Distress Characteristics in Embankment-Bridge Transition Section of the Qinghai-Tibet Railway in Permafrost Regions

The Qinghai-Tibet Railway has been operating safely for 16 years in the permafrost zone and the railroad subgrade is generally stable by adopting the cooling roadbed techniques. However, settlement caused by the degradation of subgrade permafrost in the embankment-bridge transition sections (EBTS) is one of the most representative and severe distresses. A field survey on 440 bridges (including 880 EBTSs) was carried out employing terrestrial laser scanning and ground-penetrating radar for comprehensively assessing all EBTSs in the permafrost zone. The results show that the types of distresses of EBTSs were differential settlement, upheaval mounds of the protection-cone slopes, subsidence of the protection-cone slopes, surface cracks of the protection cones and longitudinal and transverse dislocation of the wing walls. The occurrence rates of these distresses were 78.93, 3.47, 11.56, 3.36, 21.18 and 4.56%, respectively. The most serious problem was differential settlement, and the average differential settlement amount (ADSA) was 15.3 cm. Furthermore, the relationships between differential settlement and 11 influencing factors were examined. The results indicate that ADSA is greater on the northern side of a bridge than on the southern side and on the sunny slope than on the shady slope. It is also greater in the high-temperature permafrost region than in the low-temperature permafrost region and in the high-ice content area than in the low-ice content area. The EBTSs are more influenced by ice content than by ground temperature. The ADSA increases when the embankment height increases, the particle size of subgrade soil decreases and the surface vegetation cover decreases.


Introduction
The Qinghai-Tibet railway (QTR) is located in the hinterland of the Qinghai-Tibet plateau (QTP) and traverses 550 km through continuous permafrost with complex engineering and geological conditions. Approximately 40% of the permafrost in the Qinghai-Tibet engineering corridor (QTEC) is ice-rich (defined as containing > 20% ice) and around 50% is warm permafrost (Cheng 2005;Wu, Sheng, et al. 2020). However, under the global warming trend, the continuous permafrost on the QTP has degraded by 30% since 1980 and is predicted to decline by nearly half by the end of the century (Nan et al. 2005). Many factors, such as the significant thermal sensitivity of permafrost and the disturbance by human activities, have changed the regional atmosphere in certain locations (Vecellio et al. 2019). This change will affect the heat accumulation in the underlying permafrost of the embankment, which may reach a level that leads to permafrost degradation and causes railroad settlement . The QTR has been operating for more than 15 years, as an active-cooling principle was adopted in the design and construction for protecting the underlying permafrost and the roadbed stability (Cheng 2005). During the construction, sun sheds, crushed rock berms, thermosyphons, and additional insulation layers inside the roadbed can effectively protect the permafrost under the roadbed and alleviate the thawing settlement deformation (Cheng et al. 2004;Wen et al. 2005;Feng et al. 2006;Wu, Zhao, et al. 2020). However, The QTR is still facing problems caused by permafrost degradation, as the climate warming on the 1 3 plateau is about two times the average global warming rate (Ran et al. 2018). In-site investigation showed that the main distresses include thaw settlement, cracks, and others related to permafrost degradation. According to the investigation, more than 20 km of the railway embankment was remedied during 2006−2018 because of these distresses ). Among the distresses, the problem at the embankment and bridge transition section (EBTS) is prominent and serious for the remediation project of the QTR.
As a crucial component of the connection between the bridge and the embankment, the EBTS is where the problem of differential settlement most frequently occurs . The EBTS creates a sudden change in the structural properties due to variations in stiffness, geotechnical and construction issues and subgrade reactions (Choi 2013;Indraratna et al. 2019). Such abrupt changes and the properties of the construction lead to differential deformation, and the dynamic response of the train is significantly amplified (Mishra et al. 2017). In such places, the dynamic responses contribute to successive degradation of track, ballast and subgrade, ultimately, in turn, aggravating differential settlement and structural damage. In addition, differential settlement in the EBTS can lead to an increased maintenance effort and higher costs for the structure (Li and Davis 2005)-for example, engineering maintenance is carried out on the track in the transition area four to eight times more often than on the open track in the Netherlands (Varandas et al. 2011). Along with the problems of differential settlement, stiffness variations and track deterioration that exist in the non-permafrost zone, there are challenges with the freeze-thaw effect of permafrost on the EBTS in the permafrost zone. Due to the high ground temperature and ice-rich permafrost on the QTP, the severe degradation of permafrost will present serious challenges to the maintenance of the QTR, especially the stability of EBTS (Ni et al. 2021;Li et al. 2022).
According to a survey carried out in July 2007, Lin et al. (2008) found that the main symptoms of bridge abutment distresses include cracks, breaking and seepage at the bottom, and some of the protection cones have cracking and settlement. Niu et al. (2011) found that the average settlement amount was up to 7 cm and the settlement rate was 82.6%, and they analyzed the relationship between settlement and six factors, including bridge orientation, slope direction, embankment height, ground temperature, ice content and subgrade soil type, based on a survey of 164 embankment and bridge transition sections conducted 3 years after the QTR opened. According to Wang and Wu (2017), the settlement rate of EBTSs decreased from 70−250 mm/a before 2013 to 10 mm/a after 2014 based on the monitoring results of four bridges. By using the highresolution synthetic aperture radar interferometry (InSAR) technique to assess the deformation of the QTR, Zhang et al. (2019) identified that the deformation of the EBTSs was obvious and that the deformation rate was as high as 15 mm/year; the deformation of the road-bridge transition section was related to factors such as orientation, roadbed height, slope direction, permafrost types and topography. Based on the logistic regression (LR) and support vector machine (SVM) methods, combining information on ground temperature, ice content, soil type, orientation, roadbed height and precipitation, Zhang et al. (2022) carried out a risk assessment for all EBTSs along the QTR in the permafrost zone and the results showed that EBTSs with high or very high risk accounted for one-third of the whole line. To cope with the differential settlement of the EBTSs in the permafrost zone, raising the thickness of the ballast layer to control the level of the track is a solution (Wu, Sheng, et al. 2020). Flake fill was applied to the bottom layer of the EBTSs, and the inverted trapezoidal form of filling was adopted to achieve the effect of cooling the structure (Liu et al. 2004). However, as a result of the different geological conditions and permafrost types along the QTR in the permafrost region, the EBTS distress is worsening and is difficult to solve with a single measure. For this reason, compared to our 2009 survey of 164 bridges, which could not provide a comprehensive overview of the reality of EBTSs in permafrost zones, this study carried out a field survey of 440 EBTSs along the QTR, including the types of distresses and their degree of development and influencing factors. The site survey information is urgently needed to understand and evaluate the actual situation of the whole line.
The objectives of this study are (1) to investigate the distress conditions of all EBTS along the QTR in the permafrost zone; (2) to analyze the deformation and distribution characteristics of EBTS; (3) to reveal the patterns between differential settlement and its influencing factors; and (4) to discuss the distress development mechanisms and treatment measures of EBTS, which provide a reference for the maintenance of the QTR in permafrost areas and the construction of new railway lines and are the basis for further research on the mechanism of distress development as well as the employment and development of treatment techniques.

