CONFLUENCE
Cross-Scale Graph Attention Architecture for Watershed Vegetation Forecasting.
CONFLUENCE investigates whether cross-scale graph attention mechanisms improve spatial vegetation forecasting across heterogeneous watersheds and ecological drainage basins.
Accepted to ACM SIGSPATIAL GeoAI 2026 and submitted to NeurIPS TCCML Workshop.
Innovations
- Cross-Scale Graph Attention: Explicitly models hydrological flow and topological elevation relationships across hierarchical sub-catchments.
- Ecological Dynamics Forecasting: Anticipates vegetation health, NDVI progression, and canopy response under varying climate regimes.