The Identification and Influence Factor Analysis of Landslides Using SBAS-InSAR Technique: A Case Study of Hongya Village, China

Abstract

On 1 September 2022, a landslide in Hongya Village, Weiyuan Town, Huzhu Tu Autonomous County, Qinghai Province, caused significant casualties and economic losses. To mitigate such risks, InSAR technology is employed due to its wide coverage, all-weather operation, and cost-effectiveness in detecting landslides. In this study, focusing on the landslide in Hongya Village, SBAS-InSAR and Sentinel-1A satellite data from July 2021 to September/October 2022 were used to accurately identify the areas of active landslides and to analyze the landslide deformation trends, in combination with the geological characteristics of the landslides and rainfall data. The results showed that strong deformation was detected in the middle and back of the landslide in Hongya Village, with a maximum deformation rate of approximately -13 mm/year. The surface of the landslide consisted of mainly Upper Pleistocene wind-deposited loess, which is extremely sensitive to water. The deformation of the landslide was closely related to the rainfall, and the deformation of the landslide increased with the increase in rainfall. The research results prove that the combination of ascending and descending orbit data based on SBAS-InSAR technology is highly feasible in the field of landslide deformation monitoring and is of great practical significance for landslide disaster prevention and mitigation.

1. Introduction

Landslides are a common geological hazard in which soil or rock slides downhill along a slope, driven by gravity, under the influence of factors such as rain erosion and soaking [1]. Landslides are highly impactful geological hazards, characterized by their widespread occurrence and severe consequences, causing economic losses of nearly USD 10 billion and thousands of casualties each year in many countries around the world [2]. Landslides account for more than 70% of the total number of common geological hazards in China. In recent years, global climate change and harmful human activities have exacerbated the occurrence of landslide hazards, which pose a significant threat to the lives and property of residents in vulnerable areas [3]. In the face of the frequent landslide hazards in China, the construction of a monitoring, early warning, and forecasting system to mitigate or even prevent landslide-induced devastation is crucial [4]. The average number of successful geohazard forecasts in China between 2019 and 2021 was 796, preventing possibly 22,748 fatalities and direct economic losses of CNY 1.07 billion [5]. This indicates that successful forecasts of geological hazards, such as landslides, are effective in avoiding casualties and reducing the direct economic losses. However, accurate and comprehensive deformation monitoring data are the basis for successful forecasts. Therefore, the long-term monitoring of landslide deformation is crucial to understand the evolution of landslide hazards and to provide data support for the prevention and control of landslides [6].

Traditional methods for the monitoring of surface deformation include Global Positioning Systems (GPSs) and other elevation measurement approaches. However, these methods not only consume significant manpower and resources but also have a very limited measurement range; thus, they cannot meet the needs of large-scale landslide monitoring [7]. With the advancement of geographic information systems (GISs), remote sensing techniques have been widely used in monitoring and surveillance and have become an effective means of recognizing and understanding terrain deformation [8]. At present, the mainstream remote sensing techniques are categorized into optical remote sensing and radar remote sensing. Optical remote sensing can only provide the results of interpreted optical images, and it cannot reflect the specific values of ground deformation. Therefore, optical remote sensing is often used as a tool in identifying and interpreting topographic features, as well as determining the type of landslide, and it cannot provide strong evidence in the early warning stage [9,10,11]. InSAR-based radar remote sensing can be used to directly obtain terrain information and ground deformation information, which plays a significant role in hazard monitoring [12,13]. In 2018, Rudy et al. [14] used Radarsat-2 imagery from 2013 to 2015 to obtain surface deformation information for permafrost conditions via D-InSAR. The occurrence of large-scale landslide hazards is characterized by uncertainty and suddenness. In 2020, Abolghasem et al. [15] used the PS-InSAR technique to obtain surface subsidence data for Isfahan, Iran, in response to the ongoing drought events and groundwater overexploitation. They analyzed the causes of the outcomes in relation to geomorphological factors. In 2021, Han et al. [16] used the SBAS method to monitor the extent and rate of ground subsidence in the Yellow River Delta in 2018–2019, and they concluded that the most severe subsidence area was located in Xianhe Township, with an average annual subsidence rate of 204.7 mm/a. In 2021, Bao et al. [17] used PSI technology to monitor the long-term temporal deformation in the area around Dongting Lake. They also used a model considering thermal expansion and seasonal environmental factors to simulate the time-varying characteristics of the deformation and obtained the temporal deformation and thermal expansion parameters of the area. In 2024, Bayaraa et al. [18] investigated the representation of InSAR metadata as entity embeddings within a deep learning framework (EE-DL) to model the spatiotemporal deformation response. They successfully detected and predicted the fine spatial movement patterns at the Cardia Mining Proving Ground in Australia. Accurate early warning of the abnormal deformation caused by landslides requires rapid information acquisition and the analysis of the target, but conventional data acquisition and analysis methods have certain limitations and lags [19,20,21]. Consequently, it is of great scientific and practical value to study the integration of SBAS and InSAR in a landslide early warning model.

