Temperature Profiling with Neural Network Inversion of Microwave Radiometer Data

Abstract

A neural network is used to obtain vertical profiles of temperature from microwave radiometer data. The overall rms error in the retrieved profiles of a test dataset was only about 8% worse than the overall error using an optimized statistical retrieval. In certain cases, such as one with a large temperature inversion, the neural network produced better reproductions of the profiles than did the statistical inversion.