Experimental Modeling of Short-Term Effects of Rain on Satellite Link Using Machine Learning

The signal received at the ground station for a satellite link is affected by the stochastic nature of atmospheric channel. Adverse weather events such as rain attenuates the signal, increases atmospheric noise and scintillation fading leading to rapid variations in the received signal at the ground station. In order to analyze the stochastic effects of weather phenomenon on the link performance on short-term basis, both the slowly changing signal attenuation and the rapid variations caused by channel at the receiver have to be studied. In this work, we first analyze the short-term effects of rain on the statistical and spectral properties of fast varying signal component affecting the link performance. Following this, we model such parameters using several features extracted from the slowly varying signal component with machine learning (ML) algorithms. We then show an interesting result that the parameters of fast varying signal can be predicted with very high accuracy using ML models up to the following 300-s duration using the features obtained in the current time duration. The energy per symbol-to-noise power spectral density (
$E_{S}/N_{0}$
) data has been obtained at a site located in Israel with AMOS-7 satellite. The prediction of such parameters will lead to receiver design adaptive to the varying channel dynamics affecting the link performance under rainy conditions.