The InSAR technique is an advanced space-based Earth observation method that is extensively employed for the monitoring of surface deformation [22,23]. Compared to traditional methods, such as GPS measurements, the InSAR technique has the advantages of being all-day, all-weather, and wide-ranging and is unaffected by adverse weather conditions, such as clouds and rain. Moreover, it can acquire a wide area of ground deformity with precision up to the millimeter level [24,25]. However, the traditional InSAR technique also has some problems in its application. For example, the images are susceptible to spatial incoherence, temporal incoherence, and atmospheric delay effects [26,27,28]. Thus, on this basis, different time-series InSAR techniques have been developed which not only overcome these problems but also improve the precision of the monitoring results. Currently, the mainstream time-series InSAR monitoring methods include SBAS-InSAR and persistent scatterer interferometric synthetic aperture radar (PS-InSAR). SBAS-InSAR is mainly applied in the area of surface deformation monitoring [29,30]. Deformation monitoring is the measurement of a monitored object to determine its spatial location and internal morphological characteristics over time, which can be analyzed by obtaining the landslide rate before the landslide deformation and combining it with the historical landslide sites to build a better early warning model [31,32]. SBAS-InSAR is one of the most effective tools to obtain the deformation rate. In this study, based on ascending orbit data between July 2021 and September 2022 and descending orbit data between July 2021 and October 2022 from the Sentinel-1 satellite, we utilized the SBAS-InSAR technique to acquire the deformation rate of the Hongya Village landslide and analyze the relationship between the landslide and the fracture zone, stratigraphic lithology, and rainfall.

In recent years, the landslide in Hongya Village, Mutual County, Qinghai Province, has moved several times. The frequent movement of this landslide has posed a great threat to the lives and safety of the residents. In order to address the limitations of traditional monitoring methods, we apply the SBAS-InSAR technique, utilizing Sentinel-1 satellite data from both ascending and descending orbits spanning from July 2021 to September/October 2022, to investigate the deformation rates of the landslide in Hongya Village. Our primary focus is on elucidating the intricate interactions between the landslide dynamics, fault zones, stratigraphic lithology, and precipitation patterns. Specifically, this study aims to explore the correlations between the deformation characteristics of the Hongya Village landslide and the underlying geological features, with a particular emphasis on the stratigraphic lithology, as well as their interactions with rainfall. By integrating InSAR-derived deformation measurements with geological and meteorological data, this study seeks to provide a comprehensive understanding of the factors influencing landslide activity. The insights gained from this study are anticipated to contribute to the broader field of landslide studies and enhance our capacity for effective landslide hazard prediction and management.

