A Note on Joint Satellite Measurements of Cloudy Atmosphere in Microwave and Infrared Ranges

One can count on significant increase in efficiency of algorithms used for solving the inverse problem of atmospheric total water vapor (TWV) and liquid water content (LWC) microwave radiometric retrieval through introducing into the related computational models such information as cloud top height, ground-level and cloud top temperatures, binary cloud mask or also cloud phase distributions obtained in the radiometer antenna’s field-of-view. This supplementary information, in turn, may be separately pre-reconstructed by means of data on infrared (IR) observations carried out simultaneously and in accordance with microwave ones. But it also makes sense to consider the possibilities and advantages of developing unified (single) algorithms, which involve synchronous IR and microwave data processing (using all available spectrum at once) for solving various inverse problems. We suppose such algorithms could be implemented on the basis of deep convolutional neural networks (CNNs) using data assimilation approach. Herewith, to demonstrate capacities of CNNs in combining estimates made by analyzing data of various wavelength ranges, we consider the problem of cloud phase spatial distribution IR-reconstruction in more detail, because of both near and far IR subranges are sensitive to cloud phase, but two distinct and independent reconstruction algorithms are usually utilized for them.