Deformation and failure mechanisms of deep-seated landslides: insights from Guili-Baige in Jinsha River tectonic belt

Abstract

The Jinsha River tectonic mélange belt, marked by complex geological structures and large-scale deep-seated landslides, presents significant challenges in understanding deformation and failure mechanisms. This study focuses on the Guili-Baige section, characterized by deep river incision and uneven rainfall distribution. Field surveys, remote sensing, UAV photogrammetry, SBAS-InSAR monitoring, mineral composition analysis, SEM, and geotechnical testing were employed to examine deformation characteristics and distribution patterns of large-scale landslides. A thick, continuous sliding zone in the Guili landslide was identified, with a total volume of 6.55 × 10⁷ m³, revealing prone geological structures and progressive failure mechanisms. SBAS-InSAR monitoring indicated creep deformation with localized acceleration, reaching a maximum surface deformation rate of −24.29 mm/yr. Mechanical tests showed that water content significantly reduces soil cohesion in sliding zones. Geological structures, lithology, concentrated rainfall, and strong river erosion was identified as primary deformation controls. These findings enhance understanding of deformation mechanisms in mélange zones globally, emphasizing the impacts of extreme rainfall and erosion events.

Introduction

The upper reaches of the Jinsha River are located at the eastern edge of the Tibetan Plateau, within the Jinsha River tectonic mélange belt, characterized by intense fault activity, complex material composition and geological structures, as well as significant variations in engineering geological properties^1,2,3. Influenced by intense tectonic activity, dramatic crustal uplift, and rapid river incision, the region exhibits extremely complex topography and geomorphology, with numerous high mountains and deep valleys, and elevation differences ranging from 1000 to 3000 m. Under the combined influence of internal and external dynamic forces, a series of large to deep-seated landslides have developed along both banks of the main stream of the Jinsha River⁴. Historically, multiple landslide-induced river blocking events have occurred, including recent landslides such as the Baige landslide, blocked the Jinsha River, forming a landslide-dammed lake and creating a landslide-river blocking-dammed lake-flood disaster chain⁵, as well as a series of large deep-seated ancient landslides like the Woda landslide⁶, Chashushan landslide⁷, Temi landslide, and Xiongba landslide⁸. Some of these landslides are undergoing intense deformation, while others have caused severe “landslide-river blocking-dam break” disaster chains, characterized by large scale, complex causative mechanisms, and extremely serious hazards.

The study of creep deformation mechanisms and progressive failure mechanisms of large deep-seated landslides in tectonic mélange belts has received increasing attention from scholars worldwide. Previous research has extensively examined factors such as reservoir impoundment, rainfall, river erosion, engineering excavation, and tectonic activity with lithological control on the creep deformation mechanisms of these landslides. Schuster and Highland⁹ highlighted that the large-scale landslides globally, such as the Vajont Dam landslide in Italy (1963) and the Oso landslide in Washington State, USA (2014), are often associated with severe human casualties, economic losses, and ecological disruption. These catastrophic events not only reveal the destructive characteristics of large, deep-seated landslides but also emphasize the critical need for understanding their mechanisms and assessing associated disaster risks. However, the current understanding of the identification and deformation mechanisms of large deep-seated landslides in tectonic mélange belts remains inconsistent due to technical limitations.

This study employs techniques such as remote sensing interpretation, UAV aerial surveys, field geological investigations, time-series InSAR deformation monitoring, mineral composition analysis, scanning electron microscopy, and rock-soil mechanical testing to investigate the developmental characteristics of large deep-seated landslides in the Guili-Baige Section of the upper Jinsha River. It analyzes the temporal deformation patterns of typical landslides and identifies the factors influencing creep deformation. The Guili landslide, located approximately 5 km upstream of the Baige landslide, has a scarp overall slope and significant sliding mass thickness. The clay-altered rocks formed by intense tectonic activity in the mélange belt provide the material basis for the development of the Guili landslide. Influenced by short-term concentrated rainfall and strong river erosion, the strength of the sliding mass at the landslide’s front edge decreases, accelerating creep deformation and posing a risk of further instability and failure, which severely threatens the safety of residents’ lives and property both upstream and downstream. This study indicates the influence mechanisms of tectonic activity, rainfall, and river erosion on the creep deformation of large deep-seated landslides from the perspective of geological structures and specific lithological control. These findings can provide valuable references for the early warning and prevention of similar types of landslides.

The Guili-Baige Section in the upper reaches of the Jinsha River is located at the southeastern edge of the Tibetan Plateau (Fig. 1). This region exhibits complex geological tectonic evolution, resulting in the formation of multiple tectonic mélange belts such as the Jiali-Yigong, Nujiang, Jinsha River, and Ganzi-Litang mélange belts. These features contribute to the complex and unique geological conditions of the upper Jinsha River region. The area is characterized by the development of branch faults of the Jinsha River, intense tectonic activity, and regional intermittent crustal uplift accompanied by strong horizontal compression and frequent faulting^10,11. The deeply incised Jinsha River has created steep topography on both sides of the valley, typifying tectonic erosional and fluvial erosional landscapes¹². The Guili-Baige Section of the upper Jinsha River falls within a high-altitude cold temperate semi-humid climate zone, with an average annual precipitation of approximately 800 mm. In recent years, the maximum annual precipitation reached 919.2 mm, with uneven distribution throughout the year. The maximum daily precipitation recorded is 41.4 mm, and the period from May to September accounts for about 86.7% of the annual precipitation, with occasional extreme rainfall events.

