ABSTRACT
Rainfall forecasting is an important subject to consider to prevent flood, drought and other extreme hydrological conditions that could adversely affect both man and its environment. Accurate forecasting of rainfall events is an important issue in hydrological research.
The focus of the research was to investigate the capability of linear and non-linear regression techniques for short term rainfall prediction. Multiple linear regression method was employed linear regression technique. One of the non-linear regression techniques being widely used in time series prediction is Artificial Neural Networks (ANN) approach which has the ability of finding relationship in non-linear data. ANN has proven successful in modelling time dependent system. ANN was employed as a non-linear regression technique. This study was restricted to Nsukka, Enugu State, Nigeria. Daily rainfall data, wind speed, wind direction, temperature, and relative humidity for the period of 2009 to 2012 spanning to about four (4) years was collected processed and used for the analysis. Data Analysis tools, name; Python and Excel were employed to carry out the analysis. The p-value obtained based on Shapiro and Kolmogorov-Smirnov normality reveals that the data used for this were not normally distributed.
On the performance of Multi linear regression and Artificial neural network it was observed that ANN performed better than multi-linear regressions. This conclusion was based on the coefficient of determination (R2) for which ANN had 0.996 and MLR 0.333. The performance ANN compared to MLR is based on the non-linear relationship that exists between rainfall and other climatic variables such as temperature, relative humidity, wind direction and wind speed.
