Rainfall-induced shallow landslides and soil wetness: comparison of physically based and probabilistic predictions

. Landslides are an impacting natural hazard in alpine regions, calling for effective forecasting and warning systems. Here we compare two methods (physically based and probabilistic) for the prediction of shallow rainfall-induced landslides in an application to Switzerland, with a speciﬁc focus on the value of antecedent soil wetness. First, we show that landslide susceptibility predicted by the factor of safety in the inﬁnite slope model is strongly dependent on soil data inputs, limiting the hydrologically active range where landslides can occur to only ∼ 20% of the country with typical soil parameters and soil 5 depth models, not accounting for uncertainty. Second, we ﬁnd the soil saturation estimate provided by a conceptual hydrological model (PREVAH) to be more informative for landslide prediction than that estimated by the physically based coarse resolution model (TerrSysMP), which we attribute to the lack of temporal variability and coarse spatial resolution in the latter. Nevertheless, combining the soil water state estimates in TerrSysMP with the inﬁnite slope approach improves the separation between landslide triggering and non-triggering rainfall events. Third, we demonstrate the added value of antecedent soil satu-10 ration in combination with rainfall thresholds. We propose a sequential threshold approach, where events are ﬁrst split into dry and wet antecedent conditions by a N-day antecedent soil saturation threshold, and then two different total rainfall-duration threshold curves are estimated. This, among all different approaches explored, is found to be the most successful for landslides prediction.

soil, which decreases its strength (e.g., Anderson and Sitar, 1995;Highland et al., 2008). This condition is often associated with 25 intense or long lasting rainfall events that saturate the soil by vertical infiltration and lateral subsurface drainage. The wetness of the soil prior to the triggering rainfall is therefore also important (Bogaard and Greco, 2018).
Several approaches exist for the prediction of landslides that focus on one or more predisposing and triggering factors, typically classified into 3 types: susceptibility mapping, probabilistic approaches, and physically-based modelling (e.g., Aleotti and Chowdhury, 1999). 30 Susceptibility mapping assesses the susceptibility of a certain area to landsliding based on predisposing factors. In statistical susceptibility mapping, the different predisposing factors, geological, topographical, and climatological properties, are combined and used as explanatory variables in a statistical model (e.g., Reichenbach et al., 2018). Landslide hazard maps are then generated by various forms of linear and nonlinear multivariate regression models (e.g., Chung et al., 1995), logistic regression (e.g., Ohlmacher and Davis, 2003;Ayalew and Yamagishi, 2005;Lee and Pradhan, 2007;Yilmaz, 2009;von Ruette et al., 35 2011), or machine learning algorithms (e.g., Saito et al., 2009;Ermini et al., 2005;Yilmaz, 2009). Susceptibility mapping can also be achieved by applying a physically-based geotechnical model which identifies the likelihood of failure in a region based on an assessment of likely soil water distribution in space (e.g., Baum et al., 2002Baum et al., , 2008Dietrich and Montgomery, 1998;Formetta et al., 2016).
Probabilistic approaches focus mainly on the temporal component of the landslide hazard (triggering factors), rather than 40 the spatial susceptibility (predisposing factors). They are based on the assumption that rainfall is the main triggering factor, and take advantage of historical records of rainfall and landslides. These databases are combined to learn which meteorological conditions have been associated with the triggering of landslides in the past. This allows then to recognise critical conditions in weather forecasts of the coming days and estimate how likely the occurrence of landsliding is. The most common of these approaches is that of rainfall thresholds, and in particular intensity-duration or total rainfall-duration threshold curves (e.g., 45 Guzzetti et al., 2007;Leonarduzzi et al., 2017;Segoni et al., 2018). While rainfall is mostly the main triggering factor, soil wetness conditions prior to triggering rainfall can also be included in this framework (e.g., Bogaard and Greco, 2018;Marino et al., 2020). The antecedent soil wetness conditions can be derived in many different ways, each with its advantages and limitations, for example from in-situ measurements (depend on network density, e.g., Wicki et al., 2020), remote sensing of soil moisture (suffer from low resolution and insufficient penetrating depth, e.g., Brocca et al., 2012;Thomas et al., 2019), 50 through proxies of soil wetness like antecedent rainfall (miss evapotranspiration and snowmelt, e.g., Glade et al., 2000;Godt et al., 2006;Mathew et al., 2014), or by hydrological soil water balance modelling (e.g., Ponziani et al., 2012;Thomas et al., 2018).
Finally, physically-based modelling approaches are usually made up of two components to simulate slope stability in time and space: an hydrological and a geotechnical model. The hydrological model is used to estimate the condition of the soil, i.e. 55 the pore water pressure and/or saturation, which are then used in the geotechnical model for the estimation of slope stability (e.g., by the infinite slope or other hydromechanical slope failure model). These approaches are theoretically the most accurate and predict both when and where a landslide could occur, but are typically very computationally expensive and data demanding.

