An integrated framework for monitoring and predicting coastal land subsidence: a case study of Xiamen, China

• Integrated coastal subsidence framework: monitoring, risk assessment and prediction. • Novel indicators are designed to assess coastal subsidence risk. • Fusion model outperforms three base models in coastal subsidence prediction. Land subsidence in coastal areas poses significant threats, including geological disasters, seawater intrusion, and degradation of coastal wetlands. This paper presents a comprehensive research framework for monitoring, assessing, and predicting land subsidence in coastal areas. Utilizing Sentinel-1A imagery and leveling survey data acquired from May 2017 to December 2023, combined with multiple predictive models, this framework was applied to Xiamen City, China. Our findings reveal that: (1) the main subsidence velocity in Xiamen ranged from −12.5 to 7.5 mm/yr, with Xiang’an District experiencing the most significant subsidence, and Huli District the least; (2) We divided Land Subsidence Risk (LSR) into five grades, identifying 65.37 km 2 (5.73 % of the total assessment area) as higher risk, predominating in Haicang District, Jimei District, Tong’an District, and Siming District; (3) The LSR of Xiamen Island and Dadeng Island needs special attention, due to their economic importance and the ongoing infrastructure projects, respectively; (4) Model performance analysis indicated that the Informer model outperformed eXtreme Gradient Boosting (XGBoost) model and Seasonal AutoRegressive Integrated Moving Average (SARIMA) model with the fusion model further enhancing prediction accuracy. These findings offer a scientific basis for land subsidence mitigation and support decision-making for enhancing urban resilience and coastal zone management.