Evolution of Secondary Periglacial Environment Induced by Thawing Permafrost near China–Russia Crude Oil Pipeline Based on Airborne LiDAR, Geophysics, and Field Observation

Abstract

The China–Russia crude oil pipeline (CRCOP) operates at a temperature that continuously thaws the surrounding permafrost, leading to secondary periglacial phenomena along the route. However, the evolution and formation mechanisms of these phenomena are still largely unknown. We used multi-temporal airborne light detection and ranging (LiDAR), geophysical, and field observation data to quantify the scale of ponding and icing, capture their dynamic development process, and reveal their development mechanisms. The results show that the average depth of ponding within 5 m on both sides of the pipeline was about 31 cm. The volumes of three icings (A–C) above the pipeline were 133 m³, 440 m³, and 186 m³, respectively. Icing development can be divided into six stages: pipe trench settlement, water accumulation in the pipe trench, ponding pressure caused by water surface freezing, the formation of ice cracks, water overflow, and icing. This study revealed the advantages of airborne LiDAR in monitoring the evolution of periglacial phenomena and provided a new insight on the development mechanisms of the phenomena by combining LiDAR with geophysics and field observation. The results of our study are of great significance for developing disaster countermeasures and ensuring the safe operation of buried pipelines.

1. Introduction

Icing refers to the formation of ice from the successive freezing of flows of water overflowing on the surface. The development of icing conditions includes unfrozen water sources and channels [1,2]. Icing is classified as spring, river, or ground icing, depending on the water sources [3]. Its area ranges from several square meters to several square kilometers, and its thickness varies from several centimeters to several meters [4]. It forms under natural conditions, but also under the influence of engineering construction. There is extensive research on icing in permafrost regions [5,6,7]. Engineering construction and operations can interfere with drainage channels and alter hydrogeological conditions, leading to the formation and development of icing, especially in the vicinity of linear projects such as highways, railways, and pipelines [8]. For oil pipelines in permafrost regions, the ground is disturbed not only from the construction of the pipeline but also from its operation, which affects the hydrothermal state of the surrounding permafrost. This changes hydrological processes and promotes the development of secondary geological effects such as ponding and icing. In turn, these geological effects jeopardize the engineering structure, which is not conducive to the long-term stability and safe operation of pipelines.

The China–Russia crude oil pipelines (CRCOPs I and II) were buried in parallel and put into operation in 2011 and 2018, respectively. The annual oil transportation capacity reaches 30 million tons, and the oil temperature remains above 0 °C throughout the year. It spans 1030 km from Skovorodino in Russia to Daqing in China, with approximately 441 km of permafrost in the Mo’he–Jagdaqi section, including 119 km of warm and ice-rich permafrost [9]. Unlike other pipelines in permafrost regions, the CRCOPs crossed the Xing’an (Hinggan)–Baikal permafrost (XBP) area [10], an area with poor thermal stability that responds particularly severely to the thermal impact of the pipeline. Previous studies have shown that the permafrost around the pipe is rapidly warming and thawing [11,12,13], resulting in ground subsidence and water accumulation in the pipeline trench in warm and ice-rich permafrost regions [14]. However, the process by which icing develops in winter at the location of ponding remains poorly understood. There is research on related geological effects along the line, but it is mainly limited to statistical data and qualitative descriptions from on-site investigations [2,15,16]. A few studies have quantified the ponding distribution along the pipeline using Landsat satellite images [17], but there were issues with the large interpretation range, the influence of vegetation, and cloud cover. Due to the lack of sufficient on-site verification, the reliability of image interpretation results is limited. In addition, the use of satellite data requires the consideration of temporal and spatial resolutions [18]. Because the width of ponding above the pipeline is generally less than 10 m, the limited spatial resolution of satellite images creates challenges. In cold seasons, icing develops rapidly in the early stages, and the temporal resolution of images is inadequate for effectively capturing its changes. Sentinel data are increasingly being utilized for research in permafrost regions due to their high revisit rate and all-weather imaging capabilities [19,20,21,22]. Existing research has focused on the characteristics of the ground surface deformation of local sections of the CRCOP based on InSAR technology, but the dense vegetation along the pipeline has limited its further application [23]. Obtaining aerial images from airplanes can solve the above-mentioned problems to some extent, but high costs and operational requirements limit the feasibility of this approach [24]. Unmanned aerial vehicles (UAVs) equipped with lightweight sensors are increasingly used, due to their mobility, precision, and low cost [25]. Among the sensors on UAVs, light detection and ranging (LiDAR) sensors have excellent vegetation penetration and anti-interference capabilities and can be used to quantify periglacial phenomena or other geological effects.

LiDAR is an active remote sensing technology for earth observation that directly obtains 3D coordinate information about the topography. This can be used to generate a high-precision digital elevation model (DEM) by measuring multiple echo observations [26]. The platforms of LiDAR include ground-based, vehicle-mounted, airborne, and satellite-based LiDAR [27]. Airborne LiDAR technology can be used to penetrate the tree canopy. It has a short data production cycle, and it is highly automated, unaffected by terrain fluctuations, insensitive to lighting conditions, and highly accurate. It has been effectively applied to topographic mapping, 3D city modeling, disaster assessments, and other fields [25,28]. It has been used to study permafrost in northwest Canada [29], Alaska [30,31], and the Arctic [32], with high-resolution mapping of wetland topography, surface drainage potential, changes in the permafrost landscape, and estimations of active layer thickness. To our knowledge, it has not previously been used to study the effects of engineering in frozen soil, but its excellent vegetation penetration ability makes it suitable for studying areas with high vegetation coverage. The CRCOP runs underground in frozen soil across a permafrost region with dense vegetation in northeast China. Its construction and operation have changed the surface characteristics along the route, leading to the development of ponding and icing (Figure 1). However, it is difficult to quantify the scale and development of ponding and icing merely with field measurements and satellite images.