60
In this work, we conduct a comparison of a probabilistic and physically-based modelling approach to landslide prediction with the specific question of the value of the inclusion of antecedent soil wetness state in the prediction. Our scale of analysis is regional (Switzerland) instead of hillslope/catchment scale, because it is at this scale that landslide early warning systems need to be developed (e.g., Staehli et al., 2015). First we explore the regional susceptibility to landslides following the infinite slope approach (physically-based susceptibility mapping). This allows us to understand where hydrology can play a role in the 65 landscape in triggering landslides, i.e. identifying areas where the transient soil wetness results in the Factor of Safety (FoS) fluctuating above 1 (stable) and below 1 (unstable). We then explore two approaches to account for the soil wetness state for landslide prediction, taking advantage of the hydrological estimates of soil moisture provided by two different models set-up for forecasting purposes and covering Switzerland.
(1) A fully physically-based approach that takes advantage of a state-of-the-art European simulation of hydrology  Percot et al., 2019) with three physically-based coupled models (climate forecast model, land surface model, hydrological model) at a coarse resolution (12.5km×12.5km), from which we extract water table depth at daily resolutions. We downscale these simulations to a higher spatial resolution more appropriate for landslide forecasting following the Topographic Wetness Index method (e.g., Blöschl et al., 2009;Beven, 1995). The high resolution groundwater field is then used as a dynamic component in the Factor of Safety estimation in the infinite slope approach. This framework is designed based on similar 75 existing blueprints for landslide warning systems (Schmidt et al., 2008;Wang et al., 2020).
(2) A probabilistic approach in which we develop rainfall threshold curves for landslide prediction based on a combination of historical databases of rainfall and landslides for Switzerland (Leonarduzzi et al., 2017). We then combine these predictions with estimated soil saturation by a Swiss operational, semi-distributed conceptual hydrological model (PREVAH, Viviroli et al., 2009) to quantify whether there is a strong signal in antecedent soil moisture which could be used in rainfall threshold 80 curve methods for landslide prediction at this scale.
The comparison between the two approaches allows us to answer the following questions: 1) Is the infinite slope approach valuable for landslides hazard assessment at the regional scale? 2) Where does hydrology play a role in the triggering of landslides in Switzerland? 3) Can we generate a computationally feasible physically-based setup for landslide forecasting and early warning at the regional scale? 4) Does soil wetness help improve prediction of landslides based on rainfall thresholds? 5) 85 Which approach is preferable for landslide prediction and why? For the stability assessment, we choose to follow the infinite slope approach, because this is one of the most widely used 90 models for slope failure prediction (e.g., Pack et al., 1998;Iverson, 2000;Baum et al., 2008;Lu and Godt, 2008). It is based on the assumption that the thickness of the sliding mass (soil) is much smaller than the length of the slope, which is typically true for shallow landslides (up to 2m deep). The Factor of Safety (FoS) is computed as the ratio between soil shear strength and applied stress to the soil layer: , γ is soil unit weight [N/m 2 ] (computed from the bulk soil density ρ and gravitational acceleration g as γ = ρ * g) , γ w is the specific weight of water [N/m 2 ], β is the slope angle [rad], and φ is the internal friction angle [rad]. Typically FoS=1 is assumed to be the threshold failure value, with landsliding occurring when FoS<1, i.e. when the applied shear stress exceeds the soil shear strength.

