ARTFEED — Contemporary Art Intelligence

GeoAI Explanation Alignment with Domain Knowledge in Flood Mapping

ai-technology · 2026-04-30

A new framework called ADAGE (Alignment between Domain Knowledge And GeoAI Explanation Evaluation) is introduced to systematically assess whether deep learning model explanations for satellite-based flood mapping align with established remote sensing domain knowledge. The increasing number of satellites has improved temporal resolution, making satellite-based flood mapping promising for operational monitoring. Deep learning approaches, a key GeoAI application, show improved predictive performance by learning complex patterns from large remote sensing datasets. However, opaque decision-making remains a major barrier to integrating these models into critical scientific and operational workflows. The ADAGE framework addresses this research gap by evaluating explanation alignment with domain knowledge.

Key facts

  • arXiv paper 2604.26051 introduces the ADAGE framework
  • ADAGE stands for Alignment between Domain Knowledge And GeoAI Explanation Evaluation
  • The framework assesses alignment of GeoAI explanations with remote sensing domain knowledge
  • Increasing satellite numbers improve temporal resolution for flood mapping
  • Deep learning models show improved predictive performance in flood mapping
  • Opaque decision-making is a barrier to integration into critical workflows
  • The study focuses on satellite-based flood mapping as a GeoAI application
  • The framework systematically evaluates explanation alignment

Entities

Institutions

  • arXiv

Sources