dynamics in landslides in varved Characterization of groundwater dynamics in landslides in varved clays

Groundwater dynamics may play a signiﬁcant role in landslides. A detailed model is developed of the groundwater dynamics in landslides in varved clays in the Tri`eves area in the French Alps. The varved clays consist of a sequence of alternating silt and clay layers, covered by a colluvium layer and cut through by ﬁssures. The hydraulic 5 conductivity of the clay layers is negligible compared to the silt layers. It is conceptualized that ﬁssures form a hydraulic connection between the colluvium and the varved clays. Groundwater recharge ﬂows through the colluvium into the ﬁssures where water is exchanged horizontally between the ﬁssure and the silt layers of the varved clays. Groundwater ﬂow in the colluvium is simulated with the Boussinesq equation while ﬂow 10 in the silt layers of the varved clays is simulated with the Richards’ equation. Longitudinal outﬂow from the ﬁssure is simulated with a linear-reservoir model. Scattered data of relatively short monitoring periods is available for several landslides in the region. A good similarity between observed and simulated heads is obtained, especially when considering the lack of important physical parameters such as the ﬁssure width and 15 the distance between the monitoring point and the ﬁssure. A simulation for the period 1959–2004 showed some correlation between peaks in the simulated heads and the recorded occurrence of landslides while the bottom of the varved clays remained saturated during the entire simulation period. thin, conﬁned layers. The head in the ﬁssure acts as a boundary condition for both subsys-tems. The ﬁnite di ﬀ erence method is used to compute the head in the colluvium and varved clays. The water level in the ﬁssure is computed by means of a water balance. Simulated heads were compared to measured heads from four landslides in the Tri`eves area; results of the Avignonet landslide were presented as a representative example. Scattered data of relatively short monitoring periods are available. The simulated head in the colluvium is very similar to the measured head for all three monitoring periods considered. The simulated and measured heads in the varved clays do not match as well, but the general patterns show reasonable resemblance. It is di ﬃ cult to


20
The hydrological mechanisms taking place in landslides are often not investigated extensively, whereas their importance is frequently recognized (e.g. Reid and Iverson, 1992;Iverson, 2000;Wilkinson et al., 2002;Lindenmaier et al., 2005;Schulz et al., 2009;Wienhöfer et al., 2010;Bogaard et al., 2012). As a result, the effects of changes in climate or land use on landslide occurrence are difficult to predict (e.g. Buma and flow through fissures in landslides is an example of a hydrological process that is discussed frequently, but rarely quantified ( Van Beek and Van Asch, 1999;. Landslides may often be conceptualized as heterogeneous, cascading systems, which are difficult to simulate with standard hydrological models (e.g. Uchida et al., 2001;Van Asch et al., 2007). To make matters worse, there is often a significant 5 lack of subsurface data.
In the Trièves region in the French Alps, varved clays are highly susceptible to landslides (Giraud et al., 1991). At least 15 % of the area covered by the varved clays is considered as sliding (Jongmans et al., 2009). Displacement velocities range from several millimeters to several centimeters per year, with the risk that the velocities may 10 suddenly increase and lead to catastrophic failure. Infiltration of precipitation is the most important triggering mechanism for landslides in varved clays, as increased pore water pressure on the slip plane results in a reduction in stability (Van Asch et al., 1996, 2007Van Genuchten and Van Asch, 1988; Van Genuchten, 1989;Vuillermet et al., 1994).
Varved clays consist of alternating laminae of silt and clay. The hydraulic conductivity 15 of the clay laminae is extremely low, while that of the silt laminae is larger. The hydrogeology of the varved clays is strongly influenced by fissures that develop as a result of displacement. The hydrogeological system may be described as a cascading system, where fluxes from surface layers interact with the underlying varved clays through fissures ( Van Genuchten and Van Asch, 1988;Nieuwenhuis, 1991;Asch et al., 1996).