Study Area
The permafrost along the QTR distributes from the Xidatan fault valley in the north to the Anduo valley in the south (approximately 550 km) on the QTP, with 80% of the region above 4000 m in elevation (Luo et al. 2018). As shown in Fig. 1, the section of this study is from Wangkun station (bridge No. 496) to Anduo station (bridge No. 934), which traverses a landscape of mountains, high plains, basins and valleys intersecting each other. The continuous permafrost is a significant feature of the study area, also with a few seasonal permafrost and talik zones. The part of the permafrost that has melted is called talik, which is characterized by the presence of liquid water throughout the year, despite the surrounding permafrost being frozen. Talik is largely located near the Tuotuo River in this study area. The ground ice content above 50% (ice-rich, ice-saturated, and ice layers with soil) accounts for 42.8% of the whole line, with thicknesses up to 120 m in some areas. Along the QTR, the majority of areas have a mean annual ground temperature (MAGT) over − 1 °C and belong to the high-temperature permafrost region, whose active layer thickness is 1 to 4 m (Jin et al. 2008).
The mean annual air temperature in the study area ranges from − 4 to -6 °C (Chai et al. 2018;Wang et al. 2022). The statistics of the national meteorological stations in Wudaoliang (stable permafrost zone) and Tuotuo River (talik section) indicate that significant climate warming from 1980 to 2019 has led to an increase in MAGT, thickening of the active layer and melting of subsurface ice within the QTEC . The region has a cold and dry continental climate with an average annual precipitation of 50−400 mm, which is mainly concentrated between May and September, with July and August often seeing the most amounts (Luo et al. 2018;Zhang et al. 2019;Wang et al. 2022). Strong sunlight, sparse cloud cover, and swift winds together contribute to the average annual evaporation of up to 1500 mm (approximately six to eight times the precipitation) and snowfall on the ground generally lasts less than a week (Chai et al. 2018;Mu et al. 2018).
The annual growth rates of passenger kilometers and freight kilometers of the QTR were 9.1 and 9.6 times higher, respectively, in the second phase (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013) than in the first phase   (Li, Wang, et al. 2016). The QTR adopts ballasted single-track model to have better impact toughness and facilitate drainage. The structure of EBTS is generally in the form of a positive trapezoid, laying fill-in layers, in order to have better transitional variations and resist the effects of frost heave, and set up longitudinal and transverse drainage (Niu et al. 2011). The EBTSs also have a small number of inverted trapezoid structures and add rubbles at the bottom, which can serve to ventilate and cool down the temperature (Liu et al. 2004).

Survey of Embankment-Bridge Transition Sections
In 2009, three years after the opening of the QTR, we surveyed 164 bridges (Niu et al. 2011). With the degradation of permafrost, it is urgent to clarify the actual condition of EBTS, so we conducted a new survey in 2022 to include as many relevant distress types and factors as possible. To obtain more representative and comprehensive data, the surveys of EBTSs were carried out on 440 bridges within  Fig. 2, was the key survey area. Each bridge has two EBTSs, divided into the Golmud side (northward) and the Lhasa side (southward). The orientation of the railroad line is from north to south, with the sunny side as the left side of the train traveling direction and the opposite side as the shady side. The investigation was conducted by employing terrestrial laser scanning (TLS), direct measurement with a foot tower and field recording. The targets of this survey were bridge beams and protection-cone slopes as well as the differential settlement of the joint between the protection cone and the wing wall on both sides. The field's surface condition, the height and structure of the embankment, the orientation of the embankment, and the slope direction were also recorded.

