An Instrumented Flume to Investigate the Mechanics of Rainfall-Induced Landslides in Unsaturated Granular Soils

Geotechnical Testing Journal, Vol. 32, No. 2 Paper ID GTJ101366 Available online at: www.astm.org Lucio Olivares,1 Emilia Damiano,1 Roberto Greco,1 Luigi Zeni,2 Luciano Picarelli,1 Aldo Minardo,2 Andrea Guida,1 and Romeo Bernini3 An Instrumented Flume to Investigate the Mechanics of Rainfall-Induced Landslides in Unsaturated Granular Soils ABSTRACT: The mechanics of rainfall-induced flowslides in pyroclastic soils have yet to be completely clarified. The complexity of phenomena (rainfall-induced failure in initially unsaturated granular deposits, post-failure transition to flow-like landslide) requires the use of a well-equipped small-scale flume. To this aim, flume experiments at the Second University of Naples were performed to analyze the fundamental aspects of such phenomena. A new experimental program is now being carried out to assess the performance of a time domain reflectometry device and optical fibers as indicators of impending failure. The paper describes the instrumented flume and the procedures adopted for monitoring the major aspects of slope behavior. Our first experimental results are very promising in this respect. KEYWORDS: instrumented flume, flowslides, pyroclastic unsaturated soil, TDR device, optical fibers Introduction Every year, rainfall-induced flowslides cause loss of life and damage to property in Campania (southern Italy), one of the most densely populated areas in Europe. To mitigate the risk involved, which threatens thousands of people, substantial institutional and research efforts are being made, but radical mitigation of the consequences of flowslides can be obtained only when complete understanding of the causes and mechanics of movement is attained. To this aim, physical modeling is extremely useful. Due to the rapidity of movement, which makes in situ monitoring very difficult, physical modeling alone makes it possible to capture important aspects of slope behavior and to check relevant hypotheses concerning the mechanical process which governs the transition from slope rupture to flow. At the Geotechnical Laboratory of CIRIAM (Interdepartmental Research Centre in Environmental Engineering), an instrumented flume was built to investigate the mechanics of rainfall-induced slope failure and flowslide initiation in unsaturated granular soils (Damiano 2004). At present, research is being focused on those indicators of impending failure which can help set up new systems for rapid assessment of flowslide development. Flowslide Occurrence in Campania In the last ten years, a number of rainfall-induced flowslides have caused almost 180 victims and extensive destruction in urban and Manuscript received July 25, 2007; accepted for publication November 26, 2008; published online January 2009. 1 Department of Civil Engineering/Interdepartmental Research Centre in Environmental Engineering, CIRIAM, Second University of Naples, via Roma 29, 81031 Aversa (CE), Italy. 2 Department of Information Engineering/Interdepartmental Research Centre in Environmental Engineering, CIRIAM, Second University of Naples, via Roma 29, 81031 Aversa (CE), Italy. 3 IREA-CNR, via Diocleziano 328, 80124 Naples, Italy. extra-urban areas of Campania. The most damaging landslides occurred in 1998, in Sarno, Quindici, Bracigliano, and Siano, but other killer flowslides took place in 1997, 1999, 2005, and 2006. Such events are far from new: similar phenomena are recorded from earlier in the 20th century, while others are mentioned in the chronicles of previous centuries. Flowslides involve primary pyroclastic deposits consisting of alternating layers of unsaturated volcanic ash (nonplastic silty sand) and pumice (gravelly sand), which cover steep slopes. Typically, the maximum overall thickness of such deposits is a few metres, but triggered landslides can progressively incorporate the material found along their path, reaching a size of some tens of thousands of cubic metres. Movement can attain a peak velocity in the order of a few tens of metres per second (Faella and Nigro 2003), and displays a flow-like manner, as testified by eye-witness accounts and films showing that the soil mass moves like a flood, filling tracks and concavities present on the ground surface, and stopping only on flat areas or because of impact against structures. After arrest, the mud swiftly consolidates, taking the aspect of a dry soil. Several authors (Sladen et al. 1985) assume that flowslide generation is the result of soil liquefaction; i.e., of shear strength decrease due to rapid accumulation of positive excess pore pressure. However, liquefaction can develop only where the following conditions are satisfied (Olivares and Picarelli 2001): (i) the soil is either saturated or nearly fully saturated; (ii) slope failure and initial soil movement can cause the buildup of excess pore pressure; (iii) the soil is susceptible to liquefaction; (iv) the rate of soil deformation is so fast as to impede excess pore pressure dissipation or to regenerate excess pore pressure during movement. Due to the low height of the capillary fringe of pyroclastic soil, the first condition is usually found only below the water table; in the case of steep slopes in Campania, the shallow pyroclastic deposits are above the water table and unsaturated, and saturation can be reached only as a consequence of long lasting heavy rainfall. The occurrence of the second condition can be the result of several mechanisms, such as volumetric collapse due to saturation, rapid Copyright © 2009 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. 