Draft A novel technique to monitor the subsurface movements of landslides

Slope deformation sensors (SDSs) were developed to monitor profiles of soil deformation at a high frequency during slope-monitoring and landslide-triggering experiments. It was hypothesized that th...


Introduction
Combining information about the direction, magnitude, and rates of surface movements with the subsurface deformations of unstable slopes provide valuable data that can be used to estimate the possible volume of a moving landslide, and to understand the triggering mechanisms of instabilities Green 1973;Buchli et al. 2013;Buchli et al. 2016).
The instrument used most commonly to monitor the onset and continuation of deformation in slow moving slides is the slope inclinometer.The deformations are measured normal to the axes of a borehole by passing a probe inside a specially formed plastic casing Dunnicliff 1988).
However, the soil deformation profile measured using a slope inclinometer system in a zone of extreme shear, may not represent the actual soil deformations.The PVC casing has a specific stiffness, which means it may not follow the soil deformation exactly and there could be differential lateral movement between casing and soil.Moreover, significant deformation at the shear zone may distort the casing and prevent the probe from travelling easily along the tube e.g.Arenson et al. 2002).Furthermore, data is only obtained when probe is lowered and raised manually.
However, monitoring results of several landsides in sandy slopes reveal that the precursors take place within a short time before the failure, and generally, the soil movement happens at high speeds e.g.Chen et al. 2004;Cascini et al. 2005; Take and Beddoe 2014).Similar observations have also been reported for physical modelling of landslides due to rainfall on loose sandy slopes e.g.Moriwaki et al. 2004;Ochiai et al. 2004;Take et al. 2004;Picarelli et al. 2006;Take and Beddoe 2014;Askarinejad et al. 2015).Therefore, it is essential that the subsurface deformation measurements are carried out remotely, for safety reasons, and at high frequencies.

D r a f t 4
As an alternative to conventional slope inclinometers, ShapeAccelArray/Field SAAF) systems can be used.These sensors are facilitated with high frequency data acquisition Bennett et al. 2009).SAAFs are installed inside PVC tubes of 32 mm outer diameters with a thickness of approximately 7 mm, which makes them too stiff for shallow slopes that are generally prone to instability due to rainfall Arenson et al. 2002).Moreover, they cannot be easily used to monitor movements of shallow layers above the bedrock, given the segment length of 0.5 m.Therefore, low flexural stiffness, high frequency of remote sampling, and relatively simple installation procedure are not achievable with conventional subsurface slope monitoring sensors.
Soil Deformation Sensors SDS) were developed at the Institute for Geotechnical Engineering at ETH Zurich to monitor the subsurface movements of a silty sand slope, and to investigate the precursors of landslides induced by rainfall.The sensors were installed and tested in a 38° steep, instrumented forest slope, which was subjected to intense artificial rainfall events.The main goal of the slope monitoring and landslide triggering experiments was to investigate the coupled hydrogeo)logic and geo-mechanical behaviour of the slope, and to interpret results to determine the triggering mechanisms.
The stages of development and laboratory testing of the SDSs are discussed, and the results of the bending strain and deformation measurements during a slope monitoring experiment October 2008) are presented and compared to the characteristics of surface movements of the slope, ambient atmospheric and applied rain conditions.Details of the data acquired from SDSs during the landslide triggering experiment March 2009) are reported and analysed in a second paper.

Soil Deformation Sensors SDS)
A Soil Deformation Sensor is a slender aluminium ALMg1) plate with a rectangular cross section of 40 mm width) x 2 mm thickness).The bending stiffness EI) of this plate is about 300 times less than that of the PVC casing of a typical slope inclinometer or a SAAF inclinometer, without considering the stiffness of the grout needed to fill the gap between the casing and the borehole wall Table 1).Higher flexibility of the SDS improves their sensitivity to detect and resolve fine movements of the soil mass.
Several pairs of strain gauges were installed at a pre-determined spacing on the aluminium plate Figure 1).The strain gauges were connected as a "half Wheatstone bridge" Wheatstone 1843) to eliminate the temperature effects.The bending strain at each point is defined as the difference between the magnitude of positive tensile) and negative contractile) strains on the sides divided by 2. The strain gauges are covered by a thin flexible plastic layer to protect the wires and to waterproof the device.A double-axis inclinometer was installed on top of the SDS, which is located about 200 mm above the soil surface, to monitor the inclination of the upper part of the sensor.However, air temperature has a major effect on the readings of these sensors.
Two angle profiles were connected to the lower 200 mm of the SDSs with 8 bolts diameter 6 mm; length 50 mm), to increase the stability of the sensor at the base.The lower part of the SDS was grouted into the bedrock.

