Spatio‐temporal assessment of the hydrological drivers of an active deep‐seated gravitational slope deformation: The Vögelsberg landslide in Tyrol (Austria)

Spatio‐temporal variations of precipitation are presumed to influence the displacement rate of slow‐moving deep‐seated landslides by controlling groundwater recharge, pore‐water pressure and shear strength. Phases of landslide acceleration responding to long‐lasting rainfall and snowmelt events occur under site‐ and event‐specific time delays. Assessing groundwater recharge and simultaneous recording of landslide displacement in a sufficient spatial and temporal resolution is essential to deepen the understanding of mechanisms controlling a landslide's deformation behaviour and is indispensable when it comes to identifying and developing target‐oriented mitigation strategies. The objective of this study was to assess hydrological landslide drivers (solid and liquid precipitation, snowmelt and evapotranspiration) and to investigate their spatio‐temporal distribution in the context of movements recorded at the Vögelsberg landslide (Tyrol, Austria). Hydrometeorological variables were simulated using the AMUNDSEN (Alpine MUltiscale Numerical Distributed Simulation ENgine) hydroclimatological model and landslide movements were continuously monitored using an automated tracking total station. Area‐wide simulated time series of available water were used: (i) to separate them into single landslide triggering hydrometeorological events; (ii) to analyse spatio‐temporal patterns of water availability per triggering event including individual response times; (iii) to delineate an effective hydrological landslide catchment; and (iv) to identify relations between assessed water input and landslide displacement rate. For the observation period from 05‐2016 until 06‐2019 we identified three distinctive hydrometeorological events causing time‐delayed periods of landslide acceleration. Spatio‐temporal differences in water availability per triggering event result in spatially diverse response times varying from 20 to 60 days for rainfall‐triggered events and between 0 and 8 days for events triggered by snowmelt. Pronounced spatio‐temporal differences of snowmelt within the model domain were identified to offer a unique possibility to delineate the effective hydrological landslide catchment. While considering event‐specific time‐lags, logarithmic correlations between incoming water and landslide displacement rate become apparent.


Summary
Spatio-temporal variations of precipitation are presumed to influence the displacement rate of slow-moving deep-seated landslides by controlling groundwater recharge, pore-water pressure and shear strength. Phases of landslide acceleration responding to long-lasting rainfall and snowmelt events occur under site-and eventspecific time delays. Assessing groundwater recharge and simultaneous recording of landslide displacement in a sufficient spatial and temporal resolution is essential to deepen the understanding of mechanisms controlling a landslide's deformation behaviour and is indispensable when it comes to identifying and developing targetoriented mitigation strategies. The objective of this study was to assess hydrological landslide drivers (solid and liquid precipitation, snowmelt and evapotranspiration) and to investigate their spatio-temporal distribution in the context of movements recorded at the Vögelsberg landslide (Tyrol, Austria). Hydrometeorological variables were simulated using the AMUNDSEN (Alpine MUltiscale Numerical Distributed Simulation ENgine) hydroclimatological model and landslide movements were continuously monitored using an automated tracking total station. Area-wide simulated time series of available water were used: (i) to separate them into single landslide triggering hydrometeorological events; (ii) to analyse spatio-temporal patterns of water availability per triggering event including individual response times; (iii) to delineate an effective hydrological landslide catchment; and (iv) to identify relations between assessed water input and landslide displacement rate. For the observation period from 05-2016 until 06-2019 we identified three distinctive hydrometeorological events causing time-delayed periods of landslide acceleration. Spatio-temporal differences in water availability per triggering event result in spatially diverse response times varying from 20 to 60 days for rainfall-triggered events and between 0 and 8 days for events triggered by snowmelt. Pronounced spatio-temporal differences of snowmelt within the model domain were identified to offer a unique possibility to delineate the effective hydrological landslide catchment. While considering event-specific time-lags, logarithmic correlations between incoming water and landslide displacement rate become apparent.

