1. Introduction
Creating reservoirs is known to develop and support several industries: i.e., electricity, agriculture, water transport, fisheries, recreation, and tourism. At the same time, their creation leads to a fundamental restructuring of natural systems: river water flow and regime, bedload, and suspended sediment yield are changed, a unique microclimate is created and landscapes are transformed [
1,
2,
3,
4]. The intensity of exogenous processes acting on the shore and banks increases by orders of magnitude due to the water surface area increase. This very complex relief-forming process (together with erosion and gravitational processes) is often called bank transformation when dealing with water reservoirs. Among all modified shores of water reservoirs in Russia, approximately 78% are destroyed by abrasion type, and the remaining 22%—by abrasion-landslide, abrasion-karst, and other types of erosion processes [
5,
6]. When the banks are destroyed, a large sediment volume enters the reservoir. This leads to rapid siltation, reduction of water depths, and environmental degradation of the reservoir.
Exogenous geomorphic processes and their monitoring on the shores of water reservoirs have long been conducted [
7]. A large number of different research methods have been used for this purpose [
8]. Most often, the shoreline bank dynamics are evaluated, and there is limited quantitative data on the different exogenous processes, their intensity, and the particularities of soil transport on slopes. This is mainly due to poor field accessibility for making observations at the shoreline banks and for installing a reference point network (steep banks, high water level near the shore, bank collapse, crumbling, and landsliding). New opportunities for assessing the rate of shoreline modification by exogenous processes, particularly for shoreline areas that previously were difficult to access have emerged with the use of multi-source data (laser scanners, UAVs, remote sensing, etc.). Simultaneously, a large diversity of water reservoir types and natural-anthropogenic conditions shaping their banks has resulted in a specific spectrum of exogenic processes. This makes data generalization and spatial extrapolation difficult.
The objective of this study is to better understand the shoreline and bank dynamics of the Kuibyshev reservoir—one of the largest reservoirs in the world. Research focuses on shoreline abrasion and landslide processes as the most significant exogenous natural hazards of bank modification. Attention to these processes has increased because a significant number of settlements, religious buildings, cattle burial grounds, and infrastructure (oil and gas pipelines, roads, power lines, etc.) are located on the reservoir shores. The shoreline abrasion and landslide processes on the banks near settlements often lead to emergencies: i.e., destruction of buildings, infrastructures, communications, disturbance, and total loss of (agricultural) lands bordering the reservoir. Shore transformation processes are still going on at a high rate even though the engineering project predicted the attenuation of these processes in 20 years after the reservoir creation. In this connection, it is necessary to monitor these dangerous exogenous geomorphic processes. The objective of this study is therefore to assess quantitatively the rates of shoreline abrasion and landslide deformation rates using an integrated approach that involves different monitoring techniques. The study sites for shoreline process dynamics were selected by the Ministry of Natural Resources and Ecology of the Russian Federation and the Federal Agency for Water Resources. These government agencies identified areas of the Kuibyshev reservoir banks with the most significant exogenous natural hazards for human settlements and infrastructure.
1.1. Research Methods for Landslide Processes
At present, there is a large spectrum of methods for monitoring and estimation the intensity of exogenous processes [
5,
9,
10,
11,
12,
13,
14]. These mainly consist of collecting instrumental topographic surveys and remote sensing data, allowing the study of slope processes over several decades. In order to assess the rates of geomorphic processes occurring on the reservoir shores, this study applied alongside traditional methods (topographic field surveys) a set of modern methods: i.e., terrestrial laser scanning (TLS), unmanned aerial vehicle UAV, remote sensing, and a global navigation satellite system (GNSS). These techniques allow a quick recording with high accuracy for determining the shore process rates and displaced material volumes by landslides [
15]. After a critical evaluation of the various methods and a description of the study area, we report on the methods applied in this study.
1.1.1. Traditional Geodetic Methods for Studying Landslide Processes
One of the most widely used and well-known landslide processes monitoring method is the ground survey with total stations and GNSS receivers. This approach is useful for accurately measuring the position of individual points in the field. The use of traditional theodolite and total station measurements can achieve millimeter-scale errors when determining point coordinates [
9]. Total stations have been used to create a field of points with the subsequent creation of a digital model of the landslide morphology [
10] and to monitor the landslide movement [
16]. Generally, a relative coordinate system is used for the survey which creates a problem when comparing such data with geo-referenced data sources. It is therefore necessary to recalculate the position of the field points to allow plotting these into the global coordinate system by the specific position of the reference points [
13].
