Preparation of the Sentinel-2 space mission

Sentinel-2 : an ESA's operational mission

Sentinel-2 is ESA's operational mission for the observation of land surfaces with a decametric resolution.

Two Sentinel-2 satellites will soon be launched : official launch dates are end of 2013 for Sentinel 2A, and beginning of 2015 for Sentinel 2B.

They will provide a systematic coverage of all lands with a repetitivity of 10 days with one satellite, and 5 days with two. Images will have a swath of 290 km, and a resolution of 10m, 20m or 60m depending on the spectral bands (there are 13 spectral bands including 3 in the Short Wave InfraRed (SWIR). The Sentinel-2 satellite system will be largely better than LANDSAT, on all aspects except the presence of thermal infra-red band.

Sentinel-2 spectral bands

ESA is responsible for the development of the spacecrafts, instruments and ground segment. Within a cooperation agreement, CNES is in charge of :

• the development of the ground prototype processor
• the definition and prototyping of level 1 processing.
• the development of an image quality system (to calibrate the instrument and validate its performances)
• the definition of a demonstrator for level 2 processing (cloud detection and atmospheric correction). This demonstrator will be based on the Venµs Level 2 processing methods , and will be designed by CESBIO.

Level 2 processing :

CESBIO has studied the relevance of the algorithms designed for VENµS to Sentinel-2 data. The main differences between Sentinel-2 and Venµs lie in :

• the greater field of view fo S2 (290km intead of 30km)
• the better repretitivity for Venµs (2 days instead of 5 to 10)
• the availability of SWIR bands for Sentinel-2.

Venµs Level 2 algorithms have been applied to LANDSAT data to test their applicability to Sentinel-2 : using time series from two LANDSAT satellites, with a field of view of 180 km and the availability of SWIR bands, we obtain data sets rather close to those of Sentinel-2.

The applicability of Venµs algorithms has been assessed :

• the multi-temporal cloud detection method is fully applicable to Sentinel-2
• the multi-temporal aerosol detection method will only be applicable to the full Sentinel-2 system based on two satellites, because the repetitivity obtained with only 1 satellite will not be sufficient.
• a backup method has been designed based on multispectral relations to process the Sentinel-2 data for the period with only one satellite. This backup method works quite well provided there is sufficient vegetation on the site.

Legend : Example of a LANDSAT image processed using Venµs algorithm, Albuquerque, USA is in the top right of the image.
Top of Atmosphere (TOA) image after cloud detection (clouds are circled in red and cloud shadows in black).
same image after atmospheric correction (mouse over).