An Assessment of Runout Models applied to Rock Fall Events in Norway
Johannes Salthaug
A thesis presented for the degree of Master of Science
60 Credits
Study programme: Geohazards Department of Geosciences
University of Oslo Norway
28.6.2017
1 Abstract
Runout models are an essential tool in evaluating the risk of infrastructure being impacted by mass movements. Rockfalls are a particular type of gravity driven mass movements (landslide).
They tend to involve the fall of small groups or volumes of rock blocks rather than rock avalanches that involve large volumes of rock (>100000m3 ). Rock falls inherently have a high degree of variability du to block size, shape, slope and terrain, and have proven difficult to model accurately. In this study, the usefulness of four different rock fall runout models will be addressed by comparing results from each model to previous rock fall events. These models include two 3D models (RAMMS, RockyFor3D) and two 2D models (Rocfall and CRSP). Given the large effect of small terrain features on a rock trajectory, a high accuracy Digital Elevation Model (DEM) acquired from a Light Detection and Ranging (LIDAR) survey is used along with detailed field observations to generate terrain models. These terrain models are used as the basis for all the runout models. The implications of the results could possibly affect the software choices of official agents involved in infrastructure maintenance and planning.
Contents
1 Abstract 2
2 Introduction 4
3 Background 5
3.1 Theory of rock falls and general runout models . . . 5
3.2 Rockfall software . . . 8
3.2.1 RAMMS Rockfall: . . . 8
3.2.2 RockyFor3d . . . 10
3.2.3 Rocfall . . . 10
3.2.4 CRSP . . . 10
4 Methods 11 4.1 Field data collection . . . 11
4.2 Generating terrain models . . . 11
4.3 Rockfall runout simulation tools . . . 12
5 Field Locality 16 5.1 General description and setting . . . 16
5.2 Geology of the study area . . . 17
5.3 Data collection at field sites . . . 24
5.3.1 Tveit B: . . . 24
5.3.2 Tveit B: . . . 27
5.3.3 Ruafjødd . . . 33
5.3.4 Kjønnesvikvatn . . . 36
6 Runout simulations 39 6.1 Simulation workflow and results . . . 39
6.1.1 RAMMS Rockfall . . . 39
6.1.2 Rocfall . . . 48
6.1.3 CRSP . . . 55
6.1.4 Rockyfor3D . . . 58
7 Discussion 65
8 Future work 68
9 Conclusions 69
2 Introduction
This thesis is focused on better understanding rock fall events in Norway. Due to the glaciation history of Norway, it has an abundance of steep sided valleys, and is prone to damage from rock falls and landslides. Prediction of likely rock fall runout has important implications for infrastructure, building and maintenance, as well as for designing appropriate protection measures. Rock fall simulation tools can now be used to test different runout scenarios and provide important insights for rockfall prediction. The different models have different strengths and weaknesses, and can be populated with input parameters derived directly from field observations. Here, I will apply a selection of runout model simulators (RAMMS, Rocfall, CRSP, Rockyfor3D) to specific rockfall events in Norway to assess how well they predict actual known runout as measured in the field.
3 Background
This chapter will introduce the theory of rockfalls, common runout models and key physical characteristics included in different rockfall simulation software.
3.1 Theory of rock falls and general runout models
Rockfalls are due to the fall of rock blocks under gravitational forces where free fall contributes a large part of the travel path. One key difference between rockfalls and rock avalanches is that for rockfalls the main concern is rigid body motion of individual blocks or small rock volumes and their interaction with the terrain. Whereas for rock avalanches, large volumes of hundreds of cubic meters of rock are moving and interacting(De Blasio 2011).
Figure 1: This photograph shows multiple boulders that fell from an overhanging cliff onto grassy terrain in Val Venina, northern Italy. Note the grooves on the terrain left by some of the sliding boulders and mud splash onto the house walls (From (De Blasio 2011). Observations and original photograph by Giovanni Crosta)
Rockfall runout, similar to other types of landslides, can be chacterised as the ratio of Fall height (H) to Runout length (L). This is sometimes referred to as Fahrboschung (FB = H/L) e.g. (De Blasio 2011).
