Long term rainfall forecasting by integrated artificial neural network-fuzzy logic-wavelet model in Karoon basin

Abstract

Physical, mathematical models and statistical distribution are applied to forecasting, whereas in natural<br><br>resources, it is difficult to choose models that are closed to reality. Rainfall forecasting as an important<br><br>dynamic process is ever favored by the researchers. Analyzing the behavior of these phenomena by<br><br>intelligent systems is completely better than classical methods, because of high non-linear dynamic<br><br>atmospheric phenomena. In this paper, a long term forecasting method is presented by a combination<br><br>of intelligent methods with the use of the past month rainfall in karoon basin and global meteorological<br><br>signals such as southern oscillation index (SOI), north athletics oscillation (NAO), sea level pressure<br><br>(SLP), sea surface temperature (SST) and 41 years historical data. This method is obtained by the<br><br>combination of artificial neural network, fuzzy logic and wavelet functions. In this model, several<br><br>scenarios have been examined for the karoon basin of Iran, through the signals. SST and NAO signals<br><br>show the best results, and then, the long-term forecasts are done for periods of six months, one year<br><br>and two years. The results of the integrated model showed superior results when compared to the twoyear<br><br>forecasts to predict the six-month and annual periods. As a result of the root mean squared error,<br><br>predicting the two-year and annual periods is 6.22 and 7.11, respectively. However, the predicted six<br><br>months shows 13.15.