Summaries and Recommendations of the Fringe 2007 Workshop
The Fringe 2007 summaries and
recommendations have been prepared by the session chairpersons and are grouped
Methodology: General - Session
Studies for new and/or improved
algorithms are necessary to fully exploit high resolution SAR data from new
Simulations can be useful to better
understand the scattering mechanisms.
Advanced techniques like SAR
tomography and differential tomography, multi-aperture interferometry,
Pol-InSAR, etc., will become more important with the availability of SAR data
from the new missions (characterized by frequent revisit time, high resolution,
wide-band, polarimetric capabilities, etc.)
Methodology: DINSAR/PSI - Session
Session Summary &
People presented combinations of
point-based (full-res, time behaviour) techniques and (in combination with)
coherence-based (reduced res, region growing) techniques. Both seem
Partially coherent targets exploited together
with PS: better exploitation of the archives (pixels that would be discarded if
only one of the two methods was used) (Perissin)
Unwrapping techniques : 3D
(information in time and space should be used in solving the problem)
How to optimize the selection of
interferometric combinations. (Duque)
There are still problems in producing
error bars (strong dependency of data availability, distribution, signal of
interest, processing parametersâ¦etc).
Quality assessment of the model (not the data). Case study dependency?
What is the value of a priori knowledge?
Unwrapping is a problem, APS too
Riccardo Lanari: trends in estimated
velocity fields (should they be removed y/n?)
PS Quality vs. density.
APS should be made visible, how should
it be validated?
Observation: 3 categories of "users":
scientists, companies, end-users (non-radar experts) : different interests,
different knowledge levels, different objectives.
Should persistent scatterer (full
resolution) and coherence (multilooked)-based methods be considered as
independent or complementary means of extracting information from satellite SAR
What are the main bottlenecks in terms
of quality assessments for the various techniques?
Is it possible to make generic
statements on the quality of the estimated deformation parameters, independent
of the area of interest, or are these always case-study dependent?
How should the trade-off between point
density and quality be considered?
How can we parameterize the
information content of PS time series? (to interpret time series different from
linear velocity model)
How should advanced DInSAR algorithms benefit
from spatial + temporal phase unwrapping?
Significantly more experiments are
needed. The availability of the required data is paramount.
ESA should make the historical SAR archive
available with easy ftp access and no data charges.
Methodology: Atmosphere - Session
Session Summary &
Use of Numerical Weather Models (Etna, Holley)
APS-Analysis: Stochastic modeling (Knospe)
APS Analysis: ERS-Envisat (Perissin)
Discussion: Use of NWM for other regions (worldwide) should be
Do we feel that the problem of atmospheric delay signal in SAR
interferometric approaches is well-understood?
We distinguish effects of (i) vertical stratification, of
importance in case of strong topographic height differences, and (ii)
turbulent mixing. Should both effects be tackled independently?
Is it possible to uniquely identify a spatially correlated
interferometric phase error due to ionosphere using current sensors? If
yes, which empirical evidence is available?
What is the value of Numerical Weather Models for (i) local
case studies, and (ii) systematic correction of APS irrespective of
location and time
What is the value of the
interferometrically derived atmospheric phase screen for operational
meteorology and atmospheric research?
Significantly more experiments are needed. The availability of
the required data is paramount.
ESA should make the historical SAR archive available with easy
ftp access and no data charges.
Applications: Earthquakes and Tectonics - Session
Session Summary &
A full-day session was dedicated to Earthquakes and Tectonics.
14 Oral presentations and ~13 poster presentations in the
Many subjects presented, including inter-seismic, co-seismic,
post-seismic, rifting deformation, model parameter estimation methods,
Mostly data from the ERS/ENVISAT archives, but also ALOS,
One presentation focused on ScanSAR InSAR.
Long-term fault slip rates and locking depths were estimated
from InSAR in several locations, including Tibet and California.
High precision of interseismic velocities was achieved by
careful treatment of orbital and tropospheric errors, although problems
remain in areas of high topographic gradients.
Several presentations focused on coseismic deformation and
modeling, e.g. in South America, Middle East, and Africa.
