Hybrid wavelet–artificial intelligence models in meteorological drought estimation

In this study, wavelet transform (W), which is one of the data pre-processing techniques, adaptive neural-based fuzzy inference system (ANFIS), support vector machine (SVM) and artificial neural networks (ANNs) were used to develop the drought estimation models of Çanakkale, Turkey. For these models, 3-, 6-, 9- and 12-months drought indices were calculated by standard precipitation index (SPI) and by using precipitation data of Çanakkale, Gökçeada and Bozcaada stations between 1975 and 2010 years. Firstly, ANFIS, SVM and ANNs models were developed to estimate calculated drought indices. Then SPI values of Gökçeada and Bozcaada stations were divided into sub-series by wavelet transform technique and these sub-series were used as input in W-ANFIS, W-SVM and W-ANNs models. When the developed models were compared, it was determined that the hybrid models developed by using preprocessing technique performed better. Among these models, it was observed that the W-ANFIS model gave the best results for 6-months period. Research Highlights Calculating of 3-, 6-, 9- and 12- months meteorological drought index with SPI Developing ANFIS, SVM and ANNs drought models using SPI values Decomposition of SPI values into sub-series by wavelet transform technique and developing hybrid drought models (W-ANFIS, W-SVM and W-ANNs) using subseries of SPI values Comparing ANFIS, SVM and ANNs models with hybrid models Obtaining appropriate results with hybrid models in meteorological drought estimation