Abstract
Abstract. Spatiotemporal statistical learning has received increased attention
in the past decade, due to spatially and temporally indexed data
proliferation, especially data collected from satellite remote
sensing. In the meantime, observational studies of clouds are recognized as an important step toward improving cloud representation in weather and climate
models. Since 2006, the satellite CloudSat of NASA is carrying a 94 GHz
cloud-profiling radar and is able to retrieve, from radar
reflectivity, microphysical parameter distribution such as water or
ice content. The collected data are piled up with the successive
satellite orbits of nearly 2 h, leading to a large compressed
database of 2 Tb (http://cloudsat.atmos.colostate.edu/, last access: 8 June 2022). These observations offer the opportunity to extend the cloud microphysical properties beyond the actual measurement locations using an
interpolation and prediction algorithm. To do so, we introduce a statistical estimator based on the spatiotemporal covariance and mean of the observations known as kriging. An adequate parametric model for the covariance and the mean is chosen from an exploratory data analysis. Beforehand, it is necessary to estimate the parameters of this spatiotemporal model; this is performed in a Bayesian setting. The approach is then applied to a subset of the CloudSat dataset.