Application of Fuzzy Rule-Based Model for Forecasting Drought

• Published in: Integrated Drought Management, Volume 2 Vol. 2; pp. 171 - 192
• Main Authors: Lohani, A.K., Jaiswal, R.K., Galkate, R.V.
• Format: Book Chapter
• Language: English
• Published: United Kingdom CRC Press 2024; Taylor & Francis Group
• Edition: 1
• Subjects: Rain Gauge Stations; Fuzzy Clustering; Membership Functions; SPI Value; Subtractive Clustering; Ann Model Result; SPI; Cluster Center; Persistence Index; Linguistic Fuzzy Model; Drought Prone Areas; Fuzzy Logic; Ann Model Validation; Fuzzy Sets; Meteorological Drought; Fuzzy Inference System; TS Fuzzy Model; Drought Forecasting; Crisp Set; Fuzzy Model; Ann Model; NS Efficiency; Agricultural Drought; Input Data Vector; Lead Time Forecast
• ISBN: 9781032231686; 1032232781; 1032231688; 9781032232782
• DOI: 10.1201/9781003276548-12

Drought is an important hydrological extreme regularly affecting human life in various areas and regions of the world. Drought conditions appear due to a lack of adequate precipitation for a prolonged time with less than average precipitation that causes an adverse impact on crops. Planning and mitigation for droughts involve proper assessment of their severity, forecasting, and pre- and post-management practices. For forecasting the advent and severity of a drought event, there is no single indicator or index that can be used. However, the Standardized Precipitation Index (SPI) is one of the important indicators for this purpose. Therefore, in this study, the SPI has been computed for different rainfall data series available from various gauging sites in the upper Mahanadi basin. Using these SPI values, SPI forecasting models have been developed for selected precipitation gauging sites. The capability of fuzzy rule-based modeling techniques for forecasting the nonlinear and dynamic time series has been demonstrated by various researchers in recent years. Therefore, the present study forecasts the drought indicator SPI at a six-month lead period using an artificial neural network (ANN) and fuzzy inference system in the upper Mahanadi basin. Further, the ANN and fuzzy logic-based drought forecasting models performance was evaluated using various performance criteria. This study indicates that the forecasting model based on the fuzzy inference system performs slightly superior to the ANN model. Drought is an important hydrological extreme regularly affecting human life in various areas and regions of the world. Drought conditions appear due to a lack of adequate precipitation for a prolonged time with less than average precipitation that causes an adverse impact on crops. Planning and mitigation for droughts involve proper assessment of their severity, forecasting, and pre-management and post-management practices. For forecasting the advent and severity of a drought event, there is no single indicator or index that can be used. This chapter forecasts the drought indicator Standardized Precipitation Index at a six-month lead period using an artificial neural network (ANN) and fuzzy inference system in the upper Mahanadi basin. Further, the ANN and fuzzy logic-based drought forecasting models performance was evaluated using various performance criteria. The chapter explores that the forecasting model based on the fuzzy inference system performs slightly superior to the ANN model.