Researchers developed an adaptive ensemble learning framework combining hyperspectral and LiDAR data to identify vegetation species in karst wetlands with up to 92.77% accuracy, surpassing traditional models. The study emphasizes the significance of integrating optical and structural data for precise ecosystem mapping, showcasing the innovative AEL-Stacking model's superior performance in classifying species with overlapping spectral signatures. This research provides a scalable and explainable approach for high-resolution wetland mapping, supporting global biodiversity conservation and carbon neutrality efforts.