Figure 2: This figure shows block trajectories for rockfalls in the Italian Alps. Detachment points shown in red, trajectories in block final runout points as black squares. The longest runout here has FB = 0.64. (taken from (De Blasio 2011) who modified from original figure Cancelli et al. 1991)
Factors such as the steepness of slope, the shape of the boulder and type of terrain significantly affect the trajectory, path and eventual runout distance of a rockfall.
Figure 3: Still frames from movie of block moving down a slope where (a) and (b) on hard rock. In (c) the rock impacts soft rock producing sliding (from (De Blasio 2011), original source Bozzolo and Pamini 1986)
Figure 4: Schematic diagram illustrating the possible behaviour of a block falling onto hard terrain. At small slope angles rolling dominates, whereas at steeper slope angles blocks bounce.
At very steep angles, blocks will fall freely (taken from (De Blasio 2011))
3.2 Rockfall software
3.2.1 RAMMS Rockfall:
Overview of the basic physics included in RAMMS runout modeling approach (From the RAMMS Rock Manual (Schweizer 2015)). RAMMS models rock falls in 3D, and outputs trajectories, runout length, jump height, kinetic energy, translational and rotational velocity.
The model is a rigid body contact model, and also accounts for real rock shapes. Both predefined types based on field measurements, but also the possibility of constructing a 3d model of the exact rock in question. All modes of movement of a rock fall, sliding, rolling and jumping are simulated deterministically. Rocks are considered indestructible, no fragmenting. There are four basic predefined rock shapes that can be used, namely; Equant, Irregular, Platy and Elongate.
RAMMS also considers block rotation in both airborne phase and during ground contact, while also implementing gyroscopic forces that acts on a rolling block, which is important as extreme runouts are oftentimes by rocks rolling in a wheel-like mode. This motion and the block’s ground contact is determined by using quaternion algebra to track the ground impacts of the rock and its phases of rotation.
Rock shape is modelled by using a point cloud converted to a convex hull polyhedral (rubber band stretched around points), and mass center is calculated with assumed even mass distribution. In free flight the rocks velocities of translation and rotation is stored in a vector, while the gravitational force, Fg, and D, drag force from trees, undergrowth and soil acts on the rock continuously. While contact forces are made up of both frictional contact (Fc) forces and plain contact forces (lambda). The rigid body contact is found by monitoring the vertical gap length between corners of the rock and the ground, and forces are calculated where rock and ground intersects. There is a permitted limit of non-physical penetration into the ground, which is purely to assess the contact condition. Several different contact forces can act simultaneously, and alter the rocks rotation and velocities, to make up for complex runout-behavior.
Friction forces are made up of two tangential forces, in x and y directions. These forces are assumed to follow Coulumb’s friction law. Stiction forces act as long as the tangential force is less than the effective friction coefficient (sliding does not occur). To calculate the friction in both sliding and rolling modes, a friction disc is projected onto the slope surface, and the friction coefficient is applied to this.
When the rock impacts the ground and rebounds, Impulsive normal forces work so that the rock does not penetrate the ground, but rather regains velocity. The rebound is governed by Newtonian impact laws, where the normal restitution coefficient describes the share of velocity regained post-impact, where sigmaN = 1 is a complete restitution of energy. This coefficient is generally set very low in RAMMS rock fall calculations.
There are two forces that counteract the rocks falling velocity that act differently. There is the contact friction, which acts on the points of the rock that touches ground, and as it acts on one side of the rock will induce rotational torque. It is important to parametrize the friction force correctly as it governs the mode of motion, rolling, sliding or bouncing. The other force is drag, which acts on the mass center of the rock, and therefore does not induce any rotation torque on the rock. The drag is from both the vegetation and the viscoplastic drag from terrain deformation.
In the case of a rock sliding on soft soil, the friction coefficient will in reality increase with the amount of material accumulated at the front of the falling rock. To model this in a hard-contact model, a slip-dependent friction that will act during sliding and accounts for the effect of the material accumulation. This is an extension of the Coulumb friction model where the friction value is made dependent on the slip distance travelled by the mass center.
Forest and vegetation is modelled as a drag layer which acts on the rocks mass center when it is located below the height of this layer. This drag force is linearly proportional to the velocity of the rock. There is therefore two parameters going into the vegetation modelling, the effective tree height (the height which could to some extent slow a rock down) and the drag coefficient, proportional to the density of the forest. Forest types are divided into open, medium and dense
forest.
