TIME SERIES PREDICTION OF WEATHER VARIABLES AND RAINFALL FORECASTING USING THE FUZZY LOGIC BASED RAIN FORECAST MODEL.

M. B. Mu’azu and B. G. Bajoga

The Neuro-Fuzzy System developed using the Soft Computing technique of Neuro-Fuzzy is implemented as a pure Fuzzy Logic System as the Rain Forecast Model (RFM). The Rain Forecast Model (RFM) used as its inputs the weather variables: Relative Humidity, Wind Direction, and Wind Speed (from 1993 to 2002) and in order to forecast into the future (2003 to 2010), these weather variables were determined using a hybrid statistical technique comprising of Moving Averages and Exponential Smoothing. Two statistical tests (Root Mean Squared Error (RMSE) and Correlation Factor (R)) were carried out in order to measure the performance of the developed model with the validation data and with the forecasted weather variables. The developed model, using the validation data of 2003 to 2005, produced the following performance results: RMSE=28.02 and R=0.97. This formed the basis for comparing the performance of the developed model when subjected to the forecasted weather variables (2003 to 2010). The performance function results obtained were: RMSE=54.99 and R=0.88. Considering the nature of the problem, that is the unpredictability of rainfall, these are reasonable results.

Key words: Soft Computing, Rain Forecast Model, Rainfall, Wind speed, Wind direction, Relative Humidity.