Application of adaptive neuro-fuzzy inference system for physical habitat simulation

Abstract

Physical habitat simulation is especially significant for river and water resource management. High accuracy in physical habitat simulation is particularly helpful for impact evaluation of dam construction or restoration projects on river ecology. The aim of this study is to use the adaptive neuro-fuzzy inference system (ANFIS) model to simulate spawning habitat suitability of Chinese sturgeon in the middle Yangtze River, China. The proposed habitat model based on ANFIS combines the advantages of a fuzzy inference system as well as easy calibration and optimization, and adaptability nature of an Artificial Neural Network (ANN). By using the proposed method, Chinese sturgeon spawning habitat on the downstream of the Gezhouba Dam was simulated. The result shows that the optimal instream flow range for Chinese sturgeon spawning is about 10000 m3/s ~ 15000 m3/s. Compared with the habitat models based on habitat suitability criteria (HSC) and fuzzy logic, the presented model considers the nonlinear relation between habitat suitability and physical habitat variables such as velocity and water depth, meanwhile it shows superiority in parameter calibration of membership functions.