Survey Method and Processing
In the survey area, the bridges were classified as water bridges or dry bridges. The dry bridges are designed to cross extremely intense permafrost degradation areas. Among them, the Qingshui River Bridge is approximately 11.7 km long and is located over a large area of continuous permafrost. In addition, the bridges can also be classified by bridge length (BL) to facilitate subsequent analysis. The bridges in this survey include 125 small bridges (8 m ≤ BL ≤ 30 m), 110 medium bridges (30 m < BL < 100 m), 146 major bridges (100 m ≤ BL ≤ 1000 m) and 59 grand bridges (BL > 1000 m). The TLS was employed to collect the point cloud of the bridge and its surroundings for the major and grand bridges. From the point clouds, we can obtain information on the type, location, and size of the EBTS distresses, as shown in Fig. 3e. For the remaining bridges,  method; c Measurement of differential settlement from a wide range of point clouds using the Cyclone Software; d Targets with black and white side; e The Leica Scanstation P50 TSL we used the foot tower to measure the EBTS's differential settlement directly, as shown in Fig. 3a.
The TLS obtains a high-resolution point cloud of the survey scene using a near-infrared laser with high precision and speed, outperforming the traditional single-point measurement method. The Leica Scanstation P50 terrestrial laser scanner was utilized in this study with superior characteristic parameters (Table 1).
To investigate the developments of the EBTS distresses and the surrounding topography, the Leica Scanstation P50 laser scanner generally used the 120 m range mode to capture accurate and efficient data for this survey. For some bridges that are difficult to measure directly, such as high embankments or obstruction from rivers, the 270 m or 570 m range mode was employed. The angle of view, in which only the positional data of objects that are in the foreground are captured, is a distinct limitation of TLS when it comes to supplying positional (spatial) data. To eliminate mistakes caused by terrain factors and line of sight obstruction, the data splicing of multiple stations was aided by two or three reflected targets placed between the observation areas of two nearby stations. By scanning one bridge, 3 to 10 sets of point cloud data were acquired.
The software Leica Cyclone 9.2.0 1 was used to process point cloud data. Because the point cloud obtained at each station is in a coordinate system with the center of the scanner's lens as the origin, each set of coordinates is different. The calculation of the conversion of the coordinate system is performed by 2 to 3 targets (common points, as seen in Fig. 3d) between the stations so that the point clouds of all groups are obtained under the same coordinate system. As an example, Fig. 3b shows the point cloud of the No. 695 bridge that has been registered. The points with 0.05 m spatial sample spacing were extracted from point cloud data of two measurements because scattered and dense point clouds were redundant for topographical surveying and demanding for computing resources in processing. The TLS observation captured some superfluous points that need to be segmented to remove. After the process of registering, resampling, and denoising, the EBTS distresses can be identified and measured manually (Fig. 3c).

Types of the Embankment-Bridge Transition Section Distresses
The survey revealed that three types of distresses in the EBTS are prevalent in the permafrost zone along the QTR. The first type of distress is the differential settlement of EBTSs (Fig. 4a). The relative elevation difference between the top of the protection cone and the surface of the embankment-filled soil is treated as differential settlement, which is distributed in the outer portion of the EBTS and is easily observed or measured. The TLS and manual measurements were used to obtain the settlement values of this distress, which was found to be the most severe and frequent in the permafrost zone. Numerous issues, including the instability of protection cones and beams, can arise as a result of the differential settlement of the EBTSs. Figure 4a depicts the sunny-slope side of the Lhasa side of bridge No. 695. Based on the point cloud obtained from the TLS, the settlement is 70.9 cm. The second type of distress is the deformation of the protection cones, including the upheaval mounds of the protection-cone slopes (Fig. 4b), the subsidence of the protection-cone slopes ( Fig. 4c) and the surface cracks of the protection cones (Fig. 4d). The protection cone is a structure to maintain the EBTS stable and prevents water from scouring the bridge abutment, which serves to increase the safety of the bridge and embankment. The differential settlement of EBTSs is mostly accountable for the deformation of the protection cones. Penetration cracks begin to develop on the cone's surface when the differential settlement deformation reaches a certain level (Fig. 4d), usually trending in the slope's direction. Continued development of differential settlement may cause subsidence of the cone to which the wing wall is attached (Fig. 4c).
The third type of distress is the longitudinal dislocation ( Fig. 4e) and transverse dislocation (Fig. 4f) of the wing wall and bridge beams caused by the differential settlement of EBTSs. The survey's findings indicate that longitudinal deformation is more common in the EBTS than transverse deformation. The safety of train travel and the durability of

