Large-scale Surface Subsidence Monitoring for Permafrost Degradation along the Qinghai-Tibet Highway Corridor Using SBAS-InSAR and Machine Learning

Continuous permafrost degradation on the Qinghai-Tibet Plateau (QTP), driven by climate warming, and intensified by human activities such as highway construction, has triggered widespread surface subsidence. This process impacts regional ecology, increases the frequency of thermokarst hazards, and threatens the stability of the Qinghai-Tibet Highway (QTH), underscoring the need to clarify the relationship between surface subsidence, permafrost degradation, and anthropogenic influences. To monitor and evaluate these processes accurately, this study employed SBAS-InSAR technology to quantify the spatiotemporal characteristics of surface deformation along the 100-km-wide QTH corridor from July 2018 to June 2023, integrating machine learning to analyze controlling factors including human-induced disturbances. Results show linear subsidence rates of up to 40 mm/yr and seasonal deformation amplitudes of 0–30 mm, with primary displacement concentrated in northern and central sections ( e.g. , Wudaoliang, Tuotuohe). Significant settlement occurred in areas with gentle slopes (<3°), north-facing aspects, mean annual ground temperatures (MAGT) of −1 to 0 °C, active layer thicknesses (ALT) of 2–3 m, high ground ice content (GIC) of 6–9 × 10 6 m 3 , and sparse vegetation—with marked intensification near highway infrastructure and construction zones. MAGT emerged as the key factor influencing subsidence rate (r = 0.629, p ≤0.01), while slope and proximity to road embankments jointly dominated seasonal deformation (slope r = −0.488, p ≤0.01). Temporal deformation sequences correlated closely with climate change, soil hydrothermal dynamics, and highway-induced thermal disturbances. High-subsidence areas showed increased susceptibility to thermokarst landslides. Specifically, subsidence rates exceeding 10 mm/yr elevated the probability of high landslide risk by a factor of 2.1. These findings confirm that surface subsidence, modulated by both natural processes and human activities ( e.g ., pavement thermal effect, vehicle loads), serves as a robust indicator for monitoring permafrost degradation in engineering corridors.