Abstract
The distribution of Itea yunnanensis, a shrub species in the genus Itea of the family Iteaceae, is primarily concentrated in the Hengduan Mountains region of China, where it faces severe threats from global climate change. However, systematic research on the species' distribution patterns, climatic response mechanisms, and future suitable habitat dynamics remains insufficient. This study aims to assess the spatiotemporal evolution and driving mechanisms of I. yunnanensis-suitable habitats under current and future climate change scenarios to reveal the migration patterns of its distribution centroid and ecological thresholds, and to enhance the reliability and interpretability of predictions through model optimization. For MaxEnt modeling, we utilized 142 georeferenced occurrence records of I. yunnanensis alongside environmental data under current conditions and three future Shared Socioeconomic Pathways (SSPs: SSP1-2.6, SSP2-4.5, SSP5-8.5). Model parameter optimization (Regularization Multiplier, Feature Combination) was performed using the R (v4.2.1) package 'ENMeval'. The optimized model (RM = 3.0, FC = QHPT) significantly reduced overfitting risk (ΔAICc = 0) and achieved high prediction accuracy (AUC = 0.968). Under current climate conditions, the total area of potential high-suitability habitats for I. yunnanensis is approximately 94.88 × 10(4) km(2), accounting for 9.88% of China's land area, with core areas located around the Hengduan Mountains. Under future climate change, the suitable habitats show significant divergence, area fluctuation and contraction under the SSP1-2.6 scenario, and continuous expansion under the SSP5-8.5 scenario. Meanwhile, the species' distribution centroid exhibits an overall trend of northwestward migration. This study not only provides key spatial decision-making support for the in situ and ex situ conservation of I. yunnanensis, but also offers an important methodological reference for the adaptive research on other ecologically vulnerable species facing climate change through its optimized modeling framework.