1 2 GEOTECHNICAL TESTING JOURNAL progressive failure (Olivares and Picarelli 2006) and impact (Cairo and Dente 2003), which may lead to rapid soil deformation. The third condition can be tested in the laboratory through undrained shear tests. By testing natural or reconstituted specimens, it has been shown that liquefaction is typical of loose gravelly sand or silty sand (Hunter and Fell 2003) with a nonplastic fine-grained component (Yamamuro and Lade 1998). The fourth condition generally applies in the case of steep slopes, since slope rupture and liquefaction cause a dramatic acceleration of the soil mass. Therefore, excess pore pressure remains high or regenerates during movement as a consequence of internal stress changes, unless the soil is very coarse and able to dissipate excess pore pressure before arrest (Picarelli et al. 2007). It has been maintained (Iverson 1997; Musso and Olivares 2003) that such acceleration can result in transition from a purely frictional to a collisional regime, causing complete soil fluidization. Brief Considerations about the Mechanics of Flowslides in Campania Experience regarding flowslides in the region of Campania shows the following. (i) They develop along quite steep slopes: typical angles in areas subject to flowslides range between 35° and 45° (de Riso et al. 1999). (ii) The material involved consists of regularly layered air-fall pyroclastic soils: the layers are constituted by nonplastic silty sand (volcanic ash) and gravelly sand (pumice); the bedrock (typically, limestone, tuff, or flysch) is generally located at a small depth (a few metres). (iii) Despite the presence of silt, which can attain locally a percentage of 50 %, pyroclastic soils are generally completely cohesionless, with a friction angle ranging from 35° to 39°. (iv) Slope stability is assured by partial saturation of soil: suction varies with grain size, depth from the ground surface, and seasonal conditions. Peak values attain 80– 90 kPa, which corresponds to an apparent cohesion up to 10– 15 kPa (Olivares and Picarelli 2006). (v) Air-fall volcanic ash is generally susceptible to liquefaction. Slope failure is primarily caused by suction decrease during rainfall. This mechanical process is associated with increasing soil deformation which announces impending slope failure. In the hypothesis of an infinite slope, the safety factor can be evaluated by the following expression: FS = ␶lim/␶ = 兵关c⬘ + 共ua − uw兲␹ tan ␾⬘兴 + 共␴␣ − ua兲tan ␾⬘其/共␥z sin ␣ cos ␣兲 (1) where: ␶lim is the shear strength of soil along a plane parallel to the ground surface evaluated by means of the extension of the MohrCoulomb criterion (Fredlund and Rahardjo 1993) for unsaturated soil,4 c⬘ is the effective cohesion, usually negligible in such soils, ␾⬘ is the friction angle, 共ua − uw兲␹ tan ␾⬘ is the apparent cohesion due to matric suction, ␹ is a factor 共艋1兲 related to water content, 共␴␣ − ua兲 is the normal net stress along a plane parallel to the ground surface, ␥ is the unit weight, z is the depth from the ground surface, and 4 Following the extension of the Mohr-Coulomb criterion, the strength envelope is parallel to the envelope for fully saturated conditions, and the larger resistance associated to partial saturation is due to a larger intercept of cohesion related to suction through the term 共ua − uw兲␹ tan ␾⬘. TABLE 1—Main features of the equipment. Channel Upper part Lower part Length 2.10 m 1.20 m Width 0.80 m 0.80 m Height 2.50 m 1.40 m Max slope 0°/65° 0°/30° Soil layer Max length 1.90 m Max width 0.50 m Max height 0.50 m Max slope 65° ␣ is the slope angle. The evolution of the safety factor during infiltration is essentially due to reduction in the intercept of cohesion. At the beginning of the infiltration process a wetting front advances nearly vertically in the subsoil and hydraulic conductivity continuously changes as a function of the degree of saturation. Subsequently, the hydraulic conductivity attains its maximum value and the flow component parallel to the slope increases, also affecting the local value of pore pressure according to flow direction. This implies that the safety factor depends on elevation, since at any elevation, a different ␶lim / ␶ can be calculated. However, for tan ␣ ⬍ 共␥⬘ / ␥sat兲tan ␾⬘ slope rupture cannot occur. In contrast, for tan ␣ ⬎ tan ␾⬘ rupture can occur before complete saturation, say, when some apparent cohesion due to suction is still present, while for tan ␣ 艋 tan ␾⬘, failure can occur only for saturated soil. As discussed above, failure in saturated soil (or near to saturation) is a condition under which liquefaction can take place and a flowslide can fully develop (Olivares and Picarelli 2003). The Instrumented Flume and Testing Procedures The flume apparatus was designed to reproduce and investigate, in a controlled environment, rainfall-induced slope failure in unsaturated soils. The apparatus is heavily instrumented in order to capture all the fundamental mechanical aspects of slope behavior, and to assess the role played by single factors, such as geometry, initial and boundary conditions, and soil properties, on the mechanics of rupture and on the landslide pattern (slide or flowslide) and magnitude. Another important aspect of ongoing research is to test new instruments to be used in situ for early warning of landslide occurrence. The flume built at the Geotechnical Laboratory has the characteristics reported in Table 1 and in Fig. 1. A small-scale slope can be reproduced, by reconstitution of soil inside the flume. Different slope morphologies can be adopted: so far, only the case of homogeneous long deposit has been investigated (the thickness is less than 1 / 10 the length, allowing the infiltration process to be analyzed under the assumption of an infinite slope). The flume consists of two main parts (each supported by two metallic frames hinged together to permit constrained relative rotation