I. Introduction
Microwave radiometers, both from ground and satellite platforms, have been used for many years for profiling atmospheric parameters such as temperature, water vapor, and cloud liquid [1]–[6]. Unfortunately, radiometric retrievals of profiles suffer of inadequate vertical resolution and relatively poor accuracy, particularly when considering the region of sounding far from the sensor. Therefore, the results obtained from radiometric measurements by using standard retrieval algorithms, as multivariate linear regressions, do not always meet the requirements of meteorological applications. This is due to the insufficient information gathered by multifrequency or multiangle measurements and to the ill-posed nature of the inverse problem [7]. The difficulties are even increased by the unknown variations of the atmospheric conditions that occur as a function of both time and geographical location.