Hydrological conceptual model of Trièves area
Varved clays are characteristic for the geology of the Trièves area in the French Alps. These clays were deposited in a glacial lake of around 300 km 2 during the last glacial or Würm period (30 000-10 000 yr ago). The lake developed as the Isère glacier blocked the valleys of the river Drac and its tributaries (Fig. 1) ( Van Genuchten, 1989;Giraud et al., 1991). Varved clays are sediments with alternating laminae of silty and clayey material of mm to cm scale ( Van Genuchten, 1989;Giraud, 1988;Giraud et al., 1991;Huff, 1989;Vuillermet, 1994). The finer-grained clay layers were deposited when the lake surface was frozen. The coarser-grained silt layers were deposited during warmer periods. Huff 15 (1989) measured silt to clay ratios of 0.23 to 1.4 in the Trièves area. The thickness of the varved clays varies from 0 to 200 m over short distances due to the irregular subsurface. The varved clays were deposited on top of carbonate bedrock and are covered by a colluvium layer of 1-4 m thick. The colluvium layer is relatively permeable compared to the varved clays (Van Asch et al., 1996). The average hydraulic conductivity 20 of varved clay deposits is highly anisotropic due to their horizontally layered texture; the permeability of the clay laminae is negligible compared to the permeability of the silt layers.
Numerous landslides have occurred in the Trièves area in the last century. The landslides are rotational or translational slides with slip surfaces at depths ranging from 25 relatively shallow (4-8 m) to deeper (20-40 m) (Van Asch et al., 1996). This movement induces vertical fissures (Bièvre et al., 2012;Nieuwenhuis, 1991) Van Genuchten, 1989) and mostly parallel to the contour lines (Bièvre et al., 2012;Nieuwenhuis, 1991). The fissures reach different depths and are connected to the slip surfaces of the landslides (Van Asch et al., 1996;Bièvre et al., 2012;Nieuwenhuis, 1991;Vuillermet et al., 1994). The fissure density varies over the area and also within the landslides. For example, in a landslide near La Mure most fissures 5 are found in the downslope part ( Van Genuchten, 1989). In the Avignonet landslide the fissure density is largest in parts where the slope angle is relatively high (Bièvre et al., 2012). Field observations by Van Asch et al. (1996) indicate that the distance between 10-20 m deep fissures is about 40 m; fissures with depths and distances of only a few meters are found as well (Nieuwenhuis, 1991). The fissures are partly filled with clay and silt and are not always visible at the land surface. Some of the fissures are connected across the contour lines which allows downslope drainage of the landslide through the fissure network (Nieuwenhuis, 1991). A schematized cross section of part of a landslide in varved clays is shown in Fig. 2. The groundwater dynamics in varved clay landslides can be summarized as follows 15 (Van Asch et al., 1996;Vuillermet et al., 1994). Precipitation infiltrates into the colluvium layer, where perched groundwater develops above the varved clays. Water flows from the colluvium into the vertical fissures and infiltrates horizontally into the silt layers of the varved clays. Conversely, water drains from the silt layers into the fissures when the head in the silt layers exceeds the head in the fissures. Vertical flow in the varved clays 20 is considered negligible. The landslides are drained in downslope direction through a system of connected fissures, normal to the plane of flow shown in Fig. 2.
The stability of a landslide is a direct function of infiltration from fissures at the level of the slip surface. Not all fissures are connected to the slip surface. Shallow fissures mainly influence the pore water pressure in the upper part of the varved clays, and 25 the infiltration into and drainage of the colluvium. The pore water pressure at the slip surface may be influenced indirectly, due to a change in the inflow into deeper fissures. In this paper the hydrological system is approximated as shown in Fig. 2, but with all fissures at one location reaching the same depth. The distance between two fissures is called 2L and the depth of a fissure is called d (Fig. 2); both vary over the area. The exact distance and depth at the measurement locations are unknown. Two standard systems are considered: one with a large fissure distance and depth, referred to as the large system, and one with a small fissure distance and depth, referred to as the small system (Table 1). Model simulations are 5 presented for these two standard systems and compared to head observations from monitored landslides using the standard system that is most appropriate.