between the two parts of the flume) in order to simulate a change in slope at the toe of the deposit (Fig. 1(a); the deposit is reconstituted only in the first part of the flume) and a deposit with two different slope angles (the deposit is reconstituted in both parts of the flume). The sides of the flume are made of transparent plexiglass sheets. On the bottom, calcareous grains glued on a rubber sheet reproduce an impervious frictional contact (Fig. 1(c); the rubber sheet is sealed with silicone sealant). A pervious boundary can also be obtained by substituting the rubber sheet with a pervious sheet (geotextile). OLIVARES ET AL. ON FLUME TO INVESTIGATE LANDSLIDES 3 FIG. 1—The instrumented flume (LT= laser transducer; MT= minitensiometer; PT= pressure transducer; LC= load cell; TDR= TDR probe; VC= digital video camera; SF= sensing fiber). (a) cross section. (b) plan-view. (c) details of installation of pressure transducers (on the left) and load cells (on the right). An artificial rainfall system was designed to simulate uniform, continuous rainfall with an intensity between 10 mm/ h and 80 mm/ h. The rain is generated by spray nozzles installed above the lateral sides of the flume. The nozzles produce nebulized water particles 0.1 mm in diameter, in order to prevent erosion of the soil surface. The equipment allows slope behavior to be investigated for different slope inclinations, which plays a major role in the postfailure movement pattern. The influence of grain size, layering, initial soil density, and water content can be easily investigated using different materials and soil preparation procedures. Monitoring System The research focuses on the mechanics of slope behavior from prefailure to post-failure. Concerning pre-failure, different instruments are used to measure any change in soil properties (water content, degree of saturation, porosity), state of stress (suction, pore pressure and total stress), and displacement pattern (movements ux, uy, and uz along x, y, and z axes as in Fig. 1). The soil properties are measured directly, through soil sampling, or indirectly, through a time domain reflectometry (TDR) probe (water content). The state of stress is measured by micro-transducers, i.e., tensiometers, pore pressure probes, and total stress cells. The displacement field is ob- Manufacture/ Type Soil Moisture Corp/2100F Measurement Device Matric suction Tensiometer Mesurement Device Manufacture/ Type Sensor Water content profile along z TDR apparatus Tektronix/ 1502C Cable Tester Home-made metallic waveguide Mesurement Displacements uz Micro Epsilon/ ILD 1400-100 Measurement Manufacture/ Type Sensor Micro Epsilon/ displacement ILD 1400-50 Micro Epsilon/ Sensor displacement ILD 1400-1000 Device Device Sensor Porous ceramic cup Sensor Laser CCDarray Laser CCDarray 25 mm Transducer Current transducer Operating range Linearity Hysteresis Output Sampling frequency 0 – 100 Pa 0.25 % ⬍1 % 4 – 20 mA 750 Hz Nominal pulse rise time Observed pulse rise time Maximum spatial resolution Maximum temporal resolution Volumetric water content 90 ps 200 ps 2 cm at 1 cm at ␪ = 0.1 m3 / m3 ␪ = 0.2 m3 / m3 15 s 0.01 m3 / m3 Housing size (L by W by H) Mesuring range Resolution Linearity Output Operating temperature Sampling frequency 65 mm by 20 mm by 50 mm 50 mm 5 µm ±0.2 % 4 – 20 mA 0 – 55 ° C 1 kHz 65 mm by 20 mm by 50 mm 100 mm 20 µm ±0.2 % 4 – 20 mA 0 – 55 ° C 1 kHz Diameter Length 1 mm up to 95 mm Manufacture/ Type Sensor Housing size (L by W by H) Frame rate Resolution Pixel size Video output Image area Sampling frequency 45 mm by 62 mm by 62 mm 11.75 fps 1300 by 1030 6.7 µm by 6.7 µm RS-644 50 cm by 50 cm 0.1 Hz Dimension Diameter Height Operating range Resolution Linearity Output Operating temperature Natural frequency 0 – 35 kPa 0.1 % f.s. ±0.2 % 0 – 35 mV −20– 120 ° C 55 kHz Dimension Diameter Height Operating range Resolution Linearity Output Maximum overloading 4.5 mm 0 – 200 kPa 0.2 % ±0.1 % 0 – 35 mV 150 % Resolution Maximum temporal resolution 1 min Digital camera Basler/A101P 2/3 ⬙ interline transfer progressive Mesurement Device Manufacture/ Type Sensor Pore water pressure Pore pressure transducer Druck/ PDCR81 Integrated silicon strain gauge Mesurement Device Total stress Pressure gauge Mesurement Device Rainfall intensity Rain gauge Temperature 6 mm Dimension sensor Displacements ux; uy Relative humidity Sensor dimensions Diameter Height Manufacture Type TML/ miniature PDAPA Manufacture/ Type Oregon Scientific/ PCR918N Oregon Scientific/ BTHR918N Sensor 6.3 mm 11.4 mm 1 mm Dimension sensor Sensor Tlpping bucket Diameter Height Operating range 113.5 mm 145 mm 0 – 999 mm/ h 1 mm/ h −5 ° – 50 ° C 0.1 ° C 1 min 2–98 % 1% 1 min 4 GEOTECHNICAL TESTING JOURNAL TABLE 2—Main characteristics of the devices. OLIVARES ET AL. ON FLUME TO INVESTIGATE LANDSLIDES Cable Tester suction [kPa] 6
(ua-uw)=1.6kPa 5 cm Coaxial cable Time-lag=2.2min 4 9.6
(ua-uw)=2.4kPa Time-lag=2.7min 9.6 cm Time-lag=1.7min 7 1.3 cm Metallic rods 8 1 mm (diameter) 9 5
