Since we started to publish Sentinel-3 Level 2 world maps via our EO browser we took the challenge to process a land worldwide cloud free image of the Earth with an RGB natural colour composition. Indeed with a proper tailoring of our processors in order to handle the instruments radiances and cloud masks, using several months of OLCI acquisitions we managed to generate a few cloud free paintings of the World.
The image above has been processed considering a three months dataset, namely all the Sentinel-3A OLCI acquisition starting from July 1st 2018 to September 30th 2018. Both level 1 and level 2 products with reduced resolution (1.2 km) have been considered and merged in a single geotiff mosaic scene, then a post processing (stretch, contrast, brightenss) has been performed with SNAP tool. In the image above there are also some parts of Antarctica, indeed in September due to the approaching of the end of our summer (ed. northern hemisphere) the instrument duty cycle shifts southwards and more objects are visible.
We cannot figure out how to mitigate the “blue halo” in the southern part of South America, then there are some missing data in the central part of Africa (this is one of the most cloudy regions in the World!) and Indochina, anyway the final result is really fine, with a nice representation of all the land colors (forest, desert, snow).
In order to fill the no data pixel we considered afterwards a different and wider input dataset (4 months), namely all the data acquired between the 1st of May and the end of August 2018. Here below the generated world map.
The central part of Africa is more complete, even if some no data are still standing. This statement is applicable also for Indochina, where the snowed Himalaya is perfectly represented too. Higher latitudes are characterized by an higher number of snow/ice pixels and even if in Eurasia landmass the result is quite good we can see a not perfect visualization in North America, namely in the boundaries of the Hudson Bay. This is probably due to the nominal difficulties of low resolution sensors to discriminate cloudy pixels from snow/ice and/or the low number of good data, but this is just our personal understanding of the processing.
Some technical details: in both cases the worldmap has been processed calculating the reflectances, then the two uploaded images needed to be resized and compressed (wordpress restrictions) so the original resolution (input product RR having 1.2 km of res.) is lost. Using a larger stretch on the blue band the “blue halo” at extreme latitudes decreases, unfortunately impacting on the other colors thus returning an unnatural RGB composition.
Both world maps can be however represented together with the Oceans, masked out in a first place. Indeed in the Oceans we have a tangible “stripe effect” and a not homogeneous color distribution due probably to the sunglint effect too. Here below the two mosaics pictured with the Oceans: