Director: Prof. Michel Jaboyedoff
Supervisor: MSc. Benjamin Rudaz
Water erosion phenomenons are increasingly studied and understood but raindrop erosion is far more complex. Raindrop erosion includes subprocesses such as impact, cratering, rim formation, daughter drop splash and soil particle splash. This work is focused on modeling complex ballistic trajectories of soil particle splashes and particle dispersion process. The general purpose is first to recreate the splash effect in laboratory and second to provide optimal numerical models and a better understanding of the soil particle splash.
Since the complex multiphase interactions are difficult to model, it is easier to compute numerically the dispersion process. Physical based models are the most common approach in this investigation field. The first assumption is that the crater shape might be the controlling factor in the dispersion process governed by the average splash distance. Moreover, complex physical based models may govern ballistic trajectories. These assumptions have now to be proven.
Phenomenological observations are given by experiments in laboratory, on a setup inspired by Furbish et al. (2007) study. Fine soil samples are used in this work and advanced grain size analysis is performed using laser diffraction technique. High-speed camera acquisitions and micro-LiDAR records are used throughout experimental investigations. Then, impact velocities are measured as well as crater shape or particle dispersion. Measured velocities tend to be close to those computed by numerical simulations. High-speed photography analysis shows that the mean initial splash angle is dependent on the drop penetration depth. Moreover, the mean splash angle seems to be dependent on the slope at the crater edge.
Numerical computation is then performed to model the dispersion phenomenon. Using a
probabilistic algorithm, the grain size distribution can be taken into account throughout numerical simulation. The initial splash angle mean value is derived from previous assumption about the key role of the crater shape. Gravity, drag and buoyant forces are also taken into account. Model validation is performed by comparisons between experimental and numerical results using digitalized experimental dispersion photography and LiDAR scanning. Several differences between numerical and empirical results are noticed. The shape of the particle splashing distance distribution is found to be not similar. LiDAR acquisition analysis also shows a non-spherical shape for the crater. But a statistical trend exists for the mean crater gradient -mean splash distance relationship. However this has more to do with dierential initial velocities at the crater’s edge than with mean splash angle.
Further perspectives should be oriented in multiphase interactions ( fluid to soil particle) for a better understanding of the whole phenomenon.
Directors: Prof. Michel Jaboyedoff and Prof. Stefan Schmalholz
The Kilchenstock peak is located in the Swiss Alps about 15 km south of the city Glarus. Many rockfalls from the Kilchenstock have been reported since the 19th century. The first study of this rock slope instability is done by Albert Heim in the 30s. The area is mainly composed of folded flysch with a stratigraphy predominantly dipping towards SSE. Heim reveals a sliding mass near the top of the mountain. This represents an unstable volume of 2.5 million cubic meters. At the foot of this area blocks break off and fall towards the town of Linthal 1000m below.
Heim measured velocities of the sliding mass from 1927 to 1932. On two occasions the displacement’s velocity has accelerated (up to 40mm per day) suggesting an imminent large rock-collapse. No catastrophic failure has occurred so far, however such an event remains possible and its characterization consists of the main objectives of the present study.
A detailed structural study is performed based on digital elevation model (DEM) as well as field investigations. The results show that the activity isn’t confined to the area described by Heim, but rather extends to the whole slope, suggesting a larger and deeper potential rock slope deformation at the Kilchenstock. This is supported by different signs of activity and geomorphic gravitation features such as rockfalls, scarps, double ridge, cracks. The existence of several contiguous or imbricated landslides constitute a complex instability who started with the retreat of the glacier
The final results of the study is a detailed geological map of the Kilchenstock and a cross-section in order to better explain which structural parameters constrain the unstable system.
