Introduction

Steep terrain and rapid rivers in mountainous areas lead to unfavorable engineering geological and hydrological conditions for the construction of infrastructures. The cutting of slopes or the filling of roadbeds during the construction of mountainous roads may affect the stability of neighboring slopes (Conforti and Ietto 2019). Even when open to traffic following construction, the roadbeds of a mountainous road often subside unevenly as a result of the scouring of nearby rivers or deep-seated gravitational slope deformations, affecting the track of the road (Wang and Huang 2015; Vuillez et al. 2018; Meneses et al. 2019). Traffic on a mountain road can easily be blocked by rock falls, debris flows, or slope collapses, so the maintaining cost exceeds that of maintaining common highways (Fookes et al. 1985; Calcaterra and Santo 2004; Yang et al. 2012; Camera et al. 2014; Lee et al. 2018b). Despite the low traffic volume on mountain roads, large investments must usually be made in their conservation and maintenance to support social development, land conservation, and disaster prevention and mitigation.

Among the diverse types of landslide (Varnes 1978; Hungr et al. 2014), rock slope movements (rock falls, rock slides, and rock avalanches) are common and they typically cause problems for the maintenance of mountain roads (Mateos et al. 2016; Ragmi et al. 2016). Hungr et al. (1999) investigated rock slope movement-induced risks to two main transportation corridors in southwestern British Columbia using maintenance records spanning four decades. They found that the volumes of 3500 rock movements ranged from less than 1 m3 to over 4.0 × 107 m3 and most of the risk to life results from rock falls of intermediate volume 1–10 m3. Hasegawa et al. (2009) indicated that remnant large-scale landslides are the major causes of debris flows and slides along the major highway passing zones of the Nepalese Himalayas during the monsoon period. Significant hydrothermal alteration and clay mineralization in crushed and jointed rocks have been observed in the failure zones of large-scale landslides, which are related to the faults and folds by active and dynamic geological conditions of the Himalaya Range. Wang (2010) characterized crack patterns on tunnel linings that are associated with shear deformation that is induced by instability of the neighboring slopes. Chiu et al. (2017) reported that the deep-seated gravitational slope deformation induced not only constant road subsidence but also the continuous evolution of lining cracks in a tunnel that passes through a deformed area. Hearn and Shakya (2017) concluded that salient relief, dynamic geology, and heavy rainfall with earthquakes, landslides, floods, erosion, and the formation of sediments accounted for most of the highway hazards in the Himalaya region over the last century.

Slope instability–induced disasters repeatedly affect mountain roads in locations that are close to each other because road sections under unfavorable topographic and geological conditions are fragile. Repair work typically involves restoring passage through damaged sections as quickly as possible, so insufficient time is available to ascertain the factors that affected the disaster or all of the work that is required for complete rehabilitation, leading to repeated disasters (Giordan et al. 2017). Various methods for describing the slope instability–induced risks for mountain roads have been widely discussed in recent years. They include conventional mapping of geological and geotechnical conditions along a road and evaluating the engineering characteristics of rock masses (Siddique et al. 2015; Zheng et al. 2016; Francioni et al. 2019), the use of total station and drone photography to monitor road subsidence and slope displacement (Kovács et al. 2019; Eker et al. 2018), the identification of potential landslides using a high-resolution digital elevation model (DEM) based on airborne or ground-based light detection and ranging (LiDAR) (Breschan et al. 2018; Lee et al. 2018a), analysis of ground displacement using differential interferometry synthetic aperture radar (DInSAR)-related techniques (Bayer et al. 2017; Pappalardo et al. 2018), and the use of data mining models that support statistical regression and probability analyses for the spatially explicit prediction of the susceptibility of a landslide-prone landscape to landslides (Barton et al. 1974; Saha et al. 2005; Bar and Barton 2017; Singh and Kumar 2018; Fallah-Zazuli et al. 2019; Zhang et al. 2019). The relevant literature focuses on evaluation of vulnerable sections in mountain roads, clarifying the factors that influence stability of nearby slopes and developing appropriate maintenance strategies and rehabilitation methods, sometimes as part of a strategy for adapting transport infrastructure to global climate change (Strauch et al. 2015; Palomo 2017; Bordoni et al. 2018; Lentini et al. 2019).

Highway No. 8, Taiwan (referred to as the Tai-8 highway hereafter), was built in the late 1950s mostly by manpower using manual tools. The geological conditions along the route of the Tai-8 highway are complicated, with morphological features that are unfavorable for highway construction, such as cliffs, high and steep slopes, canyons, and erosion gullies. Up to 226 engineering personnel sacrifices were caused by such a difficult construction environment (TPHB 1960). In the six decades since the Tai-8 highway was opened to traffic, more than 250 disasters and road closure events have been caused by landslides, making the Tai-8 highway the highway with the highest disaster potential in Taiwan. The section of Tai-8 highway that winds along the Dasha River has a length of 9.4 km, of which more than 20% has been rerouted because of landslide-induced maintenance problems. This study gathers remote sensing images that have been taken in this area over more than 60 years and records of highway maintenance. The routes of the highway in various periods are examined with reference nearby landslide events. A field investigation is carried out to verify the types of landslide and to semi-quantitatively evaluate rock mass conditions along the route. The effect of the engineering geological conditions on the maintenance and conservation of mountain roads are investigated. Finally, the use of various maintenance strategies to mitigate the influence of slope collapses, gully developments, and river lateral erosion are discussed.

