Multi-Source Remote Sensing Investigation of Spatiotemporal Deformation and Mechanisms of the Pangcun Giant Accumulation Landslide, Southeastern Tibet

Highlights

What are the main findings?

• The Pangcun giant accumulation landslide exhibits pronounced spatial heterogeneity, with deformation primarily concentrated in the central Zone B, as revealed by long-term SBAS-InSAR analysis and confirmed by UAV and field observations.
• Landslide displacement shows an approximate 3-month lagged response to precipitation, indicating that deformation is controlled by delayed hydrological processes rather than direct rainfall triggering, with additional modulation from irrigation and other anthropogenic activities.

What are the implications of the main findings?

• Multi-source evidence clarifies a multi-factor coupling mechanism in which unfavorable topography and fragmented accumulation materials provide internal preconditions, while hydrological recharge, irrigation, slope-toe unloading, and seismic disturbance jointly regulate episodic acceleration.
• The results enhance the mechanistic understanding of giant accumulation landslides in tectonically active alpine regions and support process-based hazard assessment and early warning strategies.

Abstract

The geological environment of southeastern Tibet is characterized by complex tectonics and high climatic sensitivity, and giant accumulation landslides pose significant threats to infrastructure and human safety. This study investigates the Pangcun giant accumulation landslide using SBAS-InSAR (2017–2024), UAV photogrammetry, field investigations, and wavelet coherence analysis to examine its deformation and driving mechanisms. The landslide exhibits continuous, slow deformation with clear spatial heterogeneity, divided into two zones, with the largest displacement occurring in the middle of Zone B. Field evidence is consistent with the InSAR results. Wavelet coherence analysis reveals a lagged response of displacement to precipitation at a timescale of about three months. The landslide’s evolution is controlled by unfavorable topography and fragmented materials, with precipitation as the primary trigger. Human activities (agricultural irrigation and slope-toe road excavation) and seismic disturbances also contribute to its progressive development.

1. Introduction

Complex geomorphology, fragile geologic conditions, and intense tectonic and climatic activity characterize the southeastern Tibetan Plateau. Landslides are widespread in this region, often large in volume, frequent, and highly destructive, posing serious threats to infrastructure and human safety [1,2,3]. Understanding the spatiotemporal deformation and mechanisms of such giant landslides is therefore critical for hazard mitigation and early warning [4].

In this region, landslides often exhibit concealed deformation and complex evolutionary processes, making their identification and monitoring challenging [5]. Due to harsh alpine conditions, traditional ground-based monitoring methods (e.g., inclinometers, GNSS, and crack meters) face deployment difficulties, limited coverage, and maintenance challenges, failing to meet the demands of large-scale, long-term regional monitoring. In contrast, Interferometric Synthetic Aperture Radar (InSAR) has become an effective tool for landslide monitoring because of its wide coverage, all-weather capability, and millimeter-level sensitivity. Previous studies have applied InSAR to regional landslide detection [6,7]; time-series deformation analysis [8,9,10]; and multi-source data integration, including optical remote sensing, UAV photogrammetry, and other complementary datasets [11,12]. These studies have improved monitoring accuracy and enhanced understanding of landslide evolution in complex mountainous terrains.

Despite these advances, several limitations remain. First, most studies focus on regional-scale landslide detection, while detailed investigations into the long-term evolution of individual giant landslides remain limited. Second, previous research has mainly concentrated on rock landslides, whereas giant accumulation landslides, which are widely distributed and highly hazardous in southeastern Tibet, have received less attention. Third, InSAR-based studies often emphasize surface deformation patterns, with limited quantitative analysis of underlying mechanisms. Finally, commonly used statistical methods are insufficient to capture multi-timescale responses and time-lag effects between deformation and triggering factors [13].

Giant accumulation landslides in southeastern Tibet are typically ancient features formed under the combined influence of tectonic activity, climate change, and valley evolution [14,15,16]. Their heterogeneous materials and complex internal structures result in highly intricate deformation behaviors [17,18,19]. Owing to their relatively gentle topography, these ancient landslide deposits are often occupied by settlements, transportation routes, and engineering infrastructure, significantly increasing human exposure to landslide hazards [20]. Under the combined effects of ongoing engineering disturbances and intensifying extreme weather events, many of these landslides exhibit a high potential for reactivation. Nevertheless, systematic investigations into the spatiotemporal deformation characteristics and underlying mechanisms of giant accumulation landslides in this region remain limited, highlighting the need for in-depth case studies.

