Received: 23 May 2019 Revised: 20 April 2020 Accepted: 10 May 2020

DOI: 10.1002/navi.375

ORIGINAL ARTICLE

Characterization and mitigation of interference between

GNSS radio occultation and reflectometry signals for
low-altitude occultations

Ian Collett Y. T. Jade Morton Yang Wang Brian Breitsch

Department of Aerospace Engineering

Sciences, University of Colorado, Boulder, Abstract
CO, USA This research investigates the mutual interference between a GNSS signal
Correspondence
refracted through the atmosphere and the same signal reflected from the surface
Ian Collett, Department of Aerospace below. For low-altitude occultations, the reflected signal can be present as multi-
Engineering Sciences, University of Col- path when tracking the direct signal. This poses a problem when making remote
orado, Boulder, CO, USA.
Email: Ian.Collett@colorado.edu sensing measurements using GNSS-RO or GNSS-R techniques. Here, this issue
is addressed in the context of GNSS data collected by a mountaintop receiver on
the Hawaiian island of Maui. First, an elevation-angle threshold, below which
the reflected signal can be present as multipath on the direct signal, is calcu-
lated for the mountaintop geometry. Next, the first steps are taken to mitigate
the mutual interference between the direct and reflected signals. A frequency-
domain filtering approach and a time-domain smoothing approach are applied
to the signal intensity in an effort to separate the multipath contribution and the
direct signal contribution.

1 INTRODUCTION high impact on weather patterns. In this case, the reflected

signal will have a low-grazing angle (the angle between the
As a signal propagates from the GNSS satellite to the local surface and the incident ray of the reflection). Low-
receiver, it interacts with its environment—for exam- grazing angle reflections are well-suited for carrier phase
ple, the signal is refracted through the atmosphere and altimetry (Semmling et al., 2016). Furthermore, reflections
scattered from Earth’s surface below. Though these inter- at low-grazing angles occur significantly more frequently
actions are a nuisance when using GNSS for its originally than those at high-grazing angles (Southwell & Dempster,
intended purpose of navigation, they can be leveraged 2018). Clearly, the ability to make GNSS-RO and GNSS-
for remote sensing (Jin & Komjathy, 2010). GNSS radio R measurements from low-altitude occultations is of
occultation (GNSS-RO) obtains atmospheric parameters importance.
by measuring the amount of bending in the refracted However, these low-altitude occultations pose a techni-
signal (Kursinski, Hajj, Schofield, Linfield, & Hardy, 1997). cal challenge. In this scenario, it is possible for the path
GNSS reflectometry (GNSS-R) uses the reflected signal for length difference between the direct and reflected signals
purposes of scatterometry (e.g., to retrieve ocean surface to be small enough so that the reflected signal is present
winds or land vegetation characteristics) and altimetry as multipath when tracking the direct signal. Thus, the
(Zavorotny, Gleason, Cardellach, & Camps, 2014). One sce- possibility of making accurate remote sensing measure-
nario of particular interest to the remote sensing of Earth’s ments from low-altitude occultations hinges on the ability
atmosphere is the low-altitude occultation, enabling mea- to predict and mitigate the mutual interference between
surements of the moist lower troposphere, which has a the direct and reflected signals.

NAVIGATION. 2020;67:537–546. wileyonlinelibrary.com/journal/navi © 2020 Institute of Navigation 537

538 COLLETT et al.

F I G U R E 1 The signals are reflected from the dish and received by a horn antenna. The horizontal (H) and vertical (V) polarization outputs
of the horn antenna are converted to RHCP and LHCP using 90-degree hybrid combiners, and raw data are recorded using six separate USRPs
[Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com and www.ion.org]

