Retrieval of water vapor and temperature profiles from radio occultation measurements by combination of ANN and iterative method

摘要

A combining retrieval method of radio occultation measurements is presented. Vertical profiles of water vapor and temperature are retrieved from radio occultation bending angles, with a new method combining the artificial neural network and iterative method. We used a feedforward, full-connected network based on the back-propagation algorithm to retrieve the water vapor profiles in the troposphere. The network was trained by paired bending angle and water vapor pressure profiles from CHAMP. The month, latitude, altitude and bending angle were used as the input vectors, and the water vapor pressure as the output vector. The profile of water vapor pressure retrieved by the ANN was applied to the iterative procedure that exploits the constrains on temperature and water vapor pressure mandated by the ideal gas law and the equation of hydrostatic equilibrium. The vertical distribution of temperature was calculated.