Spatiotemporal Dynamics of Retrogressive Thaw Slumps in the Shulenanshan Region of the Western Qilian Mountains

Abstract

Climate warming is accelerating the degradation of permafrost, particularly in mid- to low-latitude regions, resulting in the widespread formation of thermokarst landscapes, including retrogressive thaw slumps (RTSs). These landforms, which are predominantly formed by the thawing of ice-rich permafrost, have been shown to impact topography, hydrology, and ecosystem dynamics. However, spatiotemporal changes in RTS distribution and development in mid- to low-latitude permafrost regions are not well understood. This study investigates RTS spatiotemporal dynamics in the Heshenling area of the western Qilian Mountains using multi-temporal PlanetScope and Google Earth imagery, along with Sentinel-1 InSAR data acquired from 2014 to 2023. The results reveal 20 RTSs, averaging 3.7 ha in area, primarily distributed on slopes of 7–23° and at elevations of 3455–3651 m a.s.l. The deformation rates of RTSs ranged from −54 to 27 mm/year. Three developmental stages—active, stable, and mature—were identified through analysis of surface deformation and geometric variations. Active RTSs exhibited accelerated headscarp retreat and debris tongue expansion, with some slumps expanding by up to 35%. This study highlights high temperatures and rainfall as potential factors contributing to the accelerated development of RTS in arid alpine environments, and suggests that RTS activity is likely to accelerate with continued climate change.

1. Introduction

Permafrost, a crucial component of the cryosphere, is experiencing rapid degradation due to accelerated global warming, particularly at its lower limit and in permafrost boundary regions [1,2,3]. Under various projected global warming scenarios, continued permafrost degradation is anticipated, with some regions potentially experiencing complete thaw within the coming centuries [4,5]. A prominent manifestation of permafrost degradation and the thawing of ice-rich permafrost is the development of thermokarst landforms, including thermokarst lakes, thermokarst settlements, and thaw slumps, among others. Retrogressive thaw slumps (RTSs), characteristic landforms resulting from the thaw of ice-rich permafrost or the melting of massive ground ice, are prevalent throughout the Arctic, the subarctic, and high mountain regions, such as the Qinghai–Tibet Plateau (QTP) [6,7,8]. Over recent decades, the number, area, and growth rate of RTSs have increased substantially across permafrost regions (e.g., Mu et al. [9]). For example, the number of RTSs in Canada increased 60-fold between 1984 and 2015, with over 4000 new RTSs initiated, indicating a rapid acceleration of thaw slump activity in response to climate change [10].

Triggered by the exposure and subsequent thawing of ice-rich sediments, RTSs can be caused by various processes, including lateral stream or coastal erosion, surface water flow melting channelized ice wedges, or active-layer detachment slides [11,12,13]. A typical RTS features a near-vertical headwall where ground ice ablates, a scar area or slump floor, and a debris tongue consisting of a supersaturated slurry that moves downslope [14]. The long-term activity of retrogressive thaw slumps (RTSs), persisting for decades after initiation by physical or thermal erosion, profoundly reshapes terrain morphology. These features expose thick layers of ground ice or ice-rich permafrost, alter slope gradients, and transport large volumes of thawed material. Consequently, RTSs are typically considered adverse engineering geological phenomena [15]. Moreover, these dynamic features not only alter local topography and hydrology processes but also mobilize vast amounts of sediment, organic carbon, and greenhouse gases, posing cascading impacts on ecosystems and infrastructure [16,17].

Situated at the northeastern margin of the Qinghai–Tibetan Plateau (QTP), the Qilian Mountains are a climate-sensitive region experiencing accelerated permafrost degradation under a warming–wetting climate [18,19]. Recent studies have documented a dramatic increase in retrogressive thaw slump (RTS) activity, with a fourfold rise in RTS numbers and a sixfold expansion of affected areas across three regions between 2008/2013 and 2021 [20]. This increase is likely attributable to high air temperatures during the thawing season, which have triggered numerous active-layer detachment slides. Furthermore, accelerated surface deformation of RTSs on the QTP has been successfully detected using remote sensing methods, particularly interferometric synthetic aperture radar (InSAR) [21,22,23]. However, existing studies have primarily focused on plateau areas, with limited attention to steep alpine slopes where complex terrain–ice interactions govern RTS initiation and evolution.

