Rain Attenuation Prediction Model for Terrestrial Links Using Gaussian Process Regression

Abstract

Rainfall is considered as one of the most crucial atmospheric elements that cause attenuation in signal propagation, especially in high-frequency range for fifth-generation (5G) and beyond wireless networks. To unveil the complex relationships between rain attenuation and other factors including rainfall rate, we propose a new rain attenuation prediction model for terrestrial line-of-sight (LoS) propagation using Gaussian Process Regression (GPR). In the proposed model the Recommendation ITU-R P.530 (called the ITU-R model) is used as the mean function in GPR, and to capture the deviation of measured rain attenuation from the ITU-R model we develop a latent function in GPR motivated from the ITU-R model. We validate with the ITU-R study group 3 databank (DBSG3) that the proposed model provides high prediction accuracy.