Study area

The Tai-8 highway passes through the Central Range with mostly steep terrain as a result of constant active endogenetic and exogenetic processes in Taiwan. The highway connects western and eastern parts of Taiwan and is also known as the Central Cross Island Highway or the East and West Cross Island Highway. The eastern section of the Tai-8 highway runs through the Liwu River basin, where Taroko, Yanzikou (Swallow Grotto), and Jiuqudong (Tunnel of Nine Turns) are characterized by sheer cliffs and wonderful views. The Liwu River is one of the major rivers in eastern Taiwan and originates between the Hehuan Mountain and the Qilai North Peak with an elevation of more than 3500 m and a length of about 55 km before it flows into the Pacific Ocean. The Liwu River basin is formed from tributaries such as the Shakadang River, the Dasha River, the Waheier River, the Cien River, the Taci Jili River, and the Sanzhan River; it drains approximately 600 km2 of steep terrain, which is underlain by metasedimentary rocks. The area has high rates of tectonic uplift in the range 3 to 6 km per million years and high sediment yield, indicating strong forcing and the presence of natural tools for incision of rocks by abrasion (Hartshorn et al. 2002; Liu et al. 2001). Approximately 107 metric tons of sediment moves through the river each year or about 0.1% of the global supply of sediment to the sea (Hartshorn et al. 2002; Milliman and Syvitski 1992). Figure 1 shows the four main sub-basins through which the Tai-8 highway passes before it runs into the canyon section of the Liwu River; the sub-basins are, from upstream to downstream, the Taci Jili River basin, the Cien River basin, the Waheier River basin, and the Dasha River basin.

Fig. 1
figure 1

Liwu River and its five major sub-basins. Study area is marked

Full size image

The Tai-8 highway makes a sharp turn at Sta. 159.5 k near the Houran Pavilion and runs into the Dasha River basin from the Waheier River basin (Fig. 2). The highway heads northward and descends along the right bank of the Dasha River, before turning southward at the Sta. 163.8 k close to the Huitouwan. After it passes through Guyuan and Wenshan, the highway crosses over the Dasha River by the Jhih-huei Bridge at Sta. 168.9 k near Tianxiang where that river converges with the Waheier River to become the Liwu River. The Tai-8 highway meets with the No. 9 highway at Sta. 189.75 k near Taroko after it passes southeast eastward through the canyon of the Liwu River. This study concerns the section of the Tai-8 highway between Sta. 159.5 k and Sta. 168.9 k with a length of approximately 9.4 km (referred to as the Dashaxi section hereafter). The elevation of the Dashaxi section at Sta. 159.5 k near Houran Pavilion is about 1000 m, that at Sta. 163.8 k near Huitouwan is 760 m, and that at Sta. 168.9 k close to Tianxiang is approximately 500 m (Fig. 2a). Topographic limitation by the watershed and confluence point of the Waheier River and Dasha River leads to a 500-m descent along the 9.4-km-long highway section. In this study, the upper road refers to the section between Sta. 161.5 k and Sta. 163.8 k and the lower road refers to the section between Sta. 163.8 k and Sta. 166.0 k; both are winding mountain roads.

Fig. 2
figure 2

Route of Tai-8 highway in the study area and regional geology. a Three zones in 2014 aerial photograph with locations of feature points used for image processing. b Topographic and regional geological map with locations of blockages of highway due to historical disasters

Full size image

Figure 2 b shows a regional geological map of an area that includes the study area (Lo 1992; 1993). Main exposed rock formations in the Dasha River basin include Tianchang Marble, Paiyang Schist, Pilu Formation (composed of phyllite and metamorphosed sandstone and metamorphosed chert), Guyuan Schist, and terrace deposits. The Guyuan anticlinorium and Tianxiang synclinorium are the major geological structures, and the inferred Tipanshan Fault passes around 0.4 km northwest of Huitouwan. The schistosities of schist formations have strikes in the northeast direction and typically dip northwest, causing the schistosity generally to intersect the slopes of the upper road and the lower road obliquely or reversely. The Dasha River is winding and narrow, and the water flow is rapid. Some undercut slopes in the attack side of the river are observed to have been formed by vigorous lateral erosion and water scouring, and slip-off slopes are present in the other side. The slopes of the right bank of the Dasha River are mostly in the range 30–60°, increasing to 50–80° in the vicinity of the study section, with local slopes even becoming steep or overhanging as a result of the road construction. Vegetation in the study area is dominated by arbors, and shrubs and herbaceous plants are more developed on the slopes that have recently collapsed. Figure 2 b shows the locations of road disasters and blocking accidents, based on maintenance records since 1990, most of which were induced by slope instability.

Method

Remote sensing images are collected and processed to map the route of the Tai-8 highway, and topographic features and variations are interpreted to investigate the track of some recent landslides and the relationship between the rerouting of the Dashaxi section and the landslides. A field investigation is carried out to verify the records concerning the landslides and to determine the engineering characteristics of the study area, with the purpose of elucidating the causal factors that affect the landslides and the effect of engineering geological characteristics on the maintenance of the mountain roads.

