SBAS Tutorial

Version 5.6.2
September 2022

1
Table of Contents
1. Introduction .................................................................................................................................... 4

2. Training Data sets description .............................................................................................................. 9

3. Processing Time ................................................................................................................................ 10

4. Before starting .................................................................................................................................. 11

5. Connection Graph Generation ............................................................................................................ 12

Input files.......................................................................................................................................... 12
Optional file ....................................................................................................................................... 12
Principal Parameters .......................................................................................................................... 12
Output Files ....................................................................................................................................... 14

6. Interferometric Process ..................................................................................................................... 17

Input files.......................................................................................................................................... 17
Optional file ....................................................................................................................................... 17
DEM/Cartographic System .................................................................................................................. 17
Principal Parameters .......................................................................................................................... 17
Output .............................................................................................................................................. 19

7. Inversion: First Step .......................................................................................................................... 23

Input files.......................................................................................................................................... 24
Optional files ..................................................................................................................................... 24
Principal parameters .......................................................................................................................... 24
Output .............................................................................................................................................. 27

8. Inversion: Second Step ...................................................................................................................... 30

Input files.......................................................................................................................................... 30
Optional files ..................................................................................................................................... 30
Principal Parameters .......................................................................................................................... 30

9. Geocoding ........................................................................................................................................ 35
Input files.......................................................................................................................................... 35
Optional files ..................................................................................................................................... 35
DEM /Cartographic System ................................................................................................................. 35
Principal Parameters .......................................................................................................................... 35
Output .............................................................................................................................................. 38

Annex 1. Auxiliary File ........................................................................................................................... 42

Annex 2. Meta viewer ............................................................................................................................ 43

2
Annex 3. Time Series Analyzer ............................................................................................................... 45

Annex 4. SBAS Edit Connection Graph .................................................................................................... 47

Input files.......................................................................................................................................... 47
Principal Parameters .......................................................................................................................... 47

Annex 5. Refinement and Re-Flattening .................................................................................................. 50

Automatic Refinement and Re-flattening – Available from version 5.6.2 ................................................ 50
Manual Refinement and Re-flattening.................................................................................................. 52
Output .............................................................................................................................................. 54

3
1. Introduction

The main goal of this tutorial is to give a basic description of the SARscape SBAS processing chain to obtain
precise topography, average displacement models and displacement time series that allows the
monitoring of the temporal evolution of surface deformations affecting distributed targets.

The SBAS approach is based on the Differential Interferometric Synthetic Aperture Radar (DInSAR) technique,
which refers to the basic principles of the interferometry. The interferometric phase is the contribution of 4
components: the topographic component, the phase noise component, the displacement component, and the
atmosphere component. A stack of images allows the separation of the different components.
The SBAS technique (Berardino et al., A New Algorithm for Surface Deformation Monitoring Based on Small
Baseline Differential SAR Interferograms, 2002) uses an effective combination of differential interferograms.
The interferograms are produced by data pairs characterized by small orbital separation (normal baseline)
and small temporal separation (temporal baseline) to limit the spatial and temporal decorrelation effects.
The algorithm exploits a singular value decomposition (SVD) inversion to retrieve the average surface
deformation velocity and residual topography (Inversion: First Step) and the temporal evolution of
surface deformations (Inversion: Second Step) of distributed targets.
The availability of a high number of valid interferograms is the basis for the reliability of the estimated
products. A minimum of 15-20 acquisitions are needed to perform SBAS analysis, possibly covering a temporal
interval of one year or more. Obviously, a higher number of input acquisitions is leading to a higher accuracy
of the results.

The SBAS approach can retrieve the following information/maps:

• land displacement maps, linear and non-linear temporal series of deformation, average
displacement models.
• digital elevation models.

The SBAS approach in SARscape can affords and manages:

• geophysical events occurring during the monitoring time interval (refer to Geophysical Model
parameters section in SBAS Interferometric Process), such as earthquakes.
• removal of atmospheric phase distortions through the use of external measurements (such as GACOS)
or through correlation with height.
• application of layover and shadow mask, external classification mask, temporal height correlated
mask.
• disconnected connection graph. Temporally irregular coherent measurement (e.g., snow coverage
variation, snow melting, vegetated areas) can be managed by the disconnected blocks SBAS feature
(by activating the Allow disconnected blocks and Interpol disconnected blocks flags): the missing dates
are temporally interpolated considering the average displacement trend. The temporal interpolation
across the disconnected blocks is performed when a minimum number of valid (coherent) acquisitions
is available (refer to FI_interf_perc, SI_interf_perc, SI_image_perc layers).
The SBAS approach can provides reliable and precise measurement for both requested target types:
• temporal coherent points, such as corner reflector, small-medium height buildings, mine slopes.
• temporal disconnected coherent targets, such as rural, snow coverage variation and low vegetated
areas (wider area).

The SBAS approach can achieve a precision of about 1-2 mm/year for temporal and spatial coherent areas,
comparable with GPS and Persistent Scatterers PS measurements. The measured precision decreases where
temporal and spatial coherence decrease, as for instance over vegetated or snowed areas.

The SBAS technique uses the input SLC data to select the pair combinations for the interferogram generation.
The Super Master is automatically chosen among the input acquisitions. The Super Master is the reference
4
image of the whole process, and all the slant range pairs (specified in the CG_report.txt) are co-registered on
this slant range reference. Once the co-registration is performed, the interferograms are generated. Using a
reference Digital Elevation Model (DEM) the flat earth and the topography are removed. The better the
reference Digital Elevation Model accuracy/resolution the better the result in terms of topography removal and
quality of co-registration. To remove the phase noise, the flattened interferograms are filtered. Starting from
the filtered interferograms, the spatial coherence is calculated for each interferogram. Once the filtering is
performed, the unwrapping procedure allows to obtain the absolute phase for those pixels whose coherence
overcomes a defined threshold value.

The next SBAS steps require a refinement and re-flattening process before proceeding to remove the offsets
and the trends from the stack of unwrapped phases (First and Second Inversion steps) or from the final
products (Geocoding step).

The first re-flattened unwrapped phases are used by the First Inversion step for the estimation of the modelled
displacement rate and of the topographic residuals with respect to the reference DEM. The modelled
displacement rate is then removed from the original phase information and a new unwrapping is performed
with the aim of improving the quality of the stack phase.

The second re-flattened unwrapped phases are used by the Second Inversion step for the estimation of
displacement time series with respect to the first acquisition date. Since the atmospheric component has a
high spatial correlation and a low temporal correlation; it can be removed applying a dedicated filter to the
time series. After that the final average modelled displacement is computed on the time-series.

All the steps of the SBAS processing are performed in slant range geometry. To be able to analyze and display
the SBAS results in geographic coordinates, geocoded SBAS raster and shape files are generated. The third
re-flattening is performed by the Geocoding step to associate the final measurement information to reference
points.

5
The SBAS performed steps and sub-steps are described in the list below and correspond to the logical workflow
shown in the Figure 1 and Figure 2:

1. Connection Graph
2. Interferometric Workflow
2.1. Co-registration
2.2. Differential interferogram generation
2.3. First Phase Unwrapping: standard 2D
3. Inversion First Step
3.1. Unwrapped phase stack Refinement and Re-Flattening
3.2. Average displacement model and residual topography estimation
3.3. Second Phase Unwrapping: standard 2D
4. Inversion Second Step
4.1. Unwrapped phase stack Refinement and Re-flattening
4.2. Displacement component estimation
4.3. Atmosphere pattern removal
4.4. Mean power image and precision component estimation
4.5. Temporal coherence and other ancillary information estimation
5. Geocoding
5.1. Final product Refinement and Re-Flattening
5.2. Average displacement model, precisions, and other layers geocoding in shape and raster format
5.3. Displacement time series geocoding in shape and raster format

6
 Start

N SAR images selection

Sample SLC stack

Download Import
selection

[Link]
1 - Connection Graph CG_report.txt
Super master pwr
plots

2 - Interferometric Processing

Co- Filtered differential First [Link]

registration Interferogram Unwrapping IS_cc.series
generation IS_upha.series

Figure 1 Logical workflow for the SBAS module.

7
 4 – Inversion: First Step IS_reflat_upha

Average model (h,v)

Refinement and Re- SVD inversion: Subtract average Second
flattening average model model (h,v) to fint unwrapping Temporal coherence
estimation (h,v) stack

5 – Inversion: Second Step

SI_second_reflat_upha.series
Refinement and Re- Add back average displacement
flattening model (v) to upha stack SI_disp_first.series

SI_disp.series
SVD inversion: Atmospheric Polyfit: average
Temporal coherence
displacement time series artefacts removals model estimation (v)

6 – Geocoding SI_reflat products

Average model (h,v) geo

Refinement and Re- Water body Raster Shape
flattening mask geocoding geocoding Temporal coherence geo

Displacement and displacement first

End

Figure 2 Logical workflow for the SBAS module.

The SBAS processing described in this tutorial was performed in a workstation with the following
characteristics:

SARscape version 5.6.2

ENVI version 5.6.2

SO 64-bit Windows 10 Pro

RAM 32 GB

Processor Intel (R) Core (TM) i9-9980HK CPU @ 2.40GHz

8
 OpenCL NVIDIA Quadro RTX 3000

HD SSD EVO 2TB (as working directory)

The entire sample data stack and the SBAS results shown in this tutorial can be found in the DataSetsTutorials
folder on our FTP ([Link]). Please contact us at support@[Link] to get the login credentials to
access the FTP providing the full name of the license owner, the name of your institution and the license
number.

