Reconstruction of seasonal mean clouds over the world ocean using probabilistic distribution of clouds and singular spectra

Development of the effective probability distribution functions for cloud cover data is critically important for the quantitative statistical description of cloud cover over the oceans, including the probabilities of various cloud regimes. We analyze applicability of probability distribution developed for visually observed cloud cover to satellite observations of the total cloud cover over the global oceans. Further we use parameters of probability distributions for quantifying cloud cover response to different factors. We utilized mixed Gamma distribution for approximation of the probability density of the total cloud cover. Further probability estimates derived from the theoretical distribution were used for developing predictive statistical metrics for five-year total cloud cover over the World Ocean. Global calculations were conducted for a 5° × 5° grid for the winter and summer seasons. Total cloud data were taken from the CLARA-A ed. 3 dataset retrieved from satellite measurements of AVHRR on the polar orbit satellites over the period from 1979 to 2023. The predictive distribution of total cloud cover was designed utilizing Singular Spectrum Analysis for N + 1 year with the test years ranging from 2019 to 2023. The forecasts were based on data records from 1979–2018, and each prediction for N + 1 year was next checked against the data for the respective test year. This procedure was applied for 5-year means for winter and summer seasons. Despite the use of a distribution function based on the incomplete Gamma function, prediction of the distribution of total cloud cover may have uncertainties of up to 2 octas, contingent to the dominant cloud cover regime. This discrepancy is due to poor capability of the distribution function to precisely capture abrupt changes in probability density of the cloud cover for specific cloud regimes.

total cloud cover, satellite data, shortwave fluxes