Land cover and flood type govern the detection limits of satellite-based flood mapping across diverse global flood events
A new study published on arXiv explores the effectiveness of satellite-based flood mapping using geospatial foundation models. Researchers found that the accuracy of these models is significantly influenced by land cover and the type of flood event, with cropland and riverine floods showing better detection. The study also highlighted that inconsistencies between different reference products can be mistaken for model errors, and identified 23 failure modes, suggesting pipeline engineering is a more critical factor than model capacity for operational reliability. AI
IMPACT Establishes environmental detection boundaries for operational satellite flood mapping, crucial for disaster response.