2. Training Data sets description

The training area is characterized by a distributed land subsidence due to natural soil consolidation of the
artificial and massive landfills in several districts.

The area of interest is Urayasu city, located in the Chiba prefecture (Japan) close to Tokyo. The dataset
includes Ascending and Descending acquisitions (Figure 3).

The Sentinel-1 Descending (track 46D) stack is characterized by 59 acquisitions in VV polarization. The
temporal coverage is from March 19, 2016, to December 22, 2018.

The Sentinel-1 Ascending (track 39A) stack is characterized by 59 acquisitions in VV polarization. The temporal
coverage is from March 31, 2016, to December 28, 2018.

The SBAS technique must be run separately for each acquisition geometry, from the Connection Graph to the
Geocoding step, to obtain Line of Sight (LOS) results and to permit a further possible displacement
decomposition along the vertical and east-west directions.

In this tutorial, only the Descending dataset is described, therefore all the images and graphs correspond to
the track 46.

9
Figure 3 The red and the purple boxes show the shapes of the 46 (Descending) and 39 (Ascending) sample
stack tracks, respectively.

3. Processing Time

The Table below intends to summarize the processing time to perform the SBAS in the Urayasu area for the
Descending dataset, considering the previously described hardware configuration. The processing time
reported, approximately 2 hours, considers only the machine time. The time required to analyze each step
results is not reported since it depends on the operator knowledge and experience in SBAS processing.

Dataset 46D (59 acquisitions)

Step Time [min]

Connection Graph 1

Interferometric Process 90

Inversion: First Step 20

Inversion: Second Step 17

Geocoding 2

10
4. Before starting

Before starting to run the SBAS processing chain, it is strongly recommended to:

• set the ENVI Preferences.

• set the SARscape Preferences (Specific) based on the Sentinel-1 TOPSAR (IW – EW).
• read the Getting Started, Data Preparation and Basic Tutorial to become familiar with the basic
processing chain that allows managing the dataset provided to run the SBAS processing.

In fact, the dataset provided was downloaded, imported, and sampled using the corresponding SARscape
tools.

The provided binary raster files are delivered with 3 auxiliary files:

• .hdr (an ASCII header for ENVI visualization)

• .sml (an ASCII header for SARscape processing with the main acquisition information)
• .kml (an ASCII file for Google Earth).

It is worth to mention that to apply the SBAS technique users have to consider suitable sensors
and acquisition modes, as well as small temporal and/or spatial separation (baseline) between
the acquisition orbits. The provided dataset is suitable for SBAS processing.

TIP This tutorial performs the SBAS processing in a defined area of interest. If users want to analyze a
different dataset, they should consider that a full frame processing is fully discourage if they do not have
the SBAS cluster version because of the dramatic computational resources needed. To reduce the
processing time a sampled dataset is strongly recommended.

To quickly analyze the results of each step, SARscape stores each output in the corresponding folders. In fact,
once the Connection Graph step is performed, the “Output Root Name”_SBAS_processing folder is created.

The principal sub-folders are:

• connection_graph
• first_inversion
• geocoding
• interferogram_stacking
• second_inversion
• work.

Note that all the parameters that need to be set or changed by the operator, different from the default values,
are highlighted in this document using orange color of the text.

All the details related to the description of the parameters are shown in grey in this document. The same
information can be found on the specific help page of each panel.

11
5. Connection Graph Generation

This functionality defines the SAR pair combination (Masters and Slaves) and the connection network, which
is used for the generation of the multiple differential interferograms. This step is mandatory and deletes any
content present in the output root folder if a process with the same root name has been performed before.

The pairs to be processed are shown as connections in a network that links each acquisition to the others.
Given N acquisitions, the maximum theoretical possible connections, connecting each acquisition with all the
others, are (N*(N-1))/2. For each pair, an interferogram is generated in the Interferometric Process step.

TIP

The connection graph can be edited using the specific tool in SARscape/Interferometric Stacking/Stacking
Tools/SBAS Edit Connection Graph (see Annex 4). The editing can be performed before or after any SBAS
steps.

The Connection Graph is performed running the /SARscape/Interferometric Stacking/SBAS/1 -

Connection Graph panel (see Figure 4).

Input files

Insert all the _slc or _slc_list files, previously prepared, in the Input file list tab from the “46D_cut” folder of
the dataset.

Optional file

Leave the Input Master file empty (this file is automatically chosen by SARscape).

Principal Parameters

Min Normal Baseline (%)

This threshold corresponds to the minimum percentage of the critical baseline value, which is considered
acceptable in the generation of the possible data connections.

Max Normal Baseline (%)

This threshold corresponds to the maximum percentage of the critical baseline value, which is considered
acceptable in the generation of the possible data connections.

Min Temporal Baseline (days)

This threshold corresponds to the minimum temporal distance, which is considered acceptable in the
generation of the possible data connections. This value is expressed in days.

Max Temporal Baseline (days)

This threshold corresponds to the maximum temporal distance, which is considered acceptable in the
generation of the possible data connections. This value is expressed in days.

Degree of Redundancy
A further degree of redundancy selection criteria can be specified:
• high, all the interferograms respecting the previous Baseline thresholds are accepted.
• low, the program tries to minimize the number of interferograms to be generated, based on the Graph
Redundancy Criteria and on Min Connection per Acquisition

12
Redundancy Criteria
The following selection criteria are foreseen for the removal of redundant connections, when the Degree of
Redundancy is low:
• Min Normal Baseline, the number of connections is minimized by keeping only those having the
smallest spatial baseline.
• Max Normal Baseline, the number of connections is minimized by keeping only those having the largest
spatial baseline.
• Min Temporal Baseline, the number of connections is minimized by keeping only those having the
smallest temporal baseline.

Max connections per Acquisition

Number of connections (interferograms) per acquisition remaining after the redundancy reduction. This
parameter indicates the average redundancy of the connection graph, and bibliography suggest setting a value
higher than 5 to obtain reliable inversion solutions.

Allow Disconnected Blocks

The standard approach of SBAS requires a fully connected graph. If this is not possible, the SBAS can also
works with temporal discontinuities and connect the time-series (blocks) before and after the discontinuities.
Flagging the ‘Allow Disconnected Blocks’ option, the SBAS works on continuous groups of acquisitions and
extrapolate the First Inversion model for the disconnected time intervals.
By setting this option to True, the program does not discard any group of images and the disconnected SBAS
functionality is automatically activated for the next inversion steps.
The choices are the following:
• False: when the graph is not fully connected the disjointed acquisitions and blocks are rejected.
• Allow: the disjointed acquisition blocks are kept.

13
Figure 4 Connection Graph panels.

Output Files
Choose the location of the processing folder and an Output Root name.

The Output Root Name is used to create the processing folder. The folder name is given by the following
composition: “Output Root Name”_SBAS_processing.

Once the SBAS Connection Graph process is completed, a report and two graphs are displayed (Figure 5). The
graphs are:
• Time-Position plot, which provides the relative normal distance between the acquisitions in meters (y
axis), and the input acquisition dates (x axis). The Time-Position plot allows users to visually
understand if the graph is fully connected and matches with required redundancy criteria. The higher
the connection the most reliable displacement estimates.
TIP The theoretical minimum average number of connections per acquisition is 5; we suggest an ideal
configuration of 10 connections per acquisition. Obviously, as you increase the number of connections,
the computational time is increasing.
• Time-Baseline plot, which provides the normal baseline in meters (y axis) and the input acquisition
dates (x axis). This graph shows the interferogram coverage for each date.

Each acquisition is represented by a diamond. The color of the diamond symbol is as follows:

14
 • Red: discarded acquisitions due to user specific constraints.
• Green: valid acquisitions.
• Yellow: Super-Master acquisition. The Super-Master is the reference image of the whole processing,
and all the processed slant range pairs are co-registered and overlayed to this reference geometry.

TIP The graphs can be reloaded at any processing step from /SARscape/Interferometric Stacking/Stacking
Tools/Plot Viewer panel by adding the auxiliary file.

Figure 5 Plots generated by the Connection Graph tool.

In addition to the plots, the multi-looked Super-Master image (suffix, "_pwr") is also generated and displayed
in ENVI (Figure 6). The multi-looked factor is automatically computed based on the grid size specified for the
sensor once the Preferences specific are set.

Figure 6 Multi-looked Super Master image.

The Connection Graph step generates the [Link] file, the connection graph folder results, and the work
folder.

15
The auxiliary file is the principal file that guide users in the SBAS processing, all the mandatory SBAS steps
are based on the auxiliary file (see Annex 1).

The Connection graph folder includes:

• CG_report.txt, which summarizes the graph pairs characteristic, such as the total number of
interferograms that shall be generated by the Interferometric Process step and involved in the
following steps. Given the actual stack of input acquisitions and constraints set, the number of
interferograms shall be generated by Interferometric Process step are 276. With an average number
of interferograms per acquisition equal to 9.4.
• plot sub-folder, which is used to store the files (CG_baseline.txt and CG_position.txt) necessary for
the visualization of the graphs in ENVI.

The work folder includes sub-folders. The subfolders are created in the work folder once each step of the
SBAS processing chain is run, and their name is automatically given by the step name.

It is worth to mention that the [Link] file is also created in the work folder. In this file all the set
parameters used in that processing are recorded.

16
6. Interferometric Process

The Interferometric Process step creates a stack of multi-looked unwrapped interferograms through a few
sub-steps: co-registration, differential interferogram generation, optional phase pattern removal and
unwrapping. The sub-steps implemented here are executed using the standard DInSAR processing approach.