The use of permanent control points allows measuring the dynamics of the landslide process with millimeter-scale accuracy. Similar error rates are provided by permanent robotic geodetic control systems on sites with high landslide process dynamics or special hazards to human life or buildings and structures [
17]. The advantage of this surveying method is its high accuracy. The disadvantage is that the surveys take a long time; therefore, a limited number of points located in accessible places are selected. The development of TLS, UAVs, allowing to take large amounts of data in a short time, precisely and in detail describing the surveyed surface, makes the classical topographic survey an outdated approach [
18].
Limited control of processes operating within the landslides and the displaced ground masses are the most significant disadvantages of traditional approaches. None of these methods provides comprehensive information on quantitative characteristics and mechanisms of erosion processes on slopes. Thus, the most effective methods for studying inaccessible slopes with landslides are the repeated morphological surveys with accurate referencing of multi-temporal data.
1.1.2. Ground-Based Laser Scanning
High accuracy and density of point clouds obtained with ground-based scanning permits the obtaining of highly detailed models of local landslide areas [
11]. Compared with traditional methods (aerial photography interpretation, ground photogrammetric survey, field studies), laser scanning has several advantages. These include the possibility of studying even small landslides (up to several m
2) and landslide bodies of different ages [
19] and investigating dangerous or inaccessible areas with direct topographic measurements. Very precise results make it possible to estimate the intensity of slope processes and estimate the age of landslide bodies [
20,
21], to build detailed three-dimensional models and evaluate the slope stability [
22].
Laser survey technologies are often used to solve practical problems. For example, for monitoring of landslides that threaten the safe use of transport infrastructure. Various techniques have been used for such monitoring, in particular, the “Stop and Go” approach, which involves the installation of a scanner on a mobile platform [
23] or the technique of spatial data referencing, based on the weighted transformation of parameters to obtain more accurate data [
24].
One of the essential uses of ground-based laser scanning is to study the dynamics of slope processes. Morphometric and structural analysis of multi-temporal TLS data allows obtaining spatially distributed characteristics of velocity, direction, and volume of slope-displaced material [
25,
26]. The study of relief dynamics requires repeated measurements of the object under study and their comparison with previous data, which requires geodetically accurate referencing provided by modern ground-based laser scanning technologies [
27,
28]. To achieve high accuracy of multitemporal scans adjustments different approaches have been applied: least-squares [
29], Iterative Closest Point (ICP) [
30] differential global positioning system (DGPS) [
31], relative coordinate system [
32,
33,
34]. One centimeter is an acceptable error for studying landslide processes due to the large volume changes occurring on the affected slopes [
35].
The need to survey from several positions to provide full coverage of the slope surface is one of the disadvantages of ground laser scanning when studying landslide processes on the nearshore slopes. In the case of the small beach width, this is difficult and often impossible. Given the inaccessibility of bank slopes, more convenient survey methods are used, such as mobile laser scanning from a ship, which allows studying extended areas from the water at small waves [
32]. A combination of high-resolution remote sensing data and several ground-based methods (Ground-Based Interferometric Synthetic Aperture Radar, TLS, and InfraRed Thermography) is effective as part of long-term monitoring and emergency management [
15].
Studies of landslides near water bodies are mainly carried out on the coasts of seas and oceans [
36]. Few scientific studies have been devoted to the monitoring of hazardous processes on slopes bordering reservoirs [
3] Several publications on the Three Gorges Reservoir in China, created in 2003, stand out [
37,
38,
39]. Here, landslides have caused significant damage to infrastructures on the shore and have threatened the life of the local population. In contrast, in Russia, there are many studies devoted to studying geomorphic processes on the banks of large reservoirs, because of the large-scale construction of hydroelectric power plants in the middle of the last century [
40,
41,
42,
43,
44,
45].