ScanSAR interferogram of the magnitude 8.0 Peru earthquake
(August 2007) was presented and showed the advantage of wide-swath for
very large events
The advantage of using multiple look directions and full data
covariances in model parameter estimations was demonstrated.
A few studies were presented on post-seismic deformation
observations and analysis, e.g. from South America, Iran, and Iceland.
Spectacular images and analysis of the 2005 Dabbahu rifting
episode were presented.
Sub-millimetre/year surface deformation rates across the Asal
rift, Djibouti, were derived from a 10-year InSAR time-series.
Visco-elastic response of the crust, due to water level
variations of lake Mead, Nevada, was constrained using a 15-year ERS/Envisat
With more than a decade of operation of C-band radar, InSAR
measurements of ground deformation have improved in accuracy and its
assessment has to be re-evaluated. What is the smallest ground deformation
signal that can be measured with InSAR? Over what spatial and temporal
The precision of ground velocities has
to be evaluated on case by case basis.
General limiting factors include data
availability and time span.
Site-specific factors include surface
coherence, topography, tropospheric conditions, and spatial scale of the
Examples shown with
sub-millimeter/year precision in ideal cases (availability of long time series)
when observing steady deformation rates.
Long temporal series of radar data allows scientists to
estimate rates and changes in rates of slow deformation processes using
approaches such as SBAS and PS. What is the accuracy on constant rate
estimates? What level of rate change can be estimated? Over what time
Quality of deformation rate-change estimations was not
With displacement and rate estimate errors we currently
achieve, what are the deformation processes that we can reliably resolve?
(Fault slip/creep, poro-elastic deformation, visco-elastic relaxation, dyke
expansion, magma chamber inflation/deflation). Can the trade-off between
competitive processes (e.g., after-slip and visco-elastic relaxation) be
resolved with InSAR monitoring of post-seismic movement?
Important to include other
information, e.g. from geology and seismology, due to non-uniqueness of models
describing geophysical processes.
Rapid response and systematic data
acquisition are needed following an event to constrain processes characterized
by different time constants.
Important to combine different look
directions (and azimuth offsets) to help constraining the 3D deformation field.
The phase propagation delay through the troposphere has been
identified as the main source of error for ground deformation
measurements. What recent developments have we achieved with the use of
ancillary data such as the MERIS or GPS nadir delay maps etc? Can long
temporal data series help characterize the tropospheric signal?
Addressed in several presentations
during the session and remains one of the main limitations to achieve high
precision deformation measurements.
Methods include phase-topography
correlations, atmospheric models, ground- and space-based meterological data
(including MERIS), APS estimation, atmospheric error characterization.
Half of the speakers clearly
demonstrated improvement in deformation retrieval by using one or more of the
How does the multi-directional observations
(ascending/descending, near/far range) help constrain the source
parameters (geometry of faults and slip direction)?
Multi-directional observations greatly
improve model parameter estimates, in particular when full data covariances are
accounted for in the model parameter optimization.
What is the advantage of ScanSAR interferometry? What can we
observe with scan-InSAR that was not accessible to conventional InSAR?
Should scan-SAR become the background mode of data acquisition for
Advantageous for large
spatial-wavelength deformation signals (large magnitude earthquakes, interseismic
Facilitates orbital corrections.
More frequent revisits at a any given
Some suggested ScanSAR should be part
of the background mission on dedicated tracks (e.g. every 4th track), in places
like Tibet, Anatolia, Iran, South America, Western North America.
Others suggested that we should
concentrate on IS2 for the reminder of the ENVISAT mission.
Limited access to software for the
scientific community is a concern
Background mission is greatly appreciated by the community and
many studies rely on these acquisitions.
The community also appreciates the extension of the ERS-2
mission, but is concerned about gap between the Envisat and Sentinel-1
Enormous potential research opportunities exist in further
exploiting ERS/Envisat archived data using newly developed analysis
methods. Important concerns include:
- Easy and free access to archived data.
- Required archive preservation and maintenance.
provide ERS and Envisat radar data to Cat.-1 users via FTP at no cost:
Keep it simple, all data online in level-zero
If there are technical obstacles, then
fund to solve them.
