POLinSAR 2007 Session Summaries and Recommendations
The POLinSAR 2007 Session Summaries have been prepared by
the session chairpersons based on the round-table discussions, and are grouped
in the following session-order:
- ALOS PALSAR First results
- Advances in POLSAR and Pol-InSAR
- Applications: Forestry
- Applications: Land-Agriculture
- Applications: Other
- Compact Polarimetry
- Polarimetry and Persistent Scatterer Interferometry (PSI)
- Theoretical Studies
ALOS PALSAR First results - Session
Chaired by M. Shimada and A. Moreira.
Presentations in the session
Initial Polarimetric Calibration of PALSAR (M. Shimada)
Stability and accuracy of the Polarimetric Calibration using the
Rio Branco sites, etc.
First Results of the ALOS PALSAR verification processor (P.
Evaluation of the Polarimetric data etc., ready for SCANSAR
The Dual Polarization H/a Decomposition: a PALSAR Case study
Highly utilization of H/a for FBD in forest monitoring.
ALOS-PalSAR Pol-InSAR Data Analysis: First results (K.
First results of the tree height estimation were discussed.
First Polarimetric and Interferometric Results from
Temporal stability of coherent scatterers has been evaluated.
Answers to Seed Questions
1) The ALOS/PALSAR commissioning phase has been completed,
including the polarimetric calibration. How well does the calibration fulfil
the user requirements for Pol-SAR and Pol-InSAR applications?
The calibration of the PALSAR sensor shows excellent results
which can be summarized as follows
Amplitude standard deviation: < 0.2 dB
|HH-VV| phase < 0.35 deg.
Cross talk: between -25 and -30 dB
- NESZ: - 28 dB
- HH-VV imbalance
< 0.02 dB
accuracy: ca. 11 m (one pixel)
The above values fulfil the user requirements for Pol-SAR
applications. For Pol-InSAR applications it is however required that the
satellite orbit is kept within the specified tube of 500 m (vertical) times 2.5
km (horizontal). This would allow achieving baseline values of ca. 1 km which
are well suited for Pol-InSAR applications.
2) The use of dual-pol mode relaxes the SAR systems
constraints caused by the increase of the PRF. Which applications require
the use of fully polarimetric data? What is the recommendation for future SAR
In general it can be stated that a fully polarimetric system is
preferred for urban areas, soil moisture retrieval (with vegetation cover),
forest structure, high biomass values (beyond the saturation level with
amplitude based retrieval) etc. The use of quad pol systems allows also the
determination of the Faraday rotation directly.
The trade-off between the use of a dual-polarized and
quad-polarized system is not trivial and has to be defined for each specific
application. In some cases a dual polarized system is preferred (spite of
reduced retrieval performance) due to the larger swath width and possibility to
have larger incident angles.
It is recommended to generate a handbook summarizing the
trade-off of dual x quad-pol systems for each specific application. This
handbook should be updated every year as new results of studies and
publications become available.
The conclusions of the session “compact polarimetry” are also
relevant for the design of future SAR systems.
3) Based on the current operation plan, the incident
angle is almost fixed to 21.5 degrees, and it is possible to select only
between 9.9 degrees and 26 degrees. For the time of the future operation
plan setting (probable 3 years after), which application needs smaller and
larger off-nadir angles?
The answer to the previous question can be adopted for this
question. Following comments can be added:
Dual-pol mode allows the use of larger incident angles (due to
less ambiguity constraints).
Less knowledge is available concerning the use of polarimetric
systems with very steep incident angles (e.g. 10 deg.)
For geological applications the use of dual-pol and larger
incident angles is preferred (instead of lower incident angles and quad-pol).
The trade-off between a dual-polarized and quad-polarized system
as a function of the incident angle should be further investigated for each
specific application (see recommendation for generating a handbook with the
Further discussion points/recommendations of this
- There is a strong requirement from the Pol-InSAR community
for the acquisition of fully polarimetric data of a few specific test
sites during the operational phase of PALSAR (as it has been the case
during the commissioning phase). The acquisition should be made for every
orbit cycle (46 days) in order to allow the further development of
Pol-InSAR applications. The selection of a few specific test sites should
be done in such a way that the operational data acquisition of ALOS is not
- The processed data from ESA’s verification processor has a
Doppler zero geometry while JAXA’s processor delivers the image data with
the geometry related to the Doppler centroid of the particular data set.