The Pangcun landslide is a typical giant accumulation landslide in southeastern Tibet, posing significant threats to both infrastructure and local communities. Previous studies on the Pangcun landslide have employed a range of methods, including InSAR-based monitoring, field investigations, and hazard assessment approaches. These works have demonstrated the ability of remote sensing techniques to capture surface deformation effectively and have provided valuable insights into the landslide’s overall deformation patterns and stability conditions. In particular, the reliability of InSAR-derived deformation has been supported through comparisons with ground-based observations [21], and the landslide has been interpreted as being in a stage of ongoing deformation influenced by evolving mechanical properties [21]. Additionally, stability evaluations under different environmental conditions have contributed to understanding its hazard potential [12].

However, existing studies primarily focus on deformation detection or stability assessment. At the same time, relatively limited attention has been paid to the long-term spatiotemporal heterogeneity of deformation processes and their time-lagged responses to hydrological factors. Moreover, the influence of anthropogenic activities, such as irrigation, remains insufficiently explored.

In this study, we investigate the Pangcun giant accumulation landslide using an integrated framework that combines time-series InSAR, UAV photogrammetry, field investigations, and wavelet coherence analysis. The objectives of this study are to (1) characterize the long-term spatiotemporal deformation patterns of the landslide; (2) quantify the spatial heterogeneity of deformation through multi-source data integration; and (3) reveal the multi-timescale and time-lagged responses between landslide deformation and external driving factors. The results provide new insights into the mechanisms of evolution of giant accumulation landslides and contribute to improved hazard assessment and early warning in southeastern Tibet.

2. Overview of the Pangcun Landslide

The Pangcun landslide (28.29°N, 92.78°E) is located in Jiayu Township, Longzi County, Shannan City, Tibet Autonomous Region. The landslide is situated adjacent to the right bank of the Jiayu River, with the G219 National Highway passing along its front edge (Figure 1). The river terrace at the landslide toe was originally a residential area of Pangcun village, comprising 90 households and 257 people. In April 2019, residents observed significant deformation, with multiple continuous cracks forming across the slope, indicating a distinct reactivation of the landslide. To mitigate geohazard risks, the entire population in this area has since been relocated.

The study area is characterized by intense topographic dissection, featuring a typical high-mountain deep-cut river valley landform. It belongs to a plateau temperate monsoon semi-arid climate zone, with an average annual temperature of 5.5 °C and an annual precipitation of 297.4 mm. Precipitation is highly seasonal, with over 80% occurring between June and September. The region experiences large diurnal temperature fluctuations and intense freeze–thaw cycles. Groundwater recharge primarily originates from atmospheric precipitation, alpine ice and snowmelt, and infiltration from agricultural irrigation. The tectonic stress field is complex, and neotectonic activity is vigorous; two recent seismic events were an Ms 5.3 earthquake on 1 December 2018 (epicenter 23 km from the landslide) and an Ms 5.6 earthquake on 19 July 2019 (epicenter 72 km from the landslide).

The boundaries of the reactivated landslide are clearly visible, exhibiting a fan-shaped planar morphology. The interface between exposed bedrock and the accumulation layer defines the left, right, and rear boundaries. At the same time, the front shear outlet is located at the G219 highway, with a primary sliding direction of NE16°. The elevation ranges from 3540–3632 m at the rear edge to 3299–3338 m at the front edge, representing a relative relief of 218–333 m. The longitudinal length of the landslide is 335–624 m, the transverse width is 294–350 m, and the total area is approximately 0.3 km². The profile exhibits a “steep-upper and gentle-lower” characteristic, with an overall average slope gradient of approximately 30°. Borehole data reveal two sliding zones: a shallow zone at a depth of 6–25 m and a deep zone at 25–55 m. The estimated volume ranges from 1.2 × 10⁷ m³ to 1.9 × 10⁷ m³, classifying it as a giant ancient accumulation landslide.

The bedrock primarily consists of Jurassic grey-to-blackish-grey, thin-layered, highly weathered metamorphic siltstone and silty slate, characterized by well-developed joints, fragmented structures, and poor integrity (Figure 2). Bedrock attitudes at the rear edge and side walls are chaotic, exhibiting pronounced toppling and bending deformation. The landslide body and its surroundings are extensively covered by Quaternary loose deposits, mainly consisting of landslide accumulations and proluvial-alluvial layers with mixed compositions. The landslide mass is primarily composed of debris-rich soil and silty clay containing breccia. The fragments are predominantly angular and subangular, with a soil-to-stone ratio of approximately 4:6 to 3:7, and exhibit a loose structure and high porosity. Proluvial–alluvial deposits are distributed on the Jiayu River terraces at the front edge, forming a binary sedimentary structure of rounded-to-subrounded cobbles and sub-clay. The sliding zone thickness is approximately 0.1–0.3 m, mainly composed of silty clay with breccia.