The research described here addresses these two needs GPS signals were collected: L1 and L2 sampled at 5 MHz,
through the analysis of data recorded by an antenna dish and L5 sampled at 25 MHz. For low-elevation satellites,
placed on the summit of a mountain and directed towards both the direct and reflected signals are within the 7-degree
GPS satellites rising over the horizon. The first goal is to half power beam width of the antenna.
calculate an elevation-angle threshold 𝑒𝑙𝑡ℎ below which USRP N200 software-defined radios are used to record
the reflected signal is present as multipath. To calculate raw data for the three frequency bands on RHCP and
𝑒𝑙𝑡ℎ , the code structure of the signals is considered, and the LHCP polarization channels. Although GPS signals typi-
path length difference between the direct and reflected sig- cally flip polarization from RHCP to LHCP upon reflection
nals is calculated considering the altitude of the mountain- from Earth’s surface, reflections for extremely low-grazing
top receiver. The second goal is to devise a technique for angles (smaller than 3 degrees) retain a significant amount
separating the direct and reflected signals when operating of power within RHCP. Thus, the geometry of this experi-
below 𝑒𝑙𝑡ℎ . Here, a frequency-domain filtering approach ment allows the direct and reflected signals to be recorded
and a time-domain smoothing approach are applied to on the same polarization channel. Due to the reflection
the tracked signal intensity (SI) in an effort to separate from the dish into a horn antenna, both the direct and
the effects of the direct and reflected signals. The smooth- reflected signals flip polarization from RHCP to LHCP just
ing approach, which outperforms the filtering approach at before reception. The data used in this paper come from
lower elevation angles, is further analyzed through a sim- tracking the data recorded on the LHCP channel (USRP 4,
ulation study. 5, and 6). The signals are tracked using a 0.5 chip corre-
lator spacing and a 100 Hz time series of the SI for each
frequency is computed in post processing from the I and Q
2 DATA COLLECTION correlator outputs. The data are divided into two datasets:
dataset 1 contains GPS L1, L2, and L5 SI from a single rising
The experiment, illustrated in Figure 1, was performed in occultation event (PRN 3 from approximately 8:00 to 8:50
May 2017 near the summit of Haleakala on the Hawaiian UTC on May 26th ) and is used to illustrate the methodol-
island of Maui (latitude: 20◦ 42’09’’ N, longitude: 156◦ 15’24’’ ogy and qualitatively compare the filtering and smoothing
W, altitude: 3060 m) (Morton et al., 2017). A 1.9-meter approaches. Dataset 2 contains GPS L1-only SI for ten addi-
antenna dish (27 dB gain at GPS L1) was steered to track tional rising occultation events (recorded May 4th through
GNSS satellites near the horizon. In addition to signals 6th ) and is used to facilitate the simulation study to quan-
from other GNSS constellations, three frequency bands of tify the performance of the smoothing approach.
COLLETT et al. 539

F I G U R E 2 Geometry used to determine the path length difference as a function of satellite elevation angle [Color figure can be viewed
in the online issue, which is available at wileyonlinelibrary.com and www.ion.org]

3 METHODOLOGY The length of the line segments are

For a 0.5 chip spacing between the early, prompt, and late ℎ + 𝑟𝑒 (1 − cos 𝜃)
𝑎=
correlators in the tracking loop of the direct signal, multi- sin(2𝜃 + 𝑒𝑙)
path effects can persist until the path length difference is
( )
1.5 chips (Misra & Enge, 2010). So, 𝑒𝑙𝑡ℎ is the elevation angle ℎ + 𝑟𝑒 (1 − cos 𝜃)
𝑏= cos(2𝜃 + 2𝑒𝑙).
of the transmitter relative to the receiver for which the sin(2𝜃 + 𝑒𝑙)
path length difference is 1.5 chips. When the transmitter is
below this elevation, the reflected signal can be present as The path length difference is therefore
multipath in the tracking results of the direct signal. This ( )
threshold depends primarily on the height of the receiver ℎ + 𝑟𝑒 (1 − cos 𝜃)
Δ𝑅 = (1 − cos(2𝜃 + 2𝑒𝑙)).
above the surface and the chipping rate of the signal. As a sin(2𝜃 + 𝑒𝑙)
first approximation, 𝑒𝑙𝑡ℎ is determined by approximating
the Earth as a sphere and assuming that the incident ray The theoretical path length difference as a function of
of the reflection is parallel to the direct signal path (see the satellite elevation angle is plotted in Figure 3.
Figure 2). The thresholds are set at 1.5 times the chip width: L1 and
Here, 𝑟𝑒 is the radius of the Earth, 𝑒𝑙 is the elevation L2 at 439.57 m and L5 at 43.96 m. Based on these approxi-
angle of the transmitter relative to the receiver, and ℎ is the mate results, 𝑒𝑙𝑡ℎ is 4.03˚ for L1 and L2 and –0.41˚ for L5.
receiver’s height above the surface. The path length differ- With these theoretical results to serve as a point of
ence is the difference between the length of the line seg- comparison, the relevant geometrical parameters are now
ments 𝑎 and 𝑏: determined for dataset 1 (see Figure 4). The path length
difference is calculated assuming specular reflection. The
Δ𝑅 = 𝑎 − 𝑏. specular point is the point on the Earth’s surface that min-
imizes the total reflected path length (Jales, 2012). This
First, the angle 𝜃 is determined by numerically solving point is determined by approximating the Earth’s surface
the equation using the WGS84 ellipsoid and solving the minimization
problem. Once the location of the specular point is known,
ℎ + 𝑟𝑒 (1 − cos 𝜃) the path length difference can be easily determined.
= tan(2𝜃 + 𝑒𝑙).
𝑟𝑒 sin 𝜃 Figure 4 also includes the time rate of change of the path
540 COLLETT et al.