100
All calculations are done at the resolution of the DEM, that is a grid of cell size 25m×25m. This resolution is a result of testing (not reported here), and a compromise between not violating the infinite slope assumptions (length scale of landslides >> their depth), keeping the grid size similar to that of a typical landslide detachment area, but also capturing local topographic gradients β, which are smoothed as resolution decreases.
To estimate the bulk soil density, cohesion, and friction angle we use publicly available datasets OpenLandMap (Open-

105
LandMap; Hengl et al., 2017), which provide global maps of a wide range of soil, land cover, hydrology, geology, climatic and relief characteristics. All the maps used here are available at a resolution of ca. 250m×250m. The soil properties are provided for 6 different depths up to 200 cm produced by machine learning algorithms trained on soil profiles globally (SoilGrids Dataset). For the estimation of the bulk soil density, we compute the thickness-weighted average for the 2m soil column. For the friction angle, we first associate a value for each soil texture class (USDA system) present in Switzerland in the OpenLandMap 110 dataset (Geotechdata.info, Angle of Friction). Then at each location, we choose the value of the friction angle at the depth corresponding to the local soil depth (e.g. if the soil is 120cm deep locally, the closest value in SoilGrids will be that at 1m depth). Finally for cohesion, we assume that the soil itself is cohesionless (c = 0), but we add the important contribution of vegetation to cohesion on slopes. From the landcover map in OpenLandMap, we identify 8 classes of tree cover, and assign them a cohesion value between c = 5 − 22 kPa. Denser tree covers and mixed forests are associated with larger values of cohesion 115 (Schwarz et al., 2012;Dorren and Schwarz, 2016). To quantify the sensitivity of the FoS estimation to the vegetation-driven cohesion, we also simulate the reference case in which cohesion is assumed c = 0 over the entire country ( Figure 1). To assess the susceptibility to landsliding in Switzerland, we compute the Factor of Safety of the two end-member scenarios: completely wet soil (h = d) or completely dry soil (h = 0), which give us the minimum and maximum FoS. This allows us to identify unconditionally stable and unstable areas in our domain just based on hydrology. Unconditionally stable areas have 120 the minimum FoS>1 for a completely wet soil and will never fail regardless of the actual hydrological state. Unconditionally unstable areas have the maximum FoS<1 for a completely dry soil and will (should) always fail according to Eq. 1. In all other areas, hydrology can play a role in the initiation of landslides according to the FoS methodology.
We then compute the dynamic Factor of Safety in time and its statistics for all cells (25m×25m) in which at least one landslide was recorded according to the landslide database (for details on the database see Section 2.2), using the water pressure 125 head h estimated from the hydrological model described in Section 2.1.3. Additionally, we also compute the average FoS during each triggering and non-triggering rainfall event which are defined in Section 2.2. These analyses allow us to observe variations in FoS at cell level and relate them to observed landslide occurrence. If these relations are found to be strong, we hypothesize that a warning system could be set up based on the estimated FoS, combined with forecasted rainfall.

Soil depth
Because soil depth is the most poorly known variable and uncertain parameter in the slope stability model in Eq. 1, here we use four different methods to estimate soil thickness distributions in space and test their impacts on FoS estimates: (1) Uniform soil depth of 1 m for the entire country.
(2) Slope-dependent model (Saulnier et al., 1997): where d max is maximum soil depth, d min the minimum soil depth (assumed to be 5cm), β i is the local slope, β max is the 135 maximum slope above which no soil layer can form (assumed to be 45 • ), and β min the minimum slope (0 • ).
(3) Elevationdependent model (Saulnier et al., 1997): where z i is the local elevation, z max is the maximum elevation, and z min the minimum elevation. (4) Steady state soil depth produced by the linear diffusion transport model (Roering, 2008) where we simulate the distributed soil depth after 15'000 140 years of soil development. This approach is based on mass conservation (Eq. 4), with soil production decreasing exponentially with soil depth (Eq. 5) and soil erosion and transport assumed to be linearly dependent on slope (Eq. 6).
where ∂di ∂t is the change of soil depth in time, q s,i the transport vector at location i, ρr ρs the ratio between the bedrock and soil density (2 as in Dietrich et al., 1995), i the soil production rate at location i, 0 the maximum soil production rate associated with 0 depth (0.000268 m/year as in Heimsath et al., 2001), β the slope at location i, µ the critical value depth (3 1/m as in Roering, 2008), K l the coefficient of linear proportionality (0.0050 as in Dietrich et al., 1995), and ∇z i the gradient of elevation at location i.

