Monitoring of the Beauregard landslide (Aosta Valley, Italy) using advanced and conventional techniques

Engineering Geology 116 (2010) 218–235

Contents lists available at ScienceDirect

Engineering Geology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e n g g e o

Monitoring of the Beauregard landslide (Aosta Valley, Italy) using advanced and conventional techniques G. Barla, F. Antolini, M. Barla ⁎, E. Mensi, G. Piovano Dipartimento di Ingegneria Strutturale e Geotecnica, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129, Torino, Italy

a r t i c l e

i n f o

Article history: Received 31 October 2009 Received in revised form 4 August 2010 Accepted 8 September 2010 Available online 16 September 2010 Keywords: Monitoring Deep-seated gravitational slope deformation Radar interferometry Beauregard dam

a b s t r a c t An advanced monitoring technique, based on radar interferometry and implemented by using a ground-based instrumentation (GBInSAR) has been applied for monitoring the Beauregard Deep Seated Gravitational Slope Deformation. This landslide is located in the Aosta Valley (on the Dora di Valgrisenche river), in northwestern Italy, and impinges on a 132 m high concrete arch-gravity dam. This is recognized to have relevant implications in terms of civil protection and poses important territorial and environmental issues. The poor rock mass conditions of the left abutment slope were reported in the fifties, during dam construction. Since 2002, additional geological, hydrogeological and geotechnical investigations have underlined the presence of a deep seated shear zone up to 20 m thick, at the landslide toe. Continuous conventional monitoring over a time span of more than 50 years of both the slope and the dam has allowed to gain insights into the understanding of the behaviour of the basal portion of the slope, with very limited and uncertain point-wise displacement monitored in the upper sector. The GBInSAR monitoring technique has allowed to obtain multi-temporal surface deformations of the upper portion of the landslide, discovering the presence of a main sector in motion, previously unknown, characterized by a total displacement of 45 mm over 4 months. The results of radar monitoring have been validated by comparing with topographic measurements carried out by an automatic total station on 4 targets located at the toe of the slope. © 2010 Elsevier B.V. All rights reserved.

1. Introduction The investigations carried out since the Beauregard dam construction (between 1951 and 1960), located in Aosta Valley (northwestern Italy), have shown that a Deep Seated Gravitational Slope Deformation (DSGSD) is present on the left slope of the valley (Barla et al. 2006). DSGSDs have been widely described by many authors (e.g. Zyschinsky, 1969; Ter Stephanian 1977; Hutchinson, 1988; Agliardi et al., 2001; Ambrosi and Crosta, 2006). Although defined in different ways, in general terms these are slope movements occurring on high relief slopes and with relatively small displacements (Agliardi et al., 2001). The movements produce distinctive geomorphologic features including scarps, counterscarps, trenches, open tension cracks and grabens often associated with double crests and toe bulging. They appear to be best developed in rocks with marked strength anisotropy, particularly in metamorphic rocks such as shales, schists, phyllites, gneiss, and paragneiss (Hutchinson, 1988).

⁎ Corresponding author. Tel.: + 39 011 5644908; fax: +39 011 5644899. E-mail address: [email protected] (M. Barla). 0013-7952/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.enggeo.2010.09.004

The surface deformations typically range from a few millimetres per year to several centimetres per year and are often close to the detection limit of conventional monitoring equipment (Bovis, 1990). Within the deformed mass of DSGSDs, the development of sudden and rapid secondary landslides (rotational and planar slides, falls, topples and flow-like movements) is a common feature. Information regarding geometrical and geotechnical characteristics of DSGSDs at depth are usually lacking, making it difficult to distinguish and correctly understand their behaviour. The development of DSGSDs is strictly influenced by the presence of major tectonic features such as faults, fractures, shear zones and other structural lineaments (Agliardi et al., 2001; Ambrosi and Crosta, 2006). Displacement monitoring of DSGSDs have generally been performed by means of conventional geotechnical monitoring techniques (inclinometers, extensometers, etc.) and topographic or GPS (Global Positioning System) surveys. The information thus provided is limited to a given number of points within the landslide area. In large landslides such as DSGSDs, which are often characterized by different and complex movement patterns, single-point data are not sufficient to evaluate their kinematics and behaviour. Even if the instrumental and topographic measurements are carried out on extensive networks, there are difficulties in achieving spatial and

G. Barla et al. / Engineering Geology 116 (2010) 218–235

continuous information on displacement pattern (Tarchi et al., 2003). This is a critical aspect for high-risk landslides. The Ground-Based Interferometric Synthetic Aperture Radar (GBInSAR) technique can overcome most of these limitations: radar sensors can operate over wide areas in almost any weather conditions, continuously over a long time, providing real time widespread information with millimetre accuracy, without the need of accessing the area. In this paper, following a general updated overview of the geological, geomorphologic and tectonic settings of the Beauregard landslide, both the conventional monitoring system installed on the landslide and the advanced GBInSAR techniques will be briefly described, pointing out the results obtained so far. A distinctive feature of the work carried out is that the GBInSAR results have been validated by comparing them with independent topographic measurements provided by an automated total station. 2. Geological and geomorphologic settings The investigated area lies in the Western Italian Alps, close to the Italian–French border, where tectonic units derived from the Briançonnais paleogeographic domain are present (Fig. 1). Prior to the Tertiary collisional phase between the Apulian and the European plates, responsible for the formation of the Alps, the Briançonnais domain was part of a micro-continent (terrane) bounded by two oceanic zones, the Valaisan oceanic Domain and the more internal Piedmont-Ligurian oceanic Domain (Dal Piaz et al., 2003; Bucher et al., 2004; Schmid et al., 2004). The metamorphic rocks cropping out all along the Valgrisenche valley belong to the Pre-Permian (Guillot et al., 2002) crystalline basement of the Ruitor Unit, which consists of predominant garnet

