The satellites that transmit in the GMI 18 GHz at those longi-
filtering for AMSR2 10.65 GHz channel. tudes are given in the bottom plot of Fig. 10. All these satellites
The 18.7 GHz channels are affected by transmissions from are used for high-speed internet or direct broadcast TV and all
geosynchronous satellites reflecting off the Great Lakes and utilize the GMI 18-GHz band. Furthermore, discussion about
ocean surrounding the continental United States and Hawaii. the reflected RFI is found in [27].

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
DRAPER: RADIO FREQUENCY ENVIRONMENT FOR EARTH-OBSERVING PASSIVE MICROWAVE IMAGERS 1921

VI. CONCLUSION [9] D. M. Le Vine and P. de Matthaeis, “Aquarius active/passive RFI envi-
ronment at L-band,” IEEE Geosci. Remote Sens. Lett., vol. 11, no. 10,
Dealing with RFI has been an on-going struggle for mi- pp. 1747–1751, Oct. 2014. doi: 10.1109/LGRS.2014.2307794.
crowave earth remote sensing applications, especially around [10] P. N. Mohammed, M. Aksoy, J. R. Piepmeier, J. T. Johnson, and
A. Bringer, “SMAP L-band microwave radiometer: RFI mitigation
populated areas. Terrestrial-based transmitters and direct broad- prelaunch analysis and first year on-orbit observations,” IEEE Trans.
cast satellite reflections from the ocean corrupt earth remote Geosci. Remote Sens., vol. 54, no. 10, pp. 6035–6047, Oct. 2016. doi:
sensing measurements from sub-Kelvin up to hundreds of 10.1109/TGRS.2016.2580459.
[11] D. W. Draper and D. A. Newell, “An assessment of radio fre-
Kelvins. The generalized RFI index method presented in this quency interference using the GPM microwave imager,” in Proc. IEEE
paper provides moderate to high sensitivity for detecting RFI Geosci. Remote Sens. Symp., Milan, Italy, 2015, pp. 5170–5173. doi:
over both land and ocean. 10.1109/IGARSS.2015.7326998.
[12] X. Tian and X. Zou, “Television frequency interference in AMSR2
Using the AMSR-E and AMSR2 combined dataset, we find K-band measurements over reflective surfaces,” IEEE Geosci. Re-
that the RFI environment has been slowly evolving since 2002. mote Sens. Lett., vol. 13, no. 11, pp. 1621–1625, Nov. 2016. doi:
During this time, for some more established areas such as the 10.1109/LGRS.2016.2598058.
[13] E. Daganzo-Eusebio, R. Oliva, Y. H. Kerr, S. Nieto, P. Richaume, and
United States, land-based RFI has gradually decreased in extent, S. M. Mecklenburg, “SMOS radiometer in the 1400–1427-MHz pas-
while in more developing areas, its extent has widened. sive band: Impact of the RFI environment and approach to its mitigation
We find that utilizing a dual-band RFI mitigation technique and cancellation,” IEEE Trans. Geosci. Remote Sens., vol. 51, no. 10,
pp. 4999–5007, Oct. 2013. doi: 10.1109/TGRS.2013.2259179.
such AMSR-2 the C-Band channels, can be an effective method [14] C. S. Ruf, S. M. Gross, and S. Misra, “RFI detection and mitigation
of removing much of the high-level RFI. Even if explicit on- for microwave radiometry with an agile digital detector,” IEEE Trans.
board mitigation is not employed, band selection should in- Geosci. Remote Sens., vol. 44, no. 3, pp. 694–706, Mar. 2006. doi:
10.1109/TGRS.2005.861411.
volve knowledge of the environment, which can impact the spe- [15] B. Guner, J. T. Johnson, and N. Niamsuwan, “Time and frequency blank-
cific requirements of environmental data products. For example, ing for radio-frequency interference mitigation in microwave radiome-
for ocean applications, it may be beneficial to avoid the bands try,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 11, pp. 3672–3679,
Nov. 2007. doi: 10.1109/TGRS.2007.903680.
from 10.7 through 12.75 GHz, 18.3 through 18.8 GHz and 19.7 [16] J. R. Piepmeier, P. N. Mohammed, and J. J. Knuble, “a double de-
through 20.2 GHz to eliminate or minimize direct broadcast tector for RFI mitigation in microwave radiometers,” IEEE Trans.
signal reflections from the oceans. Geosci. Remote Sens., vol. 46, no. 2, pp. 458–465, Feb. 2008. doi:
10.1109/TGRS.2007.909099.
Given the widespread and changing RFI environment at the [17] S. Misra, P. N. Mohammed, B. Guner, C. S. Ruf, J. R. Piepmeier, and J. T.
C-Band through the Ku-Band, a viable long-term solution for Johnson, “Microwave radiometer radio-frequency interference detection
future sensors is to build RFI mitigation into the on-board pro- algorithms: A comparative study,” IEEE Trans. Geosci. Remote Sens.,
vol. 47, no 11, pp. 3742–3754, Nov. 2009. doi: 10.1109/