150
For the soil depth models (2)-(4), to be consistent, we fix the maximum soil depth d max = 2m.

Hydrology
The water pressure head within the soil layer required for the calculation of the Factor of Safety (h in Eq. 1), is provided by an operational European forecasting system. This consists of the climatology from 1989 to current day, obtained by applying the Terrestrial Systems Modeling Platform (TerrSysMP) (Kurtz et al., 2016). This platform is made up of three physically-based 155 coupled models that solve the water and energy fluxes from the atmosphere to the groundwater: a weather prediction model, a land surface process model, and a hydrological model for surface and subsurface 3D water fluxes. TerrSysMP is produced at daily resolution over a 12.5km×12.5km grid covering Europe. Several state and flux variables are available and can be freely accessed, here we use the water table depth from the surface (wtd, positive when the water table is above the surface).
The coarse 12.5km×12.5 km simulations of wtd by TerrSysMP have to be downscaled to the 25m×25m grid of the geotech-160 nical model so that the high resolution topography which affects the lateral redistribution of soil water (groundwater) is accounted for. We do this using the Topographic Wetness Index (TWI) approach, following the blueprints of previous studies (Schmidt et al., 2008;Wang et al., 2020). This appraoch assumes that the local departure of the water table to the coarser scale average is proportional to the departure of the local TWI from the spatial average (Eq. 7 and Eq. 8): where the index i refers to the local 25m×25m cell considered, and j to the 12.5km×12.5km cell from TerrSysMP in which cell i is embedded, wtd i (t) is the T W I-redistributed local water We combine landslide inventory data in Switzerland and a daily gridded dataset of rainfall to develop rainfall threshold curves following the method of Leonarduzzi et al. (2017). The historical landslides were collected in the Swiss flood and landslide damage database (Swiss Federal Research Institute WSL, Hilker et al., 2009). This database contains floods, landslides, and 180 rockfall events which produced damages in Switzerland since 1972. We select the landslide events that had a known location and date, and were not associated with snowmelt, for a total of 1807 events between 1981 and 2016 (timeframe of the analysis).
The rainfall record is obtained as the interpolation of a network of ca. 430-460 raingauges, using the local climatology and regional precipitation-topography relationships (Shepard, 1984;Frei and Schär, 1998;Frei et al., 2006). It contains daily rainfall totals on a 1km×1km grid covering the entire country since 1961.

185
For the definition of rainfall events, we follow the procedure introduced in Leonarduzzi et al. (2017). First we select susceptible cells, that is rainfall cells (1km×1km grid cells) where at least one landslide was recorded. For those cells we separate the rainfall timeseries into events, where an event is defined as a series of consecutive rainy days with a minimum of 1 dry day in-between events. These events are then classified as observed triggering if a landslide was recorded during or immediately after them, and non-triggering otherwise.

190
We then define a power-law total rainfall-duration (ED) threshold curve E = aD b that objectively separates triggering and non-triggering rainfall events. To this end, we estimate the a and b parameters of the power law curve by maximising the True Skill Statistic (TSS=True Positive Ratio -False Positive Ratio), as in Leonarduzzi et al. (2017). This allows us to classify the rainfall events into the following groups by the calibrated ED threshold (see also Leonarduzzi and Molnar, 2020): observed and correctly predicted triggering events above the ED curve (True Positives), observed triggering events which fall below the 195 ED curve (Misses), observed non-triggering events which fall above the ED curve (False Alarms), and observed non-triggering events which fall below the ED curve (True Negatives).

