Hybrid biogeography-based optimization and Mamdani fuzzy modelling for physical habitat suitability modelling under limited data conditions

• A novel BBO-FIS hybrid model was developed to simulate physical habitat suitability in rivers. • The model overcomes data limitations by optimizing fuzzy rules and membership functions directly from limited field data. • BBO-FIS outperforms conventional univariate and multiple linear regression models in prediction accuracy. • The approach captures nonlinear and uncertain ecological relationships, providing more ecologically realistic habitat simulations. • The framework is applicable for river management, ecological flow assessment, and habitat conservation under data-scarce conditions. This study introduces a novel hybrid framework combining Biogeography-Based Optimization (BBO) with a Mamdani fuzzy inference system (FIS) to simulate physical habitat suitability in riverine ecosystems. The approach was developed to overcome two critical limitations in habitat modelling: the reliance on expert-defined fuzzy rules and the need for extensive datasets, both of which are often unavailable in many aquatic ecosystems. The model was tested using field data, where key physical habitat parameters—flow depth, velocity, and bed particle size—were measured alongside normalized Brown Trout population data to assess habitat suitability. Three modelling approaches were compared: a univariate model, a multiple linear regression (MLR) model, and the proposed BBO-FIS model. The univariate and MLR models failed to reliably replicate observed suitability patterns due to their inability to account for complex ecological interactions. In contrast, the BBO-FIS model generated optimized membership functions and fuzzy rules directly from limited data, significantly improving prediction accuracy. Evaluation using statistical metrics—root mean square error (RMSE) and Nash–Sutcliffe efficiency (NSE)—confirmed the superior performance of the BBO-FIS framework. By integrating fuzzy logic with evolutionary optimization, the model successfully captured nonlinear and uncertain relationships among habitat variables, offering a more ecologically realistic simulation. The results highlight the potential of the BBO-FIS framework for use in ecological flow assessment, habitat conservation, and riverine ecosystem management. This hybrid approach provides a promising solution for robust habitat modelling in data-scarce and complex aquatic ecosystems in rivers.