Fig. 1: Geological structure and landslide distribution map of the Guili-Baige section of the upper reaches of the Jinsha River.

a Geographical location map of the study area; b Geological structure and stratigraphic lithology map; c Remote sensing images of Xiaomojiu landslide; d Development characteristics of Baige landslide.

The main exposed strata in the Guili-Baige Section belong to the Gangtuo Formation, consisting of schist, slate, and phyllite, which are relatively weak substrates. These are intermixed with blocks of ultramafic rocks, gabbro, basalt, and marble. Influenced by the rapid uplift of the Tibetan Plateau and the intense tectonic activity of the mélange belt, the rocks in the Guili-Baige Section exhibit significant foliation, with particularly fractured rock masses along the Jinsha River. Some rock masses are characterized by weakened properties due to active faults, hydrothermal processes, and lithological control⁷. According to the Chinese Seismic Ground Motion Parameter Zonation Map (GB 18306-2015), the landslide area is located in the Zone VII seismic intensity region, indicating strong seismic activity. Historical earthquakes, such as the 1870 Batang Ms 7¼ earthquake, the 1989 Batang Ms 6.7 earthquake, and the 2006 Baiyu County Zhangdu Township Ms 4.5 earthquake, have significantly impacted the structural integrity and stability of the landslide rock masses in the study area¹³.

Results

Development and deformation characteristics of typical landslides in the Guili-Baige Section

Surface Deformation Results of Landslides: Herrera et al.¹⁴ suggested that in the processing and analysis of SAR radar data, a slope can be considered to have poor stability when the absolute value of the deformation rate in the radar line-of-sight (LOS) direction exceeds specific thresholds: >21 mm/yr for L-band radar, >14 mm/yr for C-band radar, and >6 mm/yr for X-band radar. Since the Sentinel data used in this study is C-band, a LOS deformation rate exceeding 14 mm/yr indicates that the landslide is in a stage of poor stability.

1. Yanshan Series Geposuper Unit; 2. Upper Triassic Waqiu Formation; 3. Upper Triassic Jingu Formation; 4. Proterozoic Xiongsong Group; 5. Lower Triassic Songduo Formation; 6. Upper Triassic Songduo Formation; 7. Variscan Gabbro; 8. Variscan Peridotite; 9. Triassic Adula Formation; 10. Triassic Bolila Formation; 11. Yanshan Series Zebasuper Unit; 12. Triassic Jiapila Formation; 13. Middle Carboniferous Shengpa Group; 14. Fault; 15. Drainage System; 16. Typical Landslide; 17. Residential Area; 18. Sentinel-A Ascending and Descending Orbit Data.

This study identified and interpreted 17 typical landslides in the Guili-Baige Section of the Jinsha River (Fig. 2). From the figure, it is evident that the surface deformation patterns closely match the corresponding landslide geomorphological features. The boundaries derived from deformation rates align with those derived from optical remote sensing images based on terrain features. Additionally, for each landslide, the areas with the highest deformation rates correspond to the most active regions observed in the remote sensing images (Fig. 3).

Fig. 2: Typical landslide deformation technology results for the Guili-Baige section of the Jinsha River.

a Deformation rates result for the ascending orbit; b Deformation rates result for the descending orbit.

Fig. 3: Remote sensing characteristics of active landslides in the Guili-Baige section of the upper Jinsha River.

a InSAR deformation rates map for 8 typical landslides; b Corresponding optical remote sensing images for the 8 typical landslides (white solid lines indicate the landslide boundaries, optical remote sensing images sourced from Google Earth).

Developmental Characteristics of Active Landslides in the Guili-Baige Section: Through remote sensing interpretation, field investigations, and InSAR deformation analysis, 17 landslides have been identified along the 14 km stretch of the Guili-Baige section of the Jinsha River (Table 1, Fig. 4). Based on SBAS-InSAR deformation analysis, it is observed that 8 of these 17 landslides are undergoing creep deformation, while the remaining 9 landslides show less deformation. Figure 2 illustrates the annual average deformation rate in the Line of Sight (LOS) direction from Sentinel-1 SAR data for the Guili-Baige section of the Jinsha River. Positive values (blue) indicate landslides moving towards the sensor direction, while negative values (red) indicate landslides moving away from the sensor direction. Large landslides are unevenly distributed along both banks of the Jinsha River within the 14 km stretch of the Guili-Baige section, with 12 large landslides developing on the right bank and its tributaries, and 5 on the left bank and its tributaries. The western bank of the Jinsha River, located in the anticline flank, exhibits extensive development of large to giant landslides under the control of fold structures and long-term influences. Analysis of the annual average change rate in LOS direction (Fig. 2a, b) from ascending and descending orbits of Sentinel data reveals distinct deformation trends at the forefront of typical large deep-seated landslides like the Guili landslide and the Xiaomojiu landslide, indicating a risk of further instability and sliding.

Fig. 4: Engineering geological plan and development characteristics of Guili landslide.

a Engineering geological plan; b Schist entrapped basalt; c Bedrock exposure; d Sliding zone soil exposure characteristics.

Table 1 Overview of InSAR deformation rates for typical landslides in the Guili-Baige section of the Jinsha River basic parameters of SAR image data

Developmental and Deformation Characteristics of Guili Large-scale Deep-seated Ancient Landslide

Developmental Characteristics of the Landslide: Based on remote sensing interpretation and field geological investigation, the trailing edge of the Guili landslide is at an elevation of 3508 m, while the front edge is at 2868 m. The longitudinal length of the landslide is approximately 1728 m, with a width of about 1200 m at the trailing edge and 900 m at the front edge. Considering the micro-geomorphological conditions and deformation characteristics of the landslide, it is divided into the landslide trailing area (I) and the landslide accumulation area (II) on a plan view. The accumulation area (II) includes two heavily deformed zones (II₂, II₃) (Fig. 4).

