AINFALL FORECASTING USING NEURAL NETWORKS IN THRACE (TURKEY)

Among the climatic elements rainfall data show the most temporal and spatialvariability. Rainfall prediction is the most intensely studied phenomenon,nevertheless due to its nonlinear nature it yields low predictability ratios. Artificial neural networks are increasing in importance in rainfall forecasting in recent years. In this study rainfall data are analyzed as a time series using artificial neural networks. The data set used in this study is the daily rainfall data of Edirne, Corlu,Tekirdag, Florya (Istanbul) meteorological stations during the period of 1970 -2000. The data is analyzed using an artificial neural network (ANN), trained usingfeed-forward back-propagation (FFBP) technique and the optimum network topology is determined. During the analysis, 4 years of monthly rainfall data areused for training, 4 years for testing and 3 years for running processes. Results ofdaily total values (sum of 10 days) were obtained better rather than the daily value sresults