There are certain limitations in relying on data obtained through a single method to explain secondary periglacial environments or geological phenomena in permafrost regions. Traditional methods involve field observations, measuring ground surface elevation, thawing depth, freezing depth, vegetation characteristics, the hydrothermal state of permafrost, and so on. Among them, the thermistor placed in the borehole is used to monitor the thermal state of permafrost [33]. Although traditional methods are direct and effective, they cause strong disturbances to permafrost and are limited to point scales. Geophysical techniques, such as ground-penetrating radar (GPR) and electrical resistivity tomography (ERT), are used to map the freezing–thawing state of soil, stratum information, and evaluate permafrost hazards due to their non-destructive, portable, and efficient characteristics. However, when they are applied individually, there may be issues with the ambiguous interpretation of subsurface information. Combining borehole or other prior geological information will make the interpretation results more reliable [13,34,35,36,37]. Although airborne LiDAR can accurately capture surface features, its limitations in probing underground make it difficult to independently reveal complex geoscience issues. Field observation and geophysics can effectively complement these limitations. Chasmer et al. [38] analyzed the spatial correlation between terrain and freezing depth using LiDAR and field observation. Levy et al. [39] investigated the accelerated changes in thermokarst landforms using both airborne and terrestrial LiDAR. Hubbard et al. [30] quantified and correlated surface features with the degradation of subsurface permafrost using LiDAR, GPR, and field observation. Douglas et al. [40,41] linked permafrost thawing to thermokarst development and vegetation changes using LiDAR, ERT, and field observation. These studies have successfully revealed the synergistic changes between the surface and subsurface, but the low frequency of LiDAR measurements and insufficient continuity of field observations have limited the precise depiction of surface feature changes and the revelation of geological phenomenon mechanisms. In particular, there are few studies on the evolution and formation mechanism of secondary periglacial phenomena on the local landscape scale considering engineering characteristics.

We studied a wetland site along the CRCOP and obtained multi-period point cloud data and image data from June to December 2022 using airborne LiDAR technology. We combined these with on-site observation data regarding ground temperatures, ground surface deformations, and pipe burial depth, to study the formation and development of water ponding and ponding-related icing along the pipeline. First, we quantified the scale of the pipe trench and ponding above the pipeline and analyzed the differences in ponding and non-ponding areas. Second, we quantified the scale of icing above the pipeline in winter and analyzed their distribution and development. Finally, we analyzed the formation mechanism of ponding and icing using monitoring data, geophysical data, and UAV data. Moreover, we discussed corresponding prevention and control measures.

2. Study Region and Methods

2.1. Study Site

The study site (50°28′11″ N, 124°13′06″ E; 427 m a.s.l.) is located in the isolated permafrost wetlands along the CRCOP [42], about 391 km away from the first station and about 600 m south of the Jagdaqi pumping station (Figure 2). The permafrost in the site is warm and ice-rich, with a mean annual ground temperature (MAGT; i.e., the annual temperature of permafrost at a depth of zero) of −0.8 to −0.5 °C, active layer thickness of 178–200 cm [12], and permafrost thickness of about 71.5 m (calculated according to the ground temperature curve). According to the data from the meteorological station, the annual range of daily air temperature is about 56 °C, the average warming rate is 0.40 °C/10a, which is higher than the global average, the monthly average wind speed is 0.9–2.6 m/s [12], the mean annual precipitation (MAP) is 540.4 mm, and the snow thickness is about 5–35 cm [13]. The vegetation coverage of the site is dense and mainly composed of Carex tato, mud sedge, and a few low shrubs near the pipeline. A 20 m borehole core [33] showed that the strata from top to bottom included 0.8 m of peat soil, 2.2 m of silty clay, 1.8 m of gravel sand, 3.2 m of fully weathered granite, and 12 m of strongly weathered granite. The geological conditions for engineering along the CRCOP are complicated, and different engineering measures were adopted in different sections. Both lines in the research site were treated to prevent corrosion, but no insulation layer was used. They are spaced about 31 m apart. The designed burial depth of the pipe ranged from 1.6 m to 2.0 m.

2.2. Data Acquisition

We used a DJI Matrice M300RTK UAV with a DJI Zenmuse L1 lens (https://enterprise.dji.com/cn/zenmuse-l1, accessed on 1 June 2022). See

Table 1. Parameters of the DJI Matrice M300RTK UAV and Zenmuse L1 lens.

Equipment	Parameter	Value
DJI Matrice 300 RTK	Weight (including batteries)	6.3 kg
	Max. flight time	55 min
	Max. flight speed	82 km/h
	Satellite positioning systems	GPS + GLONASS + BeiDou + Galileo
	GNSS positioning accuracy (RTK Fixed)	1 cm + 1 ppm (H) 1.5 cm + 1 ppm (V)
DJI Zenmuse L1	Weight	930 ± 10 g
	Point rate	Single return: max. 240,000 pts/s
		Multiple return: max. 480,000 pts/s
	Return number	1~3
	System accuracy (RMS 1$\mathsf{\sigma }$)	Vertical: 5 cm per 50 m
		Horizontal: 10 cm per 50 m
	Distance measurement accuracy (1$\mathsf{\sigma }$)	3 cm per 100 m
	Field of View (Repetitive)	70.4° (H) × 4.5° (V)
	Field of View (Non-Repetitive)	70.4° (H) × 77.2° (V)
	IMU update frequency	200 HZ
	RGB camera effective pixels	20 Mpix

for specific parameters. The DJI Zenmuse L1 lens integrates a laser radar module, high-precision inertial navigation, a mapping camera, a three-axis gimbal, and other modules. We carried out multiple aerial surveys in 2022, on 21 June, 30 October, 5 November, and 16 November 2022, with the same flight parameters. The route parameters were set as follows: lens inclination angle at −90°, flight altitude at 100 m, flight speed at 8 m s⁻¹, laser side overlap rate at 60%, visible light bypass overlap rate at 69%, visible light heading overlap rate at 70%, LiDAR sample rate at 160 KHZ, triple echo mode, the repeated scan mode, and RGB coloring. The obtained point cloud density was 217 points/m², and the ground sampling distance (GSD) was 2.87cm/pixel. The point cloud coloring mode was selected for route planning to obtain true-color point cloud data integrating visible light and the LiDAR point cloud. These two kinds of data were processed to generate digital orthographic maps (DOMs) and digital elevation models (DEMs), respectively, as detailed in Section 2.3.

