Testing Neural Network (NN) algorithms implemented in MATLAB for the study of rainfall distribution patterns at different locations in Nigeria

Abstract

In Nigeria, accurate information concerning rainfall distribution is generally difficult to obtain because it is a discontinuous quantity. Therefore, special tools and equipment are required for its recording and interpretation. This has made information dissemination regarding rainfall distribution across different locations in Nigeria difficult. The scenario has made everyone sorely dependent on the Nigeria Meteorological Agency (NiMet) as the only source of recording and interpreting rainfall distribution. Rainfall distributed at different locations in Nigeria at the same time and season for a particular year is not even. Therefore, it is uncertain for one to conclude that the rainfall distribution pattern at one location is or will be the same or different at another location at a particular time and season of the year. Therefore, the objective of this work is to test neural network (NN) algorithms implemented in MATLAB for the study of rainfall distribution patterns at different locations in Nigeria using satellite-based data obtained from The European Centre for Medium-Range Weather Forecasting (ECMWF) to have alternative methods of interpreting rainfall data in Nigeria. After data conversion, the designed neural network algorithms implemented in MATLAB were adopted to contour daily rainfall incidences in all the locations for the years of study. The result of the study showed that locations with latitude closer to 4o, received the highest amount of rainfall distributions and locations with latitude farther to 4o, received the lowest amount of rainfall distributions within the same years of the study. The results also showed that rainfall distribution patterns can be interpreted using NN implemented algorithms in MATLAB and they vary at different locations in Nigeria. Therefore, NN algorithms implemented in MATLAB may serve as another method for fast and reliable interpretation of rainfall distribution recorded data at any chosen location.