The trailing edge of the Guili landslide exhibits numerous steep scarps without distinct sliding walls. Both sides are mainly composed of residual slope deposits, showing clear steep scarps and tension cracks. In some local areas, bedrock is exposed (Fig. 4c), predominantly composed of highly weathered slate, shale, with extensively developed joint fissures, resulting in blocky rock masses. The accumulation area of the landslide is primarily located in the middle and front parts of the Guili landslide. The surface layer consists mainly of Quaternary residual slope debris and residual soil, with a rock content of about 30–40%. The main rock types include slate, shale, and basalt (Fig. 4b). The area exhibits significant deformation features, primarily characterized by road surface cracks, steep scarps, and localized collapses. Numerous small gullies are distributed across the slope of the accumulation area, formed by deformation and surface water erosion. Steep scarps are prominent on the surface, ranging in height from 0.5 to 2.5 m and lengths from 5 to 25 m.

Based on field surveys, the cross-section of the Guili landslide exhibits a steep rear and a gentle front. The exposed bedrock is primarily composed of metamorphic rocks such as slate, schist, and gneiss. Additionally, basalt rock blocks are interspersed and exposed in the middle and rear parts of the landslide. The rock strata mainly exhibit an orientation of 51°∠70°, forming an overall reverse slope structure (Fig. 5). Due to the long-term tectonic stress from the tectonic mélange belt, the exposed bedrock shows a fractured-laminated structure, with significant folding and schistosity, highly developed joints and fractures, and severe alteration. The rock and soil at the rear edge of the landslide consist mainly of fractured rock and weathered gravelly soil, subjected to strong compression and evident metamorphism. The landslide material is predominantly Quaternary residual slope deposits, mainly composed of rubble and stone soil, with varying degrees of metamorphism. The weathering is intense, with gravel particle sizes ranging from 2 to 20 cm and boulder sizes ranging from 30 to 200 cm, exhibiting angular shapes. The landslide thickness is approximately 15 to 65 m, with an average thickness of about 35 m. The landslide volume is 6.55 × 10⁷ m³, classifying it as a large deep-seated landslide.

Fig. 5

A-A’ engineering geological section of Guili landslide.

Based on drilling data and field investigations, the depths of the deep-seated sliding zone are 64.02 m, 57.90 m, and 54.13 m, respectively. The sliding zone predominantly consists of brownish-yellow and dark gray silty clay, composed mainly of metamorphic sandstone, basalt, schist, and mica fragments, with a significantly lower gravel content than the landslide accumulation body. Soil samples from the sliding zone were taken at depths of 20–25 m (Fig. 5). These samples are primarily composed of altered clay, with particle sizes ranging from 2 to 10 mm, angular in shape, poorly rounded, and overall brownish-yellow with high moisture content. Considering the exposed bedrock at the landslide’s front edge, it is inferred that the shear outlet of the landslide is approximately 30 m above the Jinsha River level. Laboratory scanning electron microscope (SEM) analysis of sliding zones collected in the field revealed notable microstructural differences between various locations. In some areas, SEM observations showed wide fissures filled with blocky and needle-shaped clay minerals, numerous clay dissolution pores, and a tendency for clay minerals to form spherical aggregates (Fig. 6). In contrast, other areas exhibited clay minerals filling inter-particle voids in aggregated forms, resulting in a more fragmented and loosely packed structure (Fig. 6).

Fig. 6: Structural characteristics map of the Guili landslide sliding zone.

a, b The pictures show the characteristics of the sliding zone of the Guili landslide; c, d The pictures show the SEM microstructure of the soil in the sliding zone of the Guili landslide.

Deformation Characteristics of the Landslide: Field geological surveys indicate that the Guili landslide exhibits severe deformation, with numerous tension cracks and collapses covering almost the entire landslide body. The front edge of the landslide is located on the right bank of the Jinsha River, which is a zone of intense deformation. This area shows significant deformation over a wide range, primarily manifesting as tension cracks and collapses, along with the development of multiple deformation bodies. The tension cracks extend along the outer side of the village road with a trend of 122°, almost perpendicular to the direction of the landslide movement. The cracks are approximately 10–12 m long and 10–110 cm width, most of which are currently covered by rubble and soil. The offset escarpment (Fig. 7b) is about 35 m long, 35–60 cm wide, and has a vertical displacement of 30–55 cm. Due to the steep scarps and loose materials at the front edge of the landslide, rainfall and other factors have reduced the strength of the rock and soil, causing the slope to slide downward (Fig. 7c). In the middle of the landslide, local collapses occurred due to road construction (Fig. 7a), with a collapsed volume of approximately 300 m³. Road cutting and unloading in the middle part of the landslide have led to the development of slope surface tension cracks and local collapses. This, combined with the front edge traction deformation, has further expanded the tension cracks. The tension cracks in the middle section of the landslide (Fig. 7d) extend approximately 8–10 m and are 10–20 cm wide, with small offset escarpments locally, consistent with the direction of the cracks.

Fig. 7: Deformation and failure characteristics of Guili landslide.

a Local sliding in the middle part of the landslide; b Lower escarpment; c Reactivation deformation of the front edge; d Tensile cracks develop in the middle part of the landslide.

Material composition and mechanical strength properties of sliding zone soil

Studies have shown that clay minerals play a significant role in promoting large-scale landslides⁷. The combination of active faults and hydrothermal runoff in the central part of the Guili landslide has created favorable conditions for hydrothermal alteration of the rock mass. Field investigations reveal that the sliding zone soil, composed of altered clay, exhibits prominent slickensides and shear surfaces.