2.3. Data Processing

The images from the aerial survey were in JPG format, and each image contained geographic reference information (x, y, z). DJI Terra was used for data preprocessing, followed by ENVI 5.3, point cloud post-processing software, and ArcGIS for data post-processing (Figure 3). Image data were processed as follows. First, the UAV images were imported into DJI Terra, and a DOM was derived through stitching, spatial triangulation, and 2D model reconstruction. The projection coordinate system was uniformly transformed into WGS_1984_UTM_zone_51N. Second, the DOM was imported into ENVI, and ground features were classified by image stretching, region-of-interest (ROI) creation, and Support Vector Machine (SVM) classification. Finally, the results of the first two steps were imported into ArcGIS to extract and quantify information regarding water and vegetation. Point cloud data were processed as follows. First, the original LiDAR point cloud data were imported into DJI Terra, and after setting the effective distance of the point cloud, concatenation, accuracy optimization, and an accuracy check, a LAS file was exported with the same coordinate system as the DOM. Second, we imported the LAS file into the point cloud processing software and obtained a DEM of the research area through operations such as denoising, point cloud classification, ground point extraction, and TIN triangulation filtering. During the filtering process, slope thresholds, iteration angles, and distances were set to reduce the influence of discrete points and accumulated errors, thereby enhancing the reliability of the filtering results. Notably, the precision of the original point cloud collected by the LiDAR scanner employed in this paper can be comparable to that of most advanced filtering algorithms [43]. Finally, we imported the results of the first two steps into ArcGIS to extract information regarding ponding, icing, and the terrain. By comparing the DOMs and DEMs in the research area, we analyzed and quantified the distribution, depth, and area of ponding, as well as the development characteristics, range, and volume of icing.

2.4. Accuracy Evaluation

To test the capability of this LiDAR system to maintain accuracy without ground control points (GCPs), we conducted an aerial survey without deploying GCPs. Ten ground marking points (GMPs), made of 65 × 100 cm opaque polyethylene and designed as red right-angle markers, were established (Figure 4c). A Hi-iRTK5 device was connected to a Continuously Operating Reference Station (CORS) network with the account of FindCM. The coordinate information of each GMP was measured. Subsequently, an aerial survey mission was carried out, and image data and point cloud data for the research area were collected. The route settings were the same as the route parameters in Section 2.2. The Real-time Kinematic (RTK) measurement results (Figure 4c) were compared with the GMP information in the DEM (Figure 4b) and DOM (Figure 4d). We then evaluated the plane accuracy, elevation accuracy, and global accuracy of the airborne LiDAR technology with the root mean squared error (RMSE). The calculation formulas are shown in Equations (1)–(3).

$${RMSE}_i=\sqrt{{\displaystyle \frac{\sum {(∆i)}^2}{N}}}$$

(1)

$${RMSE}_{XY}=\sqrt{{\displaystyle \frac{{RMSE}_X^2+{RMSE}_Y^2}{2}}}$$

(2)

$$RMSE=\sqrt{{\displaystyle \frac{{RMSE}_X^2+{RMSE}_Y^2+{RMSE}_Z^2}{3}}}$$

(3)

where $i$ denotes $X$, $Y$, and $Z$, $∆i$ denotes the difference between the measured coordinates of GMPs and the corresponding point cloud coordinates, ${RMSE}_{XY}$ is the plane accuracy, ${RMSE}_Z$ is the elevation accuracy, and $RMSE$ is the global accuracy.

3. Results and Analysis

3.1. Ponding Distribution

By taking an area of 250 m × 70 m within the site as the research case, the ponding distribution along the pipeline was quantified using the UAV images captured on 21 June 2022, as depicted in Figure 5. There was an obvious ponding area above CRCOP II (Figure 5a), and the distribution was concentrated (Figure 5b). The SVM method was used to classify the DOM, with a classification accuracy of 99.69% and a Kappa coefficient of 0.9943, confirming that ponding and vegetation could be accurately extracted (Figure 5c). The vegetation area in the study area was 14,861 m², and the ponding area was 2643.34 m², accounting for about 15.10% of the total area. The vegetation area within 5 m of both sides of the CRCOP II was 1792 m², and the ponding area was 694 m², accounting for about 27.92%.

LiDAR point cloud data were used to obtain precise surface elevation data (Figure 5d), with a resolution of 0.05 m. As can be seen from the figure, the ponding area was obviously lower than in the surrounding area, especially above the pipeline (Figure 5e), and there was a strip-shaped low-lying area, namely the pipe trench. The average elevation of the research area was 427.72 m: 427.75 m in the vegetation area and 427.56 m in the ponding area. The average elevation of the pipeline within 5 m is 427.63 m (Figure 5e): 427.67 m in the vegetation area and 427.51 m in the ponding area, with a difference of 16 cm. For flat areas, small topographic differences can lead to different drainage potential, resulting in different scales of ponding. The terrain data were extracted from the vertical pipeline direction (Figure 5d), and the results showed that the width of the pipe trench was 3–8.5 m, the maximum depth was about 88 cm, and the average was about 31 cm. Because the trench was completely filled with water at that time, the depth of the trench was the same as that of the ponding.

3.2. Icing Development

The dynamic development process of icing was analyzed using UAV images and LiDAR data from periods 30 October 2022 (20221030), 5 November 2022 (20221105), and 16 November 2022 (20221116), as shown in Figure 6, Figure 7 and Figure 8. The black area shows burn marks caused by burning the firebreaks at the end of October each year to meet the fire protection requirements of pipelines.

