Translator Disclaimer
Paper
1 August 1992 Neural network for retrieval of indirectly measured information
Author Affiliations +
Abstract
Microwave radiation emitted by the atmosphere contains information about the composition and temperature structure of the atmosphere. Given the vertical temperate profile and humidity structure of the atmosphere, it is possible to compute the microwave radiance received by an upward-looking surface-based radiometer operating in the 20 - 60 GHz range. The inverse problem is ill-posed and requires additional information for its solution. We have found a method for solving this problem using the techniques of neural networks. In our calculation, the network has as inputs the set of microwave brightness temperature measurements made at the surface, surface temperature, and surface pressure, or surface altitude, or both. The output is the atmospheric temperature as a function of height. The neural network computes the outputs from the inputs using its internal weights which have been established by a training process. The training consists of presenting the system with input brightness temperatures calculated from radiosonde observations, and comparison of the resulting outputs with the radiosonde temperature profiles using a backpropagation training rule. The trained system was then tested on a separate set of brightness temperatures. Similar techniques should work with many types of inverse problems.
© (1992) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Edward M. Measure, Young P. Yee, Teddy L. Barber, Wendell R. Watkins, and Dick R. Larson "Neural network for retrieval of indirectly measured information", Proc. SPIE 1688, Atmospheric Propagation and Remote Sensing, (1 August 1992); https://doi.org/10.1117/12.137903
PROCEEDINGS
5 PAGES


SHARE
Advertisement
Advertisement
Back to Top