Genetic Algorithm Optimized Rainfall-Runoff Fuzzy Inference System for Row Crop Watersheds with Claypan Soils.

Fuzzy logic algorithm is capable of describing knowledge in a human-like manner in the form of simple rules using linguistic variables. It provides a new way of modeling uncertain or naturally fuzzy hydrological processes such as non-linear rainfall-runoff relationships. The fuzzy inference system (FIS) utilizes fuzzy membership functions (MF) and fuzzy rules (FR) for decision making. Genetic algorithm (GA) which employs a natural selection method inspired by biological evolution: selection (inheritance), crossover (recombination) and mutation, has been used for optimization of MFs and FRs. The objective of this study was to develop a FIS with GA optimization, for rainfall-runoff prediction on three adjacent row crop watersheds with claypan soils at the Greenley Memorial Research Center, Knox County, Missouri. Fuzzy toolbox of MATLAB 7.10.0 was used for FIS development. Five MFs and FRs were developed based on the measured rainfall-runoff data for a 9-year period for one watershed. Mamdani type FIS system was used for this study. FIS with GA optimized MFs and FRs was used for validation using data from the other two watersheds. The FIS system predicted daily runoff with an r² value of 0.74 during calibration, and 0.72 and 0.82 values during validation for the other two watersheds. The developed FIS system automatically  adjusted by GA to the problem specific conditions of rainfall-runoff relationships. This FIS offers a valuable tool for TMDL estimations of runoff using only runoff-rainfall relationship of a representative area of the watershed rather than requiring a large amount of detail about the watershed as is typically required for physically based models.