Validation of a neural network model for the separation of atmospheric effects on attenuation

In high frequency bands, between 10 and 50 GHz. atmospheric attenuation is caused by several types of atmospheric component: gases (oxygen and water vapour), clouds and rain. Each of these components behaves quite differently, when considered in terms of its temporal and spatial variability. Separation of the different atmospheric contributions (also called separation effects) is an essential step for the improvement of propagation model. Our aim in this study is to develop and valid an artificial neural network (ANN) able to separate out the contribution of different atmospheric component. A wide simulated database, corresponding to different sets of meteorological conditions is used to train the ANN. The selection of input variables among following quantities: attenuation at one, two or three frequencies, humidity, pressure, and temperature at ground level, is performed in computing their relative contribution to output. The best ANN obtained is thus validated with actual measured attenuations performed during Olympus experiment. The validation of separation effects is performed by the comparison of rain attenuation statistics.