2. Study Area

Hongya Village is located on the northwestern side of Weiyuan Town in the Tu Autonomous County of Huzhu. It is not only the county seat but also the political, economic, and cultural center of Tu Autonomous County in Huzhu. The study area lies in the transitional zone between the Tibetan Plateau and the Loess Plateau and, at the same time, in the transition section between the Darban Mountains and the Intermountain Basin. It has a complex topography, as shown in Figure 1. The precipitation within the study area is primarily governed by the southwestern monsoon; years characterized by a robust monsoon experience heightened precipitation, whereas years with a weaker monsoon see a corresponding decrease. The distribution of the annual precipitation in this region is notably uneven, with the majority occurring during the June–September period. During this time, the precipitation accounts for over 72% of the yearly total, with July and August being the most precipitation-rich months, contributing approximately 40% of the annual rainfall. The majority of this precipitation is manifested in the form of rainstorms and showers. Statistics indicate that the region experiences maximum daily rainfall of 59.0 mm, maximum hourly rainfall of 23.3 mm, and peak 10 min rainfall of 13.4 mm. Due to this concentrated precipitation, the area is highly susceptible to landslides [33]. The study area’s geomorphic unit constitutes a transitional zone bridging the low-hill region and the river valley plain. The stratigraphic lithology’s upper layer is composed of Upper Pleistocene wind-accumulated loess, typically characterized by porosity, looseness, and susceptibility to softening upon contact with water. In scenarios of heightened precipitation, either due to climate change or extreme weather events, the loess’ water content can surge rapidly, resulting in diminished shear strength and consequently elevating the risk of landslides [34]. The lower stratum consists of Neoproterozoic mudstone, which is generally characterized by lower strength, higher plasticity, and relatively inadequate landslide stability. In instances where the overlying loess layer becomes softened due to precipitation or groundwater infiltration, the underlying mudstone may fall short in providing adequate support, thereby intensifying the risk of landslides [35].

3. Methodology

3.1. Data

The Sentinel-1 satellite, which was launched by the European Space Agency in 2014, operates in a dual-satellite orbit with a C-band synthetic aperture radar (SAR). It offers frequent satellite imagery with a revisit period of 12 days and an extensive single-image overlay of 250 × 250 square kilometers. In this study, 37 ascending orbit InSAR wide (IW) swath images from July 2021 to September 2022 and 38 descending orbit IW swath images from July 2021 to October 2022 were utilized.

Table 1. The basic parameters of Sentinel-1 data for the study area.

Wavelength	Ascending/ Descending Orbit	Time	Revisit Cycle	Resolution	Angle of Incidence	Imaging Mode	Polarization Mode	Number of Images (Scenes)
5.6 cm/C Band	Ascending track	July 2021 to September 2022	12 Days	20 m	41.7°	IW	VV	37
5.6 cm/C Band	Descending track	July 2021 to October 2022	12 Days	20 m	41.5°	IW	VV	38

lists the fundamental parameters of the Sentinel-1 satellite.

In this study, the ALOS World 3D (AW3D 30) digital surface model (DSM) from JAXA was utilized as the external digital elevation model (DEM) to correct for terrain effects. Moreover, the precision orbit data (precise orbit ephemerides (PODs)) of the Sentinel-1 satellite were used to correct the orbital error.

ENVI software (ENVI 5.6) was applied for data processing. The image from 19 September 2021 served as the reference (super-master) image for the ascending orbit data, whereas the imagery from 12 September 2021 was used as the reference (super-master) image for the descending orbit data. The remaining images were used as auxiliary images to align with the super-master image. The temporal baseline for the ascending and descending orbit data was limited to 80 days, with spatial baselines of approximately 155 m and 145 m, respectively. The independent use of ascending and descending track data resulted in 179 and 181 interferometric pairs generated, respectively. Figure 2 illustrates the temporal and spatial baseline connections for the ascending and descending orbit data. In Figure 2, the green points depict valid data pairs, while the yellow points signify the super-master images. The X-axis corresponds to the time dimension, and the Y-axis indicates the extent of the spatial baseline separating each image from the respective super-master image.

3.2. Methodological Procedure

SBAS-InSAR is an innovative method of time-series analysis, introduced by Berardino et al. for the study of low-resolution and massive deformation [36]. The method utilizes short spatiotemporal baselines to generate time-series interferograms from multiple master images through flexible combinations. It employs singular value decomposition (SVD) for matrix inversion to derive the deformation sequences and average deformation rates of the study area over the observation period [37,38,39]. Regarding SBAS-InSAR processing, SVD is used to decompose the design matrix into components, allowing for the identification of significant singular values that contribute to the stable solution, while discarding small singular values that correspond to noise or redundant information. This helps in regularizing the inversion process and ensures that the deformation signal is effectively extracted. The application of the SBAS-InSAR technique includes the following steps, as illustrated in Figure 3 [40,41,42].

