Slope stability evolution of a deep‐seated landslide considering a constantly deforming topography

Slow‐moving deep‐seated landslides are characterised by continuous deformation, constantly changing topography and sliding‐mass geometry. Deformation rates are predominantly controlled by temporal dynamics of pore pressure. Progressing movements typically cause an over‐steepening of a landslide's foot, making these areas more susceptible to secondary slope failures and piggyback slides that, once they occur, change the geometric boundary conditions of a slope. This study presents an integrated topographic monitoring and geomechanical modelling approach, which is suitable for both model‐based replication of the landslide's hydro‐meteorological drivers and assessment of the long‐term effect of topographic changes on the stability behaviour of a large deep‐seated landslide. Parametrised at the Vögelsberg landslide (Tyrol, Austria) the integrated approach quantified considerable mass relocations between 2007 and 2020 at the landslide's foot and assessed respective effects on slope stability. Additionally, scenarios of past and future topographies were reconstructed and projected. Mass relocations of the order of 25 000 m3 were assessed between multiple airborne laser scanning acquisitions covering a period of 13 years. Based on annual uncrewed aerial vehicle laser scanning campaigns, area‐wide 3D displacements were analysed, exceeding a magnitude of 200 cm a−1 at small parts (2.500 m2) on the steeper foot of the active landslide. The main landslide body (0.28 km2) moves considerably slower with movements of 2–10 cm a−1. Besides spatio‐temporally varying hydrological drivers, topographic changes can have a severe impact on slope stability and therefore modify the spatiotemporal activity of the landslide. It is shown that, besides the hydrometeorological drivers, the varying elevation of the landslide's toe is a key parameter determining the long‐term trend of slope stability. With the presented approach the formation and evolution of the Vögelsberg landslide can be understood and explained.