K E Y W O R D S
AMUNDSEN, automated tracking total station, hydrological modelling, landslide displacement, OPERANDUM, time series analysis 1 | INTRODUCTION Large mass movements such as deep-seated gravitational slope deformations (DSGSDs) are common phenomena in mountain ranges across the world, affecting entire mountain slopes from the valley bottom to the slope's ridge (Agliardi et al., 2012;Crosta et al., 2013). They are typically complex and large systems of slope failures characterized by tensional morphological features in the upper part and compressional features in the lower part. These features originate from longterm displacements of a few millimetres per year, as it was identified to be common for DSGSD (Ambrosi & Crosta, 2006;Varnes et al., 1990;Zangerl et al., 2010). In several cases DSGSDs are accompanied by secondary, more active subunits preferentially occurring on highly fractured and intensely deformed rock masses on the bulged toe of the DSGSD (Bovis & Evans, 1996;Crosta et al., 2014). Such landslide subunits can show enhanced, fluctuating displacement rates ranging from centimetres to several metres per year (Brückl et al., 2013;Pfeiffer et al., 2018;Zieher et al., 2019). In some cases they accelerate and result in catastrophic slope failures such as rockslides and rock avalanches, endangering human well-being and infrastructure situated on the top of the landslide or within its potential run-out area (Carlà et al., 2017;Ostermann & Sanders, 2017).
Changing landslide displacement rates over time are commonly associated with fluctuating pore-water pressures subject to spatiotemporally varying rainfall and snowmelt, permitting groundwater recharge. Infiltrating precipitation reduces the shear strength by increasing the pore-water pressure when reaching the landslide's governing aquifer. The pore-water pressure-controlled reduction of the effective normal stress can then lead to increased slope movement rates (Terzaghi, 1950). Several studies investigating the relations between precipitation and landslide displacement rate have been carried out. (Iverson & Major, 1987;Lacroix et al., 2020;Van Genuchten & Van Asch, 1988). Handwerger et al., (2013), for example, investigated the controls on seasonal deformation of multiple slowly moving landslides using satellite radar interferometry (InSAR) and precipitation data recorded at a rain gauge approximately 30 km away from the landslide area. Bievre et al. (2018) combined extensive geophysical, geotechnical and hydrogeological investigations towards a better understanding of the causes of seasonal variations of displacement rates. They used cross-correlations between hydrometeorological and piezometer time series to prove a preferential water infiltration path in a slow-moving clayey earthslide. Macfarlane (2009) used geodetic surveys combined with extensometers, inclinometers, piezometers, flow-measuring weirs and records of a nearby rain gauge to investigate and further predict the behaviour of large, slowly moving landslides in reservoir settings. Brückl et al. (2013) presented an integrated monitoring approach combining geodetic, hydrological and seismological methods to assess the displacement over time and its potential drivers. The authors paid particular attention to the influence of snowmelt, which was assessed on site by weekly snow water equivalent measurements at points along a slope profile. Together with liquid precipitation recorded at a nearby weather station, they estimated infiltration as a proxy for the hydrostatic water level to support early warning. Similar investigations are described in Crosta et al. (2014), where snowmelt was identified to be a major cause in the context of landslide acceleration. Osawa et al. (2018) carried out detailed investigations to retrieve a comprehensive understanding of hydrological processes (i.e., snowmelt and rainfall) in the context of a slow-moving landslide. Additional effects of evapotranspiration were accounted in the study by Coe (2012), where a regional moisture balance index was retrieved to analyse future landslide movements considering climate change projections.
Most of these studies depict the existence of manifold methods to assess landslide movements. Continuous or periodic measurements estimating displacements on or below the surface in an area-wide or at an individual point scale find application in a target-oriented measurement setup. But when it comes to the assessment of the hydrological input, most of these studies rely on precipitation records measured at single points with conventional gauge systems, either neglecting spatial differences in precipitation or neglecting the timedelayed effect of melt water being released from the snow pack after the winter season. Investigating hydrological drivers of deep-seated landslides in a large DSGSD setting demands continuous and precise information of landslide displacement together with a temporally and spatially highly resolved product of water available for groundwater recharge. Where a large variety of methods for estimating landslide displacements in either high temporal or spatial resolution are applied in several case studies, the on-site acquisition of precipitation, evapotranspiration and snowmelt is not practicable in high temporal and spatial resolution (Brückl et al., 2013).
Physically based hydroclimatological models driven with atmospheric variables recorded at well-distributed surrounding weather stations overcome this issue and allow a practicable assessment of available water in high temporal but also high spatial resolution. They provide unique information about rainfall and snowfall, including the temporary storage of snow and the delayed release of snowmelt water, which leads to very pronounced spatio-temporal differences in the availability of water in Alpine regions. These models are frequently applied in glaciological and hydrological investigations focusing on the assessment of current and future water resources in mountainous regions (Hanzer et al., 2016;Marke et al., 2015;Strasser, 2008). Furthermore, their ability to simulate potential groundwater recharge time series for every raster cell within a catchment emphasizes their importance in investigating the spatio-temporal distribution of hydrological landslide drivers. Although modelling-based approaches aiming to assess landslides influencing water availability were presented in the past, they demonstrate a lack in spatial completeness or model complexity considering the physical processes determining snowpack evolution (de Palézieux & Loew, 2019;Vallet et al., 2015).
No study so far has aimed at analysing the different spatiotemporal characteristics of available water forcing groundwater recharge associated with landslide acceleration, which is assumed to offer a novel way for creating a deeper understanding of hydrological drivers of landslides. Having a spatially distributed product of water availability in a sufficient temporal resolution derived from a physically based hydroclimatological model is supposed to create several advantages and possibilities important for identifying mitigation measures when comparing them with landslide movement records. The benefits of an area-wide product compared to non-area wide products are manifold. Among them, the mapped distribution of water availability per landslide-triggering event can help to indicate areas of higher water avilability. Performed raster cell-and event-wise comparative analysis of water input and landslide displacement time series can be used to estimate an event-specific time lag between hydrological input and landslide acceleration. This time lag, further defined as the landslide response time and assessed for every raster cell, forms maps of response times valid for each identified triggering hydrometeorological event. These maps support the delineation of the landslide's effective hydrological catchment by excluding areas where water availability cannot be responsible in forcing a landslide's acceleration (e.g., when the moment of water availability occurs after the landslide acceleration). Estimated event-specific time lags are further necessary to derive models allowing the prediction of landslide movements as a response to hydrological input. Retrieved outcomes help in understanding the landslide triggers and are therefore beneficiary in planning and locating target-oriented measures aiming to reduce a landslide's displacement rate Eberhardt et al., 2007).
In this study the hydrometeorological drivers forcing movements of the Vögelsberg deep-seated landslide were assessed and analysed in a spatio-temporal way by using the hydroclimatological model AMUNDSEN (Alpine MUltiscale Numerical Distributed Simulation ENgine) (Strasser, 2008). While paying special attention to the more complex simulation of the snow cover evolution, simulated snow depths and snow masses were validated in detail to provide an accurate information of snowmelt, which forms the water availability together with rainfall and respecting the withdrawing effect of evapotranspiration. Hydrometeorological events responsible for phases of landslide acceleration were temporally delimited using a rule-based segmentation approach and examined considering their spatial characteristics. Area-wide cross-correlations between statistics of hydrometeorological time series and monitored landslide displacement rate were derived and investigated on an event basis to derive individual landslide response times for every raster cell within the catchment. By only considering nonnegative response times, an effective hydrological landslide catchment was delineated and a time series of thereon aggregated water availability was used for detailed analysis of interactions between landslide-controlling hydrometeorological processes and displacement rates.
The objectives of the present study were: • to assess hydrological drivers controlling landslide displacement using a hydroclimatological model; • to extract hydrometeorological events causing pronounced landslide accelerations including individual response times; • to delineate the effective hydrological landslide catchment; and • to identify and analyse the temporal and spatial correlations of hydrometeorological events and related increase in landslide displacement rate.