Users should be aware of this fact and different implementations of the
geocoding software are required for each case. This is not seen as a
problem since each processor implementation has its own geocoding
- The geolocation accuracy of the ESA verification processor
and JAXA’s processor show similar results so that both processing
strategies are valid. Again, users must be aware of the difference of the
geometry in the processed data before the geocoding.
Advances in POL-SAR and Pol-InSAR - Session
Chaired by K. Papathanassiou and E. Pottier.
Answers to Seed Questions
1) Is there any demonstrated benefit of multi-baseline
Pol-InSAR (e.g. improved classification, more stable parameter estimation,
tomography etc.)? What are the prospects for spaceborne MB (ALOS-2, Tandem-X
etc.) and do they match the requirements of the algorithms?
Multi-baseline implementations in a repeat pass scenario are of
interest for evaluating the change of polarimetric signature of distributed and
point-like scatterers in time.
The lack of appropriate (temporal) data sets makes an (accurate)
evaluation of their impact at this time difficult. However, it was recognised
that the time component can be essential for several applications ranging from
classification, to …
Multi-baseline implementations in a single (or quasi single) pass
scenario are essential for structure recovery of temporal unstable volumes (as
2) ALOS-PalSAR is the first fully polarimetric space-borne
sensor that will provide polarimetric data in a systematic way.
What are the potential polarimetric (PolSAR) applications that can be
validated and established with that data? Are the associated experiments and
What are the potential Pol-InSAR applications that can be demonstrated
and established with it? Are the associated experiments and data secured?
Is the polarimetric SAR community prepared to face in a systematic way
this unique opportunity?
The systematic acquisition of Quad-pol data from ALOS-PalSAR (if
possible every 46 days) over a number of selected test sites (representative
for each application) is a high priority request and strongly recommended.
3) What are the potential new applications that can be
demonstrated / validated on the basis of multi-temporal polarimetric data sets
as provided by ALOS-PalSAR.
See Question 1
4) With ALOS on the sky and RadarSAT-2 and TerraSAR-X on the
horizon, a multi-frequency observation scenario will be feasible at least for a
set of experimental observations. What are the potential applications that can
be demonstrated / validated on the basis of multi-frequency polarimetric
There are a number of important applications (in agriculture,
forestry, ice, geology, sub-surface …) that will highly benefit from the
availability of Multi-frequency Quad-pol data. Towards the development of the
appropriate R+D the establishment of common test sites for all actual and
near-future sensors (ALOS-PalSAR, TerraSAR-X, RadarSAR-2) is recommended and
has to be initiated / coordinated.
JAXA Constraints: Common super-test site located in the Pacific
Ocean could be a good compromise (Fiji islands?)
Applications: Forestry – Session
Chaired by C. Erxue and P. Dubois-Fernandez.
Summary of the Session
Ground-base, airborne, spaceborne radar
POL-SAR, Pol-InSAR and modelling
All tools needed to characterize the vegetation
Pol-InSAR at P band
Compact Pol-InSAR results
INSAR ground base radar
Susceptibility to wind effect
Measurements of attenuation effect & comparison to a
Mr Raimadoya presented a POL-SAR analysis over tropical area
Benefit of POL-SAR acquisitions for forest type discrimination
Use of POLSARPRO…
Vertical structure of the vegetation
More ecologically correct model
How to compute the behaviour of backscatter and frequency
Analysis of the Indrex campaign P L and X band
Good consistency between P and L
Demonstration of POLINSAR at X band
Discussions around the Seed Questions
ALOS PALSAR and forestry…
Very impressive calibration assessment results
Enthusiastic user community
More data, more coverage, more often
Forest and frequency
Results ranging from X to P band
The important parameters:
Penetration, temporal decorrelation, resolution, sensitivity
Tandem-x will allow the exploration of the X band for forestry
Techniques for forestry…
Let’s have them work together…
PolInSAR and LIDAR or ….