below the surface, whereas complex internal deformation structures contribute to a uniquely and constantly changing slope geometry.
Characteristic geomorphological features are scarps, trenches, grabens and double or multiple crested ridges in the upper parts, counterscarps in the slope's mid-section and bulging, buckling folds and highly fractured rock masses at the slope's foot (Agliardi et al., 2012;Chigira, 1992;Crosta et al., 2013). Besides their characteristic landscape-shaping behaviour, slow deformations can cause serious damage to infrastructure and affect livelihoods. In order to prevent damage and mitigate the hazard, detailed knowledge of the hydromechanical factors controlling the sliding behaviour is of great relevance Eberhardt et al., 2007;Hofmann & Sausgruber, 2017).
Once a DSGSD has developed by strain localisation and consequential formation of shear zones Zangerl et al., 2010Zangerl et al., , 2015, the activity of therein hosted deep-seated landslides is generally controlled by changes in the stress fields (Eberhardt, 2008;Preisig et al., 2016). Predominantly, these changes are induced by fluctuations in pore-water pressure (PWP) and accompanying changes in effective stress in a landslide's shear zone. Several studies have demonstrated the correlation between PWP or groundwater-related variables (e.g., groundwater recharge, reservoir level) with landslide activity Eberhardt et al., 2007;Zangerl et al., 2010). Changes in PWP typically occur within short time periods, leading to seasonal or episodic phases of accelerated landslide movement after prolonged rainfall and intense snowmelt (Pfeiffer et al., 2021). Over longer periods, nested slope deformations (DSGSDs) cause significant topographic changes, leading to an over-steepening of the foot. Once a critical threshold of over-steepening is reached, secondary and spontaneous slope failures are commonly observable phenomena (Agliardi et al., 2012;Glueer et al., 2019). These short-term mass relocation processes, on the other hand, induce stress and strain relocation, which may also affect the activity of the DSGSD or enclosed deep-seated landslides. Continuous deformation degrades the rock strength by fracturing, fragmentation and alteration (Preisig et al., 2016). The longer the deformation, the more intense is the strength degradation. In this context, secondary slope failure processes at the continuously steepening and degrading foot evolve to be increasingly likely. Buttressing effects by interactions of the landslide's toe with the alluvium of a dynamic river bed additionally influence the slope's geomechanical conditions (Zangerl et al., 2015).
Several studies aim at reproducing the activity of hydrometeorologically driven deep-seated landslides using a variety of model approaches. Van Asch and Buma (1997) developed a hydrological model to describe groundwater fluctuations and their impact on the temporal frequency of instability of landslides. Corominas et al. (2005) proved that prediction of landslide displacements from groundwater level is feasible by rheological parameters and solving the momentum equation extended by a viscous term (Bingham and power law). Preisig (2020) exploited a one-way coupled hydromechanical numerical modelling strategy to assess slope stability under expected groundwater pressure changes. Whereas the hydrological component is well represented by a large number of studies (Corominas et al., 2005;Preisig, 2020;Van Asch & Buma, 1997), the geomechanical effect of changing slope geometries is only marginally investigated in the existing literature and predominantly based on theoretical assumptions (Molnar, 2004). In particular, the lack of accurate topographic long-term monitoring data is supposed to be a key issue for a precise reassessment of historic, past, or even pre-failure slope stability conditions. Zangerl et al. (2015Zangerl et al. ( , 2021 respond to these challenges by reconstructing pre-failure topographies based on present-day digital terrain models (DTMs) combined with the interpretation of the landslide's geomorphological features.
Recently, precise and detailed assessments of landslide topographies can be performed with the help of various remote sensing technologies. Laser scanning and photogrammetry performed from different platforms provide crucial topographic data for investigating a variety of landslide-associated research questions (Gigli et al., 2014;Jaboyedoff et al., 2012;Scaioni et al., 2014). Different sensor and acquisition methods acquiring 3D surface data induce a multitude of advantages, disadvantages and challenges depending on the research's aim (Salvini et al., 2013;Scaioni et al., 2014;Tofani et al., 2013). Whereas passive photogrammetric methods face challenges in densely vegetated areas, laser scanning as an active remote sensing technique is feasible for retrieving topographic data, even below the canopy (Baltsavias, 1999). Hence multitemporal airborne laser scanning (ALS) acquisitions represent an important and ready-to-use data archive. Complemented by more accurate, temporally and spatially higher resolved uncrewed aerial vehicle laser scanning (ULS) campaigns, solid time series of topographic 3D data can be made available . These datasets provide a unique archive to investigate landslide-induced topographic changes well suited to be integrated in any geomechanical modelling framework of a deep-seated landslide.
Here we present a geomechanical modelling strategy that is able to assess and replicate the activity of a subunit of a large DSGSD-the Vögelsberg landslide (Tyrol, Austria; for location see Figure 1b,d)under different hydrological and topographic conditions. Changes in the model's geometry are precisely assessed using combined ULS and ALS data covering a period of 13 years. Thus, geomorphological processes responsible for assessed topographic changes are identified and quantified. Based on these precise observations, potential future topographic scenarios and a pre-failure topography were approximated and discussed. Geomechanical modelling was carried out for both different hydrological and topographic parametrisations.
Objectives of the present study were to • assess recent topographic changes, reconstruct a probable pre-failure topography and project potential future topographic conditions of the Vögelsberg landslide, • parametrise a slope stability model accounting for both changing topographic and hydrological boundary conditions, • present and quantify the impacts of long-term topographic changes and seasonal hydrological conditions on the landslide activity, and • identify and quantify critical parameters controlling the evolution and formation of the Vögelsberg landslide.

