The Kalideos Databases: A Framework for the Scientific Analysis and Exploitation of Sentinel Images
Rabaute, Thierry1; Lafrance, Bruno1; Imbo, Patrick1; Selle, Arnaud2
1CS Systèmes d'Information, FRANCE; 2CNES, FRANCE

The Kalideos programme was born in 2001 from the CNES will to support research in remote sensing through the development of "reference remote sensing databases". Such databases aim at providing researchers with time series of multi-sensors and multi-resolution remote sensing imagery.

Up to now, four reference datasets have been developed over multi-thematic sites of about 50 x 50 km2:

  • The first one, the Adam database, has been developed over the site of Fundulea, a plain area east of Bucharest, Romania, to carry out wheat crop modeling analysis;
  • The Isle-Reunion database was later implemented over the Reunion island, a French overseas department in the Indian ocean, to study various topics (agriculture decision making, land use/land cover, risk management...) concerning this small but contrasting territory;
  • Then, the Littoral database was set up over the Arcachon basin by the Atlantic coast, France, focusing on coastal environment and forestry related studies;
  • At last, the development of the OSR MiPy database has started on the outskirts of Toulouse, France, to study environmental resources, water consumption and quality issues, as well as town and country planning in periurban areas.

    The basic principle of the Kalideos databases consists in gathering satellite images and accurately preprocess them in order to provide temporally consistent datasets, along with exogenous data (maps, physical parameters, land use information, in-situ measurements...):

  • The applied geometric correction aims at providing perfectly coregistered image series. To reach this objective, a dedicated preprocessing chain has been developed and is now operated by CS, allowing to get spatially consistent image time series with a relative positioning accuracy generally lower than 1 pixel, i.e. 2.5 meters in the case of Spot 5 THX data.
  • A radiometric preprocessing is also applied to the images entering the Kalideos databases to make them temporally coherent. In the case of optical data, it is based on the 6S radiative transfer code, improved with slope and environmental effect corrections, and aims at producing Top of Atmosphere and Top of Canopy reflectances, so that the resulting information can be directly integrated into models using biophysical parameters.

    The optical datasets integrated in the databases are essentially made of Spot and Formosat imagery, but QuickBird and GeoEye products are also available in some cases. Concerning SAR data, most of the information comes from ERS, Envisat and, more recently, TerraSAR-X sensors, but a few Radarsat data can also be retrieved.

    The Kalideos databases are then disseminated towards the scientific community through a common dedicated Web portal (http://kalideos.cnes.fr), from where image time series can be retrieved on a free basis, provided applying scientists have requested a protected access granted by CNES and that they limit the use of the data only to research purposes. This portal supplies all the information requested to join the Kalideos network and gives short descriptions of the most significant achievements of the programme and indications about the contributing scientific teams. Moreover, it offers access to documentation supplied by the involved researchers as a feedback from their work (scientific papers, presentations, technical reports, relevant Web links...).

    One of the key issues of the Kalideos programme is that the databases have now turned into focal points for the development or the improvement of Earth observation based applications, as well as the assessment of recently launched or future sensors with regard to thematic issues. As a consequence, some of the corresponding sites have been selected as priority test areas for the thematic in-flight commissioning activities of the Pleiades system. They are also among the test sites considered for the Spot 4 Take Five experiment initiated by CNES late January to supply researchers with Sentinel-2 precursor imagery during a few months.

    Looking at the variety of data available, it can be considered that the Kalideos programme has been anticipating full and open access to "GMES like" Earth observation data for scientists since 2001. Indeed, among the GMES spatial infrastructure, Sentinel-1 and Sentinel-2 are particularly awaited by the Kalideos community to get high resolution high frequency SAR or optical data in order to perform temporal analysis of several phenomena like crop systems, forest, vegetation phenology, soil moisture...

    After an overview of the current status and content of the Kalideos databases, some examples of use of these scientific reference datasets will be presented and discussed. Indications about the potential integration of Sentinel-1 and Sentinel-2 data and their use for scientific analysis will also be given.