Antecedent soil saturation
We use the values of soil saturation estimated by the Swiss operational hydrological model PREVAH (Viviroli et al., 2009) at a 500m×500m resolution to check the added value of antecedent soil saturation on the ED curve predictions. PREVAH  Finally, we test the information content of soil saturation for the ED curves, i.e. analyse whether information on soil saturation could reduce some of the misses and false alarms generated by the ED threshold curve estimated in Section 2.2.1. For each group of events (misses, false alarms, true positives and true negatives) and each rainfall event duration (1 to 6 days) we compute the mean soil saturation over 5-60 days prior to the beginning of the event. This allows us to examine if we fail to predict some triggering events (Misses) with the ED curve because saturation was very high, reducing the rainfall amount required for the initiation of a landslide, and likewise if some larger rainfall amounts were insufficient for the triggering (False Alarms) because the soil was very dry before them. We also split rainfall events according to a saturation threshold and define two different optimum ED curves for generally dry and wet conditions. Such curves could be used in a landslide warning context together with estimated current wetness and forecasted rainfall. and results in low soil depth on mountain ridges, but sometimes larger values in convergent topography right next to them.
We then compute the minimum (assuming soil completely wet, h = d in Eq. 1) and maximum (assuming soil completely 230 dry, h = 0 in Eq. 1) FoS for every 25m×25m cell in Switzerland considering all four depth maps. We group the cells as unconditionally stable (when F oS min > 1) and unstable (when F oS max < 1), and conditionally (un)stable (all remaining cells) (Figure 3). The resulting limits of the FoS over the country seem not to be affected strongly by the soil depth model chosen. This is confirmed also when looking at the fraction of cells or landslides in each condition (Table 1). Nevertheless this is not to say that soil depth is not an important parameter for the initiation of landslides, as in Figure 3 and Table 1 we 235 are not considering the interplay of soil depth and hydrology. Considering the two extreme scenarios (soil completely wet or completely dry), we are ignoring how likely these conditions are to occur, and the fact that a thicker soil will likely be more difficult to saturate. Therefore, while soil depth does not seem to impact the limiting conditions for completely wet and dry soil, it will very likely impact the landslide volume and the actual hydrological state and therefore FoS value.
Under the conditions studied here, only 22-25% of the area of Switzerland is conditionally unstable, i.e. area where hydrol-240 ogy matters for landslide occurrence according to the infinite slope model. The presence of so many landslides in unconditionally stable areas (65-66% of the total number of landslides), and the existence of some unconditionally unstable cells (10-13% of the country), are undesirable outcomes. While some inaccuracy in the location of the landslides (which might not refer to the detachment zone) could play a role, these results also suggest that either the infinite slope model is inadequate or the input parameters are inaccurate. In fact, the sensitivity of the FoS to cohesion makes the point (Figure 3 and Table 1)  (unstable or potentially unstable), and the hydrologically active portion, conditionally (un)stable, is now 40% of the country, with more than 60% of the total landslides, and only 19% of the landslides remains in unconditionally unstable areas (3 and Table 1). This is a strong indication that the infinite slope model predictions are highly sensitive to input parameters. These aspects and potential limitations of the FoS will be further discussed in Section 4.

The effect of dynamic hydrology
To address the temporal dynamics of FoS in the susceptible areas (25m×25m cells where at least one landslide was recorded) and their connection to observed landslides, we extract the daily timeseries of simulated TerrSysMP water The expectation is that landslides should occur when the FoS<1. While the value of 1 is often chosen as a theoretical threshold based on the balance of forces in a soil, several studies actually calibrate either the threshold FoS value or the critical area over which FoS<1 in a region (e.g., Casadei et al., 2003). In this work we accept that the critical value FoS=1 can vary spatially depending on the soil parameters and the performance of the hydrological model, and instead we look at the likelihood of landslide-triggering FoS values in individual cells from long-term simulations. To this end, we compute the probability of 260 the FoS being equal or smaller than the value on the day a landslide was observed in the corresponding cell from long-term simulations and repeat this for all landslide cells (Figure 4). This probability should be small, showing that landslides tend to occur when the FoS is smaller in a given susceptible cell, regardless of the actual value, which might be location dependent.
Instead we observe that the probability of the FoS being equal or smaller than the triggering value seems to be small in only few cases (upper panel in Figure 4). Considering rainfall triggering potential of events as a ratio between triggering intensity 265 and mean daily precipitation (Rtrig/mdp), which could help distinguish the events that are extreme, also does not show any signal (middle panel in Figure 4). There are many cases where the non-exceedance probability of triggering FoS equal to 1, which means that a very high FoS led to landsliding, which is contrary to expectations. However, these are also the cases in which the difference between mean FoS and triggering FoS is 0 (lower panel in Figure 4) and the standard deviation is also very small. This means that in those cells (i.e. the majority of cells), the simulated water table is constant in time or is always

