Drought forecasting using ANFIS- a case study in drought prone area of Vietnam

Abstract

Drought occurs throughout the world, affecting people more than any other major natural hazards. An important requirement for mitigating the impact of drought is an effective method of forecasting future drought events. This study investigated the applicability of Adaptive neuro-fuzzy inference system (ANFIS) for drought forecasting and quantitative value of drought indices, the Standardized Precipitation Index (SPI) and Standardized Precipitation Evapotranspiration Index (SPEI). Khanhhoa Province Vietnam with three meteorological stations was selected as the study area. The sea surface temperature anomalies (SSTA) events at NinoW and Nino4 zones were selected as input variables to forecast drought. Fifteen ANFIS forecasting models for SPI/SPEI (1, 3, 6, and 12 months) were trained and tested. The results show the performance of the ANFIS forecasting models for SPI/SPEI of all stations is equivalent and most ANFIS forecasting models for SPEI are better than SPI; the performance of the ANFIS forecasting models for SPI/SPEI-12 is better than other ANFIS models for SPI/SPEI-1 to SPI/SPEI-6; the models with high performance are M10–M13; model with the highest performance is M12 model. The results of this research showed that ANFIS forecasting models with SSTAs events as input variables can forecast longer term than SPI and precipitation as input variables. The ANFIS forecasting model with SSTA events as input variables can be successfully applied and provide high accuracy and reliability for drought forecasting.