219

micaschists and albite bearing paragneisses with abundant intercalation of metabasites (prasinites and amphibolites) and records a poliphase tectono-metamorphic evolution related to Pre-Alpine and Alpine orogeneses (Desmons and Mercier, 1993; Caby, 1996; Bertrand et al., 1998; Fugenschuh et al., 1999; Bucher et al., 2004). The rocks belonging to the Piedmont–Ligurian oceanic nappe tectonically overlap the Ruitor Unit basement and consist of calcschists interlayed with marble lenses and matabasites. These rocks crop out in the Valgrisenche area along a tight syncline oriented NE–SW, called by some authors “Avise Syncline” (Caby, 1996). Even though the two slopes of the dammed valley consist of the same rock lithology and formation, they have different characteristics. On the south-eastern slope the rock mass is of good to excellent quality, while on the north-western slope it is highly fractured, disjointed and mylonitic. As reported by Desio (1973), Fig. 2 shows that the basal portion of the left slope overlays a deep pocket of glaciofluvial sediments, which were found at depth during the left dam abutment excavation. The presence of these glacio-fluvial sediments posed very serious problems during construction of the dam and was of great concern for both water flow control and dam stability. During dam construction the pocket of glacio-fluvial sediments was completely replaced by concrete to a depth of 200 m into the left slope. In order to ensure water tightness a main grout curtain was constructed all around the dam to about 100 m below the Dora river bed. The presence of extended rock slices of the Ruitor Unit directly overlaying the fluvioglacial deposits is a clear evidence that after the glacial debutressing the left slope toe has slipped down. Fig. 3 illustrates the geological and morpho-structural map of the Valgrisenche left slope based on the studies carried out so far. Considering the geomorphologic features, the deformation zone involves a significant portion of the slope and extends approximately

Fig. 1. Geological sketch of the western Aosta Valley with indications of the paleogeographical domains. The circle shows the location of the study area (from Boucher et al., 2004, modified).

220

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 2. Geological cross section of the dam site (from Desio, 1973, modified).

1500 m in height from the toe (1700 m a.s.l.) to the mountain ridge (3200 m a.s.l.). Steeply dipping into the slope are predominant subvertical (70°–80°) discontinuities which strike parallel to the valley axis (NE–SW and NNE–SSW), associated to minor shear zones and discontinuities which are nearly orthogonal to them (ESE–WNW). The main schistosity planes along the Beauregard slope dip on average 25°–30° to S and SW. A main head scarp trending E–W with a maximum relief of about 250 m, associated with the double ridge of Becca de L'Aouille Mt. is evident in the north sector of the landslide and it is interpreted as the day-lighted weathered portion of the main scarp delimiting the DSGSD surface of rupture. By Scavarda Refuge the scarp turns to the south forming a wide niche from which a rocky ridge (Scavarda ridge) has been lowered. The sector is highly weathered and fractured and it is affected by rock falls which result in discharge of rock in two main talus at the foot of the cliff, between 2100 and 2300 m a.s.l. In the upper part of the ridge, above the main DSGSD scarp, a swarm of trenches up to 3 m wide trending parallel to the valley is present. These trenches are the continuation towards SW of an alignment of scarps, counterscarps and predominant joints cropping out between the Moriond Glacier and Becca de L'Aouille zone. In the middle and lower portion of the landslide slope it is possible to recognize two distinct ridges to the N (Bois De Goulaz) and to the S (Bochat) characterized by bulging and convex profile. Swarms of trenches up to several meters wide and oriented NE–SW are developed behind both ridges. In the Alpettaz area, between 2000 and 2200 m a.s.l., one major concave scarp is present which can be assumed to be the morphological expression of the basal sector of the landslide moving towards the reservoir. At lower elevations (1800–2000 m a.s.l.), close to the reservoir, numerous minor linear scarps and counterscarps trending parallel to the valley axis can be recognized. The detailed site investigations of the left slope carried out between 2003 and 2006 have allowed to identify a basal shear zone delimiting the more active portion of the DSGSD. As shown in Fig. 4, essential for this interpretation were the results of both seismic refraction and seismic reflection surveys and the results of two boreholes (S1/03 and S1/04) drilled respectively from the left abutment tunnel at 1680 m a.s.l to a depth of 62.7 m and from the slope surface, near Alpettaz, down to a depth of 380.8 m. Both boreholes encountered a thick shear zone (over 20 m in S1/ 04 and over 8 m in S1/03) characterized by the presence of highly fractured rock, locally reduced to a loose silty–clayey material (Barla et al., 2006). A third borehole (S2/04 – total depth 259.3 m) drilled near the Bochat area encountered a major shear zone at depth of

147 m, characterized by a lithology similar to that encountered in S1/ 03 and S1/04 boreholes. The shear zone, 12 m thick, is interpreted as the continuation of the landslide basal shear surface towards the southwestern sector of the slope. 3. Monitoring system Fig. 5 gives the layout of the monitoring system for both the dam and the slope at the site. A description of such a system with reference to both conventional and advanced monitoring (GB-InSAR) is given below. 3.1. Conventional monitoring 3.1.1. The dam Starting with 1960, topographic manual collimations are being performed on 4 targets placed on the dam crest and on additional 4 targets on the left slope, close to the dam (Fig. 5). Precision levelling measurements are also being carried out on the dam top arch and on 4 targets in the 1772 m a.s.l. left slope tunnel. Manual collimation and levelling measurements are performed monthly. Since May 2003, 30 targets placed on the downstream face of the dam are being monitored using a total station. Measurements, which are performed daily, during the night, give information on the displacements in the upstream–downstream, uphill–downhill and vertical direction. It is noted that targets D7 and D5 are placed on the same dam cantilevers as the plumb lines P0 and P4S. Since January 1958 the dam is also being monitored by means of two plumb lines installed in the central (P0) and in the 4S dam cantilever (P4S). Normal and inverted plumb lines allow for measuring relative displacements between the top and the base of the dam (normal plumb line) and between the base and the rock foundation (inverted plumb line). The relative displacements are recorded in two directions: upstream–downstream and uphill– downhill. In the normal plumb line, the upper end of the wire is fixed at the top of the dam. A heavy weight is clamped at its lower end. To prevent any oscillatory movement the weight is damped in a box filled with oil. The inverted plumb line is used for measuring the displacement between the base of the dam and the rock foundation. One end of the plumb wire is fixed at the bottom of the drilled hole in the foundation of the dam. The upper end of the steel wire is linked to a float submerged in a water box in the observation area. The water in the box acts as a damping medium to prevent oscillatory movements of

