논문 상세보기
Development of ANFIS-based model for prediction of droughts
Drought prediction is of significance importance for drought disaster risk management and mitigation. The regression based statistical models and physical process based models are commonly used for drought prediction. The statistical models assume stationarity of data which limits their ability to capture highly non-linear patterns of droughts. On the other hand, reliable long-range rainfall forecast is necessary for drought prediction using physical process based models. However, the long-range rainfall prediction especially in the Asian monsoon regions is quite challenging for climate models. In this study, the use of Adaptive Neuro-Fuzzy Inference System (ANFIS) was explored to develop a model for prediction of droughts over the East Asia monsoon region (20oN–50oN,103oE–149oE) by employing Standardized Precipitation Index (SPI) as a drought index. Most of the drought studies in the East Asia have been focused on basin or country scale. In this study, we identified homogeneous rainfall zones in the East Asia monsoon region using cluster analysis methods and analyzed the impact of global Sea Surface Temperature Anomalies (SSTA) on drought in each zone. The ANFIS-based model was developed and evaluated with different configurations to identify optimal model architecture and suitable predictor variables for drought prediction. The performance of the proposed model was assessed by comparison of observed and predicted values of the drought index using different statistical measures.