The data multi-looking factors are referred to the Super Master image.

TIP We suggest processing the distributed targets with a looks number greater than 1 look (range looks
x azimuth looks), to obtain wrapped interferograms with a higher Signal to Noise Ratio (SNR), capable to
provide more reliable coherence estimates and continuous unwrapped phases.

The Interferometric Process is performed running the /SARscape/Interferometric Stacking/SBAS/2 -

Interferometric Process panel (see Figure 7).

Input files

Insert the [Link] file.

Optional file

Leave the Geometry GCP File empty.

Leave Avoid Moving Area File empty.

Leave Optional Water Vapor List empty.

Leave Classification Mask File empty.

DEM/Cartographic System

Insert the Srtm-1_V3_dem provided in the dataset.

Principal Parameters

Rebuild All
By setting this flag the whole process is started from scratch. It is advisable to leave this flag unchecked in
case of process interruption, so that the products already generated have not to be computed and stored
again.

Range Looks
Number of looks in range referred to the Super Master.

Azimuth Looks
Number of looks in azimuth referred to the Super Master.

Grid Size for Suggested Looks.

The grid size, in meters, used to tune range and azimuth looks. If the other parameters are manually set, the
grid size will not imply a change in their values.

Number of Parallel Unwrapping

Defines the number of unwraps that will be performed in parallel (one interferogram for each thread). This
parameter must not exceed the total number of CPU threads. The user should consider the available system
memory before increasing the number of parallel processes, as the memory consumption is directly related
to the number of unwraps performed in parallel.

17
Apply Layover and Shadow Mask
Mask the results in layover and shadow areas.

Atmosphere External Sensors

By inserting this optional list, the software removes the atmosphere artefacts from the filtered interferogram
stack using external sensors such as GACOS and MERIS (only for ENVISAT data). The external layers must
be previously downloaded and imported with the appropriate tools. The algorithm checks the
correspondence between the SAR acquisition and the external sensor layer stack dates, before converting to
the slant range Zenith Path Delay (ZPD) and removing it from the corresponding interferograms.

Atmosphere Height Correlation Flag

By selecting this flag, the algorithm estimates and removes the height-related component of the atmosphere
from each interferogram, using a the “Height Correlation Window Size [m]” (Atmosphere section) parameter
to define the size of the filter. The smaller this window the more spatial (height-related) details will be
removed from the interferograms. This flag is dedicated to mountainous areas with atmospheric artifacts
strongly correlated to height.

Co-registration with DEM

By setting this flag, the input Digital Elevation Model is used in the co-registration process.

Unwrapping Method Type

The choice is given between the following unwrapping methods:
• Region Growing, the Region Growing unwrapping method is used.
• Minimum Cost Flow, the Minimum Cost Flow (square grid) unwrapping method is used.
• Delaunay MCF, the Minimum Cost Flow (triangular grid) unwrapping method is used.

Unwrapping Decomposition Level

The number of under sampling levels to be applied to the complex interferogram before performing the
phase unwrapping can be specified.

TIP At this step it is recommended to set decomposition level at 0 (expensive for the calculation times,
but able to better preserve fast spatial phase variations) or 1 (faster, more robust with noisy
interferograms, but in cases of very fast spatial phase variations may suffer from aliasing problems).

Unwrapping Coherence Threshold

Pixels with coherence values smaller than this threshold are not unwrapped and set to NaN.

Filtering method
The choice is given among the following filtering methods:
• Boxcar window.
• Goldstein.
• Adaptive window.
• Adaptive Non-Local InSAR.

18
Figure 7 Interferometric Process panels.

Considering that the area under analysis is almost flat and the interferograms are well flattened by the
reference DEM (coupled with the weak sensitivity to the topography of the Sentinel-1 orbital configuration), it
is possible to use the Delaunay MCF technique for the unwrapping. This technique offers greater robustness
around areas of low coherence while avoiding incorrect phase discontinuities compared to its MCF approach.

TIP Other important parameters that deserve to be mentioned in this step are the interferometric filter
type and the interferometric filter intensity. The Goldstein filter offers an excellent quality-performance
ratio, while the Adaptive Non-Local InSAR filter preserves more details as well as offers a better estimation
of spatial coherence, at the expense of a larger computational time; therefore the Adaptive Non-Local
InSAR filter is recommended for small datasets (in terms of number of interferograms and their pixel size,
e.g., 1000x1000 pixels) as well as being bound to the use of a graphic processing units (GPU) to exploit
its multithread optimization.

Output

Output layers are saved inside the interferogram_stacking folder. This directory contains the following
products:
• Due to the large number of output products meta (_meta) and series (.series) files are generated.
They enable to load at once all relevant output products.
• The interf_tiff sub-folder contains the tiff files of intermediate interferometric SBAS results.
19
Generated meta and series files:
• IS_cc_meta, IS_cc.series, which refers to all coherence images.
• IS_fint_meta, IS_fint.series, which refers to all flattened and filtered interferograms.
• IS_clean_fint_meta, IS_clean_fint.series, which refers to all flattened, filtered and cleaned (outputted
by optional processes) interferograms.
• IS_upha_meta, IS_upha.series, which refers to all unwrapped phases.
• IS_pwr_meta, IS_pwr.series, which refers to all slant range intensity images.

The work_interferogram_stacking folder stores intermediate processing results.

To distinguish the incoming SAR pair from which each outgoing product comes (as well as to identify it in the
connection network), a prefix containing the master-slave acquisition dates (e.g., Yyyymmdd) and the master-
slave relationships will be added.
Please note that hereinafter IS_ is the prefix used for Interferometric Process products.

After the Interferometric Process step, it is very important to check the intermediate products, in particular:
• the flattened and filtered interferograms, *_fint.tiff files (also called wrapped interferograms).
• the unwrapped phases *_upha.
• the coherence series *_cc.

TIP To check the coherence and unwrapped phase files, upload in ENVI the IS_cc.series and the
IS_upha.series from the interferogram_stacking folder. The ENVI Series Manager opens and allows users
to quickly analyze the files. The visualization of the coherence files in grey scale is recommended (see
Figure 8), the visualization of the unwrapped files in rainbow color scale is suggested (see Figure 10). For
this reason, right click on the *_upha.series file from the ENVI Layer Manager, and select Change Color
Table/Rainbow. To check the * _fint.tiff files a simple image viewer could be enough (see Figure 9).

Figure 8 coherence transect for the Unwrapping Coherence Threshold evaluation.

TIP The coherence threshold value to be set in the Interferometric Process panel can be increased or
decreased depending on the coherence values characterizing the study area (see Figure 8). This value
must be increased if there are (on average) too many noisy areas that must be removed from the
unwrapped stack not to compromise the result; anyway, it should not be increased too much, e.g., not
more than 0.4-0.5, to preserve a sufficient spatial coverage.

20
 Figure 9 Filtered interferogram wrapped phase tiff.

Figure 10 Unwrapped interferogram phase colored with the ENVI Rainbow color scale.

Observing the above-mentioned results anomalous behavior, such as inaccurate orbits phase ramp, un-
coherent pairs, atmospheric artifacts, residual topography effects, can be detected.

TIP In case of some unwrapped interferograms are still characterized by very low coherence (visible as
low coverage) or unwrapping errors, it is recommended to remove them from the interferometric stack

21
with an Edit Connection Graph (see Annex 4), aiming to improve the result of the following SBAS
inversions.

22
7. Inversion: First Step

In this step, the parameters of the model fitting the trend of deformation according to the behavior of the
area of interest are estimated. The parameter estimation is performed on the unwrapped interferogram stack
(re-flattened*_reflat_upha), by inverting the corresponding system of equations. The matrix inversion is then
carried on with a Single Value Decomposition (SVD) approach. Only the pixels characterized by a fully
connected graph (subordinated to the imposed coherence threshold value), will be considered, and solved by
the inversion. The result of this step will therefore typically be an average displacement rate (v) and a
topographic residual (h) value. The No Displacement model is a custom method, where the SBAS is used
as a high precision DEMs generator.

TIP It is advisable to start using a linear inversion model as fitting deformation model resulting the more
robust and reliable choice. Other models (e.g., quadratic, cubic) require high connection graph redundancy
and highly coherent interferograms to provide reliable results.

After the deformation trend estimation via SVD inversion, the known components are removed from the
wrapped interferogram stack, to allow a second unwrapping (see Figure 11), with the aim of improving the
unwrapped interferogram stack ingested by the Inversion: Second Step.

First
displacement
stack

Temporary
Subtract SBAS
wrapped stack
Second
Inversio
n
Residual
topography
estimation
Unwrap

Re-
SBAS
flattening
First
Inversio
n
Temporary
unwrapped
Second
stack
Unwrapped
stack

First velocity
estimation Add
Unwrapped
Wrapped stack
stack

Re-
flattening

First Wrapped First

stack Unwrapped
stack

Figure 11 Second unwrapping detailed workflow.

23
The Inversion: First Step is performed running the /SARscape/Interferometric Stacking/SBAS/4 –
Inversion: First Step panel (Figure 4 Figure 12).

Input files

Insert the [Link] file

Optional files

Refinement GCP file

Enter a previously created Ground Control Point file (Load GCP File .xml) or use the interface to create a new
Ground Control Point file (Create GCP File, refer to the "Tools>Generate Ground Control Point" for details).
This file is used for manual refinement. This file is optional and when it is not introduced, the software performs
the refinement and re-flattening in an automated way.
See Annex 5 for more details.

