The large and ever-increasing number of satellites dedicated to Earth observation provide us continuously and globally with a huge mass of information about our environnement. Synthetic Aperture Radar (SAR) is an important class of the numerous on-board remote-sensing instruments, with its allweather and day-night imaging capabilities.

In the 90s, the first generation of civilian radar satellites (the European ERS-1, ERS-2 satellites, the Japanese JERS-1 and the Canadian RadarSAT-1) have demonstrated the potential of differential Interferometry (D-InSAR) to measure Earth displacement from repeat pass SAR time series. Specific information processing techniques are requiered to derive geophysical knowledge at the end of a long and demanding data processing chain. Measuring, monitoring and modelling efficiently Earth surface displacements is a real technical and scientific challenge because their highly heterogeneous spatial and temporal distributions. Because of its large and global coverage, its high spatial and temporal resolution, SAR imagery prove to be very competitive, compared to other remote sensing or ground-based techniques, to cover a wide part of the displacement spectrum and distribution.

Spaceborne SAR images cover large areas (up to 100 km swath) with a typical decameter ground resolution and allow displacement measurement with a precision which depends on the method and the number of images stacked over a given area. Correlation techniques allow a precision of a fraction of the pixel size (typically a meter resolution), whereas the precision achieved by differential interferometry reaches a fraction of the wavelength (5.6 cm in C band) or even a millimeter accuracy with techniques requiring large series of images (more than 30). The displacement measurement potential of satellite SAR images has created strong expectation, notably among geophysicists and opened new research fields in measuring, understanding and monitoring Earth phenomena such as seismic deformations, volcanic activity, landslides and glacier evolution. However, this potential is currently not fully exploited because of various limitations of the available data processing methods.

The need of new methodological developements is also dramatically expending with the launch of new satellites: after 10 years where only one new satellite with a SAR sensor has been launched (ENVISAT, Europe), 2006 and 2007 are the beginning of a "golden age" for SAR imagery with the launch of 4 Earth observation satellites with new generation SAR sensors in 3 different bands: ALOS (L band, Japan, January 2006), TerraSAR-X (X band, Germany, April 2007), RadarSAT-2 (C band, Canada, fall 2007) and a first satellite of the COSMO-SKYMED constellation (X band, Italy, end of 2007). The new SAR sensors will offer new observation modes including higher resolution data (up to meter resolution) or fully polarimetric data which should increase the potential and the reliability of different methods dedicated to Earth displacement measurement. At this time, no comprehensive fullyor partially automated processing chain is available. This restricts the exploitation of current and future data to a few areas. The transformation of the multitemporal 2-dimensional electromagnetic ground response, stored in Terabytes of SAR data, into geophysical knowledge, namely models with a few parameters, relies on tedious processing chains: SAR processing, interferogram generation, phase unwrapping, geocoding, artefacts corrections, geophysical model inversion... A few commercial software are now available to perform some of the processing steps (Gamma, Diapason [Massonnet 1997]...).

However, these packages usually behave as black boxes and do not allow researcher to obtain, improve or combine results in difficult cases or to develop new approaches taking their experience and the application specificities into account. A few open source software are also available such as ROI-PAC [Rosen et al. 2004] for D-InSAR and PolSARpro [POLSARPRO] for Polarimetric and Polarimetric Interferometric (Pol- InSAR) data analysis. The teams of this proposal are currently using those tools (some are active in their development), but they suffer from the lack of open source, powerful processing tools able to handle the huge amount of data which are already available in space agency archives and will increase with the new SAR sensors. Many individual efforts have been done by image processing laboratories or geophysics teams to improve the existing tools and adapt them to the application requirements. However, the French D-InSAR community, one of the most active in the 90s [Massonnet et al. 1998], needs multidisciplinary projects to develop new information processing approaches adapted to the amount of SAR data which is necessary to measure displacement signals hidden by noise and artefacts and to derive geophysical knowledge on the associated phenomena.