This study investigates the dynamics of retrogressive thaw slumps (RTSs) in the Heshenling area near the Y617 Road, located in the western Qilian Mountains. Combining multi-temporal remote sensing imagery and Sentinel-1 InSAR data, we aim to: (1) characterize the geometric and topographic controls on RTS distribution; (2) investigate RTS activity across different developmental stages and quantify associated movement rates; and (3) elucidate the interplay between microclimatic factors and RTS dynamics. Located at the lower limit of permafrost regions within the Qilian Mountains, this study contributes to a better understanding of RTS development in response to permafrost degradation, as well as providing insights into improved hazard assessments and adaptive strategies for infrastructure resilience in warming permafrost landscapes.

2. Materials and Methods

2.1. Study Area

The Qilian Mountains, situated at the northeastern edge of the Qinghai–Tibetan Plateau, constitute a fragile and climate-change-sensitive region (Figure 1). This study focuses on the Shulenanshan Mountains, a western subrange of the Qilian Mountains, specifically the Heshenling area near the Y617 Kaimao Road in Tianjun County, Haixi Mongolian and Tibetan Autonomous Prefecture, Qinghai Province, China (97°9′17″–97°11′53″ E, 38°58′46″–38°59′54″ N). The terrain generally slopes from southwest to northeast, with elevations ranging from 3400 m to 3700 m a.s.l. Geologically, the area is part of the pre-Cambrian crystalline schist belt of the central Qilian Mountains, composed of Sinian siliceous limestone, dolomitic limestone, sandstone, and slate [24]. The climate of the study area is continental and arid, characterized by a large annual temperature range, intense solar radiation, and sparse vegetation [25]. Its interior Eurasian location limits water vapor transport, resulting in a fragile ecological environment. Winters are long, cold, and dry, while summers are short and relatively humid. Annual precipitation ranges from 100 mm to 300 mm, and the mean annual air temperature (MAAT) is between −6 °C and 3 °C. Permafrost conditions vary with altitude: continuous permafrost at high altitudes, discontinuous and sporadic permafrost at medium altitudes, and seasonally frozen ground at lower elevations [26].

2.2. Dataset

(1)

Digital elevation models (DEM) data

1-arcsecond (30 m) SRTM DEM data, released by the United States Geological Survey (USGS) accessed on 23 September 2014 (https://earthexplorer.usgs.gov/), were used in this study, to estimate and remove the topographic phase during InSAR processing.

(2)

Remote sensing image

Multi-source remote sensing data, including Google Earth and PlanetScope imagery, were employed for the manual delineation of RTS boundaries. PlanetScope imagery, acquired between February 2014 and October 2023, provided a spatial resolution of 3–5 m and contained red, green, and blue spectral bands, with a reported horizontal accuracy of 5–10 m (https://earth.esa.int). This PlanetScope dataset was utilized for both the delineation of RTS boundaries and subsequent temporal analysis. Google Earth provided access to archived historical imagery, primarily from 2010, 2011, and 2019, with a coarser spatial resolution of 15–30 m (https://earth.google.com/). This imagery was used in conjunction with the PlanetScope data to corroborate the identification of the landforms as RTS.

(3)

Sentinel-1 data

A total of 204 descending Sentinel-1A images, acquired from 28 October 2014 to 30 August 2023, were collected. This dataset covers a full seasonal cycle and was used for SBAS-InSAR analysis. All Sentinel-1A scenes were VV-polarized with incidence angles of approximately 39.26°. The spatial baselines ranged from −59.95 to 356.23 m.

(4)

Local precipitation and temperature data

Local precipitation and temperature data from 2015–2023 were derived from two sources: (1) the Climatic Research Unit gridded Time Series (CRU TS) monthly high-resolution gridded multivariate climate dataset [27] for the study area, and (2) meteorological measurements from two nearby national weather stations. These data were used to analyze the factors influencing landslide occurrence. CRU TS is a widely used global climate dataset with a 0.5° × 0.5° latitude/longitude grid, derived from interpolated monthly climate anomalies from extensive weather station networks. Data from two national meteorological stations, Lenghu (2772.6 m a.s.l.) and Dulan (3193.2 m a.s.l.), operated by the China Meteorological Administration (http://www.cma.gov.cn/), were also used.