Could combine InSAR at different frequencies
LIDAR (scattered points with profile measurements…compared to a
Biomass or heights: both are identified as interesting parameters
Height takes its full measning when basal area info is available
Provide the product data as close as possible to the measurements
Recommendations to ESA
- Validation of the compact Pol-InSAR on other datasets at P
and L band over hilly terrain
- Support the request for more quad-pol PALSAR data from EU
users to JAXA.
- Identify and construct several international forest super
- A consistent SAR mission design is recommended enabling
Applications: Land-Agriculture – Session
Chaired by L. Ferro-Famil and P. Lombardo.
Comments and Recommendations based on the Seed Questions
1) Which major modifications would be required to export
forest volume parameter estimation procedures based on Pol-InSAR coherence
models to the case of agricultural crops?
Existing Pol-InSAR volume scattering model, developed in the
frame of forest studies at L band, should naturally adapt to agricultural crop
analysis, provided that a higher frequency band (C- or X-band) is used.
The need for modifications is not clear yet and should be tested.
2) C-band Pol-SAR data have been widely used for crop
monitoring. Are X-band (Single pass) Pol-InSAR acquisitions well adapted to
agricultural volume analysis or should C-band data be preferred?
For PolSAR surface parameter estimation techniques, volume
response may cause severe biases and must be taken into account, or reduced
using a lower frequency (L-band).
In general, Pol-InSAR techniques require higher carrier
frequencies (C- or X-band) to analyze a media volumetric response and aim to
separate it from the ground information.
For forest the addition of the InSAR allows to separate out
contribution from the surface and estimate vegetation height. In agricultural
areas, the objective might be the reverse.
The need for Pol-InSAR still needs to be proved in terms of
usefulness and improvements w.r.t. PolSAR (AgriSAR, Aquiferex projects) in both
Airborne and Spaceborne configurations.
3) Do the outputs of current PolSAR/Pol-InSAR processing
PolSAR techniques have been developed for years and are much more
mature than Pol-InSAR ones. Benefits from Pol-InSAR in this domain still have
to be demonstrated.
The accuracy of parameter retrieval techniques is still variable
and may not satisfy potential end-users.
Some of the current Pol-InSAR/PolSAR outputs (biomass/roughness)
may not be useful to the agricultural community.
End-user requirements do not seem to be well known. Clear need to
steer analysis techniques towards the estimation of useful information over
areas of interest.
4) Classification techniques using the whole PolSAR/Pol-InSAR
information segment pixels according their global scattering behaviour. In
order to provide more specific results, would it be preferable to limit the
input information to some indicators, highly linked to physical parameters
(humidity, roughness, density …)?
Classification of agricultural scenes has reached an interesting
state of maturity. Acquisition conditions and modes may still have an important
influence on results accuracy and relevance.
We need to concentrate on classification strategies that also
take into account the influence of perturbing factors (angle of incidence, view
direction …) and reduce their effects.
Applications: Other - Session
Chaired by C. Lopez-Martinez & P. Pasquali
PolInSAR signatures of a subpolar glacier (Jayanti Sharma (DLR - Microwaves and
of surface parameters from NASA/JPL AIRSAR polarimetric SAR data (Sang-Eun Park (Seoul National University))
Extraction Beneath Canopy Using L- and P-Band Data from INDREX-II: Preliminary
Results (Bryan Mercer (Intermap Technologies Corp.))
Overview of the PolSARpro v2.0 Software. The Educational Toolbox for
Polarimetric and Interferometric Polarimetric SAR Data Processing (Eric Pottier
(I.E.T.R UMR CNRS 6164 - University of Rennes 1))
of repeat-pass SAR interferometry for search of earthquake precursory
land-cover deformation; and how Infrasonic Imaging and HF-OTHR Technology can
be implemented for detecting the On-set of and real-time spreading of Tsunamis
(Wolfgang-Martin Boerner (University of Illinois at Chicago))
Answers to Seed Questions
1) Which is the current state of the art on retrieving
physical information form snow or ice covered regions? Is this retrieval
affected by environmental conditions? Are multi-frequency techniques necessary
to retrieve reliable information?