| STUDY AREA
The Vögelsberg DSGSD is situated on a northeast-facing slope at the Lithologies favouring prevailing slope deformation processes belong to the Innsbrucker Quartzphyllite complex of the central Eastern Alps (Rockenschaub et al., 2003). Sericite phyllites, chloritesericite phyllites and quartz phyllites with intercalated calcareous marbles are apparent at the investigated slope. Greenschist (prasinites) are exposed around the summit of the Largoz, which is the highest point in the DSGSD catchment (red polygon in Figure 1a dominates the land cover. The highest areas in the catchment (1800-2200 m a.s.l.) are characterized by natural grassland scattered by single Swiss stone pine trees (Pinus cembra). The map in Figure 1b shows the the land cover types occurring in the study area. The closeby weather station 'Rinn' (960 m a.s.l.), located approximately 8 km to the east of the active landslide, records a mean annual precipitation of 896 mm between 2008 and 2018, of which 13% is snowfall, considering a temperature threshold of 1 C. Even larger amounts of snowfall are assumed to accumulate at high elevations in the catchment, with great potential to recharge groundwater during spring and early summer snowmelt periods, whose precise quantification is part of this study. The study area represents a case study investigated within the scope of the OPERANDUM project aiming to investigate naturebased solutions for hydrometeorological risk reduction (Ruangpan et al., 2020).