as a superposition of the direct and reflected signals (e.g.,

in Byun, Hajj, & Young, 2002):

𝑠(𝑡) = 𝐴𝑑 (𝑡)𝑒𝑗2𝜋𝜙𝑑 (𝑡) + 𝐴𝑟 (𝑡)𝑒𝑗2𝜋𝜙𝑟 (𝑡) ,

where 𝐴𝑑 (𝑡) and 𝜙𝑑 (𝑡) are the time-dependent amplitude

and phase of the direct signal, and 𝐴𝑟 (𝑡) and 𝜙𝑟 (𝑡) are the
time-dependent amplitude and phase of the reflected sig-
nal. The phase terms are expressed in units of cycles. This
model assumes that the signal is only reflected from the
specular point. The power of 𝑠(𝑡) is

𝑃 = 𝐴𝑑 (𝑡)2 + 𝐴𝑟 (𝑡)2 + 2𝐴𝑑 (𝑡)𝐴𝑟 (𝑡) cos(2𝜋(𝜙𝑟 (𝑡) − 𝜙𝑑 (𝑡))).

The argument of the cosine term corresponds to the inter-

F I G U R E 3 Path length difference as a function of satellite ele- ferometric frequency that appears in the spectrograms.
vation angle, calculated assuming spherical Earth and parallel inci- This frequency follows the time rate of change of the phase
dence [Color figure can be viewed in the online issue, which is avail- difference (equivalently, the time rate of change of the path
able at wileyonlinelibrary.com and www.ion.org] length difference divided by the signal wavelength):
( )
𝑑 1 𝑑(Δ𝑅)
[𝜙 (𝑡) − 𝜙𝑑 (𝑡)] = .
𝑑𝑡 𝑟 𝜆 𝑑𝑡

The time rate of change of the path length difference

depicted in Figure 4 is normalized by the carrier wave-
length of the corresponding signal so that it can be com-
pared directly to the spectrogram. Figure 8 shows the spec-
trograms for dataset 1 with the theoretical interferometric
frequency overlaid.
The other primary goal of this research is to devise a
technique to separate the contributions to the SI of the
direct and reflected signals. A simple approach is to apply
a filter to the SI with a 0.5 Hz cutoff (determined visually
from the spectrogram). Two different filters are considered:
F I G U R E 4 Relevant geometrical parameters for the time inter- a third-order Butterworth filter and a fifth-order Cheby-
val of interest in dataset 1. The times are referenced to the beginning shev type I filter. Figure 5 illustrates the L1 SI spectrogram
dataset [Color figure can be viewed in the online issue, which is avail- and the magnitude response of the filters used to remove
able at wileyonlinelibrary.com and www.ion.org]
or isolate the multipath contribution. To remove the mul-
tipath component and retain tropospheric features, a low-
length difference, which is later used to model the interfer- pass filter is used. To retain the multipath component and
ometric frequency introduced by the reflected signal acting remove tropospheric features, a high-pass filter is used.
as multipath. In all cases, the filter is applied forwards and backwards
The value of 𝑒𝑙𝑡ℎ is 3.94˚ for L1 and L2 and –0.43˚ for L5, to yield a linear phase response. The filter magnitude
which agrees well with the earlier results found using the response depicted in Figure 5 is of the forwards-backwards
spherical Earth approximation. According to these results, filters (i.e., the square of the magnitude response of the
multipath should no longer be present in the tracking filter applied once). A 1 dB ripple in the passband of the
results of the direct signal after the 30-minute mark for L5 Chebyshev filter is arbitrarily chosen—this value could be
and after the 45-minute mark for L1 and L2. tweaked if more specific requirements for the accuracy of
Spectrograms of the tracked SI are used to investigate the filtered result were known. In the end, the Chebyshev
the mutual interference between the direct and reflected filter is chosen because of its sharper cutoff.
signals. To understand the spectral content of the SI, it is The other technique devised for removal or isolation
necessary to develop a model of the power of the received of the multipath contribution is a time-domain smooth-
signal. Here, the composite received signal 𝑠(𝑡) is modeled ing approach. Here, the peaks and troughs of the SI are
COLLETT et al. 541

