Algorithm Theoretical Baseline Document for Sentinel-5 Precursor: TROPOMI CH4 Random Forest Cloud Filter

This document describes the novel approach of employing a RFC for cloud clearing in the operational TROPOMI XCH4 product. With the goal of achieving a product accuracy of <1%, robust cloud filtering is essential. Historically, operational XCH4 data processing relied on SUOMI-National Polar-orbiting Partnership (SUOMI-NPP) VIIRS cloud observations. However, with SUOMI-NPP nearing the end of its operational tenure and experiencing malfunctions in 2022 and 2023, an alternative approach was necessitated. In this investigation, we introduce a machine learning-driven cloud-clearing strategy anchored on the RFC framework. The RFC is trained on co-located TROPOMI and SUOMI-NPP VIIRS datasets to mirror VIIRS-like cloud clearing. Post-training, cloud masking necessitates solely TROPOMI data, thus ensuring operational autonomy from SUOMI-NPP.