Director: Prof. Michel Jaboyedoff
Jury: Prof. Giovanni Crosta, Prof. Yury Podladchikov, Dr. Irene Manzella, Prof. Suren Erkman
As rock avalanches are rare catastrophic events in which granular masses of rock debris flow at high speeds, commonly with unusually long runout distances, analog and numerical modeling can provide important information about their behavior. This thesis is composed of three main contributions: (1) laboratory experiments in order to demonstrate that the basal roughness and the grainsize as well as the volume and slope angle are important parameters of the motion of a dry granular mass; (2) the analysis of rock avalanche dynamics by means of a detailed structural analysis of the deposits coming from data of 3D measurements of mass movements of different magnitudes, from decimeter level scale laboratory experiments to well-studied rock avalanches of several square kilometers magnitude; (3) development of a numerical model to simulate the laboratory experiments.
Laboratory experiments are performed with a tilting plane. Granular material is released, chutes down a slope, propagates and finally stops on a horizontal surface. Different grainsizes (115, 545 and 2605 μm) and substratum roughness (simulated by sandpapers with grainsize from 8.4 to 269 μm) are used in order to understand their influence on the motion of a granular mass. This work shows that there is a logarithmic relation between the substratum roughness and the motion of the granular flow. For same volume, slope angle and fall height, the runout of the mass is comprised between 4.5 and 11 cm. The influence of the volume and the slope angle is also investigated. The runout increases from 8 to 11 cm with volumes from 300 to 600 cm3. Contrarily to the volume, the slope angle (from 35° to 60°) influences greatly the runout of the mass front (from 5 to 20 cm).
In order to emphasize and better detect the fault structures present in the deposits, we applied a median filter with different moving windows sizes (from 3×3 to 9×9 nearest neighbors) to the 3D datasets and a gradient operator along the direction of propagation. The application of these filters on the datasets results in: (1) a precise mapping of the longitudinal and transversal displacement features observed at the surface of the deposits; (2) a more accurate interpretation of the relative movements along the deposit (i.e. normal, strike-slip, inverse faults) by using cross-sections. Results show how the use of filtering techniques reveals disguised features in the original point cloud and that similar displacement patterns are observable both in the laboratory simulation and in the real scale avalanche, regardless the size of the avalanche.
To simulate the analog granular flow, a numerical model based on the continuum mechanics approach and the solving of the shallow water equations was used. In this model, the avalanche is described from a Eulerian point of view within a continuum framework as single phase of incompressible granular material. The interaction of the flowing layer with the substratum follows a Mohr-Coulomb friction law. Within same initial conditions (slope, volume, basal friction, height of fall and initial velocity), results obtained with the numerical model are similar to those observed in the analog model. In both cases, the runout of the mass is comparable and the size of deposits matches well. Moreover, both analog and numerical modeling provide velocities of same magnitudes. In this study, we highlighted the importance of the friction on a flowing mass and the influence of the numerical resolution on the propagation. The combination of the fluid dynamics equations with the frictional law enables the self-channelization and the stop of the granular mass.
Director: Prof. Michel Jaboyedoff
Co‐director: Dr. Ivanna Penna
Experts: Prof. Stuart Lane, Dr. Karen Sudmeier‐Rieux
Natural and human-induced erosive processes shape landscape by transferring masses from the mountain to downstream areas. They also impact population both located in the source areas of sediments as well as urban areas settle on the depositional area. Mountain areas in Bolivia present high surface dynamics and high rates of rural migrations, causing e.g. a significant increase of population in Cochabamba city in the last 20 years. This work aims to estimate the sediment production on the Jatún Mayu watershed in Cochabamba department taking into account the different origins of sediments.
The region of study consists of a mountain area situated in the Andes with altitudes ranging from 2500 to 4600m. Fieldwork on July 2014 and high-resolution satellite image
interpretation (2004 & 2009) allowed mapping and measuring landslides and gullies. A
hundred of landslides are recorded mostly around the river channel. Most of the gullies are
situated in the upper part of the valley where the vegetation is less abundant on low-sloping agricultural lands.