At the early stage of icing development, a thin ice layer of 2–5 cm formed at the ponding position (Figure 6a), and then the ice layer gradually thickened with a slight ice overflow (Figure 6b). By 16 November 2022, a large amount of ice overflowed, and the icing developed rapidly (Figure 6c) and spread to both sides from the ponding position above the pipeline (Figure 6d). From October 30 to 16 November 2022, the elevation of the research area increased by an average of 8.9 cm, with a maximum value of 48 cm, located directly above the pipeline. Within a range of 5 m on both sides of the pipeline (purple dashed box), the average elevation increased by 14 cm. The surface volume tool in ArcMap was used to calculate the volume of the surface bulge. The DEM difference was taken as the input upper surface (Figure 6d), the plane height was set to 0, and the Z factor was set to 1. The results indicated that the volume of surface bulge in the research area was about 1150 m³: 338 m³ within 5 m on both sides of the pipeline, 553 m³ within 10 m on both sides, and 915 m³ within 20 m on both sides.

According to the field investigation and UAV image interpretation, there were three icing regions (A, B, and C) in the research area (Figure 7a). Their dynamic development characteristics were analyzed. Icing A developed perpendicular to the pipeline direction (N–S), with an east–west direction of about 92 m and a north–south direction of 10–41 m (Figure 7c). Its average height was 7 cm, its maximum value was 35 cm, and its volume was about 133 m³. It overflowed from the ponding area above the pipeline to both sides, and the overflow range on the right side was larger, because of the difference in initial topography (30 October 2022) on both sides of the pipeline prior to the formation of Icing A. The average elevation on the left side of the pipeline was 427.12 m, while that on the right side was 426.98 m. When the water was compressed to a certain extent during the formation of the icing, it broke through a weak point of the overlying ice layer and overflowed. It then froze and the overflow was more inclined to the lower terrain. The decreasing trend of ΔDEM on the right side of the pipeline also illustrates this phenomenon (Figure 7d). By analyzing the central region of Icing A (Figure 7e,f), it is clear that it spread and developed in all directions—higher in the middle, lower in the periphery, and with a stepped character.

Icing B mainly developed along the pipeline direction, and its overflow range was limited due to the obstruction of the embankment of CRCOP I on its left side (Figure 8a). It was about 56 m wide from east to west, and 68–105 m long from north to south (Figure 8c), with an average height of 11 cm, a maximum value of 43 cm, and a volume of about 440 m³. Comparing the DEM data of the three phases (Figure 8c,e), we found that the Icing B developed continuously, and its elevation above the pipeline changed significantly. The ΔDEM map (Figure 8f) revealed the presence of multiple small icing regions above the pipeline. As they continued to develop, water overflowing from the weak points converged and then froze. Icing C was 10–40 m wide and about 80 m long (Figure 8g), with an average height of 13 cm, and a maximum value of 48 cm, and a volume of approximately 186 m³. Due to the higher terrain on the north side and lower terrain on the south side (Figure 8i), more water accumulated on the south side in summer. Therefore, the DEM changed more on the south side in winter, and icing developed more obviously (Figure 8l). Based on zonal statistics, the northern half of Icing C had a maximum height of about 43 cm and a volume of about 62 m³, while the southern half had a maximum height of approximately 48 cm and a volume of about 124 m³.

4. Discussion

Water ponding and icing along the CRCOP mostly occurred in the pipe trench. With the development of the trench settlement, this phenomenon will become more serious, threatening the safe operation of the pipeline. This section discusses the following aspects: the causes of the trench settlement, the formation mechanism of ponding and icing, countermeasures to ponding and icing, and the accuracy evaluation of airborne LiDAR technology.

4.1. Reasons for Pipe Trench Settlement

4.1.1. Permafrost Thawing around the Pipe

The CRCOP runs at above-freezing temperatures all year round and the oil temperature increases year by year, causing thermal effects on the surrounding permafrost [14]. According to the monthly average oil temperature data of the Jagdaqi Pumping Station, the outlet oil temperatures of CRCOP I and CRCOP II in 2022 reached 13.3 °C and 13.9 °C, respectively. This is significantly higher than the designed oil temperature range (−6 °C to 10 °C) [44]. Based on monitoring data from 2018 to 2021 of borehole 2-T1 (2 m from the center of the CRCOP II) and natural borehole T2 (Figure 1b), we analyzed the degree of permafrost thawing around the pipeline. The thermistor cables made by the State Key Laboratory of Frozen Soils Engineering (SKLFSE) were placed in the borehole to record the ground temperature, with a measurement accuracy of ±0.05 °C. Figure 9 shows the ground temperature profiles of two boreholes during the warm and cold seasons.

We found that the uninsulated pipeline caused strong thermal disturbance to the surrounding permafrost, with the accelerated warming and thawing of the permafrost manifested by an annual increase in maximum thawing depth and talik thickness. From 2018 to 2021, the annual maximum thawing depth of the permafrost at hole T2 remained almost unchanged at about −1.8 m, while that at hole 2-T1 decreased from −4.7 m to −9.4 m at a rate of 1.57 m/a (Figure 9a). On 15 October 2021, the average ground temperature at depths of 0–10 m in the two holes was −0.09 °C and 2.86 °C, respectively. In addition, CRCOP II only operated for more than two months from January 2018 to March 2018, and a talik with a thickness of 0.8 m appeared at hole 2-T1 (Figure 9b). Then, the talik thickened to 3.7 m, 6.2 m, and 7.6 m in 2019, 2020, and 2021, respectively, with a thickening rate of 2.27 m/a. However, there was no talik at the natural hole T2. The thermal impact caused by the pipeline was particularly significant in the range of 0–10 m. For example, on 15 March 2021, the average ground temperature at holes 2-T1 and T2 within this depth range was 0.45 °C and −0.79 °C, respectively.

