IEEE Geoscience and Remote Sensing Magazine - December 2014 - 35

observables. Brief summaries of the algorithms and techniques are compiled below, together with their range of
applicability. Only the main principles behind each of the
techniques are presented, and bibliographic references are
given to the source documents. Some of them are shown
in Fig. 25.

waveform. The "size" can be defined in different ways, such
as the extension of its delay and/or Doppler spread, or its area
and/or volume within a given threshold. These techniques
can be applied from airborne and spaceborne platforms, but
they lose sensitivity at low altitudes.
A stochastic theory that results in two algorithms to retrieve surface winds was developed in [174]. They relate the
sea roughness conditions to both the Doppler spread and
the delay spread of the reflected signals. For example, the
bandwidth at 3-dB level of the cumulative Doppler values is
proportional to the roughness MSS and the elevation angle:

Power (a.u.)

(a) Fitting theoretical models
After renormalizing and realigning the delay-waveform,
the best fit against a theoretical model gives an estimate for
the geophysical and instrumental-correction parameters.
As seen in Section IV, the geophysical parameters typically
(44)
B 3dB ? MSSsin ^ e h .
are the 10-meter altitude wind speed, or isotropic sea-surface slopes' variance (mean square slopes-MSS). Some of
The volume under the normalized or calibrated DDM,
the work done with this methodology include: (e.g., [17],
or the area under the normalized waveform up to a pre[18], [20], [167], [173]). The fit also can be constrained to
determined threshold change depending on the surface
the trailing edge of the waveform, as suggested in [19], and
roughness. This hypothesis was first suggested and tested
implemented in [17]. This technique works best for airborne
in simulated data by [169] and experimentally confirmed
altitudes and is not feasible for ground-based receivers bein [172]. This approach might be valuable for potential use
cause the effect of waveform spreading due to roughness is
of GNSS-R observations in support of oceanic L-band radiminished for low altitudes.
diometric missions, such as SMOS [175].
It can be extended to a multi-satellite common inversion
Recently, more elaborate DDM metrics have been prousing several GNSS reflections arriving from close areas charposed to retrieve ocean wind speed [176] and wind vector
acterized by the same surface roughness state. This approach
[177] from GNSS-R experiments onboard a high-altitude jet
estimates wind direction (or roughness anisotropy) with 180c
aircraft. The wind-speed retrieval based on the "taxicab disresidual ambiguity. It was suggested and tested in airborne
tance" from the DDM center of the mass to the DDM maxicampaigns in [16], [18]. It is rather straightforward for the case of
mum position gave the best result. It was found that winda single airborne platform, when different satellite reflections
speed retrievals based on "distance"-related observables
arrive to the receiver from relatively close surface areas with
provide more accurate wind-speed estimates compared to
the presumably same surface roughness state. For spaceborne
the retrieval based on the DDM-area metrics. The reason is
receiving systems, this method would lose
spatial resolution because the trailing edge
6
of the reflected waveform originates from
a very large glistening zone. Instead, one
5
D
op
DDM Volume
can look into the change of the peak power
4
pl
er
Above Certain Threshold
as suggested in [52]. The CYGNSS mission
Ba
3
nd
baseline retrieval algorithms will be mainly
w
2
id
th
based on this approach [116]. In order to
Fitting a Model
D
1
op
estimate wave anisotropy (wind direction)
pl
er
0
from space a multi-satellite constellation
Sp
ec
would be required to invert simultaneously
tru
-100
m
-50
(or close in time) several GNSS reflections
0
assuming they all sense the same surface
50
100
roughness at different angles.
375
188
0
-188
Alternatively, a possibility for obtain-375
DDM Area
at Certain Threshold
ing wave anisotropy information from a
single GNSS reflection observation by us-100
ing the Doppler domain of the signal, that
-50
is, the delay-Doppler map, was discussed in
0
50
[168], [49]. Then, the applicability extends
100
375
tScatt
to spaceborne systems (in addition to air188
0
-188
-375
Delay (m)
borne scenarios).
y
nc
ue
eq
Fr

z)
(H

(b) ddm metrics
The sea-surface roughness information
can also be obtained from the "size" of the
december 2014

Figure 25. Several of the roughness/wind-retrieval algorithms are sketched here, using

as an example an iGNSS-R DDM obtained from a real experimental flight. The data are
available in [143].

ieee Geoscience and remote sensing magazine

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - December 2014