A Heterogeneous Graph Neural Network Framework for Landslide Susceptibility Mapping Using Slope Units with Spatial and Hydrological Edge Modalities
A deep learning pipeline that uses Slope Units instead of grid cells, with heterogeneous graph construction (spatial adjacency + hydrological flow) via PyTorch Geometric for landslide susceptibility mapping in Cebu City, Philippines.