Moreover, this paper utilized the electrical resistivity tomography (ERT) technique, which boasts the advantages of convenience, non-destructive, and non-invasive detection, to detect the talik around the pipeline. The Advanced Geosciences Incorporated (Austin, TX, USA) SuperSting R8 Polarization Meter was used to measure the electrical resistivity (ER) of the stratum on 14 April 2021. The ERT measurement profile was perpendicular to the pipeline direction and passed through the borehole 2-T1, with an electrode spacing of 2.5 m, profile length of 100 m, and maximum penetration depth of over 20 m. The Wenner–Schlumberger electrode array was adopted. The EarthImager 2D software (Version: 2.4.0, Austin, TX, USA) was used to invert the ER data, employing the smoothness-constrained least-squares method for ERT inversion. The inversion result is shown in Figure 10. This paper used the good correlation between resistivity, ground temperature, and lithology to identify the talik. This method has been applied in related studies [12,14], and an ER value of 501 Ωm was used as the critical value for the frozen and unfrozen zone. The result indicated the formation of an asymmetric talik around the pipelines. Specifically, the talik around CRCOP I was about 25.0 m in the horizontal direction and 9.4 m in depth, while the talik around CRCOP II was approximately 8.0 m in the horizontal direction and 7.7 m in depth. The thawing and consolidation drainage of the permafrost around the pipeline caused ground settlement, leading to the continuous development of the pipe trench.

4.1.2. Pipeline Displacement

Permafrost thawing around pipelines leads to a decrease in the bearing capacity of the foundation soil, causing the pipelines to sink [34]. In turn, the displaced pipelines heat the surrounding permafrost, expanding their influence, and repeatedly threatening operational safety and increasing the risk of ruptures and leakage [45,46]. A level was used to measure changes in the deformation rods connected to the pipeline multiple times in order to determine the displacement of CRCOP II. Deformation rods D1 and D2 were 20 m apart and located in the regions of Icing A and Icing B, respectively (Figure 7a). As seen in Figure 11, during the period from April 2018 to April 2021, the buried depth of the pipeline at D1 decreased from −2.66 m to −3.10 m, while at D2 it decreased from −2.45 m to −2.90 m, and the pipe subsidence was 44 cm and 45 cm, respectively. In addition, the construction and maintenance of pipelines both involve damaging vegetation, which may cause strong thermal disturbances to the lower permafrost, leading to pipeline displacement [47].

4.1.3. Ground Surface Deformation (GSD)

The pipeline triggered the accelerated thawing of the surrounding permafrost, and the related thaw settlement developed, causing varying degrees of ground surface deformation that gradually evolved into surface subsidence. A 22-year observation of ground temperatures and thaw settlement in the Norman Wells pipeline in northern Canada also exhibited this phenomenon [48]. Due to the large scale of ponding in our study area, it was difficult to maintain the stability of GSD monitoring stakes. Therefore, we chose to analyze GSD and LiDAR data from another monitoring site along the pipeline.

This site is located 304 km from the first station (51°09′38″ N, 124°11′35″ E), in isolated permafrost wetlands and mainly ice-rich permafrost [13]. Twenty-four steel bars were arranged perpendicular to the pipeline as deformation monitoring points with a spacing of 1 m, all of which were on the right-of-way of the pipeline (on-ROW). Taking the ground temperature borehole as the benchmark pile (Figure 12a), regular measurements were taken using a level to obtain surface deformation data from October 2017 to October 2022. The benchmark stake location was free from permafrost and frost heave and was stable throughout the measurement period to ensure the accuracy of the deformation data.

As can be seen from Figure 12, the deformation trends of all monitoring points were consistent, with a greater deformation above the pipeline than at other positions (Figure 12b). Each point exhibited seasonal frost heave, seasonal thaw settlement, and interannual settlement deformation (Figure 12c). From October 2017 to June 2018, all points exhibited varying degrees of uplift, with an average relative elevation increase of 5.0 cm on the surface. This was due to the fact that in the study area, the thawing period of the active layer ends in October each year, and the frozen part of the previous freezing period has not completely thawed by June. This results in the seasonal uplift of the ground surface. From June to October 2018, the average relative elevation settlement of the surface was 10.4 cm, with an annual cumulative deformation of −5.4 cm. This was due to permafrost thawing from the thermal impact of the pipeline. The average surface settlement at each point from October 2017 to October 2022 was 15.6 cm, with an annual settlement of approximately 3.1 cm/a. The average settlement above the pipeline (points 3–5 and points 16–18) was 19.2 cm, while the average settlement at other points was 14.5 cm. The corresponding settlement rates were 3.84 cm/a and 2.90 cm/a, respectively. The differential thaw settlement deformed gradually with strip-shaped low-lying areas above the pipeline, eventually developing into a pipe trench. The depth of the trench at CRCOP I was significantly greater than that at CRCOP II (Figure 12b). This was because the former has been running for seven years longer than the latter, and the cumulative thermal impact of the former led to a greater surface settlement.

Although single point measurements can accurately quantify surface deformation, the limited and sparse points are insufficient to describe continuous changes in terrain. Therefore, we obtained LiDAR point cloud data of the site using the UAV on 28 October 2023. The DEM obtained after data post-processing was used to extract terrain data (Figure 13). The pipe trench was clearly observed from the UAV image (Figure 13a). Its elevation was significantly lower than in the surrounding area (Figure 13b). The elevation data of four sections (A, B, C, and D) were extracted, and the sampling interval was 0.1 m. Thus, we quantified the depth and width of the trench and its difference from the surrounding terrain. The depth and width of the trench for CRCOP I at section A were about 41 cm and 7 m, while those for CRCOP II were about 14 cm and 5.0 m, respectively (Figure 13c). Similarly, the depth and width of the trench for CRCOP I at section B were about 37 cm and 8.5 m, while those for CRCOP II were about 18 cm and 6.2 m, respectively (Figure 13d). Section C was higher than section D, and the elevation difference was 4.5–42 cm, with an average of about 25.4 cm (Figure 13e).