3.2.2 RockyFor3d
Presented here is a brief overview of the basic physics implemented in the RockyFor3D model, from the RockyFor3D user manual (Dorren 2015).
The model simulates individual falling rocks without particle interaction. It further combines physically-based algorithms and probabilistic approaches. The rockfall trajectory is simulated as a vector in three dimensions, with the two of the three modes of motion enabled (parabolic free fall and rolling). To simulate rolling, a sequence of short rebounds is used. Sliding is not modelled in RockyFor3D (Dorren 2015). Rebound is also modelled against trees, where individual trees can be placed or distributed via a separate file.
3.2.3 Rocfall
Rocfall uses a lumped mass modelling approach, where the shape and size of the block is modelled by alteration of the other parameters internally. The mass, density and volume of the rocks can be defined in the designing of the simulation. Rocfall is more probabilistic in its approach than the very physically-based RAMMS software and RockyFor3D that combines the two approaches to simulation. RocFall has both projectile and sliding algorithms, where the velocity needs to be low for the particle (rock) to enter the sliding algorithm. As rocks are not considered to have any size in the simulations, mass is only used to calculate kinetic energy of the rocks. The projectile algorithm looks for the intersection point between the rock and the slope, and calculated the rebound and remaining velocity through the defined restitution parameters (Stevens 1998).
3.2.4 CRSP
CRSP is also a model employing the lumped mass approach. The manual and documentation of the model was not available as this section was written, so information available is limited.
4 Methods
4.1 Field data collection
Figure 5: Measured or quantifiable data collected during fieldwork. (Tveit 1 = Tveit A ; Tveit 2 = Tveit B)
Figure 6: Inferred data collected during fieldwork, not inherently objective. (Tveit 1 = Tveit A
; Tveit 2 = Tveit B)
4.2 Generating terrain models
Terrain has a strong influence on rebound and the amount of kinetic energy conserved after the rock impacts the ground. Terrain models (figure 7) are generated for each field locality based on direct observations of the main types of bedrock, soil and vegetation along the slope.
Figure 7: An example of a constructed terrain model, with a hillshaded DEM and terrain boundaries inferred from fieldwork and analysis of the DEM. Showing the drawn polygons over terrain types in the Tveit B location.
4.3 Rockfall runout simulation tools
The four simulation tools in the thesis are RAMMS, RockyFor3D, Rocfall and CRSP. The table 8 show the input parameters for each model.
Figure 8: Table showing the input parameters needed to run each rock fall model.
Two of the tools are 3D (RAMMS and RockyFor3D) while the other two are 2D models (Rocfall and CRSP). Table 9 shows the main features and advantages/disadvantages of each model. RAMMS has the interesting Rock builder feature allowing you to construct rocks of many different shapes that approximate those found in the field closely, see figure 10. Both Rocfall and CRSP have the possibility of using input parameters (eg. Tangential and normal restitution and friction angle) that can be directly measured in the field. All simulations are run in Rocfall in this thesis use slope material properties that have varying restitution values, but friction angle is generally set to 5 degrees (recommendation from co-supervisor Gunne H˚aland). Slope material properties are all normally distributed with a standard deviation of 0.05. Vertical starting velocity set to 0.5 m/s. For the simulations in CRSP input parameters are set to compare closely with these, while the recommended values may differ.
Figure 9: This table outlines the pros and cons of each of the four modeling softwares.
The procedure for inputing the needed parameters for each model along with the terrain model into each model goes through the use of Arcmap. In RAMMS the rock shapes are constructed in the rock builder feature, and the nearest match of rock type is used for each other software. The polygons over terrain types are used in both RAMMS and Rockyfor3D, while the different terrain types are defined within each model to be as similiar as possible. To go from 3D to 2D a profile is drawn from the release area perpendicular on the contour lines, and the length of the profile through each terrain type polygon is measured and used in the model. In Rocfall the resolution of the DEM creates too many vertices for the model to handle, and needs to be simplified through a built in tool that simplifies the slope while retaining the most important features. In CRSP each slope must be simplified further, to have less than the maximum of 100 vertices (or cells). This is also performed through the ”simplify slope” tool within Rocfall.
Typical outputs The standard outputs of each software vary, but all have the possibility of showing the position of the deposited rocks from the simulations. As a large part of the aim of the study was to see how well these softwares compare to the actual runouts, this output is the main focus of the results in this thesis.