Processing aerial photographs

Figure 3 shows the periods during which collected aerial photographs and images were captured. The 22 photographs and images were taken between 1959 and 2019, covering a time span of 60 years, and are grouped into four classes based on the type of camera used. After 2008, images were taken using a digital camera with a focus length of 120 mm. The camera components and parameters concerning orientation during shooting are provided. Information about the ground control points (GCPs) for photogrammetry is also available. Therefore, a digital surface model (DSM) with high accuracy was produced directly. Photographs were taken during the periods 1997–2008 and 1975–1997 using RMK TOP/TOP2 and RC-10 cameras, respectively, and indicated principal point (IPP), fiducial center (FC), and measure marks (MMs) are available for each photograph. The parameters of exterior orientation, required to produce DSM, were obtained using GCPs, and stereo pairs there were formed by successively shot and partially overlapping photographs. The components of the camera that were used to capture photographs before 1975 were unavailable, so IPP, FC, and MMs could not be obtained, neither the production of DSMs. These photographs were rectified and their locations roughly determined using a limited number of zones with specific features.

Fig. 3
figure 3

Collected aerial photographs and images. Four periods are identified according to type of camera used. (Red crosses indicate the year of used photographs)

Full size image

To facilitate the analyses of topographic variations and changes in the highway route over the years, all collected aerial photos had to be processed to have the same coordinate system. The TWD97-2010 coordinate system, which is compatible with the World Geodetic System 84, was used, along with local adjustments for the effects of the 1999 Chi-Chi earthquake and the continuous collision of crustal plates near Taiwan. Feature points that exhibit no obvious change are identified within multi-temporal photos and their coordinates in a 2014 image. A field survey that was carried out in 2015 verifies these coordinates in the 2014 image. The shifting, distortion, and deformation of the images are rectified (referred to as “image modification” hereafter) to enhance their spatial accuracy for the subsequent production of the DSM. Figure 4 shows some feature points (Fig. 2a) in multi-temporal photos. For the 1986 image, for example, the root mean squared error (RMSE) of the horizontal coordinates of the 57 feature points is 1.404 m. Points a10, b2, b4, c10, c7, e1, and e2 clearly deviate from their corresponding points in the 2014 image. Image modification produces an average displacement of feature points of 1.788 m, and reduces the RMSE to 1.036 m.

Fig. 4
figure 4

Feature points in photos taken in different years. a, c, e, g, i Points a1 to a4, b1 to b4, c5 to c10, d1 to d4, and e1 to e4 in photograph taken in 1986 in that order. b, d, f, h, j Points in image taken in 2014

Full size image

Figure 5 presents the effect of image modification using a traverse net that is formed from some feature points (seven points in Fig. 5a). Figure 5 b to f show the locations of these feature points in the 2014 image (black circles) and their original locations (light blue triangles and light red pentagon) and modified locations (dark blue squares and dark red hexagons) in the 1986 photograph and 2001 photograph, respectively. The horizontal shifting of feature points is marked and enlarged. All the feature points in the 1986 and 2001 images had moved toward their corresponding locations in the 2014 image, implying that the used feature points with tiny variations to provide a reference coordinate system effectively overlap the multi-temporal aerial photos in close positions, favoring the interpretation of the landform changes and the production of a DSM for the analysis of topographic variations.

Fig. 5
figure 5

Traverse net formed by feature points (a) and direction and magnitude of displacement of points upon image modification (b to f). Black circle indicates location of point in 2014 image; light blue triangles and light red pentagon indicate locations of that point in original 1986 and 2001 photos, respectively. Dark blue squares and dark red hexagons indicate locations of corresponding point after image modification, respectively

Full size image

Figure 6 shows the topographic changes along two profiles upon image modification. The elevations of these profiles are extracted from the DSMs with and without image modification. The elevation of the slope in the 2001 photo is modified, and the average displacement of the feature points (3.625 m) (Table 1) changes the elevations of the upper road and the lower road from those in 2014 DSM, from 5.34 m and 12.07 m to 4.59 m and 10.94 m, respectively (Fig. 6b). However, the modification of the slope is uneven and insignificantly relative to the 1986 profiles. The resolution of the aerial photograph (around 4 m) and the vegetation account for such results. Nevertheless, image modification reduces the differences in the elevations of the upper road and lower road from 4.54 m and 0.94 m to 4.02 m and 0.74 m, respectively (Fig. 6b). Table 1 presents information about the collected aerial photos and RMSEs before and after image modification.

Fig. 6
figure 6

Locations of two profiles (a) and results of image modification (b and c). Three-dimensional terrain model using aerial photos taken in 1986 (d) and 2014 (e), after image modification, can be produced within precious location accordingly

Full size image
Table 1 RMSE of feature points before and after image modification
Full size table

Highway reroute and maintenance records

The highway maintenance branch office has recorded road disasters and blocking accidents since the 1990s. Possible causes of these disasters and accidents are ordinarily described as uphill slope–related damage and roadbed-related damage. Uphill slope–related damage refers to causes of disasters and accidents that are associated with the slope above the road, which include rock fall, slide, avalanche, or even the debris flow. Roadbed-related damage refers to causes that are related to the slope below the road, such as loss of the roadbed; subgrade, subsoil, or natural ground; subsidence, cracking, or displacement of the pavement; sudden depression or bulging along the longitudinal profile of the road; and damage to retaining walls or drainage down the slope from the road, among others. Rock slides, rock avalanches, and headward erosion and lateral erosion of the river bank below the road account for most roadbed-related damage. Based on the maintenance records, of the road disasters and blocking accidents in the studied Dashaxi section, 67% were caused by uphill slope–related damage, 25% by roadbed-related damage, and the remaining 8%, which were road closures, by construction and other factors.