4.1.4. Thermal Influence of Ponding

Due to its enormous heat capacity and latent heat, surface water can significantly affect the hydrothermal state of the permafrost beneath it [49,50]. A simulation of the thermal impact on the permafrost of the ponding along the pipeline showed that ponding accelerated talik development around the pipeline and deepened the maximum thawing depth of permafrost [16]. We utilized measured data from August 2019 to August 2020 to analyze the temperature differences between ponding and non-ponding areas, as shown in Figure 14.

The temperature of the ponding was measured by the hobo temperature sensor MX2201, with a measurement range of −20 to 70 °C, an accuracy of ±0.5 °C, and a measurement interval set at 1 h. The sensor was placed in the ponding area of the pipe trench (Figure 5b), 0.1 m below the water surface. Soil temperature measurements came from the monitoring data at 0 m and −0.5 m of natural hole T2. The data indicated that the ponding temperature was higher and more variable compared to the soil temperature, ranging from −11.6 to 18.2 °C. Meanwhile, the temperature ranges at 0 m and −0.5 m depths in borehole T2 were −9.5 to 8.2 °C and −5.3 to 4.2 °C, respectively. The annual average temperatures of these three locations were 2.5 °C, −0.3 °C, and −0.5 °C, respectively. The temperature of the ponding was significantly higher than the soil temperature at the same location in the natural site, and the duration of the thawing period was longer. The thermal influence of ponding and the pipeline accelerated permafrost thawing, and subsequent consolidation drainage replenished the accumulation of water. This feedback effect led to an increase in the ponding scale and the development of the pipe trench settlement [11,51]. In addition to the thawing water of permafrost, the sources of ponding in the pipe trench also include precipitation, surface water, and groundwater [34].

4.2. Evolution of Ponding and Icing

Based on previous field investigations, UAV measurements, ERT measurement, and monitoring data, we analyzed the ponding and icing near the pipeline. Furthermore, we described its formation mechanism in order to provide a scientific basis for both the operation and maintenance of the pipeline and the prevention of permafrost disasters. Figure 15 is a schematic diagram showing the development process of ponding and icing.

The first stage is the formation of the pipe trench (Figure 15a). An uninsulated pipeline serves as a heat source that continuously warms the surrounding permafrost, leading to permafrost thawing, pipeline sinking, and talik expansion. Due to the thaw consolidation and permafrost drainage, the ground surface above the pipeline settles to form a pipe trench. The second stage is water accumulation in the trench (Figure 15b). Precipitation, surface water, and water in the talik accumulate in the trench to form ponding. During the third stage, the ponding gradually bears pressure (Figure 15c). When the temperature drops below the freezing point, the ground surface and the ponding area gradually freeze. After the ponding surface freezes, water below gradually bears increasing pressure. The fourth stage is the formation of ice cracks (Figure 15d). As the ice layer thickens, pressure on ponding gradually increases. When the pressure reaches a certain threshold, the confined water migrates to weaker areas, and then it freezes, expands, forms cracks, and continues to extend. The fifth stage is water overflow and the formation of icing (Figure 15e). When the pressure is enough to break through the weak areas, water continuously overflows onto the ground from the crack and freezes, marking the beginning of the formation of icing. The sixth stage is the stage of icing development (Figure 15f). With the formation of cracks in multiple weak areas and the continuous overflow of confined water, the scale of icing gradually increases and spreads to the surrounding natural ground. Overflowing water freezes on the ice formed by the previous overflow, forming a stepped icing.

4.3. Countermeasures for Ponding and Icing

Ponding and icing occur due to thawing and frost heaving, respectively. For disaster prevention, the application effects and environmental protection need to be considered to ensure the safe operation of the pipeline. Besides the CRCOP, other countries in the world have constructed oil pipelines in permafrost regions, such as the Trans-Alaska pipeline in the USA, the Norman Wells pipeline in Canada, the Eastern Siberia–Pacific Ocean (ESPO) pipeline in Russia, and the Golmud–Lhasa pipeline in China [52]. While these pipelines have made significant contributions to social development, they have also encountered some permafrost-related hazards. Some countries have proposed targeted engineering measures and solutions [2].

The Trans-Alaska pipeline was completed in 1977, transporting oil at temperatures ranging from 38 to 63 °C. The high oil temperature quickly thaws the permafrost, posing a high risk of foundation instability or fracture in some sections [45]. To address this issue, 676 km of the pipeline along the route was laid overhead by thermal piles, which effectively controlled permafrost degradation and thawing hazards. The Norman Wells oil pipeline was completed in 1985, with oil temperatures ranging from −1 to 6 °C, and the entire pipeline was buried. There were frequent problems such as frost heaving, thaw settlement, and trench settlement along the pipeline, among which the trench settlement at KP608 exceeded 2 m, and the width of the trench expanded from 1 to 16 m [48]. Measures such as winter construction, cooling oil temperature, wood insulation, and increasing pipe wall thickness were adopted to deal with the above problems [53,54]. The ESPO crude oil pipeline was completed in 2012, with oil temperature of 15–35 °C. The whole pipeline was buried. Thaw settlement, ponding in pipe trenches, and thermalkarst occurred along the pipeline [55], and measures such as an insulation layer, hanger, and cooling measures (e.g., thermosyphon) were used to mitigate permafrost hazards. The Golmud–Lhasa oil pipeline was completed in 1977, transporting oil at temperatures ranging from −5 to 9 °C. It was also buried along the whole line. Since its operation, at least thirty leaks and four pipeline ruptures have occurred due to issues such as frost heave, thaw settlement, icing, and so on [56]. Due to economic constraints at that time, no engineering measures were taken for effective prevention and treatment, but the line was changed and re-laid in the later stage [57].

