Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Numerical modeling and characterization of a peculiar flow-like landslide

Numerical modeling and characterization of a peculiar flow-like landslide Background: On March 25th, 2015, a rapid landslide occurred upstream of the village of Gessi-Mazzalasino, in the municipality of Scandiano, affecting two buildings. Rapid landslides, due to their high velocity and mobility, can affect large areas and cause extensive damage. Considering the often unpredictable kinematics of landslides, the post-failure behavior has been studied by many authors to predict the landslide runout phase for hazard assessment. Findings: With the aim of characterizing the Gessi-Mazzalasino landslide, field surveys were integrated with the results of laboratory tests. The geometric characteristics (thickness, area and volume) and kinematic aspects of the landslide were estimated by using a laser scanning survey and geomorphological data. To model the landslide and obtain its rheological parameters, a back analysis of the event was performed by means of a depth-averaged 3D numerical code called DAN3D. The results of the back analysis of the landslide propagation were validated with field surveys and velocity estimations along selected sections of the landslide. Finally, potential areas prone to failure or reactivation were identified, and a new simulation was performed that considered the back-calculated rheological parameters. Conclusions: Rapid landslides are one of the most dangerous natural hazards and are one of the most frequent natural disasters in the world. Therefore, prediction of post-failure motion is an essential component of hazard assessment when a potential source of a mobile landslide it is located. To assess the risk affecting the area, both numerical and empirical methods have been proposed, in order to predict the runout phase of the phenomenon. For the numerical modelling of the landslide, carried out with DAN-3D code, the best results were obtained by using a Voellmy reological model, with a constant turbulence parameter (ξ) of 250 m/s and a friction parameter (μ) comprised between 0.15 and 0.19. The rheological parameters obtained through dynamic back analyses were used to evaluate the propagation phase and the deposition areas of new potential landslides, that could affect the same area of the 25th March 2015 event. The predicted runout length obtained by the DAN3D software was compared to runout lengths predicted by the Corominas (Can Geotech J 33:260–271, 1996), (Nat. Hazards 19, 47-77) and (UNICIV Report, R-416, School of Civil & Environmental Engineering, UNSW, Sydney Australia 2003) empirical relations. All the data confirm that the impact area of possible future events will be smaller than the 2015 event, probably due to the safety measures established after the landslide. Keywords: Landslide, Geotechnical characterization, Runout simulation, Scandiano * Correspondence: mattia.ceccatelli@unifi.it Department of Earth Sciences, University of Florence, Via G. La Pira, 4, 50121 Florence, Italy © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 2 of 15 Introduction At approximately 07:00 PM on March 25th, 2015 a Rapid landslides such as debris flows, debris avalanches, rapid landslide was triggered upstream of the village of rock avalanches and flow slides are instability phenom- Gessi-Mazzalasino, in the municipality of Scandiano ena that affect superficial deposits as a consequence of (Emilia Romagna region), in north-central Italy (44°34′ intense and prolonged rainfall events. Rapid landslides 45″N,10°39′18 E, Fig. 1) due to days of heavy and are one of the most dangerous and frequent natural haz- persistent rainfall. ards in the world and can cause significant damage to The landslide was triggered at approximately 220 m goods and people in their path. a.s.l. and reached the village, causing slight damage to Guzzetti (2000) showed that more than 80% of the two buildings that were evacuated. deaths and injuries due to landslides in Italy were related One of the main challenges regarding the analysis of to fast-moving failures, including debris flows, rockfalls, rapid landslides is that they are affected by different rockslides, and soil slips. mechanisms during the failure and post-failure stages Fig. 1 a Topographic map and b) geological map of the study area (from Servizio Geologico Sismico e dei Suoli, Emilia Romagna region, 2011) Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 3 of 15 (Bandara et al., 2016). Many studies can be found in the Landslide description literature that are focused on the analysis of landslides The landslide source area is located on the slope using experimental and mathematical methods. upstream of the village of Gessi-Mazzalasino (Fig. 2) at Empirical formulas derived from the statistical data of an elevation between 230 and 215 m a.s.l. The triggering past landslides can provide valuable information (Hsü, event was likely heavy rainfall that occurred a few days 1975; Corominas, 1996), but these formulas are generally before the event. The Cà de Caroli weather station, lo- approximations, and the obtained information is usually cated 1 km northeast of the study area, recorded more limited to specific contexts. To better understand the ef- than 100 mm of rainfall over a 10-day period (Fig. 3), fects of landslides, numerical modeling is a particularly with a peak rainfall intensity of approximately 60 mm on useful tool that is capable of capturing the entire land- March 25th (Fig. 4); for comparison, the annual average slide process in both space and time. In this paper, we rainfall at this station is approximately 750 mm. conduct a combined analysis, via both empirical and nu- The source area is approximately 1700 m and has an merical approaches, to characterize the 2015 irregular shape. The average slope of the source area is phenomenon and to obtain additional information for a approximately 17–20°, but the slope increases up to 30– risk assessment of the area. 35° in the triggering area. A back analysis of the post-failure behavior was con- The initial failure triggered at approximately 220 m ducted with the DAN3D code (McDougall & Hungr, a.s.l., approximately 20 m below the ridge. This altitude 2004; Hungr & McDougall, 2009), using a trial and error difference may reflect an increase in the pore pressure procedure, to obtain the rheological parameters of the due to the hydraulic head, which may represent an add- March 25th, 2015 phenomenon. itional instability parameter of the slope in addition to DAN3D was developed for the simulation of extremely the heavy rainfall. rapid landslides, even in complex topographies (McDou- The flow-like landslide, after initially spreading in a gall & Hungr 2004, Salvatici et al. 2017). Since this code flat area in the middle sector of the slope, moved is also capable of simulating material motion and its through an existing impluvium and reached the corresponding rheological changes (Hungr, 1995), inhabited area at the foot of the hill, 450 m below the DAN3D was used to study the 2015 event. The simula- source area. tion results were validated by means of runout lengths The thickness of the deposits ranges from a few and flow velocities derived from empirical runout decimeters up to 2–3 m in the most significant accumu- prediction methods. lations areas. Finally, a forecast analysis was carried out to evaluate Field evidence showed that the mass movement the characteristics of potential landslides that could occur started as a sliding mass of the surface layers at the in the area in the future, using the rheological parameters upper part of the slope and evolved into a rapid mud obtained by the back analysis of the 2015 event and the flow at the top of the impluvium, where the flow was post-event digital elevation model (DEM) of the area. channeled into, probably due to the addition and mixing of surface water during the mass movement. The source volume, approximately 10,000 m , and the planimetric Study area and landslide description area of the landslide have been identified from the field Study area observations and aerial images that were collected after The study area is located in the municipality of Scan- the event. diano, in the Emilia Romagna region. The landslide af- fected the western slope of the Tresinaro Valley, above Geotechnical characterization the village of Gessi-Mazzalasino. Two soil samples were collected from the landslide de- The area is geologically characterized by units of the posits immediately after the event to perform a geotech- External Liguride domain and the Neogene-Quaternary nical characterization of the materials and recreate the succession of the Northern Apennines. The External initial flow conditions. Liguride domain consists of thin calcareous turbiditic The first soil sample was collected in the deposit area formations known as the Palombini Shales and at the beginning of the channelized section, and the Varicolored Shales, while the Neogene-Quaternary suc- second sample was collected at the landslide toe (Fig. 5). cession in this area is represented by the Gessoso- These samples were subjected to the following labora- Solfifera Formation, the alluvial units of the Ravenna tory tests: index property testing, Atterberg limits test- Subsynthem, and the Modena Unit (Amorosi, 1999). ing, grain size analysis, and direct shear testing. Several active and inactive landslide deposits are also In addition, three geotechnical in situ tests were car- located in the area; the 2015 landslide originated from ried out with the aim of collecting further information one of these deposits (Fig. 1). about the soil in its natural condition. Specifically, two Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 4 of 15 Fig. 2 Aerial photograph of the Gessi-Mazzalasino landslide (photo by G. Bertolini) permeability measurements were carried out with a con- The BST results were interpreted using the Fredlund et stant compact head permeameter (Amoozemeter) and al. (1978) shear strength equation for unsaturated soils. one measurement of shear strength was carried out with a borehole shear test (BST). 