Validation of the Copernicus Sentinel-2C Level-2A Products

Kurzfassung

The third satellite of the Copernicus Sentinel-2 constellation, Sentinel-2C (S2C), was launched on 5 September 2024. The satellite is currently in its commissioning phase and it is planned to accompany Sentinel-2B (S2B), which was launched in 2017. During the commissioning, a tandem phase between Sentinel-2A (S2A) and S2C is implemented in order to perform a cross calibration between the instruments, before S2C replaces S2A in early 2025. The three satellites are all equipped with an optical imaging sensor Multi-Spectral Instrument (MSI) that acquires high spatial resolution images of land, islands, inland and coastal waters from the Earth. Users from all around the world have been using these images for land cover change observation in relation to agriculture, forestry, urban, rural, and coastal areas, that is also useful for risk and disaster mapping. The Level-1C (L1C) Top-Of-Atmosphere reflectance and Level-2A (L2A) surface reflectance data are available to the public through the Copernicus Data Space Ecosystem (CDSE). Sen2cor is the processing tool implemented to remove the atmospheric effects from L1C products and produce L2A products, providing surface reflectance products together with Aerosol Optical Thickness (AOT), Total Column Water Vapour (WV) and Scene Classification (SCL) maps. The tandem and drift phases between S2A and S2C provide a crucial opportunity to assess the performance of the Level-2A products. This contribution will show the commissioning phase validation results for cloud masking, AOT, WV, and surface reflectance retrieval of the newly launched S2C imageries, with associated accuracy and uncertainty scores of AOT and WV retrieval, assessed with reference measurements from Aerosol Robotic Network (AERONET) stations. BOA surface reflectance retrievals were estimated with ground reference data provided by independent measurements such as Radiometric Calibration Network (RadCalNet) and HYPERNETS. Reference data for SCL cloud masking validation was generated by visually labelling the cloud and cloud-free pixel classes of the images from the test sites.