1.1.3. The Use of UAVs
Even though many sensors are currently available on ground, airborne, and space-based platforms, methods using unmanned aerial vehicles are rapidly developing among all modern methods for exogenous processes monitoring [
46,
47]. The possibility of obtaining three-dimensional (3D) information about the terrain with high accuracy and spatial resolution opens up new horizons for studying landslide processes [
48,
49]. Further development of technologies such as robotics, photogrammetry [
50] and, in particular, Structure-from-Motion (SfM) and computer vision [
51], as well as GNSS, has given a rise to a whole range of studies using UAVs for research in the field of dynamic geomorphology [
52,
53,
54,
55,
56,
57].
Particular attention should be given to methodological and technical works in which UAVs are compared with other, already proven methods. For example, when comparing models obtained by TLS and UAV with the results of total station surveys, in grass vegetation covers, the UAV gives a model even more accurate than the TLS [
58,
59]. If the survey methodology is followed, UAVs’ use allows one to survey and create a DEM with similar accuracy as the TLS on any landscape [
60,
61,
62,
63].
It is essential to consider GCP [
64,
65,
66,
67,
68] and GNSS), the presence of which ensures the achievement of high modeling accuracy. As noted in Turner et al. [
69], the non-use of ground control points resulted in a 6–10-fold increase in model accuracy relative to studies in which ground control points were used. The newest GNSS-receivers allow achieving a positioning error not exceeding 3–5 mm in height, which is sufficient even for microrelief reconstruction [
70,
71].
2. Study Area
The Kuibyshev reservoir, built on the Volga River, is the largest in Eurasia and the third-largest in the world, after Volta (Ghana) and Smallwood (Canada) reservoirs. It is located in the central part of the Middle Volga basin at the intersection of the forest and forest-steppe landscape zones of the Volga uplands and the Low Volga [
72] (
Figure 1). The geographic coordinates of the reservoir’s borders are 56°10′–53°30′ N, 47°30′–49°30′ E. There are three orographic regions within its basin: Volga Upland, Vyatka Predkamye, Low Volga. The research was conducted within the Volga Upland on the right side of the reservoir; its banks are steep and cliffy. Absolute heights are 175–215 m in the north and 250–270 m in the south. The maximum altitudes are confined to the Zhigulevskiy Mountains (349 m). The climate is characterized as temperate continental, with relatively warm and humid summers, cool and moderately snowy winters. The annual precipitation depth is 500–540 mm, in the warm season reaching 340–360 mm. The snow cover height is 30–37 cm, with a water equivalent of 90–95 mm.
The reservoir was formed on 31 October 1955, following the Volga River damming by Kuibyshev hydroelectric complex. The reservoir reached its full reservoir level (FRL) of 53 m a.s.l. during the 1957 high waters. Its total capacity is 57.3 km3, its water surface area equals 6150 km2 and its total length along the Volga River is 510 km and 280 km along the Kama River. Its width varies between 2 and 27 km with a maximum of 38 km at Kamskoe Ustye. The average water depth is 9.4 m, maximum is 41 m. The length of the coastline is 2604 km, minimum navigation level is 49.00 m. The reservoir is a seasonal flow regulation storage: the average annual conditional coefficient of water exchange is 4.3.
There are 79 rivers more than 10 km long and 260 rivers less than 10 km long flowing into the reservoir. The Volga, Kama, and Vyatka rivers, which directly form the reservoir’s water mass, belong to the rivers with predominant snowmelt feeding. Surface and groundwater produce 99% and precipitation 1% of the water input into the reservoir. The water balance components determine the annual water level fluctuation in the reservoir. The inflow and outflow discharges are asynchronous, and their ratio determines the reservoir filling and drawdown. The reservoir fills up to the maximum level during spring floods, while in autumn and winter the level is at its lowest position. The annual amplitude of level fluctuations is about 6 m. The average runoff velocity in the reservoir is 2–10 cm/s, depending on the value of transit flow and the live cross-sectional area. The morphological structure of the reservoir is a system with lake-like expansions. This reservoir serves several economic sectors: i.e., energy, water transport, agriculture and fisheries, industrial and municipal water supply.
Two sites on the right bank of the Kuibyshev reservoir near Kamskoe Ustye village (Republic of Tatarstan) were selected to evaluate the intensity of shoreline abrasion and landsliding. Right-bank slopes of the Volga River have scars of ancient landslides. A significant rise in the water level due to the reservoir creation triggered the reactivation of landslide movements. The sites were selected based on their representativeness in terms of geological-geomorphological and landscape conditions for the study area. In addition, the economic aspect was considered since there is a threat of partial destruction of the buildings and infrastructure facilities according to the state authorities of the Russian Federation.
