Performance probability distribution function for modeling visibility for free space optical link in Nigeria

This paper attempt to assess the performance of the most appropriate distribution function for modeling atmospheric visibility over five selected locations (Akure, Enugu, Ikeja, Jos and Port-Harcourt) in Nigeria. The distribution functions are tested based on 5-year (2010-2014) visibility data obtained from Nigeria Meteorological Agency (NIMET). Five distribution functions are tested to determine most suitable one based on four different metric measures. The associated parameters of the most suitable fitted distribution are estimated and the trends in the characteristic of the visibility are deduced. The result shows that lognormal distribution present the best probability distribution function for modeling visibility over the selected locations. A typical result for one of the locations (Akure) show that the lognormal distribution has Root Mean Square Error (RMSE) of 0.0766 km, Mean Absolute Error (MAE) of 0.0095 km, Mean Absolute Percentage Error (MAPE) of 17.8 % and coefficient of determination (R2) of 0.87. When compared with other distribution functions, the same trend could be seen in other locations although with different values of RMSE, MAE, MAPE and R2. The location and the scale parameters for the total distribution varies seasonal-wise and are location depended. The overall result will be useful for predicting future visibility over the locations. It will also be a good tool for free space optical link design in Nigeria.