Mathematical model
The conceptual model consists of two subsystems: the colluvium and the varved clays (Fig. 2). The fissure connects both systems; the head in the fissure acts as a bound-10 ary condition for each system. The system between two fissures is approximated as horizontal, and horizontally symmetrical. The x-axis runs from the fissure (x = 0) to the center between two fissures (x = L) (Fig. 2). Flow in the colluvium and varved clays is approximated to occur in a vertical cross-section.
Flow in the colluvium is written in terms of the hydraulic head h c [L] in the colluvium,

Colluvium layer
The colluvium layer is an unconfined aquifer. Water level variations are the result of variations in the groundwater recharge. The water level drops as water flows out into the fissure. Only saturated flow is simulated in the colluvium. The head in the colluvium, h c , is governed by the Boussinesq equation with areal recharge (e.g. Bear, 1972).
where k c [L/T] is the hydraulic conductivity of the colluvium, S y [-] is the specific yield of the colluvium, N [L/T] is the areal recharge, and t [T] is time.
The recharge N is defined as the gross precipitation minus evaporation as discussed 5 in the next section. Boundary conditions are that the head at x = 0 is equal to the head in the fissure, and that the boundary at x = L (halfway between two fissures) is a noflow boundary because of symmetry where h f,c is the water level in the fissure relative to the varved clays-colluvium interface 10 [L]; when the water level in the fissure drops below the bottom of the colluvium, the head at the boundary is set to zero. Equation (1) subject to boundary conditions (2) is solved using a standard implicit finite difference approach (for details see Van der Spek, 2011). A dynamic equilibrium is used as initial condition. 15 Water infiltrates in horizontal direction from the fissure into the silt layers, or drains from the silt layers into the fissure, depending on the head gradient. Flow in each silt layer is modeled as one-dimensional flow independent of the other silt layers. The head h s in the silt layers is measured with respect to the bottom of the colluvium/top of the varved clays. When the silt is saturated, the pressure head in the silt layers, φ s , is governed 20 by the one-dimensional diffusion equation (e.g. Bear, 1972)

Varved clays
HESSD 10,2013 Groundwater dynamics in landslides in varved clays where k s [L/T] is the saturated hydraulic conductivity of the silt and S s [1/L] is the specific storage of the silt. When the silt is not saturated, the water content changes in time and in space. The water content and the pressure head are governed by the one-dimensional Richards' equation where θ [-] is the water content, C = ∂θ/∂ϕ s [1/L] is the moisture capacity, and K s (φ s ) [L/T] is the hydraulic conductivity of silt as a function of φ s . The unsaturated hydraulic conductivity and moisture content are represented by the the Mualem-Van Genuchten model ( Van Genuchten, 1980). The boundary conditions are the same as for the collu- 10 vium, but now written in terms of pressure head where φ f,z is the pressure head in the fissure at elevation z [L]. The pressure head φ f,z decreases linearly to a minimum of −1 m at 1 m above the water level in the fissure, which represents that the moisture content never drops below field capacity. Flow in the 15 silt layers is again solved numerically using a standard implicit finite difference scheme. As for the colluvium, a dynamic equilibrium is used as initial condition.

Fissures
The water level in the fissure is calculated by means of a water balance. There are three components (see Fig. 3): inflow from the colluvium (Q c ), inflow into the varved HESSD 10,2013 Groundwater dynamics in landslides in varved clays outflow from varved clays into the fissure. The water balance for the fissure is written as where w [L] is the width of the fissure. The water balance is solved numerically (using a constant time step) in three steps. First, the water level in the fissure resulting 5 from inflow from the colluvium is calculated simultaneously with the water level in the colluvium. Second, the new water level in the fissure is used as a boundary condition for calculating the pressure head in the varved clays. The pressure head in the varved clays is used to compute the inflow from the varved clays into the fissure (Q v ) and the head in the fissure is adjusted accordingly. Third, the longitudinal outflow from the 10 fissure (Q f ) is simulated with a linear reservoir function where K [T] is the reservoir coefficient, and the head in the fissure is updated accordingly. 15 Mean yearly precipitation in the Trièves region is 1050 mm yr −1 as measured by Mete-oFrance in Monestier de Clermont in the period 1948-2005. Average daily precipitation ranges from 2 mm day −1 in July to almost 3.5 mm day −1 in November. Monthly evaporation is approximated using the Thornthwaite equation (Thornthwaite, 1948) and a sine curve is fitted to obtain daily values. Actual evaporation is approximated as 75 % of 20 potential evaporation for summer conditions and potential evaporation in winter conditions.

