Rainfall Forecasting in Sudan Using Computational Intelligence

Abstract

Weather forecasting is the application of science and technology to predict the state of the atmosphere for a future time at a given location. Human kind has attempted to predict the weather since ancient times. Generating predictions of meteorological events is very complex process, because the atmosphere is unstable and the systems responsible for the events are the culmination of the instabilities and involve nonlinear interaction between different spatial scales from kilometers to hundreds of kilometers. The chaotic nature of the atmosphere limits the validity of deterministic forecasts, but the increasing economic cost of adverse weather events provides a strong reason to generate more accurate and updated weather forecasts. Weather forecasting (particularly rainfall prediction) is one of the most imperatives, important and demanding operational tasks and challenge made by meteorological services around the world. It is a complicated procedure that includes numerous specialized fields of knowledge. The task is complicated because in the field of meteorology all decisions are to be taken with a degree of uncertainty, because the chaotic nature of the atmosphere limits the validity of deterministic forecasts. Long term Rainfall prediction is very important for countries whose economy depends mainly on agriculture, like many of the third World countries. It is widely used in the energy industry and for efficient resource planning and management including famine and disease control, rainwater catchment and ground water management. This thesis studies long term rainfall prediction in Sudan using computational intelligence. Monthly meteorological data obtained from Central Bureau of Statistics, Sudan from 2000 to 2012, for 24 meteorological stations distributed among the country has been used. The relationship of rainfall in Sudan with some important parameters is investigated and determined the most influencing variables on rainfall among the available ones. The performance of base and Meta algorithms to deal with rainfall prediction problem is explored and, compared. A novel method to develop long-term rainfall prediction model by using ensemble technique is proposed. The new novel ensemble model is constructed based of Meta classifier Vote combined with three base classifiers. Several neuro-fuzzy Models using different types of membership functions, different optimization methods and different dataset ratios for training and testing are built. The proposed models are evaluated and compared by using correlation coefficient, mean absolute error and root mean-squared error as performance metrics. The empirical results illustrate that the ANFIS neuro-fuzzy system and the ensemble Vote+3 models are able to capture the dynamic behavior of the rainfall data and they produced satisfactory results, so they may be very useful in long-term rainfall prediction. Spatial analysis of rainfall in Sudan is conducted for the interval 2000-2012 on three levels (towns, states and regions) and rainfall maps are obtained.