| Monitoring the landslide movement
The slope's movement is monitored by an automated tracking total station (ATTS) operated by the the Department of Geoinformation They indicate the extent of the more active landslide part. Recorded displacements at the other 29 prisms are below the assumed measurement accuracy and further assumed as stable. Targets are mounted either on buildings or artificially installed piles. The hourly measurements were aggregated to daily mean coordinates, which were used to estimate cumulative displacements and displacement rates. Time series of displacement rates were further smoothed using a moving average within a considered window of 20 days. This window size turned out to be a suitable time span in which the effect of outliers could be minimized by still considering accurate temporal information of changes in displacement rate. For further analysis and correlation with hydrological time series, mean displacement rates were calculated based on the 13 prisms distributed on the active landslide part, which cover the same observation period.

| Model description, setup and input data
Rainfall and snowmelt are the main water input sources within the catchment of the DSGSD analysed in this study, and their spatiotemporal variability is essential for the quantification of the relevant hydrological drivers. But quantifying time-varying distributed water input is not feasible by point measurements of precipitation, for example, due to the spatial variability or the temporary storage in the snow cover and the delayed release during snowmelt. Therefore, we used the hydroclimatological model AMUNDSEN to simulate the water input potentially available for infiltration (i.e., rainfall and snowmelt minus evapotranspiration) at the catchment scale. AMUNDSEN is a modular fully distributed hydroclimatological model (Strasser, 2008) designed to simulate the evolution of mountain snow cover (i.e., the accumulation and ablation) and has been validated at diverse Alpine Interpolated fields of the measured meteorological variables (T a , P, RH, WS) were derived by a combined lapse rate-inverse distance weighting scheme using calculated lapse rates at each time step.
G was interpolated accounting for topographic effects. The type of precipitation was determined by the wet bulb temperature threshold (T w = 0 C) and the energy balance approach was used to simulate the snow cover evolution. Submodules 'canopy' and 'evapotranspiration' were enabled because of the high forest cover (66% in the model domain) and to subtract evapotranspiration from rainfall as potential water input during the snow-free period. We ran AMUNDSEN with a temporal and spatial resolution of 3 h and 20 m (grid cell size), respectively. To ensure a spatial comprehensive assessment of the landslide's hydrological drivers, a model domain representing the whole above area of the lowest point of the active landslide was delimited. This was done by applying an algorithm for estimating flow directions and upslope areas proposed by Tarboton (1997) (1)) was chosen to measure the goodness of fit of the snow cover extent based on true positive (TP), false positive (FP) and false negative (FN) estimated raster cells: The CSI is a sensitive contingency table-based criterion to compare observed and simulated snow cover pattern in each raster cell.
Additionally, the overall accuracy (ACC) and BIAS were used as comparable criteria to assess the quality of simulated and observed snow cover. ACC is defined as the number of correct simulations divided by the number of samples. The BIAS is defined as the frequency of correct snow cover simulations divided by the number of times where snow cover is observed. All three goodness-of-fit criteria approach a perfect score with a value of 1 (Hanzer et al., 2016;Zappa, 2008).