Principal parameters

Rebuild All
By setting the flag the whole process is performed from scratch (if not incremental mode is active).
It is advisable to leave this flag unchecked in case of process interruption, so that the products already
generated have not to be computed and stored again.

Product Coherence Threshold

Pixels with coherence values smaller than this threshold will be set to dummy (NaN) in the final products.

Displacement Model Type

The choice is given between the following models:
• no displacement, a stable area is considered, no displacement is calculated (this parameter set to true
exploit the SBAS capability to calculate precise DEMs)
• Linear, the displacement velocity [mm/year] is calculated
• Quadratic, the displacement acceleration [mm/year2], and velocity [mm/year] are calculated
• Cubic, the displacement acceleration variation [mm/year 3], the displacement acceleration [mm/year2]
and velocity [mm/year] are calculated
• Linear Periodic, the periodic delay [days], the periodic modulation [mm] and velocity [mm/year] are
calculated
The three displacement polynomial models can be synthesized as follows:

1 1
𝐷𝑖𝑠𝑝 = 𝐾1 ∙ (𝑡 − 𝑡0 ) + ∙ 𝐾2 ∙ (𝑡 − 𝑡0 )2 + ∙ 𝐾3 ∙ (𝑡 − 𝑡0 )3
2 6

where Disp is the displacement at time t; K1 is the linear term [mm/year], corresponding to the instantaneous
displacement velocity; K2 is the quadratic term [mm/year 2], corresponding to the instantaneous displacement
acceleration; K3 is the cubic term, corresponding to the instantaneous displacement acceleration variation
[mm/year3].
The displacement periodical model can be synthesized as follow:
𝐷𝑖𝑠𝑝 = 𝐾1 ∙ (𝑡 − 𝑡0 ) + 𝐾2 ∙ cos(2 ∙ 𝜋 ∙ 𝐹 ∙ (𝑡 − 𝑡0 ) + 𝐾3 )

where Disp is the displacement at time t; K1 is the linear term, corresponding to the instantaneous
displacement velocity; K2 is the modulation term [mm]; K3 delay [days]. F is the known expected frequency
(1/ Model Period).

Estimate Residual Height

If selected, it estimates the residual topography height. This parameter should set to false in two situations:
• In case of good quality high reference resolution DEM, topography component estimation is not
necessary.

24
 • In case of less than 20 images, this could lead to a pour residual height estimation, and consequently
a wrong velocity estimation, due to strong atmospheric artefacts. This could mostly happen for stack
with small normal baseline as Sentinel-1. The simplest model (with just one free term) increases the
result reliability.
When this parameter is selected, and the Displacement Model Type selected is no displacement, the SBAS is
used for the generation of precise DEMs.

Spatial Wavelet Size (m)

It preserves the estimated residual topography data resolution up to the specified value. It removes the low
pass distortion under this resolution value and preserves the spatial details by using wavelet decomposition. If
this value is smaller or equal to 0 is not considered.

Allow Disconnected Blocks

By setting this flag the program provides the solution for disconnected coherent targets over the grid point.
The full graph connectivity is not a constrain anymore (refer to the Technical Note in the Connection Graph).
The disconnected block SBAS provides a higher spatial coverage of the final results, limited by the Min Valid
Interferogram percentage threshold.
The choices are the following:
• False: when the graph is not fully connected it is rejected.
• Allow: the results are provided even if the graph is not fully connected.

Min Valid Interferograms (%)

Percentage of the minimum number of accepted SBAS inversion equations (interferograms over the coherence
threshold) for getting a result for each output grid point. This parameter is considered if it is equal or greater
than zero and should be used when the Allow Disconnected Blocks is flagged (or the Allow disconnected Blocks
has been previously flagged in the Connection Graph generation step). We suggest a threshold higher than
50% to have enough valid measurements to correctly calculate the SVD inversion.

Stop Before Unwrapping

By setting this flag, the phase unwrapping (last step of the processing sequence) is not executed.

Number of Parallel Unwrapping

Defines the number of unwraps that will be performed in parallel (one interferogram for each thread). This
parameter must not exceed the total number of CPU threads. The user should consider the available system
memory before increasing the number of parallel processes, as the memory consumption is directly related to
the number of unwraps performed in parallel.

Unwrapping Method Type

The choice is given between the following unwrapping methods:
• Region Growing, the Region Growing unwrapping method is used.
• Minimum Cost Flow, the Minimum Cost Flow (square grid) unwrapping method is used.
• Delaunay MCF, the Minimum Cost Flow (triangular grid) unwrapping method is used.
Set the Unwrapping Method Type to Delaunay MCF

Unwrapping Decomposition Level

The number of under sampling levels to be applied to the complex interferogram before performing the phase
unwrapping can be specified.

Unwrapping Coherence Threshold

Pixels with coherence values smaller than this threshold are not unwrapped and set at NaN.

Refinement Radius (m)

Maximum buffer radius for the association of Ground Control Point with respect of the slant range unwrapped
phase closer valid pixel.

Refinement Residual Phase Poly Degree

Degree of the polynomial used to estimate the phase ramp, which will be removed from the input unwrapped
phase during the Re-flattening operation. In case this value is higher that the number of input Ground Control
Points, it will be automatically decreased. The default values of 3 means that a phase ramp in range and
25
azimuth direction plus a constant phase offset will be corrected. In case only the phase offset correction is
needed, the polynomial degree will be set to 1.

Figure 12 Inversion: First Step panel.

For this process, it was decided to use the automatic refinement and re-flattening: no GPCs are needed.

Considering that the area under analysis is almost flat and the interferograms are well flattened by the
reference DEM, it is possible to use the “Delaunay MCF” technique for the unwrapping. This technique offers
greater robustness around areas of low coherence while avoiding incorrect phase jumps compared to its “MCF”
counterpart. Increase the coherence thresholds to 0.3, more suited for this high coherence area.

TIP During the second unwrapping performed in the First Inversion step it is possible to increase the
Unwrapping Decomposition Level compared to the first unwrapping (e.g., Level 2 could be afforded)
because the wrapped interferograms have been simplified (the estimated topographic residuals and the
displacement model have been removed) and therefore the risk of aliasing is lower. Increasing the
decomposition level could lead to improve the quality of the second unwrapping results as well as a
greater speedup of the process.

26
Output

first_inversion folder
Directory containing the following products:
• Due to the large number of output products meta (_meta) and series (.series) files are generated.
They enable to load at once all relevant output products.
• The “interf_tiff” sub-folder with the tiff files of intermediate interferometric SBAS results.
• The “inversion” sub-folder with the products (first model) from SBAS inversion in slant range
geometry.

Generated meta and series files:

• FI_reflat_upha_meta, FI_reflat_upha.series which refers to the re-flattened upha.
• FI_reflat_fint_meta, FI_reflat_fint.series which refers to the re-flattened filtered complex
interferograms.
• FI_model_first_meta, FI_model_first.series which refers to the residual height and displacement
velocity (plus ancillary layers), obtained from the SVD inversion, in slant range geometry.
• FI_second_fint_meta, FI_second_fint.series, which refers to all re-flattened filtered interferograms.
From these interferograms the estimated displacement model and the estimated residual topography
has been removed. Note that this same extension is used for the 1st level unwrapped products, which
are generated during the Interferometric Process step.
• FI_second_upha_meta, FI_second_upha.series, which refers to all unwrapped phases. From these
unwrapped phases, the estimated model has been removed. Note that this same extension is used
for the 1st level unwrapped products, which are generated during the Interferometric Process step.

The output results in the inversion folder consist of:

• height related products which are:

o FI_H_correction, corresponding to the correction (in meters) with respect to the input Digital
Elevation Model.
o FI_H_precision, corresponding to the estimate in meters of the height average precision.
o FI_srdem, corresponding to the corrected Digital Elevation Model in slant range geometry.
• displacement related products which are:
o FI_term_1, corresponding to the first-degree term of the inversion polynomial [mm/year].
o FI_term_2, corresponding to the second-degree term of the inversion polynomial [mm/year2].
o FI_term_3, corresponding to the third-degree term of the inversion polynomial [mm/year3].
o FI_velocity, corresponding to the average displacement velocity [mm/year].
o FI_acceleration, corresponding to the average displacement acceleration [mm/year 2].
o FI_delta_acceleration, corresponding to the variation of the average displacement
acceleration [mm/year3].
o FI_periodical_modulation, corresponding to the periodic model modulation [mm].
o FI_periodical_delay, corresponding to the periodic model delay [days].
o FI_V_precision, corresponding to the estimate in millimeter/year of the velocity average
precision.
• general products which are:
o FI_coherence, corresponding to the multi-temporal coherence. It shows how well the
displacement trend fits with the selected model.
o FI_RMSE_rad, is a measure of the inversion quality. It is the RMSE expressed in radians. The
higher this value the worse the fitting and inversion quality.
o FI_interf_perc, valid interferogram percentage.

The displacement values are reported with:

• Positive sign if the movement corresponds to a decrease of the sensor-to-target slant range
distance.
• Negative sign if the movement corresponds to an increase of the sensor-to-target slant range
distance.

27
Figure 13 FI_model_first_meta products.

TIP Figure 13 shows a warning popup that is related to the necessity of reducing the spatial size of the
sample dataset under processing. This can be avoided by reducing the Spatial Wavelet Size (m) or by
enlarging the size of the dataset but the tool is automatically adapting this parameter in case it is not set
properly.