Forest seems to be the most developed application of PolSAR and
Pol-InSAR data. Nevertheless, the study of ice covered regions is not at the
same level of development. As indicated by A. Freeman, the fist task would be
to determine which parameters we need to pay attention to. He proposed three:
Snow accumulation, Ice thickness and snow velocity/movement.
Recommendation: Define products to be retrieved in the case of
snow/ice covered areas
There exists an agreement in the community that multi-frequency
systems are one way to go in the analysis and characterization of these areas.
Scattering from ice/snow covered regions contains an important component due to
volume scattering; therefore low frequencies seem to be the best option. In
order to study the volumetric structure of the ice/snow, polarimetry seems to
offer also encouraging results. Using the PCT (Polarimetric Coherence Tomography),
J. Sharma presented results concerning volumetric structure of the ice/snow
M. Bernier indicated also the importance of polarimetry when
differentiating different types of sea ice. She also raised the important
necessity to study land ice.
2) The analysis of polar, sub-polar and snow covered areas
is critical since they are perhaps the first Earth environments affected by the
greenhouse effect. Could all these regions be analyzed using the same
approaches and scattering models? or their study should be considered
J. Sharma and others indicated that the different types of snow
and ice covered environments must be considered separated differently due to
the differences in their internal structure.
Recommendation: Particularization of scattering models for
ice/snow covered regions
E. Trouve raised the point that Temperate Glaciers are an
important environment to play attention to since they may be employed as local
indicators about the global warming effect.
3) The technique reported by B. Mercer and Q. Zhang has
considered Pol-InSAR data from densely vegetated areas. It would be possible to
use the same approach in areas with different morphology, as for instance, the
Mediterranean forest characterized by being less vegetated?
B. Mercer indicated that the proposed method also works in other
environments, but in these cases attention should be paid to the accuracy with
which the different parameters are retrieved in these environments.
Recommendation: Consider accuracy aspects on the retrieved
He also indicated that the optimum solution would be a
single-pass system to avoid the problems of temporal decorrelation. The system
should operate in L-band. He also commented that a P-band system presents the
problem of low bandwidth with the consequent lost of spatial details.
4) Considering both, the important implications an
earthquake prediction system based on multi-channel SAR data would have on the
society and the current state of this technology, without neglecting its complexity,
which could be done nowadays, from a realistic point of view, in terms of
W. Martin Boerner recommended the exploitation of all the
Recommendation: One option could be data fusion techniques?
5) Nowadays, most of the applications of multi-channel SAR
data (InSAR, PolSAR, Pol-InSAR, etc…) are still on a research stage. Taking
into account that the next generation of spaceborne SAR system present fully
polarimetric capabilities, how ready are we in terms of theory (modelling and
understanding of mechanism), algorithms, signal processing techniques and tools
to deliver fully operational applications to the industry and in general to the
6) Long term polarimetric spaceborne SAR missions will open
the door to consider time as a new axis of information. Which is the current
state of the art on the use of this new source of information for PolSAR and
The main problem considering the temporal axis is the effect of
temporal decorrelation. This source of decorrelation represents a drawback in
Pol-InSAR applications. Nevertheless, even in the presence of this
decorrelation source, results may be obtained. H. Rott indicated this
possibility in the case of snow covered areas.
Comment on POLSARPRO Software
The availability of POLSARPRO SW is one of the main contributions
of the POLinSAR conferences series. The SW has a double contribution:
scientific development & education. All singing, all dancing
Compact Polarimetry – Session
Chaired by A. Freeman and J.-C. Souyris.
Session dedicated to compact polarimetry architectures – 4
3 presentations: comparisons between FP and CP architectures @
P,L,X bands by applying reconstruction algorithms. Promising results in each
Sensitivity study to azimuth symmetry assumption FP
reconstruction algorithm compromised for urban environment
CP permits to apply polarisation synthesis on receive
Comparisons between CP and FP in the framework of multi-freq.
AIRSAR data classification using SVM techniques. CP acquisition deteriorates
the classification results by about 3%.
Assessment of CP system aspects
CP assets : PRF & Swath maintained allowing acquisitions @
high inc. angles
Assessment of P band , when circular pol. transmitted – expected
to critically reduce ionospheric effects
Demonstration of the added-value of a circular pol. transmission:
invariance conditions & unbiased polarimetric analysis from Stokes
Appropriately configured CP designs can minimize the need for
gain changes due to backscatter differential, as compared with conventional
dual-pol, e.g. HH, HV
Description of a CP design compliant with a lunar mission for
frozen ice deposits analysis.
