Linking historical vegetation changes to bamboo monodominance via multi-source remote sensing

Bamboo expansion and monodominance threaten subtropical forest biodiversity, yet large-scale spatiotemporal patterns remain poorly quantified. Here, we integrate high-resolution mapping (2-m based on Gaofen-6 and Sentinel-2), Landsat time-series analysis (2000–2020) and Extreme Gradient Boosting (XGBoost)-SHapley Additive exPlanations (SHAP) machine learning to unravel bamboo dynamics in Wuyishan National Park, China. Our results reveal a bimodal distribution of bamboo coverage fraction (BCF) with either high-dominance (>75% cover) or low-dominance (10 m from bamboo forest boundaries. This study provides the first multi-sensor framework for monitoring bamboo monodominance, offering critical insights for managing invasive bamboo ecosystems under climate change.