cessing. Current research as well as on-orbit missions such as TGRS.2009.2031104.
SMAP is showing much promise to such an end. [18] G. F. Forte, J. M. Tarongi-Bauza, V. dePau, M. Vall-llossera, and A. Camps,
“Experimental study on the performance of RFI detection algorithms in
microwave radiometry: Toward an optimum combined test,” IEEE Trans.
REFERENCES Geosci. Remote Sens., vol. 51, no. 10, pp. 4936–4944, Oct. 2013. doi:
10.1109/TGRS.2013.2273081.
[1] E. G. Njoku, P. Ashcroft, T. K. Chan, and L. Li, “Global survey and [19] J. Lahtinen, J. Uusitalo, T. Ruokokoski, and J. Ruoskanen, “Compari-
statistics of radio-frequency interference in AMSR-E land observations,” son of polarimetric algorithm with other algorithms in detecting radio
IEEE Trans. Geosci. Remote Sens., vol. 43, no. 5, pp. 938–947, May 2005. frequency interference,” IEEE J. Select. Top. Appl. Earth Observ. Re-
doi: 10.1109/TGRS.2004.837507. mote Sens., vol. 10, no. 7, pp. 3087–3095, Jul. 2017. doi: 10.1109/JS-
[2] L. Li, P. W. Gaiser, M. H. Bettenhausen, and W. Johnston, “WindSat TARS.2017.2680244.
radio-frequency interference signature and its identification over land and [20] G. F. Forte, J. Querol, A. Camps, and M. Vall-llossera, “Real-time RFI
ocean,” IEEE Trans. Geosci. Remote Sens., vol. 44, no. 3, pp. 530–539, detection and mitigation system for microwave radiometers,” IEEE Trans.
Mar. 2006. doi: 10.1109/TGRS.2005.862503. Geosci. Remote Sens., vol. 51, no. 10, pp. 4928–4935, Oct. 2013. doi:
[3] D. McKague, J. J. Puckett, and C. Ruf, “Characterization of K-band 10.1109/TGRS.2013.2267595.
radio frequency interference from AMSR-E, WindSat and SSM/I,” in [21] S. Misra, A. Tkacenko, S. Gowda, R. Jarnot, and S. Brown, “Implemen-
Proc. IEEE Geosci. Remote Sens. Symp., Honolulu, HI, USA, 2010, tation of a flexible wide-band on-board radio frequency interference mit-
pp. 2492–2494. doi: 10.1109/IGARSS.2010.5651860. igating digital back-end radiometer system,” in Proc. 2014 United States
[4] T. Lacava, I. Coviello, M. Faruolo, G. Mazzeo, N. Pergola, and V. Nat. Committee URSI Nat. Radio Sci. Meeting, Boulder, CO, USA, 2014,
Tramutoli, “A multitemporal investigation of AMSR-E C-band radio- pp. 1–1, doi: 10.1109/USNC-URSI-NRSM.2014.6928042.
frequency interference,” IEEE Trans. Geosci. Remote Sens., vol. 51, no. 4, [22] N. Skou, S. S. Kristensen, A. Kovanen, and J. Lahtinen, “Processor bread-
pp. 2007–2015, Apr. 2013. doi: 10.1109/TGRS.2012.2228487. board for on-board RFI detection and mitigation in MetOp-SG radiome-
[5] M. Aksoy and J. T. Johnson, “A comparative analysis of low-level ra- ters,” in Proc. 2015 IEEE Geosci. Remote Sens. Symp., Milan, Italy,
dio frequency interference in SMOS and aquarius microwave radiome- Jul., 2015, pp. 1445–1448. doi: 10.1109/IGARSS.2015.7326050.
ter measurements,” IEEE Trans. Geosci. Remote Sens., vol. 51, no. 10, [23] J. T. Johnson et al., “The cubesat radiometer radio frequency in-
pp. 4983–4992, Oct. 2013. doi: 10.1109/TGRS.2013.2266278. terference technology validation (CubeRRT) mission,” in Proc. IEEE
[6] X. Zou, X. Tian, and F. Weng, “Detection of television frequency in- Geosci. Remote Sens. Symp., Beijing, China, Jul. 2016, pp. 299–301.
terference with satellite microwave imager observations over oceans,” doi: 10.1109/IGARSS.2016.7729070.
J. Atmos. Ocean. Technol., vol. 31, pp. 2759–2776, 2014. doi: [24] A. J. Schoenwald, D. C. Bradley, P. N. Mohammed, J. R. Piepmeier,
10.1175/JTECH-D-14-00086.1. and M. Wong, “Performance analysis of a hardware implemented com-
[7] I. S. Adams, M. H. Bettenhausen, and W. Johnston, “The impact of radio- plex signal kurtosis radio-frequency interference detector,” in Proc. 14th
frequency interference on windsat ocean surface observations,” IEEE Spec. Meeting Microw. Radiometry Remote Sens. Environ., Apr. 2016,
Trans. Geosci. Remote Sens., vol. 52, no. 10, pp. 6665–6673, Oct. 2014. pp. 71–75. doi: 10.1109/MICRORAD.2016.7530507.
doi: 10.1109/TGRS.2014.2300345. [25] M. Takashi, T. Yuji, and K. Imaoka, “GCOM-W1 AMSR2 level 1R prod-
[8] E. Anterrieu, “On the detection and quantification of RFI in L1a signals uct: Dataset of brightness temperature modified using the antenna pattern
provided by SMOS,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 10, matching technique,” IEEE Trans. Geosci. Remote Sens., vol. 54, no. 2,
pp. 3986–3992, Oct. 2011. doi: 10.1109/TGRS.2011.2136350. pp. 770–782, Feb. 2016. doi: 10.1109/TGRS.2015.2465170.