Model evaluation
In this section, model performance is explored using rainfall and evaporation data from 1958, which is representative for the   an equilibrium situation was reached, where the total yearly recharge equals the total yearly outflow from the fissure, and the head variation in the colluvium and clays is the same from year to year.

Dynamic equilibrium 1958
The dynamic equilibrium for the large system is shown in Fig. 3. In the upper graph, the 5 recharge of 1958 is shown with bars (left vertical axis), and the simulated water level in the fissure is shown with a line (right vertical axis). In the lower graph the variation of the outflow from the colluvium (Q c ), the water exchange with the varved clays (Q v ), and the outflow from the fissure (Q f ) are shown. The total inflow into the varved clays is equal to the total outflow from the varved clays, and constitutes 7.6 % of the total 10 outflow from the colluvium.

Dynamics of individual systems
The where T is the period of the fluctuation. When β v decreases, the damping and phase 20 shift increase (e.g. Haitjema, 1995). Consider, for example, the case where the saturated hydraulic conductivity of the varved clays is decreased tenfold. The head variation at the bottom of the varved clays (which is always saturated) is plotted against time for the large system for the year 1958 in Fig. 4 damp out quicker when the hydraulic conductivity is smaller (dashed lines). Head variations in the colluvium react in a similar manner. In the colluvium, head variations are governed by Eq. (1) and the damping and phase shift are controlled by where h c is an average head in the colluvium.