Mineral Composition Analysis of Sliding Zone Soil Based on XRD Testing: Samples of clay with particle sizes smaller than 0.075 mm were selected for quantitative determination of both whole rock minerals and clay minerals. The results indicate that the sliding zone soil (altered clay) near the Guili landslide has a high clay mineral content, primarily composed of Illite, Illite/Smectite mixed-layer (I/S) minerals, Kaolinite, and Chlorite, with Illite being the dominant component, followed by Illite/Smectite mixed-layer minerals.

The GL01-02 sample is predominantly composed of clay minerals, accounting for 37.47%. Specifically, it contains 61% Illite (It), 29% Illite/Smectite mixed-layer (I/S), 7% Chlorite (C), and 3% Kaolinite (K), with some occurrences of Illite/Smectite mixed layers making up about 35%. Additionally, there are signs of sodium metasomatism, with albite content at approximately 27.64%. The GL03 sample mainly consists of quartz and clay minerals, with clay minerals being 46% Illite (It) and 54% Illite/Smectite mixed-layer (I/S), and a high quartz content of 69.91%. The GL04 sample’s clay minerals are primarily 41% Illite (It), 52% Illite/Smectite mixed-layer (I/S), 5% Chlorite (C), and 2% Kaolinite (K), with some occurrences of Illite/Smectite mixed layers comprising about 35%. The GL05 sample mainly consists of quartz and clay minerals, with quartz content at 55.41% and clay minerals making up 25.71%, primarily composed of 60% Illite and 13% Chlorite (Tables 2 and 3, Fig. 8).

Fig. 8

XRD quantitative test results of the material composition of the sliding zone soil (GL01-02).

Table 2 X-ray diffraction whole rock mineral component analysis results

Table 3 X-ray diffraction clay mineral component analysis results

Mechanical Strength Properties of Sliding Zone Soil: Due to the swelling and shrinking properties and structural orientation of clay minerals, the cyclic action of water can reduce the shear strength and structural integrity of sliding zone soil. Clay minerals affect the strength of sliding zone soil in two ways: (1) Clay minerals with inherent expansibility, such as Illite, Montmorillonite, and Chlorite, form numerous cracks within the sliding zone as the moisture content changes, thereby reducing the strength of the sliding zone; (2) Clay minerals filling the spaces between coarse particles in the sliding zone soil decrease the friction between them, reducing the internal friction angle (φ). When the Illite content in the sliding zone soil is high, it easily swells upon contact with water, causing soil particles to align directionally, which is the decisive reason for the reduction in the drained shear strength of the sliding zone soil¹⁵.

This study mainly explores the water sensitivity of different clay mineral contents and their impact on the mechanical strength of the sliding zone soil. Considering the high clay mineral content in the sliding zone during the formation of the Guili landslide, undrained triaxial shear tests were conducted on sliding zone soil samples with different moisture contents (10%, 13%, 16%) using a GDS large automatic triaxial apparatus. The results show that the moisture content significantly affects the strength of the sliding zone soil, including both its friction strength and cohesion strength. The internal friction angle of the sliding zone soil is relatively low under different moisture contents, and cohesion decreases as the moisture content increases. When the moisture content of sliding zone soil GL01-02 increases from 10% to 16%, the cohesion decreases from 61.11 kPa to 12.54 kPa, a reduction of 70.8%. When the moisture content of sliding zone soil GL03 increases from 10% to 16%, the cohesion decreases from 25.98 kPa to 11.4 kPa, a reduction of 49.2%.

The clay mineral content in sliding zone soil GL01-02 is significantly higher than in GL03, which further indicates that the cementation between minerals in GL01-02 is stronger and the structure is denser. As the moisture content continues to increase, the cementation structure between clay minerals is destroyed, and cohesion significantly decreases. As the moisture content increases, the internal friction angle of the sliding zone soil GL01-02 gradually decreases from 21.11° to 15.89°, a reduction of 16.7%. For sliding zone soil GL03, the internal friction angle first increases and then decreases, reducing from 25.72° to 21.94°, a reduction of 10.6%. The difference in internal friction angles among soil samples with different moisture contents fluctuates within a range of ± 4° (Fig. 9).

Fig. 9: Shear strength evolution curves of sliding zone soil under different moisture content conditions.

a GL01-02 strength envelope; b GL03 strength envelope.

Analysis of InSAR deformation characteristics of the Guili landslide

InSAR Deformation Characteristics of the Guili Landslide: Due to the overall SE orientation of the Guili landslide and the terrain slope primarily ranging from 20° to 35°, ascending orbit data is suitable for analyzing surface deformation in the east, southeast, and northeast directions. Descending orbit data is not visible for slope deformation in these directions¹⁶. Therefore, the deformation analysis of the Guili landslide in this study uses the results from Sentinel-1A ascending orbit data. The deformation rate distribution map of the Guili landslide from January 2016 to February 2022 was obtained using SBAS-InSAR technology (Fig. 10). The results show that during the monitoring period, deformation mainly occurred in the central and left-front accumulation areas of the landslide, with deformation direction predominantly along the main sliding direction of the landslide. The maximum annual deformation rate reached −24.29 mm/yr.

Fig. 10

Deformation rate distribution map of the Guili landslide based on SBAS-InSAR technology.

