Machine Learning for Cloud Detection in IASI Measurements: A Data-Driven SVM Approach with Physical Constraints

Cloud detection is fundamental for the interpretation and operational exploitation of hyperspectral infrared sounders, yet the capability of infrared radiances alone to provide reliable cloud information remains insufficiently assessed. We introduce the Cloud Identification Support Vector Machine (CISVM), a supervised framework for global clear and cloudy classification from Infrared Atmospheric Sounding Interferometer (IASI) Level 1C observations. The analysis spans four seasons and compares radiances and brightness temperatures, alternative spectral reductions, and stratifications by surface type and climate zone. Using AVHRR-derived IASI cloud labels for training, the classifier operates exclusively on hyperspectral infrared radiances and is evaluated against independent IASI and collocated MODIS observations. The best-performing configuration, based on radiances and principal component analysis, achieves 88.52 percent of agreement with the operational IASI cloud reference. The results show that infrared radiances alone can reproduce the large-scale behaviour of an operational cloud product while providing physically interpretable insight into the influence of surface properties, seasonality, and geography on cloud detection performance. The proposed framework provides an operational baseline for future hyperspectral infrared missions, including ESA's Far-infrared Outgoing Radiation Understanding and Monitoring (FORUM).