Dynamics of entire system
The behavior of the entire system is considered next. The fluctuation of the water level in the fissure, and hence the pressure variation in the varved clays, is an interplay of the three main processes: inflow from the colluvium, exchange with the varved clays, and longitudinal outflow from the fissure. Three factors are discussed here: the effect 10 of the distance between two fissures, the effect of the model parameters of the fissure, and the effect of unsaturated conditions in the varved clays.
(1) The main effect of a larger distance between two fissures is a larger head halfway the fissures (x = L) and an increased outflow from the colluvium into the fissure. The increased flow into the fissure results in a higher water level in the fissure and higher 15 heads in the varved clays.
(2) The fissure hydrology is modeled as a linear reservoir (Eq. 7) and the water level in the fissure is computed with a water balance (Eq. 6). A smaller fissure width w or a larger reservoir constant K results in a larger water level in the fissure. These two parameters are correlated. Water level fluctuations in the fissure show more short-term 20 variations when w and K are smaller.
(3) Not surprisingly, pressure propagation in the varved clays is much slower when the varved clays become unsaturated near the fissure. The fissure is never completely empty in the model of 1958 for either of the standard systems ( water level in the fissure in the large system never drops below 8.2 m. In the small system it never drops below 0.4 m. The pressure head in the varved clays is contoured at the beginning of 1958 (t = 0) and at the moment when the water level in the fissure reaches a minimum in Fig. 5. In the large system, a relatively small part of the varved clays becomes unsaturated (negative pressure heads, Fig. 5b). Unsaturated flow oc-5 curs only close to the fissure in the upper part of the clays. In the small system, a large part of the varved clays is unsaturated at the moment of a minimum water level in the fissure (Fig. 5d). The influence of unsaturated flow is larger for the small system than for the large system. 10 Data of four landslides in the Trièves area are available: La Mure, Avignonet, St. Guillaume, and Monestier du Percy (Fig. 1). The data is very scattered (both temporally and spatially) as long-term monitoring has not been performed in this region on any of the landslides. Sparse data orginate from relatively short duration projects, mainly during or after increased landslide activity. As a further complication, the distance between an 15 observation point and the closest fissure is not known for any of the monitoring points. Parameter values for varved clays were obtained with laboratory or field experiments as reported in the literature. Parameters for the colluvium and fissures are not available and were determined through calibration. The saturated hydraulic conductivity of the silt is set to k s = 0.001 m day −1 , the specific storage is set to S s = 0.001 m −1 , the poros-20 ity is set to n p = 0.45, and the Van Genuchten parameters are set to α = 0.28 m −1 , and n = 1.16 (e.g. Nieuwenhuis, 1991;Vuillermet et al., 1994). Four parameters are obtained from calibration: the hydraulic conductivity of the colluvium k c = 0.1 m day −1 , the specific yield of the colluvium S s = 0.1-0.3, the width of the fissure w = 0.1 m, and the reservoir constant K = 5-28 days. Because of the calibration data quality, a qualitative, expert-driven calibration was performed instead of a formal best-fit approach. The large range for the specific yield of the colluvium may have several reasons. First, Introduction what is called the colluvium here may contain parts of alluvial or morainic origin, which have different properties (Van Asch et al., 2009). Second, the specific yield may be influenced by the presence of a number of (small) fissures in the colluvium. The model was calibrated for four different landslides in the Trièves region (Fig. 1). Model initiation was done using the 1958 precipitation and evaporation data. Results of 5 the Avignonet landslide are discussed here as an example. Simulations of groundwater dynamics in the other landslides gave similar results (Van der Spek, 2011). Heads were measured during three periods. From July 1985 until January 1987 the head was measured in fourteen piezometers at different locations and depths ranging from 4.5 to 30 m below ground surface (Blanchet, 1988) penetrating both colluvium and 10 varved clays. From January 2004 until September 2005 the head was measured in two piezometers: one in the colluvium and one in the varved clays, at a depth of 43 m. From July until September 2008 the head was measured in two piezometers (Bièvre et al., 2012): one at a depth of 5 m and one at a depth of 47 m. The piezometers deeper than 20 m are located below the maximum fissure depth used in this study, 15 and therefore not included in the calibration. The simulated groundwater levels depend on the distance between the fissure and the piezometer. As this distance is unknown, heads are simulated at the fissure (x = 0) and halfway two fissures (x = L).

Application to landslides in the Trièves area
The Avignonet landslide has a thickness of 4-5 m and slip surfaces at different depths: a few meters, 10-17 m, and 42-47 m (Jongmans et al., 2008b). This means 20 that the situation of this landslide may be comparable to either of the standard systems of Table 1. The small standard system gave the best results for the first monitoring period. The large standard system gave the best results for the other two monitoring periods. Results are shown in Fig. 6. The top row of graphs represents the recharge during the monitoring periods; the second row of graphs represents the head vari-25 ation in the colluvium; the third row of graphs represents the head variation in the varved clays. Measurements in the fourteen piezometers of the first monitoring period are summarized by the grey area in the first column of graphs in Fig. 6 measurements that were taken in the colluvium or the varved clays. For the second and third monitoring periods, measured heads are compared only to measurements in the colluvium (black dots and black line). Good results are obtained for the colluvium. It is more difficult to obtain a good match in the varved clays. This is partly due to a lack of information, especially the depth of 5 the fissures and the distance between the fissures and the monitoring locations. Furthermore, variations of the head measured in open standpipes may show a delayed response caused by the low hydraulic conductivity of the varved clays (e.g. Van Asch et al., 2009). Nonetheless, the measured and simulated heads show significant similarities.

Groundwater dynamics compared to landslide activity 1959-2004
In this section, groundwater dynamics are simulated for the period from 1959 to 2004. Data on landslide activity in the Trièves region are compared to simulated groundwater dynamics; long-term groundwater records are not available for such a comparison. The model was run with the precipitation data of Monestier de Clermont from 1956 to 2004. 15 The first three years are used for model initiation.