Photogrammetric reconstructions using camera and drone were the main method used to
characterize some strategic points along the river in order to get dimensions of landslides,
gullies, as well as the riverbed roughness, as the final goal was to model floods at the mouth of the watershed, where migrants have been settling for the last years. A total of 9 points of interests along the riverbed were surveyed and for each of them a square 5x5m surface was analyzed. Approximately 250 pictures by area were needed to estimate roughness along the channel. A flood model has been performed, by using the Riverflo-2D Plus software, to produce a model of the downstream region.
Directors: Prof. Michel Jaboyedoff, Serigne Faye
The Dakar region experiencing recurrent floods in the years 1989 to 2012. These recurrent flooding particularly affect the urban area of Pikine and Guediawaye. Previous studies show that these floods have a positive development in the sector. The various complex causes of these events can be summarized in urban causes, demographic, topographic and hydrogeological. The departments of Pikine and Guediawaye have depressions high endoreic character. These interdunals depressions are called Niayes. They are outcrop areas of the aquifer of Quaternary sands of the Cap-Vert peninsula. So these localities where floods occur are naturally areas for the development of this hazard. These agglomerations of Pikine and Guediawaye were born following the drought of the 1970s. During the drought the rural exodus has led to migration and therefore the urbanization of these areas. They were areas of wet and swampy depressions seasonally. The drought led to believe that they had become dry and therefore they became apt for human habitat. They were so rapidly urbanized, the Dakar region saw its population in the peri-urban area increased from 23’000 in 1960 to 1’044’814 inhabitants in 1998. On the other hand the area of the Cap-Vert Peninsula home under the dune sands of the Quaternary, a free aquifer : the aquifer of Thiaroye. Thus Niayes are outcrops of this aquifer. The aquifer of Thiaroye was exploited in 1950, contact with the ocean has created a risk of causing salt intrusion and stop its pumping between 1959-1961. After the withdrawal of the bevel, the pumping restarted with a lower rate. Thus the lower pumping and the rarity of rainfall in the 1970s, have maintained the level of the low table. These two factors contributed to drought wetland making think that these areas have become dry. Uncontrolled urbanization in towns of Pikine and Guediawaye was done anarchic way without respect of any construction standard. This has led to the return of the aquifer wastewater. Very high levels of nitrate in groundwater have trained to the decline in pumping rate and eventually stop pumping. The population is growing, the region of Dakar imported its water from the Guiers lake in addition to operating in the aquifer of infrabasaltiques sands (confined aquifer) located at the head of the Cap-Vert peninsula. These imported wastewater is an additional recharge for the aquifer. The decrease in the flow of rates, returns to the aquifer of wastewater; makes the most superficial aquifer. The low permeability of the sands as that the surface of the aquifer becomes closer to the ground. The recovery of rainfall since the 1995-1999 in a context of aquifer very close to the ground and urbanization in depression areas mean that floods have become recurrent in the Dakar region. The total cessation of drilling operations is a decision that may reinforce the scale of such disasters.
Director: Prof. Michel Jaboyedoff
Assistant: MSc. Antoine Guérin
The winter 2011-2012, relatively rich in snow, was very marked in the spring by the avalanches of wet snow especially on the Swiss part of the Portes du Soleil (Valais). The following winters also had a lot of heavy snow avalanches in the spring. The first findings of the zones exposed to failure were made during the winter 20I3-20I4.
ln order to characterize snowpack movements and identify areas of rupture, a series of LiDAR scans were carried out at various locations in the Crosets (Portes du Soleil – Valais) with the help of the domain trackers. For logistical and practical reasons, only the Vaillime sector was monitored thereafter. Several scans were carried out at regular intervals throughout the sector. During the non-consecutive 4 days of monitoring, 2 avalanches occurred and a creeping zone was identified. With georeferencing on a digital terrain model (DTM), it was possible to identify the slopes in which these events occurred. The volumes in motion could also be calculated.
ln addition to LiDAR investigation, the thermal imaging technique was used to detect warmer areas in the snowpack before a crack could open. Unfortunately, this was not possible, but the thermal imaging analysis still showed differences in snow temperatures between the shaded areas protected from rocks and open areas.