0 0 0 The in situ tests were carried out both outside of the τ ¼ c þðÞ σ−u tanΦ þðÞ u −u tanΦ ð1Þ a a w landslide area (test number 3) and within the landslide deposit (test number 4). where τ is the shear strength, c’ is the effective cohe- The BST tests were performed on soils in unsaturated sion, σ is the total normal stress, u is the pore air pres- conditions; at an equivalent depth, matric suction values sure due to surface tension, φ′ is the effective friction (u − u ) were measured with tensiometers. angle, u is the pore water pressure, and φ is the angle a w w b Fig. 3 Daily rainfall intensity from 13/03/2015 to 30/03/2015 Fig. 4 Hourly rainfall intensity on 25/03/2015 Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 5 of 15 Fig. 5 Location of the soil samples (in red) and the geotechnical field investigation (in cyan); point 3: permeability measurement; point 4: permeability measurement and BST expressing the rate of the increase in strength related to The measured saturated hydraulic conductivity ranges −6 −7 matric suction. The BST test results show that the in- from 2.19 × 10 m/s to 5.19 × 10 m/s, corresponding ternal friction angle equals 33.8°. to samples 3 and 4. The procedure used for measuring k in the field is Samples 1 and 2 are primarily unsorted silty soils (Fig. 6) called the constant-head well permeameter technique and are classified as a clay of low plasticity (CL) and a silt (Philip, 1985), and it is carried out in a borehole. This (ML), respectively, following the Unified Soil Classification procedure allowed us to measure the amount of water System (USCS, Wagner, 1957). The samples have plasticity flowing through the soil in a given time interval under index (IP) values ranging from 8 to 11. soil-saturated conditions. The saturated permeability of Direct shear tests were performed on reconstituted sam- the soil is evaluated with the Glover solution: ples, using normal stresses between 40 and 80 kPa, deter- mined from the in situ characteristics. The internal friction angle ranges from 29.1 to 30.1°, while the cohesion (c’)is 2 2 −1 Q sin ðÞ h=r − þ 1 þ r=h very low. The results of the laboratory and in situ tests are showninTable 1. k ¼ ð2Þ 2πh In Fig. 7, the matrix compositions of the two soil sam- ples from the 2015 landslide are compared with the where Q is the steady-state rate of water flow from the compositions of earth flows, debris flows and mud flows −1 permeameter into the soil, sinh is the inverse hyper- from several areas of the world (Hungr et al., 2001). bolic sine function, h is the depth of water in the bore- Hungr et al. (2001) distinguished different materials hole, and r is the radius of the borehole. involved in flow-like landslides on the basis of several Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 6 of 15 Fig. 6 Granulometric curves for the two soil samples (black line for sample and red line for sample 2) material geotechnical properties. Both the samples from the sources may have clay contents greater than 10% and 2015 landslide deposit had low plasticity indices (IP =11 for plasticity indices of more than 10 (Jordan, 1994). sample 1 and IP = 8 for sample 2) and the liquidity index Therefore, the distinction between “mud” and “earth” (IL) values were approximately 0.6 and 0.5, respectively. should not be based solely on grain size distribution but As shown in Fig. 7, the matrix compositions of the should instead be derived from the context of each land- 2015 landslide samples fall in the textural field of earth slide class. flows and mud flows, while debris flows typically contain Specifically, earth flows and mud flows may involve less than 30% silt and finer particles. material of similar texture but are significantly different A comparison of colloidal indices does not allow a in other ways; in particular, the velocity of movement clear distinction to be made between these two different during an earth flow differs from that of a mud flow. classes. Earth flows have clay contents ranging from 10% to 70%, averaging approximately 35%, while debris and Velocity analysis mud flows are usually not plastic or are only weakly There are many equations in the literature for estimating plastic. However, some mud flows derived from volcanic the velocity of the frontal part of flow-like landslides (Hungr et al. 1984). These relations provide a useful par- ameter to validate the back analysis results (Salvatici et al., 2017, Nocentini et al., 2015). Table 1 Geotechnical parameters obtained from laboratory and In this work, flow velocity was estimated in the channel- in situ tests ized section of the landslide, along the cross sections shown Sample 1 2 3 4 in Fig. 8, by using two methods: the superelevation of the USCS classification CL ML –– debris surface in the channel belt (Johnson & Rondine, Porosity [%] 40.1 39.7 –– 1984) and the Poiseuille equation (Hungr et al., 1984). Void ratio [−] 0.67 0.7 –– The Johnson & Rondine (1984) relation is based on the difference in the splash heights on the inside and outside Saturation degree [%] 118.3 117.9–– of the bends in the flow path (Nocentini et al., 2015). Total unit weight [kN/m ] 20.2 19.9 –– The superelevation of the debris wave around the Statured unit weight [kN/m ] 19.5 19.3 –– channel bends tends to be higher than that on the op- Liquid limit 34 33 –– posite side due to the centrifugal force (Fig. 9). Plastic limit 23 25 –– Thus, in cross sections 1 and 2, the velocity can be cal- Plasticity index [%] 11 8 –– culated by using the following equation: Liquidity index [%] 0.6 0.5 –– pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi v ¼ gR cosδ tanβ ð3Þ Friction angle [°] 30.1 29.1 33.8 – Cohesion [kPa] 2 5 0 – where β is the angle between the line connecting the −6 −7 Permeability [m/s] –– 2.19 × 10 5.19 × 10 top of the debris waves at both sides of the section and a Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 7 of 15 Fig. 7 Ternary plot of the two soil samples with textural classification (from Hungr et al., 2001) horizontal line, δ is the slope angle of the flow path, R is to determine the position of up to 12,000 points per sec- the radius of curvature and g is the gravity acceleration. ond, with a maximum angular resolution of 0.008° and The radius of curvature of the channel was obtained an accuracy of ±10 mm from a maximum distance of graphical processing using a 1:5000 topographic map. 800 m. The Hungr et al. (1984) relation, which is based on the To completely cover the intervention areas and avoid Poiseuille equation, can be used to evaluate the flow vel- the shadow areas, were captured four scans from differ- ocity in the straight sections (cross sections 3 and 4). ent positions (Fig. 10). This equation relates the velocity to the geometric char- Several laser cylindrical reflectors were placed on the acteristic of the path, the unit weight of the flow mass, hill slopes, and their coordinates were defined by per- and the viscosity of the flow mass: forming a GPS survey. These tie points were later used to align the point clouds. This process is required for correctly georeferencing the point cloud on a chosen ref- γ sinδH v ¼ ð4Þ erence system and for merging two or more scans of the lv same object realized from different points of view. where y is the unit weight of the material and is ob- On April 22nd, 2015, another GPS survey was carried tained by laboratory test, δ is the slope angle of the flow out to reconstruct the exact geometry of the landslide body, path, H is the flow depth, l is a constant based on the define thesource areaand identify trenches that could de- cross-sectional shape of the channel (3 for a broad velop into the edges of potential detachment areas. channel and 8 for a semicircular channel) and ν is the The data obtained from the laser scanning surveys dynamic viscosity of the flow (assumed to be 3, as indi- have been processed to obtain a high-resolution DEM of cated by Hungr et al., 1984). the area (Fig. 11). The results of the estimated velocity and the geometric Some DEM sectors outside of the landslide were not parameters of the path are summarized in Tables 2 and 3. acquired due to the presence of buildings and dense vegetation, particularly near the toe portion; therefore, it Laser scanning survey was necessary to integrate the model with an existing New high-resolution surveying techniques, such as ter- 1:5000 topographic map. restrial laser scanning, quickly obtain detailed 3D terrain During the data processing, the safety works on the models that can be employed in runout analyses (Gigli landslide started. These works, in an initial phase, in- et al., 2014). cluded the construction of an earthfill dam (Fig. 12) to A laser scanning investigation was performed during prevent the excessive expansion of future landslides and two field surveys, on April 1st, 2015, and April 16th, to channel the flow of those potential landslides towards 2015, by means of a long-range 3D terrestrial laser im- the existing channel. One last GPS survey, on September aging sensor (RIEGL LMS-Z420i device), which is able 9th, 2015, was carried out to detect the geometry and Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 8 of 15 Fig. 8 Location of the cross sections for flow velocity estimation. a) Location of the cross sections for flow velocity estimation and b) Sections profiles location of the earthfill dam, which was then imple- DAN3D numerical model mented in the digital terrain model, to capture the DAN3D is a 3D numerical model that uses the continu- modified shapeof theslopeand thepost-landslide ous Lagrangian approach for integrating the equations conditions. of Saint-Venant with depth. The mass conservation equation governs the model: ∂h ∂v ∂v ∂b Runout simulation methods x y þ h þ ¼ ð5Þ Methods to predict landslide runout were grouped ∂t ∂v ∂y ∂t into two categories by (Rickenmann 1999): the first group includes empirical methods that are