Figure 10: This figure shows the constructed rocks from the RAMMS Rock Builder tool. Volumes and shapes are either measured or inferred from images of the rock.
5 Field Locality
5.1 General description and setting
There are many potential areas in Norway where it is possible to examine several rock fall events within a concentrated area, and for this thesis the field locations are situated in the area surrounding Setesdalen, in the county of Aust-Agder. This area has had several rock fall events the last few years, and the main supervisor, Jone Strømsv˚ag - from the Norwegian Public Roads Administration (Statens Vegvesen), was among the first geologists on site after the rock falls, and had useful prior knowledge about several of the rock falls.
Figure 11: The figure shows a map over the field area with all the four locations marked with flags and one image is shown for each location. Basic properties for each site is also incuded below the images. Map fromtopowms
In the top left corner of figure 11 is Ruafjødd which is a mountain located along the river Otra, which runs from the lake of Bykil 5 kilometers upstream from Ruafjødd, and as far south as Kristiansand. The mountain top rises 1058 meters above sea level. The rock fall started at
approximately 800 m.a.s.l. and ran down to a little below 400 meter elevation. The Runout length was measured to 534 meters, and with a fall height of 371 meters. No rocks impacted the road, but large boulders stopped against trees only a few tens of meters from the road. It was a high volume rockfall probably well above 1000 cubic meters. The rock fall started in the vertical wall above the rock fall path visible in the top left image in lighter gray.
Moving to the right top corner of figure 11 and Kjønnesvikvatnet, which is situated along FV 45, 15 km west of Bykle. It is in the south part of a small north-south stretching mountain belt of approximately 80 km length, which divides Setesdalen from the Telemark region further east. The lake of Kjønnesvikvatnet lies at an elevation of 800 m, which is much higher than the base of the three other sites, which are all situated along the Setesdalen valley. Although it is at a high elevation, the runout length and height of this rock fall was small, at 107 and 95 meters respectively. The rock fall volume was also low, at a few cubic meters, and as one single block. The rock fall started in the steep cliffs above the scree, as seen in the top right image (fig. 11), and it stopped directly at the shoulder of the road.
The location at Tveit located furthest west was named Tveit A and is seen in the bottom left of figure 11. This is the rock fall with the shortest runout length and fall height, of 89 and 46 meters respectively. The rock fall volume was approximately equivalent to that of the Kjønnesvikvatnet event.
The second Tveit location, named Tveit B, is seen in the lower right corner (fig. 11). It is located less than a kilometer east of Tveit A, but was a very different event. It had a rock fall volume equivalent to that of the Ruafjødd rock fall. It’s runout length an height also came close to that of Ruafjødd, with a runout length approx. 100 meters shorter, but a slightly higher total fall height. The rock fall started in the rock wall directly above the track of the rock fall seen in the lower right image of figure 11.
5.2 Geology of the study area
Regional Geology: Norway is a part of the Baltic shield, which is a flat bedrock area (craton) that stretches from the Kola Peninsula of Russia, through most of Scandinavia. In Norway the craton is intersected in the west part by the Caledonian mountain belt. In Southern Norway the Caledonides cuts through the craton in the central-western parts of Hardangervidda. In the remainder south east part of Norway the bedrock of the Baltic shield is exposed (fig 12). The bedrock in this part of the shield is of Proterozoic age and is predominantly granitic in type. This bedrock type covers the south of Norway from Stavanger to top of the Oslo-field and stretches further through the South-eastern parts of Sweden, as seen in figure 12 with orange color. The shield has a complex history of origin, and also became a part of the Rodinia supercontinent, which emerged approx. 1300 mya (Slagstad et al. 2013). During this period the Sveconorwegian orogeny occurred, from about 1050-1000 mya. This was caused by a collision between the
continents of Baltica, Laurentia and Amazonia (Slagstad et al. 2013). Rodinia broke apart in the period around 850 mya, and this caused widespread stretching and faulting. During the period following the break up of the supercontinent, the glaciation often called Varangeristiden or more popularly; ”Snowball earth”, followed. The bedrock of southern Norway has also been affected by an even older orogen, namely the gothic orogeny. This occurred approx 1,66-1,5 mya.
