Comparative analysis of Biomod2 statistical and machine learning methods for Lactarius deliciosus distribution in Refahiye, Turkiye.

The spatial prediction of edible fungi is essential for the conservation and sustainable use of non-wood forest products (NWFPs) and contributes to the understanding of fungal biodiversity in forest ecosystems. This study compares multiple species distribution modeling (SDM) techniques to predict the spatial distribution of Lactarius deliciosus (L.) Gray in the Refahiye and Tekçam Forest Planning Units (FPUs) in Türkiye. Using the Biomod2 platform, we implemented five modeling algorithms: generalized linear models (GLM), multivariate adaptive regression splines (MARS), classification tree analysis (CTA), boosted regression trees (BRT), and random forests (RF). Among these, the RF model outperformed the others, demonstrating superior accuracy across all performance metrics, likely due to its ability to handle non-linear relationships, categorical predictor variables, and complex interactions without requiring extensive parameter tuning. The resulting RF-based suitability map provides valuable guidance for sustainable mushroom harvesting, forest management planning, and the conservation of mycological resources.