Distribution of the Embankment-Bridge Transition Section Distresses
In the permafrost region, the QTR travels through basins, valleys, high plains and mountains, which are four different types of landforms. The railroad surroundings are separated into 18 landform units from the Xidatan fault basin to the Anduo valley, as shown in Fig. 5. Following the survey, it was discovered that the differential settlement issue was severe in the EBTS. For the measurements of differential settlement, 880 EBTSs were surveyed and for each EBTS two sites were measured (Fig. 3c), and there was a total of 1760 measurement sites. In the surveyed area, the average differential settlement amount (ADSA) and settlement rate of EBTSs were 15.3 cm and 78.75%, respectively. The ADSA and settlement rate of each geomorphic unit are marked on the map. The ADSA and settlement rate of the 164 bridges (bridge No. 496-659) surveyed by the team in 2009 were 7.0 cm and 82.60%, respectively, and for comparison, the ADSA and settlement rate of these EBTSs in this survey were 19.4 cm and 85.06%, respectively, which are greater than the values obtained in the 2009 field survey. Almost all EBTSs were affected by differential settlement in some sections, such as 97.92% in the Kunlun Mountains area, 95.45% in the Tanggula Mountains area and 95% in the Wuli basin (Fig. 5). In the Kunlun Mountains area, the ADSA was as high as 27.11 cm. The purple dots show the EBTSs in a bridge with an ADSA higher than 40 cm. There are 25 such bridges, and their bridge numbers are indicated in Fig. 5. The No. 511 bridge, which is located in the Kunlun Mountains area, has an ADSA of 107.43 cm at its four measurement points and is the bridge with the largest settlement on the whole line. The differential settlement of EBTS can be classified into four grades according to the magnitude of settlement (S): slight settlement (S < 10 cm), medium settlement (10 cm ≤ S < 20 cm), serious settlement (20 cm ≤ S < 30 cm) and very serious settlement (S ≥ 30 cm). As shown in Table 2, they accounted for 29.66, 22.73, 11.42 and 14.94% of the total, respectively. The EBTS experienced medium and below settlement total accounting for 73.64% and they occupied the majority. However, the survey found that 263 measurement points showed a differential settlement of over 30 cm, which needs special attention. Particularly, 14 measurement points on seven bridges with a differential settlement of at least 80 cm have been monitored and appropriately reinforced. According to the railway subgrade design code of China, the accumulative settlement shall not exceed 200 mm or the settlement rate shall not exceed 50 mm/year. As a result, the distress problem of EBTS still meets the specification requirements and the QTR is safe and stable.
The differential settlement of EBTSs causes damage to adjacent structures, including protection cones and beams, in addition to deteriorating tracks and train jumps. In this investigation, the distresses of the protection cone are divided into three categories: upheaval mounds of the protection-cone slopes, cracks on the surface of the protection cones and subsidence of the protection-cone slopes. This survey focused on the four cones of each bridge. Crack damage affected 203 cones in varying degrees, frequently in the direction of the slope surface, making it the most common distress of the protection cone (Fig. 6b). Despite occurring less frequently, upheaval mounds and subsidence of the protection-cone slopes have a greater impact on the structure of the protection cone (Fig. 6a and c).
During the field investigation, it was observed that there were more occurrences of distresses at the junction of the beams and wing walls. The direct contact between the wing wall and the beam results from the movement of the wing wall toward the beam body and even the subsidence joints vanish. Figure 6d demonstrates the prevalence of this type of distress along the QTR in the permafrost zone, with an occurrence rate of up to 21.18%. Another type of distress that could result in rail bending is the transverse displacement of the wing wall. Compared with the first type of distress, the transverse displacement of the wing wall occurs less frequently and is mostly concentrated from the Hoh Xil Mountain area to the Chiqu valley (Fig. 6e). It can be seen that the cracks of the protection-cone surface and the longitudinal dislocation of the wing walls are widely distributed along the QTR. The Beilu River basin to the Chiqu valley, the Kaixinling Mountains area and the Tanggula Mountains

Results
Based on the survey findings, the differential settlement of EBTS, which is the most serious and prevalent distress, is analyzed. Five major categories of factors, including railroad direction and bridge, embankment, permafrost and geographical characteristics, are revealed to influence the differential settlement pattern.