2.3. RTSs Geometric Characteristics

RTS areas were manually delineated within ArcMap (Version 10.8) based on Google Earth and PlanetScope imagery, as well as the InSAR-derived surface deformation results. Two experienced researchers independently conducted the delineation, using surface deformation and geomorphological features (e.g., ground collapse, flow, and absence of vegetation) as criteria. To analyze the distribution of RTS occurrences and their relationship with environmental factors, terrain characteristics (elevation, slope, and aspect) and geometric characteristics (area, length, and width) were calculated for each RTS.

2.4. InSAR Processing

Given the seasonal dynamic changes of the permafrost surface in the study area, the interferogram pair selection was carefully considered based on the spatiotemporal baseline distribution and coherence to mitigate spatiotemporal decorrelation. After the co-registration of all SAR images, interferograms with perpendicular baselines < 5% and temporal baselines < 180 days were checked for coherence, and only those with high coherence were selected. Using the SAR data from 12 April 2020, as the super master image, a total of 1013 interferometric image pairs were generated, with an average temporal baseline of 47 days, an average spatial baseline of 61 m, and a maximum spatial baseline of 260 m (as shown in Figure 2).

The azimuth and range resolution ratio of the Sentinel-1 IWSLC data is approximately 4:1. To achieve a more uniform resolution, a 4x multi-look processing was applied in the range direction during interferometric processing. Following interferogram generation, the flat-earth and topographic phases were removed (flattening), and Goldstein filtering was employed to reduce phase noise. Minimum Cost Flow was used for phase unwrapping. The filtered interferograms, unwrapped phase images, and coherence coefficient images were manually inspected to exclude image pairs with poor coherence or unwrapping errors. Image pairs exhibiting high coherence and accurate unwrapping were used for track refinement and re-flattening. Ground control points (GCPs) were manually selected for track refinement and re-leveling. Finally, time-series analysis was conducted on the unwrapped and corrected phase to retrieve displacement changes. The accuracy of vertical movement rates ranged from 0.35 mm to 2.43 mm, with a mean value of 1.60 mm. The accuracy estimation incorporated temporal baseline and multitemporal coherence analysis, despite which subsequent field monitoring should be conducted to validate these measurements.

3. Results

3.1. Terrain and Geometric Characteristics

Integrating optical imagery and InSAR-derived surface deformation data, a total of 20 RTSs were identified in the study area. The spatial distribution of these RTSs, mapped based on PlanetScope imagery from 11 October 2023, is shown in Figure 3a. The results indicate that the areas of RTSs in the study region range from 0.3 to 19.4 ha, with an average area of 3.7 ha (

Table 1. Geometric characteristics of the retrogressive thaw slumps in the study area.

No.	Area/ha	Length/m	Width/m	Slope/°	Aspect	Elevation/m	Type
1	0.8	200.9	71.3	20.55	N	3618.334	Stable
2	0.7	160.9	44.3	10.15	NW	3474.288	Mature
3	4.6	548	114.2	11.58	NW	3501.196	Stable
4	1.4	391	54.1	9.39	NW	3487.292	Stable
5	3.6	633.1	68.9	7.22	NE	3537.859	Active
6	2.3	372.9	100.6	13.35	N	3598.773	Active
7	5.3	519.8	218.7	10.81	NE	3601.044	Stable
8	0.3	93.3	42.2	2.37	W	3472.859	Mature
9	0.3	127.3	43.5	15.15	NE	3607.837	Mature
10	19.4	1312.1	211.1	7.57	NW	3589.668	Active
11	6.2	683.1	92.9	16.26	NE	3605.761	Active
12	2.6	316.1	88.1	15.43	NE	3570.772	Stable
13	14.6	763.3	191.9	17.89	N	3638.459	Active
14	3.8	595.1	85.1	17.39	NW	3520.616	Mature
15	0.6	207.1	46.3	17.25	NE	3651.339	Stable
16	1.4	313.3	76.4	16.74	NE	3581.576	Stable
17	1.0	150.2	74.3	11.35	E	3594.876	Active
18	0.7	167.4	60.1	16.47	NE	3640.511	Stable
19	1.1	399.5	31.5	11.34	NW	3455.1	Active
20	3.3	722.1	71.8	11.12	W	3501.483	Active
Average	3.7	433.83	89.37	12.90		3562.482

). The majority of RTSs (80%) were less than 5 ha, with two (10%) surpassing 10 ha (Figure 3b). The largest slump, No. 10, covers 19 ha with a length of 1.31 km, while the smallest, No. 8, measures 0.3 ha with a length of 0.093 km.

