IEEE Geoscience and Remote Sensing Magazine - March 2020 - 83

CHALLENGING ISSUES AND FUTURE RESEARCH

potentially be mitigated by selecting a reference point within
the local region [9]. In practice, however, we generally extract
mining displacement measurements from an InSAR LOS displacement map over a much wider area than the region of interest so as to mitigate large, spatial-scale errors of interferometries (e.g., due to orbital uncertainties) and select a reference
point in the far field for phase unwrapping. Consequently, APS
mitigation is necessary for accurate estimation of InSAR-based
mining displacements, as the APS in a wide area is considered
spatially uncorrelated in general.
In recent years, several effective methods were developed
to reduce the APS from TS interferometries without auxiliary
or external data sets, such as by temporal filtering [36], [131]
and high- and low-pass filtering in the space and temporal
dimensions [33], [38]. However, for a single interferometry,
the straightforward methods for APS mitigation based on
auxiliary atmospheric data sets (e.g., GPS atmospheric measurements [13], [18], multispectral images [16], [132], and
weather or meteorological models [133], [134]) have recently
been developed and, how to effectively limit the APS is still an
open problem, especially from a single interferogram over areas without dense and continuous GPS stations [135]. Given
the fact that continuous GPS stations are usually absent in
most mining areas, it is challenging task to effectively mitigate
the APS from each InSAR interferogram over mining areas.

CHALLENGING ISSUES
A prerequisite of InSAR-based 3D mining displacement
reconstruction, model parameter inversion, and forward
prediction is that the surface 1D/2D displacement can be
reliably measured by InSAR techniques. This raises three
challenging issues-interferometric-phase temporal decorrelation, the mitigation of the APS, and the retrieval of large
mining deformation-that are particularly relevant from the
point of view of InSAR-based 1D/2D mining displacement
detection. These issues are discussed in the following section.

RETRIEVAL OF LARGE MINING DEFORMATIONS
The magnitude of mining-induced surface deformation can range from millimeters to dozens of meters,
depending on the scale of the underground mineral extraction [3]. Generally, InSAR techniques can successfully

N

#5
#4

#1

INTERFEROMETRIC-PHASE TEMPORAL
DECORRELATION
Coherence is an important indicator for the reliability of interferometric phases [9]. The higher the coherence, the more
reliable the interferometric phase, and vice versa. This implies
that, for those coregistered SAR pixels with a loss of coherence (i.e., decorrelation), InSAR cannot accurately detect the
ground surface deformations. Therefore, phase decorrelation
is a typical challenging issue for InSAR-based deformation
measurement. This is particularly true in InSAR mining deformation detection, mainly because most mining areas are
located in rural regions covered by vegetation (e.g., trees,
crops, and grasses). As a result, physical changes (e.g., seasonal changes) to the ground surface of mining areas can be
significant during the time period of SAR acquisitions, possibly causing temporal decorrelation [128]. This is particularly true for short-wavelength interferometry because short
wavelengths (e.g., X bands) usually have a lower penetration
ability than do long wavelengths (e.g.,
L bands). Therefore, a current, possible option is to select SAR images
with long wavelengths and short time
separations to reduce the influence
of severe temporal decorrelation on
mining deformation detection using
InSAR methods.
#2

6 cm

#3

MITIGATION OF THE
ATMOSPHERIC PHASE SCREEN
Spatial movements of atmospheric
water vapor in the atmosphere cause
variations in InSAR interferometries
[9] [referred to as the APS]. Such atmospherically induced variations may
cause large errors in InSAR-retrieved
LOS displacement measurements if
the APS cannot be removed or effectively mitigated [130]. Mining-induced
deformations are localized (e.g., within an area of a few square kilometers),
whereas the APS in a local region
(e.g., within 1km 2) can be considered
spatially correlated in general and can
MARCH 2020

A
(b)
B

Category 1
Category 2
Category 3
Buildings

0 km 0.5
(a)

1

1.5

6 cm
(c)

FIGURE 5. (a) The assessed building damage caused by extracting five planned working panels
#1 to #5 (marked by the black dashed rectangles) in the Qianyingzi coal mining area of China,
obtained using (b) the temporal PIM and its InSAR-derived model parameters. (c) Field images of
the building damage caused by the underground extraction [126].

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

83
IEEE Geoscience and Remote Sensing Magazine - March 2020

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