Groundwater dynamics
Results for the large system are shown in Fig. 7. The graphs in Fig. 7

Comparison to recorded landslide activity
Increased pore water pressure on the slip surface of a landslide results in a stability reduction. High landslide activity is expected to coincide with very high heads in the varved clays. Data on landslide activity in the Trièves region is compared to the simulated groundwater dynamics of Fig. 7. Dates at which first signs of movement of several 5 landslides were observed in the Trièves, mainly slides and debris flows, are recorded in the landslide database of BRGM (http://www.bdmvt.net/). The landslides were selected while the movements labeled as "mud-debris flow", "collapse", and "stream erosion" were not considered. Records for which the date of occurrence has an accuracy of approximately 1 month are used for a qualitative analysis. A subset of 19 landslides 10 for which the date of occurrence is known with an accuracy of approximately one day is used for a tentative quantitative analysis. Most landslides in the Trièves start to move between February and April. Annual peaks in the simulated heads in the varved clays occur in these same months (Fig. 7). The magnitude of the simulated head is, however, not the only trigger, as high heads 15 of the same or higher magnitude are also simulated in other years where no significant landslide activity was recorded. This indicates specific local circumstances are important to trigger a landslide. Interestingly, the model results show clear cycles of wetter and drier years, resulting in higher and lower heads in the system. These trends, however, do not emerge from the database of landslide occurrence in the Trièves. simulated heads are highest. The simulated heads for the years in which these events occurred are relatively high compared to those for other years, although heads of the same magnitude or higher also occur at other times.

Discussion
The presented conceptual model is based on a number of approximations. Three is-5 sues are discussed here. First, the conceptual model presented here is compared to the conceptual model of Van Asch (1996). Second, the processes causing a timevarying fissure width and their possible effects on groundwater dynamics are described. And third, the possibility of air entrapment in silt layers is discussed. 10 Van Asch et al. (1996) describe a hydrological conceptual model equivalent to the model used in this research, but use different equations to represent the groundwater dynamics. They use a combined linear reservoirs model to simulate the water fluctuations in the colluvium and fissures and use an approximate formula for infiltration into the varved clays rather than formally solving Richards' equation. One of their conclu-15 sions is that the mean annual residence time of water in the fissures is not sufficient to fully saturate the varved clays each year. This is based on the assumption that the varved clays are unsaturated (pressure head between −1.5 and −0.5 m and associated degree of saturation of approximately 0.9) before a rise in head in the fissure occurs. Simulations of the two standard cases in this paper show that the silt layers at 20 fissure depth are always saturated during an average year. Assuming that the fissures reach the slip surface, this means that the pore water pressure at the slip surface is determined by pressure propagation in saturated soil, which is a faster process than saturation. Not only the head in the fissures, but also the head in the varved clays varies considerably and may influence the stability of the landslide. The extent to which the variations are damped and delayed depends on the distance from the fissures.

Fissure geometry and time-varying fissure width
Fissures are a crucial part of the conceptual model, as they constitute the connection between the colluvium and the varved clays (e.g. Nieuwenhuis, 1991;Vuillermet et al., 5 1994;Van Asch et al., 1996;Bièvre et al., 2012). In this study, longitudinal drainage from the fissure is a function of the water level in the fissure only, not of the water levels in the remainder of the fissure network. Head fluctuations in the fissures are the result of fluctuating precipitation and evaporation during the year, water exchange with the varved clays, and drainage from the fissure. Nieuwenhuis (1991) reports a dynamically changing fissure width at the La Mure landslide. He describes how water collects in channels above the fissures at the downslope edge of the landslide, which results in increased infiltration time for water to reach the slip surface. The stability of the downslope part of the landslide decreases due to the increased pore water pressure. New fissures develop between the moving and 15 stagnant parts of the landslide, or existing fissures open up to the slip surface. Water can reach the slip surface via these fissures. In this way the instability of the landslide develops retrogressively, so in upslope direction. In this paper, the width of the fissure is constant through time. A time-varying fissure width may be required in case of highly dynamic interactions between fissure geometry and mass movement, as system be-20 havior is very sensitive to the width of the fissure.

