Sentinel-2 Level 3A

Description

Levels 3A are a monthly syntheses of cloud surface reflectances. A short documentation about the product is available at CESBIO’s multitemp blog.

Also, the ESA’s document provides explanations about algorithms used to generate L3A products.

Download data

Data are downloadable from : https://geodes.cnes.fr/ .

iota2’s cook

Configuration file parameter

chain.s2_l3a_path is the parameter to enable the use of L3A products.

chain:
{
    ...
    s2_l3a_path:"/absolute/path/to/Storage_directory"
    ...
}

Data storage

Data must be stored by tile as the in following tree :

└── Storage_directory
    ├── T31TCJ
    │   ├── SENTINEL2X_20181015-000000-000_L3A_T31TCJ_D_V1-1
    │   │   ├── MASKS
    │   │   │   └── *.tif
    │   │   └── *.tif
    │   └── ...
    ├── T31TDJ
    │   └── ...
    └── ...

Where tiles are T31TCJ, T31TDJ …

Data usage

Time series

All dates are chronologically stacked together in order to provide a temporal series to the learning / classification system. Then, if the configuration file parameter sensors_data_interpolation."use_gapfilling is set to True, the gapfilling is processed.

Note

Every band is resampled to a 10m resolution.

Gapfilling

Gapfilling allows iota2 to interpolate cloudy dates and resample every tile’s time series on the same temporal grid. The cloud interpolation is based on masks provided by L3A products.

Masks usage

A BINARY_MASK raster is generated for each L3A date and placed next to native data. BINARY_MASK is the raster provided to the gapfilling algorithm to know which pixels to interpolate by date. iota2 uses _FLG_R1.tif raster to determine pixels to flag as to be interpolated. Every pixel under 0 and 1 respectively NODATA and Cloud / Cloud's shadow will be interpolated.

Features

Once the gapfilling is done, if the configuration file parameter arg_train.features is different from [], then features are computed from the the gapfilled time series. NDVI, NDWI and brightness are automatically computed. Also, users could provide a set of features described as a BandMath expression using : Sentinel_2_L3A.additional_features configuration file parameter. By default, every band is used as a feature and used by the classification process to find pixels label, but with the joint use of Sentinel_2_L3A.keep_bands and iota2_feature_extraction.extract_bands only targeted bands will be used.

Available bands

All bands provided by THEIA L3A products are available.

Sentinel_2_L3A:
{
    keep_bands:["B2", "B3", "B4", "B5", "B6", "B7", "B8", "B8A", "B11", "B12"]
}