Principles of deformation measurement
The Soil Deformation Sensors can be analysed as encastre vertical cantilever beams Figure 2).
The sign of the bending moment is chosen such that a positive value leads to a tensile stress in the D r a f t 6 downslope face.The sign of the shear force has been chosen such that it matches the sign of the bending moment.
It is assumed that the transversal plane sections remain plane and normal to the longitudinal face during bending Euler-Bernoulli assumptions in beam theory).Experimental measurements show that these assumptions are valid for long, slender beams made of isotropic materials with solid cross-sections Bauchau & Craig 2009).
The angle of the deformed SDS to the initial condition, θ Figure 2b) is calculated as follows: where, ܴ ଵ is the radius of curvature at the upslope face Figure 2b), ܴ ଶ is the radius of curvature at the downslope face Figure 2b), S 1 is the length of upslope face of the bent section in contraction) Figure 2b), S 2 is the length of downslope face of the bent section in extension) Figure 2b) and ߠ is the angle of deformation measured relative to the vertical.The following equation can be written to relate S 1 , S 2 , and the longitudinal strain: where ‫ܮ‬ is the initial length of the bent element Figure 2b), and ߝ is the axial strain at the outer faces of the section.As the line with angle θ to the vertical is the tangent to the curved section, the deformation profile will be derived by integration of the θ -curve.Boundary values are needed to carry out the integrations, which were defined by the fixed base of the SDS, i.e. y = 0 displacement), and θ rotation) = 0.The integration will be performed from the encastre base of the SDS to the top.
where y i is the amount of deformation at strain gauge i, and ∆݈ is the distance between strain gauges i-1 and i.
The radius of curvature at the neutral axis of the section, according to Equations 1 and 3, will be: where ߢ is the curvature of the beam.The bending moment M) is equal to: where, E is Young's modulus of aluminium 69 GPa), I is the second moment of inertia of the cross section 26.67 mm 4 , Table 1).

D r a f t 8
The shear force is: And the distributed lateral load per unit area of the probe is: The total horizontal soil pressure acting on the probe can be calculated based on the distributed load per unit area.
There is a change in the sign of bending strain close to the point where the failure surface crosses the SDS Figure 2a).Therefore, the depth of the failure surface can be estimated by determination of the depths of adjacent strain gauges with different bending strain signs.A pair of strain gauges must be installed at the lowest part of the SDS in the soil above the fixed section in the mortar, to gain information about the deflection of the sensor in case the failure surface develops at the interface of the soil and bedrock.
Bending strains at different points along the Soil Deformation Sensor and the inclination at the top were sampled at a frequency of 100 Hz.These were input into an algorithm based on Equations 1 to 4 to determine the relationship between deformations and rotations with depth, so that the initiation of slow movements, and propagation of a failure zone during fast soil mass movements could be monitored.

D r a f t
9

External bending work
The external work done on a beam with length L due to bending can be defined as Freudenthal 1966): where ߪ ௭௭ and ߝ ௭௭ are the axial stress and strain due to bending, respectively, ‫ݒ‬ is the volume SDS and A is the area of the cross section of the bent element Figure 3).
The bending strain at each depth of the cross section y) is related to the inverse of the radius of curvature ߢ) according to beam theory for pure bending: where R is the radius of curvature due to bending.Equation 13 is derived by combining the three Equations 10 to 12: where, n is the number of strain gauges, and Δ‫ܮ‬ is the spacing of the strain gauges.The value of the external bending work per unit volume of SDS is calculated as:

Laboratory tests on Soil Deformation Sensors
A 1 m x 1 m x 1 m wooden box with a transparent side window was constructed to prove the concept and validate the performance of the SDSs, according to geometries illustrated in Figure 4.A 0.8 m long SDS was fixed in the box by two angle profiles to guarantee zero movement and zero rotation at the base, and installed behind a laterally moveable wooden plate.The inclination of the upper part, and the bending strains along the SDS, were measured and saved with a frequency of 10 Hz.
The box was filled with dry Perth sand at a void ratio of 0.65.Perth sand is a uniform sand with critical state internal friction angle of 33° Nater 2005) Table 2).The wooden plate was pushed towards the SDS by a hydraulic jack and the movements of the soil grains were tracked using the digital image analysis technique of Particle Image Velocimetry White et al. 2003).The initial distance between the wooden plate and SDS was 70 mm.The location of the passive failure wedge was compared to the position of neighbouring strain gauge bridges, showing different signs of bending strains.The intersection of the shear zone with the SDS was in agreement with D r a f t the location of maximum deflection of the inclinometer and a change in the bending sign is observed below and above the shear zone Figure 5).

Numerical modelling of the laboratory test
The Finite Element Method was used to simulate the laboratory tests in plane strain using  2).The Soil Deformation Sensor was modelled as a plate structural element with isotropic elastic constitutive model with geometrical and mechanical properties as listed in Table 1.
A predefined displacement boundary condition was used to simulate the moving wooden plate in the box.The results of bending strains of SDS are compared to the measurements in Figure 5.
These results confirm the change in the sign of the bending strain at the shear zone.However, the values of bending strain are slightly different from those measured in the laboratory test.This discrepancy can be due to the differences in the simulated plane strain condition with the real three-dimensional nature of loading and the effects of a protective rubber cover of the SDS on the bending stiffness of the sensor.
The results of both physical and numerical models confirmed the assumption of the change in the bending direction of the SDS at the shear zone and hence the ability of the Soil Deformation Sensor to detect the depth of the failure surface.However, back-calculations of the sub-surface soil displacements are based on the assumption that the response of the Soil Deformation Sensor D r a f t 12 remains elastic.Therefore, the elastic bending limit of the Soil Deformation Sensor must be determined, so that the device can be applied within the acceptable range.
The bent part of the SDS is approximated by two circular arcs with radius of R and an opening angle of α Figure 7a).The yielding strain of the AlMg1 material is 1.43x10 -3 [-] and the minimum radius of curvature for the SDS to remain in an elastic state is calculated to be 0.7 m Equation 6).The corresponding opening angle and radius of deflection arcs can be calculated based on the thickness of the shear band t) and the shearing displacement ∆ߴ).
Knowing the values of t thickness of the shear band, Figure 7a) and ∆ߴ, the values of R and ߙ can be calculated.A new parameter of t* is defined, for simplicity, which is equal to ∆ߴ ‫ݐ‬ ⁄ .The value of ߙ can be calculated using the following equations: The value of R also can be calculated from Equation 15, when ߙ is known.The value of critical shear displacement can be determined for each thickness of the shear band, based on the minimum radius curvature 0.7 m) and Equations 21 and 15.The critical values of ∆ߴ, as a function of the thickness of the shear band, are illustrated in Figure 7b.

Test area
The experimental test site was located on an east facing slope on the banks of the river Rhine in Four Soil Deformation Sensors were installed in different instrumentation clusters, two in the upper cluster SDSs 3 &4), one in the middle SDS 1), and one in the lower cluster SDS 2)).The length of the SDSs was selected based on the depth of the bedrock, which was determined by dynamic probing Figure 8).The length of the SDSs, and position of the strain gauge pairs on each inclinometer, are summarised in Table 3.These sensors were constructed in the laboratory and transported carefully to the field with minimum vibration and bending.
A mortar was prepared with water, cement, super-plasticiser, micro-silica, fine sand and gravel aggregates to have high workability and rapid hardening.Boreholes were drilled at the predefined locations on the slope to a depth of 200 mm inside the weathered bedrock.The Soil Deformation Sensors were installed inside the borehole and fresh mortar was guided through a 20 mm inner diameter) pipe to the lower end of the sensors to fill the lower 200 mm with mortar.
The gap between the sensor and the borehole wall was then filled by pouring dried, and sieved soil from the field with a grain size less than 2 mm Figure 11).