The average velocity, shown in Figure 14, is calculated in the following way: the displacement calculated over
the considered period over the temporal interval. If the model chosen by the user is of the first order, the
average velocity corresponds to the linear term K1. The mean acceleration and the change in mean
acceleration are calculated in the same way in case of quadratic or cubic models.

28
Figure 14 Average velocity first estimate (v) [mm/year] left, residual height estimation (h) [m] right.

29
8. Inversion: Second Step

The Inversion: Second Step aims to estimate the final displacement time series [m], from the stack of
unwrapped interferograms. The unwrapped interferograms are obtained from the previous SBAS step and
should no longer contain any residual topography component. The unwrapped interferograms are firstly
refined and re-flattened before being used as input to the SVD matrix inversion tool. The SVD inversion is
used to solve the system of equations that describe the stack of unwrapped interferograms obtaining the
displacements solution corresponding to each acquisition date. The displacement time series values are
referred to the oldest date which is the reference zero.

Date-by-date displacements are filtered to remove atmospheric phase components: the filter preserves the
displacement model component found during the Inversion First Step. Once the SVD matrix inversion has
provided the “first estimate” of the displacement time series, the software temporarily removes the
displacement model component, to preserve it during the Atmospheric Phase Screen (APS). The spatial LP and
temporal HP filters are applied to the displacement time series. The resulting LP-HP layers (one for each
acquisition) correspond to the contribution of atmospheric artifacts. These components can be checked using
the corresponding series file in the work folder. The software removes atmospheric artifacts from the "first
estimate" of the displacement time series, to obtain the final displacement time series. To increase the power
of the HP-LP filter: decrease the size of the LP window (meters) and increase the size of the HP window (days).

The final step is to derive the new displacement model as a regression of the filtered time series.

The Inversion: Second Step is performed running the /SARscape/Interferometric Stacking/SBAS/5– Inversion:
Second Step panel (Figure 4 Figure 15).

Input files
Insert the [Link] and the refinement GCPs file.

As for the generation of GCPs for refinement and re-flattening, refer to the considerations previously made,
or simply use the same set of points previously generated.

TIP For this step it is possible to create a new set of GCPs to re-flat the unwrapped interferogram stack
generated in the previous Inversion: First Step.

Optional files

Refinement GCP file

Enter a previously created Ground Control Point file (Load GCP File .xml) or use the interface to create a new
Ground Control Point file (Create GCP File, refer to the "Tools>Generate Ground Control Point" for details).
This file is used for manual refinement. This file is optional and when it is not introduced, the software performs
the refinement and re-flattening in an automated way.
See Annex 5 for more details.

Principal Parameters

Rebuild All
By setting the flag the whole process is performed from scratch (if not incremental mode is active).
It is advisable to leave this flag unchecked in case of process interruption, so that the products already
generated have not to be computed and stored again.

Product Coherence Threshold

Pixels with coherence values smaller than this threshold will be set to dummy (NaN) in the final products.

30
Interpol Disconnected Block
By setting this flag to true the program provides the solution for disconnected coherent targets over the grid
point. The temporal gaps, where the displacement measurement does not exist or have been rejected, are
interpolated using the average trend estimated during the first inversion. The disconnected SBAS provides a
higher spatial coverage of the solution, limited by the Min Valid Interferogram and Min Valid Acquisition
percentage thresholds.
The choices are the following:
• False: the temporal series with gaps are rejected.
• Interpol using model: the gaps are interpolated using the average trend estimated during the
Inversion: First Step.
• Interpol starting from zero: the gaps are filled with zero. The temporal series start from zero at the
end of each gap.

Min Valid Interferograms (%)

Percentage of the minimum number of accepted SBAS inversion equations (interferograms over the coherence
threshold) for getting a result for each output grid point. This parameter is considered if it is equal or greater
than zero and should be used when the Interpol Disconnected Blocks is flagged (or the Allow disconnected
Blocks has been previously flagged in the Connection Graph generation step or First Inversion Step). We
suggest a threshold higher than 50% to have enough valid measurements to correctly calculate the SVD
inversion.

Min Valid Acquisitions (%)

Percentage for the minimum number of valid acquisitions (covered by at least a certain amount of
valid interferogram over the coherence threshold) for getting an acceptable result, at each output grid point.
This parameter is considered if is equal or greater than zero and should be used when the Interpol
Disconnected Blocks is flagged (or the Allow Disconnected Blocks has been previously flagged in the
Connection Graph generation step or First Inversion Step). We suggest a threshold higher than 80% to have
enough valid measurements to correctly calculate the SVD inversion.

Atmosphere Low Pass Size (m)

Enter the window size, in meters, to apply the spatial distribution related filter.

Atmosphere High Pass Size (days)

Enter the window size, in days, to apply the temporal distribution related filter.

Refinement Radius
Maximum buffer radius for the association of Ground Control Point with respect of the slant range unwrapped
phase closer valid pixel.

Refinement Residual Phase Polynomial Degree

Number of polynomial coefficients used to estimate the phase ramp, which will be removed from the input
unwrapped phase during the Re-flattening operation. In case this value is higher that the number of input
Ground Control Points, it will be automatically decreased. The default value of 3 means that a phase ramp in
range and azimuth direction plus a constant phase offset will be corrected. In case only the phase offset
correction is needed, the polynomial degree will be set to 1.

Set the Refinement Residual Phase Polynomial Degree to 1

31
Figure 15 Inversion: Second Step panel.

For this process, it was decided to use the automatic refinement and re-flattening: no GPCs are needed.

Output

second_inversion folder
Directory containing the following products:
• Due to the large number of output products meta (_meta) and series (.series) files are generated.
They enable to load at once all relevant output products:
• The “inversion” sub-folder with the products (second model) from SBAS inversion in slant range
geometry.
• The internal copy of the input refinement and re-flattening GCP (.xml).
• The input refinement and re-flattening GCP converted in shape format in slant geometry (.shp).
• The input refinement and re-flattening GCP converted in shape format in geocoded geometry (.shp
and .kml).

Generated meta and series files:

• SI_model_meta, SI_model.series, which refers to the residual height and displacement velocity (plus
ancillary layers), obtained from the time series polynomial regression, in slant range geometry.
• SI_disp_first_meta, SI_disp_first.series, which refers to the date by date displacements, measured in
slant range geometry, without atmospheric correction.

32
 • SI_disp_meta, SI_disp.series, which refers to the date by date displacements, measured in slant range
geometry, after the atmospheric correction.
• SI_interf_counter_meta, SI_interf_counter.series, which refers to valid measure (interferograms with
coherence higher than the Product Coherence Threshold) counter for each acquisition.
• SI_only_disp_upha_meta, SI_only_disp_upha.series, which refers to all unwrapped phases flattened
by the residual height estimated component
• SI_second_reflat_upha_meta, SI_second_reflat_upha.series, which refers to all unwrapped phases
re-flattened for the second SVD inversion.

The output results in the inversion consist of:

• displacement related products which are:
o SI_term_0, corresponding to the zero-degree term of the inversion polynomial [mm].
o SI_term_1, corresponding to the first-degree term of the inversion polynomial
[mm/year]. SI_term_2, corresponding to the second-degree term of the inversion polynomial
[mm/year2].
o SI_term_3, corresponding to the third-degree term of the inversion polynomial [mm/year3].
o SI_velocity, corresponding to the average displacement velocity [mm/year].
o SI_acceleration, corresponding to the average displacement acceleration [mm/year2].
o SI_delta_acceleration, corresponding to the variation of the average displacement
acceleration [mm/year3].
o SI_periodical_modulation, corresponding to the periodic model modulation [mm].
o SI_periodical_delay, corresponding to the periodic model delay [days].
o SI_disp_first_est, corresponding to the date specific displacement (in millimeters) measured
with respect to the first acquisition date, to which a null displacement is assigned, without
atmospheric correction.
o SI_disp, corresponding to the date specific displacement (in millimeters) measured with
respect to the first acquisition date, to which a null displacement is assigned, after the
atmospheric correction.
• general products which are:
o SI_coherence, corresponding to the multitemporal coherence. It shows how much the
displacement trend fits with the selected model.
o SI_RMSE_rad, which is a measure of the inversion quality. It is the RMSE expressed in radians.
The higher this value the worse the fitting and inversion quality.
o SI_RMSE_mm, which is a measure of the fitting quality. It is the RMSE expressed in
millimeters. The higher this value the worse the fitting and inversion quality.
o SI_norm_L1, corresponding to the cumulated summation of the date-by-date absolute value
displacements (in millimeters).
o SI_interf_counter, counter with the valid measure number for each acquisition.
o SI_image_perc, valid acquisitions percentage (covered by at least one valid measure).
o SI_interf_perc, valid interferogram percentage.

The displacement values are reported with:

• Positive sign if the movement corresponds to a decrease of the sensor-to-target slant range
distance.
• Negative sign if the movement corresponds to an increase of the sensor-to-target slant range
distance.

33
Figure 16 Average velocity final estimate (v) [mm/year] upper left, temporal average intensity upper right,
root means square error [mm] lower left, temporal coherence lower right.

After the conclusion of the Inversion Second Step it is possible to analyze the meta file containing the principal
products: SI_model_meta (see Figure 16), in addition to the time series before (SI_disp_first_meta) and after
(SI_disp_meta) the removal of the atmospheric contribution, using the dedicated tool Time series Analyzer
Raster.