(ua-uw)=5.2kPa FIG. 3—Diagram and photograph of the TDR experimental apparatus. 3 2 0 5 10 15 20 25 30 35 40 45 50 time [min] FIG. 2—Typical response of submerged small-tip tensiometers. tained through laser transducers and video cameras. Optical sensing fibers are currently being adopted to assess their reliability for rapid field alerting of landslide occurrence. As regards post-failure, monitoring was mostly designed to measure excess pore pressures induced by rupture and early postrupture soil deformation (static liquefaction and possible fluidization 5). To this aim, miniature pore pressure transducers and highdefinition mobile video cameras have a prominent role. The characteristics of each device, reported in Table 2, are described below. Since TDR and optical fibers are not usually adopted in Geotechnical Engineering, some additional information is provided. Tensiometers—Suction is measured by conventional jet-fill type tensiometers (model 2100F—Soilmoisture Equipment Corp.) (Table 2). These consist of a high air entry 共100 kPa兲 porous ceramic cup connected to a pressure-measuring device through a small-bore tube filled with de-aired water. The cups can be freely located everywhere in the soil by manual installation at the selected depth. The instruments are connected to a PC for data acquisition and storage. The response of the adopted tensiometers was tested, proving fast enough to follow the expected changes in suction during infiltration in the flume. Figure 2 illustrates the response of three tensiometers during immersion in water under suction increases between 1 kPa and 5 kPa. The time-lag (about 1 min per 1 – 2 kPa of suction change) is a function of the applied increment of suction. During the test, as discussed below, suction progressively changes, and a time lag of 1.5 min may be considered an acceptable time lag for the infiltration process. TDR Device—Time domain reflectometry (TDR) has been widely used in recent decades to measure soil water content both in the laboratory and in the field. The experimental setup consists of a Tektronix 1502C cable tester and a home-made three-rod metallic waveguide (Fig. 3). The probe, vertically installed in the soil and connected to the cable tester, allows measurement of the volumetric water content profile within the entire soil layer. The main characteristics of the experimental apparatus are given in Table 2. TDR measurement is based on the strong correlation between bulk dielectric permittivity ␧r of wet soil and its volumetric water 5 The process of static liquefaction corresponds to a progressive decrease in shear strength to zero while pore pressure monotonically increases. When the strain rate assumes high values, the interaction between grains and water causes a further pore pressure increase up to a value where the submerged weight of the soil column is entirely balanced by excess pore pressure. In this case, the state of complete fluidization is reached (Musso and Olivares, 2003). content ␪ (Campbell 1990). Usually, soil permittivity is estimated by measuring the mean speed of an electromagnetic pulse travelling along a metallic probe buried in the soil, providing the mean water content in a cylindrical volume around the probe, with diameter comparable to probe external rod spacing. Several expressions of the ␧r共␪兲 relationship are available in the literature, empirically stated (Topp et al. 1980) as well as based on a semi-analytical approach to dielectric mixing models (Roth et al. 1990; Whalley 1993; Gong et al. 2003), applying to many soils without the need of calibration, with absolute errors in estimated water content smaller than 0.03 m3 / m3. The accuracy of water content assessment may be improved up to 0.01 m3 / m3 by experimentally determining specific calibration relationships through gravimetric tests coupled with TDR measurements carried out over small soil samples. Recently, several studies have been devoted to developing innovative methods to extract more information from TDR measurements, in order to estimate non-homogeneous moisture profiles along the probe axis (Oswald et al. 2003; Greco 2006; Moret et al. 2006). In this study, we use the inverse profiling method proposed by Greco (2006), recently applied to the retrieval of water content profiles in reconstituted sandy soil samples in the laboratory. Rainfall and Temperature Gauges—Rainfall intensity and air temperature are measured, respectively, by an Oregon Scientific PCR918N rain gauge and a standard thermocouple, controlled with a WMR928NX weather station. Another thermocouple is installed within the soil layer to measure possible temperature changes during the test. Cumulative rain height is measured every minute. Rainfall intensity is calculated by differentiating the hyetogram. Laser Sensors and Particle Image Velocimetry System—Soil surface displacements are monitored by means of laser sensors and a particle image velocimetry (PIV) system. The laser sensors (models ILD1400 and ILD1500—Micro Epsilon) allow the settlement of the ground surface to be monitored (displacement component along axis z in Fig. 1). The device exploits the principle of optical triangulation: a modulated light is projected onto the target (soil surface), and the distance is evaluated by the intensity of the reflected diffused light (measured by CCD-array sensors) captured by a receiving lens arranged at a certain angle with respect to the optical axis of the laser beam. The instruments are located above the ground surface with the optical axis 共z兲 perpendicular to the monitored surface. Orthogonal deviations between 5° and 15° correspond to errors in the evaluation of the distance of about 0.5 % of the measurement field. The roughness of the target surface and the presence of water drops produce a noise that in the current applications is nevertheless acceptable: the records plotted in Fig. 4 show that the noise is in the order 6 GEOTECHNICAL TESTING JOURNAL 1.6 0 noise pore pressure [kPa] 0.0 2 0.2 3 4 5 6 0.4 0.6 0.8 time [min] 1.0 7 15 16 17 5 10 15 20 time [min] 25 FIG. 4—Typical measurement of settlement of the ground surface by laser transducers. of one tenth of a millimetre. The PIV system, consisting of three high definition digital videocameras (model A101P—Basler) connected to digital image acquisition boards (National Instruments hardware components), was designed to monitor and store the images with 0.1 Hz frequency sampling. Two cameras are located over the ground surface to monitor the components of displacements along axes x and y. The other is placed on the