signal that incorporates realistic tropospheric scintillation

is simulated by selecting a portion of the low-pass filtered
SI where the multipath is clearly removed. Then, the SI
of the combined signal is simulated using the aforemen-
tioned model of the composite received signal and com-
puting its power. In this case, 𝐴𝑑 (𝑡) is the square root of the
simulated direct SI, 𝐴𝑟 (𝑡) is 𝐴𝑑 scaled by a constant factor
determined by the Fresnel reflection coefficient, and the
phase difference between 𝜙𝑑 (𝑡) and 𝜙𝑟 (𝑡) is determined by
ΔR for the given event. The performance of the smoothing
technique is measured by how well the simulated direct
signal can be extracted from the simulated combined sig-
nal. This is quantified by calculating the percent error over
a range of elevation-angle bins. The simulation procedure
is illustrated in Figure 7.

F I G U R E 5 The L1 SI spectrogram from dataset 1 and the mag-

nitude response of the forwards-backwards filters (Butterworth and 4 RESULTS AND ANALYSIS
Chebyshev type I) [Color figure can be viewed in the online issue,
which is available at wileyonlinelibrary.com and www.ion.org] First, the single event from dataset 1 containing SI for
GPS L1, L2, and L5 is used to qualitatively compare the
located using a wavelet-based peak-finding algorithm (Du, interference mitigation techniques. The spectrograms of
Kibbe, & Lin, 2006). Splines are then used to interpo- the raw SI and the theoretically determined interferomet-
late the points of the maxima and minima to yield the ric frequency curves are depicted in Figure 8. (Although
upper and lower envelopes of the SI. As illustrated in Fig- only the spectrograms shown here are for dataset 1, the
ure 6, which uses an example of the L1 SI from dataset L1 SI spectrograms for the ten events in dataset 2 exhibit
1, the smoothed SI is the average of the upper and lower the same behavior.) Recall that the multipath should drop
envelopes. out by 45 minutes for L1 and L2 and by 30 minutes for
The later results show that the filtering technique suc- L5. In all three cases, the multipath contribution fades
cessfully removes the multipath contribution when the out before this point. For L5, the multipath contribution
interferometric frequency is clearly distinguishable. The is hardly distinguishable from the fluctuations caused by
smoothing approach shows promise at lower elevations tropospheric scintillation (the faint trend in L5 after 30
where the filtering technique begins to fail. A simulation minutes that matches the interferometric frequency curve
study using dataset 2 is performed to quantify the perfor- is likely due to sidelobes in the autocorrelation function
mance of the smoothing approach. First, the SI for a direct of the code). So, a threshold elevation angle has been

F I G U R E 6 The time-domain smoothing technique involves finding the peaks and troughs of the SI, interpolating curves through these
points, and taking the average of the upper and lower envelopes [Color figure can be viewed in the online issue, which is available at wileyon-
linelibrary.com and www.ion.org]
542 COLLETT et al.

F I G U R E 7 Steps of the simulation procedure. This is illustrated here using the L1 SI from dataset 1; the direct signal is simulated using
the low-pass filtered SI between minutes 28 and 38 [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com
and www.ion.org]

F I G U R E 8 Spectrograms of the raw SI for GPS L1, L2, and L5 signals for dataset 1 [Color figure can be viewed in the online issue, which
is available at wileyonlinelibrary.com and www.ion.org]

successfully established for the geometry of the moun- bution between 20 and 25 minutes. After 25 minutes, the
taintop receiver. GNSS satellite is higher above the horizon, and the fre-
There are several other interesting features of the quency of the multipath contribution is well-predicted by
spectrograms that should be noted. Between 20 and the interferometric frequency curve. At this point, it is also
25 minutes, the GNSS satellite is barely above the horizon. easier to distinguish between the multipath contribution
During this time interval, the direct and reflected signals and the direct signal contribution.
undergo a significant amount of bending, resulting in a dis- The effectiveness of the separation techniques is ana-
crepancy between the true interferometric frequency and lyzed first by looking at the resulting SI spectrograms.
the theoretical curve. A model that incorporated bending Figure 9 shows the SI spectrograms after the low-pass
effects would be able to more faithfully reproduce the fre- Chebyshev filter is applied. The obvious multipath con-
quency of the multipath contribution. Also note the diffi- tribution has been removed after minute 25. However,
culty of distinguishing between the frequency fluctuations between 20 and 25 minutes, the frequency of the multi-
due to tropospheric scintillation and the multipath contri- path contribution is below 0.5 Hz. Here, the filter cannot
COLLETT et al. 543