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.
1922 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 11, NO. 6, JUNE 2018

[26] J. R. Piepmeier et al., “Radio-frequency interference mitigation for the [34] D. W. Draper and E. Stocker, “A comparison of radio frequency inter-
soil moisture active passive microwave radiometer,” IEEE Trans. Geosci. ference within and outside of allocated passive earth exploration bands at
Remote Sens., vol. 42, no. 1, pp. 761–775, Jan. 2014. 10.65 GHz and 18.7 GHz using the GPM microwave imager and Wind-
[27] D. W. Draper, “Report on GMI special study #15: Radio frequency in- Sat,” in Proc. IEEE Geosci. Remote Sens. Symp., Fort Worth, TX, USA,
terference,” Ball Aerosp. Technol. Corp., Boulder, CO, USA, Tech. Rep. 2017, pp. 2731–2733.
2444347, Jan. 2015. [Online]. Available: Goddard Space Flight Center.
[28] R. Castro, A. Gutierez, and J. Barbosa, “Detection and mitiga-
tion of radio frequency interference in SMOS data,” in Proc.
IEEE Geosci. Remote Sens. Symp., Jul. 2012, pp. 4575–4578. doi:
10.1109/IGARSS.2012.6350451.
[29] D. Truesdale, “A probability distribution method for detecting radio- David W. Draper (M’12) received the B.S. degree
frequency interference in windsat observations,” IEEE Trans. Geosci. and the Ph.D. degree from Brigham Young Univer-
Remote Sens., vol. 51, no. 6, pp. 3780–3788, Jun. 2013. doi: sity, Provo UT, USA, both in electrical engineering
10.1109/TGRS.2012.2223473. in 2000 and 2003, respectively.
[30] X. Zou, J. Zhao, F Weng, and Z. Qin, “Detection of radio-frequency Since 2004, he has been with Ball Aerospace and
interference signal over land from FY-3B microwave radiation im- Technologies Corp., Boulder, CO, USA, as a Sys-
ager (MWRI),” IEEE Trans. Geosci. Remote Sens., vol. 50, no. 12, tems Engineer on the Ozone Mapping and Profiler
pp. 4994–5003, Dec. 2012. doi: 10.1109/TGRS.2012.2191792. Suite (OMPS), GMI, Weather System Follow-on –
[31] S. W. Ellingson and J. T. Johnson, “A polarimetric survey of radio- Microwave (WSF-M), and various other programs.
frequency interference in C- and X-bands in the continental united states He was the calibration lead for the GMI instrument.
using WindSat radiometry,” IEEE Trans. Geosci. Remote Sens., vol. 44, He currently is the Mission Performance Lead for
no. 3, pp. 540–548, Mar. 2006. doi: 10.1109/TGRS.2005.856131. the WSF-M mission. His research interests and experience from industry and
[32] I. S. Adams, M. H. Bettenhausen, P. W. Gaiser, and W. Johnston, “Iden- academia include estimation theory, active and passive microwave remote sens-
tification of ocean-reflected radio-frequency interference using windsat ing, microwave radiometer calibration, and radiometer RFI mitigation.
retrieval chi-square probability,” IEEE Geosci. Remote Sens. Letters, Dr. Draper is a member of Tau Beta Pi, and was a fellow of that society in
vol. 7, no. 2, pp. 406–410, Apr. 2010. doi: 10.1109/LGRS.2009.2037446. 2000–2001. He was the recipient of the IEEE Geoscience and Remote Sensing
[33] D. W. Draper, “Terrestrial and space-based RFI observed by the Society Transaction Prize paper award in 2004. He was the recipient of the
GPM microwave imager (GMI) within NTIA semi-protected passive Engineering Excellence Award from Ball Aerospace and Technologies Corp
earth exploration bands at 10.65 and 18.7 GHz,” in Proc. Radio in 2015. He serves as part of the of the Precipitation Measurement Missions
Freq. Interference 2016, Socorro, NM, USA, 2016, pp. 26–30. doi: (PMM) Intersatellite Calibration (X-CAL) working group, which received the
10.1109/RFINT.2016.7833526. 2015 PMM Science Team Award from NASA.

Authorized licensed use limited to: Institut français l'exploitation Mer (IFREMER). Downloaded on January 26,2024 at 23:15:03 UTC from IEEE Xplore. Restrictions apply.

You might also like