TIP The analysis and the comparison between the “first estimation” time series and the final time series,
could be useful to calibrate the intensity of the atmosphere filtering. This analysis is also useful to highlight
any anomalies on the spatial and temporal deformations that are not properly treated from the
atmospheric filter. This could happen in case of strong wide and sudden events such as an earthquake (a
phenomenon which behavior is similar to that of the atmosphere, and therefore could be removed or
degraded by the APS filter).

34
9. Geocoding

SBAS products are firstly refined and re-flattened before proceeding to the geocoding. Geocoded products can
be displayed in two types of formats: Shape and/or Raster based on the flag selected in the parameters.

The Geocoding step is performed running the /SARscape/Interferometric Stacking/SBAS/6– Geocoding panel
(Figure 4 Figure 18).

Input files

Insert the [Link] file.

Optional files

Refinement GCP file

Enter a previously created Ground Control Point file (Load GCP File .xml) or use the interface to create a new
Ground Control Point file (Create GCP File, refer to the "Tools>Generate Ground Control Point" for details).
This file is used for manual refinement. This file is optional and when it is not introduced, the software performs
the refinement in an automated way.
See Annex 5 for more details.

Shape Mask File

The user can create a shapefile (.shp) that contains the area to discard, this shapefile must be geocoded (with
the same carto-system of the reference DEM). The subsequent step will take this shape into account to remove
the pixel inside the mask area from the geocoding step. This file is optional.
DEM /Cartographic System

Insert the Srtm-1_V3_dem provided in the dataset.

Principal Parameters

Rebuild All
This flag refers to all SBAS process. By setting the flag the whole process is performed from scratch (if not
incremental mode is active).
It is advisable to leave this flag unchecked in case of process interruption, so that the products already
generated have not to be computed and stored again.

Height Precision Threshold

Upper threshold of the height measurement average precision values (refer to the Phase to Height
Conversion tool for more details). This must be provided in meters. This threshold is used together with the
Velocity Precision Threshold and the Product Temporal Coherence Threshold for the masking process.

Velocity Precision Threshold

Upper threshold of the velocity measurement average precision values (refer to the Phase to Displacement
Conversion tool for more details). This must be provided in millimeter/year. This threshold is used together
with the Height Precision Threshold and the Product Temporal Coherence Threshold for the masking process.

Product Temporal Coherence Threshold

Lower threshold of the temporal coherence measurement values. This threshold is used together with the
Height Precision Threshold and the Velocity Precision Threshold for the masking process.

Water Mask (dB)

Value in dB used as threshold to retrieve a water body mask from the average intensity image. All the points
under this threshold will be removed from the output layers. Active if smaller than 0. We suggest to set a
value of -10 or smaller.
Set this parameter to -12
35
Make Geocoded Raster
By setting this flag the output raster file/s are generated.

Make Geocoded Shape

By setting this flag the output shape file/s are generated.

Shape Max Number of Points

If values other than -1 are entered, the output shape file is split in several parts (each marked by a progressive
numbering - _1; _2; etc.); each part of the shape contains a portion of the total number of points, which
corresponds to the value entered. It is suggested not to exceed 200000 points to avoid visualization problems.

Generate Shape Time Series

By setting this flag the output shape file/s, with the displacement temporal evolution, are generated.

Generate Shape Counter Series

By setting this flag the output shape file/s, with the valid interferogram counter for each acquisition date, are
generated. When the counter is zero, the correspondent acquisition date has been temporally interpolated.
Higher the counter higher is the reliability of the corresponding estimated displacement measure.

Vertical Displacement
By setting this flag the displacements and velocity products are projected on the vertical direction. Do not use
the reprojection if the real displacement direction is unknown.

Slope Displacement
By setting this flag the displacements and velocity products are projected on the vertical direction. Do not use
the reprojection if the real displacement direction is unknown.

Displacement Custom Direction

By setting this flag any vector can be specified, in terms of azimuth (Azimuth Angle, measured in degrees from
the North - clockwise direction) and inclination (Inclination Angle, measured in degrees from the horizontal
plane). The map showing the displacement values projected on the specified direction is generated among
the output products. Do not use the reprojection if the real displacement direction is unknown.

X Dimension (m)

The grid size of the output data in Easting (X) must be defined; the default unit of measure is meters.
Note that - for the Geographic projection - if values higher than 0.2 are entered they will be considered as
metric units and then automatically, and roughly, converted from meters to degrees; if values lower than 0.2
are entered they will be considered as degree and used as such without any conversion.

Y Dimension (m)
The grid size of the output data in Northing (Y) must be defined; the default unit of measure is meters.
Note that - for the Geographic projection - if values higher than 0.2 are entered they will be considered as
metric units and then automatically, and roughly, converted from meters to degrees; if values lower than 0.2
are entered they will be considered as degree and used as such without any conversion.

36
Figure 17 Slant range geometry layers used for masking geocoded products. FI_H_precision (first_inversion
folder) upper left, FI_V_precision (first_inversion folder) upper right, SI_coherence (second_inversion folder)
lower left, SI_mean_pwr (second_inversion folder) lower right.

37
Figure 18 Geocoding panels.

For this process, it was decided to use the automatic refinement and re-flattening: no GPCs are needed.

To improve the masking of the geocoded products, it was decided to keep the Height Precision, the Velocity
Precision, and the Product Temporal Coherence thresholds unchanged, and to act only on the Water Mask
threshold. After evaluating the backscatter values on the water drawing a transect profile (see Figure 17) on
the SI_mean_pwr over the water, it was decided to use 0.06 as backscatter threshold to mask the water areas,
which corresponds to -12dB (10*log0.06).

Output

Geocoding folder
Directory containing the following products:
• Due to the large number of output products meta (_meta) and series (.series) files are generated.
They enable to load at once all relevant output products:
• The “raster” sub-folder with the products in raster version.
• The “vector” sub-folder with the products in shape version.

Generated meta and series files:

38
• G_model_geo_meta, which refers to the height and displacement velocity measurements (plus
ancillary layers) in the output cartographic projection.
• G_disp_first_geo_meta, which refers to the date-by-date displacements, measured in the output
cartographic projection, without atmospheric correction.
• G_disp_geo_meta, which refers to the date-by-date displacements, after the atmospheric correction,
in the output cartographic projection.
• G_interf_counter_geo_meta, which refers to valid measure counter for each acquisition.

The output results in “raster” folder consist of:

• displacement related products which are:
o G_FI_velocity_geo, corresponding to the first inversion mean displacement velocity
[mm/year]. This raster is the result when G_SI_velocity_geo is not present.
o G_FI_acceleration_geo, corresponding to the first inversion mean displacement
acceleration [mm/ year2]. This raster is the result when G_SI_acceleration_geo is not
present.
o G_FI_delta_acceleration_geo, corresponding to the first inversion mean displacement
acceleration variation [mm/ year3]. This raster is the result when
G_SI_delta_acceleration_geo is not present.
o G_SI_velocity_geo, corresponding to the second inversion mean displacement velocity
[mm/year].
o G_SI_acceleration_geo, corresponding to the second inversion mean displacement
acceleration [mm/ year2].
o G_SI_delta_acceleration_geo, corresponding to the second inversion mean displacement
acceleration variation [mm/ year3].
o G_SI_term_0_geo, corresponding to the zero-degree term of the second inversion
polynomial [mm].
o G_SI_term_1_geo, corresponding to the first-degree term of the second inversion
polynomial [mm/year].
o G_SI_term_2_geo, corresponding to the second-degree term of the second inversion
polynomial [mm/year2].
o G_SI_term_3_geo, corresponding to the third-degree term of the second inversion
polynomial [mm/year3].
o G_FI_V_precision_geo, corresponding to the velocity measurement average precision
[mm/year] (refer to the Phase to Displacement Conversion for more details).
o G_SI_DDDDDDD_X_disp_geo, corresponding to the date specific displacement [mm]
measured with respect to the first acquisition date, to which a null displacement is
assigned. Where DDDDDDDD is the acquisition date and X the incremental date ID.
o G_SI_DDDDDDD_X_disp_first_geo, corresponding to the date specific displacement [mm]
measured with respect to the first acquisition date, to which a null displacement is
assigned, without atmospheric removal. Where DDDDDDDD is the acquisition date and X
the incremental date ID.
• general products which are:
o G_FI_H_correction_geo, corresponding to the first inversion correction [m] with respect
to the input Digital Elevation Model. This raster is the result when G_SI_H_corretion_geo
is not present.
o G_SI_H_correction_geo, corresponding to the second inversion correction [m] with
respect to the input Digital Elevation Model.
o G_FI_H_precision_geo, corresponding to the height measurement average precision [m]
(refer to the Phase to Displacement Conversion for more details).
o G_FI_DEM, corresponding to the first inversion corrected Digital Elevation Model [m]. This
raster is the result when G_SI_DEM is not present.
o G_SI_DEM, corresponding to the second inversion corrected Digital Elevation Model [m].
o G_SI_RMSE_mm_geo, which is a measure of the fitting quality geocoded. It is the RMSE
[mm]. The higher this value the worse the fitting and inversion quality.
o G_SI_ DDDDDDD_X_interf_counter_geo, corresponding to the valid measures counter
per each date.
o G_SI_interf_perc_geo, valid interferogram percentage.

39
 o G_SI_image_perc_geo, valid acquisitions percentage (covered by at least one valid
measure).
o G_SI_disc_counter_geo, disconnection counter (SBAS disconnected blocks disconnection
counter).
o G_ALOS_geo, Azimuth Line of Sight. Positive angles are measured clockwise from the
North; negative angles are measured counterclockwise from the North.
o G_ILOS_geo, Incidence angle of the Line of Sight. The angle is measured between the
Line Of Sight and the vertical on the ellipsoid (flat earth).