based on where b is the bed-normal erosion-entrainment depth, statistical analyses of past events (Iverson, 1997; Cor- v and v are local flow velocities, and t is time. x y ominas, 1996; Hunter & Fell, 2003), and the second DAN3D employs a simple semi-empirical approach group includes analytical methods that account for based on the concept of “equivalent fluid”, as defined by conservation of momentum and energy to simulate (Hungr 1995). the propagation of flow using 2D or 3D models In this method, the landslides are considered one (Hungr, 1995; McDougall & Hungr, 2004; Hungr & material governed by simple rheological relations. There- McDougall, 2006). In this work, the runout distance fore, an internal frictional rheology is considered, as well obtained by the DAN3D code was compared with as a basal rheology that depends on one or two parame- that from empirical methods. ters (depending on the chosen rheological model) that Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 9 of 15 Fig. 9 Empirical formula parameters for flow velocity estimation: a) splash heights on the inside and outside of the bends of the flow path (cross section 2); b) radius of curvature; and c) slope angle are established with a calibration procedure by using during the dynamic modeling of a landslide is related to back analysis. the expected event type and depends on the rheological The model requires three input files that describe the characteristics of the landslide material. topography (path file), source area (source file), and The post-failure phase processes that are triggered dur- number of materials used with their rheologies and ing the movement of rapid landslides are extremely com- erosion parameters (erosion file). plex, and the direct measurement of either the parameters With the aim of simulating different types of fast land- of the involved materials or of the characteristics of the slides, DAN3D can implement the following rheological landslide is impossible. Therefore, the best rheological relations: frictional, plastic, turbulent, Bingham and model is determined by performing a back analysis for to Voellmy. The selection of rheological model to use the investigated case by using similar phenomena. Table 2 Velocity values of the flow obtained according to the Table 3 Velocity values of the flow obtained according to the Johnson & Rondine (1984) formula Hungr et al. (1984) formula Section gR δβ v Section γδ Hl ν v 2 3 [m/s ] [m] [°] [°] [m/s] [kN/m ] [°] [m] [−] [kPa] [m/s] 1 9.81 13.7 12 4.17 3.10 3 9.81 12 3.3 8 3 1.89 2 9.81 15.9 12 2.75 2.71 4 9.81 12 3 8 3 1.56 Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 10 of 15 Fig. 10 Laser scanner point cloud from four scan positions Empirical relations 0:16 0:83 L ¼ 1:9V H ð7Þ There are several empirical methods that relate the geometric parameters of landslides to the runout dis- 0:69tana þ 0:086 ð8Þ tance; in this case study, three methods, presented by (Corominas 1996), (Rickenmann 1999) and Hunter & Fell (2003), were used to correlate the angle of reach where A and B are coefficients that depend on the (fahrboschung) with the volume. landslide types and α is the slope inclination. The fahr- boschung was defined by (Heim 1932) as the inclination of the line connecting the crest of the landslide source Log ¼ BlogV þ A ð6Þ with the toe of the deposits and can be evaluated by the Fig. 12 Earthfill dam, built during safety operations in the Fig. 11 DEM of the area landslide area Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 11 of 15 ratio between the elevation difference of the highest and ρgv τ ¼ − f σ þ ð10Þ zx z lowest points of flow (H) and the corresponding hori- zontal distance (L): where f is the frictional coefficient; ξ is a turbulence parameter representing all possible sources of velocity- tanα ¼ ð9Þ dependent resistance in landslide dynamics; and ρ, g and These methods may be applied in preliminary hazard assess- v are the density, gravity and velocity, respectively. ments, and they may be compared with the dynamic analysis. For the determination of the rheological model param- eters, a back analysis was performed based on the study Findings and Results of the deposits from the March 2015 collapse. Two main Back analysis phases of motion were simulated, according to the field The calibration procedure is based on a trial and error evidence, assuming that the mass movement started as a back analysis of the Gessi-Mazzalasino landslide, to sliding mass involving the surface layers of the upper identify the most suitable rheological parameters for de- part of the slope and evolved into a rapid mud flow as it scribing the flow motion. The path file used for the back entered the impluvium, where the flow was channelized. analysis was the DEM of the slope, obtained from a Two basal shear resistances were chosen, according to 1:5000 topographic map of the area. The source file was the landslide dynamics; one rheology material was used defined by the triggering area obtained from the field for the source area between 226 and 173 m a.s.l., and an- survey. Since no significant erosion was observed during other was assumed for the channelized area between the field surveys, the erosion file was neglected. 173 and 131 m a.s.l. The Voellmy resistance parameters The numerous study cases analyzed with the DAN3D were adjusted by trial and error to achieve the best code have shown that the Voellmy rheological model is simulation match in terms of velocity, thickness of particularly suitable to describe this type of phenomena deposits and runout distance. and that, accordingly, it should be used for modeling the The best match between the actual and simulated Gessi-Mazzalasino landslide (Nocentini et al., 2015; Gigli material distributions was obtained using a frictional et al., 2014). coefficient of f = 0.19 for the upper material and a fric- This model, introduced by (Voellmy 1955) for snow tional coefficient of f = 0.15 for the lower material, while avalanches, contains a friction term and a turbulence the turbulence term of ξ = 250.00 m/s remained term: constant throughout the event (Fig. 13). Fig. 13 Comparison between the maximum runout distances obtained from the empirical and numerical methods with the field data (in red) Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 12 of 15 Additionally, DAN3D requires the internal frictional Table 5 Comparison between the DAN3D and empirical velocity values angle and the unit weight of the material as input pa- rameters. On the basis of the performed laboratory tests, Cross section DAN3D Johnson & Rondine Hungr et al. (1984) (1984) we used an average internal frictional angle of 30° and a [m/s] [m/s] [m/s] unit weight of 20 kN/m . The simulation results show that the landslide reaches 1 7.1 3.10 – the flat area located in the middle sector of the slope, 2 6.9 2.71 – where most of the mobilized material was deposited in a 3 5.1 – 1.89 layer up to 2 m thick, and then moves through the im- 4 4.6 – 1.56 pluvium until it reaches the inhabited area. Runout distances decreased. These results agree with the testimonies of As show in Fig. 13, there is good agreement between the local residents, who evaluated the flow velocity in the simulation output and the field data obtained by the final meters of movement at approximately 0.2–1 m/s. GPS survey (landslide path), for both the average thick- Examples of velocity observations from various ness and the planar extension of the deposits, especially sources are shown in Fig. 14, together with the velocity in the lower part of the slope. values calculated in this paper. These velocities represent The results are summarized in. point observations or maximum values at randomly Table 4 and show some differences between the chosen locations and are not necessarily maxima for a methods used to evaluate the runout distance. given event. The (Corominas 1996) and (Rickenmann 1999) rela- A clear distinction can be made between extremely tions, which produced predicted runout distances of rapid processes such as debris flows, mud flows and deb- 311 m and 379 m, respectively, underestimate the max- ris avalanches and slow process such as earthflows imum runout distance of the landslide. However, the (Hungr et al., 2001). Hunter & Fell (2003) relation, with a predicted runout distance of approximately 431 m, agrees with the Landslide characterization modeling results and field data; therefore, this is the As shown in the previous paragraphs, the distinction be- most suitable empirical method to describe the Gessi- tween “mud flows” and “earth flows” cannot be based Mazzalasino landslide. solely on grain size distribution but can instead be de- rived in other ways, in particular, from the velocity of Velocity calculations movement. By comparing the flow velocity values obtained by using From the velocities obtained with three direct methods the empirical equations and the numerical model results, (including resident testimonies and empirical methods), a similar trend along the flow travel distance was ob- the Gessi-Mazzalasino landslide has an estimated vel- served (Salvatici et al., 2017). ocity ranging between 0.2 and 3 m/s, while the velocities The flow velocity was obtained with the models along from the numerical simulation are higher, ranging be- the four cross sections, the DAN3D results were slightly tween 4.5 and 12 m/s. higher than the velocities obtained by the empirical rela- These differences derive from the assumptions made tions (Table 5). during the simulation phase, resulting in an overesti- These differences derive from the assumptions made mation of the flow velocity. during the simulation phase, making it difficult to model Analysis of the velocities allow better classification of both the kinematic and the depositional parameters of the Gessi-Mazzalasino event, which has characteristics the flow, which were the focus of our study. of a landslide with behaviors between those of mud flow According to the model results, the flow reached a and earthflow phenomena. maximum velocity