The last technique used during the same winter is Structure-from-Motion. A large number of photographs were taken in areas where an avalanche was possible. At first, the expected results were not sufficient, it is only during the following winter that new reconstruction tests were more relevant, even if the technique still needs some improvements.
During the winter 20I5-2016, that work was reoriented toward the study of snow melt. An
experimental table was built since it is neither possible to predict how the winter will be (e.g the number of wet snow avalanches) nor possible to control the necessary parameters continuously. lt was a tilting table and allowed the control of meltwater, climate data from the weather station (that collected data continuously) and the change in snow volume through the use of a LiDAR. ln parallel, an experiment took place to continuously measure the temperature of the snow through a probe coupled to a logger. The latter resulted in the loss of data due to a power failure. Coupling of the weather station, LiDAR and meltwater volume has been successful, although the system can be improved, e.g snowfall harvesting.
Director: Dr. Marc-Henri Derron
Expert: Dr. Clément Michoud
The coastal chalk cliffs in Normandy located in the North-West of France, is an area especially affected by the erosion. lt inducts a rapid retreat of coastline. Due to its important anthropization, this coastline is an essential area in risk management and land use planning.
Coastal erosion is influenced by many interrelated factors. Due to its complexity, the coastal erosion is still a relatively unknown phenomenon.
This work aims to study the modalities of the erosion of the coastal cliffs of Normandy usíng data acquired between 2012 and 2015, thanks to a LiDAR (Light Detection and Ranging) installed on a boat. This study aims to:
1) Define the limitations of acquired data
2) Locate and characterize the main detected collapses
3) Study the role of lithology in the dynamics of erosion
4) Study the role of human activities in the dynamics of erosion
5) Perform a diachronic analysis of a major landslide near Dieppe
5) To use the LIDAR intensity for detecting the lithology and the ground movements
7) Define the role of lithology and general orientation of the cliff on fracturing.
Director: Prof. Michel Jaboyedoff
This study aims to identify the risk of rockfall alongside a road on a regional scale. On first step, rockfall source areas was identified, using slope angle distribution (SAD) procedure and then it was cross checked with orthophotos and field observation. Later, three software l) CONEFALL (Jaboyedofl 2003) 2)Flow-R (Horton, et al., 2013) and 3) Rockyfor3D (Dorren, 2015) were used to investigate the maximum possible of runout distance of rockfalls. Except CONEFALL, the two-other softwares provide the reach probability of rockfall to the road, made it possible to analyze the rockfall risk alongside the road.
The study area is mainly covered by a forest, and the effect of the forest on rockfall propagation have been tested by help of Rockyfor3D. To define the characteristics of forest (DBH and forest density), field observation and a software called FINT(Donen, 2014) were used. The results of the study showed that forest has a big protective effect on rockfall propagation specially on big and medium size of rocks.
On the study area of Aigle, it was not clear to find a rockfall with a distinct source areas to calibrate rockfall propagation using Rockyfor3D, to solve this problem another similar site was chosen (Frenieres sur Bex) and the position of rocks based on their volume were marked by GPS, then calibration was done on this area and parameters were changed to conespond the observed rocks positions on the field. Parameters used for simulation in area of Frenieres sur Bex were used for rockfall propagation in Aigle area.
For validation of the results from Rockyfor3D, two other softwares were used 1) RockforNET (Berger & Dorren, 2007) and 2) RocFall (rocscience, 2017). RockforNET showed that the protection effect of forest can stop the rocks with 5 m3 volume, confirming the simulations results from Rockyfor3D, and RocFall results for maximum possible of runout distance were corresponding the results from simulations done by Rockyfor3D.
Risk analysis along road showed that small rocks cost more than the medium and big rocks per year, and risk of driving heavy cars is higher than driving motorcycles or normal cars on the road between Aigle and Le SéPey.
Director: Dr. Marc-Henri Derron
Experts: Prof. Michel Jaboyedoff, MSc. Antoine Guerin
In the field of risk analysis related to rock instabilities, temperature measurement can yield useful information in terms of stability. Indeed, temperature differentials can sometimes indicate elements that are invisible to the naked eye (eg presence of discontinuities, voids, fractured zones or exfoliation layers in certain geologies). Therefore, measuring large-scale temperatures could be beneficial in this area.