(1)

$N+1$ SAR images of the same region arranged chronologically were obtained. One image was selected as the reference for alignment. The remaining SAR imagery were aligned to the reference image by setting suitable temporal and spatial baseline thresholds, and we generated up to ${\displaystyle \frac{N\left(N+1\right)}{2}}$ interferometric pairs of the $N+1$ SAR images.

(2)

The combined image pairs were subjected to differential interferometry to generate time-series interferograms of multiple primary images. The interferometric phase composition is described as follows:

$$\mathit{\Delta }{\varphi }_i\left(x,y\right)=\mathit{\Delta }{\varphi }_{disp}+\mathit{\Delta }{\varphi }_{topo}+\mathit{\Delta }{\varphi }_{atm}+\mathit{\Delta }{\varphi }_{res}$$

(1)

where $\mathit{\Delta }{\varphi }_{disp}$ denotes the deformation phase (along the radar line-of-sight direction), $\mathit{\Delta }{\varphi }_{topo}$ denotes the topographic phase, $\mathit{\Delta }{\varphi }_{topo}$ denotes the air phase, and $\mathit{\Delta }{\varphi }_{atm}$ denotes the disturbance phase. $\mathit{\Delta }{\varphi }_{disp}$ and $\mathit{\Delta }{\varphi }_{topo}$ can be expressed as follows:

$$\mathit{\Delta }{\varphi }_{disp}={\displaystyle \frac{4\pi }{\lambda }}\left[d_{t_2}\left(x,y\right)-d_{t_1}\left(x,y\right)\right]\mathit{\Delta }{\varphi }_{disp}={\displaystyle \frac{4\pi }{\lambda }}\left[d_{t_2}\left(x,y\right)-d_{t_1}\left(x,y\right)\right]$$

(2)

$$\mathit{\Delta }{\varphi }_{topo}={\displaystyle \frac{4\pi }{\lambda }}{\displaystyle \frac{B_{\perp i}}{r sin\theta }}\mathit{\Delta }Z\left(x,y\right)$$

(3)

where $\lambda$ and $\theta$ are the wavelength and angles of incidence, $d_{t_2}\left(x,y\right)$ and $d_{t_1}\left(x,y\right)$ are the accumulated deformations (relative to the reference time $t_0$) from $t_1$ to $t_2$ along the radar line-of-sight (LOS) direction, and $\mathit{\Delta }Z\left(x,y\right)$ is the terrain error.

The terrain phase was corrected using an externally referenced digital elevation model (DEM), while noise was effectively reduced through multi-looking processing and filtering. To calculate the accumulative deformation in the radar line-of-sight (LOS) direction, phase unwrapping was conducted using the minimum cost flow (MCF) method.

(3)

Ground control points (GCPs) were selected for orbital refinement and re-deplatforming to correct the orbital parameters, and to estimate and remove residual phases and phase changes.

(4)

Based on a linear model, the residual elevation and deformation rates were estimated first. Unwrapped phases underwent secondary unwrapping, followed by spatial high-pass filtering and temporal low-pass filtering to remove atmospheric phases. This process inversely deduces the temporal deformation sequences and average deformation rates. Finally, geographic encoding was performed to obtain the temporal deformation sequences and average deformation rates in geographic coordinates.

This method boasts a high level of precision in surface deformation monitoring, achieving accuracy to the level of centimeters or even millimeters. This capability is vital in promptly detecting subtle deformation indicative of landslides. Furthermore, SBAS-InSAR excels in terms of its extensive coverage, enabling the efficient monitoring of landslide hazards across wide areas. This provides robust data support for early warning systems and disaster management efforts. Additionally, the technology’s immunity to weather constraints allows for uninterrupted, all-weather, and all-day monitoring, thereby ensuring the consistency and completeness of the collected data. While SBAS-InSAR offers numerous benefits in landslide identification, it is not without its challenges. Specifically, the costs associated with data acquisition and processing are considerable. Moreover, the identification accuracy can be compromised by the interference of complex geological environments and external factors. Simultaneously, this technique exhibits limitations in precisely delineating landslide boundaries, necessitating the integration of other factors for a comprehensive assessment.