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 3. Geological and morpho-structural map of the Beauregard DSGSD. Rose diagrams show orientations of structural lineaments and the mean planes of the main discontinuity sets. 221

222

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 4. Site investigations results with indication of the main shear zones identified across boreholes S1/03, S1/04 and S2/04 and interpretation of seismic refraction tomographic lines 3 and 4. For the location of the boreholes and the seismic lines see respectively Figs. 3 and 5.

the plumb wire. In order to obtain the total displacement of the dam crest, the measurements taken at several elevations are cumulated. In order to monitor a visible crack on the dam downstream face, two multipoint extensometers were installed in 2003. One of them was placed in the 2D/1bis hole drilled in the dam from the downstream footbridge (1700 m a.s.l.) up to the pulvino (see the inset in Fig. 5). The multipoint extensometer comprises 6 anchors. Data acquisition takes place 6 times a day by means of vibrating wire displacement transducers and an automated data acquisition system. The cracks in the dam are monitored also by means of two extenso-inclinometers which were installed in 2003. These instruments record the three main displacement components along the vertical hole every meter. Measurements take place 6 times a year between June and November. 3.1.2. The left slope A topographic monitoring system was installed between the time of the completion of the dam to 2009. The most accurate record of the slope movements which extends from 1967 to the present, consists of

geodetic measurements on 22 topographic targets, mostly located on the basal portion of the left slope (Fig. 5). Measurements were usually taken twice a year (June–July and September–October). In May 2003, a total station was installed on the right slope close to the dam guard house. Actually the total station is monitoring 19 points on the left slope; some of them coincide with the manual topographic targets. Measurements are performed daily, during the night. The values obtained for each target are used to calculate the Cartesian coordinates x,y,z referred to the conventional system with axes placed in the K24bis fixed vertex of the manual geodetic measurements. The displacements are referred to the upstream–downstream (x), uphill–downhill (y) and vertical (z) directions. In 1995 a GPS monitoring network was installed on the slope. The network comprises 16 targets placed on the Beauregard landslide, on the upstream side of the dam. The reference position of the network was determined on June 1995 while the more recent GPS measurements were performed respectively in 2000, 2004 and 2007, once or twice a year (in spring–summer and in autumn).

G. Barla et al. / Engineering Geology 116 (2010) 218–235

223

Fig. 5. Monitoring system installed at the Beauregard site. The box on the left shows the position of the plumb lines (P0 and P4S), multipoint extensometer (2D/1bis) and total station benchmarks on the downstream face of the dam.

The target relative positions were determined with reference to a fixed point (FIX 2), located on the right side of the valley using at the same time 4 geodetic dual frequency GPS receivers. Observation lasted 1.5 h for each target and the recording interval for the data collection was kept at 15 s. Starting in 1968, five plumb lines PR1, PR2, PR3 PR4 and PR5 were installed in order to measure horizontal displacements at different depths (Fig. 5). At present only the PR1, PR2, PR3 and PR4 plumb lines are active. The plumb line installed in the slope are a sequence of inverted and normal plumb lines. Each plumb line provides a relative displacement value and total displacement is obtained by summing up these values. In October 2003, 10 standpipe piezometers which were installed at the time of dam construction, on the left slope, were equipped with automatic sensors to measure the ground water level continuously. Two piezometers were added in 2005 and one in June 2006. Measurements take place 3 times per day.

moving, via a linear positioner, a motorized sled hosting the radar head along a straight rail 2 m long. The whole system is controlled by a shock proof PC equipped with specific software. The GBInSAR system was installed at an elevation of 1775 m a.s.l., on the right abutment of the dam. This has ensured a good visibility and a frontal view of the landslide. The mean distance between the GBInSAR system and the centre of the landslide is about 1.3 km, while the maximum distance in the radar image is about 2.8 km. An artificial radar reflector was installed on the basal portion of the left slope to set an easily detectable target to be used for georeferencing the displacement maps on a DEM (Digital Elevation Model) of the area. Monitoring with the GBInSAR system was carried out over 4 months, between the 18th of June and the 17th of October 2008, acquiring about 8300 SAR images in a continuous sequence. Data acquisitions were exactly repeated from the same position, at intervals of about 20 min. The main radar parameters adopted during monitoring are summarized in Table 1. Under these operational conditions the spatial resolution in range was 0.5 m, while the

3.2. Advanced monitoring with the GB-InSAR system Over the last decade, radar interferometry has become a useful technique for landslide displacements monitoring and has demonstrated its capability in the assessment of ground surface displacements over wide areas (e.g. Massonet et al., 1993; Zebker et al., 1994; Tarchi et al., 2003, Antonello et al., 2004; Luzi et al., 2006), also given the improvement of specific ground-based sensors. The radar system adopted at the site is composed of a coherent microwave trans-receiver unit working at Ku band (central frequency 16.75 GHz) which generates a set of sweep of electromagnetic waves of proper duration at different frequencies (Continuous Wave – Stepped Frequencies). The generated signal is then amplified and transmitted to the antennas. The synthetic aperture is realized by

Table 1 Summary of the main operational parameters of the GBInSAR radar measurements campaign. Target distance Central frequency Bandwidth Polarisation Synthetic aperture Linear scansion point number Range resolution (theoretical) Azimuth resolution (theoretical) Antenna gain