| STUDY AREA
The northeastern-facing slope at the entrance of the Watten valley from 750 to 2200 m a.s.l. is subject to a large DSGSD (Figure 1b).
Located within the Lower Austroalpine Innsbruck Quartzphyllite complex (Haditsch & Mostler, 1983;Rockenschaub et al., 2003), varieties of quartzphyllite characterise the geological setting of the DSGSD. An active sliding mass (i.e., deep-seated Vögelsberg landslide) is situated on the foot of the DSGSD. Here, two varieties of quartzphyllite are noticeable: in the hanging a quartz-rich sericite phyllite containing metacarbonatic intercalations and in the lying a chlorite-sericite-phyllite (Engl, 2018). Prominent, narrow and isoclinal folded schistosity slightly dipping towards WNW is apparent at outcrops around the DSGSD . A high degree of fragmentation and decomposition can be recognised at the Vögelsberg landslide's foot. (Engl, 2018).
The bulged and compressed toe at the Wattenbach and the extensionally imprinted ridge in the upper and southernmost part of the study area indicate complex, large-scale and deep-seated slope deformations covering an area of approximately 4 km 2 . These morphological features of deep-seated and nested slope deformations overprint a glacially shaped valley morphology and hence provide evidence for post-glacial deformation activities. Situated between 750 and 1050 m a.s.l., the Vögelsberg landslide is known to cause damage to buildings and infrastructure . The Vögelsberg landslide covers an area of approximately 0.28 km 2 and is subdivided into two interacting slabs based on morphological indications ( Figure 1a; (Engl, 2018)). The higher lying crown of the northwestern slab (Slab A), as well as its very pronounced bowl-shaped central part, indicate a more intense deformation history compared to the southeastern slab (Slab B). Recent automated tracking total station (ATTS) measurements prove this observation and show the highest (up to 10 cm a À1 ) and most dynamic velocity on a reflector target situated on slab A (Figure 1c). An ATTS target mounted in the intersection zone (Slab AB) of the two slabs shows smaller (up to 7 cm a À1 ) but still temporally varying velocity. In contrast, an ATTS target mounted on Slab B shows almost constant velocity of 2 cm a À1 without distinct temporal variations. As a consequence of intense precipitation in October 2008, the foot of the over-steepened Vögelsberg landslide failed with a lateral extent of up to 100 m and failure depths of multiple metres (Klebinder & Graf, 2012). The onset of this slope failure occurred at an elevation of 860 m a.s.l., which simultaneously marks a prominent change in the slope's gradient. Whereas a steeper slope towards the Watten River is located below the onset of failure, a moderately inclined slope characterises the area above and is assumed to result from combined deep-seated slope deformation and glacial overprinting. In spring 2009, this piggyback landslide slowed down but episodically reactivated after distinct precipitation events. Overall, it can be assumed that this secondary process has relocated a significant amount of material and therefore explicitly modified the topography of the Vögelsberg landslide. The nearby Wattenbach, draining an approx. 73 km 2 big catchment, has shown significant fluvial and torrential activity during the Holocene (Patzelt, 1987). Fluviatile influence at the Vögelsberg could be mapped from the recent Watten River elevation (ca. 750 m a.s.l.) to the change in the slope's gradient and onset of recent spontaneous slope failures at 860 m a.s.l. Moreover, recent fluvial dynamics were observed to play a crucial role in the context of the secondary failure processes on the foot of the Vögelsberg landslide (Grafenauer, 2014).

| Modelling strategy and background
In order to assess and replicate the geomechanical behaviour of the Vögelsberg landslide under changing hydrological and topographic conditions ( Figure 2), a reproducible and purpose-fitting strategy of model parametrisation was established based on following background assumptions. As conceptualised in Figure 2, multiple geomorphological processes are known to shape a slope's topography.
First, in the case of a continuously moving deep-seated landslide, the time-dependent deformation itself can induce topographic changes, becoming more pronounced the longer the process continues. Typical resultant morphologies are a convex/bulge-shaped foot and a subsidence/concave-shaped middle and upper part (Agliardi et al., 2012;Zangerl et al., 2015). Second, spontaneous slope failure at the foot of such deep-seated slides becomes more likely the longer the slow deformation process lasts. Major mass relocation at the foot can have a significant impact on the superordinated sliding geometry, induce stress relocations and affect the deformation behaviour of the deepseated landslide. Third, relocated material gets deposited in the valley and thus exposed to fluvial or torrential activity. Depending on local river characteristics, alluvial levels and associated buttressing effects can change over time and significantly affect the activity of deepseated slides (Korup et al., 2010;Zangerl et al., 2015).
In order to assess topographic changes and consider them in a geomechanical model, topographic monitoring techniques as well as reconstruction and projection approaches were applied. Each topographic time step (see time series in Figure 2

| Subsurface monitoring
The subsurface of the active part of the landslide is investigated by a network of monitoring instruments in place since spring 2018. Knowledge about the subsurface is based on three core drillings (KB1, KB2 and KB3) conducted within the active landslide ( Figure 1). Retrieved material samples were used for geotechnical characterisation of the sliding mass. Obtained cores were recorded and documented regarding their geological appearance (Engl, 2018). After core withdrawal, two boreholes (KB1 and KB2) were extended as groundwater monitoring wells continuously recording the piezometric head. One borehole (KB3) was equipped with an inclinometer. The screen of KB1 well was between 16 and 49 m depth, whereas KB2 was between 21 and 39 m below the surface. Periodic inclinometer measurements record the deformation pattern along a vertical profile and therefore provide information about the depth and range of current shearing.
Coring and the installation of instruments were commissioned by the Austrian Service for Torrent and Avalanche Control. The measurements of each prism's 3D position results in a time series providing insights into the landslide-induced displacement at the respective location. Besides the derived 3D displacement vectors, the dense temporal resolution of these measurements enables a precise determination of the landslide's velocity, indicating phases of landslide acceleration or deceleration as a response to snowmelt or rainfall (Pfeiffer et al., 2021). The measurement accuracy is specified by the standard deviation of the decomposed time series (AE 0.54 cm a À1 ).

