Neural Networks for tropospheric profiling from GPS-LEO radio occultation

In this work a method based on neural networks is proposed to retrieve profiles of refractivity, temperature, pressure and humidity in the troposphere from GPS-LEO radio occultation. To overcome the constraint of temperature profile availability at each GPS occultation, we have trained a neural network with refractivity profiles as input computed from the geometrical occultation parameters of the CHAMP LEO satellite, while the outputs are the dry and wet refractivity profiles and the dry pressure profiles obtained from the contemporary ECMWF data.