Author(s): Yuki Matsuzawa, Mitsuru Ohira, Shinji Fukuda
Linked Author(s): Shinji Fukuda
Keywords: Conservation, endangered freshwater fish, habitat model, microhabitat, urban stream.
Abstract: This study assessed microhabitat conditions of an endangered freshwater fish, Lefua echigonia, using field observed ecohydraulic data and a machine learning-based habitat model. We made a series of monthly field surveys in a spring-fed, small urban stream in Japan: one for understanding longitudinal distributions of the fish and the other for understanding species interactions among fish fauna in the target river. Random forests (RF) was applied as a tool to analyze the relationship between physical habitat conditions and the presence/absence of L. echigonia. As a result, 12 freshwater fish species were observed in the river, of which longitudinal distributions of these species were relatively stable across the year. This may be partially due to its nature of spring water having stable temperature regime (around 18 �C) within a year except summer when water temperature rises up to 25 �C. RF-based habitat model showed high performance for modelling longitudinal distribution of L. echigonia, with two kinds of ecological information, namely variable importance and response curves. Variable importance suggested the importance of hydraulic parameters of flow velocity and water depth, and the presence/absence of aquatic vegetation. Response curves illustrated the important instream habitat conditions such as shallow water with low flow velocity, and larger proportion of vegetation and medium- to large-sized gravels. The habitat information can be used to identify potential habitats for L. echigonia. Future works should consider seasonal dynamics of habitat conditions and their suitability to various life-stages of L. echigonia such as spawning and interactions with other competitive aquatic species
Year: 2017