Deep learning and geospatial approaches for post-flood disaster damage assessment: a systematic review

Abstract

Abstract Extreme precipitation and flooding have demonstrated nature’s overwhelming force against human resilience, triggering geohazards such as landslides. These geohazards frequently cause infrastructure damage, resulting in power outages and loss of communications. This situation highlights the critical need for geospatial technologies to bridge data gaps in assessement of disaster-related damage. However, current literature typically examines landslides, infrastructure damage, and power outages as separate incidents rather than as cascading failures. To address this research gap, our study aimed to synthesize the existing body of knowledge on multi-hazard interactions using a systematic review approach. Adhering to PRISMA guidelines, the study analyzed 162 articles from 2010 to June 2025, sourced from Google Scholar, Springer, and Web of Science. Our findings indicate a growing trend toward the application of machine learning and deep learning in geospatial approaches for damage assessment. Notably, significant scientific contributions in this field predominantly originate from Asia and North America, revealing a geographic gap in research from Africa. In addition, the study discusses the challenges associated with interferometry methods for landslide analysis, classification techniques for building damage assessment, and the use of nighttime light data concerning damage-related power outages. Our findings reveal a significant gap in the absence of an integrated multi-hazard assessment. To bridge this gap, we propose a novel framework for the comprehensive evaluation of the interconnectedness between extreme precipitation, landslides, and infrastructure-induced power outages. This framework provides a roadmap for more resilient disaster management strategies.