lateral side of the flume to monitor the components along axes x and z (Fig. 1). The cameras placed above the flume are motorized and can move with pre-defined acceleration or velocity to follow the movement of soil particles. Each camera is placed at such a distance from the soil surface as to capture a 50 cm⫻ 50 cm image, thus exploiting the maximum definition 共3 megapixels兲. The image contains a number of particles (about 0.1 mm in diameter) sufficient to correctly define the displacement field through a PIV technique. These characteristics allow proper investigation of both the pre-failure and post-failure stage up to a displacement rate of 1 m / s. The system is designed to activate the storage of images for set values of movement (threshold of activation). The image shown in Fig. 5 highlights the quality of the digital tension crack 10 20 30 1.24 4 1 31.1 2 -2
u 31.1 10 -30 0.8 0.6 0.4 0.2 18 8 0 0 rottura 1.4 1 uz [mm] displacement uz at ground surface [mm] -1 40 50mm FIG. 5—Image captured by the PIV system during rainfall. 0 1850 1855 1860 1865 1870 1875 time [s] FIG. 6—An example of pore pressure measurement. video system, which is able to recognize single soil particles and the development of cracks, and demonstrates the negligible influence of background noise due to artificial rainfall (interference of water drops). Pore Pressure Transducers—Positive pore pressure is measured by miniature silicon diaphragm pressure transducers (model PDCR81—Druck) with an integral semiconductor straingauge bridge (Hight 1982) installed at the base of the flume. The devices have a very low volume capacity (the volume of water between the high air-entry porous stone and the diaphragm is 1 mm3) with a high resolution 共0.03 kPa兲 and short time lag 共0.05 s兲. The full scale operating pressure 共35 kPa兲 was chosen with reference to the maximum expected pore pressure during the tests. An example of monitoring is shown in Fig. 6. Optical Fibers—With distributed optical fiber sensing, which has been successfully employed in monitoring steel beam deformations (Niklès et al. 1997), strain and temperature can be captured along the entire length of a fiber embedded in the soil through stimulated Brillouin scattering (SBS) measurements. (Thevenaz et al. 1998; Aulack et al. 2004; Fujihashi et al. 2003). The experimental setup is illustrated in Fig. 7. Light emitted by a distributed-feedback diode laser is first split into two arms by a fiber-fused coupler. An acousto-optic modulator (AOM) is used to provide pulses with widths down to 20 ns (corresponding to 2 m spatial resolution), whereas the cw (continuous wave) probe signal is generated by the electro-optic modulator using the sideband technique (Bernini et al. 2004). The detector consists of an InGaAs photodetector, whose output is sent to an A/D converter connected to a notebook via a USB port. Data are periodically transferred to the notebook, which also performs processing. A frequency shift of 300 MHz is induced by the AOM on the pump optical frequency, such that only one of the two sidebands can effectively interact with the pump wave for Brillouin scattering generation. This provides inherent stability to the system, as it is totally immune to any drift of the source wavelength. The system is operated with a 2 m spatial resolution, while 2000 averages are taken for each fixed pump-probe frequency shift in order to increase the signal to noise (S/N) ratio of the measurements. The acquired waveforms are processed by a standard Lorentzian fitting technique. Given the time required to perform the whole process 共acquisition+ processing兲, the system is capable of providing a new updated Brillouin frequency shift distribution OLIVARES ET AL. ON FLUME TO INVESTIGATE LANDSLIDES 7 FIG. 7—Experimental setup for SBS-based fiber-optic measurements (C = optical circulator; PC= polarization controller). along the sensing fiber about every 15 min. The sensitivity of the sensor to the longitudinal strain experienced by the fiber is ±20µ␧, while the sensitivity to temperature is 1 ° C. The fiber employed for the experiments is a single-mode standard optical fiber, protected by an outer PVC jacket with a diameter of 900 µm. The total length of the fiber is about 35 m. Two 1-m-long strands of fiber are placed within the soil layer, in the longitudinal direction of the flume. The two strands are separated by a fiber spoil of about 15 m, placed outside the soil and not subjected to strain. Soil Preparation and Test Procedures Different techniques (pluviometric deposition, moist-tamping, etc.) can be used to reproduce natural field conditions. In the case of the air-fall pyroclastic soils found in Campania, the main problem is to obtain a porosity as high as 70–75 %, which characterizes the deposits northeast of Naples (Picarelli et al. 2006). From our experience, the moist-tamping technique is the easiest way to obtain high porosity, provided a very low compaction energy and a water content between 30 % and 50 % are used. In order to obtain a uniform distribution of porosity and water content, the soil is laid on the flume in thin layers with a thickness of around 0.5 cm. During formation of the layer, the optical fiber is placed along the slope at a selected depth (Fig. 8). Small geo-grids (1 cm by 1 cm) are glued to the fiber in order to prevent relative movements between the latter and the surrounding soil. In contrast, the tensiometers and the TDR device are installed only after formation of the layer. To prevent evaporation, after deposition the soil layer is covered by an impervious membrane. Suction and water content are then monitored by tensiometers and TDR. In general, a time of about 48 h is sufficient to attain an equalization of suction. An example is reported in Fig. 9: initially, at