Geomorphologically, the RTSs in the study area are predominantly characterized by retrogressive sliding, resulting in elongated shapes. Some slumps exhibit multiple heads or debris tongues. Notably, thermocirque or thermoterrace-developed slumps were not observed in the study area. The slumps primarily expand through upward retrogressive erosion, with limited lateral expansion, contributing to their generally long and narrow forms. Most slumps (80%) have lengths between 100 m and 700 m, averaging 391 m. One slump (No. 10) exceeds 1000 m in length, whereas the shortest, No. 8, measures 93.3 m.

The headwall elevations of RTS in this study range from 3455 m a.s.l. to 3651 m a.s.l., with an average of 3562 m. The lowest headwall elevation was observed for No. 19 (3455 m), and the highest for No. 15 (3651 m). Approximately 80% of the slumps (16 in total) have headwalls situated between 3500 m and 3650 m (Figure 3d), proximate to the local permafrost lower limit. Slump slopes vary from 7.22° to 22.55°, averaging 12.89° (Figure 3c), slightly steeper than those documented in the Beiluhe region of the Qinghai–Tibet Plateau [7]. Slopes exceeding 30° were not identified, as such steepness is more characteristic of rock and rock–ice avalanches [28]. The average elevation difference across the slumps is 60 m, ranging from 3 m for No. 8 to 196 m for No. 10. Furthermore, the majority of slumps (95%) are oriented towards shaded slopes (NE, NW, N, and W), with respective proportions of 40%, 30%, 15%, and 10% (Figure 3e). Southern aspects were not observed in the study area.

3.2. Spatiotemporal Changes of RTS Areas

Based on visual interpretation of historical PlanetScope remote sensing imagery, the spatiotemporal evolution of the thaw slumps was analyzed from February 2014 to October 2023. Based on area changes, the RTSs were categorized into three types: active, stable, and mature. Active RTSs exhibit continuous expansion; stable slumps show minimal area change; and mature slumps experience the surface progressively stabilizing, particularly in the upper thaw–collapse zones.

(1)

Active RTSs

There are eight active RTSs in the study area, accounting for 35% of the total. These slumps are characterized by ongoing headscarp melting and collapse, with saturated loose material sliding or flowing along the thaw interface, resulting in expansion of the head and/or accumulation zones. Over the period 2014–2023, the total area occupied by active RTSs in the study region changed from 41.22 to 50.45 ha, with a mean size of 6.31 ha. The active slumps demonstrated an average rate of area change of 40%. For instance, No. 10, a multi-headed, tongue-shaped RTS, displayed varying retrogressive rates due to heterogeneous surface vegetation, material composition, and ground ice distribution. Between September 2020 and May 2021, the northern and southern headwalls of No. 10 retrogressed more rapidly than the central portion, resulting in a 500 m advance of the debris tongue and a 5 ha area increase, representing a 35% change (Figure 4). In contrast to No. 10, RTS No. 11 is a long slump with a multi-debris tongue. Between September 2020 and May 2021, the headwall area changed minimally, but the debris tongue developed additional branches due to topographic constraints and local factors, with a 23% area increase.

(2)

Stable RTSs

Eight stable RTSs (45%) were identified, characterized by decelerating collapse rates and stabilization as the slumps approached hilltops or the margins of thick ground ice. For example, No. 4, east of Kaimao Road, showed negligible area change from December 2014 to October 2023. The melting and collapse in the head area largely ceased, with the surface morphology approaching stability and some areas exhibiting new vegetation cover. However, the debris tongue continued to show notable changes in surface morphology, indicating that while sliding has not completely ceased, the deformation rate has gradually decreased, and the step-like and imbricate landforms are diminishing (Figure 4).