of approximately 8–12 m/s in the The Gessi-Mazzalasino landslide material is an un- upper part and then slowed down as it entered the im- sorted deposit composed of a mixture of sand, gravel pluvium, and it finally stopped when the slope and cobbles as well as varying proportions of silt and clay, and the landslide event is characterized by low Table 4 Comparison between the calculated and measured plasticity and an intermediate velocity. runout distances Corominas Rickenmann Hunter & Fell DAN3D Field data Future conditions (1996) (1999) (2003) To evaluate the characteristics of potential landslides [m] [m] [m] [m] [m] that could occur in the area, another numerical simula- 311 379 430 426 412 tion was performed with DAN3D using the rheological Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 13 of 15 Fig. 14 Range of velocities for various types of flow-like landslides; the range of the estimated Gessi-Mazzalasino landslide velocity is shown in red (square symbols for resident testimonies, triangular symbols for empirical methods and circle symbols for numerical simulation) parameters obtained by the back analysis of the 2015 To assess the risk affecting the area, different methods event. have been proposed to predict the runout phase of the For this analysis, the most recent DEM obtained by phenomenon. laser scanner survey was used, representing the topog- A back analysis based on the path and the deposits of raphy of the slope after the landslide. the 2015 event was performed in order to identify the During the analysis, a single volume of 10,000 m was optimal rheological model and to simulate the behavior investigated based on the field survey and on the conser- of potential landslides. vative assumption that failure occurred for a single The dynamic modeling was carried out by using the volume. DAN3D code, which estimated the extent of the impact Figure 14 shows the maximum runout distances ob- area and mapped the distribution of landslide tained from the numerical methods used in this study. parameters. The results of the potential landslide show that, due to The predicted runout length obtained by the DAN3D the safety works, the landslide impact area was predicted software was compared to runout lengths predicted by to be smaller than that of the 2015 event (Fig. 15), with the (Corominas 1996), (Rickenmann 1999) and Hunter the mass stopping 70 s after the start of the simulation & Fell (2003) empirical relations. at the beginning of the impluvium, where the earthfill There is good agreement between the simulation out- dam was constructed. A maximum thickness of approxi- put and the observed field data for both the average mately 2.8 m was reached in the flat area at the middle thickness and the planar extension of the deposits, of the slope, and a maximum velocity of approximately especially in the lower part of the slope. 12 m/s was predicted. The Hunter & Fell (2003) results agree with the numerical results, but the (Corominas 1996) and (Rickenmann 1999) equations seem to underestimate Conclusions the runout distance. Rapid landslides represent one of the most dangerous To obtain more information about the 2015 landslide, natural hazards and are one of the most frequent natural the flow velocity was calculated along four cross sections disasters in the world. by means of the superelevation of the debris surface in Therefore, prediction of post-failure motion is an the channel belt (Johnson & Rondine, 1984) and the essential component of hazard assessment when a po- Poiseuille equation methods. tential source of a mobile landslide can be located. We classified the Gessi-Mazzalasino landslide into an On March 25th, 2015, at approximately 07:00 PM, a intermediate category between mud flow and earthflow rapid landslide was triggered upstream of the village of phenomena on the basis of the velocities and textural Gessi-Mazzalasino, in the municipality of Scandiano composition. (Emilia Romagna), and it reached the village, causing All the data, obtained by using a range of methods, slight damage to two buildings that were evacuated for confirm that the impact area of possible future events many days. will be smaller than that of the 2015 event, since a Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 14 of 15 Fig. 15 Deposit flow thickness from the DAN3D simulation potential landslide should stop upstream of the village of Received: 7 June 2017 Accepted: 17 October 2017 Gessi-Mazzalasino due to the safety works constructed after the landslide. References The methodology presented in this paper could Amorosi, A., M.L. Colalongo, F. Fusco, G. Pasini, and F. Fiorini. 1999. Glacio-eustatic become a standard procedure in areas affected by differ- control of continental-shallow marine Cyclicity from late quaternary deposits ent types of flow-like landslides, providing a complete of the southeastern Po plain, northern Italy. Quat. Res 52 (1): 1–13. Bandara, S., A. Ferrari, and L. Laloui. 2016. Modelling landslides in unsaturated description of hazards. slopes subjected to rainfall infiltration using material point method. Int. J. Numer. Anal. Meth. Geomech 40: 1358–1380. doi:10.1002/nag.2499. Corominas, J. 1996. The angle of reach as a mobility index for small and large Acknowledgments landslides. Can Geotech J 33: 260–271. This work was carried out within a research contract between the municipality Fredlund, D.G., N.R. Morgenstern, and R.A. Widger. 1978. The shear strength of of Scandiano and the Department of Earth Sciences, University of Florence, unsatured soil. Can. Geotech. J 15 (3): 312–321. entitled: “Studio e Monitoraggio della frana di Gessi-Mazzalasino”.The authors Gigli, G., W. Frodella, F. Garfagnoli, et al. 2014. 3-D geomechanical rock mass would like to thank the local people of the study area for their assistance during characterization for the evaluation of rockslide susceptibility scenarios. fieldwork. We would also like to thank Mr. Giovanni Cantoni for his help during Landslides 11: 131. doi:10.1007/s10346-013-0424-2. the entire project and Dr. Giovanni Bertolini for providing the aerial photos of Gigli G; Morelli S, Fornera S, Casagli N (2014) Terrestrial laser scanner and the landslide. geomechanical surveys for the rapid evaluation of rock fall susceptibility scenarios. Landslides vol. 11(1), pp. 1–14, ISSN:1612-510X DOI. Guzzetti, F. 2000. Landslide fatalities and evaluation of landslide risk in Italy. Eng. Authors’ contributions Geol 58: 89–107. MC, LL, MN and TS contributed to the fieldwork and were responsible for Heim, A. 1932. Bergstruz und Menschenleben, 218. Zurich: Fretz und Wasmuth. collecting, integrating and interpreting the field data, as well as preparing Hsü, K.J. 1975. Catastrophic debris streams (sturzstroms) generated by rockfalls. the manuscript. GG gave technical support and conceptual advice and Geol. Soc. Am. Bull 86 (1): 129–140. contributed to the preparation of the manuscript. All authors read and Hungr, O. 1995. A model for the runout analysis of rapid flow slides, debris flows approved the final manuscript. and avalanches. Can. Geotech. J 32: 610–623. Hungr, O., and McDougall. 2009. S. Two numerical models for landslide dynamic analysis. Comput. Geosci 35: 978–992. doi:10.1016/j.cageo.2007.12.003. Competing interests Hungr, O., G. Morgan, and R. Kellerhals. 1984. Quantitative analysis of debris torrent The authors declare that they have no competing interests. hazards for design of remedial measures. Can. Geotech. J 21: 663–677. Hungr, O., S.G. Evans, M. Bovis, and J.N. Hutchinson. 2001. Review of the classification of landslides of the flow type, 221–238. VII: Environmental and Publisher’sNote Engineering Geoscience. Springer Nature remains neutral with regard to jurisdictional claims in Hunter GJ, Fell R, (2003) The deformation behavior of embankment dams. UNICIV published maps and institutional affiliations. Report, R-416, School of Civil & Environmental Engineering, UNSW, Sydney Australia. Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 15 of 15 Iverson, R.M. 1997. The physics of debris flows. Rev. Geophys 35 (3): 245–296. doi:10.1029/97RG00426. Johnson, A.M., and J.R. Rodine. 1984. Debris flowin slope instability. Prior DB, 257–361. Chichester: Wiley. Jordan, R.P. 1994. Debris flows in the southern Coast Mountains, British Columbia: Dynamic behavior and physical properties, 258. Canada: Department of Geography, University of British Columbia. McDougall, S., and O. Hungr. 2004. A model for the analysis of rapid landslide motion across three - dimensional terrain. Can. Geotech. J 41 (6): 1084–1097. doi:10.1139/T04-052. Nocentini, M., V. Tofani, G. Gigli, F. Fidolini, and N. Casagli. 2015. Modeling debris flows in volcanic terrains for hazard mapping: the case study of Ischia Island (Italy). Landslides 12 (5): 831–846. Philip, J.R. 1985. Approximate analysis of the borehole permeameter in unsaturated soil. Water Res 21: 1025–1033. Rickenmann, D. 1999. Empirical relationship for debris flows. Nat. Hazards 19: 47–77. Salvatici, T., S. Morelli, V. Pazzi, et al. 2017. Debris flow hazard assessment by means of numerical simulations: Implications for the Rotolon creek valley (northern Italy). J. Mt. Sci 14 (4). doi:10.1007/s11629-016-4197-7. Voellmy A (1955). Ueber die Zerstoeerunskraft von Lawinen Schweizerische Bauzeitung. English version “on the destructive force of avalanches” translated by Tate R.E. (1964), ed. US Department of Agricolture Forest Service. Wagner, A.A. 1957. The use of the unified soil classification system by the bureau of reclamation, 125. London: Proceedings of Fourth International Conference on Soil Mechanics and Foundation Engineering. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Geoenvironmental Disasters Springer Journals

Numerical modeling and characterization of a peculiar flow-like landslide

Download PDF
15 pages

Loading next page...
 
/lp/springer-journals/numerical-modeling-and-characterization-of-a-peculiar-flow-like-W0BxLrUZvs

References (23)

Publisher
Springer Journals
Copyright
Copyright © 2017 by The Author(s).
