Forecasting Seasonal Rainfall in Zambia – An Artificial Neural Network Approach

Abstract

Weather forecasting is an ever-challenging area of investigation for scientists. It is the application of science and technology in order to predict the state of the atmosphere for a given time and location. Rainfall is one of the weather parameters whose accurate forecasting has significant implications for agriculture and water resource management. In Zambia, agriculture plays a key role in terms of employment and food security. Rainfall forecasting is one of the most complicated and demanding operational responsibilities carried out by meteorological services all over the world. Long-term rainfall prediction is even more a challenging task. It is mainly done by experts who have gained sufficient experience in the use of appropriate forecasting techniques like modelling. It is mainly done by experts who have gained sufficient experience in the use of appropriate forecasting techniques like modelling. In this paper, a rainfall forecasting model using Artificial Neural Network is proposed as a model that that can be 'trained' to mimic the knowledge of rainfall forecasting experts. This makes it possible for researchers to adapt different techniques for different stages in the forecasting process. We begin by noting the five main stages in the seasonal rainfall forecasting process. We then apply artificial neural networks at each step. Initial results show that the artificial neural networks can successfully replace the currently used processes together with the expert knowledge. We further propose the use of these neural networks for teaching such forecasting processes, as they make documentation of the forecasting process easier and hence making the educational process of teaching to forecast seasonal rainfall easier as well. Artificial Neural Networks are reliable, handle more data at one time by virtual of being computer based, are less tedious and less dependent on user experience.