3. Data and Methods

3.1. Remote Sensing and Ancillary Data

In this study, Sentinel-1A satellite data from the European Space Agency (ESA) Copernicus Program were utilized for the long-term deformation monitoring of the Pangcun landslide. Sentinel-1A is equipped with a C-band Synthetic Aperture Radar (SAR), providing all-weather, day-and-night imaging capabilities suitable for continuous observation in high-altitude alpine canyon regions. A total of 80 descending Single Look Complex (SLC) images covering the study area were acquired between March 2017 and June 2024. To eliminate topographic phase effects, the USGS-released SRTM DEM (30 m) was used as an external elevation reference. Simultaneously, data from the Generic Atmospheric Correction Online Service (GACOS) were introduced to correct for atmospheric phase delays caused by tropospheric water vapor. Detailed parameters of the InSAR dataset are summarized in

Table 1. InSAR data information.

Parameter Category	Detailed Information
Satellite	Sentinel-1A
Orbit Direction	Descending
Band	C-band
Imaging Mode	IW (Interferometric Wide Swath)
Polarization	VV
Radar Wavelength	5.6 cm
Spatial Resolution	15 m
Look Angle	39.5°
Revisit Cycle	12 d
Number of Images	80 scenes
External DEM	SRTM DEM (30 m)
Atmospheric Correction Data	GACOS (90 m)
Time Period of Atmospheric Correction Data	March 2017–June 2024

.

To analyze the relationship between landslide deformation and environmental factors, monthly precipitation and temperature data from 2017 to 2024 were collected from the Longzi County meteorological station. Furthermore, to finely identify current deformation features and validate the InSAR inversion results, UAV aerial surveys and field investigations were conducted on 21 August 2024. These efforts provided critical data on fissure distribution, geomorphological features, and various signs of deformation across the landslide body.

The selected time period (2017–2024) was chosen to capture the reactivation phase of the landslide, which initiated around 2018–2019, and its subsequent evolution. This time span ensures continuous observation before and after reactivation, enabling a more comprehensive analysis of spatiotemporal deformation characteristics and underlying mechanisms.

3.2. InSAR Processing and Time-Series Deformation Retrieval

In this study, the Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) technique was employed to invert the surface deformation of the Pangcun landslide. Given the complex terrain of the study area, the SBAS-InSAR workflow was optimized. First, orbital refinement was performed on the 80 acquired Sentinel-1A SLC images. An interferometric network was then constructed based on the short-baseline principle, with the perpendicular baseline threshold set to ≤45 m and the temporal baseline threshold to ≤365 days. The master image was selected using the principle of coherence maximization: by calculating the average coherence coefficient for each image pair, the image acquired on 15 June 2019 was identified as the super-master image. Ultimately, 210 valid interferometric pairs were generated, and precise image registration was completed. Based on these thresholds, the resulting spatiotemporal baseline network of the interferometric pairs used for time-series inversion is shown in Figure 3b,c.

During the interferometric processing stage, multi-looking (4 × 1 in azimuth × range) was applied to the interferograms to improve the signal-to-noise ratio (SNR) and suppress phase noise. Goldstein filtering was subsequently performed to smooth the phase. The topographic phase was removed using the SRTM DEM to complete the differential interferometric processing. To address the atmospheric phase delay caused by tropospheric water vapor in the alpine valley, GACOS data were introduced for atmospheric correction. Additionally, a temporal high-pass filter (with a cutoff period of 365 days) was applied to separate and reduce the influence of long-period non-deformation signals.

In the phase unwrapping and deformation inversion stage, the Minimum Cost Flow (MCF) algorithm was used for 2D phase unwrapping. Based on this, a set of equations for the deformation rate was established and solved via SVD (Singular Value Decomposition) to invert the mean annual velocity field of the landslide area, with the root-mean-square error (RMSE) of the residual phase kept within 1.5 mm. Furthermore, the line-of-sight (LOS) deformation was projected onto the slope direction, and the Kriging interpolation method was employed to generate a continuous deformation field. Finally, a high-temporal-resolution surface-deformation dataset for the Pangcun landslide from 2017 to 2024 was obtained. The detailed data-processing workflow for the SBAS-InSAR technology is illustrated in Figure 4.

To relate the InSAR measurements to the actual landslide kinematics, the geometric relationship for projecting 1D LOS deformation ($d_{\mathrm{s}\mathrm{l}\mathrm{o}\mathrm{p}\mathrm{e}}$) onto the steepest slope direction ($d_{\mathrm{s}\mathrm{l}\mathrm{o}\mathrm{p}\mathrm{e}}$) is defined as follows:

$$d_{\mathrm{s}\mathrm{l}\mathrm{o}\mathrm{p}\mathrm{e}}={\displaystyle \frac{d_{\mathrm{L}\mathrm{O}\mathrm{S}}}{\cos \beta }}$$

(1)

where $\beta$ is the angle between the LOS and the slope direction, determined by the satellite incidence angle ($\theta$), heading angle ($\alpha$), slope angle ($\gamma$), and slope aspect ($A$). This projection assumes that the landslide primarily undergoes slope-parallel translational movement. A key limitation is its inability to resolve complex 3D movements, such as deep-seated rotational sliding. In this study, to avoid potential numerical artifacts and magnification errors in steep terrain, the results are presented primarily along the LOS direction. Subsequently, ordinary Kriging interpolation was applied to generate continuous deformation maps. An exponential semi-variogram model was selected based on the spatial autocorrelation of the monitoring points. While Kriging provides a statistically robust continuous surface, its inherent limitation lies in the smoothing effect, which may slightly underestimate localized extreme deformation gradients.