400–2800 m 16.75 GHz 100 MHz VV 2m 401 0.5 m 4.5 mrad 20 dB

224

G. Barla et al. / Engineering Geology 116 (2010) 218–235

azimuth resolution was 4.5 mrad (i.e. 2.25 m at 500 m and 9 m at 2 km). The SAR images were processed using a specific software called GRAPeS, based on the PS (Permanent Scatterers) technique (Ferretti et al., 2001). The processing algorithm provides accurate targets displacement assessment, estimating and removing the atmospheric phase screen (APS) effects. Since the reliability of the atmospheric correction is influenced by the sensor-scenario distance, pixels placed at a distance greater than 2 km from the sensor were discarded in the processing stage. The processing procedure is based on a three step approach as follows: • Raw data focusing; • Calibration set generation; • Motion estimation. The first step converts the acquired raw data into focused (or single look complex) images that are defined in range (distance) and in aperture angle domain. The calibration set is a collection of the most stable points available in the scenario. These points are identified as those that will be used for displacement estimation. The identification of the stable points is done through the analysis of the focused images selecting the pixels which show a stable and reliable amplitude and phase (i.e. keeping only the pixels whose coherence is above a specific threshold). In order to correctly locate the displacement maps obtained, it was necessary to determine the spatial transformation from the local coordinate system of the GBInSAR images into the geographic coordinate system (UTM WGS84). The geocoding process can be directly accomplished inside the GRAPeS software using a DEM of the scenario (20 m resolution cell) and at least one point and one known angle in the two coordinate systems. Since the z axis of the SAR images and the DEM are the same, the spatial transformation between the two coordinate systems requires only a rotation parameter and two translation parameters. In the geocoding process the mid-point of the radar rail and the pointing angle (i.e. the angle between N and the line of sight of the sensor) have been used as reference parameters. The former (which is also the origin of the SAR image) has instead been determined in the DEM coordinate system through GPS measurements while the latter has been measured in the field using a compass. The artificial radar reflector previously installed at the foot of the slope has been used as an optional control point in the geocoding process. The corner reflector appears as a bright point in the radar power images and it is possible to easily determine its location in the radar coordinate system.

• Statistical method that establishes a cause-effect relationship in empirical or numerical form. The statistical method consists in approximating the shape of the deterministic indicators by a simple function easier to manipulate. The statistical analysis was carried out for the displacements of the crest of the dam recorded at plumb lines P0 and P4S and at D7 and D5 total station targets. The statistical model defined is a function of the following governing variables: • Water level; • Concrete temperature (related to year and water temperature); • Age of the structure (since its construction). The fundamental formulation of the statistical method is of the form M(t) = P(t) + D(t). M is the observed value of an indicator at a given time t, P is the predicted value for the same indicator and D the difference between the observed and predicted value (i.e. D is the sum of measurement errors). The prediction term P can be separated into a reversible component (hydrostatic component and thermal component) and into a an irreversible component (age of the structure, errors of observation, modelling and numerical calculation). The measured values were approximated by a polynomial function. The analysis consists of estimating the value of the adjustable parameters aj (unknown coefficients of the polynomial function) using a multiple least squared regression analysis procedure. Fig. 6 shows the results of the statistical analysis of plumb lines total displacements. The cumulated displacements at P0 and P4S plumb lines are plotted together with the thermal component, irreversible component and deviation. The least squared regression

4. Monitoring results 4.1. Conventional monitoring In order to analyze the behaviour of the dam versus time a statistical analysis of the monitoring data was performed according to the methods given by the Swiss Committee on Dams (Wasser Energie Luft, 2003). These consist in comparing “measured values” (M), of behaviour indicators, with “predicted values” (P), based on a behaviour model for these same indicators. The measurement is done at a given time t and the prediction must be made for this same time t accounting for the environmental conditions. Different physical quantities can affect the behaviour of a structure. They depend on the type of structure and on the local conditions. Those for which a variation induces a significant change of the behaviour indicator are defined as the governing variables. In order to predict a variable (behaviour indicator) from other variables (governing variables) two fundamental methods are available: • Deterministic method, that links cause to effect considered;

Fig. 6. P0 and P4S plumb lines: comparison between predicted and measured displacements based on statistical prediction model, thermal and irreversible components of displacements, standard deviation.

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 7. P0 and P4S plumb lines versus D7 and D5 total station targets: comparison between predicted and measured displacements based on statistical prediction model, thermal and irreversible components of displacements, standard deviation.

225

was performed with data recorded in an adjustment period of 5 years (2000–2004) and verified in the following 5 years (2005–2009). The prediction curve obtained is plotted together with the measured data. The statistical analysis allows one to observe that the behaviour of the dam is independent from the reservoir level (this is reasonable during the last 10 years when the reservoir level has been always managed around 1700–1702 m a.s.l.). Moreover, the irreversible component of displacement, due to the action of the slope on the dam, is also well highlighted. In fact, the irreversible component of the P0 and P4S plumb lines displacement shows a progressive movement of the dam cantilevers in the upstream direction. The amount of movement in 10 years (2000–2009) is about 23.3 mm for P0 and about 4.9 mm for P4S. The displacements monitored at P0 and P4S are compared in Fig. 7 respectively to those recorded at the D7 and D5 total station targets. Also in this case a statistical approach was used to analyze the total station data, using an adjustment period (2003–2006) and a control period (2007–2009) of 3.5 years. A satisfactory agreement is found between the data recorded at the plumb line station and those obtained by the total station monitoring. A further step in the statistical analysis of data involved the displacements recorded at some total station targets used for monitoring the slope behaviour. Fig. 7 shows the results obtained. 4 targets are considered and for each point the reversible and irreversible effects are plotted together with the measured and predicted values. Considering that the uphill–downhill component is plotted, the irreversible displacement of the slope is about 14.2 mm for K1, 24.4 mm for K4, 14.4 mm for K13 and 33.5 mm for K19 in the downhill direction (Fig. 8). The period taken into account is 6.5 years. With reference to GPS monitoring, the total displacement vectors measured from August 2000 to June 2007 on 9 targets of the GPS network and the related ellipse error are given in Fig. 9. The data indicate that Scavarda and gps8 targets are not affected by significant

Fig. 8. K1, K4, K13 and K19 total station targets installed on the slope: comparison between predicted and measured displacements based on statistical prediction model, thermal and irreversible components of displacements, standard deviation.