Subject
Environment; Environment, general; Earth Sciences, general; Geography, general; Geoecology/Natural Processes; Natural Hazards; Environmental Science and Engineering
eISSN
2197-8670
DOI
10.1186/s40677-017-0087-8
Publisher site
See Article on Publisher Site

Abstract

Background: On March 25th, 2015, a rapid landslide occurred upstream of the village of Gessi-Mazzalasino, in the municipality of Scandiano, affecting two buildings. Rapid landslides, due to their high velocity and mobility, can affect large areas and cause extensive damage. Considering the often unpredictable kinematics of landslides, the post-failure behavior has been studied by many authors to predict the landslide runout phase for hazard assessment. Findings: With the aim of characterizing the Gessi-Mazzalasino landslide, field surveys were integrated with the results of laboratory tests. The geometric characteristics (thickness, area and volume) and kinematic aspects of the landslide were estimated by using a laser scanning survey and geomorphological data. To model the landslide and obtain its rheological parameters, a back analysis of the event was performed by means of a depth-averaged 3D numerical code called DAN3D. The results of the back analysis of the landslide propagation were validated with field surveys and velocity estimations along selected sections of the landslide. Finally, potential areas prone to failure or reactivation were identified, and a new simulation was performed that considered the back-calculated rheological parameters. Conclusions: Rapid landslides are one of the most dangerous natural hazards and are one of the most frequent natural disasters in the world. Therefore, prediction of post-failure motion is an essential component of hazard assessment when a potential source of a mobile landslide it is located. To assess the risk affecting the area, both numerical and empirical methods have been proposed, in order to predict the runout phase of the phenomenon. For the numerical modelling of the landslide, carried out with DAN-3D code, the best results were obtained by using a Voellmy reological model, with a constant turbulence parameter (ξ) of 250 m/s and a friction parameter (μ) comprised between 0.15 and 0.19. The rheological parameters obtained through dynamic back analyses were used to evaluate the propagation phase and the deposition areas of new potential landslides, that could affect the same area of the 25th March 2015 event. The predicted runout length obtained by the DAN3D software was compared to runout lengths predicted by the Corominas (Can Geotech J 33:260–271, 1996), (Nat. Hazards 19, 47-77) and (UNICIV Report, R-416, School of Civil & Environmental Engineering, UNSW, Sydney Australia 2003) empirical relations. All the data confirm that the impact area of possible future events will be smaller than the 2015 event, probably due to the safety measures established after the landslide. Keywords: Landslide, Geotechnical characterization, Runout simulation, Scandiano * Correspondence: mattia.ceccatelli@unifi.it Department of Earth Sciences, University of Florence, Via G. La Pira, 4, 50121 Florence, Italy © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 2 of 15 Introduction At approximately 07:00 PM on March 25th, 2015 a Rapid landslides such as debris flows, debris avalanches, rapid landslide was triggered upstream of the village of rock avalanches and flow slides are instability phenom- Gessi-Mazzalasino, in the municipality of Scandiano ena that affect superficial deposits as a consequence of (Emilia Romagna region), in north-central Italy (44°34′ intense and prolonged rainfall events. Rapid landslides 45″N,10°39′18 E, Fig. 1) due to days of heavy and are one of the most dangerous and frequent natural haz- persistent rainfall. ards in the world and can cause significant damage to The landslide was triggered at approximately 220 m goods and people in their path. a.s.l. and reached the village, causing slight damage to Guzzetti (2000) showed that more than 80% of the two buildings that were evacuated. deaths and injuries due to landslides in Italy were related One of the main challenges regarding the analysis of to fast-moving failures, including debris flows, rockfalls, rapid landslides is that they are affected by different rockslides, and soil slips. mechanisms during the failure and post-failure stages Fig. 1 a Topographic map and b) geological map of the study area (from Servizio Geologico Sismico e dei Suoli, Emilia Romagna region, 2011) Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 3 of 15 (Bandara et al., 2016). Many studies can be found in the Landslide description literature that are focused on the analysis of landslides The landslide source area is located on the slope using experimental and mathematical methods. upstream of the village of Gessi-Mazzalasino (Fig. 2) at Empirical formulas derived from the statistical data of an elevation between 230 and 215 m a.s.l. The triggering past landslides can provide valuable information (Hsü, event was likely heavy rainfall that occurred a few days 1975; Corominas, 1996), but these formulas are generally before the event. The Cà de Caroli weather station, lo- approximations, and the obtained information is usually cated 1 km northeast of the study area, recorded more limited to specific contexts. To better understand the ef- than 100 mm of rainfall over a 10-day period (Fig. 3), fects of landslides, numerical modeling is a particularly with a peak rainfall intensity of approximately 60 mm on useful tool that is capable of capturing the entire land- March 25th (Fig. 4); for comparison, the annual average slide process in both space and time. In this paper, we rainfall at this station is approximately 750 mm. conduct a combined analysis, via both empirical and nu- The source area is approximately 1700 m and has an merical approaches, to characterize the 2015 irregular shape. The average slope of the source area is phenomenon and to obtain additional information for a approximately 17–20°, but the slope increases up to 30– risk assessment of the area. 35° in the triggering area. A back analysis of the post-failure behavior was con- The initial failure triggered at approximately 220 m ducted with the DAN3D code (McDougall & Hungr, a.s.l., approximately 20 m below the ridge. This altitude 2004; Hungr & McDougall, 2009), using a trial and error difference may reflect an increase in the pore pressure procedure, to obtain the rheological parameters of the due to the hydraulic head, which may represent an add- March 25th, 2015 phenomenon. itional instability parameter of the slope in addition to DAN3D was developed for the simulation of extremely the heavy rainfall. rapid landslides, even in complex topographies (McDou- The flow-like landslide, after initially spreading in a gall & Hungr 2004, Salvatici et al. 2017). Since this code flat area in the middle sector of the slope, moved is also capable of simulating material motion and its through an existing impluvium and reached the corresponding rheological changes (Hungr, 1995), inhabited area at the foot of the hill, 450 m below the DAN3D was used to study the 2015 event. The simula- source area. tion results were validated by means of runout lengths The thickness of the deposits ranges from a few and flow velocities derived from empirical runout decimeters up to 2–3 m in the most significant accumu- prediction methods. lations areas. Finally, a forecast analysis was carried out to evaluate Field evidence showed that the mass movement the characteristics of potential landslides that could occur started as a sliding mass of the surface layers at the in the area in the future, using the rheological parameters upper part of the slope and evolved into a rapid mud obtained by the back analysis of the 2015 event and the flow at the top of the impluvium, where the flow was post-event digital elevation model (DEM) of the area. channeled into, probably due to the addition and mixing of surface water during the mass movement. The source volume, approximately 10,000 m , and the planimetric Study area and landslide description area of the landslide have been identified from the field Study area observations and aerial images that were collected after The study area is located in the municipality of Scan- the event. diano, in the Emilia Romagna region. The landslide af- fected the western slope of the Tresinaro Valley, above Geotechnical characterization the village of Gessi-Mazzalasino. Two soil samples were collected from the landslide de- The area is geologically characterized by units of the posits immediately after the event to perform a geotech- External Liguride domain and the Neogene-Quaternary nical characterization of the materials and recreate the succession of the Northern Apennines. The External initial flow conditions. Liguride domain consists of thin calcareous turbiditic The first soil sample was collected in the deposit area formations known as the Palombini Shales and at the beginning of the channelized section, and the Varicolored Shales, while the Neogene-Quaternary suc- second sample was collected at the landslide toe (Fig. 5). cession in this area is represented by the Gessoso- These samples were subjected to the following labora- Solfifera Formation, the alluvial units of the Ravenna tory tests: index property testing, Atterberg limits test- Subsynthem, and the Modena Unit (Amorosi, 1999). ing, grain size analysis, and direct shear testing. Several active and inactive landslide deposits are also In addition, three geotechnical in situ tests were car- located in the area; the 2015 landslide originated from ried out with the aim of collecting further information one of these deposits (Fig. 1). about the soil in its natural condition. Specifically, two Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 4 of 15 Fig. 2 Aerial photograph of the Gessi-Mazzalasino landslide (photo by G. Bertolini) permeability measurements were carried out with a con- The BST results were interpreted using the Fredlund et stant compact head permeameter (Amoozemeter) and al. (1978) shear strength equation for unsaturated soils. one measurement of shear strength was carried out with a borehole shear test (BST). 