Climate Signal Clustering Using Genetic Algorithm for Precipitation Forecasting: A Case Study of Southeast of Iran
Banafsheh Zahraie, Abbas Roozbahani · World Environmental and Water Resources Congress 2007 · 2007
In this paper, an innovative method for clustering of climate signals is developedusing Genetic Algorithm (GA). In this model, the relation of the signals with thevariations of another climatic variable is considered in clustering algorithm. In thecase study, the model is used for clustering the Sea Surface Temperature (SST) datain Omman Sea, Arabian Sea, and Indian Ocean considering the precipitationvariations in Sistan-Balouchestan Province in Southeast of Iran. For this purpose, theprecipitation data is classified to the three categories of below normal, normal, andabove normal and the fitness function of the GA model is formulated to minimize thevariance of the precipitation for each selected cluster. The results show that the modelcan be effectively used for prediction of low and high precipitation seasons in thestudy area using the SST variations in the defined clusters.