Spatial organization of InSAR deformation-rate extremes across permafrost-transition and thermokarst contexts on the Tibetan Plateau

Ground motion in degrading permafrost terrain is uneven, and regional averages obscure the rare pixels—those with the most negative rates or the sharpest neighborhood contrasts—that matter most for infrastructure screening. Whether these extremes form organized, recurrent spatial groups or scattered anomalies has remained open. We analyzed the approximately 2015–2021 Sentinel-1 InSAR-estimated up-component rate and its spatial gradient along the Qinghai–Tibet Engineering Corridor, defining extremes separately within the mapped permafrost (PF) and non-permafrost (NPF) domains. The extremes are spatially organized. Mapped transition segments contain 68.0% of NPF rate extremes and 71.3% of gradient extremes at 6.6 and 5.1 times interior prevalence, and PF rate extremes occur 2.4–6.5 times as frequently near mapped lakes, thaw slumps, or both as far from either inventory. Across 12 spatial evaluations, neighborhood differences in environmental variables consistently improved prediction of extremes at withheld pixels only within PF in the primary models. Among pixels retained after temporal quality control, roughly half of 2015–2021 extremes recurred in both 2015–2018 and 2019–2021, rising to 59–62% in the NPF transition, and within the same record, rate-selected pixels accumulated more negative line-of-sight displacement than matched non-extreme pixels in both viewing geometries. These map-referenced populations provide a transparent basis for screening and follow-up monitoring while supporting association-based interpretation rather than pixel-level attribution to thaw, thermokarst, or any single process.