IEEE Geoscience and Remote Sensing Magazine - March 2020 - 105

A general framework for a systematic integration of the results of geomorphological investigations or (GeoG method),
geotechnical monitoring (or Geot method), and DInSAR
data (or Sat method) for landslide characterization in urban
area was proposed in [50] via the a posteriori integration
(aPosIn) procedure. aPosIn consists of three phases, which
can be summarized as follows [50].
1)	 First, landslides mapped using the GeoG method are
classified based on their maximum thickness (via geomorphological criteria), lithology (from geological surveys), kinematics, and state of activity (from geomorphological features).
2)	 Then, displacement information derived from the Sat
and Geot (from GPS and inclinometer monitoring networks) methods are combined.
3)	 Finally, landslides, for which the results obtained via
the full integration of the GeoG, Sat, and Geot methods are available, are typified in terms of their geometric
(i.e., boundary and thickness) and kinematic (i.e., minimum, average, and maximum velocity and average displacement direction) characteristics.
In phase 2, the obtained typified landslide map is used
jointly with the built-up area map to identify the exposed
elements (i.e., single buildings) located within or close
to the boundary of landslide-affected areas. For each of
them, the cumulative settlements in the SAR observation
period are estimated by multiplying the average velocity
along the vertical direction (derived from the LOS-sensor-target direction) pertaining to each coherent DInSAR
measurement point, hereafter referred to as the benchmark, covering the buildings for the observation period of
the considered SAR sensor.
Phase 3 focuses on the estimation, for each identified element (i.e., building) at risk, of the suffered damage severity
level and the magnitude of the intensity parameter (selected
as representative of the building movement interacting with
the landslide). The former is assigned to the building during
in situ damage surveys in six classes (D0 = negligible; D1 =

very slight; D2 = slight; D3 = moderate; D4 = severe; and D5
= very severe), reported in Table 2. These adapt the ranking
proposed in [129] and [130], using ad hoc predisposed fact
sheets [49]-[132] that allow the collection of information-
in addition to that concerning the location and urban fabric
characteristics (e.g., structural type, foundation typology,
number of floors, and occupancy type)-about the crack
patterns, including their widths, exhibited by the building facades to clearly distinguish the structural pathologies
pertaining to each exposed building. The intensity parameter-assumed to be the differential settlement ^D h of a
given building-is calculated as the maximum difference of
the cumulative settlements recorded by the pertaining coherent pixels derived by DInSAR data processing [38], [131],
[133]-[137].
Because a differential measurement is important for the
estimation of the intensity parameter, a pair of measurement points, at least, is required to be detected over the
single structure. To comply with this requirement, the processing of data at the highest resolution (i.e., PSI or SAR tomography) should be considered. Even better performances are achieved if, in addition to full-resolution processing,
VHR data (such as those from the TerraSAR-X and COSMOSkyMed missions) are exploited.
The information gathered on both the damage severity and magnitude of the selected intensity parameter,
homogenized according to the structural typology [e.g.,
reinforced concrete (RC) or masonry], are used in phase
4 to retrieve threefold end products to be used in procedures for investigating the damage suffered by structures
at risk and to forecast their vulnerability. First, by merging the building damage levels recorded during in situ
surveys with the intensity parameter, relationships between cause [DInSAR-derived differential settlement ^D h
] and effect (building damage level) are obtained. Then,
starting from the obtained relationships and adopting a
proper probabilistic model [44]-[46], [48], [136]-[140],
empirical fragility curves are generated. In particular,

TABLE 2. DAMAGE CLASSIFICATION.
DAMAGE CATEGORY

DESCRIPTION OF TYPICAL DAMAGE

D 0 (negligible)

Hairline cracks are mainly caused by shrinkage and thermal gradients. Typical widths are less than 0.1 mm.

D 1 (very slight)

Cracks are fine and rarely visible; they can be easily treated via normal decoration. Typical crack widths are up to 1 mm.

D 2 (slight)

Visible cracks can be masked by suitable linings or easily filled. Doors and windows might stick slightly. Typical crack
widths are up to 5 mm.

D 3 (moderate)

Cracks are very visible; they require some opening and can be patched by a mason. Doors and windows might stick; service pipes might fracture. Typical crack widths are 5-15 mm.

D 4 (severe)

Widespread cracks and extensive damage require breaking out and replacing sections of walls, especially over doors and
windows. Settling might cause a slight tilt to walls and the onset of fractures to structural elements. Typical cracks widths
are 15-25 mm, but the number of cracks must also be taken into account.

D 5 (very severe)

Extensive cracking with structural damage requires a major repair job involving partial or complete rebuilding. Settling
might cause a tilt to walls and instability requiring building evacuation. Typical cracks widths are greater than 25 mm, but
the number of cracks must also be taken into account.

Adapted from [49] and [130].

MARCH 2020

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

105
IEEE Geoscience and Remote Sensing Magazine - March 2020

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