Assessing transferability of genetic algorithm-optimized fuzzy habitat preference models for Japanese medaka (Oryzias latipes)

The present study assessed the transferability of genetic algorithm-optimized fuzzy habitat preference models for Japanese medaka (
Oryzias latipes
), a freshwater fish dwelling in agricultural canals in Japan. By using three independent data sets from field surveys, three models were developed and evaluated with respect to model accuracy and transferability. As a result, all the habitat preference curves showed similar preference at deeper depth, lower velocity, larger cover and smaller vegetation. Despite this similar trend, the model accuracy and transferability differed among the models, which would be ascribed to site-specificity in physical environment and quality of data. Further studies would be necessary to reduce the effects of site-specificity and data quality for improving the present models.