3.3. UAV Survey and Field Investigation

To overcome the challenges of inaccessible terrain in high-steep landslide areas and the spatial resolution limitations of satellite remote sensing imagery, this study employed a DJI Mavic 3E Unmanned Aerial Vehicle (UAV) (SZ DJI Technology Co., Ltd., Shenzhen, China) for photogrammetric operations to acquire multi-angle high-resolution imagery of the landslide area. The UAV is equipped with a 4/3 CMOS wide-angle camera (20 million effective pixels; SZ DJI Technology Co., Ltd., Shenzhen, China) and integrated with Network Real-Time Kinematic (RTK) positioning technology. This configuration enables high-precision, high-efficiency surveying and mapping, ensuring stable operation under the high-altitude, complex-terrain conditions of southeastern Tibet. On 21 August 2024, four flight missions were conducted over the Pangcun landslide area, capturing a total of 171 images with a single-frame resolution of 5280 × 3956. The UAV imagery clearly captured micro-geomorphological features, including tension cracks, step-like scarps, and headscarp displacements on the landslide surface. These data provided a reliable surface basis for identifying landslide deformation zones and validating the InSAR inversion results.

Simultaneously, a systematic field investigation was conducted in the study area, focusing on the detailed identification of signs of deformation and geomorphological characteristics. Through fixed-point photography and on-site measurements, the spatial locations and scale information of deformation features, such as cracks and scarps, were obtained. This fieldwork further clarified the spatial distribution range of the accumulation body and the structural characteristics of the primary rock and soil masses, providing necessary physical constraints for the analysis of spatiotemporal deformation characteristics and the interpretation of landslide mechanisms.

3.4. Wavelet Coherence Analysis

Landslide deformation and triggering factors, such as rainfall and temperature, typically exhibit certain nonlinear, non-stationary, and time-varying characteristics. Traditional linear correlation methods (e.g., Pearson correlation coefficient) are based on the assumption of stationarity in time series and can only reflect the overall average degree of correlation. Consequently, they struggle to characterize the local evolutionary features and lag effects of deformation responses across different time scales. To overcome these limitations, Continuous Wavelet Transform (CWT) and Wavelet Coherence (WTC) were introduced, enabling systematic analysis of multi-scale, time–frequency relationships between landslide deformation and environmental factors while preserving temporal localization (Figure 5) [22,23]. First, CWT time–frequency analysis using the Morlet wavelet was performed on both the cumulative displacement and rainfall time series. A cubic polynomial fit was applied to detrend the original sequences, thereby reducing the influence of long-term cumulative trends on periodic fluctuation signals and enabling the identification of dominant cycles and energy distribution characteristics of landslide deformation across various evolutionary stages [24].

Building upon the univariate analysis, Cross-Wavelet Transform (XWT) and Wavelet Coherence (WTC) were further utilized to explore the response mechanisms of landslide deformation to variations in rainfall and temperature. Specifically, XWT was employed to identify significant regions in the time–frequency domain where the two time series share high common energy. WTC, by eliminating the influence of amplitude differences, quantitatively characterized the local degree of correlation (ranging from 0 to 1) between the two variables across different spatiotemporal scales. Throughout the analysis, a Cone of Influence (COI) was introduced to mitigate edge effects at the ends of the finite-length time series. Only regions within the COI that passed the 95% confidence level test were subjected to interpretation.

4. Results

4.1. Field Deformation Characteristics

Field investigations and UAV photogrammetry reveal intense overall surface deformation of the Pangcun landslide. Numerous cracks are distributed across the landslide’s front, middle, and rear sections, accompanied by localized slumping, indicating continuous sliding. At the rear edge of the landslide, prominent longitudinal tension-settlement cracks are visible, with a maximum vertical displacement of approximately 7 m and an average displacement of 2 m (Figure 6). These rear-edge cracks are essentially continuous, extending for about 400.0 m with a zigzag planar morphology, forming the posterior boundary of the landslide (Figure 6c). This settlement crack was generated by the overall downslope sliding of the landslide mass; consequently, the exposed landslide headwall is relatively smooth and planar, with a dip angle of approximately 47°. On the surface of the landslide headwall, multiple straight, parallel linear striations (oriented at 15°) are visible due to sliding friction, alongside distinct steps on the sliding wall. These features indicate that the landslide undergoes staged, non-uniform deformation movements with variable strain rates. The bedrock behind the landslide headwall (Figure 6d) is fragmented but shows no obvious signs of active deformation.