Applied rainfall
Artificial rainfall was applied by means of 10 oscillating garden sprinklers Gardena Aqua-zoom 250), which were aligned on the middle longitudinal line of the slope with an on-surface spacing D r a f t of 2.5 m Figure 10).The lower sprinklers experienced higher hydraulic heads as the water was supplied from water tanks above the slope.The applied rain intensity, measured at Clusters 1 and 3 are shown in Figure 12).

Field monitoring results
The slope was subjected to a 4.5-day intense artificial rainfall event, with the intention of causing a landslide of about 150 -200 m 3 in volume.Although this experiment was unsuccessful in terms of triggering a landslide, valuable data and information, regarding the hydro-mechanical responses of the slope during this severe rainfall event, were obtained and are reported here.

Surface and subsurface movements
Photogrammetric analysis showed the greatest movements occurred in the upper right quarter of the slope Figure 13), where bedrock has been found to be shallower 0.5 to 1 m deep) and less permeable than in the lower half of the slope Figure 8) and the soil was less reinforced by the roots than in the bottom half of the slope Figure 9).Therefore more movements had been expected in this area.The extents of the eventual failure in 2009 are also shown in Figure 13, which matches well with the concentration of movements in this experiment.The accuracy of the photogrammetric measurements in the Z direction refer to Figure 2 for coordinates) is about 5 times higher than that in the horizontal directions, ± 3.4 mm vs. ± 16.5 mm, respectively).The maximum movement that was measured by this method is less than the accuracy in the horizontal directions.Therefore, the movements in the Z direction are translated into "on-surface" movements ∆ℎ ) using the Equation 22: where, ߸ is the average angle of the slope surface 38°).It is assumed that the target points follow a trajectory parallel to the soil surface.These values are used in the contour plot of Figure 13.
The " on-surface" movements of the slope, at the location of the SDS 1, 3, and 4, determined by this method are shown in Figure 14 and compared to the measurements of SDSs.The surface movements of the slope measured using the Soil Deformation Sensors are smaller than those measured using the photogrammetry method by an average ratio of 0.75.This difference be an indication of the effect of the stiffness of the sensors on the measurements, which could hinder them from moving freely together with the surrounding soil, although the bending stiffness of these sensors is about 300 times less than that of the conventional slope inclinometers.Moreover, it should be noted that the measured values of the surface deformation are smaller than the accuracy of the photogrammetry method.
Five out of 8 strain gauge bridges installed on SDS2, which was located in the lower part of the slope, showed faulty records; therefore, the results are not presented here.The deformed shapes of the 3 functioning Soil Deformation Sensors 1, 3, and 4 are presented in Figure 14 at different stages of rainfall.
The Soil Deformation Sensor located in the upper right quarter of the slope SDS 3) showed downslope movements at all depths from the surface to 1.23 m Figure 14b), until the peak rainfall event shown by the vertical dashed line 1 in Figure 15a at time 2:53 on 30.03.2008, approximately 61 hours after the start of the rainfall).The measurements show a gradual increase in movement with cumulative rainfall.However, the strain gauges at depths of 890 mm and 720 mm measured divergences in the sign of the bending strain increments about 5 hours after the D r a f t 18 sudden increase in the rain intensity dashed vertical line 2 in Figure 15a).This divergence in the bending strain is an indication of the formation of a slip surface at this depth, based on the results of the scoping physical and numerical models.The location of this secondary and shallower slip surface is shown in the dashed box in Figure 14b and an enlarged view is depicted in Figure 15b shown with large arrow.
Consistent downslope movements of the neighbouring SDSs 3 and 4 showed a linear displacement profile with depth, with a maximum displacement of between 2.4 and 1.7 mm at the top strain gauges, respectively.SDS1 exhibited a double-parabolic form, showing notably lower downslope displacements that were nonetheless pronounced at a depth of 1.3 m.This shear surface caused negative bending strains in the strain gauges located below it, and positive strains in the upper adjacent ones sign convention: Figure 2a).The absolute value of positive bending strain can be significant due to passive pressures imposed on the inclinometer above a shear zone.
These localised significant positive strains introduce bending in the reverse direction in the Soil Deformation Sensor.This is seen in SDS1, where upslope deflections are calculated between depths of 0.40 m and 1.10 m after the initial generation of a shear zone.A shallow slip surface is also detected at a depth of around 0.40 mm.