| Hydrometeorological event detection
Hydrometeorological events define time spans of increased water availability for infiltration and groundwater recharge. Intensity and duration of prolonged rainfalls were identified to be important parameters controlling the evolution of shallow landslides (Caine, 1980;Guzzetti et al., 2008). Assuming that hydrometeorological events with a critical intensity duration relation also trigger acceleration phases of more complex, slow-moving and delayed responding deep-seated landslides, an approach for estimating their thresholds was deployed.
Time series of modelled rainfall, snowmelt and evapotranspiration were consolidated to a common time series of water potentially available for infiltration while neglecting the unknown portion of surface runoff. A segmentation approach, as successfully applied in (Meißl et al., 2020) to investigate runoff response to rainfall events, was We used the area-wide simulated time series of water availability to assess response times for every raster cell in the predefined model domain. Assuming different response behaviours referring to different event characteristics, additional knowledge of the temporal boundaries of triggering events was considered for a catchment-wide response time estimation. We used a cross-correlation approach to estimate respective response times, as it is proven to be an appropriate way for an automated assessment of time lags between precipitation and landslide displacement rate time series of hydrologically driven landslides (Bievre et al., 2018;Lollino et al., 2002Lollino et al., , 2006. The cross-correlation function (R xy ) describing the correlation between the function of available water (x t ) and landslide displacement rate (y t ) under varying time lags (τ) is expressed as The occurrence of a maximum value within the respective crosscorrelation function (R xy ) locates the time lag indicating the best conformity of hydrological and geodetic time series, which in our case is interpreted to be the landslide response time. We used prior identi- since the focus of this study was to investigate causes of varying movements in time as a response of hydrological input. Figure 3 shows the displacement rates and position of the 13 targets used for landslide movement interpretation. These targets cover the same observation period; hence their mean displacement rate (black line in  (Table 1).
The spatio-temporal validation of the snow cover extent derived from simulated snow depth grids compared to snow cover maps based on Sentinel-2 imagery is shown in Figure 5   closer to the active landslide body release higher amounts of water previously stored in the snow cover than the surrounding forested areas. Similar effects can be recognized in the annual mean distribution of available water (Figure 7d). A more balanced distribution pattern of simulated available water is observable during the rainfall-dominated event (event #2, Figure 7b). Derived response times for event #1 are in the range of approximately 20-60 days, where a general trend of increasing response time with increasing elevation and distance to the active landslide is obvious (Figures 9a and 10a). The landslide's response to the rainfalldominated event #2, on the other hand, is spatially indifferent and can uniformly be approximated with a response time of 45 days (Figures 9b and 10b). The most recent event #3 triggering a pronounced acceleration of the landslide mainly due to snowmelt stands out concerning low response times in the order of À60 to 8 days (Figures 9c and 10c). Negative response times indicate raster cells where the landslide acceleration occurred prior to water becoming available for infiltration. Commonly, this is the case for snowmelt occurring later in spring at higher elevations when the landslide had already accelerated due to meltwater released earlier at lower elevations.

| Hydrometeorological events
A heterogeneous response of landslide acceleration to variable water availability caused by different land cover types is apparent when investigating event #1 (Figure 9a). During this event the

| Assessment of the landslide's hydrological catchment area
The results obtained from the area-wide cross-correlation offer essential information for delineating the landslide's effective hydrological catchment. Neglecting raster cells with response times less than 0 days results in a catchment extent ranging approximately from 750 to 1700 m a.s.l. (see Figure 9d). Within this range the highest values of the mean correlation strength of the three analysed events occur between 1250 and 1700 m a.s.l., indicating this range to be the most probable source of hydrometeorological water, causing pore pressure changes in the landslide's aquifer and leading to accelerated slope movements (Figure 9h).
To further delineate a laterally limited hydrological catchment, the upslope area of the lowest point in the active landslide was calculated. Additionally, the percentage contribution of every upslope raster cell to this initial raster cell was assessed (Tarboton, 1997). When only respecting raster cells potentially contributing more than 0.01% to the lowest point, the lateral extent of landslide hydrological  (Figure 11b,c). The model derived for event #1 indicates a more sensitive response of landslide acceleration caused by water availability. Here the increase in displacement rate is more pronounced and ranges from 2 to 5.3 cm/a between 0 and 90 mm of cumulated water availability.
From 90 mm towards higher water availability values the displacement rate increase flattens as in the models for events #2 and #3 but still reaches higher absolute displacement rate values.
The more sensitive response of landslide to water input during event #1 may be associated with different preconditions in terms of pre-event water. Within a 30-day time span 64 mm of water was available prior to event #1, but only 14 mm prior to event #3 and even negative amounts of À9 mm caused by distinct evapotranspiration prior to event #2 characterize the preconditions before the start of the succeeding triggering event ( Figure 12).

