International Journal of Electrical and Electronic Science 2019; 6(1): 1-7 http://www.aascit.org/journal/ijees ISSN: 2375-2998 Rain Attenuation Prediction in Nigeria Using Artificial Neural Network (ANN) Ibukun Daniel Olatunde, Kazeem Oladele Babatunde, David Oluwarotimi Afolabi Department of Physics with Electronics, Oduduwa University Ipetumodu, Ile Ife, Osun State, Nigeria Email address Citation Ibukun Daniel Olatunde, Kazeem Oladele Babatunde, David Oluwarotimi Afolabi. Rain Attenuation Prediction in Nigeria Using Artificial Neural Network (ANN). International Journal of Electrical and Electronic Science. Vol. 6, No. 1, 2019, pp. 1-7. Received: January 12, 2019; Accepted: March 26, 2019; Published: April 9, 2019 Abstract: The prediction of rain rate and rain attenuation plays an essential role in the fields of communications, agriculture, military services, etc. This work presents rain attenuation prediction in Nigeria using Artificial Neural Network (ANN). Rainfall data of ten years were collected from measurements made in six different geographical locations. The locations include Enugu (east), Ikeja (south-west), Kano (north-west), Lokoja (north-central), Maiduguri (north-east) and Port- Harcourt (south-south). These locations represent all geographical areas in Nigeria. ANN was trained to predict rain attenuation in these locations using the annual rainfall data given from 2007 to 2016. Conversely, the ANN was trained with sets of data from the year 2007 to 2013, thus, the result of the training was used to predict rain attenuation from the year 2014 to 2016. The rain attenuation results given by ANN were compared to the results given by the International Telecommunication Union (ITU) model which is a well-established model. The results in terms of the mean squared error (MSE) performance show that ANN predicted attenuation agrees closely with the ITU model prediction. Conversely, the resulted ANN training is a useful tool for communication engineers and expatriates to predict rainfall attenuation of subsequent years and to proactively solve the inherent signal attenuation problem facing satellite-to-earth path operation above 10GHz. Keywords: Rain Attenuation, Rain Rate, ANN, ITU Model, Mean Squared Error 1. Introduction Rain is an essential substance to human life. Most of the fresh water on earth is deposited by rain. Rain attenuation affects the design of satellite-to-earth path that operates at frequencies above 10GHz [1, 2]. Raindrops absorb and scatter radio waves, thus resulting in signal attenuation, consequently, ensures the reduction in the reliability and availability of the system. The attenuation caused by rain increases with increased frequency [3]. The extent at which rain attenuates electromagnetic signals varies with frequency and climate [4]. It is important to make an accurate prediction of rain attenuation for effective planning of microwave satellite and terrestrial line-of-sight links [5]. The major problem faced by microwave engineers working on higher frequency bands is balancing the compromise between availability of bandwidth and issues of rain-induced attenuation. Although the International Telecommunication Union radio (ITU-R) has created a meteorological approach that predicts rain attenuation, this approach does not work well in some regions like the tropical climates as it is based on data collected from temperate regions [6]. Moreover, atmospheric processes from which rainfall is formed are complex and cannot be accurately predicted using mathematical or statistical models [7]. There are several procedures for rain attenuation prediction on earth-space links, thus, these procedures are grouped into two classes: Empirical and Physical. Empirical classes depend on the measurement of databases from stations in different locations within a particular region, and physical classes attempt to reproduce the physical behavior involved in the process of attenuation [8]. The frequently used method is the empirical method because it makes use of equations and certain variables including rain rate, rain height, latitude and longitude of the earth-station [5]. Although there are some other factors affecting rain attenuation such as size distribution of raindrop [9], it is very difficult to provide the neural network with proper information about the size distribution, conversely, it is not included as the input of the neural network. An empirical method provides an appropriate distribution of rainfall rate at one minute integration time