Abstract
Two types of numerical methods are examined for reproducing the distribution pattern of plant-communities on a gravel bar from habitat attributes. The first is a model that was previously proposed by the authors on the basis of the concept of PHABSIM (Physical Habitat Simulation System) developed for estimating microhabitat availability of fishes. The second is a neural network model. These models are applied to the case of an alternate bar formed in a downstream reach of Yoshino River, Shikoku, Japan. Detailed consideration is given, from an ecological point of view, to habitat attributes to be used for judging the dominant species in small (10m by 10m) mesh areas. Eight habitat attributes are empirically chosen from eleven ideas thought out in the ecological aspects. As for the first model, it is pointed out that species-specific probability density curves of the habitat attributes may be successfully used as preference curves. The validity of both models is found to be similar.