Figure 2 b shows the locations of recorded road disasters and blocking accidents that were caused by uphill slope–related damage and roadbed-related damage. Types of landslide were identified using interpretation of aerial photos and field checks. Since maintenance records before 1990 were lost, highway rerouting and changes in road structures before that time were also interpreted from aerial photos and are marked in figures.

Field investigation and rock mass rating

Locations and types of landslide that are interpreted from aerial photos are verified by field investigation, and the records are modified if necessary. Large-scale landslides and major maintenance events, such as highway rerouting and structural change, are identified by interviewing with local residents and engineers who participated in the relevant construction and have retired.

This study adopts the Q-method that was proposed by Barton et al. (1974) for rating the rock mass along the Dashaxi section. In the rock mass rating, an outcrop with a width of approximately 10 m is chosen as the evaluation unit, of which ten segments with similar widths of roughly 1 m are rated separately; the average of the ten results is taken as a representative Q value. If some of the ten segments have very different Q values, the Q-method is applied repeatedly for rating the rock masses in the 10 1-m-wide units to the left and right of those segments, and another one whose middle part involving those segments (Fig. 7). Furthermore, the rating for the water inflow and pressure effects (Jw) and the rating for strength/stress ratio in hard massive rocks, for faulting, squeezing, or swelling (SRF), have less of an effect on slope stability than during tunneling. Therefore, the ratio Jw/SRF is ignored and set to 1 in this study.

Fig. 7
figure 7

Outcrop Q value evaluation. Values separately obtained in ten segments at 1-m intervals determine average Q value for outcrop. When Q values in some part of segments are particularly low, values are obtained to the left and right of those segments, and another one whose middle part involved those segments

Full size image

The inferences drawn from the aerial photos are compared with those drawn from the field investigation. The use of Q-method semi-quantitates the mechanical characteristics of rock masses, combined with the interpretation results of multi-temporal aerial photos, and the comparison of the long-term maintenance records of mountain roads, providing valuable information for investigation of the influence of geological conditions on highway maintenance in mountainous areas, and a reference for studying sustainable development maintenance strategy.

Results

Based on the rerouting of the Dashaxi section of the highway, relevant maintenance records, and the locations of landslides interpreted using aerial photos, the study area is divided into three zones to facilitate subsequent discussion. Figure 2 a shows the three zones. The rock mass ratings are obtained using the Q-method. The Q values of 62 evaluation units are in the range 0.14–355.56, with an average of 19.33.

Zone A

Figure 8 shows the highway alignments in zone A of the Dashaxi section over the years. The maintenance history can be classified into four broad phases based on the highway alignments (Figs. 6 and 8). The first phase started with the completion of the highway and continued until around 1980. During this phase, the highway wound along the contour of the slope and suffered from landslide-related problems. In the second phase, 1980–1986, part of the highway in zone A was substantially rerouted, and the sections near Sta. 163.3 k and Sta. 164.3 k with lengths of 300 m and 350 m, respectively, were moved by about 35 m and 80 m toward the hillside, respectively. During the third phase, 1987–1996, the section in the vicinity of Sta. 163.3 k was further moved toward hillside and the structural type was changed from a slope-cut halfway road to the Xibao Tunnel. Approximately 250 m were rerouted. The fourth phase started in 1996 and has continued to the present; during this phase, the structural type of the section near Sta. 164.2 k was changed from a slope-cut halfway road to the Guyuan Tunnel, and the length of the rerouting was around 350 m. The Xibao Tunnel and Guyuan Tunnel were up to 60 m and 70 m from the original locations of the halfway road, respectively. Some traces of the road sections that were built in the second and third phases are still visible in the aerial photos. Zone A can be split into two subzones. Subzone A1 is located near Guyuan, where the highway was rerouted toward hillside from 1980 to1986 (second phase). The major structure of the highway remains the as-original slope-cut halfway road. Subzone A2 is located to the northwest of A1. The Dashaxi section within the subzone has a large longitudinal gradient. During 1986–1996, the slope-cut halfway road was moved into the neighboring mountain, with critical sections made into tunnels.

Fig. 8
figure 8

Highway in zone A of Dashaxi section

Full size image

The locations of landform change or terrain variation, interpreted from aerial photos, were initially sketched and then preliminarily checked using aerial photos subsequently taken to reduce the probability of misinterpretation. Figure 9 shows the interpreted changes in landform and terrain variation in some years. Aerial photos that were taken in 1959 and 2015 and corresponding interpreted results are also given. Areas with significant landform changes and terrain variations were further sketched, and topographical contours were generated using the 2015 DSM (Fig. 10). Simple codes are assigned for comparison in the figure. The results of the field investigation verify the types of these landslides. The Q values in zone A are in the range 0.26–92.19, with an average of 10.76.