If there are bandwidth issues, then
introduce quotas (# of scenes/month).
If commercial restrictions on new data
exist, then only provide data older than for example 6 months.
Make the ESA processing software for
SLC generation available.
The important background mission should be extended to more
areas, if possible, and should have a higher priority, i.e. becoming a
The community recommends as frequent and uniform data
acquisitions as possible until the end of the Envisat mission (i.e. the
IS2 mode and consistent polarization).
However, with Envisat wide-swath InSAR being now demonstrated,
some selected tracks/orbits should be dedicated to wide-swath acquisitions
(but only if burst synchronization can be achieved)
Applications: Volcanoes - Session
- The presentations demonstrated a
significant variety of styles and scales of deformation from large
deformation events on shield volcanoes to essentially no significant
deformation at large strato-volcanoes. In addition, InSAR revealed rifting
- The importance of having in place
a robust background mission that allows the detection of unforeseen events
has been also demonstrated.
- In cases where there is
significant decorrelation at C-band or very intense deformation, L-band
data (JERS or ALOS) were particularly important.
- The atmosphere tends to be
particularly problematic for volcanoes. The use of regional weather models
may significantly reduce the atmospheric effects.
Summary of Discussion
The number of volcanoes studied has
increased with respect to past Fringe workshops; the coverage is worldwide (the
oral presentations or posters report the results of studies on volcanoes of all
continents, except Asia).
Is SAR Interferometry ready for an effective global monitoring of
volcanic activity? If not, what improvements are desirable (e.g. new missions
to reduce revisit times? Improvements in the algorithms? Improvements in the
Yes, but data are often insufficient
to effectively image and separate different volcanic deformation events (more
frequent revisit times). In many cases this can be compensated for by acquiring
data for a given volcano over multiple tracks from different incidence angles,
but for detailed time series analysis this is not ideal.
Background missions to maximize the
possibility to study unforeseen events are recommended on specific areas (e.g.
East Africa Rift).
Several papers report studies based on
the integration of multi-sensors (e.g. ERS1/2 and ENVISAT or ERS1/2, ENVISAT,
RADARSAT, ALOS) or multi-swaths dataset.
Are the current datasets satisfactory? Are multi-sensor and/or
multi-swaths analyses required to improve our knowledge of the volcanoes? Is it
necessary that the Agencies modify their data acquisition "policy" in order to
expand the type of the images available to study/monitor the volcanoes?
These questions were not significantly
addressed in the discussion.
Presentations on Piton de la Fournaise
and Mt. Etna highlighted the importance of multi-swath data both in terms of
achieving maximum temporal sampling and spatial resolution, and in terms of
using high incidence angle data (IS4+) to reduce layover effects.
In the case of L-band data (ALOS) the
current data acquisition strategy is often insufficient or unknown (i.e. East
Africa Rift) and a robust background mission for these data is desired.
In recent years several techniques to
process InSAR time series have been implemented.
Is this now a standard technique that for volcano analysis? If not,
what are the main limitations to these methods? And what are their requirements
in terms of data acquisition strategies?
InSAR time series (SBAS, StaMPS) are
well suited for volcano deformation analyses since they allow the determination
of time variable deformation. StaMPS has been applied less extensively than
SBAS. L-band would still provide significant improvement in spatial coverage.
Another key point in this discussion
was the current limited use of InSAR time series in understanding the temporal
evolution of volcanic sources. This is often caused by the rapid temporal
evolution of the volcano source (i.e. dike intrusion) that happens on a time
scale that is shorter than the InSAR repeat interval.
What is the current state-of-the-art
in volcano modelling? Is it time to put together a "cookbook" set of software
for routinely generating 3D FEM solutions for open-source software packages,
such as PyLith? (i.e. input topo, structure, boundary conditions, source
properties and easily generate Greens functions?). More generally, has
numerical modelling definitively replaced the analytical approaches?
Volcano modelling is essentially data
Spatial and temporal sensitivity of
the InSAR data are often satisfied by relatively simple models.