0 0 0 The in situ tests were carried out both outside of the τ ¼ c þðÞ σ−u tanΦ þðÞ u −u tanΦ ð1Þ a a w landslide area (test number 3) and within the landslide deposit (test number 4). where τ is the shear strength, c’ is the effective cohe- The BST tests were performed on soils in unsaturated sion, σ is the total normal stress, u is the pore air pres- conditions; at an equivalent depth, matric suction values sure due to surface tension, φ′ is the effective friction (u − u ) were measured with tensiometers. angle, u is the pore water pressure, and φ is the angle a w w b Fig. 3 Daily rainfall intensity from 13/03/2015 to 30/03/2015 Fig. 4 Hourly rainfall intensity on 25/03/2015 Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 5 of 15 Fig. 5 Location of the soil samples (in red) and the geotechnical field investigation (in cyan); point 3: permeability measurement; point 4: permeability measurement and BST expressing the rate of the increase in strength related to The measured saturated hydraulic conductivity ranges −6 −7 matric suction. The BST test results show that the in- from 2.19 × 10 m/s to 5.19 × 10 m/s, corresponding ternal friction angle equals 33.8°. to samples 3 and 4. The procedure used for measuring k in the field is Samples 1 and 2 are primarily unsorted silty soils (Fig. 6) called the constant-head well permeameter technique and are classified as a clay of low plasticity (CL) and a silt (Philip, 1985), and it is carried out in a borehole. This (ML), respectively, following the Unified Soil Classification procedure allowed us to measure the amount of water System (USCS, Wagner, 1957). The samples have plasticity flowing through the soil in a given time interval under index (IP) values ranging from 8 to 11. soil-saturated conditions. The saturated permeability of Direct shear tests were performed on reconstituted sam- the soil is evaluated with the Glover solution: ples, using normal stresses between 40 and 80 kPa, deter- mined from the in situ characteristics. The internal friction angle ranges from 29.1 to 30.1°, while the cohesion (c’)is 2 2 −1 Q sin ðÞ h=r − þ 1 þ r=h very low. The results of the laboratory and in situ tests are showninTable 1. k ¼ ð2Þ 2πh In Fig. 7, the matrix compositions of the two soil sam- ples from the 2015 landslide are compared with the where Q is the steady-state rate of water flow from the compositions of earth flows, debris flows and mud flows −1 permeameter into the soil, sinh is the inverse hyper- from several areas of the world (Hungr et al., 2001). bolic sine function, h is the depth of water in the bore- Hungr et al. (2001) distinguished different materials hole, and r is the radius of the borehole. involved in flow-like landslides on the basis of several Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 6 of 15 Fig. 6 Granulometric curves for the two soil samples (black line for sample and red line for sample 2) material geotechnical properties. Both the samples from the sources may have clay contents greater than 10% and 2015 landslide deposit had low plasticity indices (IP =11 for plasticity indices of more than 10 (Jordan, 1994). sample 1 and IP = 8 for sample 2) and the liquidity index Therefore, the distinction between “mud” and “earth” (IL) values were approximately 0.6 and 0.5, respectively. should not be based solely on grain size distribution but As shown in Fig. 7, the matrix compositions of the should instead be derived from the context of each land- 2015 landslide samples fall in the textural field of earth slide class. flows and mud flows, while debris flows typically contain Specifically, earth flows and mud flows may involve less than 30% silt and finer particles. material of similar texture but are significantly different A comparison of colloidal indices does not allow a in other ways; in particular, the velocity of movement clear distinction to be made between these two different during an earth flow differs from that of a mud flow. classes. Earth flows have clay contents ranging from 10% to 70%, averaging approximately 35%, while debris and Velocity analysis mud flows are usually not plastic or are only weakly There are many equations in the literature for estimating plastic. However, some mud flows derived from volcanic the velocity of the frontal part of flow-like landslides (Hungr et al. 1984). These relations provide a useful par- ameter to validate the back analysis results (Salvatici et al., 2017, Nocentini et al., 2015). Table 1 Geotechnical parameters obtained from laboratory and In this work, flow velocity was estimated in the channel- in situ tests ized section of the landslide, along the cross sections shown Sample 1 2 3 4 in Fig. 8, by using two methods: the superelevation of the USCS classification CL ML –– debris surface in the channel belt (Johnson & Rondine, Porosity [%] 40.1 39.7 –– 1984) and the Poiseuille equation (Hungr et al., 1984). Void ratio [−] 0.67 0.7 –– The Johnson & Rondine (1984) relation is based on the difference in the splash heights on the inside and outside Saturation degree [%] 118.3 117.9–– of the bends in the flow path (Nocentini et al., 2015). Total unit weight [kN/m ] 20.2 19.9 –– The superelevation of the debris wave around the Statured unit weight [kN/m ] 19.5 19.3 –– channel bends tends to be higher than that on the op- Liquid limit 34 33 –– posite side due to the centrifugal force (Fig. 9). Plastic limit 23 25 –– Thus, in cross sections 1 and 2, the velocity can be cal- Plasticity index [%] 11 8 –– culated by using the following equation: Liquidity index [%] 0.6 0.5 –– pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi v ¼ gR cosδ tanβ ð3Þ Friction angle [°] 30.1 29.1 33.8 – Cohesion [kPa] 2 5 0 – where β is the angle between the line connecting the −6 −7 Permeability [m/s] –– 2.19 × 10 5.19 × 10 top of the debris waves at both sides of the section and a Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 7 of 15 Fig. 7 Ternary plot of the two soil samples with textural classification (from Hungr et al., 2001) horizontal line, δ is the slope angle of the flow path, R is to determine the position of up to 12,000 points per sec- the radius of curvature and g is the gravity acceleration. ond, with a maximum angular resolution of 0.008° and The radius of curvature of the channel was obtained an accuracy of ±10 mm from a maximum distance of graphical processing using a 1:5000 topographic map. 800 m. The Hungr et al. (1984) relation, which is based on the To completely cover the intervention areas and avoid Poiseuille equation, can be used to evaluate the flow vel- the shadow areas, were captured four scans from differ- ocity in the straight sections (cross sections 3 and 4). ent positions (Fig. 10). This equation relates the velocity to the geometric char- Several laser cylindrical reflectors were placed on the acteristic of the path, the unit weight of the flow mass, hill slopes, and their coordinates were defined by per- and the viscosity of the flow mass: forming a GPS survey. These tie points were later used to align the point clouds. This process is required for correctly georeferencing the point cloud on a chosen ref- γ sinδH v ¼ ð4Þ erence system and for merging two or more scans of the lv same object realized from different points of view. where y is the unit weight of the material and is ob- On April 22nd, 2015, another GPS survey was carried tained by laboratory test, δ is the slope angle of the flow out to reconstruct the exact geometry of the landslide body, path, H is the flow depth, l is a constant based on the define thesource areaand identify trenches that could de- cross-sectional shape of the channel (3 for a broad velop into the edges of potential detachment areas. channel and 8 for a semicircular channel) and ν is the The data obtained from the laser scanning surveys dynamic viscosity of the flow (assumed to be 3, as indi- have been processed to obtain a high-resolution DEM of cated by Hungr et al., 1984). the area (Fig. 11). The results of the estimated velocity and the geometric Some DEM sectors outside of the landslide were not parameters of the path are summarized in Tables 2 and 3. acquired due to the presence of buildings and dense vegetation, particularly near the toe portion; therefore, it Laser scanning survey was necessary to integrate the model with an existing New high-resolution surveying techniques, such as ter- 1:5000 topographic map. restrial laser scanning, quickly obtain detailed 3D terrain During the data processing, the safety works on the models that can be employed in runout analyses (Gigli landslide started. These works, in an initial phase, in- et al., 2014). cluded the construction of an earthfill dam (Fig. 12) to A laser scanning investigation was performed during prevent the excessive expansion of future landslides and two field surveys, on April 1st, 2015, and April 16th, to channel the flow of those potential landslides towards 2015, by means of a long-range 3D terrestrial laser im- the existing channel. One last GPS survey, on September aging sensor (RIEGL LMS-Z420i device), which is able 9th, 2015, was carried out to detect the geometry and Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 8 of 15 Fig. 8 Location of the cross sections for flow velocity estimation. a) Location of the cross sections for flow velocity estimation and b) Sections profiles location of the earthfill dam, which was then imple- DAN3D numerical model mented in the digital terrain model, to capture the DAN3D is a 3D numerical model that uses the continu- modified shapeof theslopeand thepost-landslide ous Lagrangian approach for integrating the equations conditions. of Saint-Venant with depth. The mass conservation equation governs the model: ∂h ∂v ∂v ∂b Runout simulation methods