To further analyze the deformation and failure mode of the Guili landslide, two cross-sections A-A’ and B-B’ were selected along the main sliding direction in the severely deformed area at the front edge of the landslide to plot the deformation rate profiles (Fig. 11). The V_LOS deformation rates are generally negative, mainly distributed at the landslide front edge. In the severely deformed area at 1.0–1.05 km of section II₂, there is a local uplift phenomenon (Fig. 11), possibly related to new intense deformation at the front edge, internal creep within the slope, leading to significant differential deformation across the slope. Field investigations revealed that this area experienced local collapses, tensile cracks, and offset escarpments. Considering the topographical conditions, geological structure, and spatial deformation characteristics of the landslide, it can be inferred that the Guili landslide involves multiple stages of sliding and is currently in a progressive creeping deformation stage, with deformation rates ranging from −24.29 to 15.09 mm/yr.

Fig. 11: Typical profile of the InSAR deformation rate of the Guili landslide.

a Sectional view of the deformation rate of the II2 deformation body; b Sectional view of the deformation rate of II3 deformation body.

Analysis on Deformation Trend of Guili Landslide: Based on the cumulative displacement time series diagram from different SAR images between January 2016 and February 2022 for the Guili landslide (Fig. 12), the results indicate that the deformation at the front edge of the Guili landslide continuously accumulated during the monitoring period from 2016 to 2022. The cumulative displacement of the Guili landslide increased gradually over time. The maximum displacement at the monitoring point increased from 0.4 mm on February 4, 2016, to 272.5 mm on February 20, 2022. During the 2016–2021, the maximum displacement increments at the monitoring point were 12.8 mm, 34.4 mm, 37.1 mm, 52.0 mm, 72.1 mm, and 64.1 mm, respectively. By February 20, 2022, the cumulative displacement reached 272.5 mm, showing a significant increasing trend compared to the cumulative displacement in January 2020. After that, the rate of displacement growth tended to stabilize.

Fig. 12: Cumulative deformation (LOS) of the Guili ancient landslide in different periods.

a 1 January, 2016; b 23 April, 2017; c 19 December, 2017; d 16 August, 2018; e 24 June, 2019; f 26 January, 2020; g 16 August, 2020; h 2 April, 2021; i 20 February, 2022.

Discussion

Effect of Rainfall on Deformation of Guili Landslide: Rainfall is one of the primary factors inducing landslide deformation and failure. Its mechanisms mainly include loading the landslide rock and soil mass and altering their structure and mechanical properties, thereby reducing the stability of the landslide body. For instance, the Jiangdingya large deep-seated landslide in Zhouqu County, located on the eastern edge of the Tibetan Plateau, was reactivated by heavy rainfall. The increased weight of the slope and the weakened mechanical strength of the rock and soil mass led to the reactivation of the ancient landslide¹⁷. Similarly, the Baige large deep-seated landslide in the upper reaches of the Jinsha River was affected by rainwater infiltration, which reduced the strength of the slope body. The slope was characterized by a high degree of rock fragmentation and the development of weak interlayers, which facilitated the Baige landslide and subsequent river blockage event.

The Guili landslide area is located within the Jinsha River tectonic mélange belt, formed by the subduction and collision of the Tethys Ocean^18,19. The strata are extremely complex and have undergone long-term tectonic evolution. The slope rock and soil mass develop along joints and fractures, often presenting as layered fractured structures. The clay minerals contained within these structures are highly water-sensitive, meaning their strength easily weakens upon contact with water. As the moisture content increases, the cementation between clay minerals is destroyed, leading to directional alignment and elongation, significantly reducing the friction coefficient and cohesion of the sliding zone²⁰. This plays a controlling role in the expansion, penetration, and transformation of potential sliding surfaces into sliding zones^21,22.

Field investigations have shown that there is significant groundwater exposure in the central part of the Guili landslide. Under the coupling effect of rainfall and groundwater, the engineering geological properties of the sliding zone soil containing clay minerals change. Especially under alternating wet and dry conditions, the cyclic swelling and shrinkage lead to further deterioration of the sliding zone soil structure and a further reduction in strength. Thus, the hydrophilic clay minerals and their interaction with rainfall are crucial factors contributing to the creeping deformation and progressive failure of the Guili landslide.

By analyzing the rainfall data from the Jiangda meteorological station and the InSAR deformation monitoring points (Fig. 13), it is evident that the Guili landslide was in a slow deformation phase before October 2018, with a cumulative deformation of less than 100 mm (Fig. 13a). From June 1, 2020, to October 1, 2020, the cumulative rainfall in the area exceeded 900 mm (Fig. 13b). After entering the rainy season in November 2020, the deformation rate of the landslide significantly increased, indicating an exacerbation of the landslide deformation. Monitoring point J1, located in the mid-front part of the landslide, exhibited the highest deformation, with a cumulative deformation reaching 230 mm. The deformation rates of monitoring points J1, J2, and J3 in the mid-front part of the landslide increased much more than those of monitoring points J4 and J5 in the upper part of the landslide. Compared to February 2019, the deformation amount and area had significantly expanded by February 2021, and the deformation rate had started to accelerate (Fig. 13a). This indicates that the Guili landslide, influenced by continuous rainfall, shows a clear trend of creeping deformation.

Fig. 13: Time series diagram of regional rainfall and cumulative deformation at the Guili landslide monitoring point (rainfall data according to Jiangda Weather Station).

a the relationship between the cumulative deformation and rainfall of the Guili landslide; b the daily rainfall and cumulative rainfall from June to September 2020; c the statistics and proportion of rainfall days.

Effect of River Erosion on Deformation of Guili Landslide: Rivers play a significant role in landslide stability during their formation and evolution through erosion processes. Riverbed erosion can undermine the rock and soil layers at the slope base, reducing slope support and creating steeper, exposed surfaces. This progression increases the likelihood of landslide instability, particularly in areas with high river flow or intense erosion. While often gradual, the cumulative impacts of this process may eventually trigger sudden slope failure. Rivers can also exert counter-pressure on slope toes, especially during periods of high water levels. The interplay between counter-pressure and erosion is dynamic, often creating a delay in landslide initiation. Over time, river erosion generally surpasses the stabilizing effects of counter-pressure, leading to eventual slope destabilization.