At site 1 not only landsliding but also abrasion processes at the shoreline have developed. The site is located on the right bank 30 m upstream of the Volga River pier (
Figure 2). The landslide and abrasion process acting on the bank are observed here. In the upper part of the slope, composed of heavy and middle deluvial-solifluction loams, collapse and sliding of earth blocks of different volumes towards the shoreline occur. At the foot of the abrasion ledge in some places, there are outcrops of the Kazanian age deposits, i.e., clayey-melty gray pack of rocks of the upper Permian system.
The earth slide type landslide [
73] has a frontal shape, its length is 32 m, and its width is 55 m (total area is 1760 m
2). An abrasion scarp has formed in the lower part of the slope with traces of landslides and washouts. In a slope profile, the landslide body itself is well identified in the upper part with a head scarp and in the lower part a terrace-shape section. The beach is 2.5–5.0 m wide and is composed of sand-and-shingle material. The abrasion section at the lower part of the slope erodes and collapses; soil blocks slide through cracks to the shoreline where they are eroded by water. Wave breaking caves are practically absent. Other exogenous processes are acting on the slope in addition to the dominant landslide process. There are rills and ephemeral gullies in the upper part of the slope and gravitational processes (collapse and crumbling) in the middle and lower part. However, they are of subordinate importance in terms of active slope processes.
At site 2, a large landslide of earth slide type [
73] has formed as a result of mass movement processes. Here, the landslide-abrasion type of coastal escarpment is observed. Due to changes in groundwater outflow caused by water level rise, a sliding landslide develops (
Figure 3), forming a large landslide cirque. The upper part of the landslide process led to the destruction of the old cemetery. Landslide length is 173 m, and its width is 110 m (total area is 13,900 m
2). The height of the landslide edge is 14.7 m; the steepness is 90 degrees. The body of the landslide is hilly and canopy-shaped. The upper landslide scarp is close to the buildings.
The lower northern part of the landslide body is affected by very active slope subsidence of the earth block type. The slope is composed of deluvial-solifluction loams with the vertical type of clastic-block structure; blocks with shrub and woody vegetation are collapsing along the cracks.
The height of the blocks is 5–6 m, width up to 7–8 m. The soil surface of the old cemetery is deformed due to cracking and subsidence. The landslide bodies are located further down the slope, moving towards the Volga at low speed; their hilly surfaces are overgrown with willow, American maple, saltbush, common nettle, and coltsfoot. Behind the edge of this landslide body is the landslide cirque. Its southwestern cliff destroys the old cemetery. Numerous humans remains emerge from a depth of 1.5–2 m in the landslide escarpment. In the lower part of the slope, an abrasion scarp with traces of breakwater caves can be observed.
In 2018 to the south of the study Site 2, a shore reconstruction was carried out to organize a recreational area near the camping base (
Figure 4). As a result of excavation works and re-organization of access roads to the camping base pier, the south-eastern part of the studied landslide cirque and landslide body was leveled and sodded (
Figure 4B). As a result, this fragment of the landslide ledge was excluded from further analysis. However, the works carried out to improve the sloping area did not reduce the intensity of geomorphic processes, and the works on slope stabilization had to continue.
3. Materials and Methods
Various sources of information and observation methods were combined to estimate the intensity of abrasion and landslide processes at the reservoir shores. Shoreline positioning for the past periods (1958, 1985, 1987, and 1993) was obtained from archived historical aerial images (Kazan University Library); data for 1975, 2010, were obtained from satellite imagery interpretation. Field measurements at the study sites were conducted in 2002, 2003, 2005, and 2006 using Trimble M3 total station (2014). Since traditional methods mainly provide information about the planar displacement of the landslide scarp edge, the laser scanning method was additionally used in 2012–2014 to study the volume changes in hard-to-reach and hazardous areas. A Trimble GX TLS used as scanning equipment (
Figure 5,
Table 1). From 2019 onwards, field surveys include a DJI Phantom 4 UAV survey.
Since the slope processes at the studied sites develop differently—a block landslide at Site 1 and a landslide cirque at Site 2, the methodological approaches for their study were also different.