(3)

Mature Slumps

Four mature slumps (20% of the total) were observed, characterized by the cessation of melting and collapse in the head areas, a trend towards surface stabilization, gradual reduction in slump area, and the development of new vegetation cover in some areas. For instance, a comparison of historical PlanetScope remote sensing imagery from September 2020 and February 2017 reveals an approximately 34% reduction in the overall area of RTS No. 9. The head area is largely covered by secondary vegetation, with some areas exhibiting exposed bedrock, indicating the cessation of melting and collapse and a move towards surface stability. Due to the diminished material supply, the debris tongue is gradually becoming smoother, with vegetation encroaching on the margins, further indicating that the RTS is in a mature phase and the surface morphology is progressively stabilizing.

3.3. Surface Deformation Characteristics

(1)

Active RTSs

The average annual movement rate for active RTSs is 7.1 mm·a⁻¹, with the maximum rate of 54 mm·a⁻¹ observed at slump No. 10. These slumps are primarily characterized by higher deformation rates in the upper source areas, with vertical displacement predominantly negative (subsidence), while the accumulation zones exhibit lower deformation rates and predominantly positive vertical displacement (uplift). For instance, the average annual rate of RTS No. 10 from 2014 to 2023 was 11.4 mm·a⁻¹. The average movement rate gradually increased from 3.93 mm·a⁻¹ in 2014 to 15.23 mm·a⁻¹ in 2020, peaking in that year. Subsequently, the annual deformation rate decreased from 2020 to 2023. Historical PlanetScope remote sensing imagery corroborates this, showing that slump No. 10 experienced significant sliding in 2020, with the highest rate of area change.

Spatially, the highest deformation rates are observed in the upper source areas. Three characteristic points (P1, P2, and P3) were selected at different elevations within RTS No. 10 as shown in Figure 5. P1 exhibited an average deformation rate of 33.6 mm·a⁻¹ and a cumulative displacement of −267 mm; P2 had an average deformation rate of 44.5 mm·a⁻¹ and a cumulative displacement of −363.1 mm; and P3 showed an average deformation rate of 1.10 mm·a⁻¹ and a cumulative displacement of 8.4 mm. This indicates that the highest thaw slump movement rates occur in the upper and middle sections of the slump body, with the deformation rate in the source area significantly greater than that in the accumulation zone at the toe.

(2)

Stable RTSs

The average annual movement rate of stable RTSs is 5.68 mm·a⁻¹, approximately 20% lower than that of active RTSs. Over the period from 2016 to 2023, the annual movement rate of stable RTSs showed minimal variation and remained relatively constant. Spatially, stable RTSs exhibit a distinct deformation pattern compared to active slumps, with the highest deformation rates observed in the debris tongue, while the upper source areas remain relatively stable. Taking slump No. 4, located near Y617 Kaimao Road, as an example, three characteristic points (P1, P2, and P3) were selected at different elevations (Figure 5). P1 exhibited an average deformation rate of 2.10 mm·a⁻¹ and a cumulative displacement of −5.4 mm; P2 had an average deformation rate of 8.09 mm·a⁻¹ and a cumulative displacement of −62.5 mm; and P3 showed an average deformation rate of 14.83 mm·a⁻¹ and a cumulative displacement of −126.6 mm. These data indicate that the overall deformation of the slump body is not significant, with minimal deformation observed in the upper source area, while the debris tongue at the toe exhibits slightly higher movement rates.

(3)

Mature RTSs

The average annual movement rate of mature RTSs in the study area is 1.31 mm·a⁻¹, with a maximum rate of 7.7 mm·a⁻¹ observed at RTS No. 9. Deformation rates across all areas of mature RTSs are significantly lower than those of both active and stable RTSs, similar to the deformation observed in the surrounding stable permafrost regions. For instance, three characteristic points (P1, P2, and P3) were selected within the RTS No. 9. P1 exhibited an average deformation rate of 2.04 mm·a⁻¹ and a cumulative displacement of −12.8 mm; P2 had an average deformation rate of 2.29 mm·a⁻¹ and a cumulative displacement of −9.1 mm; and P3 showed an average deformation rate of 6.86 mm·a⁻¹ and a cumulative displacement of −53.3 mm. The movement rate in the upper source area of the slump is low, while the accumulation zone exhibits slightly higher rates. A nearby flat area was selected as a control area. This area showed an average annual movement rate of 1.77 mm·a⁻¹, similar to the deformation rates observed in slump No. 9, suggesting that the surface deformation in mature slumps is primarily driven by seasonal freeze–thaw processes.