Demonstration that CP modes can be calibrated from raw data
The effect of terrain slope on CP signals must be carefully
Faraday effects: it seems that some CP options (circular pol.
transmission) would tolerate Faraday rotation to up to 10-15°. Issue of 1st
concern as P band is a good candidate for CP
Comparison between CP and FP must include both POLSAR &
Antenna trade-offs must be investigated
Technologically speaking, the design of a quad-pol mode and of a
CP mode are compatible : CP can be easily integrated on FP architectures
Earth observation : “working point” should be identified in the
framework of biomass – vegetation applications (P band) – assumptions for FP
reconstruction usually verified
Referring to the earlier presentation of P. Dubois-Fernandez, it
seems that switching from FP to CP has minimal effect on forest height
estimations using Pol-InSAR, at least for the Landes forest test site.
Ice applications : geophysical parameters not known precisely.
Relations with POL-INSAR parameters to be established first
Surface-Subsurface scattering @ large incidence angles appear to
be compliant with CP : need more investigation
Polarimetry and Persistent Scatterer Interferometry (PSI)
Chaired by K. Raney and F. Rocca.
Summary of the Session Presentations
In the first presentation, results from the Technical
University of Catalonia were presented using ground based X band polarimetric
radar. The goal of the observations was the monitoring of the subsidence due to
mining in a village surrounded by a salt mine in Catalonia.
- A 6 months survey with repeated monitoring several times a
day for six hours, every month, was carried out and very interesting
results were presented, consistent also with previous results obtained
from ERS data.
- Indeed the polarimetric characteristics of some stable
scatterers appeared to change at given times, but the different
polarimetric channels indicated very similar subsidence.
In the second presentation from Tohoku University in Japan, the effect of the orientation angle of the scatterer on the polarimetric signature
- Based on airborne JAXA L-band and NICT X-band Pi-SAR data,
experimental results were presented, over a site comprising gentle hills
in a suburban area of the city of Sendai. Indeed, the orientation angle
appeared to have different impact on the signal.
- In particular L band was more sensitive to main walls
orientations compared to X band and X band appeared more sensitive to the
random surface of the imaged objects.
A third paper, presented by Schneider and Papathanassiou
from DLR, discussed alternate ways to describe the orientation of a complex
scatterer as derived from its polarimetric response.
- While the orientation angle is dependent on the
description technique chosen, its change with the rotation of the
scatterers is not, so that the polarimetric technique can be used to
detect objects rotations.
- Symmetric and asymmetric scatterers were identified in
data acquired using DLR’s E-SAR L-band airborne system over Berlin and Dresden.
- Indeed, polarimetry allowed the identification of many
more coherent scatterers with respect to conventional techniques.
The fourth paper, from Norut IT and the University of Tromsø in Norway, discussed the application of polarimetry to the retrieval of
scatterers in natural scenes in fjords in Norway, to be able to predict
landslides that are known for causing small but very dangerous tsunamis.
- The rocky scatterers have different polarimetric
characteristics, and many coherent scatterers were detected from L band
airplane EMISAR data. Further analyses will be carried out to verify their
Answers to the Seed Questions
1) In which ways the polarimetric information increases the
detectability and the utility of the PS?
- This first seed question aimed at identifying the synergy
between polarimetry and PS interferometry. Indeed the polarimetric
characteristics are very useful for the identification of coherent
scatterers, i.e. scatterers that have a very uniform response across angles
so that a stable temporal behaviour can well be expected. However, long
surveys will be needed to assess this synergy.
2) Is the Faraday rotation detectable and is it possible to
find countermeasures, using Persistent Scatterers?
- A general discussion on the effects of the Faraday
rotation was then opened and further analyses will be welcome: it appears
that the temporal stability of the scatterers could be of use to
counterbalance its effects.
- Polarimetric returns from a rotational – invariant PS could
be analyzed to measure the 2 way Faraday rotation, and then used to assess
methods to compensate scattering matrix data from Faraday rotation.