226

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 9. GPS monitoring results. Vectors represented indicate the cumulated displacements measured from August 2000 to June 2007 and the related ellipse error.

movements while gps3 and gps2, located close to the main scarp of the landslide, are characterized respectively by displacement of about 10 mm (1.4 mm/year) and 17 mm (2.4 mm/year). The displacements of the other GPS targets installed on the Bochat and Bois de Goulaz ridges are between 32 and 37 mm (4.6–5.3 mm/ year) with vectors oriented along the slope gradient (ESE–SE). However the comparison of the most recent (2007) short term GPS measurements (spring versus autumn campaigns) highlights the presence of a systematic component which masks the annual displacement pattern. This can be related to seasonal variations of both temperature and troposphere conditions that cannot be removed during the compensation phase. Hence, GPS monitoring at the site is able to identify the slope displacements only over a long term (pluriennial) period; on the contrary, the system proved to be unable to measure the seasonal movements because of the geometric and atmospheric site peculiarities. Fig. 10 shows a comparison between different monitoring instruments and helps in the understanding of the slope–dam interaction. The slope movement recorded by the total station is cyclic. The seasonal response of the slope may be outlined only by the plumb lines measurements. In Fig. 10 (a) the cumulated displacements measured at the head of PR3 and PR4 plumb lines are plotted together with the snow height. It is clear that during the spring– summer time the rate of movement increases (see for example the displacements occurred from May to the end of August 2006). Such a movement is a consequence of snow melting and the raising of the piezometric level. This is well shown in Fig. 10b where piezometric level and snow height are plotted. Following snow melting, the standpipe piezometers show a significant increase of piezometric level, up to 15–20 m, which appears to be well correlated

with the displacement triggered in the slope. The raising of the groundwater level is dependent on the amount of snow cumulated during the winter period. This is true only for piezometers located at the toe of the slope, while for piezometers located at elevation greater than 2000 m a.s.l. (S1/04 and CL1) the increase of groundwater level is not influenced by the thickness of snow cover. Considering that the P0 plumb line displays a progressive upstream movement strictly related to the landslide deformation (Fig. 10c), the main crack in the downstream face of the dam is subjected to a progressive opening movement. Fig. 10d shows that the total crack opening monitored between the b1 and b2 anchors of the multipoint extensometer (2D/b2–b1), since the installation of the instrument in 2003, is about 2 mm. Measurements also show that the annual movement cycle is strongly correlated to the seasonal temperature variations. Additional information on the slope behaviour is given in Fig. 11, where the plumb lines data and the topographic measurements from 1977 to 2008 are shown. The cumulated displacements of the P0 plumb line are in good agreement with those obtained by the collimation measurements (COA02), although the vectors referring to collimation show a slight counter–clockwise rotation. A downhill slope movement is noted, except for the collimation vectors exhibiting a greater component downstream. As a result of the slope action, the dam chord becomes shorter and the central dam cantilever undergoes an upstream movement. Furthermore, the leveling measurements of the dam crest show clearly the slope action. The left abutment is lifted up with a maximum upward movement of about 28.4 mm. It is observed that by comparing the displacement rates measured on the same targets by the automatic total station and manual topographic system, some differences arise between the two systems

G. Barla et al. / Engineering Geology 116 (2010) 218–235

227

Fig. 10. Slope cumulated displacements at the top of PR3 and PR4 plumb lines and snow height (a); snow height and water level in piezometers (CL2, M11, PZ1 and S1/04) (b); reservoir level and cumulated displacements at the top of P0 plumb line installed in the main cross section of the dam (c); cumulated displacements at P0 plumb line, 2D/b2–b1 multibase extensometer displacements and air mean temperature (d).

(Fig. 11). The total station highlights the downslope displacement component while the manual topographic measurements indicate both a downslope and downstream component of the movement. Referring to the slope monitoring data, the average displacement rates of targets located in Bonne and Alpettaz areas are about 4– 6 mm/year. In the Bochat area, 1 km upstream from the dam, the average displacement rate increases to 10–13 mm/year. The horizontal downslope displacement represents the most significant component of the slope movement and it is 5 times the vertical component. The average displacement rate is 4.4 mm/year for P0 and 2.7 mm/year for P4S. 4.2. GBInSAR monitoring The results of monitoring with the GBInSAR system provide sequential displacements maps and velocity maps of the scenario

under observation, which allow for a spatial and temporal analysis of the landslide behaviour. A sequence of 6 displacement maps, covering a time span of 121 days, is shown in Fig. 12. The interferometric maps, which are referred to the same master image and to the same calibration set, were obtained through a masking procedure that does not consider low coherence zones. The colour scale of the displacement maps gives the ground movements along the LOS (Line Of Sight) direction: negative values indicate a displacement towards the observer i.e. a sensor to target distance reduction. The colour scale in all the images is saturated towards +/−4.5 mm displacement. All deformation maps were unwrapped during processing and are not affected by the intrinsic ambiguity of phase. The cumulated displacement map, previously geocoded, was then projected on a DEM of the slope and completed with high resolution aerial photographs. As illustrated in Fig. 13, the displacement sequences and their projection on the DEM show an elongated sector

228

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 11. Comparison between the displacement vectors recorded by the head of the plumb lines (continuous lines), manual collimation measurements (dashed lines); vertical displacements of the dam crown between October 1977 and October 2008.

