The data are organized in a gridded data structure, i.e. data cubes. The tiles manifest as directories in the file system, and the images are stored within.
Check out this tutorial, which explains what a datacube is, how it is parameterized, how you can find a POI, how to visualize the tiling grid, and how to conveniently display cubed data.
Data Cube definition
The spatial data cube definition is appended to each data cube, i.e. to each directory containing tiled datasets, see Data Cube definition.
The basename of the output files can be defined in the parameter-file. The basename will be appended by Module ID, product ID, and the file extension.
The images are provided with signed 16bit datatype and band sequential (BSQ) interleaving in one of the following formats:
This is the recommended output option. Images are compressed GeoTiff images using LZW compression with horizontal differencing. The images are generated with internal blocks for partial image access. These blocks are strips that are as wide as the
TILE_SIZEand as high as the
ENVI Standard format
This produces flat binary images without any compression. This option might seem tempting as there is no overhead in cracking the compression when reading these data. However, the transfer of the larger data volume from disc to CPU often takes longer than cracking the compression. Therefore, we recommend to use the GeoTiff option.
Metadata are written to all output products.
For ENVI format, the metadata are written to the ENVI header (
For GeoTiff format, the metadata are written into the GeoTiff file.
If the metadata is larger than allowed by the GeoTiff driver, the excess metadata will be written to an “auxiliary metadata” file with
FORCE-specific metadata will be written to the FORCE domain, and thus are probably not visible unless the FORCE domain (or all domains) are specifically printed:
gdalinfo -mdd all 20160823_LEVEL2_SEN2A_BOA.tif
Currently available metrics are unique patch ID, weighted mean patch area, weighted mean fractal dimension index, edge density, number of patches and effective mesh size. Additionally, arithmetic mean, geometric mean, standard deviation, maximum value and the area of the foreground class within the kernel are available.