To achieve this difficult task, this project puts together researchers from 4 laboratories specialized in information processing (polarimetric and interferometric SAR image analysis and data fusion) and from 2 laboratories specialized in geophysics with a strong experience in using SAR imagery to measure and understand crustal deformations. Three of these laboratories were part of a previous 3-year French national project supported by "ACI - Masse de données" in 2004. This project entitled entitled MEGATOR1 obtained significant results in two directions (see here):

• in the development of processing techniques in the context of Alpine glacier motion analysis [Trouvé et al. 2007, Nicolas et al. 2007],
• in creating over the Chamonix Mont-Blanc glaciers a test site with a large multi-sensor data base: ERS archives, recent ENVISAT images, soon the results of two airborne E-SAR campaigns and futur images from the new generation SAR satellites (4 proposals accepted by space agencies).

This new "Masse de données" project includes 3 new partners who bring their experience in new complementary directions. Firstly, in the use of fully polarimetric SAR data which starts being available on spaceborne SAR sensors. The polarimetric information increases the amount of data (2 or 4 complex values per pixel) but Pol-InSAR methods as developed by IETR-UdR1 allows a better understanding of the ground properties. Secondly, in covering a wide range of displacement including seismic (LG-ENS) or volcanic (LGIT-UdS) deformations, subsidence (LG-ENS) and landslides (LGIT-UJF). The complementarity of the 6 laboratories allows us to address the three main steps of the processing chain in 3 subprojects:

1. the characterization of coherent scatterers and distributed media for displacement monitoring using polarimetric interferometric SAR data. Most of the Pol-InSAR methods are dedicated either for human made target analysis (vehicles, buildings in urban areas...) or for forestry and agricultural applications (biomass from tree height, crop monitoring...). The potential of Pol-InSAR data for displacement analysis or "Differential Pol-InSAR" is almost unexplored mostly because there were no data sets available from satellite sensors. The airborne data sets acquired during the MEGATOR project on Alpine glaciers, the arrival of fully polarimetric satellite data and the multidisciplinary experience of the teams gathered in this project make significant progress in this domain possible.
2. the extraction of the displacement measurements from large time series or new generation SAR data. Conventional two passes interferometry [Massonnet et al., 1993] remains restricted to more or less desert areas and its accuracy is severely limited by tropospheric artefacts [Hanssen, 2001]. Therefore, InSAR is still mainly used to detect relatively large signals (~10 cm or more) as coseismic, post-seismic, or co-eruption deformation. Furthermore, the whole radar data archive is still far underexploited and little is done to reconstruct deformation time series from InSAR. This subproject will show how using the whole radar data archive at a study site may improve InSAR accuracy far beyond its generally accepted limit. It will also investigate the potential and the complementarity of newly available data (polarimetric or high resolution images) for the measurement of large displacement on smaller areas such as alpine glaciers or landslides.
3. the fusion of displacement measurements from different images, different methods and complementary sources such as in-situ measurements by differential GPS. The goal of the fusion is to provide more reliable displacement fields together with confidence levels that will lead to better constrained geophysical models of tectonic and volcanic deformation. One of the difficulties due to the SAR measurements is the need to reconstruct the 3D displacement fields by using ascending and descending passes or prior information on the displacement geometry. An open issue which will be investigated is to benefit from the knowledge introduced in the geophysical model to refine the extraction of the displacement from SAR data and the fusion step in an iterative process. The methodology and the tools delivered by these 3 subprojects will be developed in the "Masse de données et connaissance" context, with signal processing and data mining approaches to handle time series and multi-channel SAR data over large areas (cf. Announcement of Opportunity, Axis 2), with information extraction and multi-source spatio-temporal data fusion (cf. AO, Axis 3) and finally with the introduction of models and knowledge from the geophysical applications (cf. AO, Axis 4). The results of this project will provide the scientific community with a complete and coherent set of Multi-Baseline processing methods, experimental data sets and results for displacement measurement from SAR imagery. By grouping information processing specialists and geophysicists, this project guarantees the developed methods to be driven by the application requirements and operational for the end-users. EFIDIR researchers share the same goal: performing and making available methodological advances to transform the archived and new generation SAR data into geophysical knowledge.