To deal with ponding and icing, thaw settlement and frost heave should be prevented and controlled. The essence of treating icing is to treat the water, and treating water is the key [58]. This is handled from the following two aspects. On the one hand, permafrost thawing is controlled around the pipe. Firstly, from the perspective of management, oil temperature is controlled to reduce the heat transfer from pipelines to surrounding permafrost; secondly, from the perspective of regulating permafrost temperature, measures [59] such as vegetation restoration, thermosyphon cooling, air ventilation duct cooling, insulation layers, and composite approaches can be adopted to slow down permafrost thawing, soil settlement, and pipe trench development. In this way, the development of ponding and icing can be effectively alleviated. On the other hand, measures are also taken to prevent and control frost heave. Firstly, soil that is sensitive to frost heaving can be replaced by soil that is insensitive to frost heaving, so as to improve the bearing capacity of the pipe foundation and reduce water migration. Secondly, measures such as trench backfilling, and building drainage ditches, underground pipes, and blind ditches can prevent water accumulation [2]. In addition, from the perspective of the pipeline structure, increasing the pipeline wall thickness can enhance its ability to resist deformation and damage. For long-distance linear projects such as the CRCOP, the engineering geological conditions along the route are complex and diverse. In preventing and controlling ponding and icing, it is necessary to consider the geological conditions, permafrost environment, the scale of ponding or icing, and the applicable conditions of measures. The optimal countermeasures or comprehensive solutions should be selected to achieve the best effect.

4.4. Accuracy Evaluation of Airborne LiDAR Technology

We evaluated the accuracy of airborne LiDAR technology without image control points using 10 GMPs (Section 2.4). By comparing the aerial survey results with the RTK measurement results of each point (

Table 2. Aerial and ground measurement results.

Points	UAV Measurement Results			Ground Measurement Results
	X/m	Y/m	DEM/m	N/m	E/m	H/m
GMP-1	586449.4467	5591661.7710	431.120	5591661.7871	586449.5035	431.162
GMP-2	586521.8039	5591620.1844	427.482	5591620.2161	586521.8523	427.530
GMP-3	586600.8975	5591581.1776	427.274	5591581.2222	586600.8958	427.313
GMP-4	586393.2438	5591544.3120	430.422	5591544.3426	586393.261	430.483
GMP-5	586462.6533	5591493.4846	427.010	5591493.5142	586462.6863	427.064
GMP-6	586563.5692	5591465.0525	426.858	5591465.0728	586563.6345	426.876
GMP-7	586339.9731	5591433.0862	429.823	5591433.1361	586339.9763	429.885
GMP-8	586414.226	5591392.2576	426.845	5591392.2874	586414.2726	426.867
GMP-9	586511.398	5591350.8587	426.531	5591350.9268	586511.3661	426.580
GMP-10	586482.6986	5591537.1151	427.216	5591537.1720	586482.7053	427.243

), the root mean square errors (RMSEs) between them were calculated according to Equations (1)–(3) (

Table 3. RMSEs between aerial and ground measurement results.

Flight Altitude	RMSEx/m	RMSEy/m	RMSEz/m	RMSExy/m	RMSE/m
100 m	0.0380	0.0409	0.0447	0.0395	0.0413

). The results showed that the DJI M300RTK with the DJI Zenmuse L1 LiDAR lens had a horizontal accuracy of 3.95 cm, vertical accuracy of 4.47 cm, and global accuracy of 4.13 cm, all of which were better than the values declared by the manufacturer (10/5 cm horizontal/vertical). The results have demonstrated the ability of this LiDAR system to maintain high accuracy without GCPs, which can significantly reduce fieldwork efforts and improve the efficiency of data processing. Štroner et al. [60] also measured the accuracy of the system, and the results showed that the accuracy in all directions was 3.5 cm after removing the global geo-referencing error. Its excellent surveying and mapping accuracy make it good for depicting topography and ground surface deformation.

Exploring the mapping capabilities and error sources of airborne LiDAR systems for vegetation, water, snow cover, etc., is crucial for its application prospects. Airborne LiDAR is an active detection method of determining surface information by actively emitting lasers and detecting ground echoes [61]. With excellent vegetation penetration ability, it can be used to obtain multi-layer elevation data regarding the vegetation canopy, branches, and ground, enabling the precise detection of the actual ground information [62]. However, when it is applied to dense and opaque vegetated areas, the absence of ground point clouds can reduce its measurement accuracy [63,64]. Since the accuracy test in our study was conducted in May during the early stages of vegetation growth (Figure 4), there are still certain limitations, despite the excellent test results we obtained.

Meanwhile, airborne LiDAR plays a significant role in water depth measurements and underwater topographic mapping. However, due to the great energy loss of laser pulse propagation in the air–water interface and water, the weak echo signal, and the influence of suspended particles in water, its measurement ability at depths is subject to various limitations [65,66]. This was also confirmed by the test results in our study. The average depth of ponding obtained from LiDAR point data was about 31 cm (Figure 5d), with a maximum value of about 88 cm. This was lower than the water depth measured on site (over 1 m in some areas). Due to turbid water accumulation in the study area, it was difficult for the airborne LiDAR to accurately obtain underwater information, resulting in an underestimation of the water depth. Therefore, only the ponding area was counted in the article, and its volume was not quantified.

Furthermore, airborne LiDAR plays an important role in estimating snow depth. Snow depth was estimated using the DEM difference between snow-free and snow-covered areas. Hopkinson et al. [67] pioneered the application of airborne LiDAR technology to the York region of Canada to map the distribution of snow depth under the forest. Subsequently, some scientists continued to focus on snow depth estimations and error-source research [18,68,69,70,71,72]. The main error sources include the interpolator type and resampling techniques used in data post-processing to obtain DEMs, the positioning type and inertial navigation system (INS) of the airborne platform, the influence of airflow during the aerial survey, canopy density, and snow characteristics. The volume of the icing was quantified in Section 3.2 (Figure 6c), but due to the increased elevation of the ice surface caused by the overlying snow cover, there was a certain overestimation of the icing volume. The snow depth in the study area was only 3 to 5 cm on 6 November 2022 (Figure 7a), resulting in a relatively minor overestimation of the icing volume. With the development of point data post-processing algorithms, it is anticipated that the snow layer can be removed and the icing can be accurately quantified.