in motion beneath the Scavarda ridge (between 2300 and 2500 m a.s.l. elevation) where the total displacements are in the range of −45 mm over 4 months. The geomorphologic analysis indicates that the sector which is mostly affected by movements belongs to the ridge which has been lowered along the DSGSD main scarp. The ridge consists of altered and fractured Ruitor crystalline rocks. The displacement pattern in this area is due to various instability phenomena (rockfalls, block toppling, and sliding) which result in collapsed materials which tend to accumulate in two main talus, developed between 2100 and 2300 m a.s.l. elevation. The interpretation of the high resolution aerial photographs and the field surveys allowed one to map a well defined trench system

above the area undergoing the most significant movements, between 2500 and 2700 m a.s.l. elevation. These linear and deeply cut morphologic structures are the expression of extensional openings of a vertical or a downward dipping surface (Fig. 14). The fracture system shows a NE–SW trend parallel to the valley axis and can be related to the regional structural lineament already described in Fig. 3. Significant movements characterise also a further sector of the slope between the Alpettaz and Bois de Goulaz ridge, where the total displacements reach a maximum value of −15 mm over 4 months. This area is in motion, as already noted for the sector beneath the Scavarda ridge, and is bounded uphill by a series of minor scarps while a talus is present in the distal portion.

G. Barla et al. / Engineering Geology 116 (2010) 218–235

229

Fig. 12. Sequence of cumulated displacement maps between 18/06/2008 and 17/10/2008 obtained with radar survey. The days elapsed from the start of the sequence are shown on the bottom left of each image. In the top left map some references to topography are reported (to be reproduced in colour in print).

The middle sector of the landslide, between 2100 and 2300 m a.s.l., is characterized by the radar echoes originated from the talus at the base of the Scavarda ridge. The displacements in this area are nearly

−8 mm over 4 months. In the lower part of the slope, between 1800 and 2100 m a.s.l., the lack of coherent radar reflectors limits the interpretation of the data. It is only possible to distinguish the

230

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 13. Cumulated displacement map referred to the monitoring period (8/06/2008–17/10/2008) projected on DEM and high resolution ortho-photos with indication of the total station and GPS benchmarks. The dashed line indicates the DSGSD boundaries (to be reproduced in colour in print).

backscattering echoes due to both the artificial corner reflector and the isolated rock boulders showing displacements within −6 and −11 mm in 4 months. By comparing consecutive displacement maps step by step it is possible to follow the spatial and temporal landslide evolution in detail. The distribution of movement versus time shows a progressive spatial extension of the sector affected by the maximum displacements, beneath the Scavarda Refuge (Fig. 13). The analysis of the rates of displacement points out an increase in the deformation velocity localized in the upper sector of the slope from the end of August to the first ten days of September. 4.3. Comparison between GBInSAR monitoring and topographic measurements The results of GBInSAR monitoring are validated by comparing the pixel displacements with the total station measurements. As shown in Fig. 13, only 4 targets (K1, K4, K13 and K19) are in areas with coherent radar backscattering. In order to compare directly with the radar measurements, the displacement history of the pixels corresponding to each topographic target was taken into account. The total station displacements were then projected along the LOS of the GBInSAR system. The results obtained by comparing the radar and the optical data referred to the same topographic targets are plotted together with the

displacement differences. As depicted in Figs. 15 to 18, the controlling parameters such as rainfall and mean air temperature are also plotted. Moreover each graph includes the distribution of differences and the mean and the standard deviation values. The comparison between the two techniques is acceptable for K1, K4 and K13. If the attention is posed on Fig. 15, which compares the displacements of K1 with the 1066 pixel, a trend of displacement towards the lake is noted, with a mean difference which reaches 3.7 mm. The displacement trend of K4 (482 pixel) and K13 (435 pixel), shown in Figs. 16 and 17 respectively, is nearly the same, with a mean difference not exceeding 2.5 mm. The comparison between the two techniques results to be less favourable for K19 (1220 pixel), in the upper part of the slope, 1.7 km far from the total station. The maximum difference between the radar and the total station measurements increases up to 10.5 mm with high frequency oscillations observed in the optical data. It is clear that the standard deviation of the difference between optical and radar data increases progressively with the distance from the measurement station. Optical data are more affected by dielectric effects which result in high frequency data oscillations. This is thought to be related to daily temperature, humidity changes and rainfall which modify the atmosphere refraction index. It is worth pointing out that the differences between the GBInSAR and optical techniques can be related both to inaccurate spatial correspondence between the radar and the total station targets and to

G. Barla et al. / Engineering Geology 116 (2010) 218–235 Fig. 14. Map showing displacements vectors measured on the landslide monitoring network by GPS (August 2000–June 2007), total station (May 2003–June 2007), manual topographic system (May 2003–June 2007) and the results of the GBInSAR monitoring (June–October 2008). All the measures are expressed in mm.

231

232

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 15. Comparison between displacements of K1 total station target and the related GBInSAR pixel plotted together with mean air temperature and rainfall; the normal distribution of difference between K1 and pixel 1066 is shown in the lower left corner.

their intrinsic differences. Total station displacements are recorded only on a reference point (target) while radar interferometer displacement refers to an area. Based on the work done so far, for distances greater than 1.5 km one may conclude that the results obtained with the total station are less reliable than the GBInSAR measurements.

5. Implications on the landslide mechanism The occurrence of a non-homogeneous deformation pattern along the Beauregard DSGSD, characterized by higher displacement rates in the upper slope (Scavarda ridge) and slow and steady deformations along the basal portion of the landslide, suggests two different trends of behaviour for the two sectors. Morpho-structures such as faults, open cracks and trenches, widespread along Scavarda ridge, undergo

Fig. 16. Comparison between displacements of K4 total station target and the related GBInSAR pixel plotted together with mean air temperature and rainfall; the normal distribution of difference between K4 and pixel 482 is shown in the upper right corner.

