What question did this study set out to answer?

This study aims to synthesize global cases of land subsidence to understand its drivers and monitoring advancements.

July 5, 2026Open Access

A global systematic review of land subsidence drivers, technologies, and future directions from monitoring to intelligence

YZYuqing ZhangSun Yat-sen University YZYongzhang ZhouSun Yat-sen University LNLujia NiuSun Yat-sen University

Key Points

This study aims to synthesize global cases of land subsidence to understand its drivers and monitoring advancements.
Analyzed 483 subsidence cases from 43 countries through quantitative analysis and case studies.
Assessment included spatial distribution, driving mechanisms, and trends in monitoring technologies.
Evaluated integration of AI methods like deep learning and computer vision for enhanced monitoring.
Human activities dominate subsidence hotspots, with groundwater overexploitation at 31.18%.
Composite aquifers are most susceptible to differential subsidence due to their variability.
InSAR monitoring is primary but requires improvement through integration with hydrogeological models.

Abstract

Land subsidence is a critical geological phenomenon affecting coastal deltas, megacities, and agricultural regions worldwide. It causes ground elevation loss and infrastructure damage, exacerbates coastal flood risks when combined with sea-level rise, and threatens ecological security and sustainable development. This study synthesizes 483 subsidence cases across 43 countries from 408 publications, combining quantitative analysis with case studies to assess spatial distribution, driving mechanisms, and evolving trends in monitoring and modeling technologies. Results show that human activities dominate subsidence hotspots, primarily located in coastal plains and river deltas (e.g., Shanghai, Jakarta, and Mexico City). Groundwater overexploitation (31.18%) contributes substantially more than other factors. Climatic factors do not directly explain subsidence variations but indirectly affect the balance between groundwater extraction and recharge. Composite aquifers (72.8% of all aquifers) are most susceptible to differential subsidence due to their thickness variability and structural complexity. Regarding monitoring technologies, InSAR has become the primary method for millimeter-scale subsidence monitoring, though its accuracy requires enhancement through integration with hydrogeological models and ground measurements. From the perspective of AI-geoscience integration, this review identifies a paradigm shift from single-method monitoring toward multimodal data fusion and intelligent modeling, including deep learning for deformation forecasting (LSTM), computer vision for InSAR quality enhancement and trigger recognition, and multimodal models for heterogeneous data analysis. Developing intelligent analytical frameworks that integrate multimodal AI with mechanistic models, along with promoting multi-source data sharing and interdisciplinary collaboration, represents a key pathway for mitigating subsidence risks. These findings provide a theoretical foundation for advancing geohazard research within the emerging paradigm of AI-driven geoscience.

AI에게 질문

Bookmark

View Full Paper