Air entrapment in silt layers
Air entrapment in thin silt layers is not included in the model presented in this paper. Vuillermet et al. (1994) performed a field experiment to measure the horizontal infiltration rate in varved clays. When water infiltrates, air may get trapped in unsaturated silt 25 layers, which are confined between two clay layers. Pneumatic pressures develop at the infiltration fronts and inhibit the horizontal infiltration. Air may only escape through the infiltrating water, which is a slow process. In the model presented here, the pressure propagation in the silt layer at the slip surface is not directly influenced by air entrapment, as the bottom of the varved clays is always saturated. There is an indirect effect, however, as air entrapment of silt layers may slow down infiltration from the fissure, 5 which in turn means that the fissure fills up quicker.

Conclusions
In this research the hydrogeology of landslides in varved clays is conceptualized as follows. The varved clays and the colluvium are two subsystems, which are connected by vertical fissures. Precipitation infiltrates into the colluvium and results in a perched water table in this layer. Outflow from the colluvium enters the fissure. Water infiltrates horizontally from the fissure into the silt layers of the varved clays, resulting in a pressure rise. When the water level in the fissure drops, water drains from the varved clays to the fissure. The system drains in downslope direction by a network of connected fissures.

15
The conceptual model is translated into a mathematical model to simulate the groundwater dynamics in landslides in varved clays. The colluvium is modeled as an unconfined aquifer. The silt layers in the varved clays are modeled as separate, thin, confined layers. The head in the fissure acts as a boundary condition for both subsystems. The finite difference method is used to compute the head in the colluvium and 20 varved clays. The water level in the fissure is computed by means of a water balance.
Simulated heads were compared to measured heads from four landslides in the Trièves area; results of the Avignonet landslide were presented as a representative example. Scattered data of relatively short monitoring periods are available. The simulated head in the colluvium is very similar to the measured head for all three monitoring 25 periods considered. The simulated and measured heads in the varved clays do not match as well, but the general patterns show reasonable resemblance. It is difficult to obtain a good match in the varved clays as the distance between the observation well and the fissure, and the distance between fissures, are not known for any of the monitoring locations. The good match in the colluvium and reasonable match in the varved clays are evidence that the most important hydrological mechanisms are included in the model, especially when considering the limitations of the available data.

5
The fissure geometry plays a major role in the hydrological system of the varved clays in the Trièves region. Measurement of the width and depth of the fissures and the distance between the fissures are needed to improve the accuracy of the model simulations for specific locations. Geophysical methods can be used to determine the fissure geometry (Bièvre et al., 2012). Measurement of the width of the fissures at different moments during a year may elucidate the effect of changing fissure conditions, as discussed by Nieuwenhuis (1991). Analysis of the model results leads to increased insight into the most important factors affecting the groundwater dynamics in varved clay landslides. Both the width and reservoir coefficient of the fissure have a large effect on the magnitude of the head 15 in the fissure and the varved clays. Changes in fissure width during slope deformation may affect the pressure development in the varved clays. The head variation in the fissure is a boundary condition for the head variation in the varved clays. The (unsaturated) hydraulic conductivity and specific storage of the silt layers affect the extent to which the head variation in the varved clays is delayed and damped. A proper 20 estimation of the values of these parameters is important in simulating the pressure distribution and hence in assessing the stability of the landslide. Model simulations for the period 1959-2004 indicate that the heads are largest during the first months of the year. Landslide activity is generally largest during the same months. The effect of wet and dry years is visible in the model results. A combination of the hydrological model 25 with slope stability and slope displacement analyses may provide further insight in the effects of groundwater dynamics on landslide activity.