Fig. 17. Comparison between displacements of K13 total station target and the related GBInSAR pixel plotted together with mean air temperature and rainfall; the normal distribution of difference between K13 and pixel 435 is shown in the upper right corner.

a brittle deformation, while the basal portion of the landslide exhibits a creep behaviour. Thus, based on the displacement field obtained with the GBInSAR monitoring, a novel interpretation of the upper and intermediate slope sectors is illustrated in Fig. 19. It is worth to say that these displacement patterns represent currently the only nearly continuous observation of the Scavarda ridge. The elongated sector with displacements up to −45 mm over 4 months is located along a portion of the Ruitor unit, where sub vertical faults oriented NNE– SSW (parallel to the valley) are present. It is observed that the displacements decrease towards the top of the Scavarda ridge, where the gps2 target shows less than 5 mm annual displacement. Also, at the base of the ridge, the K19 target shows an annual displacement of about 13.5 mm. Since direct and indirect information at depth are lacking, the more significant displacements detected by the radar survey across the fault zone could be related both to surface differential movements of rocks blocks and to displacements occurring along the active faults and fractures.

Fig. 18. Comparison between displacements of K19 total station target and the related GBInSAR pixel plotted together with mean air temperature and rainfall; the normal distribution of difference between K19 and pixel 1220 is shown in the upper right corner.

G. Barla et al. / Engineering Geology 116 (2010) 218–235

The extension of movements at depth may be related to the widening of the persistent trenches in the upper Scavarda ridge. This displacement pattern, characterised by the opening of tension cracks and showing slope bulging, is thought to be evidence of a progressive translational movement of the upper portion of the slope (Scavarda ridge) along an incipient basal rupture surface. At this stage of the DSGSD evolution, this rupture surface could not be totally developed but characterised by a mode of failure involving stress path and intact rock fractures (Stead et al., 2006). As shown in Figs. 19 and 20, a stepped geometry of the rupture surface may be controlled by the sub-vertical faults and by the schistosity planes dipping towards E–SE. This surface is bounded upslope by the main fault which has lowered the Scavarda ridge along the DSGSD scarp. In particular, based on the analysis of the P-wave velocity profile along the slope sector above the Alpettaz (Fig. 20), one may suggest that this rupture surface extends upwards. The behaviour of the slope at the toe, which impinges directly on the dam, is instead related to the geotechnical characteristics of the cataclastic rock forming the basal shear zone of the landslide, at a depth up to 240 m (Fig. 4), which is known to exhibit a time dependent behaviour (Barla, 2009). The movement patterns of this sector of the DSGSD show a small strain rate of 4–6 mm/year, i.e. creep movements which are likely to be related to the seasonal fluctuations of the water table in the slope, resulting in changes of the effective stresses along the basal shear zone. 6. Conclusions Both conventional and advanced monitoring systems were used to monitor the displacements of the Beauregard landslide, located in the

233

Aosta Valley (northwestern Italy). Since dam construction, a conventional monitoring system, including piezometers, plumb lines, extensometers and topographic measurements of surface targets, has been installed in the lower left slope and in the dam. Monitoring data indicate that the slope is characterized by displacements occurring from spring (May–June) until the early autumn (September–October). In the same period the piezometric levels show a considerable increase, up to 15–20 m, as a consequence of snow melting. This strictly influences the sliding phenomena recorded by the monitoring network. In the basal slope sector the mean displacement rate at surface is 4–6 mm/year. The slope movement is correlated to the dam response and causes a state of damage of the structure. The central part of the dam crest moves upstream as the left slope moves towards the right abutment and the dam arch closes progressively. The dam displacement is partially recovered during the winter when the slope shows negligible movements. In 2008, the application of the GBInSAR technique has allowed to obtain the displacement field of the upper part of the landslide, which is characterized by a good radar reflectivity and coherence. The cumulated displacement maps, covering a time span of 121 days, have evidenced an elongated sector in motion, previously unknown, between 2300 and 2500 m a.s.l. elevation, where a maximum downward displacement of nearly 45 mm over 4 months is reached. The up-to-date displacement data represent a new tool for the global interpretation of landslide mechanism. The comparison between the ground based interferometry and topographic measurements on 4 targets has shown that the precision of the two techniques is almost the same but beyond distances of 1.2– 1.5 km the results obtained by using the automated total station are

Fig. 19. Interpreted geological section E–E' with indication of the cumulated displacement measured along the plumb lines installed on the slope.

234

G. Barla et al. / Engineering Geology 116 (2010) 218–235

Fig. 20. Geomorphologic sketch and interpretative geologic section with indication of superficial deformation pattern along the Scavarda ridge.

markedly affected by atmospheric noise. GBInSAR has confirmed to be a powerful tool for monitoring slope displacements over large landslide areas, even in sectors difficult or impossible to reach. Based on the displacement field obtained with the GBInSAR monitoring, a novel interpretation of the slope sectors has been given. This displacement pattern, characterised by the opening of tension cracks and showing slope bulging, is evidence of a progressive translational movement of the upper portion of the slope along an incipient basal rupture surface. The behaviour of the slope at the toe, which impinges directly on the dam, is instead related to the geotechnical characteristics of the cataclastic rock forming the basal shear zone of the landslide, at a depth up to 240 m (Fig. 4), which is known to exhibit a time dependent behaviour. The results so far obtained and the continuation of the monitoring activities will provide an overall view of the landslide deformations versus time and will contribute to improve in the understanding of

the slope behaviour for landslide hazard management. In particular, numerical modelling is being used to assess the merits of a water drainage adit to be excavated within the slope as a means to permanently maintain lower ground-water levels and minimize slope deformations.