Probabilistic approach 280
The role of antecedent wetness and the information content of the saturation estimates provided by the hydrological model PREVAH (Viviroli et al., 2009) for landslide prediction is tested in a similar way to the physically-based approach, but because we do not have an estimate of water table depth in PREVAH, we use the soil saturation instead. We expect patterns opposite to that of the FoS: the saturation to be exceptionally large on landslide days and generally larger during triggering than nontriggering events. This is exactly what we observe considering the probability of the saturation being smaller or equal to that 285 of the landslide day ( Figure 6) and the distribution of average saturation during triggering and non-triggering events ( Figure   7). The probability of a cell being wet at the time of landsliding is very high in most susceptible cells (upper and middle panel in Figure 6). Also the average saturation during triggering rainfall events is clearly larger than during non-triggering events, especially the maximum saturation reached (Figure 7). bining spatial (i.e. differences between landslide locations) and temporal (i.e. differences between events in the cells) differences.
These results suggest that the saturation estimate provided by PREVAH, might contain information useful for the prediction 290 of landslides. While no significant signal in precipitation combined with soil saturation, i.e. triggering rainfall being larger for events with smaller antecedent soil saturation, could be identified in the previous results for individual events ( Figure   6), we choose to combine saturation estimates with ED thresholds for landslide prediction. We first define the optimum ED threshold for landsliding by maximising the TSS (E = 20.1D 0.74 , TSS=0.656), and then compute the average saturation for each duration and class of events: false alarms, misses, true positives, and true negatives. If saturation could improve the ED 295 threshold predictions it would reduce the number of misses and false alarms. Indeed this is what we observe in the mean antecedent saturation trends: regardless of the number days prior to the beginning of the rainfall event over which the mean saturation is computed, the misses (T below in Figure 8) are always associated with the highest antecedent saturation, and the false alarms (NT above in Figure 8) with the lowest saturation. This confirms that at least some of the misses were triggered by a smaller rainfall amount than expected due to exceptionally high antecedent soil wetness, whereas sometimes although the 300 ED threshold was exceeded, no landslide event was observed due to exceptionally low saturation prior to the rainfall event.
While generating hydrometereological thresholds (rainfall-saturation thresholds) does not lead to improved performances compared to the ED threshold, separating the events depending on the antecedent saturation and then optimising separately two ED thresholds does. We obtain a slightly larger threshold (E = 22.1D 0.76 ) for events with antecedent 5 days average saturation below 0.3 than for those with prior saturation antecedent wetness than 0.3 (E = 18.1D 0.74 ), confirming that when 305 the soil is drier prior to the rainfall event, more rainfall is required to trigger a landslide. These combined thresholds lead to an overall TSS of 0.664, so a small improvement in model predictions.