UAVs, equipped with multi-sensors, can obtain multi-scale and multi-dimensional geoscience data, thereby eliminating the errors and uncertainties of a single sensor. They have been applied in various fields, including climate change, geological disasters, and ecological environments. The combination of visible light and thermal infrared sensors was utilized for the inversion of surface temperature, surface soil moisture content, as well as for the analysis of urban heat island effects [73,74]. Furthermore, this combination has been applied in research on snow depth measurement in alpine regions, as well as the hydrothermal deformation of slopes and thawing features in permafrost regions [75,76,77]. The combination of thermal infrared and LiDAR sensors has been employed in geothermal monitoring, yet its application in permafrost regions remains to be explored. The combination of visible light and LiDAR sensors, in addition to providing high-precision topographic and geomorphic data, also offers more detailed surface texture and spectral information, enabling its application in some respects in permafrost regions, including vegetation degradation, wetland terrain mapping, and the impact of vegetation and terrain on snow distribution [29,78,79,80]. In general, research on this combination in permafrost regions is limited. This paper expands its application in capturing and quantifying the evolution of secondary cryogenic phenomena, and further explores its application potential in frozen soil engineering.

There are still some limitations in this study. Firstly, when using UAV images for land cover classification (Figure 5c), the ponding area was underestimated due to the shelter caused by vegetation near the ponding. Secondly, there was an underestimation in the extraction of ponding depth, which could potentially be improved by using bathymetric LiDAR sensors with blue-green wavelengths [31]. Thirdly, when extracting the surface elevation with LiDAR data, some ground point cloud data were missing because of the presence of clustered meadows close to the ground in local areas of the site (Figure 5d). This affected the accuracy of the DEM data. Finally, the icing volume was calculated using the difference between the DEMs of 20221030 and 20221114 as the input for the upper surface (Figure 6d). This assumes that all terrain changes were due to icing development, but the changes caused by the re-freezing of the active layer were not excluded, resulting in a certain overestimation of the calculated icing volume. Perhaps this issue can be resolved by deploying ground surface deformation points and installing a GNSS deformation displacement monitoring system. Firstly, the relative elevation of the ground surface can be measured synchronously during LiDAR measurement. Secondly, the surface changes during multiple LiDAR measurements can be calculated. Finally, the lifting volume of the active layer at the icing area can be subtracted from the icing volume obtained through LiDAR measurements.

5. Conclusions and Prospects

In this study, we used airborne LiDAR technology to map the secondary periglacial phenomenon along the CRCOP. Based on visible images, LiDAR point cloud data, geophysical data, and observation data, the scale of ponding and icing was quantified, their development and formation were analyzed, and corresponding countermeasures were discussed. The main conclusions are as follows:

(1) The CRCOP operates at positive temperatures that cause the permafrost surrounding it to thaw rapidly, leading to pipeline sinking and trench settlement. The maximum thawing depth of permafrost at 2 m away from the center of CRCOP II, which has been in operation for only three years (2018–2021), has decreased by 4.7 m, and the talik has thickened by 6.8 m. The pipeline has sunk to 3.1 m from its designed burial depth of 1.6–2.0 m.

(2) The average depth of water ponding in the pipe trench was approximately 31 cm on 21 June 2022, with a maximum depth of 88 cm. Suspended particles in turbid ponding limited the sounding capability of LiDAR, leading to an underestimation of the water depth.

(3) The dynamic development process of icing was captured using multi-temporal UAV images and LiDAR data. Icing overflowed preferentially to the low-lying areas. It was higher in the middle and lower in the periphery, with a stepped pattern. Three icings regions (volume: 133 m³, 440 m³, and 186 m³) developed above the pipeline and they were gradually converging together.

(4) We revealed the evolution of water ponding and icing and put forward preventative measures. A two-pronged approach should be adopted that aims to control permafrost thawing and prevent frost heave. The focus should be on slowing down the permafrost thawing and promoting water drainage and diversion.

Although previous studies have conducted relevant work on ponding and icing, qualitative research has not clearly revealed their formation mechanism. However, our study used airborne LiDAR technology to capture their dynamic development process and provided a strong explanation for their causes by combining geophysical data and observation data. In this paper, the ground surface changes (ponding and icing) were obtained by airborne LiDAR, and the engineering characteristics, subsurface permafrost variations, and local landscape changes were linked by geophysics and field observation data. Furthermore, the penetration capability of LiDAR in detecting water and snow cover, as well as its measurement accuracy under various noise and weather conditions, requires further assessment. The accuracy of the aerial survey should be evaluated again in the most developed period of vegetation in order to determine its ability to penetrate vegetation. The point cloud data also contain reflectance information of ground features. It would be intriguing to investigate how to integrate visible light data with reflectance information to improve the reliability of ground feature classification. Meanwhile, improvements to the accuracy of point cloud data and the development of post-processing algorithms will also contribute to precise quantifications of the scale of ponding and icing. With the light-weighting, intelligence, and decreasing cost of airborne LiDAR systems, more LiDAR data will be obtained and used in future studies of glacier changes, ecological evolution, periglacial environment, geological effects, and engineering challenges. This study provides a new insight into the formation mechanisms of periglacial phenomena by combining airborne LiDAR with geophysics and field observation. It also offers a framework for investigating the current situation of engineering geo-hazards and evaluating the impact of engineering on the permafrost environment. Furthermore, the combination of airborne LiDAR technology and satellite-borne imagery shows promise for studying the dynamic evolution processes of thermokarst lakes, frost mounds, landslides, and so on.