| Monitoring the landslide's topography
In contrast to the ATTS measurements at single points, complementary laser scanning acquisitions provide an area-wide assessment of topography and landslide-induced surface changes (Hu et al., 2020;Pfeiffer et al., 2018). Three airborne laser scanning campaigns from  Table 1. Multitemporal registration was carried out by applying the iterative closest point (ICP) algorithm (Besl & McKay, 1992) using stable areas (e.g., non-deformed or stable infrastructure, roads and buildings) within the point clouds. The point clouds were classified into ground and non-ground points using the TIN densification algorithm proposed by Axelsson (2000). were derived by subtracting subsequent multitemporal DTMs and used to assess topographic volumetric changes, which are particularly evident at the over-steepened foot due to secondary and spontaneous slope failures. For the area-wide assessment of coherent 3D displacements, an image correlation (IMCORR) approach that is able to assess vertical and horizontal displacement components was applied (Bremer, 2012;Fahnestock et al., 1992). For this analysis, hillshade images derived from the DTMs were used as input. In the next step, two selected cross-sections were extracted from the multitemporal point clouds in order to reconstruct a detailed surface representation for the slope stability analysis. Multitemporal profiles were further used (i) to assess recent incision or filling at the alluvium, (ii) to assess changes in the landslide's geometry due to secondary slope failures and (iii) as a basis to derive scenarios of potential future topographies.

| Reconstruction of past and scenario building of potential future topographies
A pre-failure topography was approximated using a reconstruction approach presented and applied to other deep-seated rockslides by Zangerl et al. (2015Zangerl et al. ( , 2021. This approach utilises present topographic information of assumed stable areas around the active and depressed sliding zone. Corresponding and hypothetical contour lines were mapped by inferring lines within the active, bowl-shaped and mass-

| Geotechnical material characterisations
Material obtained from core drillings and representing the assumed shearing depth was analysed with regard to its grain size distribution, density (γ: specific weight) and shear parameter (φ: angle of internal friction and c: cohesion). Sieving and sedimentation tests were carried out to determine the grain size distributions of each sample per borehole. Immersion weighting was done with sealed and undisturbed samples to estimate the density. Five consolidated drained triaxial shear tests considering each of three different load steps (100,200 and 300 kPa) with a shear velocity of 0.003 cm min À1 were carried out to assess the material's strength. Three disturbed samples (KB1, KB2 and KB3) were prepared in a 7 cm diameter cell. One additional sample obtained from KB3 was first prepared as an undisturbed sample and analysed in a 10 cm diameter cell applying a multi-step experiment. Subsequently, the same sample was re-prepared and investigated in a disturbed multi-step experiment in order to assess the potential effect of structures that could not be considered in disturbed samples. Table 2 summarises the installation properties of the samples analysed by triaxial shearing. The angle of internal friction and cohesion were analysed considering both the concepts of residual shear strength and peak shear strength.

| Geomechanical modelling: Slope stability assessment and model calibration
The analytical slope stability model GGU Stability (Bu, 2020)    Consecutive inclinometer measurements performed in KB3 localise the deformation within À40 to À50 m depth below the surface.
Within this range, most of the deformation focuses between À48 and À50 m (Figure 6c). Three core drillings revealed a sequence of disintegrated rock (i.e., soil) with depths of À53 m in KB2, À60 m in KB1 and À70 m in KB3 (Figure 5b,c; Engl, 2018). The volume of the disintegrated and loose rock mass is approximated by 10 million m 3 (Vecchiotti et al., 2022). Spring districts indicating intersection zones of groundwater and topography are tied to more inclined areas of the slope (Figures 1a and 6a,b), located right below the crown (920-1000 m a.s.l.) and at the over-steepened foot (800-850 m a.s.l.).

| Geotechnical parameter
The specific weight of the sliding mass material was determined between 20.3 and 21.2 kN m À3 . Supported by the observed continuous landslide deformation and indicated long history of deformation, the principle of residual shear strength (r0) was applied, assuming a cohesion of 0 kN m À2 . Analysing the peak shear strength (peak) indicates a cohesion between 0 and 38.4 kN m À2 and a friction angle between 26 and 32.6 valid for the analysed samples ( Figure 10a and Table 3). Grain size distributions show well-graded, (extremely) poorly sorted and non-uniformly distributed characteristics and a clay content of less than 9% (10b).

