IEEE Geoscience and Remote Sensing Magazine - March 2020 - 124

CHALLENGES TO MONITORING WETLAND
WATER-LEVEL CHANGES
RELATIVE WATER-LEVEL CHANGES
InSAR essentially measures the modulo-2r wrapped phase.
Therefore, 2D phase unwrapping (PU) is required to obtain
2h/2t. However, PU is an ill-posed inverse problem because
there are two unknowns (the absolute phase and integer
ambiguity) to be solved with one equation [35], such as

} = { + 2kr, (6)

where } is the unwrapped phase, { is the wrapped phase,
and k is the integer ambiguity.
In the solid-Earth deformation application, the two unknowns can be solved by selecting a far-field point as a reference, with the assumption that it has no deformation [31].
However, this premise does not apply to wetlands because
levees and channels can discontinue water-level changes. In
addition, the water depth fluctuates across the whole wetland [20], [31]. In the wetland-hydrology application, an
arbitrary reference point must first be selected for unwrapping, and, thus, the unwrapped phase has a constant difference from the true unwrapped phase, such that

} = }l + } 0, (7)

where } is the true unwrapped phase, }l is the unwrapped
phase with respect to a reference point, and } 0 is the absolute phase of the reference point.
After converting the phase into water-level changes using (3), an offset is needed to convert the InSAR-derived
relative water-level changes from }l into absolute waterlevel changes, such as

2h/2t = 2h InSAR /2t + 2h 0 /2t, (8)

where 2h/2t is the absolute water-level change, 2h InSAR /2t
is the relative water-level change from InSAR with an arbitrary point as a reference, and 2h 0 /2t is a constant offset
equivalent to the water-level change at the reference point.
Currently, there are three methods to solve this problem.
The first is to use in situ gauge measurements to estimate
2h 0 /2t . This method's concept is relatively straightforward.
The 2h 0 /2t is estimated by taking the differences between
2h/2t and 2h InSAR /2t at one or more points using (3). Alsdorf
et al. [9] compared 2h InSAR /2t from JERS-1 interferograms in
the Amazon wetlands and 2h/2t from in situ gauges in the
river. They found that the relative increase from 2h InSAR /2t
agreed with that from the gauges. They added one constant
to 2h InSAR /2t so that the calibrated interferometric measurements became equivalent to the gauge measurements.
Wdowinski et al. [16], [17] compared 2h InSAR /2t from JERS1 interferograms with in situ gauge measurements in the
Everglades and found that the two measurements showed
a similar trend with a constant gap between them. An offset was calculated by taking the difference between the two
measurements and adding it back to 2h InSAR /2t to obtain
2h/2t. Using in situ measurements is ideal to obtain 2h/2t.
124

However, not all wetlands are equipped with gauges, and
data availability becomes the main limitation.
The second method uses the measurements from radar
altimetry instead of in situ gauges to obtain 2h 0 /2t . Alsdorf
et al. [12] compared the measurements from a JERS-1 interferogram and Topex/Poseidon radar altimetry across an Amazon lake and found that the two measurements generally
agreed. Lu et al. [36] used the measurements from Envisat
altimetry to resolve the integer-phase ambiguity in the PALSAR interferogram across the Helmand wetlands. Kim et al.
[20] used Envisat radar-altimetry measurements to calibrate
PALSAR and Radarsat-1 interferograms across the Louisiana
wetlands. The generated 2D 2h/2t maps agreed well with the
in situ gauge measurements. Zhang et al. [37], [38] and Yuan
et al. [13] compared 2h InSAR /2t from PALSAR interferograms
with the Envisat radar-altimetry measurements (2h alt /2t) at
the Liao River delta and Congo wetlands, respectively. Yuan
et al. [13] showed that 2h InSAR /2t and 2h alt /2t revealed a
similar spatial variation with a constant bias between them,
similar to the comparison with in situ measurements [16],
[17]. This method can be useful when there are no in situ
gauge measurements. However, due to the fact that radar
altimetry is a 1D profiling instrument, it is not applicable
to wetlands located between its ground tracks [39]. It is also
challenging to use altimetry data as the reference across
highly dynamic coastal wetlands and wetlands that humans
intensively manipulate. This is because SAR images and altimetry data are usually obtained at different times, and the
interpolation of altimetry measurements with respect to the
SAR acquisition times cannot correctly represent the actual
temporally dynamic water levels since the revisiting period
of radar altimetry is longer than 10 days [40].
The third method is to use the SAR backscattering coefficient (v 0) to estimate the integer ambiguity for 2h InSAR /2t .
Kim et al. [40] found the JERS-1 SAR v 0 correlated with the
tide-level changes at a fish farm that had vertical wood poles
installed at the sea bottom. The correlation was then used to
solve the integer ambiguity during the PU process. Although
the study site in [40] was a fish farm, it had a vertical structure to enable the SAR signals to follow the double-bounce
backscattering, which is similar to the backscattering mechanism in wetlands. Currently, there is no study using SAR v 0
to solve the integer ambiguity in wetlands. The relationship
between SAR v 0 and water-level changes is mostly found
from studies using L-band SAR [41]. The challenges of using
SAR v 0 to solve the InSAR phase ambiguity relate to constructing the relationship between SAR v 0 and water-level
changes, which significantly depends on the spatial variation of the vegetation type and density [42]-[44].
DECORRELATION IN WETLANDS
The temporal dynamics of the water level and vegetation
changes are the main factors that cause InSAR decorrelation in wetlands. Several studies have developed algorithms to cope with this unique challenge to InSAR
processing. Hong et al. [31] proposed the small temporal
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

MARCH 2020

IEEE Geoscience and Remote Sensing Magazine - March 2020

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2020