European Commission logo
MenuForest Resources and Carbon Emissions (IFORCE)

Sentinel2 L1C cloud-free composites 2015-2017, 2018, 2019 and 2020

Click the map to browse the Sentinel2 composite
Mato Grosso State (Brazil)
Kilimanjaro (Tanzania)
Phnom Penh City (Cambodia)

Two Sentinel-2 L1C cloud-free composites over the tropical belt have been produced based on all Sentinel 2A and 2B imagery from October 2015 to October 2017 and entire 2018 (January-December) respectively. The spectral value of each composite pixel is calculated as the 'median' of all Sentinel-2 data (imagery available for each respective pixel) after cloud masking using:

with additional masking (applied only on 2015-2017 data) on
A RGB false color composite is created based on TOA-Reflectance of bands B11, B08 and B04 at a spatial resolution of 20m (10m bands are resampled to 20m using NN approach) and converted to 8bit (Byte) using a scale factor of 0.051 for optimal visualization.

Prototyping and processing has been done in Google Earth Engine.
Options for downloading the S2 mosaic (GeoTiff) will be available soon.

2020 improvements:

- An innovative Sentinel2 L1C cloud/shadow mask has been developed and implemented at local/regional scale, leading to a substantial improvement in sharpness and global quality. New composites over the tropical belt are under processing and will be uploaded in the next days.

Highlands of Peru' (a) and Papua (b) are some of the cloudiest place of 2018 in the tropics:
comparison of online composite (left) and with the new cloud/shadow mask (right, soon available)
- Indication for potential change between 2015-7 / 2018 composites computed 'on the fly' based on simple spectral distance of SWIR1 bands. The product may serve as quick "alert" for potential forest and land cover change, to be then confirmed by visual verification. Violet and green colors correspond to an increase (e.g. soil component) and a decrease (absorption e.g. due to vegetation growth or water) in the SWIR1 band, respectively. Due to different processing buffers, artifacts may occour along edges.

From left: Mato Groso (Brazil), Sanga (Congo), Manyoni (Tanzania), Stung Trene (Cambodia)

The Sentinel-2 cloud-free pan tropical mosaic is accessible via:

Example of WMS integration are provided at following links:

Technical References

[1] Simonetti D., Simonetti E., Szantoi Z., Lupi A., and Eva H. D., 2015
First results from the phenology-based synthesis classifier using Landsat 8 imagery
IEEE Geoscience and remote sensing letters
[2] Simonetti D., Marelli A., Rodriguez D., Veselin V., Strobl P., Burger A., Soille P., Achard F., Eva H., Stibig H.J., Beuchle R., 2017
Sentinel-2 web platform for REDD+ monitoring,online web platform for browsing and processing Sentinel-2 data for forest cover monitoring over the Tropics.
Luxembourg: Publications Office of the European Union, EUR28658; doi:10.2760/790249
[3] Simonetti D., Marelli A., Eva H.D., 2015
IMPACT: Portable GIS toolbox image processing and land cover mapping
Luxembourg: Publications Office of the European Union, EUR 27358 EN, ISBN 978-92-79-50115-9