Fig. 9
figure 9

Interpretation of aerial photo for zone A. Aerial photos that were taken in 1959 (a) and corresponding interpreted results (b). Interpreted results of 1980 photo with contours from orthophoto base map (c); results displayed using contours extracted from DSM that is produced using corresponding photos taken in 1986 (d), 1996 (e), 2009 (f), and 2014 (g). Aerial photos that were taken in 2015 (i) and corresponding interpreted results (h) are also given

Full size image
Fig. 10
figure 10

Variations of terrain in zone A. Contours extracted from DSM, produced using 2015 aerial photo. L denotes landslide, G denotes gully development, LE denotes lateral erosion and undercut slope along a river bank, and S denotes pile-up along a river bank, including a slip-off slope

Full size image

Figure 11 shows the inferred times of landform changes. In the initial stage after completion of the Tai-8 highway, a large proportion of bare slopes was observed and interpreted as shallow slope failures or debris on slopes (Fig. 9a). The slope was cut by blasting without careful consideration of blast holes drilling and charges, and the excavated debris was arbitrarily piled up on the hillside below the road and even dumped down the slopes, affecting the stability of the slope on both sides of the road. In the vicinity of subzone A1, LA5 was distributed between the upper road and the lower road (Fig. 9b). LA3, LA2, and LA1 were located along the undercut slope of the left curved bank. LA2 was significant in the 1980 aerial photo (Fig. 9c), which shows debris following the collapse piled up within the Dasha River, causing the river to become narrow to a minimum width of approximately 20 m. Where a slip-off slope had been located, the water flow turned to wash the right bank, causing significant lateral erosion. GA1 near the boundary between subzones A1 and A2 developed upward from the river side, reaching above the upper road (Fig. 9c). The highway in subzone A1 was rerouted to its current course during 1980–1986 (Fig. 9d). The interpreted LA4 during the initial stage after completion of the Tai-8 highway was completely exposed. The landslide developed upward above the upper road, with debris piled up to the other side (left bank) of the Dasha River. Several years later, except for the enlargement of LA2, the slopes in subzone A1 were relatively stable and vegetation recovered. The nearby river was obviously widened in 1996 due to the reactivating of LA1 and its length and width increased to 400 m and 100 m, respectively (Fig. 9e). Reactivating of LA1 was observed in the 2000–2003 images. Although the highway was rerouted far from the river and was insignificantly affected by the reactivating of LA1, the collapsed debris piled up within the river temporally, leading to the development of a new gully (GA2) from the side of the river. The gully quickly eroded upward, meting GA1 around 2005 (Fig. 9f–h).

Fig. 11
figure 11

Diachronic variation of interpreted landslides and gullies in zone A

Full size image

LA4 and the gullies significantly affected the highway in subzone A2. LA4 was distributed widely along the highway following its construction, with two gullies north of its (Fig. 9b). Gully GA1 continuously developed and formed a boundary on the downstream side of LA4. While LA4 shrank in downslope side of the lower road, GA1 developed 10 m above the upper road. The debris of LA2 piled up in the river and worsened the lateral erosion of the side of the upper stream. LA4 reactivated in 1986 and became significantly larger (Fig. 9d). The bare slope near LA4 was steep with debris scattered along the highway. It is believed that the 1986 photo was taken during the highway repair work following the collapse. LA4 became larger again in 1987, causing the roadbed of the upper road to disappear completely. Around 1987, the upper road was rebuilt in a neighboring mountain, passing through this section in the Xibao Tunnel. Reactivating of LA4 continued until 1991. During this period, the lower road was rerouted toward hillside and passed through the Guyuan Tunnel. Even after the highway was rerouted, LA4 was reactivated intermittently. GA1 met GA2 in 2005. An opened crack on the vault of the unlined tunnel at Sta. 163 between GA1 and GA2 was reported in 2013 (Fig. 12). The recovery of vegetation is observed along GA2 after mitigation was carried out near the lower road in 2012. However, the scope of bare slope near the area where these gullies met has gradually become larger since then.

Fig. 12
figure 12

Images of unlined tunnel near Sta. 163 k. Opened fracture observed on tunnel vault (a) and developed to slope surface above tunnel (b)

Full size image

Zone B

Figure 13 shows the highway alignments in zone B over the years. Except for the sections between Sta. 165.4 k and Sta. 165.9 k, the highway alignment has varied insignificantly (Fig. 13a). The highway alignment between Sta. 165.4 k and Sta. 165.9 k occurred in three phases in zone B. The first phase started with the completion of the highway and ended around 1964. During this period, the highway wound along slope contours, except for the section between Sta. 165.8 k and Sta. 165.9 k, where an old tunnel with a width of about 4.2 m and a length of around 70 m had been built. This tunnel was rebuilt twice during the second phase, 1964–1993, according to maintenance records. The first rebuilding shifted the section between Sta. 165.7 k and 165.8 k toward the hillside by up to 40 m horizontally, and the slope-cut halfway road section was replaced by a tunnel during 1964–1970. The second rebuilding during 1991–1993 widened the existing tunnel to 7.5 m and rerouted the highway northwest of the tunnel into the current Taishan Tunnel with a length of 304 m. The maximum horizontal movement was 45 m. The third phase of highway alignment continues from 1993 to the present. The alignment remains similar to that at the end of the second phase, except that some parts were widened or modified to increase the horizontal curvature enlargement along the original course.