The upper reaches of the Jinsha River, where the Guili landslide is located, have experienced significant landslides such as the Suwalong landslide and the Xuelongnang landslide, both of which caused river blockages. According to regional river erosion and crustal uplift data, the river erosion rate in the upper reaches of the Jinsha River is relatively high^23,24. During periods of concentrated precipitation, the river flow increases significantly, resulting in strong scouring and erosion at the toe of the landslide’s front slope. This increases the exposure of the slope face, causing tension cracks on the slope surface to expand continuously, thereby reducing the stability of the landslide. In October and November 2018, the Baige landslide triggered two landslide-river blockage-dam breach disaster chains. The resulting landslide-dammed lake submerged the toe of the Guili landslide. After the dam breach, the rapid drop in water level caused intense short-term scouring and erosion of the front edge of the Guili landslide, leading to local deformation and sliding. Following the Baige landslide-river blockage and the formation of the landslide-dammed lake, the deformation rate and cumulative displacement of the Guili landslide showed a significant increasing trend (Fig. 13a). Additionally, nearby large deep-seated landslides, such as the Xiaomojiu landslide, also exhibit similar traction-type progressive failure mechanisms.

The geomorphological type of the landslide area is primarily river erosion landforms. The river erosion rate in the Baiyu section of the upper Jinsha River is relatively high^25,26. Using a terrain analysis toolkit, the steepness index of the upper Jinsha River was calculated²⁷. An analysis of the relationship between the steepness index of the Baiyu section of the Jinsha River and landslide distribution indicates that the higher the k_s value, the more frequent the landslides occur (Fig. 14b). The Guili landslide is located in a river section with high erosion rates (0.04-0.06 mm/yr) and a steepness index greater than 150 (Fig. 14). During periods of concentrated precipitation, the flow of the Jinsha River increases significantly, leading to strong scouring and erosion at the toe of the landslide’s front slope. The surface tension cracks on the slope continue to expand, reducing the stability of the Guili landslide and causing the deformation rate of the landslide to increase periodically. This indicates that the formation of the Guili landslide is closely related to the river erosion processes.

Fig. 14: Map of river erosion in the upper reaches of the Jinsha River.

a Erosion rate of Baiyu - Batang section of Jinsha River; b Channel steepness of Baiyu section of Jinsha River (Based on the revision of reference²³).

The Jinsha River tectonic mélange belt exhibits strong tectonic activity and severe surface erosion. Tectonic uplift and river erosion are closely related to the occurrence of large deep-seated landslides in the upper Jinsha River. Intense river erosion and short-term extreme river erosion events (such as landslide-dammed lakes and dam-break floods) can significantly increase the potential instability of landslides. In the upper reaches of the Jinsha River, many large to gigantic deep-seated landslides with similar geological backgrounds have developed, such as the Xiongba landslide⁸ and the Sela landslide. Therefore, special attention should be paid to the possibility of large-scale landslides and the associated disaster chains in river sections with strong tectonic uplift and high river erosion.

Large-scale Deep-seated Landslides in the Tectonic Mélange Belt Formation and Deformation Mechanism: A Case Study of the Guili Landslide: Fault activity and significant tectonic deformation within the Jinsha River tectonic zone can intensify the creeping behavior of deep-seated landslides^17,24. The Guili landslide located in the Jinsha River tectonic mélange belt, is characterized by extremely complex stratigraphy and long-term tectonic evolution, often presenting as layered fractured structures. The active fault in the middle of the slope is one of the five regional branch faults of the Jinsha River fault zone, showing significant Holocene activity. The long-term activity of this fault has reduced the integrity and quality of the rock mass, weakened its resistance to erosion, and lowered its stability²⁸. Repeated tectonic processes in active faults lead to structural damage and strength degradation in the faults and adjacent rock and soil masses, facilitating the development of these large landslides. Moreover, the influence zones of active faults often serve as critical boundary conditions, material source areas for landslide initiation, and focal points of external dynamic forces²⁹.

Rainfall infiltration and river erosion play crucial roles in the formation, evolution, and deformation processes of the Guili landslide. The two instances of instability, river blockage and dam breach of the Baige landslide, combined with heavy regional rainfall, caused the Jinsha River’s water level, flow, and velocity to increase. This led to intense scouring and erosion of the toe at the front edge of the Guili landslide, enlarging the exposed face of the slope and providing favorable topographic conditions for the landslide. Additionally, changes in the rock and soil structure of the slope caused the slope to be in an unloading state under gravitational forces, providing favorable stress conditions for instability and sliding. Consequently, the stability of the Guili landslide decreased, leading to subsidence and collapse at the front edge. The traction-type failure at the front also induced deformation in the middle and rear parts of the landslide, resulting in an overall traction-type deformation pattern.

The analysis of the material composition, slope structure, and geomorphological conditions of the Guili landslide indicates that its formation and creeping deformation are part of a multi-phase progressive deformation failure process. This can be divided into the following major stages (Fig. 15):

• (a) After experiencing plate collision and compression, the residual oceanic crust formed the Jinsha River tectonic mélange belt. Intense tectonic deformation and rapid regional crustal uplift provided the necessary material conditions for landslide occurrence. Rapid crustal uplift has significantly altered the surface stress field in the Guili section, promoting fault formation and rock mass fragmentation. These processes provide the essential material and structural conditions for the development of the Guili landslide. Furthermore, this compressive regime often leads to the generation of abundant fault gouge and clay minerals, which serve as key controlling factors during the initial creep stage of the landslide (Fig. 15a).