In this perspective, infrared thermography appears to be a particularly suitable tool. Indeed, the infrared thermography is a method of remote sensing that allows the instantaneous measurement of the temperature over an entire surface. This one results from the recording of a signal which is a function of the amount of infrared radiation received at the surface of the sensor. By transforming the latter into temperature using calibration curves specific to each thermal camera, this technique makes it possible to measure the surface temperature of an object remotely and rapidly.
However, this method is very sensitive to many parameters which, if not taken into account, can distort the measurement and cause artefacts. Therefore, a good understanding of the influencing parameters is necessary in order to bring the measured temperature closer to the actual temperature of the object under study. To do this, a good parameterization of the camera is necessary and this results from a good understanding of the phenomena that can influence the thermographic measurement. These may be related to 1) the nature of the surface studied (type of rock, alteration, roughness, oxidation) 2) the state of the environment in which the measurement is carried out (influence of the environment) and 3) The geometrical arrangement between the camera and the object studied (distance, angle of incidence).
The objective of this work is to: understand the theoretical notions specific to infrared thermography, assimilate the operation of the camera used and propose an adequate acquisition protocol, evaluate the impact of different influential parameters with the achievement of some experiments, propose a three-dimensional model of correction of apparent temperatures using a coupling between the thermal camera and a terrestrial laser scanner and finally propose a method to achieve thermal panoramas on a large scale thanks to the coupling between the thermal camera and a “GigaPan” panoramic head device.
Director: Prof. Michel Jaboyedoff
Expert: Dr. Marc-Henri Derron
Supervisors: MSc. Dario Carrea, Dr. Antonio Abellan
Landslides are complex phenomena. Despite the number of studies that have been conducted on this topic, they still remain unpredictable. It is particularly difficult to predict them because of the many factors that can impact the stability of a slope. This thesis will focus mainly on two of these factors: water and the slope inclination. Its goal is then to identify their impact on the characteristics of a landslide such as its geometry, its volume, its speed and its response time. For this purpose we have created an analogue model of slope in laboratory, made with sand of a diameter smaller than 2 millimeters. The level of water injected into the model has at first been varied whilst keeping the slope inclination steady, then we changed the slope inclination whilst maintaining a stable water level. The slope surface has been scanned thanks to the scanner Konica Minolta Vivid 9i with time intervals of one minute. For each experiment, the volume has been determined based on the failure surfaces visible through the transparent glasses that form the outside borders of the model. A MATLAB code has been developed to allow their upload from pictures of failure surfaces taken sideways as well as for the calculation of the landslide volume for each experiment. The MATLAB code developed by Carrea et al. (2012) has been used to follow the trajectories of the small balls set on the slope in order to calculate their speed based on LiDAR scans. As last, an analysis of the geometry and of the calculation of response time have been realised thanks to video recordings during the experiments.
The results show a significant impact of the water height on the extent of the landslide resulting in a Iinear increase of the landslide volume as well as its speed with the water height. As regards the geometry, a variation of the water implies little difference. Indeed, it has been observed that a landslide with a low water height is comparable to a landslide with a higher water height, but that it interrupts at an early stage, which implies that the final scarp is located further down the slope. Finally, the proximity of the toe of the slope from the water input has certainly prevented the determination of the real influence of the water height over the response time.
The change in slope inclination has shown major differences regarding to the geometry as the experiments with a weak slope inclination have mainly been characterised by a viscous gravity flow, whereas when the slope inclination is high, topples phenomena and rock falls show up. It has also been demonstrated that the landslide volume increases with the slope inclination. However, for the velocity, no trend has been identified due to the high uncertainty of the results. As response time strongly depends on the distance between the toe of the slope and the water input, and as this declines with the slope inclination in our experiments, it hasn’t been possible to highlight any relationship between these two variables.