Overall, the Pangcun landslide body exhibits negative relief (a concave landform). Both the left and right boundaries are defined by the contact interface between bedrock ridges and the accumulation layer, with gullies developed along these interfaces. The deformation at the left boundary is notably more intense than that on the right. The left boundary cracks extend from the rear toward the front along the base of the left ridge (Figure 7a), reaching the middle section of the landslide boundary with an extension length of approximately 98 m. In contrast, the right boundary cracks develop only at the base of the ridge along the landslide’s rear edge (Figure 7b) and extend for about 25 m. The developmental status of these lateral boundaries indicates that deformation is greater in the western (left) portion of the landslide than in the eastern (right) portion, suggesting lower stability in the left sector.

Overall, the spatial heterogeneity of deformation within the Pangcun landslide is significant. The landslide can be divided into two secondary zones: Sub-landslide A (Zone A) and Sub-landslide B (Zone B). The boundary between the two zones is defined by Gully No. 3, a major through-going gully in the central part of the landslide body that naturally separates the eastern and western sectors. This division is supported by differences in slope extent, proximity to the national highway, land use, and the spatial distribution of deformation features, with the western sector showing more pronounced cracking and fissuring than the eastern sector (Figure 1c).

In Zone A, a clear undulating (wave-like) rear-edge scarp is visible. Multiple sets of tension cracks (Figure 8a) are developed near gully #C2, accompanied by a vertical settlement of approximately 0.3 m. In the middle section of Zone A, a small number of arcuate tension cracks (Figure 8b) are present on the slope surface, with lengths of about 50 m and widths of 0.2 m. The geomorphology at the front edge of Zone A is highly fragmented (Figure 8c), characterized by several deeply incised “V-shaped” gullies and longitudinal cracks (approximately 20 m in length). Localized collapsing and slumping phenomena are particularly significant near the #C2 area.

In Zone B, tension cracks are widely distributed across the rear edge of the landslide (Figure 9a). Most of these cracks exhibit an arcuate morphology curving toward the sliding direction, with individual lengths ranging from 5 to 15 m and an average depth of approximately 0.5 m, indicating significant tensile stress at the rear boundary. Under this tensile stress, a graben (pull-apart trough) has developed at an elevation of 3423 m (Figure 9b), covering approximately 25 m². The central subsidence depth is about 1.5 m, with crack widths ranging from 0.1 to 0.5 m. The overall morphology of this feature is spider-web-like, characterized by “central collapse with peripheral radial distribution.” The middle section of the landslide features relatively steep topography, where cracks primarily manifest as longitudinal tension fissures (Figure 9c), with widths of about 0.3 m and lengths ranging from 15 to 35 m. The most intense deformation occurs at the front edge of the landslide, particularly at the shear outlet, where sliding failure has already occurred (Figure 9d). The failure zone measures approximately 150 m in width and 50 m in length, covering an area of about 750 m². This movement has caused severe damage to the G219 National Highway pavement and the overturning of the shoulder retaining walls (Figure 9e). On the landslide crown of this frontal failure, radial tension cracks (Figure 9f) are distributed with lengths varying from 3.0 to 9.0 m.

4.2. Spatiotemporal Deformation Characteristics

Figure 10 presents the annual deformation rate and cumulative deformation of the Pangcun landslide in the LOS direction from 2017 to 2024. The monitoring results reveal significant spatial heterogeneity in the surface deformation of the landslide body. Except for the frontal edge, which exhibits uplift deformation, the remaining areas are characterized by subsidence. The deformation magnitude in Zone B is greater than that in Zone A. The annual average line-of-sight (LOS) deformation rate of the landslide ranges from −15.0 mm/a to 6.0 mm/a. Specifically, the maximum negative LOS deformation (manifested as subsidence or downslope movement away from the satellite) is located in the middle–front section of Zone B, with an annual average rate reaching −15.0 mm/a and a maximum cumulative displacement of −87.3 mm. Conversely, the maximum positive LOS deformation (manifested as uplift or movement toward the satellite) is located on the right side of the frontal edge in Zone B, with a maximum rate of 6.0 mm/a and a maximum cumulative displacement of 33.6 mm.

Figure 11 illustrates the semi-annual incremental line-of-sight (LOS) deformation maps from March 2017 to June 2024. Overall, the Pangcun landslide exhibits continuous and distinct deformation signals. In contrast, the areas outside the landslide boundary (the surrounding background region) are predominantly light blue (−1.4 to 0.8 mm/a), indicating minimal variations. This demonstrates that the InSAR inversion results possess excellent spatial consistency and that the landslide deformation is subject to clear boundary constraints. The landslide as a whole exhibits long-term, progressive, cumulative deformation rather than the planar diffusion characteristics of a sudden global failure, conforming to the typical slow-creeping evolutionary pattern of accumulation landslides.