• (b) Continuous tectonic compression caused deformation, destruction, and dynamic metamorphism of the slope rock mass, gradually increasing topographical differences, and forming a “V” shaped canyon. Simultaneously, the rock mass of the slope unloaded due to exposure, leading to slope deformation and internal stress adjustment (Fig. 15b).

• (c) Continuous erosion and incision by the Jinsha River at the slope toe caused instability and downward sliding of the slope front, with multiple sliding characteristics, continuously reducing slope stability. River scouring progressively weakens the slope toe, acting as a key driver of instability at the landslide front. As the river cuts deeper, the landslide exhibits multi-stage sliding behavior, where the displacement of frontal material elevates shear stress on the upper slope, triggering further sliding. This dynamic illustrates the nonlinear intensification of river erosion’s impact on slope toe stability. With deeper erosion, the overall stability of the slope deteriorates at an accelerating rate, especially when coupled with rainfall infiltration, which amplifies the destabilizing effects of river incision (Fig. 15c).

• (d) Under the coupling effect of heavy rainfall infiltration and river erosion, the weak interlayers formed by fault gouge and clay minerals in the slope rock and soil mass accelerated slope deformation, gradually expanding the structural surfaces. Heavy rainfall serves as a major trigger for landslides, transitioning into the acceleration phase. Infiltrated rainwater raises pore water pressure, substantially weakening the shear strength of weak interlayers and undermining the structural integrity of the slope. Concurrently, river erosion exacerbates instability by deepening the slope toe and amplifying the traction effect on the landslide mass. Moreover, the shear rheology of weak interlayers under intense rainfall accelerates the propagation of structural planes, creating key pathways for the eventual full-scale sliding of the landslide (Fig. 15d).

• (e) Discontinuous surfaces gradually developed into sliding surfaces. The rock mass at the slope toe cut along the gently inclined structural surface, while the rear of the landslide fractured along the steeply inclined structural surface due to tension. As the lower rock mass slid, the upper rock mass collapsed, scraping the middle and lower surface rock and soil, significantly increasing the landslide volume and blocking the Jinsha River (Fig. 15e).

Fig. 15: Deformation model diagram of a large-scale deep landslide in the Guili-Baige section of the Jinsha River.

a Crustal extrusion and uplift process; b Strong earthquakes and faults cause disturbance to the slope; c River erosion causes the front accumulation to become unstable and slide; d Rainfall infiltration causes the slope front to A strong deformation area is formed; e the strong deformation area loses stability and slides and blocks the river channel; f the slope changes before and after instability sliding.

The Baige and Xiaomojiu landslides in the upper reaches of the Jinsha River near the Guili-Baige section exhibit similar deformation characteristics. Before the Baige landslide-river blockage event, the severely deformed area was heavily eroded by the Jinsha River. After sliding, dynamic shearing led to the formation of a river blockage disaster^30,31. The severely deformed area of the Guili landslide is located in the mid-front part of the slope, with the shear outlet approximately 30 m above the river surface, making it highly likely to experience traction sliding and block the Jinsha River, forming a landslide-dammed lake (Fig. 15f).

Methods

Landslide remote sensing interpretation and field investigations

UAV aerial photography and remote sensing technology offer significant advantages in landslide research due to their high efficiency, speed, and cost-effectiveness, particularly in challenging high mountain and canyon terrains. UAVs provide high-resolution imagery, allowing for precise identification of landslide boundaries, and multi-temporal analysis reveals the dynamic processes of landslide evolution. Remote sensing extends this capability by delivering large-scale monitoring data based on surface changes.

Traditional geological survey methods excel in providing detailed and direct geological insights, especially in understanding landslide mechanisms, material composition, and structural characteristics. These methods allow for a deeper exploration of landslide internal structures and detailed assessments of soil and rock types and slip planes.

The integration of UAV and remote sensing technologies with traditional geological surveys enables a synergistic approach, combining large-scale, rapid landslide identification with in-depth, localized verification and analysis. Thus, employing remote sensing as the primary tool, complemented by traditional methods for targeted validation, is a practical and effective research strategy.

This study builds on the analysis of existing geological structure, stratigraphy, lithology, and geological hazard survey data for the Guili-Baige Section in the upper reaches of the Jinsha River. We collected multiple sets of satellite imagery, including Spot and GeoEye-1, and used 1:50,000 topographic maps and 1:200,000 geological maps to conduct remote sensing interpretation, UAV mapping, and field investigation of the large deep-seated landslides in the Guili-Baige Section. Special attention was given to the development and deformation characteristics of the Guili and Baige landslides, focusing on the landslide head scarps and deposition areas.

Furthermore, we employed InSAR deformation monitoring techniques and analyzed the resulting deformation data. This was combined with regional rainfall data and rock-soil mechanical testing results to investigate the deformation and instability mechanisms of large deep-seated landslides in the area (Fig. 16).

Fig. 16

Deformation analysis process of large-scale deep-seated landslides in structural mélange belts.

Test on material composition and mechanical strength of sliding zone soil

Investigation indicates that the rock-soil masses of the Guili landslide exhibit significant foliation, characterized by weakened properties due to active faults, hydrothermal processes, and lithological control, with evident disturbance during sampling. To address this issue, we conducted tests using an Empyrean X-ray diffractometer at an ambient temperature of 25 °C and a relative humidity of 60%. Clay samples with particle sizes smaller than 0.075 mm were selected and processed using three methods: oriented slides prepared from suspensions, glycolated oriented slides, and oriented slides heated at 550 °C, to quantitatively determine the clay minerals.