The output results in “vector” folder consist of:

• G_vVdV_hHdH_cCdC_X.shp where VdV is the precision velocity threshold, HdH the height precision
threshold, CdC the temporal coherence threshold and X the shape part id. Shapefile of the SBAS with
the associated files (.shx, .dbf and Google Earth .kml). If the number of points is larger than the Shape
Max Nr of Points, several shapefiles are created.

Figure 19 Geocoded mean power G_SI_mean_pwr_geo image upper left, geocoded final average displacement
rate G_SI_velocity_geo upper right, ellipsoidal DEM G_FI_dem lower left and Root Mean Square Error [mm]
G_SI_RMSE_mm_geo lower right.

The analysis of the time series can be performed both on the displacement stack in raster format (using the
corresponding meta file) and on the generated shapes. The visualization tools are found in the General Tools
menu: Time Series Analyzer Raster and Time Series Analyzer Vector, respectively. The shapes provide the
product measures in point cloud format with a precise geolocation in 3D (i.e., at the maximum resolution
obtainable from the available multi-looked slant range geometry dataset), while the meta raster provide a grid
40
of product measures regularized in 2D (at the sampling specified by the user with X Dimension (m) and Y
Dimension (m)), averaged and interpolated. See Annex 2 and Annex 3 to have more details about the plot of
deformation time-series.

The analyses of the different output layers can help on understanding the deformation behavior of the study
area and its temporal evolution.

In particular, the mean velocity of deformation is helping on identifying unstable areas, but is not providing
information on the evolution of deformation, that is provided from each pixel time series. An additional
information can be retrieved from the RMSE, describing how the deformation model (in this case a linear
model) is able to fit the deformation behavior. Therefore, in case of high RMSE values, a certain divergence
from the deformation model can be identified.

Figure 20 show two exemplificative stable points characterized but a low and a high RMSE (0.92 and 1.76
respectively).

Figure 20 Time-series of two stable points characterized by a low RMSE (blue curve) and high RMSE (green
curve).

41
Annex 1. Auxiliary File

The auxiliary file is used by all stacking interferometry steps. This file is the stacking central thread; it reports
the updated list of steps and sub-steps performed so far by the SBAS and its path location describes the
internal path folders and sub-folder locations to the software interfaces. In case a step of the stacking
interferometry processing is stopped for any reason, the same can be started again and the SBAS will be able
to continue from the last processed pair on, without re-starting from scratch. At present time this feature is
limited in the cluster version, where only the steps performed are considered, not the single pair counter.

The "Auxiliary file" ([Link]) is saved in the root output directory and it is updated during the execution
of the different processing steps. It is generated by the Generation Connection Graph step. It is the input for
the Interferometric Process step onwards, throughout the whole processing chain; it is important to note that
the first input to enter, in any processing panel, is the [Link].

TIP To further simplify this procedure, the steps to be executed can be stored in a batch sequence so that
it will be very easy to stop them and let them continue again by exploiting the Batch Browser tool (see
the Getting Started tutorial for more information.

It is possible to edit this file when the user decides to perform again a particular operation. For example: while
looking at intermediary results, it was discovered that a new tuning of the parameters should be done. To
remake a sub-step, the auxiliary file must be edited by setting “NotOK” the corresponding sub-step and setting
to 0 the relative counters. Once the SBAS step is run again, it will start from the step edited to “NotOK” in the
Auxiliary files and will go on, by skipping the previous sub-steps and hence saving time. Consider that the sub-
steps are ordered according to their execution sequence. It means that all the following sub-step after the
“NotOK” will be performed again.

Example for SBAS: the interferogram generation (and the next steps) must be re-done:

Portion of the Auxiliary as before re-starting:

< interf_stack>OK</interf_stack>
< interf_stack_master>20</interf_stack_master>
< interf_stack_slave>0</interf_stack_slave>

Same portion after proper editing, that will force the re-execution of the interferogram generation when
the tool is re-started:

< interf_stack>NotOK</interf_stack>
< interf_stack_master>0</interf_stack_master>
< interf_stack_slave>0</interf_stack_slave>

TIP The exact counters of “interf_stack_master” and “interf_stack_slave” can be retrieved in the
CG_report.txt file (_SBAS_processing/connection_graph folder). This Report file contains the list of
images to be processed. The processed pair nomenclature is the following:
_MasterDate_m_ID_SlaveDate_s_ID. In this way, it is easy to retrieve which are the counters of the
pair under analysis by looking at the acquisition dates and/or the acquisition IDs.

TIP In case the user might also decide to run an entire stacking step again, no modification to the
auxiliary file is needed. It is possible to run the desired SBAS tool again setting the Rebuild All flag to
True.

42
Annex 2. Meta viewer

Here is described the correct procedure to open, visualize and use a meta file.

Load for example the meta file corresponding to the final SBAS displacement time-series results:
“G_disp_geo_meta”, located in the “geocoding” SBAS folder. The file contains the displacement time series,
the average velocity, and other ancillary geocoded layers.

TIP Whenever the user moves the SBAS folder location, the meta internal links (used to load the layers in
the ENVI viewer) refer to a wrong absolute path and the meta would not be opened. For this reason, there
is another meta file (with the same name) directly inside the sub-folders containing the files to be opened
(in this case the “raster” sub-folder).

TIP The SBAS meta files are mainly used by the Time series Analyzer Raster tool. For a visual check
purposes (for example, visual analysis of the unwrapped phase stack) the use of .series files is
recommended.

Notice that opening a meta file in ENVI, will cause the automatic RGB visualization of the first 3 bands in the
meta list. This coloured map has not a physical meaning, so the operator must remove it from the “Layer
manager” column tab (right click of the mouse over the visualized meta layer name): see Figure 21.

Figure 21 Open “G_disp_geo_meta” file in ENVI and remove the automatic.

Once the RGB has been removed from the screen, open the Data Manager, and load the desired layer
(“SI_velocity_geo”) in the ENVI viewer (see Figure 22). Stretch its histogram for a better visualization.

43
Figure 22 Open “SI_velocity_geo” file in ENVI and correctly stretch between -10 and 10 mm/year.

To visualize the time series (and the ancillary info) for any pixel, just open the /SARscape/General Tools/Time
Series Analyzer/Raster tool and select the meta layer in the “Layer Manager” column tab, click on the desired
pixel and plot the time series by clicking “Plot” on the “TS Raster Analyzer” panel (see Figure 23).

Figure 23 Time series Analyzer Raster.

44
Annex 3. Time Series Analyzer

The Time Series Analyzer, which shows the displacement information extracted for a selected point, can be
found in /SARscape/General Tools/Time Series Analyzer/Vector (to manage shape files, see Figure 24)) or
/SARscape/General Tools/Time Series Analyzer/Raster (to manage raster meta files). In this tutorial, only the
analysis of vector results is considered.

To analyze the displacement time series of a SBAS, upload all the .shp results as well as the G_mean_pwr_geo
image (as background) in ENVI.

After setting the display range of the SBAS shapefile on the Time Series Analyzer panel, to activate the color
scale press the Color Apply button. Once the desired point has been selected directly clicking on it, click on
Plot Time Series in the Time Series Analyzer panel to open a plot showing the history of movements over time
(see Figure 24 and Figure 25).

TIP It is worth to mention that due to limitation on the shapefile format, that it cannot exceed 2 GB,
approximately 70-million-point features are allowed at most. For this reason, SBAS points are split into
different shapefiles. To avoid it, increasing the Product Coherence Threshold (or reducing the velocity and
elevation precision thresholds) will reduce the number of points, so they could be stored in one single .shp
file. However, the point density of the results will be lower.

TIP When the result is split into more shapes, opened simultaneously, before using the TS Vector Analyzer
tool, make the desired shape active: right click on the selected .shp file and select Set as Active Layer.
This can be done even leaving the TS Vector Analyzer window open. Choose the same data range to
homogeneously stretch the color visualization for all the different shapefile (see Figure 24).

The results of the SBAS processing in the Urayasu area can be seen in the Figure 24. The figure shows that
the processing results highlight a widespread stable behavior (green areas). However, it is characterized by a
significant subsiding area, mainly focused on the red boxes in the Figure 24.

Figure 24 Time Series Vector Panel, G_mean_pwr_geo image and G_v8d0_h10d0_c0d1_0.shp shapefile with
measurements.

45
The displacement in the area ranges between -10 mm (blue) and +10 mm (red). The green and yellow place
marked of the selected points show the location of the corresponding time series presented in the plot.

The plot below (see Figure 25) shows that the selected green point is characterized by a cumulated
displacement of about 30 mm. On the other hand, the inland area can be considered stable.

Figure 25 Time series over two subsiding points in the Urayasu area.

TIP We point out that depending on the video card, the use of a remote desktop connection does not
allow to show different colors on the shapefile. To remedy this limitation, connect directly to the remote
location and restart the machine or copy the results to a local computer.

46
Annex 4. SBAS Edit Connection Graph

This tool is used to remove or add connections or acquisitions from the Connection Graph defined by the
Connection Graph Generation tool.

Input files

Insert the [Link] file.

Principal Parameters
Add Pair
The pair selected in the "Master List" and "Slave List" is added to the original connection network. The process
must be re-run from the Interferometric Process step. The interferometric pairs already processed in the
previous iteration are not generated again; only the new entries will be processed.