| Slope stability
Under present conditions, results of geomechanical modelling indicate the slope to be either slightly above or slightly below its limit equilibrium, depending on the applied setting of material parameters, phreatic surface and the investigated cross-section. Resulting FoS calculated for each material parameter set (Table 3) and at static pore pressure distribution adapted to present observations are slightly above equilibrium limit. Differences can be observed between the two considered cross-sections. Profile 2 features generally lower FoS values compared to Profile 1 (Figure 11c). The OAT-sensitivity T A B L E 3 Geotechnical characteristics of the landslide mass F I G U R E 1 1 Resulting slope stability expressed by FoS and parametrised by a set of geotechnical parameters shown in Table 3. Boxplots (box: lower and upper quartile, whiskers: 1.5 * interquartile range) show differences between (a) deep (i.e., sliding surface close to lower boundary) or shallow (i.e., sliding surface close to upper boundary) sliding surfaces, (b) residual or peak parametrisation (only profile 2) and (c) the geometry of the modelled cross-section (Profile 1 vs. Profile 2). The lower graph (d) shows the effect of alluvial height deviations on changes in slope stability for a parametrisation using residual shear strength values (φ ¼ 32:6 , c ¼ 0 and γ ¼ 21:3 kN m À3 ) and peak shear strength values (φ ¼ 31:5 , c ¼ 32:7 kN m À2 and γ ¼ 21:3 kN m À3 ) [Color figure can be viewed at wileyonlinelibrary.com] parametrisation, it becomes obvious that alluvium uplift enhances stability and alluvial erosion reduces calculated stability. This nonlinear relation becomes even more pronounced while investigating the effects of reconstructed and projected topographic scenarios specifying a wide range of different alluvial levels (Figure 11d).
A hypothetical severe secondary slope failure event as represented by Scenario 1 would lead to an increased slope stability of +8.6% (residual) and +12.5% (peak). Potential future fluvial erosion to the alluvial level of 2020 (Scenario 2) would implicate equal stability conditions as they prevailed prior to Scenario 1. Continuous fluvial erosion as hypothetically pictured by Scenario 3 would negatively influence the stability (À35.5% for residual and À29.6% for peak shear strength) and therefore increase the activity of the Vögelsberg landslide. A distinct change in the sliding geometry is likely at an alluvial level between 5 and 10 m lower than the recent level ( Figure 14). Geomechanical analysis of driving and resisting forces of the Vögelsberg landslide has been shown to be close to limit equilibrium.  Assuming that present-day measured average velocities (5 cm a À1 ) have prevailed since the formation of the Vögelsberg landslide and that the total experienced 3D displacement accounts for 240 m, a formation age of 4.8 ka before present could be estimated.
Deceeding the limit equilibrium at the assumed critical alluvial level, simultaneous sliding and ongoing fluvial erosion are seen as interlinked dominant processes that led to today's topography. Although continuous and slow movement varies with a magnitude of several centimetres depending on the state of the hydrological forcing (Pfeiffer et al., 2021, its persistency throughout time is assumed to induce following process cascade. Preceding movements lead to recurring over-steepening of the foot, consecutive secondary slope failures spontaneously relocate material from the upper foot to the toe, and succeeding fluvial erosion removes temporarily deposited landslide material. Such a geomorphic feedback between landslides and river channels is well known from many tectonically active mountain belts (Korup et al., 2010).
Based on this, we conclude that recently observed short-term variations in landslide activity are directly associated with groundwater level variations. From a long-term perspective, recently explored effects of climate change on the activity of the Vögelsberg landslide indicate a reduced frequency of accelerating events . On the other hand, topographic changes, in particular the interactions of the landslide foot with the Watten River's alluvium, are seen as a crucial variable regulating the landslide's response to adjustments of its geometrical boundaries.

| CONCLUSIONS
The present study investigates the effects of a changing topography Raising the landslide's toe by, for example, enforced alluvial sedimentation enhances the stability by 0.7% per metre. Whenever conclusions about topographic effects controlling the activity of deep-seated landslides are tainted by stabilising three-dimensional effects, the presented integrated topographic monitoring and geomechanical modelling approach is transferable to other case studies featuring similar data availability.