x y þ h þ ¼ ð5Þ Methods to predict landslide runout were grouped ∂t ∂v ∂y ∂t into two categories by (Rickenmann 1999): the first group includes empirical methods that are based on where b is the bed-normal erosion-entrainment depth, statistical analyses of past events (Iverson, 1997; Cor- v and v are local flow velocities, and t is time. x y ominas, 1996; Hunter & Fell, 2003), and the second DAN3D employs a simple semi-empirical approach group includes analytical methods that account for based on the concept of “equivalent fluid”, as defined by conservation of momentum and energy to simulate (Hungr 1995). the propagation of flow using 2D or 3D models In this method, the landslides are considered one (Hungr, 1995; McDougall & Hungr, 2004; Hungr & material governed by simple rheological relations. There- McDougall, 2006). In this work, the runout distance fore, an internal frictional rheology is considered, as well obtained by the DAN3D code was compared with as a basal rheology that depends on one or two parame- that from empirical methods. ters (depending on the chosen rheological model) that Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 9 of 15 Fig. 9 Empirical formula parameters for flow velocity estimation: a) splash heights on the inside and outside of the bends of the flow path (cross section 2); b) radius of curvature; and c) slope angle are established with a calibration procedure by using during the dynamic modeling of a landslide is related to back analysis. the expected event type and depends on the rheological The model requires three input files that describe the characteristics of the landslide material. topography (path file), source area (source file), and The post-failure phase processes that are triggered dur- number of materials used with their rheologies and ing the movement of rapid landslides are extremely com- erosion parameters (erosion file). plex, and the direct measurement of either the parameters With the aim of simulating different types of fast land- of the involved materials or of the characteristics of the slides, DAN3D can implement the following rheological landslide is impossible. Therefore, the best rheological relations: frictional, plastic, turbulent, Bingham and model is determined by performing a back analysis for to Voellmy. The selection of rheological model to use the investigated case by using similar phenomena. Table 2 Velocity values of the flow obtained according to the Table 3 Velocity values of the flow obtained according to the Johnson & Rondine (1984) formula Hungr et al. (1984) formula Section gR δβ v Section γδ Hl ν v 2 3 [m/s ] [m] [°] [°] [m/s] [kN/m ] [°] [m] [−] [kPa] [m/s] 1 9.81 13.7 12 4.17 3.10 3 9.81 12 3.3 8 3 1.89 2 9.81 15.9 12 2.75 2.71 4 9.81 12 3 8 3 1.56 Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 10 of 15 Fig. 10 Laser scanner point cloud from four scan positions Empirical relations 0:16 0:83 L ¼ 1:9V H ð7Þ There are several empirical methods that relate the geometric parameters of landslides to the runout dis- 0:69tana þ 0:086 ð8Þ tance; in this case study, three methods, presented by (Corominas 1996), (Rickenmann 1999) and Hunter & Fell (2003), were used to correlate the angle of reach where A and B are coefficients that depend on the (fahrboschung) with the volume. landslide types and α is the slope inclination. The fahr- boschung was defined by (Heim 1932) as the inclination of the line connecting the crest of the landslide source Log ¼ BlogV þ A ð6Þ with the toe of the deposits and can be evaluated by the Fig. 12 Earthfill dam, built during safety operations in the Fig. 11 DEM of the area landslide area Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 11 of 15 ratio between the elevation difference of the highest and ρgv τ ¼ − f σ þ ð10Þ zx z lowest points of flow (H) and the corresponding hori- zontal distance (L): where f is the frictional coefficient; ξ is a turbulence parameter representing all possible sources of velocity- tanα ¼ ð9Þ dependent resistance in landslide dynamics; and ρ, g and These methods may be applied in preliminary hazard assess- v are the density, gravity and velocity, respectively. ments, and they may be compared with the dynamic analysis. For the determination of the rheological model param- eters, a back analysis was performed based on the study Findings and Results of the deposits from the March 2015 collapse. Two main Back analysis phases of motion were simulated, according to the field The calibration procedure is based on a trial and error evidence, assuming that the mass movement started as a back analysis of the Gessi-Mazzalasino landslide, to sliding mass involving the surface layers of the upper identify the most suitable rheological parameters for de- part of the slope and evolved into a rapid mud flow as it scribing the flow motion. The path file used for the back entered the impluvium, where the flow was channelized. analysis was the DEM of the slope, obtained from a Two basal shear resistances were chosen, according to 1:5000 topographic map of the area. The source file was the landslide dynamics; one rheology material was used defined by the triggering area obtained from the field for the source area between 226 and 173 m a.s.l., and an- survey. Since no significant erosion was observed during other was assumed for the channelized area between the field surveys, the erosion file was neglected. 173 and 131 m a.s.l. The Voellmy resistance parameters The numerous study cases analyzed with the DAN3D were adjusted by trial and error to achieve the best code have shown that the Voellmy rheological model is simulation match in terms of velocity, thickness of particularly suitable to describe this type of phenomena deposits and runout distance. and that, accordingly, it should be used for modeling the The best match between the actual and simulated Gessi-Mazzalasino landslide (Nocentini et al., 2015; Gigli material distributions was obtained using a frictional et al., 2014). coefficient of f = 0.19 for the upper material and a fric- This model, introduced by (Voellmy 1955) for snow tional coefficient of f = 0.15 for the lower material, while avalanches, contains a friction term and a turbulence the turbulence term of ξ = 250.00 m/s remained term: constant throughout the event (Fig. 13). Fig. 13 Comparison between the maximum runout distances obtained from the empirical and numerical methods with the field data (in red) Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 12 of 15 Additionally, DAN3D requires the internal frictional Table 5 Comparison between the DAN3D and empirical velocity values angle and the unit weight of the material as input pa- rameters. On the basis of the performed laboratory tests, Cross section DAN3D Johnson & Rondine Hungr et al. (1984) (1984) we used an average internal frictional angle of 30° and a [m/s] [m/s] [m/s] unit weight of 20 kN/m . The simulation results show that the landslide reaches 1 7.1 3.10 – the flat area located in the middle sector of the slope, 2 6.9 2.71 – where most of the mobilized material was deposited in a 3 5.1 – 1.89 layer up to 2 m thick, and then moves through the im- 4 4.6 – 1.56 pluvium until it reaches the inhabited area. Runout distances decreased. These results agree with the testimonies of As show in Fig. 13, there is good agreement between the local residents, who evaluated the flow velocity in the simulation output and the field data obtained by the final meters of movement at approximately 0.2–1 m/s. GPS survey (landslide path), for both the average thick- Examples of velocity observations from various ness and the planar extension of the deposits, especially sources are shown in Fig. 14, together with the velocity in the lower part of the slope. values calculated in this paper. These velocities represent The results are summarized in. point observations or maximum values at randomly Table 4 and show some differences between the chosen locations and are not necessarily maxima for a methods used to evaluate the runout distance. given event. The (Corominas 1996) and (Rickenmann 1999) rela- A clear distinction can be made between extremely tions, which produced predicted runout distances of rapid processes such as debris flows, mud flows and deb- 311 m and 379 m, respectively, underestimate the max- ris avalanches and slow process such as earthflows imum runout distance of the landslide. However, the (Hungr et al., 2001). Hunter & Fell (2003) relation, with a predicted runout distance of approximately 431 m, agrees with the Landslide characterization modeling results and field data; therefore, this is the As shown in the previous paragraphs, the distinction be- most suitable empirical method to describe the Gessi- tween “mud flows” and “earth flows” cannot be based Mazzalasino landslide. solely on grain size distribution but can instead be de- rived in other ways, in particular, from the velocity of Velocity calculations movement. By comparing the flow velocity values obtained by using From the velocities obtained with three direct methods the empirical equations and the numerical model results, (including resident testimonies and empirical methods), a similar trend along the flow travel distance was ob- the Gessi-Mazzalasino landslide has an estimated vel- served (Salvatici et al., 2017). ocity ranging between 0.2 and 3 m/s, while the velocities The flow velocity was obtained with the models along from the numerical simulation are higher, ranging be- the four cross sections, the DAN3D results were slightly tween 4.5 and 12 m/s. higher than the velocities obtained by the empirical rela- These differences derive from the assumptions made tions (Table 5). during the simulation phase, resulting in an overesti- These differences derive from the assumptions made mation of the flow velocity. during the simulation phase, making it difficult to model Analysis of the velocities allow better classification of both the kinematic and the depositional parameters of the Gessi-Mazzalasino