IEEE Geoscience and Remote Sensing Magazine - March 2020 - 125

baseline subset (STBAS) method to generate interferogram
subsets based on the fact that InSAR coherence across
wetlands can be better maintained with shorter temporal
baselines, especially for C-band SAR data. Hong et al. [31]
used 13 in situ gauge measurements across a water-conservation area in the Everglades to calibrate 2h InSAR /2t . The obtained 2D 2h/2t was then tied to the vertical data of the in
situ gauges and, finally, converted to absolute water-level
maps. Later, Hong and Wdowinski developed a multipath
STBAS algorithm using ALOS PALSAR data to generate water-level maps of the Everglades, located at the intersection
of four ALOS paths [45]. By using the interferograms from
those paths, this method improves the temporal resolution
of PALSAR InSAR from 46 days in a single-path algorithm
to as short as seven days in a multipath STBAS algorithm.
Xie et al. [46] proposed a distributed-scatter method that
formed an optimum interferogram network and applied
spatial adaptive filtering to map 2h/2t at the coastal wetlands in the Yellow River deltas. The InSAR observation was
tied to a reference frame to generate the 2h/2t time series,
which was converted to water-depth time series.
These three methods expand the ability of InSAR to extract information across wetlands by forming an optimum
subset, using interferograms from as many paths as possible, and applying a filter. They are useful when InSAR can
maintain an adequate coherence but challenging when the

68° W

67° W

66° W

coherence is lost. For instance, InSAR coherence can barely
be maintained when vegetation is submerged beneath water
and little energy is backscattered to the antenna [34], [42].
PHASE UNWRAPPING
To obtain the correct and unique PU result, the traditional
2D method must obey the phase-continuity assumption,
which is also called the Itoh condition (that is, the phase difference between neighboring pixels must be fewer than r
[47]). In other words, the phase-continuity assumption essentially requires the observed phase to have spatial continuity to estimate the phase gradient between each pair
of neighboring pixels. Because of the phase-continuity assumption, the traditional 2D PU applied to InSAR processing becomes a challenging step where strong phase discontinuities occur in the area of interest. The interferometric
fringes in wetlands can be very complicated due to a high
phase gradient and the phase discontinuities caused by rivers and levees. For example, Figure 4(a) shows an interferogram generated from PALSAR-2 scanning-SAR (ScanSAR)
images acquired on 28 June and 26 July 2015 around the
confluence of the mainstem Amazon and Japura River [48].
The InSAR fringes are bordered by the unflooded upland
area and segmented by the rivers, with a low InSAR coherence. The spatial discontinuity and complexity of the phases brought challenges to the traditional 2D PU methods.

65° W

1° S

2π

Japura
2° S

(b)
Reference Point
High: 10 (rad)

150
km

Range
Change in LOS Direction
0 cm

12.1 cm

4° S

rua

3° S

tem

s
Main

Low: -25

zon
Ama

(a)

(d)

FIGURE 4. (a) The interferogram across Amazon wetlands around the confluence of the mainstem Amazon and Japura River, generated

from PALSAR-2 ScanSAR images obtained on 28 June and 26 July 2015. (b) A close-up examination of the black-box region in (a). (c) The
unwrapped interferogram without masking. (d) The unwrapped interferogram with masking. The black dots represent the reference points
for unwrapping (with the phase set to zero) (from [48]).
MARCH 2020

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

125

IEEE Geoscience and Remote Sensing Magazine - March 2020

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