CHIRPS Precipitation, Copernicus DEM, (Ag)ERA5 Meteorological Data, GEOS-5 Meteorological Data, IMERG Precipitation, Landsat satellites, MODIS sensors, MSG satellites, Sentinel-2 satellites, VIIRS sensors and WorldCover Land Cover

Sentinel-2

The Sentinel-2 satellites are together with Landsat the primary source for optical data at 100m (L2) and 20m (L3). Sentinel-2 is a European wide-swath, high-resolution, multi-spectral imaging mission. Currently, two satellites are in orbit (Sentinel-2A and Sentinel-2B), flying in the same orbit but with a 180 degrees phase difference. The MSI sensor samples 13 spectral bands, with four bands at 10m, six bands at 20m and three bands at 60m spatial resolution. Sentinel-2A has been in orbit since 23 June 2015, and Sentinel-2B launched on 7 March 2017, together providing a five-daily global coverage.

Purpose of data

Sentinel 2 satellites are used for both WaPOR L2 (resampled to 100m) and L3 (resampled to 20m) to produce the following data components:
‐ NDVI/fAPAR composites for Level 2
‐ NDVI/fAPAR composites for Level 3
‐ Surface albedo for Level 2
‐ Surface albedo for Level 3
‐ Thermal sharpening) at Level 2

Approach

Both Sentinel-2A and 2B data is used.

Level 2 thermal sharpening: Sentinel-2 L2A data (Bottom-Of-Atmosphere reflectances in cartographic geometry) at 10-20m spatial resolution is used to produce 100m resolution features (both individual bands and indices that combine 2 or more Sentinel-2 bands). These features are used in the thermal sharpening procedure to predict land surface temperature. The individual bands are masked with the standard Sentinel-2 Sen2cor mask and smoothened and interpolated by applying the Whittaker smoother (Eilers, 2003, Eilers et al., 2017) prior to calculating the indices.

Level 2 data production: Level 2 uses Sentinel-2 L2A. These images are already atmospherically corrected with the Sen2Cor atmospheric correction methodology. The data is resampled to 100m and then used for the calculation of NDVI and albedo after cloud masking, which is based on the standard Sen2cor mask and a neural network approach (KappaMask). KappaMask is a cloud detector developed by KappaZeta LTD for Sentinel-2 Level-1C and Level-2A input products. The project was funded by European Space Agency, Contract No. 4000132124/20/I-DT (Domnich et al, 2021).

Level 3 data production: The Sentinel-2 satellites are together with Landsat the primary source to create 20m optical data. L2A bottom of atmosphere reflectances data is cloud masked with Sen2cor and Kappa masking techniques.

Challenges

The operational use of Sentinel-2 at a continental scale comes with a few challenges. One is the large amount of data. A total of 4250 tiles need to be processed to cover Africa, the Near East and Sri Lanka With each tile approximately 800 MB large, and with approximately 2 images are available per dekad, a total of 6.8 TB of input data is generated per dekad. To handle these data volumes, the resampling of 10m to 100m data is done in the cloud. Several cloud-based platforms are available, such as Google Engine, Amazon AWS, and the Copernicus Data Space Ecosystem. Another challenge when producing time series is the high cloud cover that occurs over certain areas.

Alternative sources

For level 2 the Sentinel-2 satellites are the best currently available at the sub-100m resolution. No alternatives are currently foreseen. Proba-V has been discontinued. Landsat is not considered as a viable high resolution alternative, due to its low revisiting time. The other alternative would be to resample a lower resolution source to a higher resolution, for example MODIS, VIIRS or Sentinel-3. However, this would result in reduced quality of the final product.

For level 3 alternative sources are Landsat, and potentially ASTER or CBERS.