Regarding the spatial distribution in each monitoring period, the high-deformation zone consistently appears in the middle section of the landslide, particularly in the central units of Zone B, maintaining a stable spatial location and presenting a continuous yellow-orange anomalous area (−15.0 to −9.1 mm/a). The rear edge of Zone B exhibits weak-to-moderate anomalies during certain periods (−7.4 to −5.3 mm/a), but its overall intensity is lower than that of the middle section. Although the frontal units of Zone B are located at the lower part of the landslide, their deformation signals remain weak over the long term, failing to exhibit the typical spatial pattern of strong slope-toe extrusion or significant frontal acceleration. These characteristics indicate that the landslide deformation possesses significant spatial heterogeneity. The middle section of the landslide acts as the core area of long-term activity, and the primary deformation is likely dominated by internal shear creep within this central region.

Chronologically, from July 2017 to December 2019, the high-deformation zone initially emerged within the landslide, but its intensity was relatively limited; the maximum deformation remained below 27.0 mm. During the period from July 2020 to December 2021, the yellow-orange areas in the middle of the landslide showed a significant intensification (the maximum incremental deformation increased sharply to 12.1 mm), indicating accelerated deformation or an increased cumulative rate during this stage. From June 2022 to June 2024, the high-deformation zone persisted and further strengthened. Notably, a more pronounced red concentrated area was identified in June 2024, where the maximum cumulative deformation reached −87.3 mm. This demonstrates an obvious characteristic of “localized continuous intensification” rather than uniform deformation across the entire landslide body. Overall, the InSAR analysis results are highly consistent with the spatial distribution of landslide deformation observed during the field investigations.

To further reveal the spatial heterogeneity of deformation, a longitudinal profile (1–1′) was established along the primary sliding direction in Zone B. The cumulative deformation of eight characteristic points along this profile from March 2017 to June 2024 was extracted (Figure 10b). The InSAR time-series results for the Pangcun landslide indicate an overall background of continuous, slow deformation from 2017 to 2024. Yet, the deformation exhibits distinct episodic acceleration characteristics. Starting in 2018, the negative cumulative deformation deepened significantly, with the cumulative deformation increasing rapidly from −2.5 mm at the beginning of the year to −23.2 mm by the end of the year, followed by long-term accumulation superimposed on short-term fluctuations. In the first half of 2023, a relatively consistent “step-like” deepening process occurred. Furthermore, in the spring of 2024 (approximately March to April), the strongest short-term acceleration pulse in the entire sequence occurred. This reflects a time-series structure where long-term creep coexists with episodic responses driven by hydro-meteorological processes.

Spatially, the eight monitoring points exhibit a significantly nonuniform deformation distribution along the sliding direction (from the frontal Point 1 to the rear Point 8). The cumulative deformation is strongest in the middle of the landslide and weaker at both ends. As of 7 June 2024, Point 3 (approximately −71.7 mm) and Point 4 (approximately −58.1 mm) represent the primary zones of deformation concentration, followed by Points 5 and 7. In contrast, the cumulative deformation at the frontal Point 1 (approximately −6.2 mm) and the rear Point 8 (approximately −16.0 mm) is notably smaller. This forms a “central-peak” profile, indicating that the main sliding zone, or primary shear deformation band, is more likely located in the middle section of the landslide. That coupled acceleration is more pronounced toward the middle–lower and rear sections during specific periods. This spatial pattern is consistent with the deformation zoning characteristics controlled by a water-rich weak zone in the middle of an accumulation landslide. Overall, the landslide can be characterized by a typical spatiotemporal evolution model of “long-term creep + episodic acceleration in the central primary-control deformation zone.”

Field investigations revealed severe collapse at the front edge (Figure 1), whereas InSAR shows relatively weak signals in this area (Figure 10, Figure 11 and Figure 12). This discrepancy is mainly attributed to phase decorrelation and geometric distortions (e.g., shadow and layover) caused by rapid, large-gradient deformation, which limit the detection capability of InSAR. In addition, although the cumulative deformation shows an overall increasing trend, minor short-term fluctuations are observed (Figure 12), likely related to atmospheric effects and temporal variations in surface coherence.

4.3. Deformation Response Characteristics to Meteorological Factors

Based on the aforementioned spatiotemporal deformation analysis, the Pangcun landslide exhibits pronounced spatial heterogeneity in its deformation. To comprehensively investigate the driving mechanisms, monitoring points 3, 6, and 9 (their spatial distribution is shown in Figure 13) were selected as representative characteristic points across different structural zones (Zones A and B). Wavelet coherence analysis was employed to quantitatively characterize the time–frequency correlation between landslide displacement, precipitation, and temperature.

