A regression model for forecasting microwave radio fading at Palmetto, GA

Abstract

Refractive fading degrades telecommunications transmissions on microwave radio systems. Since fading occurrence is highly variable from one night to the next, forecasts of the probability of fading occurring at a given location on the upcoming night would be useful for reducing the adverse impact of fading on transmission performance. In this work logistic regression analysis was used to develop an empirical formula for forecasting refractive fading probability from meteorological parameters forecast by the National Weather Service's numerical forecast models. The formula was developed using 40 months of daily meteorological and fading measurements collected near Palmetto, GA. The performance of the method was evaluated using eight and one-half months of meteorological forecast data. For days when fading was forecast to be relatively unlikely (0-40 percent probability), 90 percent of those days experienced less than 3 s of fading below -20 dB. For days when fading was forecast to be relatively likely (80-100 percent probability), 75 percent of those days experienced more than 3 s of fading below - 20 dB. These results demonstrate the feasibility of forecasting microwave refractive fading a day in advance.