Additionally, we performed triaxial consolidated undrained shear tests on remolded samples of the sliding zone soil under different moisture conditions. These tests were conducted using the GDS large automatic triaxial apparatus at the Ministry of Natural Resources Key Laboratory of Active Tectonics and Geological Safety. Shear strength tests were performed on sliding zone soil samples with moisture contents of 10%, 13%, and 16%. The samples measured 80.0 mm in height and 39.1 mm in diameter. The pressure control frame had a maximum range of 500 kN, and the strain-type load cell also had a maximum range of 500 kN. The displacement device, a digital displacement meter, had a range of 0–50 mm and a minimum resolution of 0.01 mm. The shear rate during the tests was 0.8 mm/min. Under the same moisture conditions, confining pressures of 100 kPa, 200 kPa, and 300 kPa were applied, with a maximum shear displacement of 20 mm.

SBAS-InSAR landslide deformation technology

SAR Data: The Synthetic Aperture Radar (SAR) satellites commonly used in geoscience analysis include Japan’s J-RES, ALOS, the European Space Agency’s (ESA) ERS1/2, Envisat, Sentinel-1, and Germany’s Terra SAR-X. The Sentinel-1 satellite, launched by ESA in 2014, features various polarization modes and operates in TOPS (Terrain Observation with Progressive Scans) mode. Due to its high spatial resolution (5 m × 20 m) and short revisit time (12 days), Sentinel-1 can accurately capture minor deformations associated with landslides³². This data has been widely used in landslide deformation analysis studies with considerable success^33,34. For this study, Sentinel-1 ascending and descending radar satellite imagery was selected, covering the period from January 2016 to February 2022 (Table 4).

Table 4 Basic parameters of SAR image data

SAR Data Processing and Error Analysis: We utilized the SBAS-InSAR (Small Baseline Subset Interferometric Synthetic Aperture Radar) technique to process and analyze the development and deformation characteristics of the landslides in the Guili-Baige Section of the upper Jinsha River. SBAS-InSAR is an interferometric SAR technique based on small baseline subsets³⁵, which uses multi-temporal radar data for surface deformation monitoring. This method provides high spatial resolution and low noise levels, making it suitable for studying surface deformations in complex terrain areas such as landslides³⁶. The SBAS-InSAR technique is advantageous in landslide deformation monitoring due to its high measurement accuracy and extensive coverage^35,37.

Generation of Small Baseline Interferograms: The SAR images underwent preprocessing steps such as radiometric, geometric, and Doppler corrections to minimize systematic errors and enhance interferogram quality. Images from January 11, 2016, and February 11, 2016, were chosen as reference images for precise alignment with the remaining dataset. The co-registered images were then conjugate-multiplied to produce interferograms representing phase differences between image pairs. To ensure high coherence and improve data quality, a thresholding method was used to select well-matched interferogram pairs for subsequent differential interferometric analysis.

Processing Differential Data: For the selected high-quality interferograms, flat-earth and topographic phase effects were removed to obtain differential phases. Adaptive filtering was applied to refine the differential interferograms, and a 12.5 m DEM was used to mask shadowed and overlapping areas. The Minimum Cost Flow (MCF) algorithm, combined with coherence masking, was employed to avoid unreliable low-coherence regions. The high-quality unwrapped phases were then modeled to generate differential phase maps, and atmospheric delays were mitigated using high- and low-pass filtering, resulting in unwrapped differential phases.

Deformation Time Series Analysis: Deformation rates were calculated using the Singular Value Decomposition (SVD) method and integrated over time to produce deformation maps. These maps reveal surface deformation patterns and provide critical insights into ground motion³⁴.

The study utilized the differential interferometry processor and interferometric point target analysis modules within the GAMMA software platform. Using the SBAS-InSAR processing framework, we conducted radar image interferometric pairing analysis on SAR raw images by setting thresholds for temporal and spatial baselines. For the registered data, we used a time baseline threshold of 48 days and a spatial baseline threshold of 200 meters (Fig. 17). The Sentinel-1A ascending track imagery data covered the period from January 11, 2016, to February 20, 2022, comprising a total of 144 scenes (Fig. 17a). The descending track imagery data covered the period from January 11, 2016, to February 15, 2022, with a total of 157 scenes (Fig. 17b). The data processing involved geocoding, differential interferometry, and phase unwrapping^38,39. After removing flat earth effects, topographic phase, and atmospheric phase errors, the line-of-sight (LOS) deformation rate field for the ascending and descending deformation rates in the Guili-Baige Section of the upper Jinsha River landslide study area was obtained.

Fig. 17: Temporal and spatial baseline diagram of sentinel-1 A data for the Guili-Baige section of the upper Jinsha River.

a Temporal and spatial baseline of ascending orbit Sentinel-1 A data; b Temporal and spatial baseline of descending orbit Sentinel-1 A data.

River channel steepness index analysis

The channel steepness index is often used to analyze the spatial distribution characteristics of regional tectonic activity. It can be calculated from river longitudinal profile data using the steady-state river flow erosion equation. By representing the morphological characteristics of the river longitudinal profile, it reveals uplift rates under fluvial erosion conditions. To facilitate calculations, a new fluvial erosion model has been further optimized and widely applied in tectonic geomorphology studies, as shown in Eq. 1:

$${K}_{s}={\left(\frac{U\left(x,t\right)}{K}\right)}^{\frac{1}{n}}$$

(1)

Among them, k_s is the channel steepness index, K is the erosion coefficient related to lithology, active faults, base level and climate, x is the location, time is t, and n is a constant.