Using explainable artificial intelligence (XAI) as a diagnostic tool: An application for deducing hydrologic connectivity at watershed scale

2025-09-02

AI Summary

Hydrological connectivity mechanisms at the watershed scale remain poorly understood and difficult to interpret mechanistically. Method: We propose a novel analytical framework integrating physics-based hydrological modeling with eXplainable Artificial Intelligence (XAI). A high-fidelity soil moisture dataset, generated by a process-based hydrological model, trains an LSTM network; pointwise XAI techniques—such as SHAP or Gradient-weighted Class Activation Mapping—are then applied to disentangle input–output relationships, and local attributions are spatially aggregated to the sub-watershed scale. Contribution/Results: This study pioneers XAI-driven hydrological functional zoning and quantitative characterization of connectivity development. It identifies functionally heterogeneous units—including source areas, transport zones, and response zones—and reveals the spatial pattern of hydrological connectivity and its regulatory role in runoff dynamics through aggregated attribution maps. The framework establishes a new, interpretable, verifiable, and generalizable paradigm for data-driven hydrological attribution.

Technology Category

• Humans and AI: Explainable AI (XAI) for Human Understanding
• Machine Learning: Transparent, Interpretable, Explainable ML
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Explainable artificial intelligence (XAI) methods have been applied to interpret deep learning model results. However, applications that integrate XAI with established hydrologic knowledge for process understanding remain limited. Here we present a framework that apply XAI method at point-scale to provide granular interpretation and enable cross-scale aggregation of hydrologic responses. Hydrologic connectivity is used as a demonstration of the value of this approach. Soil moisture and its movement generated by physically based hydrologic model were used to train a long short-term memory (LSTM) network, whose impacts of inputs were evaluated by XAI methods. Our results suggest that XAI-based classification can effectively identify the differences in the functional roles of various sub-regions at watershed scale. The aggregated XAI results provide an explicit and quantitative indicator of hydrologic connectivity development, offering insights to streamflow variation. This framework could be used to facilitate aggregation of other hydrologic responses to advance process understandings.

Problem

• Applying XAI to interpret hydrologic connectivity at watershed scale
• Integrating XAI with hydrologic knowledge for process understanding
• Using XAI to identify sub-region functional roles in watersheds

Innovation

• XAI method applied at point-scale
• LSTM network trained with soil moisture data
• XAI-based classification identifies functional roles