The Cryosphere (Jul 2025)

Detection and reconstruction of rock glacier kinematics over 24 years (2000–2024) from Landsat imagery

  • D. Cusicanqui,
  • P. Lacroix,
  • X. Bodin,
  • B. A. Robson,
  • A. Kääb,
  • S. MacDonell,
  • S. MacDonell

DOI
https://doi.org/10.5194/tc-19-2559-2025
Journal volume & issue
Vol. 19
pp. 2559 – 2581

Abstract

Read online

Rock glacier velocity is now widely acknowledged as an Essential Climate Variable for permafrost. However, representing decadal regional spatiotemporal velocity patterns remains challenging due to the limited availability of high-resolution (<5 m) remote sensing data. In contrast, medium-resolution satellite data (10–15 m) covering several decades are globally available but have not been widely used for rock glacier kinematics. This study presents a robust methodological approach combining pairwise feature-tracking image correlation with medium-resolution Landsat 7/Landsat 8 optical imagery, surface displacement time-series inversion and the automatic detection of persistent moving areas (PMAs). Applied to rock glacier monitoring in the semiarid Andes of South America, this methodology enables the detection and quantification of the surface kinematics of 153 rock glaciers, 124 landslides and 105 unclassified landforms over 24 years across a 2250 km2 area. This is the first time that Landsat images have been used to quantify rock glacier displacement time series. The study estimates an average velocity of 0.30±0.07 m yr−1 for all PMAs, with rock glaciers moving 23 % faster (0.37 m yr−1) over the 24-year period. Some large rock glaciers and debris-frozen landforms exhibit surface velocities exceeding 2 m yr−1. The results align well with high-resolution imagery, recent Global Navigation Satellite System measurements and previous inventories. However, the Landsat 7/Landsat 8 (L7/8) imagery-derived velocities are underestimated by approximately 20 %–30 % on average. High uncertainties between consecutive image pairs limit the reliability of interpreting annual velocity variations. However, decadal velocity changes exceeding the uncertainties were observed in only 2 % of PMAs, with two (one) rock glaciers exhibiting significant acceleration (deceleration) over the past two decades. Our calculations show that decadal velocity changes <0.4 m yr−1 are generally within the uncertainty range when using L7/8 data, with sensitivity depending on the reference period. Despite these limitations, our results highlight the correlation between velocity trends and topographic parameters such as PMA size, orientation, slope and elevation. These relationships suggest that permafrost thaw may influence the occurrence of high-altitude landslides. Overall, this study demonstrates the feasibility of using medium-resolution optical satellite imagery for monitoring rock glacier velocity over several decades.