External bending work
The evolution of the external bending work per unit volume of the Soil Deformation Sensors is illustrated in Figure 16a.The general trend of these curves shows an increase in the applied work after the start of rainfall.These graphs illustrate higher amounts of applied mechanical work from the surrounding soil to the SDSs in the upper part of the slope, where higher values of surface movements were observed and also the eventual failure occurred in the landslide triggering experiment in March 2009 Askarinejad et al. 2012).

D r a f t 19
The changes in the ground water level measured by a piezometer installed at a depth of 3 m in the middle part of the slope location shown in Figure 13) are also reported in Figure 16a.A similar trend is seen in the measurements of the bending strains Figure 15a), external bending work and the water table fluctuations Figure 16a) in the upper and middle sections of the slope.The absolute value of these measurements are either constant or increase very gently until approximately 5 hours after the sudden increase in the rain intensity about 61 hours after the start of rainfall.All of these parameters increase for approximately 24 hours shown by vertical dashed line 3 in Figure 15a and Figure 16).The decreasing trend of the absolute bending strains and external bending work measured by the SDSs indicate a decrease in the downslope velocity of the soil mantel.This observation confirms the coupling of the hydro-mechanical responses of the slope.

Summary and Conclusions
Simple and low cost Soil Deformation Sensors were developed to monitor surface and subsurface movements of a steep forested slope in Northern Switzerland that was subjected to artificial intense rainfall events.The data from these sensors can be sampled remotely at high a frequency.
The performance of SDSs in detecting the failure surface depth was validated by laboratory experiments and numerical modelling.Thereafter, four of them were installed in loose silty sand at different locations in the slope.
The general trend of movements, based on analysis of the measurements from the SDSs during the course of the 4.5 day rainfall experiment, indicated that the slope has experienced a downslope surface movement of a magnitude of 0.5 to 2.5 mm, with more pronounced deformations in the upper part of the slope.Comparing these results to the surface deformation D r a f t 20 measured by a 4-camera photogrammetry method, confirmed that the measured surface deformations are similar, although the values are not exactly the same.
The results of the movement measurements were analysed together with the external bending work per unit volume of each Soil Deformation Sensor, which was determined as an indication of the mechanical energy transmitted from the surrounding soil to the sensor.Hence, the locations with more transmitted mechanical energy are identified in less stable areas.The results are in agreement with the measured surface displacements, and eventual failure area of the landslide that occurred 5 months later, due to artificial rainfall applied on the same slope.
It can be concluded that the Soil Deformation Sensors are able to detect the depth of the failure surface, based on the results of extensive laboratory physical modelling, numerical simulations.
They are very responsive to fine soil movements since their bending stiffness is lower by a factor of 300, compared to the conventional slope inclinometers.Enhanced sensitivity, high frequency of remote sampling, and relatively simple installation procedure make them suitable for the early warning systems of shallow less than 2.5 m depth) landslides triggered by rainfall.
Calculation of the deformation profile using the measurements of Soil Deformation Sensors is based on the assumption of an elastic bending along the main body of the sensor.This criterion limits the maximum reliably measurable deformation profile of the slope and is governed by the thickness of the shear band.However, the measured bending strains along the sensor are not limited by this criterion and therefore can be used as indications of the state of the slope at large deformations prior to the onset of failure.
The limit states and levels of bending strains and external bending work measured by these sensors need to be set for different alarm statuses based on the tolerable risk, as well as the D r a f t 21 geotechnical and hydro-geological properties of each individual site.Moreover, the data from these sensors need to be combined with measurements of pore water pressure and, ideally, with the horizontal earth pressure in an effective EWS as the safety of a slopes subjected to pore water pressure increase is significantly influenced by the stresses parallel to the slope measurements Picarelli 2000;Leroueil et al. 2009).-2085-1770-1455-1140-825-51-195

