16 January 2026
10 May 2021
The study defines requirements, and both defines and implements (codes) algorithms for CLIM-based cloud screening to aid highly accurate retrievals of atmospheric concentrations of key greenhouse gases XCO₂, XCH₄, and NO₂. Screening algorithms detect clouds and quantify deviations from clear-sky conditions via cloud-effective obstruction—a concept from the Sentinel-3 Synergy Cloud Mask Development (S3-SYN-CM) project.
Objectives
The study aims to define and implement initial cloud masking and obstruction algorithms based on both established methods and the S3-SYN-CM framework. It will refine the algorithms and develop functional, operational-grade scientific core software for integration into the ground segment. Demonstrating the software’s functional performance precedes comprehensive scientific validation as the ultimate objective, which ensures cloud screening accuracy and reliability. These objectives collectively define the entire scope of work needed to enable cloud screening for accurate greenhouse gas concentration retrievals by the CO2M mission.
The initial algorithm definition (Phase 1) of this work was conducted as a sub-study within the Copernicus Sentinel-3 Synergy Cloud Mask Development project. The follow-up work (Phase 2) is organized as a dedicated contract.
Overview
Phase 1 overview
This phase developed the initial Algorithm Theoretical Baseline Document (ATBD) and coded an algorithm demonstrator. Both were based on the re-use of the framework of the EUMETSAT study Sentinel-3 Synergy Cloud Mask Development (S3-SYN-CM) which is based on radiative transfer physics and therefore relatively sensor-agnostic and applicable to both atmosphere and surface use cases.
The initial ATBD carefully addressed several factors. It considered the specific requirements of cloud detection needed to support the CO2 mission, ensuring alignment with mission objectives. The algorithm was also designed with consideration of the CLIM instrument's spectral and geometric characteristics, particularly the need to establish a cloud detection approach tailored to the three channels (670 nm, 752 nm, and 1370 nm) and the native spatial resolution of CLIM. Additionally, the development process incorporated user requirements regarding the format and content of the CLIM cloud mask product to ensure usability and relevance for greenhouse gas concentration retrievals facilitated by the main CO2M instruments.
Phase 1 outcomes
Several traditional cloud tests were specified based on spectral homogeneity (670 nm, 752 nm), spatial homogeneity (670 nm, 752 nm), and brightness—particularly at 1370 nm for detecting optically thin ice clouds. A feasibility study using approximate CLIM-equivalent Sentinel-3 OLCI and SLSTR spectral channels demonstrated the utility of an artificial neural network to combine these traditional tests. The CLIM obstruction algorithm provided cloud-effective obstruction values for each spectral channel and cloud optical depth (at 550 nm, facilitating direct comparison with aerosol optical depth) as a byproduct. An experimental algorithm produced a proxy for cloud top height, offering additional information potentially beneficial for greenhouse gas concentration retrievals. The ATBD also included recommendations on interpreting the algorithm outputs in the context of the wider fields of view of the CO2M main instruments. Algorithm demonstrator code was applied to proxy Sentinel-3 SLSTR data.
Phase 2 overview
Phase 2 provides Scientific Service Support for the CO2M CLIM L2 processor evolution and maintenance toward operational use. Support covers scientific development and evolution, software development and maintenance, and scientific validation.
Scientific development and evolution focuses on several key activities. The ATBD undergoes amendment to include the remapping of CLIM cloud detection products onto the CO2M spectrometer field of view, alongside ongoing updates and clarifications as needed. Continuous interaction with the CLIM Test Data Set (TDS) study supports preparation for using the test data in performance modelling. Regular collaboration with the GHG and NO2 study activities also ensures the CLIM cloud detection focus stays aligned with end-user needs, responding promptly to questions and feedback from these activities.
The development and maintenance of the CLIM L2 processor software involves defining and maintaining the interface between this software—responsible for generating the CLIM L2 cloud product—and the Instrument Processing Facility (IPF) within the ground segment. The code is developed according to requirements specified in the EUMETSAT Instrument Processor Facility Requirements Document (IPFRD) and maintained in line with the ATBD, the IPFRD and EUMETSAT’s generic coding standards for operational ground segments.
Scientific validation of the CLIM L2 cloud product includes defining methodologies, generating synthetic or proxy datasets where CLIM TDS data prove insufficient, providing software for scientific validation, and resolving scientific issues in iteration with software development and maintenance activities. Leading into the commissioning period, contributions are made to defining scientific validation tests for CLIM and SCENE-aggregated L2 cloud products, along with the software needed to execute the validation plan.
Phase 2 outcomes to date
Scientific and software development tasks have proceeded as planned. Scientific highlights include incorporation of atmosphere radiative transfer modelling into traditional cloud tests, which made the problem of tuning threshold values more tractable. The pixelwise propagation of radiometric uncertainty to uncertainty of retrieved cloud parameters (cloud optical depth, proxy cloud top height, and obstruction) and cloud mask confidence is ongoing. Development of the CLIM L2 processor software started from scratch and has progressed well. The software was developed using an object-oriented programming approach, with a modular architecture featuring operators handling input/output/breakpoint datasets and algorithms applied to images. Release 2.3 of the processor software fulfils all technical and functional requirements. Scientific verification of CLIM L2 cloud products has been performed and all issues identified were resolved. Comprehensive scientific validation using Sentinel-3 SLSTR data as proxy is pending. Examples below illustrate typical algorithm performance on central European synthetic (simulated) TDS granules.