Figure 14 presents the wavelet coherence analysis results between cumulative displacement and meteorological factors. The coherence coefficient ranges from 0 to 1, with colors from blue to red indicating weak to strong coherence.

As observed in the displacement-precipitation coherence spectrum, the high- and low-coherence regions are unevenly distributed across time and period scales, indicating that the correlation between displacement and precipitation exhibits clear non-stationarity and intermittency.

Specifically, high coherence (reddish areas) is primarily concentrated in two periods: the 3-month scale from March 2017 to May 2018 and the 3-month scale from January 2022 to June 2024. This indicates that precipitation primarily affects landslide deformation during the spring-to-summer transition, with a characteristic time lag of approximately 3 months. During other periods, landslide deformation may be predominantly controlled by factors other than precipitation, rendering precipitation’s contribution relatively weak.

In contrast, the overall coherence in the displacement–temperature coherence spectrum is significantly lower than that of displacement–precipitation. Low-to-moderate coherence regions (coefficients of 0.5–0.7) only appear during limited periods, such as the 3-month scale from 2017 to 2018 and the approximately 3-month scale in 2023. For most of the time series, the coherence coefficient remains below 0.4, persistently low, especially from 2019 to 2022. This demonstrates that the coupling relationship between temperature and landslide displacement is generally weak and lacks a stable, continuous dominant period.

5. Discussion

5.1. Influence of External Triggering Factors on Temporal Deformation

Precipitation: The displacement of the Pangcun landslide shows a significant coupling with precipitation at the monthly scale, with an approximate 3-month lag. This lagged response suggests that deformation is controlled by time-dependent hydrological processes, such as infiltration, groundwater recharge, and pore-pressure diffusion, rather than by immediate rainfall triggering. Notably, the observed lag is interpreted as an indicator of delayed hydrological response, rather than definitive evidence of the underlying mechanism.

Temperature: The correlation between temperature and displacement is weak and unstable. The study area experiences large seasonal and diurnal temperature variations, which promote evaporation, drying–wetting cycles, and potentially freeze–thaw processes. These processes are inferred to mainly affect the long-term physical properties of slope materials rather than directly controlling short-term deformation.

Irrigation and canal leakage: A canal traverses the middle portion of the landslide (Figure 15a), and agricultural terraces (Figure 15d) at the middle–front slope are cultivated with seasonal irrigation (Figure 15b), mainly during March–April. Field observations indicate seepage at the landslide front (Figure 15c), suggesting relatively abundant groundwater conditions. The irrigation period is temporally distinct from the main rainfall season (June–August), thereby distinguishing anthropogenic from climatic water inputs. These observations suggest that irrigation and potential canal leakage help maintain elevated soil moisture and groundwater levels, thereby influencing the slope’s hydrological state. This temporal coupling between artificial recharge and natural climatic cycles explains why the landslide remains active even during periods of moderate rainfall.

Anthropogenic land-use change (village relocation and agricultural activity): From Figure 12, the landslide deformation rate decreased significantly between 2022 and 2023, followed by a renewed increase after 2023. This variation may be related to land-use changes caused by the relocation of Pangcun residents. The village, originally located at the landslide toe, was relocated between 2021 and 2022, as indicated by the satellite imagery in Figure 16, which reduced human activities and surface loading in this area. In addition, field investigations reveal a decline in agricultural activities within the landslide area, particularly in Zone B (Figure 15a,d). This is partly due to ongoing deformation, which makes it difficult to retain irrigation water and reduces the suitability for cultivation. The reduction in residential and agricultural activities may have temporarily reduced anthropogenic water inputs and surface disturbance; however, it did not alter the landslide’s long-term evolutionary trend.

Road excavation: The G219 National Highway runs along the toe of the landslide, where road excavation and associated engineering works have modified the original slope geometry. Such activities may have altered local boundary conditions, induced toe unloading, and affected stress redistribution and drainage conditions. These engineering modifications represent a persistent anthropogenic influence on the stability of the slope toe.

Seismic activity: Two earthquakes occurred in the region during the study period, including an Ms5.3 event in 2018 (epicentral distance ~23 km) and an Ms5.6 event on July 19, 2019 (epicentral distance ~72 km). To evaluate their potential influence on landslide deformation, a comparative analysis of the InSAR time series before and after the 2019 earthquake was conducted. The results show that the deformation rate in Zone B increased following the seismic event. For example, the annual deformation velocity at monitoring point P3 accelerated from −7.7 mm/yr (July 2019) to −21.0 mm/yr (July 2020). Although the landslide is located at a considerable distance from the epicenter, this temporal correspondence suggests that seismic shaking may have enhanced rock mass fragmentation and permeability, thereby increasing the landslide’s sensitivity to subsequent hydrological forcing.