Acknowledgements The work described in this paper was performed with the financial support of the Italian Civil Protection Agency within the framework of the “Monimod” research project, coordinated by Professor Giovanni Barla. The authors would also like to acknowledge the help of Aresys S.r.l. and IDS – Ingegneria Dei Sistemi S.p.a. CVA (Compagnia Valdostana delle Acque), Owner of the Beauegard dam, is also to be thanked for providing the conventional monitoring data.

G. Barla et al. / Engineering Geology 116 (2010) 218–235

References Agliardi, A., Crosta, G., Zanchi, A., 2001. Structural constraints on deep-seated slope deformation kinematics. Eng. Geol. 59, 83–102. Ambrosi, C., Crosta, G.B., 2006. Large sackung along major tectonic features in the Central Italian Alps. Eng. Geol. 83, 183–200. Antonello, G., Casagli, N., Farina, P., Leva, D., Nico, G., Sieber, A.J., Tarchi, D., 2004. Ground-based SAR interferometry for monitoring mass movements. Landslides 1, 21–28. doi:10.1007/s10346-003-0009-6 Barla, G., 2009. Long term behaviour of the Beauregard dam (Italy) and its interaction with a deep-seated gravitational slope deformation. 5th Colloquium. Rock Mechanics-Theory and Practice. Vienna University. 26 Novemeber 2009. Barla, G., Ballatore, S., Chiappone, A., Frigerio, A., Mazzà, G., 2006. The Beauregard Dam (Italy) and the deep seated gravitational deformation on the left slope. Hydropower 2006 – International Conference, October 23–25, Kunming, China, pp. 99–112. Bertrand, J.M., Guillot, F., Leterrier, J., Perruchot, M.P., Aillères, L., Macaudière, J.M., 1998. Granitoides de la zone Houillère Brianconnaise en Savoie (Alpes Occidentales) et en Val d'Aoste: géologie et géochronologie U–Pb sur Zircon. Schweiz. Mineral. Petrogr. Mitt. 80, 225–248. Bovis, M.J., 1990. Rock-slope deformation at Affliction Creek, southern Coast Mountains, British Columbia. Can. J. Earth Sci. 27 (2), 243–254. Bucher, S., Ulardic, C., Bousquet, R., Ceriani, S., Fugenschuh, B., Gouffon, Y., Schmid, S.M., 2004. Tectonic evolution along a transect (ECORS-CROP) through the Italian– French Western Alps. Eclogae Geol. Helv. 97, 321–345. Caby, R., 1996. Low-angle extrusion of high pressure rocks and the balance between outward and inward displacements of Middle Penninic Units in the Western Alps. Eclogae Geol. Helv. 89/1, 229–267. Dal Piaz, G.V., Bistacchi, A., Massironi, M., 2003. Geological outline of the Alps. Episodes 26 (2003), 175–180. Desio, A., 1973. Geologia applicata all'ingegneria. 3a edizione. Hoepli editore. Milano, p. 1193. Desmons, J., Mercier, D., 1993. Passing through the Briançon Zone. In: Von Raumer, J.F., Neubarer, F. (Eds.), Pre-Mesozoic Geology in the Alps. Springer-Verlag, Heidelberg, pp. 279–295.

.

235

Ferretti, A., Prati, C., Rocca, F., 2001. Permanent Scatterers in SAR Interferometry. IEEE Trans. Geosci. Remote Sens. 39 (1), 8–20. Fugenschuh, B., Loprieno, A., Ceriani, S., Schmid, S., 1999. Structural analysis of the Subbriançonais and Valais Units in the area of Moûtiers (Savoy, Western Alps) paleogeographic and tectonic consequences. Int. J. Earth Sci. 8, 201–218. Guillot, F., Schaltegger, U., Bertrand, J.M., Deloule, E., Baudin, T., 2002. Zircon U–Pb geochronology of Ordovician magmatism in the polycyclic Ruitor Massif (internal W Alps). Int. J. Earth Sci. 65, 814–828. Hutchinson, J.N., 1988. General Report: Morphological and geotechnical parameters of landslides in relation to geology and hydrogeology. In: Bonnard, C. (Ed.), Landslides. Proceedings of the 5th international symposium on landslides, Lausanne 10–15 July 1988, pp. 3–35. Luzi, G., Pieraccini, M., Mecatti, D., Noferini, L., Macaluso, G., Galgaro, A., Atzeni, C., 2006. Advances in ground based microwave interferometry for landslide survey: a case study. Int. J. Remote Sens. 27 (12), 2331–2350. Massonet, D., Rossi, M., Carmona, C., Adragna, F., Peltzer, G., Feigl, K., Rabaute, T., 1993. The displacement field of the Landers earthquake mapped by radar interferometry. Nature 364, 138–142. Schmid, S.M., Fugenschuh, B., Kissling, E., Schuster, R., 2004. Tectonic map and overall architecture oh the Alpine orogen. Eclogae Geol. Helv. 97, 93–117. Stead, D., Eberhadt, E., Coggan, J.S., 2006. Developments in the characterization of complex rock slope deformation and failure using numerical modelling techniques. Eng. Geol. 83, 217–235. Tarchi, D., Casagli, N., Fanti, R., Leva, D., Luzi, G., Pasuto, A., Pieraccini, M., Silvano, S., 2003. Landslide monitoring by using ground-based interferometry: an example of application to the Tessina landslide in Italy. Eng. Geol. 68, 15–30. Ter Stephanian, G., 1977. Deep-reaching gravitational deformation of mountain slopes. IAEG Bull. 16, 87–94. Wasser Energie Luft, 2003. Methods of analysis for the prediction and the verification of dam behaviour. 21st Congress of the International Commission of Large Dams, Montreal, June 2003. Zebker, H., Rosen, P., Goldstein, R., Gabriel, A., Werner, C., 1994. On the derivation of coseismic displacement fields using differential radar interferometry: the Landers earthquake. J. Geophys. Res. 99, 19617–19634. Zyschinsky, U., 1969. Uber Sackungen. Rock mech. 1 (1), 30–52.