Adaptive Neuro-Fuzzy Inference System for drought forecasting in the Cai River basin in Vietnam

Abstract

In order to achieve effective agricultural production, the impact of drought must be mitigated. An important requirement for mitigating the impact of drought is an effective method of forecasting future drought events. This paper presents the correlations between sea surface temperature anomalies (SSTA) and both the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI) at four areas monitoring El Nino-Southern Oscillation (ENSO) activities at the Cai River basin in Vietnam. The correlation analyses for selecting potential variables serves as a forecasting mechanism, and SSTAs events in NinoW and Nino4 zones are used to construct Adaptive Neuro-Fuzzy Inference System (ANFIS) forecasting models. Different ANFIS forecasting models for SPI and SPEI (1-, 3-, 6-, and 12-month) are trained and tested. The results of our research show that the best performing models are M5, M11, and M13. For drought forecasting in the short-term (1- or 3-month models), the SPI should be used, because it has a better performance than the SPEI. Drought forecasting with seasonal or long-term indexes (6- or 12-month models) should use the SPEI, because the SPEI performs better than SPI in these cases. We find that the ANFIS forecasting model (M11) for SPEI-12 is the best forecasting model. Furthermore, the ANFIS method with input variables constituting SSTA events can be successfully applied in order to establish accurate and reliable drought forecasting models.