Quantifying vegetation species functional traits along hydrologic gradients in karst wetland based on 3D mapping with UAV hyperspectral point cloud

Karst wetlands, recognized for their unique hydrology and remarkable biodiversity, play a crucial role in global carbon sequestration and the terrestrial carbon cycle. However, understanding the relationships between hydrology and the spatial distribution, functional traits, and diversity of karst wetland vegetation is challenging. This study proposes a novel self-supervised deep learning method, the Hyperspectral Point cloud Projection model (HPProj), for creating a Hyperspectral Point Cloud (HSPC) using the asynchronously acquired Hyperspectral Image (HSI) and LiDAR Point Cloud, which further conducts a fine-grained 3D mapping of vegetation species in karst wetland. We also establish a new 3D Vegetation-parameter Quantitative Analysis Framework (3D-VQAF). This framework is the first to quantify variations in vegetation functional traits along gradients of flooding frequency and distance-to-water based on a 3D vegetation map. This study confirmed that the HPProj fusion generated high-quality HSPC with a high consistency of spectral reflectance with the original HSI (R2 = 0.99). We demonstrated that HSPC improved the accuracy of 3D vegetation mapping for 24 categories compared to traditional 2D methods and achieved an 80.47% mean F1-score. 3D-VQAF revealed the distribution of functional traits for each species along hydrologic gradients and indicated that Cladium mariscus in high-flood-frequency areas displayed elevated levels of chlorophyll, carotenoids, and nitrogen indices. Besides, its volume distributions were roughly normal and centered around 10 m offshore. We found that vegetation diversity exhibited a rapid increase followed by a gradual decrease with rising flooding frequency and distance-to-water, peaking at 6% flooding frequency and 7 m to water, which is critical for karst wetland protection and sustainable development.