Do absence data matter when modelling fish habitat preference using a genetic Takagi-Sugeno fuzzy model?

• Published in: International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems
• Main Authors: Fukuda, Shinji, De Baets, Bernard
• Format: Article in Journal/Newspaper
• Language: English
• Published: 2012
• Subjects: Biology and Life Sciences; preference modelling; Species distribution model; data characteristics; predictive performance; transferability; genetic fuzzy systems; SPECIES DISTRIBUTION MODELS; MEDAKA ORYZIAS-LATIPES; IMBALANCED DATA-SETS; CLASSIFICATION SYSTEMS; SALMO-SALAR; PREDICTION; CRITERIA; OPTIMIZATION; Salmo salar
• Online Access: https://biblio.ugent.be/publication/2998065; http://hdl.handle.net/1854/LU-2998065; https://doi.org/10.1142/S0218488512400223; https://biblio.ugent.be/publication/2998065/file/2998072

Information on species distributions is of key importance when designing management plans for a target species or ecosystem. This paper illustrates the effects of absence data on fish habitat prediction and habitat preference evaluation using a genetic Takagi-Sugeno fuzzy model. Three independent data sets were prepared from a series of fish habitat surveys conducted in an agricultural canal in Japan. To quantify the effects of absence data, two kinds of abundance data (entire data and presence data) were used for developing a fuzzy habitat preference model (FHPM). As a result, habitat preference curves (HPCs) obtained from presence data resulted in similar HPCs between the three data sets, while those obtained from entire data slightly differed according to the data sets. The higher generalization ability of the FHPMs obtained from presence data supports the usefulness of presence data for better extracting the habitat preference information of a target species from field observation data.