Exploring the optimal fuzzy rule-based modeling procedure to assess habitat suitability of indicator Collembola species in forest soils

In the face of escalating anthropogenic fragmentation and habitat destruction, research on soil habitat disturbance using indicator species is increasingly critical to conserve and maintain the ecological functions of forest ecosystems. The modeling methodology for habitat suitability is a valuable tool for assessing habitat conditions based on the ecological preferences of indicator species; however, its application to such species in forest soils remains unexplored. Therefore, this study aimed to fill this gap by identifying an optimal procedure for developing a fuzzy model to evaluate the habitat suitability of indicator species based on their abundance classes. Fuzzy models were developed for assessing the habitat suitability of Folsomia quadrioculata and F. octoculata based on data collected from seven mountains using three types of selected variable numbers (3-, 4-, and 5-variable) for two input variable selection methods (statistics-based variable selection, SVS; knowledge-based variable selection, KVS), and their performance was compared. Our results indicate that the SVS-fuzzy model performed better than the KVS-fuzzy model in both the model training and testing phases. As the number of input variables increased, the performance of the KVS-fuzzy model improved; however, it still exhibited lower performance compared to the SVS-fuzzy model. Meanwhile, the optimal SVS-fuzzy model effectively explained the abundance classes of the two collembolan species based on the environmental conditions of their habitats (F1 score > 0.743, Matthews correlation coefficient > 0.520). The findings of this study provide a solid foundation for developing effective models to understand the habitat suitability of soil indicator species. Expanding the application of fuzzy modeling to diverse species in forest soils will improve our understanding of habitat disturbance and degradation, contributing to the development of conservation strategies for forest ecosystems.