The caladonian orogeny occurred much later than the gothic and sveconorwegian orogenies, about 400-500 mya. This thrusted the caledonian nappe over the precambrian basement of southern Norway.
Figure 12: Simplified Geological map over the Baltic shield, with the large scale rock types indicated. Illustration from the book ”Making of a land” (Ramberg, Brynhi, and Arve 2006).
Local Geology: The study area in Setesdalen lies roughly parallell to one of Norways largest fault zones, the Mandal-Ustaoset fault-zone (Gabrielsen et al. 2002). The fault crosses Setesdalen in the small village of Austad (Ettner 1989) which is where two of the field locations in Tveit (A and B) is located. The fault zone curves eastwards north of the village of Valle, and also
intersects FV45 close to the field location of Kjønnesvikvatn. The fault was formed in the middle to late Precambrian period (Proterozoic) between 1500 and 1000 mya (Sigmond 2002). In the report from Bergverket on mineral exploration along the fault, Ettner (1989), the fault zone is described as a deep transcrustal fault. It was described to have deep ductile deformation on the southern part, and more brittle deformation further north. In the southern part of setesdalen including Tveit, the deformation was said to be represented by low angled thrust faults. In the section of the fault zone further north, which will include Kjønnesvikvatnet, the faults had a higher angle (Ettner 1989). The fault seems to have been active between 1509 and 540 mya (Ragnhildstveit and Tucker 1994). The fault has been interpreted to be a large-scale transverse shear zone that separates the Vest-Agder and Rogaland block from the Telemark - block (Haas, Andersen, and Vestin 1999). See figure 13.
Figure 13: A from a exploration report from bergverket Ettner (1989) with the Mandal-Ustaoset fault zone outlined with a orange dotted line
Faults are easily visible on the 1:250 000 geological map over Tveit from NGU (14), and the boundaries between the different lithologies here are complex. The three main lithologies in the area of the rock fall at Tveit are; 1. quartzite on the western higher-elevations of Tveitfjellet, 2.
granite and granodiorite on the eastern flank of the mountain and 3. amphibolite, hornblende- and mica dominated gneisses on the lower parts of Tveitfjellet.
Figure 14: This figure shows a geological map over the Tveit area, from the online map service from NGU, and the ”Berggrunn 1:250 000” map. The rock fall location is shown with an X.
The two main lithologies in the Ruafjødd locality is visible on the geological map on figure 15, and are the granite and granodiorite indicated by red color and arrow to the legend on the geological map (figure 15). The other type is the augen gneiss, granite and foliated granite, indicated by light pink on the aforementioned map and. This rock type is further described in the online mapservice of NGU (2017) as: ”Coarse granitic-granodioritic gneiss, migmatite with an older main mass of granitic rocks, often with augen-formation” (freely translated from norwegian by author). The latter rock type was most commonly found while on site, and most of the rock fall debris was of this lithology. This is expected, as the field area lies within this lithology on the geological map (N250 NGU). Other notable features of the lithology was very distinct fracture planes along the foliations of the gneiss. The foliation was determined to be accretions of mica within the gneiss, which provides a weak plane in the rock which is very prone to fracturing. A large amount of the rock fall debris had a rectangular shape, and seemed
to have fractured along these mica-dominated foliations. Also, there is much less of the faulting associated with the Mandal-Ustaoset fault zone found in Tveit. This corresponds well to the trace of the fault seen in figure 13, where the fault bend eastwards south of Ruafjødd.
Figure 15: This figure shows a geological map over the Ruafjødd area, from the online map service from NGU, and the ”Berggrunn 1:250 000” map.
Ruafjødd was the only locality where fracture measurement was practical to do, and also where the fractures were likely to have a direct link to the mode of initial movement of the rock mass. Fracture measurements are presented in the next chapter.
Kjønnesvikvatnet has four distinct lithologies associated with the closest proximity of the rock fall (see figure 16). The same granite and granodiorite rock group is present here also, in the lower part of Berhomsfjellet. The rock fall commenced and terminated within this lithology.
Above this on Berhomsfjellet is a meta-sandstone and mica schist rock group which have been intruded by near vertical basalt dikes. Further east is a slab of the rock group ”dioritic to granitic gneiss, migmatite”. Which indicates that it is a gneiss that originated as a feldspar rich
granite, or a mix of a granitic and dioritic rock which has undergone strong metamorphosis at a high temperature to develop the migmatitic facies of gneiss. Kjønnesvikvatnet seems to have much more influence from the Mandal-Ustaoset fault zone, most evidently the basalt intruded meta-sandstone on top of Berhomsfjellet see figure 16).
