Cross-Mission Validation of Cloud Masking for Optical Remote Sensing Missions Using Sentinel-5P Cloud Products

Kurzfassung

The validation of cloud masking algorithms for optical Earth observation imagery requires a precise reference, often referred to as ground truth. Traditionally, most validation exercises rely on manual annotations, which demand significant expertise and are prone to misinterpretation of class definitions. In addition, in-situ measurements are limited in spatial and temporal coverage, making global-scale validation both challenging and expensive. This work proposes a cross-mission validation methodology that leverages globally available, physically based reference products to assess the quality of masking outputs from EnMAP and Sentinel-2. Specifically, we validate the cloud-related classes of EnMAP and cloud masks in the Sentinel-2 Scene Classification Layer (SCL) against cloud fraction products retrieved from the TROPOspheric Monitoring Instrument (TROPOMI) onboard the Sentinel-5P mission, which is designed to provide high-sensitivity atmospheric measurements. To enable direct comparison, the Sentinel-2 SCL-derived and EnMAP cloud mask are re-gridded to the spatial resolution of TROPOMI (5.5 km × 3.5 km), aligned with pixel geolocation, and expressed as fractional cloud cover. Only matching overpasses with short acquisition time differences are selected, minimizing discrepancies between observations. This approach enables scalable validation of Level-2 EnMAP and Sentinel-2 products using external data and helps identify systematic differences. The methodology is extendable to other SCL classes-such as snow, water, or vegetation-by incorporating appropriate reference data from other specialized missions. The results contribute to ongoing efforts to improve data quality assurance and support new methodologies for the validation of EnMAP and Sentinel-2 products.