InSAR and machine learning reveal new understanding of coastal subsidence risk in the Yellow River Delta, China

Coastal subsidence is a geological disaster that has devastating consequences. However, an accurate understanding of its risks involves more than simply assessing the amount or rate of land subsidence. The existing methods used to evaluate geological disaster risks depend on extensive data collection, entail substantial workloads, suffer from error estimation challenges, and lack regional adaptability. These limitations prevent us from fully understanding coastal subsidence risks in estuarine deltas. Therefore, in this study, we propose a new subsidence risk assessment method that addresses the challenges of traditional geological risk assessments in terms of spatial coverage, spatiotemporal resolution, and data collection difficulty. First, Sentinel-1 multitemporal interferometric synthetic aperture radar (MT-InSAR) and cluster analysis were used to estimate the subsidence hazards. Subsequently, Landsat-8 imagery and a random forest (RF) classifier were used to obtain land use and land cover (LULC), and the analytic hierarchy process (AHP) was used to obtain settlement vulnerability. Thereafter, subsidence susceptibility was derived from the sediment layer thickness. By combining subsidence hazard, vulnerability, and susceptibility, the first subsidence risk map with a 30 m resolution was generated. The results showed that 4.54 % of the Yellow River Delta (YRD) area was high-risk, 8.75 % was medium -risk, and 10.14 % was low -risk. Notably, the risk map shows a clear overlap between high -risk and saltwater mining areas in the YRD. The proposed method is expected to improve our understanding of the coastal subsidence risk in estuarine deltas. Considering that the risk in high -value economic areas in the YRD is increasing, whereas the risk in low -value economic areas may change owing to human activity, early preventive measures are required.