Drought forecasting using novel heuristic methods in a semi-arid environment

Abstract

Abstract The accuracy of four evolutionary neuro fuzzy methods, adaptive neuro-fuzzy inference system with particle swarm optimization (ANFIS-PSO), ANFIS with genetic algorithm (ANFIS-GA), ANFIS with ant colony algorithm (ANFIS-ACO) and ANFIS with butterfly optimization algorithm (ANFIS-BOA), is investigated and compared with classical ANFIS method in forecasting various time scales of standard precipitation index (SPI). Monthly precipitation data of Abbasabad, Biarjmand and Ebrahim-Abad stations, Iran are used in the case study. The comparison is made according to the three indexes, root mean square error (RMSE), mean absolute error and index of agreement. It is observed that the evolutionary neuro fuzzy methods perform superior to the classical ANFIS in forecasting all SPI indexes (SPI-3, SPI-6, SPI-9 and SPI12) in all three stations. The RMSE of the classic method is increased roughly by 11.4–16.7%, 11.3–32.4%, 9.8–34.4% 30.6–46.7% for the SPI-3, SPI-6, SPI-9 and SPI-12, respectively. In Ebrahim-Abad Station, the best accuracy is observed from the ANFIS-PSO method for all drought indexes while there is not a dominant method in Abbasabad and Biarjmand stations.