Characterization of Antarctic clouds using infrared spectral radiance data

Clouds play a crucial role in the Earth's radiation budget, making their detection and characterization essential for meteorological research, climate modeling, and weather forecasting. In polar regions, their importance is even more critical due to their role in shaping the global climate. This study aims to characterize cloud occurrence and properties using ground-based spectral measurements collected by the REFIR-PAD spectro-radiometer at the Concordia station (Dome C) on the Antarctic Plateau from 2012 to 2020. Different sky conditions are classified using the CIC algorithm, adopted as the official software in the ESA FORUM End2End simulator. The algorithm is trained and tested on a subset of the dataset, and results are compared with those from a traditional threshold algorithm based on three brightness temperature differences. The CIC algorithm outperforms the traditional method in identifying optically thin ice clouds. The CIC classification results are analyzed to derive cloud occurrence statistics and examine correlations with surface atmospheric parameters. The classification shows clear skies occur 67.71% of the time, ice clouds 29.63%, and mixed-phase clouds 2.63%. Ice clouds and mixed-phase clouds are linked to higher surface temperatures than clear skies. A temporal analysis reveals a pronounced annual cycle in cloud occurrence. Using spectra classified as cloudy, the radiative and microphysical properties of Antarctic clouds are derived with the SACR inversion algorithm. Mean optical and radiative properties for different cloud types are computed: ice clouds have a mean optical depth of 0.577, and an effective ice crystal diameter of 25.33 μm. Mixed-phase clouds have a higher optical depth of 2.35, and a smaller water droplet diameter of 8.55 μm. A parameterization of clouds' microphysical properties in terms of layer's temperature and ice mass concentration is provided for potential application in climate and numerical weather prediction models.