Evaluating the Alignment Between GeoAI Explanations and Domain Knowledge in Satellite-Based Flood Mapping

A new arxiv preprint introduces ADAGE, a framework for evaluating whether XAI explanations from deep learning flood-mapping models align with established remote sensing domain knowledge. The work targets GeoAI practitioners deploying satellite-based flood detection in operational workflows where opaque decisions are a liability. ADAGE provides systematic alignment scoring between model saliency outputs and known spectral-spatial flood signatures, directly addressing the trust gap blocking operational adoption. Remote sensing engineers and hydrology teams relying on SAR or multispectral flood classifiers should prioritize this as a validation layer before deployment.