Fig. 13
figure 13

Variations of highway alignment in zone B (a). Topographic feature varied in subzones B1 (b) and B2 (c)

Full size image

The gully is the most significant morphological features that are observed in aerial photos. In the first stage after completion of the highway, six gullies, GB1 to GB6, were present between the side of the Dasha River and the upper road (Fig. 13b, c). GB2 and GB3 developed southwestward from the sharp turning point of the Dasha River near Sta. 165.6 k of the Tai-8 Highway, and the other four gullies developed northwestward and were approximately perpendicular to the boundary of the strata in the study area. In Fig. 13 b, the slope in the vicinity of the intersection of highway in the first phase with GB3 and the original tunnel is high and almost overhangs to form cliff, and frequent rock falls have been reported. The modification of this section into a tunnel was once suggested following the completion of the Tai-8 highway (Lee 2011). The aerial photos show frequent collapses close to the intersections of the highway with GB2 and GB3, among which the landslide LB2 was insignificant in 1986 and appeared only in the slope toe near the right bank of the Dasha River in 1987. LB2 was larger in the next six aerial photos (1988–2001), resulting in the rerouting of the nearby slope-cut halfway road into a tunnel in the neighboring mountain. Landslide LB1 was observed after the completion of the highway but was not visible from 1980 to 2013, because of the recovery of vegetation. The reactivating of LB1 was detected near the bottom of gully GB1, and a corresponding countermeasure was performed in 2015.

In Fig. 13 c, GB4 to GB6 are markedly longer than GB1 to GB3. Collapses are identified on the slopes along both sides of GB4 to GB6 in successive aerial photos. The spatial distributions of these collapses are typically northeast-southwest. Landslide LB4, which occurred in the head area of GB5 in 1964, is representative. Since mitigation measures were taken in response to LB4, its activity was far less than that of LB2 and the slope near LB4 has been stable for more than five decades. In contrast, in the vicinity of the intersection of GB4 with the Dasha River, the river bank has gradually shrunk over a long period and has exhibited features of potential collapse in recent aerial photos. Figure 14 shows the diachronic variation of interpreted landslides and gullies in zone B. The Q values in zone B are in the range 2.38–25.10, with an average of 7.05.

Fig. 14
figure 14

Diachronic variation of interpreted landslides and gullies in zone B

Full size image

Zone C

Figure 15 shows the highway alignments in zone C. The alignment of the Tai-8 highway as built in the 1960s had not changed, and the road winds along the slope contours, but the width of the highway and its structural type in some sections have changed. Figure 15 b plots the profile near Sta. 167.8 k, where the Dasha River flows southward, making a short turn to flow northeast-eastward. Highway maintenance is divided into three phases, with different widths and structures. In the first phase following the completion of the highway to 1986, the road was mostly 5–6 m wide. The Tianxiang Tunnel near Sta. 167.8 k, almost perpendicular to the Tai-8 highway, was built in 1986. The Tai-8 highway in the vicinity of Sta. 167.8 k was then widened to approximately 8 m. The second phase started in 1986, when the hillside slope of the highway was cut to further widen the road to over 10 m, and limited slope protection measures were used for the excavation. In 2001, gallery tunnel with a length of around 180 m was built close to the sharp turning point of the Dasha River. The gallery tunnel was designed using a conventional method with a foundation depth of 1.5–2.0 m. The hillside slope had to be cut to build the foundation (the enlarged view of Fig. 15b). This gallery tunnel was damaged in Sep. 2008 by collapse of the slope (Fig. 16a). Repairing work was promptly carried out at the end of 2008. However, more than 40 m of the gallery tunnel was destroyed by another severe slope in Oct. 2009 (Fig. 16b). The tunnel was reconstructed and extended to 270 m, which is its length today. After the reconstruction work had been completed, highway maintenance entered the third phase, which continues to today.

Fig. 15
figure 15

Variations of terrain in zone C (a) and profile of EE’ (b)

Full size image
Fig. 16
figure 16

Gallery tunnel successively damaged in Sep. 2008 (a) and Oct. 2009 (b)

Full size image

Gully GC1 is visible in the aerial photo before highway construction and was larger immediately after completion of highway but could not be seen in 1986 since the recovery of vegetation. The bare slope near the construction site of Tianxiang Tunnel at Sta. 167.8 k, visible in 1986, was invisible in the 1988 photo. Landslide LC2 could be seen in 1991 because of the nearby construction of a parking area at the riverside. Local drainages were modified accordingly, causing the lower part of gully GC2 to be gradually shifted from the northern boundary of LC2 southward to its middle area. The highway in zone C was widened in 1996 by cutting into the hillside slope. Local collapses can be seen near the cut slope in the 1998 photo. The collapse near GC1, which developed upward, was significant, reaching the local hill ridge in 2000. In the 2001 aerial photo, the gallery tunnel was wider than the highway. The hillside slope in the gallery tunnel had been cut to maintain the width of the road and to enable the construction of the gallery foundation. The bare slope along GC1 was enlarged and interpreted as landslide LC1 or the reactivating of piled debris along GC1. LC1 was larger in 2003 and again in 2009. Meanwhile, two new gullies developed from the middle part of the slope and two landslides (LC1A and LC1B) were observed; gullies GC3 and GC4 developed upward to their corresponding upper cliffs. The landslide in Oct. 2009 was large and merged with LC3, LC4, and their neighboring area to form the huge LC5. Mitigation measures were completed at the end of 2012. However, new bare slopes above the gallery tunnel and close to the hill ridge are observed in the 2019 image, in which debris was scattered in GC2 and had moved downward. Figure 17 shows the diachronic variation of interpreted landslides and gullies in zone C. The Q values in zone C are in the range 0.14–355.56, with an average of 21.98.