4. Results

This study utilized the SBAS-InSAR technique to derive deformation results along the line-of-sight (LOS) direction. The analysis utilized ascending orbit data from July 2021 to September 2022 and descending orbit data from July 2021 to October 2022, obtained from the Sentinel-1 satellite, as illustrated in Figure 4 and Figure 5. The maximum deformation rate identified from both the ascending (Figure 4) and descending (Figure 5) orbits was -13 mm/year.

The Hongya Village landslide developed on the right coast of the Nanmenxia River (west bank) in the transition zone of low hills and valley plains northwest of Hongya Village. The landslide was a “traction–creep slip–push” composite landslide, and the plane morphology was “triangular”. The landslides on the northern and western sides were bounded by the back wall of the watershed and the steep and gentle junction, respectively, and the shear exit was the foot of the slope. The comprehensive analysis of the ground and borehole data demonstrated that the landslide was a “traction–creep slip–push” composite mega-landslide.

The average length, width, and thickness of the landslide body were approximately 640 m, 1000 m, and 40.0 m, respectively. The main sliding direction was 95–170°, the area was approximately 64.0 × 104 m², and the preliminary estimation of the square volume was 2560.0 × 104 m³. The elevation at the front and rear edges of the landslide was 2549.0 m and 2710.0 m, respectively, with an overall slope of approximately 20°. The back wall was smooth, with a circular chair shape and visible scratches. The back wall of the present landslide was 20–40 m high, with a steep slope, forming a two-stage platform with the front edge of the platform being 15–22 m high. The primary platform was 25–38 m long and approximately 220 m wide, with an overall slope of approximately 15°. The secondary platform was 25–35 m long and approximately 400 m wide, with a gradient of 10–20°. The surface of the landslide protruded from the slope and was slightly curved, and its overall slope morphology was convex, with an irregularly curved leading edge (Figure 6).

Regarding the spatial morphology, the western side of the driving school was “semi-circular” and of the traction mudstone creep–slip type. The shear exit was situated at the base of the incline, which featured a short range and low speed. The northern side of the three middle schools was of the “long tongue” type, with the first crack creeping–sliding after the push bulge. The shear exit was located in the original ditch channel gully, consisting of a high medium-range high-speed airwave that destroyed and buried the school building, with an accumulation thickness of 13.5 m (ZK28). The trees on the southern side of the slope were broken in a “drunkard” shape. The instability of the high, steep critical surface formed by tension crack creeping–sliding led to a situated push. The undisintegrated slider was partially attached to the back wall to form a stacked tile after the middle Range-Gorge groove source broke the wall. The original cistern was relocated by 72 m. Pushing, bulging, and sliding occurred on the western side of the slide in Community 2 of Hongya Village, burying the houses. The region from the hill to the north of the mixing plant was pushed and bulged, and the shear surface morphology varied significantly, with leading-edge inversion and roadway bulging. An “L”-type debris flow occurred along the original channel from the landslide break wall on the northern side of the Xiao-caizi ditch to the mouth of the ditch, where the slide was plowed. In the mid- to lower sections of the multi-stage ridges, the mouth of the ditch was scraped with a shovel.

5. Discussion

5.1. Relationship between Rainfall and Landslides

The studied landslide was located in Hongya Village, Weiyuan Township, Tu Autonomous County, Huzhu, and it had a width of 745 m and a length of 883 m. Optical remote sensing imagery from Google Earth 7.3.6 indicated that the base of the landslide was adjacent to man-made structures such as schools and residential buildings. Figure 7a presents the yearly average deformation rate of the Hongya Village landslide from July 2021 to September 2022, obtained using the SBAS-InSAR technique based on the ascending orbit data of the Sentinel-1 satellite. Figure 7b,c illustrate the time-series deformation at points P1 and P2 (indicated by the white circles in Figure 7a) in response to the monthly rainfall. Figure 7a reveals that the areas of greater deformation were situated in the central and rear sections of the Hongya Village landslide, with a maximum deformation rate of approximately -13 mm/yr.

