Evolutions of Sentinel-2 Level-2A Cloud Masking Algorithm Sen2Cor Prototype First Results.

Reliable cloud screening remains a critical issue for the analysis of optical imagery. In this work we present the recent evolutions of the Cloud Screening and Scene Classification module of Sen2Cor processor for the Copernicus Sentinel-2 mission. In addition to a Level-2A surface reflectance product, Sen2Cor provides a Scene Classification (SCL) map divided into 11 classes. The information provided by this map is of great interest for automated processing chains. It can be used to mask out clouds, cloud shadow, water, snow/ice from the Sentinel-2 imagery, so that downstream processing can be performed on clear land pixels, suitable for time-series analysis and quantitative remote sensing. The performance and limitations of the current algorithm are recalled. The updates aimed at improving the overall accuracy of the cloud screening are described (efficient topographic shadows computation, mitigation of false snow detection in clouds and mitigation of false clouds detection on bright targets). Preliminary results based on a Sen2Cor prototype are discussed.