Development of drought hybrid forecast model using optimization techniques

One of the factors caused by climate change is drought. Drought is a natural disaster that occurs when the precipitation value in a certain region is less than the average precipitation and occurs when underground and surface water resources are insufficient. Drought forecasting is important for the management and planning of water resources. Drought Indices are used to define drought. In this study, drought analyses were carried out with Standardized Precipitation Index and Standardized Precipitation Evapotranspiration Index in 3, 4, 6 and 12-month periods with 14 meteorological station data covering the period January 1989-July 2020 of Seyhan-Ceyhan Basins. Drought prediction model was created by calculating Standardized Precipitation Index values using Genetic Algorithm. Then, drought forecasts were made using Artificial Neural Network, Random Forest and Support Vector Machine methods based on Standardized Precipitation Index and Standardized Precipitation Evapotranspiration Index with different training, test and validation rates. In the last stage of the study, five different hybrid models were created with Artificial Neural Networks, Extreme Gradient Boosting, Support Vector Machine, Random Forest and Nearest Neighbour algorithms using Discrete Wavelet Transform and Kalman Smooth Filtering preprocessing techniques for drought prediction based on Standardized Precipitation Evapotranspiration Index. Support Vector Machine and Extreme Gradient Boosting hybrid models were found to perform better than the hybrid models created with other preprocessing and filtering techniques, respectively.