| DISCUSSION
Applying the hydroclimatological model AMUNDSEN parametrized with meteorological data acquired at well-distributed surrounding weather stations allowed us to retrieve a well-validated spatiotemporal dataset of water availability in the study area around the Vögelsberg landslide. Validation of simulated snow depths, snow masses and spatial snow cover indicate a good model performance in simulating the evolution of the snowpack throughout the investigation period. Additionally, rainfall and evapotranspiration grids were used to retrieve the spatio-temporal pattern of water availability. The lack of on-site evapotranspiration reference data makes an evaluation of simulated evapotranspiration time series challenging. However, annual evaporation is in accordance with the long-term water balance of the Watten valley and similar to the regional estimate of Duethmann and Blöschl (2018).
Three outstanding landslide acceleration phases were observed within our study, each of them being triggered either by snowmelt, rainfall or both. Their temporal occurrence does not show a periodicity or seasonality. We quantified the hydrological drivers that control the individual response times and related them to the movement of the landslide. We also found distinct spatio-temporal patterns for each of the events. Derived landslide response times vary from 20 to 60 days for the combined rainfall-and snowmelt-induced event #1.
We attribute the 40-day range of response times to the combined water input of snowmelt and rainfall, as well as to the long duration of the event (209 days We found logarithmic relationships between the displacement rate and water input for each of the events, but their differencei.e., the event-to-event variabilitysuggests that the response of the landslide varies, depending on hydrological preconditions and system states. Nevertheless, hydrogeological conditions in nature are complex and require further on-site investigations (e.g., long-term measurements of groundwater level, geological mapping, estimations of groundwater recharge elevation) to validate established assumptions and to create a more consistent groundwater model. When parametrizing numerical groundwater models, additional knowledge of surface runoff has to be acquired to complete necessary variables for groundwater recharge estimation.
Although the quality of displacement rates and simulated water availability time series were well assessed (Sections 4.2 and 4.1), quantifying the uncertainty of the derived response times remains challenging. Temporal aggregation of the highly resolved displacement time series is necessary to retrieve valuable information about the landslide displacement rates within the observation period, but may also be consistent with a loss of quality in terms of the temporal prediction of the respective start of acceleration and deceleration phases.
An additional influence on response times is attributed to the temporal accuracy of simulated water availability time series. For simulated snow depth times series we see that snowmelt is predicted earlier by the hydroclimatological model compared to the observed snowmelt reaching a maximum shift of 9 days at the lowest validation weather station during winter 2019. Accounting for this offset would cause a small shift of respective response times towards lower values in the range of a maximum of 9 days.
Limitations due to data availability are apparent within our study.
In particular, missing displacement records prior to the beginning of the first acceleration phase may cause biased response times.
Recently collected groundwater-level measurements were not yet available for this study, but very likely may strengthen our prior assumption that changes in landslide displacement rate are driven by pore-water pressure changes.

| CONCLUSIONS
Applying the AMUNDSEN hydroclimatological model allowed us to accurately assess hydrological drivers (i.e., rainfall, snowmelt), which are important for groundwater recharge and hence control the pore-water pressure and landslide movement. Point validation of simulated snow depths and snow masses as well as a spatial validation of the snow cover evolution throughout the winter months using satellite-supported reference data indicate a good model performance. Simulated rainfall, snowmelt and evapotranspiration were further used to derive time series of available water for every raster cell of the landslide catchment. A rule-based segmentation approach applied on time series of available water was used to extract hydrometeorological events, whose magnitude and duration are consistent with the subsequent magnitude of landslide displacement rate increase.
Three major phases of increased landslide displacement rate can be associated with three outstanding hydrometeorological events lasting for 106-209 days but implicating between 68% and 95% of water that is normally received during a whole year. Potential landslide response to the triggering event was derived for every raster cell and event by cross-correlating a 30-day running sum of water input time series with a mean displacement rate time series. Landslide acceleration in response to snowmelt nearby occurs in a couple of days (0-8), whereas the response to rainfall takes 45 days. Reasonable response times were used to delineate the landslide's effective hydrological catchment.
Spatially aggregated hydrological time series within this catchment boundaries were used to derive statistical models for the three events describing the behaviour of landslide displacement rate in response to water input. Fitted logarithmic models seem to describe this relation quite well, when considering respective time lag between the two time series. For two events where no significant amounts of water were available prior to the event, we observed a logarithmic increase of displacement rate until a steady level of 3-4 cm/a. No further acceleration in response of water input can be observed after reaching this level. For an event with considerable amounts of water already being feed to the slope's hydrological system beforehand, the displacement rate increase is observed to be more sensitive until a quasi-steady level of 5.5 cm/a is reached.
The present study shows how hydrological landslide drivers can be accurately assessed in both high temporal and spatial resolution and how this information, combined with continuous landslide displacement rate records, can be used to draw conclusions about hydrological processes causing increased landslide displacement rates.
Combining these results with further on-site hydrogeological investigations and monitoring may help in validating, clarifying, strengthening and deepening the understanding of the involved processes.