3.1. Site 1
At Site 1, an analysis of the abrasion-slope DEM obtained from ground-based laser scanning in 2012–2014 and UAV imagery in 2019 was chosen as the main method of shoreline deformation (
Table 2).
The results of scanning were processed in the Trimble Realworks software. The obtained point clouds were transformed to one coordinate system; next, the scans were cleaned from artifacts. The obtained point clouds were exported to the Golden Software Surfer software, whereby triangulation with linear interpolation of the DEM with a spatial resolution of 0.1 m was created. The calculation of volume change was performed using the Volume tool. In addition, qualitative analysis—creating maps of surface changes and analysis of landslide slope profile changes—was performed here. According to laser scanning results, a specific indicator such as the volume of eroded soil per unit area (V− S−1, 0.03–0.04 m3 m−2) was used to estimate the intensity of landslide processes. Calculation of soil volumes between the scanned surfaces was carried out in automatic mode.
Photogrammetric processing of UAV survey results was performed in Agisoft Photoscan software. A combined approach was used to align the point clouds obtained by TLS and UAV surveys. Since the main survey using terrestrial laser scanning was carried out in the relative coordinate system and the UAV data in the global coordinate system, it was decided to recalculate the latter in the relative coordinate system. At the first stage, common stable areas were singled out on the point clouds obtained from different sources: i.e., elements of the pier, corners of buildings, and bank protection structures. Based on the obtained matches, the point clouds were re-aligned using the ICP method to minimize the georeferencing error (
Table 3).
The obtained errors are considered to be permissible in landslide process studies [
74,
75].
To calculate quantitative characteristics and assess the intensity of the processes at Site 1, a series of maps of surface changes were constructed for the abrasion scarp and the upper part of the landslide slope. Changes on the slope were also studied by cross-section profiles along the slope surface.
3.2. Site 2
Analysis of landslide scarp retreat was chosen as the primary method for studying bank transformation at Site 2. A wide range of data was used to perform this task (
Table 4).
Archived historical aerial and satellite images were used to study the changes in the shoreline position, and modern satellite images were used as a base for their spatial reference. Since 2002, a topographic survey was carried out in a relative coordinate system. A network of reference points was created, to provide a multi-term survey. The reference points’ coordinates were determined using a Trimble M3 total station. The use of geodetic class marks provides high accuracy of the inter-seasonal and annual survey results referencing and subsequent aligning of the multi-temporal scans. A GNSS receiver using real-time kinematic corrections was used to transform the obtained results into the global coordinate system in 2012.
The Digital Shoreline Analysis System (DSAS), as an extension module of the ArcGIS software, was used to quantify shoreline displacement. The main application of DSAS is the use of polyline layers as a representation of a particular shoreline object at a particular point in time. Several statistical measures of shoreline change are compiled by comparing shoreline positions: net shoreline movement (NSM), shoreline change envelopes (SCE), endpoint rate (EPR), linear regression rate (LRR), and weighted linear regression rate (WLR) [
76]. This module is effective for simplifying the analysis of shoreline position changes [
77,
78]. Shoreline’s geodatabase was created in ArcGIS. The 1958 shoreline (immediately after filling the reservoir) was taken as a baseline. Following parameters were selected: distance between the transects is 15 m, search radius is 300 m. Based on the obtained transects, shoreline indicators such as linear retreat rate (m yr
−1), shoreline displacement (m) were automatically calculated.
5. Discussion
Using different remote sensing methods allows quantitative assessments of several exogenous processes at different scales and qualitative levels. Landslide processes may have different appearances, and their assessment requires different approaches [
80,
81]. Each tool solves the problem differently. Volumetric landslide changes evaluation is possible by calculating the difference of 3D models obtained by the outcomes of TLS and UAV. Assessment of planar displacements is possible by visual interpretation of remote sensing data and field monitoring of the boundaries of the study object using a total station survey. Satellite images, as well as simple aerial photos, do not permit the production of a volumetric picture. However, TLS does not permit the establishment of the exact position of the landslide body boundaries. Scanning total stations, permitting the solving of classical tachymetry tasks—that give the most exact position of object borders under research; and to make scans of landslide object for the subsequent three-dimensional reconstruction. TLS helps to estimate volumetric changes with millimeter accuracy. However, when estimating landslide structures located on the banks of reservoirs, access from all sides is difficult. It may give so-called shadows—areas with no information on three-dimensional data [
82]. The UAV method does not have such a disadvantage; even though there are differences in the level of accuracy achieved, UAV data can significantly complement scanning results in areas with no data. Multi-temporal measurements alignment with only UAV is still tricky because of the UAVs navigation system inaccuracies.