Figure captions
the external bending work per unit volume of the bending part, and ‫ݒ‬ is volume of the SDS.The changes in U are directly related to the changes in applied energy by the soil mass acting on the Soil Deformation Sensors.Greater work exerted by the surrounding soil mass might also be an indication of either higher unbalanced pressures on both sides of the sensors, or a sign of larger movements in the surrounding soil.
PLAXIS softwareBrinkgreve et al. 2014) and to numerically investigate the hypothesis of the ability of SDS in detection of the shear band depth.The geometry of the model, boundary conditions, meshing and displacement field are shown in Figure 6a.Elastoplastic Mohr-Coulomb parameters were used for Perth sand Table Figure 8 a).The cross section of the slope along the middle longitudinal line of A-A' Figure 8 a)

Figure 1 :
Figure 1: Sketch of Soil Deformation Sensors and arrangement for the strain gauges, which were connected as a half Wheatstone bridge to eliminate the temperature effects.The lower 200 mm part of the SDS was fixed into the bedrock using a mortar.

Figure 2 :
Figure 2: a) Coordinate system and sign convention.b) The initial and deformed shapes of the sensor.

Figure 3 :
Figure 3: Cross section of a Soil Deformation Sensor after Askarinejad 2013).

Figure 4 :
Figure 4: Schematic view of the calibration setup of the Soil Deformation Sensor after Askarinejad 2009).

Figure 5 :
Figure5: The measured and numerically calculated bending moments along the SDS at different stages of shearing the length of the SDS is 800 mm and the bending moment is determined from the strain gauge measurements, the legend represents the amount of the lateral movement of the wooden plate).

Figure 6 :
Figure 6: Geometry, boundary conditions, meshing and the displacement field of the finite element model of the SDS in the wooden box after 1 mm of horizontal moving of the pushing plate.

Figure 7 :
Figure 7: a) The schematic profile of a deformed Soil Deformation Sensor within a shear band with thickness of t. b) The critical values of shearing displacement of the Soil Deformation Sensors SDS) as a function of the thickness of the shear band t).

Figure 8 .
Figure 8. Shape of the bedrock, positions of the sensor clusters and the Soil Deformation Sensors SDSs).b) Cross section of the slope and geological lithology A-A').c) Geological layers of the slope after Brönnimann 2011).

Figure 9 .
Figure 9. Spatial distribution of maximal root reinforcement in the study area after Schwarz 2011).

Figure 11 .
Figure 11.Installation sequence of soil deformation probes: 1) Installation of SDS in the borehole, 2)Placing mortar in the bedrock zone, 3) Waiting for 30 minutes for the mortar to gain the initial strength 4)

Figure 12 .
Figure 12.Artificial rainfall applied to the upper and lower and parts of the slope during the slope monitoring experiment in October 2008.

Figure 13 :
Figure 13: Contours of "on-surface" movements, after the slope monitoring experiment, based on the photogrammetry results, location of the SDSs and piezometer Pz).The extends of the eventual landslide in 2009 are marked by dashed line.

Figure 14 :
Figure 14: Deflection calculated for the Soil Deformation Sensors during the slope monitoring experiment, a) SDS1 cluster 2, middle of slope), b) SDS3 Cluster 3 RHS, top of slope), and c) SDS4 Cluster 3 LHS, top of slope).

Figure 15 :
Figure 15: a) Bending strains for SDS3 at depths of 720 and 890 mm below the surface plotted with rain intensity, b) Enlarged view of the profile of movement of SDS3, Figure 14b, showing the development of a secondary shallower failure surface.

Figure 16 :
Figure 16: a) External bending work per unit volume of Soil Deformation Sensors and water table rise in the middle of the slope, b) Applied rain intensity and cumulative rainfall.

Figure 1 :Figure 2 Figure 3 :Figure 4 :Figure 5 :Figure 6 :Figure 8 .Figure 9 .Figure 10 .Figure 11 .Figure 12 .Figure 13 :Figure 14 :
Figure 1: Sketch of Soil Deformation Sensors and arrangement for the strain gauges, which were connected as a half Wheatstone bridge to eliminate the temperature effects.The lower 200 mm part of the SDS was fixed into the bedrock using a mortar.

Table captions Table 1 :
Bending stiffness of the Soil Deformation Sensor compared to that of the PVC casing of a slope inclinometer after Askarinejad 2009).

Table 2 :
Geotechnical properties of Perth sand.

Table 3 .
The length and the location of the strain gauge pairs for each SDS.

Table 3 .
The length and the location of the strain gauge pairs for each SI.