Ai-driven disaster management: a comprehensive review of deep learning for social media and geospatial data integration

Natural disasters devastate communities, causing casualties and significant economic losses. Although considerable research has been conducted using social media data as well as satellite and UAV imagery, much of this work has primarily focused on image processing and classification of disaster related content. This study provides a comprehensive analysis of existing disaster datasets, offering researchers a rich archive for future work. We examine a wide range of deep learning models employed in disaster data analysis, conduct a comparative review of their applications, and trace the evolution of research trends in the field. Despite substantial methodological advances, the majority of existing implementations remain challenging to deploy in real-world disaster response settings due to constraints related to timeliness, scalability, and operational integration. In response to these challenges, recent research has increasingly shifted toward object detection and real time human monitoring tasks, which offer more actionable and deployment oriented capabilities in disaster prone areas. The study further integrates research from two traditionally separate domains being social media data and geospatial data (satellite, remote sensing and UAV imagery), under a unified AI driven disaster data perspective. We identify four categories of limitations spanning both domains and propose a conceptual framework that consolidates their processing pipelines into a single, cohesive generative ecosystem. By promoting the concurrent use of inter domain disaster data, the proposed framework aims to enable AI systems to develop a more comprehensive, human like understanding of disaster scenarios, thereby facilitating faster and more effective solutions to real world disaster management problems.