4. Discussion

4.1. Triggering Mechanisms of RTS Activities

The observed increase in RTS dynamics in this study correlates with concurrent upward trends in temperature and precipitation. The active layer, overlying permafrost with its characteristically low hydraulic conductivity, functions as a shallow aquifer, controlling surface and subsurface hydrological processes [29]. Heavy precipitation, snowmelt, and high summer air temperatures contribute to increased thaw of ice-rich permafrost and the active layer, resulting in a substantial increase in liquid water content within the active layer. This increased water storage within the active layer leads to elevated pore water pressures, which can destabilize slopes and intensify landslide activity [30,31,32,33].

Given the lack of national weather stations within the study area, temperature and precipitation data were derived from CRU TS monthly high-resolution gridded multivariate climate data [27]. To validate the reliability of these gridded data, a comparative analysis was performed using data from the two nearest national weather stations (locations shown in Figure 1). While the national weather stations are geographically distant and at different elevations, they recorded similar temperature trends. All the meteorological data reveal temperature extremes in 2016 and 2021/2022, corresponding with periods of extreme deformation rates in the study area’s RTSs (Figure 6). Comparable observations have been reported across the Qilian Mountains and other regions of the Qinghai–Tibet Plateau, notably, in 2016, extreme thermal anomalies were associated with significant increases in the frequency, spatial extent, and displacement rates of RTSs [7,34]. While extreme rates of RTS movement occurred in 2016 and 2021/2022, the most significant area changes, suggesting the concentration of initial RTS development or reactivation, were observed in 2017, 2020, and 2021. This clustering of developmental activity is linked to prolonged periods of elevated temperatures from 2013 to 2016 and 2020 to 2022, indicating a potential lag in the RTS response to warming.

To investigate the influence of temperature and precipitation on landslide movement rates, statistical tests were conducted to assess the correlations between meteorological parameters (mean temperature from May to October and annual precipitation) and movement rates. As depicted in Figure 7, a statistically significant correlation (p < 0.05) was observed between the mean temperature during the warm season and landslide movement rates. Although annual precipitation did not exhibit a statistically significant correlation with movement rates, a positive trend was noted, indicating a general increase in movement rates with higher precipitation levels. The lack of a statistically significant correlation between precipitation and movement rates can potentially be explained by two primary factors: (1) the region’s arid climate, characterized by consistently low annual precipitation, which may limit the influence of rainfall as a primary trigger for slope instability; and (2) the restricted sample size of rainfall-induced landslide events within the dataset, which may diminish the statistical power necessary to detect subtle relationships between precipitation and movement. Notably, the study area exhibited the lowest MAAT in 2019 within the 2015–2022 period, yet InSAR-derived deformation measurements demonstrate anomalously high deformation rates, which consistently coincide with extreme rainfall events. Heavy precipitation events drive permafrost degradation and slope instability through coupled hydrological–thermal mechanisms. Substantial rainfall infiltration raises the water table, destabilizes slopes, and simultaneously transfers latent heat to the permafrost table. This dual process accelerates ground ice ablation and increases pore water pressure within the active layer, promoting both active-layer detachment failures and RTS initiation. Intense precipitation, particularly when coinciding with early snowmelt, further amplifies thermal erosion and accelerates headwall retreat rates [30]. The study area in the western Qilian Mountains exhibits characteristics of a continental arid climate, with mean annual precipitation of ~100–200 mm concentrated in the warm season (June–August). The shallow snowpack and limited snow water equivalent suggest minimal spring meltwater influence on slope deformation. Notably, 2019 experienced significant precipitation anomalies, recorded in both CRU TS4.06 gridded data and observations from the Lenghu meteorological station, although temperatures were not extreme. This hydrometeorological event coincided with peak retrogressive thaw slump (RTS) activity, as evidenced by the maximum mean deformation rate of −9.62 mm/yr observed between 2014 and 2024. This acceleration occurred despite lower mean annual temperatures, highlighting the dominant role of liquid water input over thermal controls in driving short-term RTS dynamics.

