Development of Rainfall–Runoff Models Using Mamdani-Type Fuzzy Inference Systems

This study explores the application of Mamdani-type fuzzy inference systems (FIS) to the development of rainfall–runoff models operating on a daily basis. The model proposed uses a Rainfall Index, obtained from the weighted sum of the most recently observed rainfall values, as input information. The model output is the daily discharge amount at the catchment outlet. The membership function parameters are calibrated using a two-stage constrained optimization procedure, involving the use of a global and a local search method. The study area is the Shiquan-3 catchment in China, which has an area of 3092 km2 and is located in a typical monsoon-influenced climate region. The performance of the fuzzy model is assessed through the mean squared error and the coefficient of efficiency R2 performance indexes. The results of the fuzzy model are compared with three other rainfall–runoff models which use the same input information as the fuzzy model. Overall, the results of this study indicate that Mamdani-type FIS are a suitable alternative for modelling the rainfall–runoff relationship.