Fig. 17
figure 17

Diachronic variation of interpreted landslides and gullies in zone C

Full size image

Discussion

The rerouted sections of the highway in subzones A1 and B1 and the section with the modified structure in zone C are all neighbored to the undercut slopes along the meanders of the Dasha River. Gullies had developed upward from the river bank before the construction of the Tai-8 highway. After the completion of the highway, slopes frequently collapsed close to the gullies that had developed near the undercut slopes of the Dasha River. Obviously, the rerouting and changing of the structure of the Dashaxi section of the highway related to neighboring slope collapses, which were further influenced by the development of gullies and erosion by flowing water in the meanders of the Dasha River.

The rerouted section toward hillside in subzone A2 locates in the slip-off slope of the Dasha River, that is different from the aforementioned subzones of the neighboring undercut slopes. The landslide in the upper slope of this section was not significant. The highway was rerouted in this subzone because of the loss of its lower slope as a result of bank erosion that occurred when the direction of water flow was changed by the mass of debris in the river. This debris was associated with the large landslide on the opposite side of the Dasha River in the 1960s, where an undercut slope was also present. Riverside undercut slopes in the study area clearly affect the stability of their hillside slopes by lateral erosion of water flow and the development of accelerating gullies. When a landslide is large enough to leave debris in the river that will also change the direction of water flow.

The gullies were distributed mostly in the northeast-southwest direction and, subsequently, in the northwest-southeast direction. The northeast-southwest gullies typically develop along various lithological boundaries and are short, steep, and densely distributed. Turowski et al. (2008) indicated that in addition to the regional anticlinoria and synclinoria, river erosion and the variability of flow influence the development of gullies. In contrast, the northwest-southeast gullies are relatively long and gentle. GB5 and GB6 are representative of the northwest-southeast gullies, which are thought to be associated with the fractural structures that are perpendicular to the regional anticlinoria and synclinoria (Lo 1992).

The effects of river erosion and scouring on slope stability are determining by examining the collapses in the vicinity of GB2 and GB5. GB2 is close to an undercut slope of the Dasha River. Successive collapses near the GB2, visible in the multi-temporal aerial photos, typically had moderate scales. However, such collapses inhibited the maintenance of the slope-cut halfway road until the highway was rerouted toward hillside and through the Taishan Tunnel. The meander where GB5 intersects with the Dasha River had a large radius and was wide, with relatively little bank erosion, and insignificantly affected the development of GB5. The slope close to the head area of GB5 slid more than 100 m wide and 200 m long in the 1960s. However, it has been stable for more than five decades after mitigation. Obviously, the stability of the slopes near the head areas of the gullies in the study area can be maintained for a relatively long time when an exogenic force acting on the gully, for example the river erosion and scouring, is not significant. Strong erosion and scouring of the river bank account for the fact that the development of the northeast-southwest gully GA1 in the subzone A2 is much faster than that of the northwest-southeast gully GA2. Accordingly, the mitigation measures associated with GA2 were only carried out in its lower part, effectively slowing the head-ward development of GA2 in the past 7 years.

The distribution of the Q values of rock masses along the Dashaxi section provides further information about factors that may affect the development of gullies and landslides. The Q values along the highway in the study area are in the range 0.14–355.56 representing a variation of about 2540 times. However, the Q values near both sides of the gullies or lithological boundaries are interesting. Figure 18 a shows the lithology that is exposed along the highway between Sta. 159.4 k and Sta. 160.2 k and the obtained Q values. Figure 18 b shows the distribution of Q values perpendicular to the boundaries of strata: the first, second, and third quantiles of the ten Q values for each outcrop, based on an interval of about 1 m, are plotted. The Q values that are obtained in a stratum are mutually inconsistent, and the ten Q values for each outcrop vary much more. Such variations are most significant in the vicinity of strata boundaries or even the boundaries between various lithologies in a stratum. Figure 7 shows a representative result of Q-method from an outcrop near a boundary of strata. The average Q values of the rock masses to the left and right of the boundary are 85.28 and 46.73, respectively. However, the Q values of the 6 units involving the boundary are in the range 0.35–3.42, with an average of 1.20. Among the six rated items of the Q-method, the rated results for RQD and Jn vary obviously near the lithological boundaries.

Fig. 18
figure 18

Locations where Q values are obtained between Sta. 159.4 k and Sta. 160.2 k (a) and obtained Q values of rock masses (b)