In cases where very large and
temporally varying deformation occur (e.g. Sierra Negra, Piton de la Fournaise;
Lake Natron), the InSAR data modeling requires more complex deformation
scenarios able to analyze the interaction of magmatic and volcano structures.
In the cases where the volcano
deformation is large and rapidly evolving the InSAR data lack the temporal
resolution required for a truly dynamic modeling interpretation of these
Another area where volcano modelling
is limited by the data is with regard to deeper deformation sources in which
the expected deformation signal is broader and weaker. The broad nature of
these source means that they often extend into areas around the volcano that
have higher decorrelation and therefore cannot be detected.
With L-band systems such as ALOS the
decorrelation in the area around the volcano
may be reduced, but then there is the problem of deformation
resolution (i.e. 1 fringe=12cm).
- What are the limitations for
detecting small time/spatial scale events with InSAR (TerraSAR
Not covered due to lack of time.
- What is the current
state-of-the-art in the integration between the DInSAR measurements of
deformations and the results of geodetic surveys or permanent networks?
Not deeply discussed, mentioned briefly in terms of modelling.
Applications: Terrain Subsidence and Landslides
Nine papers and 23 posters were presented in the terrain subsidence
and landslide session. Various applications fields, ranging from mining to
ground-water induced subsidence, from peat drainage to geotechnical
investigations on structures, from regional scale landslide mapping to
rockglaciers activity, were considered. Methodologies and techniques varied
with regard to the application field and the expertise and objectives of
authors. On the evidence of this meeting the user community has evolved to
cover a considerable number of countries around the world, including Argentina,
Australia, Belgium, China, France, Germany, Greece, Hong Kong, Indonesia, Iran,
Israel, Italy, Korea, Mexico, The Netherlands, Norway, Poland, Switzerland,
Turkey. According to the papers, the posters and the plenary discussion,
interferometric SAR techniques can be considered mature for terrain subsidence
and landslide applications.
There is a general consensus that levelling or GPS measurements are
not mandatory anymore for the validation of the SAR-derived displacement data;
on the contrary, SAR-derived measurements start to be used as validation for
other data. On the other hand, it is considered very important to increase the
maturity of the users, via training, in terms of understanding the information
content of the InSAR displacement data.
Although standardization of similar products provided by different
processing chains or providers may be considered, in every other domain it is
quite common that different suppliers provide measurements not necessarily
consistent or homogeneous.
For many uses it is more important to minimize the number of
"false alarms" instead of improving the accuracy toward smaller
One very important feature of future missions shall be data
continuity and consistency. A good trade-off may be obtained by paying some
spatial resolutions to obtain finer temporal resolution. This will also
increase the range of velocities that might be monitored.
Requests are addressed to ESA toward maintaining the ERS historical
archive and possibly to open it to the science community. Some homogeneity
between the data archives of different ESA and non-ESA missions would be
Applications: PSI Validation Session
presentations have been made covering three main areas of validation:
Firstly, the PSIC4 validation exercise of 2005-2007 with
analysis results from BRGM and TU-DELFT.
Secondly, the TERRAFIRMA Validation experiment over
Amsterdam/Alkmaar with investigations by TU-DELFT, Institute of Geomatics
(Spain) and DLR.
Thirdly, investigations focusing on artificial corner
reflectors with results from TU-DELF, University of Nottingham and the
Alaska SAR Facility (ASF).
validation exercise is still of value. It is now recognised that despite the
test site limitations for validation purposes (rapid and large deformation),
PSIC4 has stimulated further research on key issues such as:
Extension to non-linear deformation models in PSI technique,
validation exercise delivers first results. There are now quantified results of
the comparison of PSI with levelling data, e.g.:
The standard deviation of velocity differences (PSI-levelling)
in the Alkmaar test site and using ERS data ranges from 1 to 1.5mm/yr.
Using both ERS and ASAR data a new validation method, called
"validation in the parameter space" was presented by TU Delft,
demonstrating that PSI measurements are equivalent to levelling for the
estimation of the parameters of a geo-physical model used for monitoring
An extensive inter-comparison analysis of the TERRAFIRMA PSI
chains has been conducted by DLR (using the DLR results as reference);
0.35mm/y is the empirical estimation of the best attainable standard
deviation of velocity differences.