Overall, the temporal deformation of the Pangcun landslide is primarily controlled by precipitation. At the same time, irrigation and canal leakage, anthropogenic land-use changes (including village relocation and reduced agricultural activity), road excavation, and seismic events jointly modulate the slope’s hydrological and mechanical conditions. These interacting factors result in lagged, episodic, and spatially heterogeneous deformation behavior.

5.2. Mechanism Interpretation Based on Geological and Field Evidence

The Pangcun landslide exhibits a typical progressive deformation pattern under strong structural control. Field evidence shows multiple rear-edge scarps and step-like features, indicating repeated episodic sliding. At the same time, the highly fractured and heterogeneous middle section corresponds to the main deformation zone identified by InSAR. At the slope toe, intensive slumping, seepage, and infrastructure damage reflect long-term weakening and active deformation transfer toward the front edge.

The deformation patterns obtained in this study are generally consistent with previous results (Xu et al., 2021 [21]; Jiang et al., 2024 [22]), particularly regarding deformation concentration and long-term continuous movement. Xu et al. (2021) [21] demonstrated that SBAS-InSAR results are in good agreement with GNSS and field monitoring data, thereby supporting the reliability of InSAR-derived deformation estimates. Therefore, although direct in situ monitoring data are limited in this study, the consistency with previously validated results provides indirect support for the accuracy of our analysis.

Rather than being controlled by a single triggering factor, the deformation of the Pangcun landslide is governed by a lag-controlled hydrological mechanism under multi-source recharge conditions. The ~3-month delay between precipitation and displacement indicates that deformation is regulated by pore-pressure diffusion rather than direct rainfall triggering. Importantly, irrigation and canal leakage provide additional water inputs that are temporally independent of natural precipitation, thereby maintaining a persistently elevated groundwater background. This distinguishes the Pangcun landslide from rainfall-dominated systems and explains its continuous creep and episodic acceleration behavior. The deformation observed during both rainy and non-rainy seasons is consistent with the findings of Jiang et al. (2024) [22], who suggested that the landslide has entered a stage of continuous deformation. In this context, long-term deformation reflects the landslide’s intrinsic evolutionary trend, while precipitation serves as a modulating factor.

Human activities further modify this mechanism by simultaneously affecting both hydrological and mechanical conditions. Changes in land use, including village relocation and reduced agricultural activity, alter surface loading and decrease anthropogenic water inputs, resulting in a temporary reduction in deformation rates. In contrast, road excavation at the slope toe promotes unloading and facilitates the propagation of deformation. These observations indicate that human activities in Pangcun not only serve as external disturbances but also regulate the landslide’s internal hydrological state.

Seismic events play an amplifying role by enhancing material fragmentation and permeability, thereby increasing the landslide’s sensitivity to subsequent hydrological forcing. This effect does not directly trigger displacement but contributes to the transition from slow creep to episodic acceleration.

Compared with previous studies [12,21,22], this work emphasizes the coupled role of lagged hydrological response and anthropogenic modulation, rather than treating precipitation as the sole driving factor. The integration of long-term InSAR observations (2017–2024) and field evidence reveals that a combined process of sustained water recharge, slope-toe weakening, and human-induced hydrological adjustment controls the Pangcun landslide.

In summary, the deformation of the Pangcun landslide is governed by a hydrologically driven, time-lagged, and human-modulated progressive failure mechanism, resulting in long-term creep, intermittent acceleration, and strong spatial heterogeneity.

6. Conclusions

(1)

SBAS-InSAR results from 2017 to 2024 show that the Pangcun giant accumulation landslide is undergoing long-term slow deformation with strong spatial heterogeneity, and the most active deformation is concentrated in Zone B, which represents the principal deformation zone.

(2)

UAV imagery and field investigations reveal typical deformation features, including rear-edge tension cracks, middle-section fissures and scarps, and frontal slumping. These features are consistent with the InSAR-derived deformation pattern, confirming the reliability of the monitoring results.

(3)

Wavelet coherence analysis reveals a non-stationary and intermittent relationship between displacement and precipitation, with a characteristic lag of approximately 3 months, indicating that landslide deformation is governed by delayed hydrological processes rather than direct rainfall triggering.

(4)

The correlation between displacement and temperature is weak and unstable, suggesting that temperature does not directly control short-term deformation but may exert an indirect influence through long-term processes such as freeze–thaw effects.

(5)

A lagged hydrological mechanism under multi-source water recharge conditions controls the Pangcun landslide. In addition to precipitation, irrigation and canal-related water inputs play an important role in maintaining a high groundwater background. At the same time, road excavation, land-use change, and seismic disturbances further modify the mechanical and hydrological conditions. This coupling results in long-term creep with episodic acceleration and strong spatial variability.