Relationship with Slope Direction and Orientation of Embankment
The alignment of the railroad greatly affects the underlying permafrost due to the different solar radiation and absorbed heat. The orientation of the QTR from Xidatan to Anduo is generally south-north. The difference in solar radiation on the surface of the embankment slope causes a strong asymmetry in the deformation of the embankment on both sides of the railroad, giving rise to the shady and sunny slope effect ). The annual average settlement rate of the sunny side of the embankment is more than twice that of the shady side, according to the field monitoring of three embankments (Tai et al. 2020). Because of the significant dependence of permafrost on temperature, the degree of thawing varies with the heat absorption of the slope. In addition, similar effects occur on the southern and northern sides of the bridges. Therefore it is crucial to take into account the railroad orientation and the direction of the embankment slope. Figure 7 shows the relationship between the differential settlement of EBTSs and the slope direction and orientation of the embankment using three metrics: ADSA, total differential settlement amount (TDSA) and settlement rate. The TDSA is the total amount of the settlement at all measurement points. On the northern side of bridges, the ADSA and TDSA were 15.6 cm and 13,681.5 cm, respectively, both larger than the average for the entire line (Fig. 7a). Differential settlement occurs more frequently and severely on the northern side of the bridges than on the southern side because the abutments and cones on the northern side receive solar radiation for a longer time and more intense, causing permafrost degradation. For bridges No. 496-659, the TDSA of the EBTSs on the southern and northern sides were 6530.4 cm and 6193.0 cm, respectively, which is an increase of 163% and 193%, respectively, compared to the 2009 survey. In addition, the two values of these 164 bridges (from the Xidatan fault basin to the Chiqu valley) accounted for almost half of that of the EBTSs of the entire line. Compared to EBTS on the shady slopes, the differential settlement was not much more significant on the sunny slopes. However, on the sunny slope, the settlement rate of EBTSs was 82.92%, which was greater than that on the shady slopes. The differential settlement on the sunny slopes persists to be more severe despite the frequent daily maintenance and reinforcement of the railroad. Figure 7b depicts the relationship between differential settlement and the subdivision orientation of embankments. A statistical unit of 15 degrees is used, where 0° indicates northward. The orange bars indicate the value of the average differential settlement and the green sections represent the probability of settlement within that statistical unit. For embankment orientations between 60 and 90° and between 250 and 270°, the differential settlement of the EBTSs was above the average. The ADSA of EBTSs with the northwestsoutheast embankment orientation was also more severe compared to the other orientations. This is because the overall orientation of the railway line is northeast-southwest. The sample size in this direction is small and there are some EBTSs with high settlement. In particular, bridge No. 511, one of the bridges with the most severe differential settlements, is oriented at 135°, with the ADSA as high as 107 cm.

Relationship with Bridge Characteristics
Based on the design information and field survey of the QTR, bridges can be classified into non-prestressed concrete bridges and prestressed concrete bridges. There are 228 prestressed concrete bridges in the survey area, with major and grand bridges making up about 75% of the total. Small and medium bridges are commonly made of non-prestressed concrete. Figure 8 demonstrates a positive correlation between the bridge length and the ADSA for non-prestressed concrete bridges. The ADSA for the major Fig. 7 Relationship between the differential settlement of embankment-bridge transition sections (EBTSs) and slope direction and orientation of embankment. a The blue and red dashed lines represent the average differential settlement amount (ADSA) and the total differential settlement amount (TDSA) for all EBTSs, respectively, and the blue, red and light red bars represent ADSA, TDSA, and settlement rate of different EBTSs; b The inner and outer circles indicate the ADSA and the settlement occurrence rate, respectively bridge and the grand bridge was 18.9 cm and 21.7 cm, respectively, both of which far exceeded the overall average settlement. However, for prestressed concrete bridges, the effect of bridge length on the differential settlement of EBTSs was essentially the same and below average. This is attributed to the fact that prestressed reinforcement increases the tensile strength of concrete to a great extent, inhibits the development of cracks and thus features superior resistance to freeze-thaw erosion. In addition, compared to the non-prestressed concrete bridge, its lower self-weight and building height have beneficial effects on foundation settlement and sunny-shady slope effect. The employment of prestressed concrete for major bridges and grand bridges has a positive impact on EBTS. The ADSA was particularly high for small prestressed concrete bridges, due to the small number of small prestressed concrete bridges and the large variation in settlement data. Considering the high altitude construction factor of prestressed bridges, increasing EBTS stiffness difference and high construction cost, it is not suggested to use prestressed bridges in all conditions and small and medium bridges can use non-prestressed concrete bridges due to the more optimistic settlement phenomenon.