4.2. Development of RTSs

RTSs can exhibit various developmental cycles, progressing through active, stabilized, and mature phases, or even a polycyclic pattern involving reactivation [14]. In this study, RTSs predominantly occur on high slopes ranging from 7.22° to 22.55°. But no statistically significant differences were observed in slope gradient (χ² = 1.72, p = 0.423) or aspect (χ² = 8.02, p = 0.432) across the examined states. These findings may be influenced by sample size limitations and the continuous nature of stage transitions. Although no waterbodies were observed near the RTS bodies, the initial development of RTSs often coincides with surface water enrichment, which can reduce effective stress and soil strength due to elevated pore water pressures on the sloping terrain. Our findings indicate that 40% of the observed RTSs remain in the active developmental stage, characterized by head scarp retreat and debris tongue expansion driven by ice-rich permafrost ablation. These active RTSs are typically triggered by extended warm periods and heavy precipitation/snowmelt events, which generally lead to a deepening of the active layer and subsequent thawing of ice-rich deposits or massive ground ice, as discussed previously.

While the activity characteristics and triggers of RTSs on the Qinghai–Tibet Plateau (QTP) have been extensively studied in the context of global warming, the stabilization, degradation, and polycyclic behavior of RTSs on the QTP remain underexplored. Our results reveal that 40% of the studied RTSs exhibit stabilization patterns. RTS stabilization is primarily attributed to the slumps reaching hilltops or the margins of thick ground ice, resulting in the complete melt of exposed ice or its reburial by sediments, effectively insulating it from further thawing. In addition, the exposed ice is re-buried by sediments and thermally fully insulated from further melting could also cause RST into the stable stage, which may be reactively triggered by extended warm periods and heavy precipitation/snowmelt events since the thick ice was not completely melted. Furthermore, new active RTSs can form within the boundaries of a stabilized RTS, and neighboring RTSs can expand and coalesce. Therefore, the stabilized RTSs identified in this study have the potential for reactivation under extreme temperature or precipitation events, necessitating further detailed monitoring to quantify reactivation thresholds.

5. Conclusions

This study systematically investigated RTSs in the Heshenling area of the western Qilian Mountains using an integrated approach combining remote sensing analysis (PlanetScope imagery and InSAR-derived deformation data) and multivariate climate datasets. Twenty RTSs, with an average area of 3.7 ha, were identified within the study region, primarily located on slopes ranging from 7° to 23° and at elevations between 3455 and 3651 m a.s.l. Three developmental phases—active, stable, and mature—were distinguished, primarily influenced by geomorphic and soil conditions. Active RTSs exhibited accelerated headscarp retreat and debris tongue expansion, with some slumps experiencing expansion of up to 35%. Stabilization occurred upon reaching ice margins or hilltops, while the potential for polycyclic reactivation persisted due to incomplete ground ice ablation and climatic extremes. Both elevated temperatures and increased precipitation were found to induce RTS reactivation, propagation, and accelerated displacement. These results underscore that RTS evolution on the arid Qilian permafrost margin is governed by coupled thermohydrological forcing, with stage-specific deformation patterns reflecting ice ablation dynamics and climatic effects. Furthermore, the potential for rapid RTS development in the context of continued climate warming is highlighted.

However, this study is subject to limitations, including a restricted sample size (n = 20 RTSs), reliance on remote sensing data without in-situ validation, and a lack of information on internal structure, such as permafrost distribution. These limitations collectively impede a robust assessment of long-term RTS-climate feedback. Future research should expand the study area across the Qilian Mountains and incorporate enhanced UAV and InSAR monitoring across elevation gradients to accurately capture RTS deformation. Additionally, high-frequency ground-penetrating radar surveys should be conducted to quantify ground ice depletion rates across developmental stages. These advancements will improve the predictive capacity for RTS impacts on alpine ecosystem stability and infrastructure resilience under climate change.