F I G U R E 9 Spectrograms of the SI from dataset 1 after applying the low-pass Chebyshev type I filter [Color figure can be viewed in the
online issue, which is available at wileyonlinelibrary.com and www.ion.org]

F I G U R E 1 0 Spectrograms of the smoothed SI for dataset 1 [Color figure can be viewed in the online issue, which is available at wileyon-
linelibrary.com and www.ion.org]

remove the multipath, and thus, the SI still contains con- the SI after applying the high-pass Chebyshev filter are
tributions from the direct and reflected signals. depicted in Figure 11. Again, the filtering approach is suc-
Figure 10 shows the spectrograms for the smoothed SI. cessful after 25 minutes where the multipath contribution
The smoothing technique is analogous to low-pass filter- is more easily distinguishable from the direct signal contri-
ing because it attempts to remove the multipath contribu- bution. Between 20 and 25 minutes, the filter fails to isolate
tion. Again, the multipath contribution is removed after 25 the multipath contribution, removing it instead.
minutes. Between 20 and 25 minutes, it appears that this The multipath contribution can also be isolated by
approach has had more success in removing the multipath subtracting the smoothed SI from the raw SI. Figure 12
than the filtering approach. The extent of this success is depicts the spectrograms for the smoothing approach
analyzed later via simulation. for multipath isolation. In this case, it appears that the
Next, the performance of the mitigation techniques for multipath contribution is retained even between 20 and
multipath isolation is considered. The spectrograms of 25 minutes.
544 COLLETT et al.

F I G U R E 1 1 Spectrograms of the SI for dataset 1 after applying the high-pass Chebyshev type I filter [Color figure can be viewed in the
online issue, which is available at wileyonlinelibrary.com and www.ion.org]

F I G U R E 1 2 Spectrograms of SI minus smoothed SI for dataset 1 [Color figure can be viewed in the online issue, which is available at
wileyonlinelibrary.com and www.ion.org]

The difference between the filtering and smoothing results for the filtering and smoothing approaches are sim-
approaches can be further illustrated by analyzing the SI ilar.
time series. The time series between 22 and 22.5 minutes Finally, the ten events from dataset 2 are used to per-
is shown in Figure 13(a). Here, the low-pass filter fails to form the simulation study to further explore the perfor-
remove the fluctuations corresponding to the multipath mance of the smoothing technique. Each data point in
because the frequency of these fluctuations is smaller than Figure 14 represents the average percent error between
0.5 Hz. The smoothing approach, on the other hand, suc- the smoothed result and the original simulated direct sig-
ceeds in removing the fluctuations. Figure 13(b) shows the nal across a 0.1-degree elevation-angle bin. The thicker
L1 SI time series between 28 and 28.5 minutes. At this point, black curve is the average across all events. In the -1-
the SI fluctuations occur at a high enough frequency to to 0-degree elevation-angle range, the smoothing tech-
be removed with the filtering approach. In this case, the nique can recover the direct signal SI with less than 5%
COLLETT et al. 545

F I G U R E 1 3 The filtering and smoothing approaches are compared by looking at the L1 SI time series from dataset 1 between (a) 22 and
22.5 minutes, and (b) 28 and 28.5 minutes [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com and
www.ion.org]

F I G U R E 1 4 The percent error quantifies how well the smoothing technique can retrieve the direct SI. Each of the lighter colored curves
represents the results for an event from dataset 2. The thicker black curve is the average across all events [Color figure can be viewed in the
online issue, which is available at wileyonlinelibrary.com and www.ion.org]

