Short-term prediction of rain attenuation using financial time series models

The constant demand for increased capacity of communication channels has led the SATCOM industry to develop new satellite systems operating at frequencies above 20 GHz where large bandwidths are available (EHF band – Extremely High Frequencies). Nevertheless, attenuation effects of atmospheric gases, clouds and rain can reach significant levels at these frequencies and it is no longer cost-effective to use a fixed power margin. Rain attenuation, due to scattering and absorption by water droplets, is the major limitation for satellites links in the EHF band. As rain events have a very restricted extension in time and space, various adaptive Fade Mitigation Techniques (FMT) have been developed to make communications at these frequencies possible. Whatever the FMT used, short-term prediction of rain attenuation is necessary to track the propagation channel variations and trigger power compensation only when needed. Typically, the forecast interval is 10 seconds, corresponding to the control loop reaction time. Unfortunately, rain attenuation behaves almost like a random walk (Brownian motion) and existing prediction models do hardly better than considering that the process remains constant over the forecast interval. Nevertheless, as rain attenuation time series show periods with high ‘volatility’ or strong variations, its statistics are similar to some stocks and currency exchange rates, therefore suggesting that the use of models originally developed for financial applications might be appropriate [1].