Assessment of Chinese sturgeon habitat suitability in the Yangtze River (China): Comparison of generalized additive model, data-driven fuzzy logic model, and preference curve model

- Three habitat suitability models were built and compared from different aspects.
- GAM performed better than the other two models in accuracy.
- Reasons for the different performances were discussed.
- The habitat maps under seven scenarios were drawn.
- The suitable conditions for spawning were obtained by the three models.

SummaryTo date, a wide range of models have been applied to evaluate aquatic habitat suitability. In this study, three models, including the expert knowledge-based preference curve model (PCM), data-driven fuzzy logic model (DDFL), and generalized additive model (GAM), are used on a common data set to compare their effectiveness and accuracy. The true skill statistic (TSS) and the area under the receiver operating characteristics curve (AUC) are used to evaluate the accuracy of the three models. The results indicate that the two data-based methods (DDFL and GAM) yield better accuracy than the expert knowledge-based PCM, and the GAM yields the best accuracy. There are minor differences in the suitable ranges of the physical habitat variables obtained from the three models. The hydraulic habitat suitability index (HHSI) calculated by the PCM is the largest, followed by the DDFL and then the GAM. The results illustrate that data-based models can describe habitat suitability more objectively and accurately when there are sufficient data. When field data are lacking, combining expertise with data-based models is recommended. When field data are difficult to obtain, an expert knowledge-based model can be used as a replacement for the data-based methods.