Zero-Shot Mapping of War-Damaged Infrastructure Under Satellite Embargo

A recent preprint on arXiv (2608.00119) introduces a novel technique for assessing infrastructure damage in conflict areas, circumventing satellite data restrictions. This method reinterprets the mapping of affected buildings as a zero-shot geometric projection challenge, utilizing historical pre-strike maps. By harnessing coordinates and incident information from LiveUAMap and ArcGIS, Large Language Models (LLMs) determine weapon payloads (W) to calculate kinetic blast zones using Hopkinson-Cranz scaling (R_base = Z * W^(1/3)). To estimate exposed structures without post-strike images, the authors propose Adaptive Field-of-View to mitigate resolution bias in 2D segmentation (SAMGeo) and employ 2.5D pseudo-height depth maps with segmentation masks, enhancing the ability of Large Vision-Language Models (LVLMs) to differentiate overlapping rooftops. This method was tested on 2026 Middle East conflict data, showing improved accuracy with depth-augmented LVLMs. This study addresses a significant challenge in humanitarian efforts by facilitating quick damage assessments in the absence of satellite imagery. The preprint is classified as a cross-type announcement and is accessible on arXiv.