midheight of the layer, the suction ranges between 43 kPa and 49 kPa with a maximum variation less than 13 %; this becomes about 3 % after 10 h. Finally, the flume is tilted to the prefixed slope angle and the test can start. of 38° and a nil value of cohesion c⬘, measured through CID and CIU triaxial tests on saturated natural samples (Olivares and Picarelli 2003). The other soil properties, the geometrical features of the model slope, and the adopted rainfall intensity are reported in Table 3. For the water content measurement by TDR technique on this soil the ␪共␧r兲 relationship was experimentally determined on reconstituted samples (Damiano et al. 2008) since the “universal calibration relationship” by Topp et al. (1980) leads to soil water content underestimation for such a loose soil. Soil behavior in the pre-failure stage can be investigated by comparing the measured values of suction with the water content profile and with the displacement of the ground surface. The post-failure stage can be examined through the correlation between the pore pressures measured at the base of the slope and the velocity attained by the mobilized soil mass. The trend of suction with time is shown in Fig. 12(a). Indeed, all the tensiometers record a strong decrease in suction (20 kPa in about 20 min) and the infiltration front gradually moves from the ground surface to the base, as indicated by the delay of the devices located near the bottom of the model slope. Only 10 min after the beginning of the test, the deep tensiometers start to record major changes in suction. Due to progressive soil saturation, the volumetric water content also increases (Fig. 13) and the soil exhibits intense volumetric collapse due to high porosity, as revealed by soil surface settlement increase (Fig. 12(b)). Figure 13 shows some of the volumetric water content profiles measured at various times during the experiment. The shape of profiles is consistent with the expected infiltration process: in the very first stage, the water content shows a difference between the values recorded in the most superficial zone and those recorded in the deepest one, which confirms wetting of the uppermost soil layer, while it finally tends to a nearly constant vertical profile because of full saturation corresponding to a water content close to porosity. The last profile shown in Fig. 13 corresponds to a few seconds before slope failure when soil saturation is reached; the lower water content recorded near the base of the slope can be related to a decrease in porosity at the base of the layer due to volumetric collapse. In the same phase (around 30 min from the beginning of the test), suction vanishes practically everywhere. As revealed by the course of displacements, in this late Results Figure 10 shows the experimental setup of a calibration test (Test A1) performed on the Cervinara ash subjected to uniform rain. The slope angle is 40°, very close to the friction angle of soil and an impervious boundary condition at the bottom of the slope is reproduced by means of a sealed rubber sheet (see Fig. 1(c)). The soil is a nonplastic silty sand (Fig. 11) characterized by a friction angle ␾⬘ TABLE 3—Geometrical features of the model slope and soil initial conditions—Test A1. Slope Water Degree of Matric Rainfall angle Length Height content Porosity saturation suction intensity 40° 1.00 m 0.10 m 50 % 75 % 43 % 9 – 20 kPa 40 mm/ h 8 GEOTECHNICAL TESTING JOURNAL 60 initial water content
40% 55 15cm 20cm 15cm suction [kPa] 50 45 40 35 20cm Sensing fibre Sensing fibre 30 0 2 4 6 8 time [h] 10 FIG. 9—Suction equalization after formation of the layer as revealed by tensiometers. FIG. 8—Installation of the sensing fiber. stage slope instability (inability of soil to sustain the stresses associated with gravity) is imminent. Failure is announced by pore pressures (Fig. 14), which become positive: this means that there is ponding at the impervious bed of the flume. A few seconds after failure, a sudden increase in pore pressure occurs: as a consequence, the shear strength at the base of the slope experiences a drop, which results in an acceleration of the mobilized soil mass and the attainment of a very high velocity. As the transducers are left partially or fully uncovered by the soil, a reduction in pore pressure is recorded. The whole failure process can be re-examined through the images obtained with the digital video system. The first stage of infil- tration causes a practically uniform displacement field throughout the slope (section A-A⬘ in Fig. 15(a)), given by the component of soil deformation in the direction of the slope (Fig. 15(b)). In the early post-failure stage (Fig. 15(c)), the rate of displacement measured every 0.18 s along the same cross section reveals such high acceleration as to validate the idea that an undrained deformation mechanism is established as a consequence of failure. The rate of movement calculated to account for the displacement of a soil particle (point A on the surface of the sliding mass in Fig. 15(a)) attains a maximum greater than 0.6 m / s in only 1.2 s (Fig. 15(d)). In this phase, the movement takes on a flow style, while the material displays a high degree of fluidity. Slope instability is also reported by the SBS-based setup. In the diagram shown in Fig. 16, the two dashed lines indicate the position of the two strands of fiber embedded into the soil. The first measurement taken before the experiment serves as a reference, such that any subsequent measurement indicates ongoing deformation. Indeed, the second profile recorded 18 min after the beginning of the test shows that significant changes in fiber response are localized in the fiber spoil between the two embedded regions. Since no FIG. 10—Instrumentation adopted in the calibration test