Figure 7b indicates that the landslide-induced deformation was strongly influenced by precipitation, and it increased as the amount of rainfall increased. The correlation values between the time-series deformation of the points marked P1 and P2 and the monthly rainfall are 0.64 and 0.72, demonstrating that rainfall had a significant influence on the landslide’s movement [43,44]. For instance, the Hongya Village landslide experienced linear deformation from July 2021 to September 2022. Morphing acceleration occurred between July 2021 and November 2021, when the rainfall exceeded 100 mm, while it decreased between December 2021 and March 2022 due to reduced rainfall. However, when precipitation exceeding 40 mm reoccurred in April 2022, the deformation accelerated markedly. In conclusion, the analysis confirmed that rainfall was a critical factor influencing the landslide activity in Hongya Village. Moreover, the increment in rainfall can be seen in Figure 7b in terms of summer precipitation, especially in the month of August, when comparing 2021 to 2022. This phenomenon may result from the effects of climate change, and it will have a lasting influence on the occurrence of geological disasters in the near future.

5.2. Relationship between Stratigraphic Lithology and Landslides

The landslide surface is dominated by Upper Pleistocene aeolian loess (Q₃^eol) (Figure 8a). The overall structure of loess is loose, with large porosity, low strength, vertical joints, and developed fissures. The shear and compressive strength are sensitive to water, and the strength when in contact with water decreases significantly. The loess is underlain by Neoproterozoic mudstone, which is poorly lithogenic and broken. These two sets of strata not only have low strength but also are extremely sensitive to changes in water content. They are highly susceptible to softening when they encounter water, which makes it easy to form a weak sliding surface and create a slippery layer. Concerning water from rainfall, landslides in such stratigraphic environments are susceptible to sliding due to the weak slippery layer [45]. Moreover, other studies have confirmed that the stratigraphic lithology can control and facilitate the progression of a landslide [46,47].

According to the site investigation and drilling engineering exposure (Figure 8b–d), the topsoil layer consists of mainly arable soil or slope deposit loess. The middle and lower soil layers are mainly aeolian loess with a relatively greater depth of approximately 2 to 30 m. This layer is loose in texture, with large pores and developed vertical joints, and it is characterized by wettability, low strength, poor engineering geological properties, and susceptibility to water saturation. The underlying bedrock is of the Xining Formation, and the exposed strata are highly weathered or even almost fully weathered mudstone, with extremely developed fissures. The porosity of the upper loess is large, which makes it easy for surface water to penetrate. The weathered mudstone is softer in texture, and it can easily form an impermeable layer when it is softened by water, resulting in the long-term saturated state of the slope. Slopes are highly susceptible to sliding–deformation damage over time.

In summary, this study determined the deformation characteristics of the Hongya Village landslide and analyzed the deformation mechanism, considering the factors of rainfall and the stratigraphic lithology. Building on this information, this study conducted a detailed analysis of the natural factors influencing the stability of the landslide, showing an advancement from other published material which only considers one of these factors.

6. Conclusions

In this study, the monitoring results of the Hongya Village landslide along the LOS direction were acquired according to the SBAS-InSAR technique, using data from both the ascending and descending orbits of the Sentinel-1 satellite. Moreover, the spatial deformation characteristics of the landslide were analyzed with the aid of displacement derived from SBAS-InSAR and optical imagery. The time-series analysis of the deformation revealed a positive correlation between the rate of landslide deformation and rainfall, with correlation values of 0.64 and 0.72 for two representative monitoring points. In addition, the stratigraphic lithology with loess and mudstone is extremely sensitive to changes in water content, particularly considering water from precipitation, thus forming a weak sliding surface and a slippery layer. According to the combined effects of natural factors and the interpretation of the InSAR monitoring results, the Hongya Village landslide remains unstable to some extent. Therefore, it is crucial to continue conducting high-frequency monitoring of the Hongya Village landslide and to focus on the prevention and early identification of geological hazards.