Explainable artificial intelligence (XAI) for scaling: An application for deducing hydrologic connectivity at watershed scale

XAI can bridge AI and domain expertise: The study demonstrates that explainable AI methods aren't just tools for transparency—they can be combined with established hydrologic knowledge to deepen scientific understanding of how water moves through watersheds, showing practical value beyond model interpretation alone.

AI helps solve long-standing scientific challenges: The research addresses "scaling problems," a fundamental challenge in hydrology where understanding local processes at small scales and translating that to larger watershed-scale predictions has historically been difficult. XAI methods provided a quantitative way to identify which sub-regions play different functional roles in water movement.

Interpretability enables better aggregation of complex systems: By using XAI to classify different areas based on their hydrologic roles, researchers could effectively combine point-scale observations into watershed-scale insights about "hydrologic connectivity"—how different parts of a landscape connect and influence water flow—offering what the authors call "explicit and quantitative indicators" of these relationships.

Framework is transferable to other scientific domains: The authors suggest this XAI-based approach could be adapted beyond hydrology to help understand other geophysical processes, indicating potential broader applications for responsible AI in environmental science and natural resource management.