event, which has characteristics the flow, which were the focus of our study. of a landslide with behaviors between those of mud flow According to the model results, the flow reached a and earthflow phenomena. maximum velocity of approximately 8–12 m/s in the The Gessi-Mazzalasino landslide material is an un- upper part and then slowed down as it entered the im- sorted deposit composed of a mixture of sand, gravel pluvium, and it finally stopped when the slope and cobbles as well as varying proportions of silt and clay, and the landslide event is characterized by low Table 4 Comparison between the calculated and measured plasticity and an intermediate velocity. runout distances Corominas Rickenmann Hunter & Fell DAN3D Field data Future conditions (1996) (1999) (2003) To evaluate the characteristics of potential landslides [m] [m] [m] [m] [m] that could occur in the area, another numerical simula- 311 379 430 426 412 tion was performed with DAN3D using the rheological Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 13 of 15 Fig. 14 Range of velocities for various types of flow-like landslides; the range of the estimated Gessi-Mazzalasino landslide velocity is shown in red (square symbols for resident testimonies, triangular symbols for empirical methods and circle symbols for numerical simulation) parameters obtained by the back analysis of the 2015 To assess the risk affecting the area, different methods event. have been proposed to predict the runout phase of the For this analysis, the most recent DEM obtained by phenomenon. laser scanner survey was used, representing the topog- A back analysis based on the path and the deposits of raphy of the slope after the landslide. the 2015 event was performed in order to identify the During the analysis, a single volume of 10,000 m was optimal rheological model and to simulate the behavior investigated based on the field survey and on the conser- of potential landslides. vative assumption that failure occurred for a single The dynamic modeling was carried out by using the volume. DAN3D code, which estimated the extent of the impact Figure 14 shows the maximum runout distances ob- area and mapped the distribution of landslide tained from the numerical methods used in this study. parameters. The results of the potential landslide show that, due to The predicted runout length obtained by the DAN3D the safety works, the landslide impact area was predicted software was compared to runout lengths predicted by to be smaller than that of the 2015 event (Fig. 15), with the (Corominas 1996), (Rickenmann 1999) and Hunter the mass stopping 70 s after the start of the simulation & Fell (2003) empirical relations. at the beginning of the impluvium, where the earthfill There is good agreement between the simulation out- dam was constructed. A maximum thickness of approxi- put and the observed field data for both the average mately 2.8 m was reached in the flat area at the middle thickness and the planar extension of the deposits, of the slope, and a maximum velocity of approximately especially in the lower part of the slope. 12 m/s was predicted. The Hunter & Fell (2003) results agree with the numerical results, but the (Corominas 1996) and (Rickenmann 1999) equations seem to underestimate Conclusions the runout distance. Rapid landslides represent one of the most dangerous To obtain more information about the 2015 landslide, natural hazards and are one of the most frequent natural the flow velocity was calculated along four cross sections disasters in the world. by means of the superelevation of the debris surface in Therefore, prediction of post-failure motion is an the channel belt (Johnson & Rondine, 1984) and the essential component of hazard assessment when a po- Poiseuille equation methods. tential source of a mobile landslide can be located. We classified the Gessi-Mazzalasino landslide into an On March 25th, 2015, at approximately 07:00 PM, a intermediate category between mud flow and earthflow rapid landslide was triggered upstream of the village of phenomena on the basis of the velocities and textural Gessi-Mazzalasino, in the municipality of Scandiano composition. (Emilia Romagna), and it reached the village, causing All the data, obtained by using a range of methods, slight damage to two buildings that were evacuated for confirm that the impact area of possible future events many days. will be smaller than that of the 2015 event, since a Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 14 of 15 Fig. 15 Deposit flow thickness from the DAN3D simulation potential landslide should stop upstream of the village of Received: 7 June 2017 Accepted: 17 October 2017 Gessi-Mazzalasino due to the safety works constructed after the landslide. References The methodology presented in this paper could Amorosi, A., M.L. Colalongo, F. Fusco, G. Pasini, and F. Fiorini. 1999. Glacio-eustatic become a standard procedure in areas affected by differ- control of continental-shallow marine Cyclicity from late quaternary deposits ent types of flow-like landslides, providing a complete of the southeastern Po plain, northern Italy. Quat. Res 52 (1): 1–13. Bandara, S., A. Ferrari, and L. Laloui. 2016. Modelling landslides in unsaturated description of hazards. slopes subjected to rainfall infiltration using material point method. Int. J. Numer. Anal. Meth. Geomech 40: 1358–1380. doi:10.1002/nag.2499. Corominas, J. 1996. The angle of reach as a mobility index for small and large Acknowledgments landslides. Can Geotech J 33: 260–271. This work was carried out within a research contract between the municipality Fredlund, D.G., N.R. Morgenstern, and R.A. Widger. 1978. The shear strength of of Scandiano and the Department of Earth Sciences, University of Florence, unsatured soil. Can. Geotech. J 15 (3): 312–321. entitled: “Studio e Monitoraggio della frana di Gessi-Mazzalasino”.The authors Gigli, G., W. Frodella, F. Garfagnoli, et al. 2014. 3-D geomechanical rock mass would like to thank the local people of the study area for their assistance during characterization for the evaluation of rockslide susceptibility scenarios. fieldwork. We would also like to thank Mr. Giovanni Cantoni for his help during Landslides 11: 131. doi:10.1007/s10346-013-0424-2. the entire project and Dr. Giovanni Bertolini for providing the aerial photos of Gigli G; Morelli S, Fornera S, Casagli N (2014) Terrestrial laser scanner and the landslide. geomechanical surveys for the rapid evaluation of rock fall susceptibility scenarios. Landslides vol. 11(1), pp. 1–14, ISSN:1612-510X DOI. Guzzetti, F. 2000. Landslide fatalities and evaluation of landslide risk in Italy. Eng. Authors’ contributions Geol 58: 89–107. MC, LL, MN and TS contributed to the fieldwork and were responsible for Heim, A. 1932. Bergstruz und Menschenleben, 218. Zurich: Fretz und Wasmuth. collecting, integrating and interpreting the field data, as well as preparing Hsü, K.J. 1975. Catastrophic debris streams (sturzstroms) generated by rockfalls. the manuscript. GG gave technical support and conceptual advice and Geol. Soc. Am. Bull 86 (1): 129–140. contributed to the preparation of the manuscript. All authors read and Hungr, O. 1995. A model for the runout analysis of rapid flow slides, debris flows approved the final manuscript. and avalanches. Can. Geotech. J 32: 610–623. Hungr, O., and McDougall. 2009. S. Two numerical models for landslide dynamic analysis. Comput. Geosci 35: 978–992. doi:10.1016/j.cageo.2007.12.003. Competing interests Hungr, O., G. Morgan, and R. Kellerhals. 1984. Quantitative analysis of debris torrent The authors declare that they have no competing interests. hazards for design of remedial measures. Can. Geotech. J 21: 663–677. Hungr, O., S.G. Evans, M. Bovis, and J.N. Hutchinson. 2001. Review of the classification of landslides of the flow type, 221–238. VII: Environmental and Publisher’sNote Engineering Geoscience. Springer Nature remains neutral with regard to jurisdictional claims in Hunter GJ, Fell R, (2003) The deformation behavior of embankment dams. UNICIV published maps and institutional affiliations. Report, R-416, School of Civil & Environmental Engineering, UNSW, Sydney Australia. Ceccatelli et al. Geoenvironmental Disasters (2017) 4:23 Page 15 of 15 Iverson, R.M. 1997. The physics of debris flows. Rev. Geophys 35 (3): 245–296. doi:10.1029/97RG00426. Johnson, A.M., and J.R. Rodine. 1984. Debris flowin slope instability. Prior DB, 257–361. Chichester: Wiley. Jordan, R.P. 1994. Debris flows in the southern Coast Mountains, British Columbia: Dynamic behavior and physical properties, 258. Canada: Department of Geography, University of British Columbia. McDougall, S., and O. Hungr. 2004. A model for the analysis of rapid landslide motion across three - dimensional terrain. Can. Geotech. J 41 (6): 1084–1097. doi:10.1139/T04-052. Nocentini, M., V. Tofani, G. Gigli, F. Fidolini, and N. Casagli. 2015. Modeling debris flows in volcanic terrains for hazard mapping: the case study of Ischia Island (Italy). Landslides 12 (5): 831–846. Philip, J.R. 1985. Approximate analysis of the borehole permeameter in unsaturated soil. Water Res 21: 1025–1033. Rickenmann, D. 1999. Empirical relationship for debris flows. Nat. Hazards 19: 47–77. Salvatici, T., S. Morelli, V. Pazzi, et al. 2017. Debris flow hazard assessment by means of numerical simulations: Implications for the Rotolon creek valley (northern Italy). J. Mt. Sci 14 (4). doi:10.1007/s11629-016-4197-7. Voellmy A (1955). Ueber die Zerstoeerunskraft von Lawinen Schweizerische Bauzeitung. English version “on the destructive force of avalanches” translated by Tate R.E. (1964), ed. US Department of Agricolture Forest Service. Wagner, A.A. 1957. The use of the unified soil classification system by the bureau of reclamation, 125. London: Proceedings of Fourth International Conference on Soil Mechanics and Foundation Engineering.

Journal

Geoenvironmental Disasters – Springer Journals

Published: Oct 24, 2017

There are no references for this article.