average error—this is a clear improvement over the fil- In the future, a more robust mitigation technique
tering technique, which cannot reliably perform below 0 will be developed and applied to other signal parame-
degrees elevation. ters in addition to the SI. For example, performing car-
rier phase altimetry requires knowing the phase differ-
ence between the direct and reflected signals. A more
5 CONCLUSION advanced technique would ideally be able to simultane-
ously estimate the amplitudes and phases of the direct and
In this paper, an elevation-angle threshold was defined and reflected signals that make up the composite received sig-
calculated as a means of determining when the reflected nal. The effects of tropospheric scintillation on the accu-
signal will be present as multipath in the tracking results racy of the amplitude and phase estimation will also be
of the direct signal. Furthermore, the first steps were taken considered.
in devising techniques to mitigate the mutual interfer-
ence between these signal contributions that occurs below
this threshold. In general, the filtering approach is only AC K N OW L E D G M E N T S
successful when the interferometric frequency is well- The data collection experiment was funded by AFRL grant
separated from the frequency contributions caused by tro- #FA8650-14-1735. The data was collected by Steve Taylor
pospheric scintillation. The smoothing approach shows and Harrison Bourne from the Satellite Navigation and
promise when the frequencies are not easily separable, Remote Sensing Lab at the University of Colorado Boul-
a result corroborated by the simulation study. Still, fur- der. Dr. Frank van Graas from Ohio University and Mr.
ther analysis is required to more fully characterize its Neeraj Pujara from AFRL also assisted in the data collec-
performance. tion experiment.
546 COLLETT et al.

ORCID Morton, Y., Bourne, H., Taylor, S., Xu, D., Yang, R., van Graas,
Ian Collett https://orcid.org/0000-0002-6948-888X F., & Pujara, N. (2017, May 1–4). Keynote: Mountain-top radio
Y. T. Jade Morton https://orcid.org/0000-0001-9173-2888 occultation with multi-GNSS signals: Experiment and preliminary
results. In Proceedings of the ION 2017 Pacific PNT Meeting, Hon-
Yang Wang https://orcid.org/0000-0001-9169-2863
olulu, Hawaii (pp. 490–499). https://doi.org/10.33012/2017.15073
Brian Breitsch https://orcid.org/0000-0003-4041-1827
Semmling, A. M., Leister, V., Saynisch, J., Zus, F., Heise, S., & Wick-
ert, J. (2016). A phase-altimetric simulator: Studying the sensi-
REFERENCES tivity of earth-reflected GNSS signals to ocean topography. IEEE
Byun, S. H., Hajj, G. A., & Young, L. E. (2002). Development and Transactions on Geoscience and Remote Sensing, 54(11), 6791–6802.
application of GPS signal multipath simulator. Radio Science, https://doi.org/10.1109/TGRS.2016.2591065
37(6), 1098. https://doi.org/10.1029/2001RS002549 Southwell, B. J., & Dempster, A. G. (2018). A new approach to
Du, P., Kibbe, W. A., & Lin, S. M. (2006). Improved peak detection determine the specular point of forward reflected GNSS signals.
in mass spectrum by incorporating continuous wavelet transform- IEEE Journal of Selected Topics in Applied Earth Observations and
based pattern matching. Bioinformatics, 22(17), 2059–2065. https: Remote Sensing, PP(99), 1–8. https://doi.org/10.1109/JSTARS.2017.
//doi.org/10.1093/bioinformatics/btl355 2775647
Jales, P. (2012). Spaceborne receiver design for scatterometric GNSS Zavorotny, V. U., Gleason, S., Cardellach, E., & Camps, A. (2014).
Reflectometry. [Doctoral dissertation, University of Surrey]. Surrey Tutorial on remote sensing using GNSS bistatic radar of oppor-
Research Insight Open Access. http://epubs.surrey.ac.uk/807880/ tunity. IEEE Geoscience and Remote Sensing Magazine, 2(4),
Jin, S., & Komjathy, A. (2010). GNSS reflectometry and remote 8–45. https://doi.org/10.1109/MGRS.2014.2374220
sensing: New objectives and results. Advances in Space Research,
46(2), 111–117. https://doi.org/10.1109/TGRS.2016.259106510.1016/j.
asr.2010.01.014
Kursinski, E. R., Hajj, G. A., Schofield, J. T., Linfield, R. P., & Hardy, How to cite this article: Collett I, Jade Morton
K. R. (1997). Observing Earth’s atmosphere with radio occulta- YT, Wang Y, Breitsch B. Characterization and
tion measurements using the Global Positioning System. Jour-
mitigation of interference between GNSS radio
nal of Geophysical Research: Atmospheres, 102(D19), 23429–23465.
https://doi.org/10.1029/97JD01569
occultation and reflectometry signals for
Misra, P., & Enge, P. (2010). Global positioning system: Signals, mea- low-altitude occultations. NAVIGATION.
surements, and performance (revised 2nd ed.). Lincoln, Mass: 2020;67:537–546. https://doi.org/10.1002/navi.375
Ganga-Jamuna Press.