Remove Pair
The pair selected in the "Master List" and "Slave List" is removed from the original connection network.

Add Image
The input image list is added to the original connection network. The process must be re-run from
the Interferometric Process step. The interferometric pairs already processed in the previous iteration are not
generated again; only the new entries will be processed.

Remove Image
The image selected in the "Master List" or "Slave List" is removed from the original connection network.

Automatic
The remove pair editing is done automatically accordingly with the Edit Reject Automatic Strong Discontinuities
(%), Edit Reject Automatic High Std (%) and Edit Reject Automatic Poor Coverage (%) parameters. The reject
and keep stacks meta/series are generated to visually check the automatic editing behavior. The corresponding
last processed unwrapped meta/series is updated as well.

Edit Reject Automatic Strong Discontinuities (%)

By setting a reject threshold higher than zero and clicking the Automatic button of the SBAS Edit Connection
Graph panel, the worst interferograms (a percentage of the whole interferogram stack) are discarded from
the stack. The metric for the unwrapping interferogram sorting is the number of wrong discontinuity phase
cycles with respect to the number of valid pixels.

Edit Reject Automatic High Std (%)

By setting a reject threshold higher than zero and clicking the Automatic button of the SBAS Edit Connection
Graph panel, the worst interferograms (a percentage of the whole interferogram stack) are discarded from
the stack. The corresponding last processed unwrapped meta/series is updated as well. The metric for the
unwrapping interferogram sorting is the unwrapped phase standard deviation.

Edit Reject Automatic Poor Coverage (%)

By setting a reject threshold higher than zero and clicking the Automatic button of the SBAS Edit Connection
Graph panel, the worst interferograms (a percentage of the whole interferogram stack) are discarded from
the stack. The corresponding last processed unwrapped meta/series is updated as well. The metric for the
unwrapping interferogram sorting is the valid pixels coverage.

By adding connections or acquisitions, it will start an Incremental Mode process: the previous interferograms
will be preserved, while the new ones will have to be generated. For this reason, the incremental mode requires
the user to run again from the Interferometric Process step onward, as evidenced by the pop-up displayed.

47
 TIP It is good practice to store the processing steps in the Batch Browser to process new interferograms
with the same parameters as those previously made. Eventually the user can find the parameters
previously used in the file "_SBAS_processing\work\work_parameters.sml".

TIP Remember that if the user goes through the Connection Graph Generation step, the whole old process
will be lost, thus losing the benefit of using incremental processing.

This tool can be used whenever the user decides to remove one acquisition, or more frequently some pairs
that have been considered incorrectly unwrapped (e.g., after the Interferometric Process step or after the
Inversion First step) or characterized by a low spatial coverage. This editing is finalized to obtain a better
result from the SBAS inversions, starting from less but more coherent input unwrapped interferograms. This
tool should be used after the unwrapped phase image stack visual check. The pairs to be discarded can be
manually identified by opening the corresponding “.series” file as shown in figure Figure 26.

Typically, the main reasons to discard an interferogram are the followings:

• Poor coverage due to the low coherence characterizing the interferometric pair (phenological changes,
snow coverage, etc.)
• Spatial phase discontinuities caused by the limit of unwrapping tool (fringe aliasing, poor coherence,
etc.)

Figure 26 Unwrapped phase series ("_SBAS_processing\interferogram_stacking\IS_upha.series") with

RAINBOW colored scale.

To select the pairs-connections to be removed, just select the master and one or more slaves, and click on
the Remove pair button, as shown in Figure 27. The software will check if the operation can be performed,
depending on the graph connectivity (Allow disconnected Blocks option activated or not) and on the Super
Master requirements (the Super Master cannot be removed and must have at least one connection). It is up
to the user to decide which method is easier to identify the connections: by looking at the acquisition dates of
master and slave, or at their ID (xx_m_ID__s_ID_).

48
No interferograms should be removed in this tutorial because no relevant lack of coverage or unwrapped
phase discontinuities are identified. Only as an example of use of the tool we remove the following
interferogram: master 18 - slave 12.

Figure 27 SBAS Edit Connection Graph dedicated tool.

The actual editing step and the number of added and discarded acquisitions are reported in the panel. An
acquisition can be discarded because explicitly removed or because all its connections have been removed.

TIP It is recommended to remove not more than 20% of the starting interferometric stack in order not to
reduce too much system redundancy.

49
Annex 5. Refinement and Re-Flattening

Automatic Refinement and Re-flattening – Available from version 5.6.2

The refinement and re-flattening are automatically performed at the beginning of the “Inversion: First Step”,
the “Inversion: Second Step” and the “Geocoding” steps. These operations are performed before the inversions
to estimate and remove the remaining offsets and ramps from the ingested unwrapped phase stack and final
calibration is applied to the slant range products (velocity, height and displacement) during the geocoding
step.

The goal of the automatic Refinement and Re-Flattening, for the first and second inversion, is the definition
of the stable areas. The algorithm will use the coherence, mu-sigma, modeled displacement rate and residual
topography layers to define stable areas, with high coherence and small residual topography; a different mask
file will be created for each inversion; the files is located within the work folder of the first and second inversion
(project_name_SBAS_processing/work/…..inversion):

Figure 28 Stable areas mask

(project_name_SBAS_processing/work/work_first_inversionFI_initial_refinement_mask)

The created mask is applied to each unwrapped interferogram in order to select valid GCPs according to the
coverage of the single interferogram.

In the first inversion the ingested unwrapped interferogram stack must be reflattened to remove residual
phase offset and trends. The stable areas mask is defined iteratively based on the results of the first inversion
itself.

In the second inversion the ingested second unwrapped interferogram stack must be reflattened to remove
residual phase offset and trends that could have been introduced by the unwrapping algorithm; see chapter
8. Inversion: Second Step for more details. The stable areas mask is improved based on the results of the first
inversion:

50
Figure 29 Comparison between the first and second inversion mask.

The final Reference Point is generated during the geocoding process and it is applied on the final slant range
products (velocity, height and displacement); this point is characterized by zero displacement and zero residual
topography, all the pixels within the area of interest will be calibrated to this final GCP. This point is saved in
the geocoding folder of the SBAS processing folder
(project_name_SBAS_processing/geocoding/G_refinement_GCP.kml):

Figure 30 Reference Point location

51
Manual Refinement and Re-flattening
The user can decide to run the Refinement and Re-flattening manually by inserting the optional Ground Control
Points (GCP) in the “Refinement GCP file” or let the software create masks to automatically identify which
areas are stable and without residual topography (as described in the previous section). In this annex we add
some details to the tutorial dedicated to the creation of GCPs (see the Tutorial “Ground Control Point”),
oriented to the SBAS processing.

The GCP for refinement can be selected both in slant range geometry and in geocoded geometry and can be
unique for all SBAS steps that require it.

The location of the GCPs should respect some general criteria (see the Tutorial “Ground Control Point”) applied
to most of the ingested unwrapped interferograms or to the slant range products. It is suggested to select a
representative image, during the analysis of the unwrapped interferograms stack or slant range products, in
terms of coverage (pixels over coherence threshold) and stable areas identification capability.

It is difficult to find the perfect set of points that can work in the same way for all pairs, having different
coverage given by the different coherence values. For this reason, it is suggested to introduce many points
(at least 20-30) well distributed. In this way, it is more likely that almost all the interferograms will be well re-
flattened and, above all, that all unwrapped interferograms match at least one GCP in a valid position (where
the coherence is above threshold).

TIP The GCPs are generally chosen in slant ranges, but if they are used as common anchor points between
several datasets (different sensors or different acquisition geometries), they must be generated in a
common geocoded geometry. The software that will automatically reproject the points from the geocoded
geometry to the corresponding slant range geometry in use. The dataset in descending geometry of this
tutorial has been designed to be used in combination with the ascending geometry, and it is in this
perspective that we have decided to generate the points in geocoded geometry.

To generate the GCPs in the geocoded geometry: choose one or two representatives unwrapped
interferograms (or slant range product such as the “SI_velocity” layer) from each available processed stack
(in this tutorial only one interferogram will be taken) and geocode them using the dedicated Geocoding and
Radiometric Calibration tool found in the basic module. These will be the references to define the position of
the GCPs.

52
Figure 31 Panels to geocode a representative unwrapped interferogram.

Once the reference layers have been geocoded, proceed with the generation of the GCP file using the dedicated
workflow (see Figure 32).

53
Figure 32 Generate Ground Control Points GCP selection workflow. This interface can be accessed from the
binoculars button or directly from the dedicated tool in the General Tool menu section.

Figure 33 GCP (shape) position in geocoded geometry over coherent areas (above the coherence threshold)
and over probably stable points.

Output

The corresponding processing folder includes the following products:

• _refinement_GCP.xml, which is a copy of the input refinement GCPs used.
• _refinement_GCP.kml, which is a Google Earth visible version of the input refinement GCPs used.
• _refinement_GCP_geo.shp, which is a slant range shape version of the input refinement GCPs used.
• _refinement_GCP_slant.shp, which is a slant range shape version of the input refinement GCPs
used.

Generated meta and series files:

• _reflat_fint_meta, _reflat_fint.series, which refers to all re-flattened wrapped interferograms.
• _reflat_upha_meta, _reflat_upha.series, which refers to all re-flattened unwrapped phases.
• _reflat_, which refers to all slant range re-flattened products.

TIP It is very important to check the re-flattened unwrap phases and the re-flattened slant range products
detecting the residual inaccurate orbits ramp which may require a better positioning of the refinement
GCPs.

54