Relationship with Embankment Characteristics
Box plots are employed to illustrate the relationship between the differential settlement of EBTSs and the embankment and geographical location. Additionally, the relationship between this influence and settlement classification is depicted using a heat map. Since bridges are rigid constructions, they do not deform significantly when subjected to the action of vehicles and self-weight over long periods. The deformation of an embankment, however, differs from that of a bridge because an embankment is a flexible structure and massive compression deformation is easily produced under the action of long-term external force. The relationships between embankment height, subgrade soil type, embankment reinforcement measures, and differential settlement of EBTSs are depicted in Fig. 9a-c, respectively. These basic data were accessed from the design of the QTR, while some reinforcement measures were installed later during maintenance and related data were obtained through the on-site survey.
To facilitate the discussion in this study, the embankment heights of the QTR were divided into four classes that range from low to high. As shown in Fig. 9a, the median differential settlement amount (MDSA) of EBSTs for embankments higher than 8 m was 15.4 cm, with the values of other embankment heights being 9.2, 11 and 11.8 cm in the box plots. The ADSA is greatly impacted by the few embankments along the QTR with embankment heights over 8 m. The average values in the box plots demonstrate that when embankment height increases, the differential settlement of EBTSs also rises. Furthermore, embankments with heights over 6 m are prone to greater differential settlement of EBTSs, with average differential settlement over 30 cm accounting for 22% and 27% of those over 6 m and 8 m, respectively. Increasing the embankment height can remarkably reduce the additional dynamic stress of the permafrost layer and have some insulating effects to keep it safe (Zhu et al. 2022). However, when the height of the embankment increases, the self-weight and thawing settlement also increases, leading to a significant cumulative settlement of the embankment (Tang et al. 2021).
In this study, the subgrade soils of the QTR in the study area were classified into four categories according to soil particle size, including fine particle soil, coarse-gravel soil, crushed-boulder stone and weathered rocks. Figure 9b shows that the ADSA in the fine particle soil zone, which includes clay and silty clay, reached 16 cm, while the values in the other regions were 15.4, 14 and 13.8 cm, respectively. In addition, the box plot of the weathered rocks type shows that the average data was smaller than the average of the whole line, and the median was only 9.3 cm, much lower than other areas. The results demonstrate that the differential settlement of EBTSs increases with decreasing soil particle size. This is attributed to the fact that, under the same conditions, the low content of bound water and low unfrozen water content in the soils with big particle sizes results in less freezethaw effect and greater soil strength, which leads to less settlement of the embankment (Wang et al. 2014). On the contrary, soils with smaller particle sizes are more sensitive to the freeze-thaw cycle, thus bringing greater instability to the embankment (Konrad and Lemieux 2005;Zhang et al. 2017).
The reinforcement measures for the deformed embankment of the QTR include additional block-stone slopes, crushed-rock slopes, and thermosyphons. The MDSA for a Fig. 8 Relationship between the differential settlement of embankment-bridge transition sections (EBTSs) and the length of the bridge and whether the bridge is prestressed or not. The expression of the solid black line is y = 2.432x + 11.171, with a correlation coefficient of R 2 = 0.675 . The expression of the dashed black line is y = 1.665x + 7.1, with a correlation coefficient of R 2 = 0.654 . ADSA average differential settlement amount, MDSA median differential settlement amount pure-soil embankment is 9.8 cm and the MDSA for employing reinforcement measures such as block-stone slope, crushed-rock slope and thermosyphon are 11.3, 16.0 and 20.0 cm, respectively, as shown in Fig. 9c. The heat map shows that approximately 20% of the EBTSs with reinforcing measures have differential settlement exceeding 30 cm. The reinforcement measures were installed at a later period, and the differential settlement had developed during the time of pure-soil embankments. Since the survey data reflect the accumulated settlement, it is impossible to determine which reinforcement measures are effective.

Relationship with Geographical Characteristics
Different landforms generally have unique topographic features, soil properties, vegetation cover conditions and corresponding depositions that can modify how water and heat are exchanged between the surface and the subsurface, affecting the distribution and properties of permafrost (Li,Fig. 9 Box plots and heat maps of the relationship between the differential settlement of embankment-bridge transition sections (EBTSs) and embankment and geographical characteristics. The solid and dashed lines represent the mean and median of the data, respectively. ADSA average differential settlement amount Sheng, et al. 2016;Li, Wang, et al. 2016). Figure 9d shows the relationship between the settlement of EBTSs and different topography, including valleys, basins, mountains, and high plains. In the basin and mountainous areas, the ADSA of EBTSs was 12.6 cm and 13.8 cm respectively, while it was relatively large in the valley area, at 15.5 cm. In addition, the ADSA of more than half of the EBTSs in these three topographic regions was less than 15 cm. The ADSA and MDSA in the high plains reached 25.7 cm and 22.2 cm, respectively, which were higher than that in other regions, despite that the settlement statistics varied greatly there. On the high plains, 36% of the EBTSs experienced settlements exceeding 30 cm. Because rivers, swamps, and thermal thaw lakes and ponds are widely distributed on the high plains, the underground ice is developed in these areas, which combined with climate change and engineering disturbances leads to permafrost degradation (Jin et al. 2008).
Ground surface condition is also a factor affecting permafrost stability. In this survey, the ground surface conditions within 500 m of the EBTSs were investigated, including semi-buried checkerboard sand barriers, rivers, thermokarst lakes, weathered gravel-filled ground surface and vegetation cover types (sparse grassland, grassland meadow, alpine steppe and swamp meadow). Figure 9e shows that the ADSA of the EBTSs in sparse grassland and grassland meadow areas was 17.7 cm and 15.2 cm, respectively, which was larger than the ADSA of 12.4 cm and 13.9 cm in alpine steppe and swamp meadow areas. The ADSA of the EBTSs on the weathered gravel-filled ground and near the semi-buried checkerboard sand barriers area was 15.1 cm and 16.1 cm, which were similar to the less vegetated areas. The reduction of surface vegetation will lead to higher ground temperatures and thicker active layers, which will then affect the stability of EBTS (Lin et al. 2015). The bridge planning has taken into account the influence of rivers, so rivers had a low impact on the settlement of EBTSs. However, the number of thermokarst lakes along the QTR is increasing and the thermokarst effect can lead to the degradation of nearby permafrost . The ADSA of EBTSs near the thermokarst lakes reached 22.9 cm and the ADSA of nearly half of the EBTSs was up to 30 cm.