A1. OLIVARES ET AL. ON FLUME TO INVESTIGATE LANDSLIDES Clay Silt Sand Gravel 0 Block 100 t = 0min 1 90 t = 5min t = 10min 2 80 t = 16min 70 3 60 4 z [cm] percent finer by weight [%] 9 50 40 t = 20min t = 27min t = 31min 5 30 6 20 7 t = 35min 10 8 0 0.0001 0.001 0.01 0.1 1 10 100 9 particle size [mm] 10 0.1 FIG. 11—Particle size distribution of soil used in test A1. suction (ua-uw) [kPa] 0 failure superficial tensiometers (depth z=5cm) T2 10 T7 15 deep tensiometers (depth z=10cm) T1 20 0.4 0.5 0.6 0.7 0.8 FIG. 13—Test A1: vertical water content profiles estimated by TDR during the flume test. Conclusions An instrumented flume, designed and built at the Geotechnical Laboratory of CIRIAM (Interdepartmental Research Centre of Environmental Engineering at the Second University of Naples), has been successfully used to investigate the mechanics of flowslides in cohesionless pyroclastic soils (Damiano 2004; Olivares and Picarelli 2006; Olivares and Damiano 2007). The flume is currently being used to evaluate the performance of a TDR device and optical fibers to recognize, respectively, the evolution of the water content profile and any significant changes in the displacement field in unsaturated soil slopes subjected to continuous rainfall. The additional information provided by such low-cost instruments can indeed be of great help for the rapid assessment of flowslide development as well as to further clarify the physical process leading to slope failure. Preliminary experimental results look promising in this respect. Significant improvements in the fiber-based sensor can be achieved by making slight modifications to the setup shown in Fig. 7. For example, the use of an electro-optic modulator and an optical filter, instead of the acousto-optic modulator, may improve the sensor spatial resolution down to tens of centimetres (Niklès et al. 1997). Moreover, the use of erbium-doped fiber amplifiers would allow the intensity of the interacting optical beams to be increased, and hence improve the S/N ratio of the measurements. Note that T3 0.7 0 L3 uz uw [kPa] 5 1.5 x 0.6 settlement 0.5 pore pressure
u 0.4 0.3 z P3 L3 1 failure 10 uw [kPa] settlement uz at the ground surface [mm] 25 0.2 35.90 tf 35.95 36.00 10 5 0 -5 -10 -15 -20 -25 -30 -35 -40 36.05 settlement uz at the ground surface [mm] 2 (a) tf = 35.96min failure time [min] 0.5 15 0 (b) 0.3
m3/m3 strain is applied in this section, this phenomenon can be attributed to the change in temperature along this part of the fiber which is directly exposed to rainfall (it is evident from Fig. 16, where the Brillouin frequency decreases according to a temperature change from about 20° C to 10° C). The profile recorded 15 min later, reveals a net Brillouin frequency increase around the position z = 28 m of the sensing fiber, corresponding to one of the two embedded fiber strands. A smaller Brillouin frequency shift increase can also be recognized corresponding to the first immersed fiber strand, but it is barely visible due to noise. During the experiment the temperature of the soil was continuously monitored by means of a thermocouple and it remained almost constant. The increase in Brillouin frequency can be interpreted as a consequence of tensile strain caused by soil displacement. Importantly, as failure occurred 5 min after the appearance of the strain peak, the sensor seems to be able to reveal impending failure. The final measurement taken after failure shows that the Brillouin frequency in the embedded regions is returning to the initial value. 5 0.2 5 10 15 20 25 30 35 time [min] P3 P4 P5 5 10 15 0 0 20 25 30 35 time [min] FIG. 12—Test A1: (a) evolution of suction, uw − ua; (b) evolution of settlements, uz, during infiltration. FIG. 14—Test A1: pore pressures measured at the base of the slope. downslope 10 GEOTECHNICAL TESTING JOURNAL 1000 1000 A’ a) b) c)
t = 30s
t = 0.18s 900 0.7 d) 800 0.6 600 500 400 0.1 t = 31' 30'' t = 32' 30'' t = 33' 30'' t = 34' 30'' 100 A tf = 35' 57.60'' t = 32' 00'' t = 33' 00'' t = 34' 00'' t = 35' 00'' 0.0 tf t = 35' 57.60'' t = 35' 57.96'' t = 35' 57.78'' t = 35' 58.14'' +0.5 +1.0 t [s] +1.5 rate [mm/s] 40 30 20 10 0 .1 0 0 .0 8 0 0 .0 2 0 plan-view 0 .0 6 upslope 0.3 tf = 35' 57.60'' 300 200 0 0.4 0.2 0 .0 4 A 0.5 rate [m/s] position along x-axis [mm] movement 700 rate [mm/s] FIG. 15—Test A1: displacement of the ground surface in the direction of the slope: (a) plan-view; (b) displacement rate at different times along section AA⬘ during the pre-failure stage; (c) displacement rate at different times along section AA⬘ in the post-failure stage; (d) rate of displacement measured at point A against time. 10.55 Brillouin frequency shift [GHz] 10.545 t10:56 = 0 a.m. p.m. t2:36 18min t == 18min p.m. t2:51 33min t = 33min t3:01 t > tff p.m. First strand in the soil l=1m Second strand in the soil l=1m 10.54 10.535 10.53
T=10°C 10.525 10.52 10.515 5 10 15 20 25 30 Position along the fiber [m] FIG. 16—Test A1: records taken by the SBS-based setup. such an S/N ratio increase would also reduce the number of waveforms to be acquired and averaged for each measurement, with the added advantage of significantly reducing the acquisition time. Finally, processing time could also be reduced by using dedicated electronics for data processing. The combination of acquisition and processing speed-up could allow a new updated Brillouin frequency shift profile to be available in a much shorter (tens of seconds) time. References Aulackh, N. S., Chhabra, J. K., Singh, N., and Jain, S., 2004, “Microbend Resolution Enhancing Technique for Fiber Optic Based Sensing and Monitoring of Landslides,” Experimental techniques, Vol. 28, pp. 37–42. Bernini, R., Minardo, A. and Zeni, L., 2004, “Accuracy Enhancement in Brillouin Distributed Fiber-Optics Temperature Sensors Using