Artificial neural networks as rain attenuation predictors in earth-space paths

Satellite communication services have grown in recent years and the radiowave spectrum to support them is saturated. So, it is necessary to search for frequency bands higher than the presently used ones, to allocate new services. But the problem of radiowave degradation by rain is critical for communication links operating above 10 GHz, and a precise knowledge of rain attenuation is important to design reliable satellite communication links, considering that it must operate under all atmospheric conditions. Several phenomenological models have been developed to predict the rain attenuation in earth-space paths, but these models show poor accuracy for higher frequencies. In order to improve the prediction, this paper introduces a new method to evaluate the rain attenuation in satellite communication links using a specially designed neural network. The results show that this new model performs much better than the classical ones.