Relationship with Permafrost Characteristics
Permafrost is an essential factor for the deformation of the EBTS in permafrost zones. Moreover, ice content (i v ) and the MAGT are key factors affecting the stability of permafrost (Jin et al. 2008). The permafrost along the QTR was classified into five categories according to the ice content, including ice-poor permafrost ( i v ≤ 10%), icy soil (10% < i v ≤ 20%), ice-rich permafrost (20% < i v ≤ 30%), saturated permafrost (30% < i v ≤ 50%) and thin ground ice ( i v > 50%) (Ma et al. 2011). They are labeled as S, D, F, B and H, respectively. In addition, areas with no permafrost or full thawing of permafrost are called unfrozen regions (talik). The permafrost was also classified into four categories according to the MAGT, including badly unstable warm permafrost (Tcp-I, Tcp ≥ − 0.5 °C), unstable warm permafrost (Tcp-II, − 1.0 °C ≤ Tcp < − 0.5 °C), basically stable permafrost (Tcp-III, − 2.0 °C ≤ Tcp < − 1.0 °C), and stable low-temperature permafrost (Tcp-IV, Tcp < − 2.0 °C) (Ma et al. 2011). Figure 10 describes the relationship between the ADSA of EBTSs and the ice content and MAGT. The results indicate that the ADSA of EBTSs increases with the increase of the ice content of permafrost. Particularly, the ADSA in saturated permafrost and thin ground ice regions was 17.4 cm and 18.9 cm, respectively, which were significantly higher than the overall average. The lowest ADSA of EBTSs was noticed in the talik region with a value of 11.9 cm because of the absence of freeze-thaw effects of permafrost. The highest average settlement occurred in the Tcp-II region, followed by the Tcp-III region. The lowest ADSA was found in the Tcp-IV region with a value of only 11.5 cm.
Tcp-I and Tcp-II are called high-temperature permafrost zones, whereas Tcp-III and Tcp-IV are called low-temperature permafrost zones (Niu et al. 2011). In the high-temperature permafrost zones, the ADSA increased in the F+B, B, B+H, and H zones. Especially in the B+H and H zones, the ADSA was up to 20 cm. However, in the low-temperature permafrost zones, the ADSA was lower in the areas with less ice content. In regions with high ice content, the settlement was still higher than the overall average settlement. This indicates that the ice content of the permafrost has a great influence on the stability of EBTS. Fig. 10 Relationship between the differential settlement of embankment-bridge transition sections (EBTSs) and ice content types and mean annual ground temperature (MAGT) types. ADSA average differential settlement amount. S ice-poor permafrost, D icy soil, F icerich permafrost, B saturated permafrost and H thin ground ice 1 3

Discussion
Based on the above findings, in this section, the formation and development of distresses by using the GPR technique are elaborated and the applicability and limitations of the existing engineering measures in embankment-bridge transition sections are discussed.

Formation and Development of Distresses
Warm season heat can easily pass through the wing wall, protection cone concrete, and other auxiliary parts of the bridge due to their excellent heat transfer qualities into the permafrost at the bottom of EBTS and inside the embankment (Chai et al. 2019). The permafrost around the EBTS is in a long-term heat-absorbing or degrading process. Therefore, the heat absorption of the embankment near the EBTS is much larger than that of the normal embankment, which is the most significant reason for the melting and sinking of the fill behind the abutment. In turn, a "wedge" is formed between the abutment and the foundation, resulting in the accumulation of warm-season precipitation behind the abutment. A field survey of selected EBTSs suffering from severe distresses was conducted using Crossover CO730 GPR (Impulse Radar, Sweden) with 70 and 300 MHz shielded antennas, and the longitudinal and transverse profile arrangements are shown in Fig. 2. The detection time window was set to 93 ns and 375 ns and the sampling spacing was 5 cm triggered by a wheel. Figure 11a shows the transverse profile of bridge No. 695 EBTS, whose differential settlement amount was up to 69 cm, collected in the warm season by the software Reflexw 2 for processing. The reflections were relatively heterogeneous in the range of 20-48 m in the vertical coordinate, which is the transition zone, indicating that moisture has entered the transition section. It is shown that the extent of the weak area on the sunny side is larger than that on the shady side, and the weak area on the shady side will continue to develop, which in turn will lead to the continued downward movement of the permafrost table. During the freezing and water expansion in winter, the resulting frost-heaving force promotes differential settlement and wing wall displacement. By the next melting period, the ice in the region melts to produce a large amount of moisture and collects here with precipitation. Figure 11b can also confirm the existence of a large water-rich area in the transition section area (0-30 m on the vertical axis), resulting in subsidence in the region. The weak layer is located 2 m below the original natural ground surface, which indicates that this area is a new thawing layer converted from a high ice content permafrost layer, and the layer will become the main deformation layer in the transition section after about 3-5 freeze-thaw cycles ). This problem can cause deformation of the rails, which leads to increased dynamic loads of the trains, in turn exacerbating the expansion of the weak zone, as shown by the yellow arrows (Indraratna et al. 2019). The accumulation of water and thermal erosion can also lead to the thickening of the active layer in the region (Doré et al. 2016).

