Fish Habitat Assessment Based on Community Analysis and Habitat Models

Abstract

Fish habitat assessments in an agricultural irrigation system was performed based on community analysis using modified TWINSPAN and species-specific habitat models using random forests. The results were compared with respect to the consistency in habitat conditions suitable for a community and species. As a result, 6 fish communities and 10 indicator species were identified from the community analysis. Cluster-specific or species-specific habitat suitability was found to be similar among the fish species within the same cluster. Thus, it was revealed that community analysis together with physical habitat surveys is capable of assessing species-specific habitat suitability. Habitat models were more useful when assessing habitat suitability of generalist species when habitat assessment results of the two methods were different. This study establishes the effectiveness of community analysis as a basic tool for species-based habitat assessment. Furthermore, machine learning-based habitat models for a target species can inform studies involving detailed ecological characteristics and predicting species distribution patterns. Further work is needed to develop frameworks for environmental assessment which balance the productivity of an irrigation system and management actions for the conservation of aquatic ecosystems.