Development of Multiple Linear Regression Model for Prediction of Temperature Levels in Ogun Basin, Abeokuta, Nigeria

Abstract

The study centered on developing empirical equations to predict maximum and minimum temperatures, using multiple linear regressions. The technique utilizes previous data in the study area in order to effectively predict the future outcome. To facilitate the research work, 29 years information of the observed minimum and maximum temperatures of Ogun basin were obtained from the Nigeria Meteorological Agency, Federal Ministry of Aviation, Abeokuta, Nigeria. The data collected covers the periods between 1982 and 2009 and were preprocessed to extract some features or statistical indicators embedded in the time series. Features extracted are; the moving averages, exponential smoothening , oscillators and rate of change. The multiple linear regressions for each climatic element under consideration were generated using first part of the extracted features, and the remaining was used for validation of the models. R-square values obtained are 0.98 and 0.92 for maximum temperature respectively, which signifies a good correction between the actual and predicted values of the weather parameters. It was therefore concluded that applying soft computing approach for weather modeling to forecast minimum and maximum Temperature is most feasible rather than any other short term weather forecasting approach.