Engineering Measures
Due to the particular characteristics of permafrost, the distresses of EBTSs in the permafrost region also have special characteristics not found in other areas. The permafrost protection principle is upheld by employing a variety of cooling techniques, which can help maintain stable foundation conditions (Cheng et al. 2008). Therefore, the current challenge is to protect the warming permafrost around EBTSs and control the distress development. The QTR has been operating well since its construction in 2006, but the survey revealed some potential instabilities in the permafrost zone (Lin et al. 2008), among which the distresses of EBTSs are becoming more and more prominent. Despite the reinforcement measures used on the EBTSs, such as block-stone slope, additional thermosyphon, truss reinforcement and stone-dam reinforcement (Fig. 12), the challenges posed by the distresses of EBTS in the permafrost zone remain significant.
Our survey found that block-stone slope is one of the most common means in the treatment of EBTS (Fig. 12a). This measure is most frequently installed on the protection cones with high differential settlement to change the temperature boundary conditions of slopes by convection of air and finally achieve the cooling effect to maintain the stability of the underlying permafrost (Goering 2003;Cheng et al. 2012). However, being filled with gravel or aeolian sand will reduce the effect of convective cooling and cause unfavorable internal heat dissipation during the cold season (Yu et al. 2016;Zhao et al. 2019). Figure 5 shows that a large number of EBTSs with severe differential settlement are unevenly distributed along the QTR, and the biggest problem currently is the high cost of treatment, such as high processing, transportation, and labor costs. There is an urgent need to develop new methods to accommodate the EBTS distresses illustrated in Figs. 5 and 6.
Thermosyphons are also commonly used in the treatment of EBTS to effectively reduce the internal temperature of the structure and improve the bearing capacity of the frozen soil foundation (Mei et al. 2021). However, it is difficult to use thermosyphons on a large scale to treat EBTS and there is also the problem that the range of effect is limited because it is difficult to cool the core area of the EBTS. In comparison to block-stone slopes, it is considerably more challenging to install for existing structures. Truss reinforcement and stone-dam reinforcement (Fig. 12c-d) are two methods that have only been rarely used, primarily to prevent the longitudinal displacement of the wing walls and to ensure the expansion joints between the wing walls and beams. These measures are post-reinforcement measures, but they cannot fundamentally address the problem and cannot solve the transverse displacement of the wing walls. Fig. 12 The embankmentbridge transition section (EBTS) reinforcement measures along the Qinghai-Tibet Railway (QTR). The picture of the block-stone slope was adapted from Zhang et al. (2022)

Prospect
This study mapped the current distress condition of EBTSs along the QTR based on a large amount of field-measured data. The most common and serious distress of EBTSs was found to be differential settlement. The relationship between the differential settlement of EBTSs and its influencing factors was revealed. However, only a few qualitative relationships were derived, and the present investigation was not only unable to obtain changes in differential settlement or settlement rates but also insufficient to predict the development of distresses or their deterioration under the tendency of permafrost degradation. The maintenance data of the railroad is also a very vital part, but these data are difficult for us to access, which may affect the accuracy of the results, especially for the deformation of the protection cones. Given the GPR probing results within EBTS (Fig. 11), the next step is to use geophysical techniques to deeply analyze the developmental mechanism of EBTS differential settlement. More frequent field monitoring of EBTS distresses and systematic research on the deterioration mechanisms of EBTS is urgently needed to develop new railroad reinforcement and maintenance measures in the future.

Conclusion
Based on the survey of the embankment-bridge transition section (EBTS) and the analysis of each influencing factor, the following conclusions can be drawn: (1) The 880 EBTSs in the permafrost zone are frequently distressed and the common types of distresses are differential settlement (78.93%), upheaval mound of the protection-cone slope (3.47%), subsidence of the protection-cone slope (3.36%), surface cracks of the protection cones (11.56%) and longitudinal dislocation (21.18%) and transverse dislocation (4.56%) of the wing wall and bridge beams.
(2) Compared with other distresses, the most common problem along the QTR is the differential settlement of EBTSs, with an average differential settlement amount of 15.3 cm. Among the 1760 measurement points, the ADSA of 14.98% of them was greater than 30 cm, and there were 14 points where the ADSA was greater than 80 cm. The maximum ADSA was 107.43 cm, at bridge No. 511. The differential settlement is related to slope direction and orientation of embankment and bridge, embankment, permafrost and topographical characteristics.
(3) The settlement of EBTSs is more severe on the northern side than on the southern side of the bridges, and it is greater on the sunny slope than on the shady slope. Within 60-90° and 250-270° of the railroad orientation, the ADSA of EBTSs is high. For minimizing the differential settlement of EBTS, prestressed concrete bridges are recommended for major bridges and grand bridges, but non-prestressed concrete bridges are more appropriate for small and medium bridges. (4) The higher the embankment height, the higher the ADSA of EBTSs. When the embankment height exceeds 6 m, EBTS tends to suffer from more severe differential settlement. Smaller soil particle size and lower vegetation cover will cause greater differential settlement. In particular, the ADSA of EBTSs near thermokarst lakes is up to 22.9 cm, and the settlement of EBTSs in high plains is generally greater than in other areas. (5) The settlement of EBTSs is greater in high-temperature permafrost areas than in low-temperature permafrost areas and in high-ice content areas than in areas with low ice content. The ADSA exceeds 20 cm in high temperature and high ice content areas, but is relatively stable in low temperature, low ice content areas or the talik region. The ice content has a greater effect on EBTS than ground temperature.