With pure inter-comparison in radar space, the standard
deviation of velocity differences is 0.5mm/y for all TERRAFIRMA PSI chains
(ERS dataset over Alkmaar).
The standard deviation of topographic correction differences is
between 1.3 and 2.7m (i.e. 3 to 6m geocoding errors, in E-W direction).
The use of artificial Corner Reflectors provides valuable
Experiments show 1.6mm standard deviation between ASAR and levelling
(2.8mm concerning zero-gyro ERS-2 data).
Current validation activities are also useful for the
preparation of future EO missions (e.g. Sentinel-1)
Can we define the &"error bars&"
to be associated with deformation velocity; deformation time series; and PS
A lot of
progress is shown concerning the validation of PSI measurement
(inter-comparison of chains and correlation between PSI and reference data);
representative statistical results are now available (rather concerning average
velocities than time series). More documented results will come with the
TERRAFIRMA validation report will be issued in Q1 2008.
Can we define the conditions to
be fulfilled to achieve the above "error bars": number of images, deformation
pattern and magnitude, deformation rates, etc?
How can we handle the spatially
wide trends (tilts) in the data, which are due to residual orbit?
It is overall
agreed that the tilt effects should be separate from the main PSI measurements.
Possibly, in the case of expert users it could be relevant to provide an
estimation of the trends (tilts) as a by-product.
Can we characterize the PSI
capability to detect deformation phenomena (e.g. in urban areas) in terms of
omission and false alarms?
capacity of the technique is linked to the PS density which has its limitations.
Areas for progress in addressing this limitation include the contribution of
the PSI technique using VHR SAR (e.g. TERRASAR).
Can we expect significant
improvements in the measurement of non-linear deformations?
A part of the
on-going research on the PSI technique is looking at approaches not based on
the linear model. In particular several research groups are active with the
PSIC-4 dataset with the objective to elaborate improved PS estimation, see
ESA is recommended
to support the PSI research community to get access to the Gardanne levelling
Airborne and Ground-Based InSAR - Session
Papers in the Session
- The X-band RAMSES multi-pass
- Monitoring of Belvedere Glacier
using a wide angle GBSAR interferometer
- Subsidences estimation from ground
based SAR: techniques and experimental results
- Bistatic Interferometric toolkit
for campaign evaluation and DEM generation
Future research & recommendations:
- Availability of wide bandwidth,
frequent revisit systems
- Integration of multiple
frequencies and multiple sensors (Spaceborne, airborne, Ground Based,
- APS is still one of main limitation,
- Tomography is one of the most
important themes for future researches
Interferometry and New SAR Missions - Session
Presentations were made on the results of new missions recently put
in orbit including ALOS (JAXA), TerraSAR-X (DLR) and COSMO-Skymed (ASI). Also ESA's
Sentinel-1 and DLR's Tandem-X were presented, fully approved missions in an
advanced stage of development.
For the radar interferometry community these new missions signify
what is sometimes referred to as the "Golden Age of SAR". Experimental mission
commonly in a single waveband are being replaced by satellite constellations at
several wavelength and a wide range of observation characteristics.
New trends include in particular:
- L-band with improved coherence over
time and more canopy penetration as compared to C-Band.
- X-band/High Resolution showing
higher persistent scatterer density and still with interesting
polarimetric information content as compared to lower spatial resolution
- C-band with largely improved
revisit and conflict-free systematic data acquisition. Improved
geographically synchronous SCANSAR (TOPS).
- For the future no detailed
requirements were suggested for new systems for two reasons. First it was
felt that the new missions offer excellent opportunities and that in
general the right options have been chosen. Second the knowledge exists
about what can be gained by adding more resources in terms of resolution,
revisit, sensitivity and what this would cost. More money buys more
- As an exception to the above,
based on the excellent results of ALOS, it is recommended to consider for
the future an operational long-term mission in L-band featuring
- In spite of the advances of
interferometry applications it was agreed that further penetration into
the end-user market was a key issue for the space agencies, value-added
companies and the scientific community.