Nevertheless, the error of georeferencing of multitemporal models is acceptable when assessing rapid landslides. Returning to the most optimal method, it can be noted that the use of UAV to assess planar and volumetric changes on slopes allows obtaining both a three-dimensional model and an ultrahigh-resolution orthophoto, with which the retreat of the slope scarp can be monitored. However, as noted earlier, the construction of accurate difference-time maps requires high-precision positioning of the aircraft itself, which is achieved using high-precision GNSS receivers that allow making corrections in real-time kinematics or post-processing. As the established practice of using such airborne systems has shown, the best results in comparing multi-temporal data can be achieved by simultaneous use of UAV image data with high-precision positioning, taking into account corrections from the base station and ground reference points [
83].
The use of modern unmanned systems allows taking images with multispectral cameras that provide additional possibilities for analyzing the relationship between the spectral characteristics of rocks composing the landslide body and the intensity of the processes. However, comprehensive mathematical and statistical modeling is possible only with spatial data on fundamental factors of landslide formation—the steepness of slopes, soil texture and type of rocks, air temperature and precipitation data, especially for the winter period, groundwater depth data, and reservoir level fluctuations. However, for verification of modeling results, one needs data on currently developing landslides along the Kuibyshev reservoir banks, which is to be collected in the near future.
Active shoreline erosion currently occurs in almost all water reservoirs in Russia. In total, 36% of the shoreline of all water reservoirs in Russia is affected by these processes [
84]. Shoreline erosion is most active within large reservoirs with a capacity of more than 10 million m
3. For such water bodies in the European part of Russia about 40% of shores is transformed, in Siberia—36%, in the Far East—35%, while in small reservoirs, located in the same regions, only 13–15% of the total shoreline is affected by shoreline erosion [
85].
Among the large water bodies of the European part of Russia, the most affected by bank destruction are the Kuibyshev (75%), Volgograd (72%), Saratov (70%), and Gorky (65%) reservoirs, located in forest-steppe and steppe zones. More than 50% of the banks are actively retreating at the Kama and Novosibirsk reservoirs. For other water bodies in Russia, shoreline degradation usually occurs on less than 40% of the shoreline.
Analysis of actual data on linear shoreline retreat rates showed that the maximum average annual retreat rates during the first 10 years of development of water reservoirs, i.e., during the first, active stage of the process development, amount to 10–20 m yr
−1. In some years, the indicated rates of bank erosion at the first stage can be significantly higher, up to 100 m yr
−1. Thus, in the eastern part of Kamskoye Ustie (Sites 1–2), a 70–90 m wide strip of coastline was destroyed by landslides during the first 30 years of the reservoir’s existence. The average annual intensity of shoreline processes of the largest reservoirs of Russia at the second stage of development in the regime of steady-state or slowing-down deformations is, as a rule, significantly smaller and usually does not exceed 1–2.5 m yr
−1. The data obtained at the studied site are comparable with the data of previous studies [
86], conducted mainly by standard surveying methods.
Nevertheless, at some sites, recent studies show much larger values of shoreline displacement, mainly in the areas with loams. This, for example, occurs near the village of Izmeri in Spassky District, Republic of Tatarstan (55.130452 N, 49.469740 E), where maximum retreat values reach 9 m yr
−1, with average values of 1- 6 m yr
−1 [
5].
Observations of coastal retreat and landslide processes have been made at the Kuibyshev reservoir [
87,
88,
89,
90], mainly focusing on the assessment of coastal retreat and land losses. Long-term observations of landslide processes are conducted mainly by state agencies by installing ground control points at random sections. Unfortunately, these data are not available for scientific studies and are not published.
The use of modern methodological approaches allows obtaining information on landslide processes independently, quickly, and efficiently. These allow estimating changes overall front of the landslide slope and dangerous areas, inaccessible to observations by traditional methods.