An Integrated PS-InSAR and K-Means Clustering Framework for Roadbed Deformation Susceptibility Assessment Supporting Sustainable Urban Road Management in Daegu, South Korea

Long-term deformation of urban road networks can compromise infrastructure performance and requires spatially continuous methods for identifying segments that warrant targeted monitoring. This study integrates Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR), geotechnical data, multivariate k-means clustering, and GIS road-network analysis to evaluate roadbed deformation susceptibility in Daegu, South Korea. A total of 184 ascending Sentinel-1 SLC images acquired between June 2018 and December 2025 yielded 32,590 roadbed persistent scatterers (PS) and 247,734 surrounding-ground PS. Because a single ascending geometry was used, all measurements represent relative line-of-sight (LOS) motion rather than absolute vertical displacement. Roadbed relative LOS velocities ranged from −8.63 to + 4.87 mm/yr. Roadbed and surrounding-ground velocities correlated moderately within a 200 m neighborhood (r = 0.59), while percentile- and exceedance-based comparisons showed slightly heavier deformation tails on the roadbed. Susceptibility was assessed by clustering relative LOS velocity, SPT-derived VS30, and effective soil depth for 31,263 PS with complete data. The three resulting classes were defined by combinations of deformation and subsurface conditions rather than by velocity thresholds. The critical class exhibited the most negative mean relative LOS velocity (−0.60 mm/yr), lower mean VS30 (382 m/s), and relatively deep soil conditions, whereas the concern and stable classes had similar mean LOS velocities but were differentiated primarily by their subsurface profiles. Of the approximately 883.7 km major road network, 668.8 km was classified, comprising 161.8 km (24.2%) critical, 284.1 km (42.5%) concern, and 222.8 km (33.3%) stable roads, while 215.0 km remained unclassified because of insufficient PS coverage. Comparison with 19 documented ground-subsidence incidents yielded critical-class enrichment ratios of 0.34–0.49, but the limited sample was insufficient to either confirm or reject the susceptibility classification; 14 of the 19 incidents were associated with abrupt underground-utility or construction-related failures rather than gradual deformation processes. The proposed framework can support targeted monitoring, preventive maintenance, and sustainable urban road management.