Discussion
The results presented here suggest that the probabilistic approach with rainfall and saturation thresholds is superior to the physically-based approach with the Factor of Safety calculation. It is important to stress that this is not a general conclusion, 310 but rather a conclusion drawn from the specific models which we compared. In fact, if a physically-based approach would accurately capture the pore water pressure variations at the required high resolution scales and therefore reproduce and predict slope failure with the FoS (or another geotechnical) model, we maintain that it would be superior to a probabilistic approach. It is therefore worthwhile to discuss the limitations of the tested physically-based approach and the results obtained with regards to the geotechnical component (i.e. the infinite slope approach and FoS calculations) and those related to the hydrological   are practically homogeneous over the country. We expect that this property is in reality much more heterogeneous and, together with cohesion, is affecting the unconditionally stable area. The sensitivity of the FoS estimates to the uncertain parameters can be examined by Monte Carlo simulations, provided that parameters distributions are known (Hammond et al., 1992;Pack et al., 1998;Griffiths et al., 2011). Soil depth is also a very uncertain and influential parameter. The UU areas, in the alpine region, are very likely steep locations were the soil is absent (exposed bedrock). This aspect is missed by most soil datasets as well as 335 soil depth models.
Another important aspect to consider for the FoS calculation is the spatial resolution. Higher resolutions allow to better capture the local heterogeneities (if data is available), most importantly the topography (i.e. slope). On the other hand, at high resolutions, the assumption of slope length much greater than soil depth becomes invalid and if the cell size becomes much smaller than the typical detachment area of landslides, the interactions between neighbouring cells become even more critical.
The limitations of the hydrological component regardless of the geotechnical model, are evident from the very small information content in the water table depth estimation produced by TerrSysMP. In our analysis we focused on the temporal variability of the FoS (i.e. difference in time relative to the mean for each susceptible cell, or comparison of mean maximum 345 and minimum FoS during triggering and non-triggering events), and the only variable in the FoS calculation which can vary in time is the soil water pressure. This means that the lack of temporal variability in the FoS is a direct consequence of the lack of temporal variability in the water pressure head. The hydrological estimates from the physically-based framework wouldn't be successful even if used with rainfall in hydrometereological thresholds, regardless of the geotechnical component. The signal is instead evident when considering soil saturation obtained from PREVAH. Theoretically, a physically-based model should 350 be better capable of simulating the movement of water in the soil and therefore predicting the saturation or water table depth more accurately.
We believe these results are a direct consequence of the resolution. In fact, at such coarse resolution, the model is simulating the large scale fluctuations in ground water depth, which are indeed generally very small, and does not capture the local changes, driven by higher resolution topography. The downscaling with the topographic wetness approach cannot compensate 355 for this. While it can account for local topography by lowering/raising the water table depth, this correction is constant in time. Therefore, if the coarse hydrological variable does not include enough temporal variability, so will the higher resolution downscaled estimate.
For the specific cases presented here, having a higher spatial resolution (500m×500m rather than 12.5km×12.5km) in a conceptual hydrological model is more beneficial than the gain in accurate physical representation of the processes. This 360 stresses once more the importance of adequate spatial resolution of hydrological models, especially for the assessment of slope and soil saturation dependent natural hazards such as landslides.

Conclusions
We explore two approaches for the prediction of landslides and the value of soil wetness in these predictions applied to a regional scale case study in Switzerland. In the first approach we use the water table depth estimates from a coarse-resolution 365 physically-based model (TerrSysMP) dowscaled by the topographic wetness approach and slope stability assessment using the infinite slope approach. In the second appraoch we use rainfall-duration threshold curves informed by soil saturation obtained by a high resolution conceptual hydrological model (PREVAH).
Our main findings are: the infinite slope approach for quantifying slope instability is largely affected by the accuracy of input soil parameters, 370 in particular cohesion in our case (removing cohesion doubled the area where hydrology mattered in FoS prediction) soil depth does not seem to affect the estimate of (un)conditionally (un)stable areas, although it is an essential parameters for the estimate of local wetness and determines the landslide volume according to the infinite slope approach and the parameters considered here, hydrology can play a role in the initiation of landslides over only ca. 20% of Switzerland (the conditionally (un)stable area, where about 30% of all observed 375 landslides have occurred) saturation estimates from a high resolution conceptual hydrological model (PREVAH) are useful in improving landslide predictions based on rainfall only and can be exploited in combined saturation and ED thresholds while we maintain the best approach would be physically-based modelling of hydrology combined with geotechnical modelling, resolution is a very important aspect and a semi-conceptual model at a high resolution (500m×500m) pro-380 vides more dynamic and useful estimates of the soil status than a physically-based hydrological model run at a much coarser resolution (12.5km×12.5km) despite static downscaling.