Full size image

Figure 19 b plots the variation of Q values in different strata along the direction that are perpendicular to the strata boundaries (Fig. 19a). The Q values of the rock masses in the upper road are typically greater than those in the lower road. Again, the Q values which vary significantly are generally close to the boundaries of strata or lithological boundaries in a stratum. Among the four major strata that were exposed in this study, Tianchang Marble has the highest Q values, in the range 0.14–355.56, with an average of 30.52. The Q values of Paiyang Schist, Pilu Formation, and Guyuan Schist are in the ranges 0.27–124.14, 0.26–92.19, and 2.29–16.67, respectively, with their respective averages of 17.64, 13.06, and 6.38 (Fig. 19c, d). The Q value that is evaluated using an interval of 1 m can be as low as 0.14 close to the boundary between Tianchang Marble and Paiyang Schist. The Tianchang Marble stratum exhibits the greatest variation in Q values. The Q values close to the section Sta. 166.5 k are lower than those in the vicinity of Sta. 159.4 k, Sta. 159.8 k, and Sta. 167.7 k (all marked in Fig. 19a). The rock masses in the lower road were significantly affected by river erosion and scouring, and the Tianchang Marble was relatively thin near Sta. 166.5 k, making the rock masses of the Marble vulnerable to fracture under strong tectonic folding, accounting for the lower Q values in this section. Additionally, the average Q values near the axes of the Guyuan anticlinorium and the Tianxiang synclinorium are low, and those evaluated using 1-m intervals vary negligibly. With the exception of the LA2 in subzone A2 near the axis of the Guyuan anticlinorium, few landsides were observed near the axes of the anticlinorium and synclinorium. The impact of geological folds on landslides is not as significant as that of river erosion, scouring, or gully development.

Fig. 19
figure 19

Locations where Q values are evaluated in the study area (a) and obtained Q values of rock masses (b). Q values of rock in various strata (c) and at various locations in Tianchang Marble (d)

Full size image

Various metamorphic rocks in the study area have been deformed into a number of open, asymmetrical, or overturned folds and intricate flow folds as a result of multiple phases of orogenic deformation and metamorphism. The studied section of the Tai-8 highway which winds around the slope of the right bank of the Dasha River is downgraded and crosses the axes of the Guyuan anticlinorium and Tianxiang synclinorium. The strikes of the strata and the cleavage of the lithology are typically northeast-southwest and close to the plunge of the fold axes. The strata have various lithologies with different competences of rocks, and they sustain a wide range of degrees of deformation under tectonic and regional geological activities, causing the rock that is exposed in the slope to exhibit fractures with various appearances. The geometric characteristics of the fractures, such as their spacing, persistency, and roughness, vary substantially. Therefore, the rating of RQD and Jn for the Q-method is low on the boundaries among the strata or the lithological boundaries in a stratum. The Dasha River causes erosion and scouring, deteriorates nearby rock formations associated with other factors, and forms steep slopes along its bank. Cut slopes are inevitably formed in the construction of a mountain road and are protected using artificial countermeasures, which may suffice in the short term to mitigate landslides as a result of rainfall, earthquakes, and water scouring. After many years, sections with unfavorable engineering characteristics fail to withstand long-term river erosion, scouring, and gully development. Rerouting the highway toward the hillside and modifying is structural type from a slope-cut halfway road to a tunnel to reduce the effects of these strong exogenic forces are effective in the study area, according to the few maintenance records from the last two decades. A gallery tunnel is effective for mitigating rock fall from a hillslope of a highway. The overcutting of a roadside slope should be avoided during the construction of a gallery tunnel to prevent artificial excavation-induced instability.

Conclusion

This study concerns the section of the Tai-8 highway that winds along the right bank slope of the Dasha River, and establishes inventories of highway alignments and nearby landslides from multi-temporal aerial photo and highway maintenance records. Types of landslide and topographic features are verified by field investigations, and the engineering characteristics of rock masses that are exposed on outcrops along the highway are evaluated using the Q-method semi-quantitatively. The effects of engineering geological conditions on the maintenance and conservations of mountain roads, and the related effectiveness of maintenance strategies in mitigating neighboring slope collapses, gully development, and river erosion and scouring, are investigated. The following conclusions are drawn:

  1. 1.

    Multi-temporal aerial photos and highway maintenance records indicate that, following the completion of the 9.4-km-long Dashaxi section of the Tai-8 highway in 1960, three sections with a total length of about 1.9 km were rerouted or their structural types were modified. These three sections are all close to the undercut slopes of the Dasha River, and the boundaries among strata where water erosion and scouring are active and gullies develop strongly.

  2. 2.

    The Q values of rock masses in the lower road are typically less than those in the upper road. The Q values near the boundaries of the Tianchang Marble, Paiyang Schist, Pilu Formation, and Guyuan Schist strata and those of the lithological boundaries in each stratum are generally small. In addition, rock mass with a low Q value is observed in the thin section of competent Tianchang Marble that is interbedded in relatively incompetent Guyuan Schist.

  3. 3.

    The aforementioned strata have various lithologies with different competences of rocks and sustain a wide range of degrees of deformation under tectonic and regional geological activities. The differential erosion that is caused by the various competences of the strata generates meanders in the Dasha River with steep slopes. Rock masses in the vicinity of strata boundaries and lithological boundaries in each stratum have a large range of Q values, corresponding to a large variation in their engineering characteristics. The cut slope and associated protection measures suffice in the short term to maintain stability under the effects of rainfall, earthquakes, and water erosion and scouring. However, after many years, the sections with unfavorable engineering characteristics eventually fail to withstand river erosion and scouring and gully development.

  4. 4.

    Based on the few maintenance records from the last two decades, rerouting the highway toward the hillside and changing its structural type from a slope-cut halfway road to a tunnel to reduce the effects of strong exogenic forces on highway are effective in the study area. A gallery tunnel mitigates the impacts of rock fall. Its construction should not cause overcut-induced slope instability, especially where water erosion and scouring are strong.