Assessing Nonlinearity in Fish Habitat Preference of Japanese Medaka (Oryzias latipes) Using Genetic Algorithm ^|^ndash; Optimized Habitat Prediction Models

Abstract

The present study assessed nonlinearity in habitat preferences of Japanese medaka (Oryzias latipes) using genetic algorithm-optimized fuzzy habitat preference models incorporating the environmental factors of water depth (depth), current velocity (velocity), lateral cover ratio (cover), and percent vegetation coverage (vegetation). A linear relationship appeared between habitat preferences for cover and vegetation, which suggest the independent relationship between two factors. The habitat preference for velocity was found to have a nonlinear relationship with those for cover and vegetation, which appeared as different preference curves evaluated by the single-factor model. The habitat preference curves were affected by interactions between the environmental factors, especially those with nonlinear relationships, which resulted in different predictions of spatial distribution of the fish. Based on the present results, the use of a multiplex modeling approach, as applied in this study, would be appropriate to achieve better prediction accuracy, thus leading to more reliable habitat evaluation.

Abundance versus presence/absence data for modelling fish habitat preference with a genetic Takagi–Sugeno fuzzy system

“…An FHPM is a zero-order Takagi-Sugeno model (Takagi and Sugeno 1985) that relates habitat variables to habitat preference by considering uncertainties such as fish behaviour and measurement errors of the habitat variables. FHPMs have previously been applied to the modelling of habitat use by Japanese medaka in a laboratory experiment (Hiramatsu et al 2003) and at microhabitat scale (Fukuda and Hiramatsu 2008;Fukuda and Okushima 2008;Fukuda 2009). FHPMs have also been applied to habitat evaluation at a larger spatial scale ).…”

“…An FHPM can be viewed as a fuzzified version of a GA-optimised piecewise constant model (Fukuda 2009). Simultaneous optimisation of all model parameters enables an FHPM to evaluate habitat preference in an interpretable way despite the presence of nonlinear, complex interactions between habitat variables and habitat preference (Fukuda and Okushima 2008). The resulting HPCs obtained from an FHPM illustrate the response of a target species to a given habitat condition, which is an important information on how the model works (Elith et al 2005;Elith and Graham 2009).…”

“…Field surveys were conducted on three sunny days: 14 October and 5 and 9 November 2004 (Fukuda and Hiramatsu 2008;Fukuda and Okushima 2008;Fukuda 2009Fukuda , 2011. Two study reaches were established in the same canal: a 50-m-long study reach (1.6-2.0 m in width, 0.3% gradient) was surveyed on 14 October and 5 November, and a 30-m-long study reach (0.8-1.4 m in width, 0.3% gradient) was surveyed on 9 November.…”

“…This approach enables qualitative information consisting of linguistic terms to be used for the quantitative evaluation of target systems that are required for decision making at an ecosystem level. The expert knowledge-based models have been further improved with respect to predictive performance by incorporating optimisation approaches such as heuristic hill-climbing algorithms and genetic algorithms (GAs : Goldberg 1989; see Mouton et al 2008Mouton et al , 2009aMouton et al , b, 2011 for the application of hill-climbing algorithms; see Hiramatsu et al 2003;Van Broekhoven et al 2007;Fukuda and Hiramatsu 2008;Fukuda and Okushima 2008;Fukuda 2009;Pereira et al 2009;Fukuda et al 2011 for the application of GAs). Specifically, fuzzy systems optimised using GAs are called genetic fuzzy systems (Cordon et al 2001;Herrera 2008).…”

“…The presence-only and P/A data have been widely used for species distribution modelling because of their availability at a larger spatial scale (Elith et al 2006;Elith and Leathwick 2009). The abundance data have been used in habitat studies at a relatively smaller spatial scale (Hiramatsu et al 2003;Fukuda and Hiramatsu 2008;Fukuda and Okushima 2008;Fukuda 2009Fukuda , 2011Beck et al 2010) because they require intensive field surveys to obtain. Importantly, the abundance data can be used for the monitoring and assessment of the population dynamics of a target species (Beck et al 2010).…”