Abstract
It is known that habitat use of riverine fish depends on multiplicity among physical environments and waterquality. Habitat selections even differ between conspecific fish in rivers and hatcheries. For this reason, it is necessary to develop habitat preference models in the field studies. In this study, a modeling technique for instream fish habitat preference was proposed. The preference intensity of Japanese Medaka (Oryzias latipes), dwelling in an agricultural canal with the three environmental factors of water depth, current velocity and cover ratio, was quantified using the on-the-spot examination data. For explicitly taking essential vagueness of fish behaviour into consideration and for searching for an optimal model structure, two artificial intelligent techniques were introduced: a simplified fuzzy reasoning method and a simple genetic algorithm, respectively. Therewithal, the uncertainty contained in measurement as errors or dispersions of physical environment were positively taken into the model using symmetric triangular fuzzy numbers. The fuzzy preference intensity model was thus constructed, which resulted in a positive agreement between predicted and observed distribution of the Japanese Medaka resident in an agricultural canal.