3) In order to maximize the number of PS, which is the
optimum balance between resolution and polarimetric information? And is the
number of PS a reasonable figure of merit, or do we have a better one, in the
4) Apart calibrating the gain and the antenna directivity,
could we use the PS to calibrate the polarimetric matrix of the system too?
- The utility of compact polarimetry schemes for PS analyses
was also underlined, as the temporal stability to be expected for those
targets allows a multi pass polarimetric analysis along with the
- There was a general consensus on the utility of the
evaluation of the persisting polarimetric characteristics of the
scatterers, even if with compact schemes.
Theoretical Studies – Session
Chaired by S. Cloude and I. Hajnsek.
Comments to Seed Questions
1) Surface and Subsurface Scattering Theory: One key
motivation for using radar polarimetry is the ability to employ ratios for
robust surface parameter estimation, primarily the decoupling of surface
roughness and moisture/salinity in a single pass sensor. What is the current
status of the estimated accuracy and market impact of near and deep surface
2) Are the requirements of SP algorithms compatible with
existing sensor geometries, for example: a) Quad vs. Dual-Pol, b) those that
use cross polarisation vs. copolar ratios c) high angle of incidence
requirements for dynamic range and d) spatial resolution requirements vs.
bandwidth restrictions of low frequency sensors)? Is there any mismatch
between planned sensor deployments and requirements of scattering models?
There is a need to extend surface scattering models from single
to multi-layer to correctly account for polarimetric phase shifts
HHVV Phase and Coherence are important in surface as well as in
urban and forest applications, especially at large angles of incidence, hence
supporting the concept of compact-polarisation (e.g. mixed circular/linear
polarisation) for combined coverage and good parameter estimation at large
Support further studies for coherent polarimetric multilayer
subsurface/soil moisture modelling
High incidence angle dual-pol modes should be investigated for
subsurface studies (using existing sensors such as ALOS/PALSAR)
Need further investigation of the potential advantage of using
various compact-polarimetry modes for subsurface parameter estimation
3) Vegetation Scattering: Most quantitative inversion
algorithms for vegetation parameter estimation employing polarimetric
interferometry require a dominance of volume over temporal and environmental
effects. What are the prospects for space borne single pass polinsar systems
and can theory help inform the potential new products and increased accuracies
achievable using such systems. Is there a preference for standard or ping-pong
interferometry for vegetation applications?
4) Multidimensional Filtering: are the current filters we
use adequate for quantitative inversion, especially for algorithms employing
coherence optimisation and eigenvalue analyses? In particular what is the
effect of coherence bias on algorithm inversion accuracies and are there better
techniques we could use to remove bias, reduce speckle effects and improve
retrievals in multi-parameter surface and vegetation polinsar applications?
Single pass interferometer designs are well suited to vegetation
parameter retrieval, but care is required over choice of single versus dual
transmitter configurations for best retrieval performance.
While vector spatial filters for POLSAR are now quite mature, a
full and complete approach to multitemporal filtering is currently missing from
Future constellations should have a dual-transmit mode in order
to simplify vegetation modelling/inversion scenarios
Inversion constraints should be considered in design process of
future active/passive satellite constellations
Further studies are required to incorporate the multi-temporal
dimension in polarimetry filtering approaches (PolSAR)
5) Bistatic Polarimetry Theory: Does the use of bi-static
polarimetry offer new products or better accuracy in surface and volume
scattering applications? In particular does it allow better estimation of
vegetation biomass and surface moisture/roughness? What is the importance of
the HV-VH channel in bi-static polarimetry and how do we calibrate such new
information channels if reciprocity can no longer be used?
6) Optimisation Algorithms: there are now many forms of
optimisation used in radar polarimetry and polinsar. Can these be used for
practical product development or are they more useful in the algorithm
development/research phase. If so, given their increased complexity, what
software tools are available to implement these ‘optimum’ processing strategies
in the application sector?
Experimental bi-static polarimetric data are missing for quantitative
Bistatic modelling should be presented in terms of decomposition
A link between optimisation algorithms and parameter retrieval
need to established
Recommend acquisition of Pol-InSAR bi-static airborne data within
a campaign to support future bi-static algorithm developments