Signal Processing Techniques,” IEEE Photonics Technol. Lett., Vol. 16, No. 4, pp. 1143–1145. Cairo, R., and Dente, G., 2003, “A Flow Slide in a Pyroclastic Soil Fill,” International Conference on Fast Slope Movements— Prediction and Prevention for Risk Mitigation, Naples, Italy, pp. 399–404. Campbell, J. E., 1990, “Dielectric Properties and Influence of Conductivity in Soils at One to Fifty Megahertz,” Soil Sci. Soc. Am. J., Vol. 54, pp. 332–341. Damiano, E., 2004, “Meccanismi d’innesco di Colate di Fango in Terreni Piroclastici,” Ph.D. thesis, Second University of Naples, Italy. Damiano, E., Greco, R., Guida, A., and Olivares, L., 2008, “Early Warning of Fast Landslides Triggering Based on Instrumented Slope Data Analysis,” IV Biennial Meeting of iEMSs, Barcelona, Spain. de Riso, R., Buretta, P., Calcaterra, D., and Santo, A., 1999, “Le Colate Rapide in Terreni Piroclastici del Territorio Campano,” Conference on ‘Previsione e Prevenzione dei Fenomeni Franosi’, Ass. Georisorse e Ambiente, Trento, Italy. Faella, C., and Nigro, E., 2003, “Dynamic Impact of the Debris Flows on the Constructions During the Hydrogeological Disaster in Campania-1998: Failure Mechanical Models and Evaluation of the Impact Velocity,” International Conference on Fast Slope Movements—Prediction and Prevention for Risk Mitigation, Naples, Italy, pp. 179–186. Fredlund, D. G., and Rahardjo, H., 1993, Soil Mechanics for Unsaturated Soils, Wiley & Sons, Toronto. Fujihashi, K., Uehara, H., Okutsu, M., and Komatsu, K., 2003, “Development of a Road Disaster Monitoring System Using Optical Fiber Sensing Technology,” NTT Tech. Rev., Vol. 1, No. 9, pp. 48–52. Gong, Y. S., Cao, Q. H., and Sun, Z. J., 2003, “The Effects of Soil Bulk Density, Clay Content and Temperature on Soil Water Content Measurement Using Time-Domain Reflectometry,” Hydrolog. Process., Vol. 17, pp. 3601–3614. Greco, R., 2006, “Soil Water Content Inverse Profiling From Single OLIVARES ET AL. ON FLUME TO INVESTIGATE LANDSLIDES TDR Waveforms,” J. Hydrol., Vol. 317, pp. 325–339. Hight, D. W., 1982, “A Simple Piezometer Probe for the Routine Measurement of Pore Pressure in Triaxial Tests on Saturated Soils,” Geotechnique, Vol. 32, No. 4, pp. 396–401. Hunter, G. J., and Fell, R., 2003, “Mechanics of Failure of Soil Slopes Leading to Rapid Failure,” International Conference on Fast Slope Movements—Prediction and Prevention for Risk Mitigation, Naples, Italy, pp. 283–290. Iverson, R. M., 1997, “The Physics of Debris Flows,” Rev. Geophys., Vol. 35, No. 3, pp. 245–296. Moret, D., Arrúe, J. L., López, M. V., and Gracia, R., 2006, “A New TDR Waveform Analysis Approach for Soil Moisture Profiling Using a Single Probe,” J. Hydrol., 321, pp. 163–172. Musso, A., and Olivares, L., 2003, “Flowslides in Pyroclastic Soils: Transition From Static Liquefaction to Fluidization,” International Workshop on Occurrence and Mechanisms of flow-like landslides in natural slopes and earthfills, Sorrento, Italy, pp. 117–128. Niklès, M., Thévenaz, L., and Robert, P., 1997, “Brillouin Gain Spectrum Characterization in Single-Mode Optical Fibers,” J. Lightwave Technol., Vol. 15, No. 10, pp. 1842–1851. Olivares, L., and Damiano, E., 2007, “Post-Failure Mechanics of Landslides: Laboratory Investigation of Flowslides in Pyroclastic Soils,” J. Geotech. Geoenviron. Eng., Vol. 133, No. 1, pp. 51–62. Olivares, L., and Picarelli, L., 2001, “Susceptibility of Loose Pyroclastic Soils to Static Liquefaction: Some Preliminary Data,” Proc. Int. Conf. Landslides—Causes, Impacts and Countermeasures, Davos, M. Kühne, H. H. Einstein, E. Krauter, H. Klapperich, and R. Pöttler, Eds., pp. 75–85. Olivares, L., and Picarelli, L., 2003, “Shallow Flowslides Triggered by Intense Rainfalls on Natural Slopes Covered by Loose Unsaturated Pyroclastic Soils,” Geotechnique, Vol. 53, No. 2, pp. 283–287. Olivares, L., and Picarelli, L., 2006, “Modelling of Flowslides Behaviour for Risk Mitigation,” General Report, Int. Conf. on 11 Physical Modelling in Geotechnics, Hong Kong, Vol. 1, pp. 99– 113. Oswald, B., Benedickter, H. R., Bächtold, W., and Flühler, H., 2003, “Spatially Resolved Water Content Profiles From Inverted Time Domain Reflectometry Signals,” Water Resour. Res., Vol. 39, No. 12, 1357. Picarelli, L., Evangelista, A., Rolandi, G., Paone, A., Nicotera, M. V., Olivares, L., Scotto di Santolo, A., Lampitiello, S., and Rolandi, M., 2006, “Mechanical Properties of Pyroclastic Soils in Campania Region,” 2nd International Workshop on Characterisation and Engineering Properties of Natural Soils, Singapore, Vol. 1, pp. 2331–2384. Picarelli, L., Olivares, L., Comegna, L., and Damiano, E., 2007, “Mechanical Aspects of Flow-Like Movements in Granular and Fine Grained Soils,” Rock Mech. Rock Eng., Vol. 41, No. 1, pp. 179–197. Roth, K., Schulin, R., Fluhler, H., and Attinger, W., 1990, “Calibration of Time Domain Reflectometry for Water Content Measurement Using a Composite Dielectric Approach,” Water Resour. Res., Vol. 26, pp. 2267–2273. Sladen, J. A., D’Hollander, R. D., and Krahn, J., 1985, “The Liquefaction of Sand: A Collapse Surface Approach,” Comput. Graph. Image Process., Vol. 22, pp. 564–578. Thevenaz, L., Nikles, M., Fellay, A., Facchini, M., and Robert, P., 1998, “Applications of Distributed Brillouin Fiber Sensing,” Proc. SPIE, Vol. 3407, pp. 374–381. Topp, G. C., Davis, J. L., and Annan, A. P., 1980, “Electromagnetic Determination of Soil Water Content: Measurement in Coaxial Transmission Lines,” Water Resour. Res., Vol. 16, pp. 574–582. Yamamuro, J. A., and Lade, P. V., 1998, “Steady-State Concepts and Static Liquefaction of Silty Sands,” J. Geotech. Geoenviron. Eng., Vol. 124, No. 9, pp. 868–877. Whalley, W. R., 1993, “Considerations on the Use of Time Domain Reflectometry (TDR) for Measuring Soil Water Content,” Soil Sci., Vol. 44, pp. 1–9.