Subsidence Risk Assessment in Guangdong, China via SBAS-InSAR and LightGBM: Dongguan Metro Line 2

SBAS-InSAR technology was used to obtain ground subsidence data along Dongguan Metro Line 2 in Guangdong Province, China, covering the period 2023–2025. A LightGBM machine learning algorithm was applied to perform sensitivity analysis on nine factors influencing subsidence. The analysis identified the following factors in descending order of importance: distance to river, distance to fault, distance to railway, distance to road, slope, digital elevation model (DEM), topographic wetness index (TWI), land use, and geomorphology. Based on a risk matrix approach combining subsidence intensity and factor sensitivity, the study area was classified into five risk levels. The results reveal that Metro Line 2 lies predominantly in low-to-moderate risk zones. This integrated SBAS-InSAR and LightGBM approach provides a novel framework for regional subsidence risk assessment.