Figure 16: This figure shows a geological map over the Kjønnesvikvatn area, from the online map service from NGU, and the ”Berggrunn 1:250 000” map.
5.3 Data collection at field sites
Figure 17: Table showing the data used for calculating the fall height versus runout ratio for each site.
Figure 18: Diagram that plots the fall height versus runout distance for every rock fall.
5.3.1 Tveit B:
This field site lies dirctly south of the church at Austad. The rock fall started in in the lower part of Tveitfjellet, as seen in the images in figure 19.
The starting point of the rock fall was accessible by hiking, and the ledge where the block started rolling was easily recognizable. This was at 270 m.a.s.l and ca 50 vertical meters above the road below. The release area had a inclination of 35 degrees, while the slope was steeper closer to the road below. On average the slope had an average inclination of 20.6 degrees,
measuring from the starting point to the stopping location of the rock on the road. The ground consists of moraine and scree deposits. The matrix material in the moraine was soft and had a high water content during the fieldwork on the 14th of September 2016. There were three days with no precipitation before the 14th, but it was following a heavy rainfall on the 11th.
There was also a thick moss cover that holds on to the moisture for periods after rainfalls. The moraine was a mix of coarse large blocks, ranging to very fine matrix material. The poor sorting of the material likely leads to the ground to have poor drainage capabilities. In this case it is very likely that this nature of the soil contributed to the release of the rock, as it became oversaturated with water and the ledge of soft moraine which the block was resting on started to yield. Above the release area the ground was covered with larger angular blocks, while below it more rounded and smaller blocks dominated the surface. The percentage of blocks compared to fine matrix is also greater above the release area, seen in yellow on figure 21. The ground is softer near the bottom of the hill, as there is more fine moraine material in relation to large rock fragments, figure 21. This was visible as the falling rock was able to cut deeper into the ground further downslope. This is also be in part caused by the larger velocity of the falling rock, but the ground did overall seem to be softer downslope. The vegetation is evenly distributed, and mainly consists of tightly spaced birch trees with stem diameters of 5-15 cm.
Figure 19: Two pictures from fieldwork at Tveit A. The picture to the right is taken from the church, while the left is taken from the top of the runout track.
Figure 20: This image from fieldwork shows the starting location at Tveit 1.
Figure 21: Showing the drawn polygons over terrain types in the Tveit A location.
5.3.2 Tveit B:
This rock fall location is located approximately one kilometer east from Tveit A along RV9. The rockfall occurred south of RV9 as for Tveit A, and the slope is part of a continuous mountain side. The rock fall started 520 m.a.s.l. in the slope in a steep cliff section, where a large volume of rock likely toppled from. The brighter color of the freshly exposed rock face reveals the starting point.
Figure 22: Images A. and B. are taken mid-slope in the track of the rock fall in Tveit B, and B.
shows the starting area of the rock fall as a lighter colored rock face. Image A. show the track of the rock fall and the forest and road below it.
Figure 23: This image shows the rock that travelled furthest at the Tveit B location.
Figure 24: An example of georeferenced points to be used for Structure from Motion photogram- metry from the Tveit B location.
The road below is situated at an elevation of 220 m.a.s.l, while the furthest-traveling rock stopped at 227 m.a.s.l and 360 meters in horizontal distance from the starting area. Which corresponds to 132 meters in straight-line distance from the road. Furthermore, the average inclination of the slope from the starting area to the road (going through the furthest traveling rock) is 27.8 degrees. Note that the distance from the furthest runout to the road has a very close to horizontal inclination. The average inclination to this point is, in contrast 36.6 degrees.
The site has a near vertical cliff at the top of the slope, which is where the release area is found.
The release area is situated a 2-5 meters up in the cliff face, where a large block toppled from.
The slope below has a thin soil cover, with hard bedrock below. The bedrock is cut into sharp ledges, and has an uneven surface. Further downslope there is a thicker soil cover, and more scree dominates the soil. Where the slope flattens out, the ground is again even softer, but still with a portion of scree material. Further north, the road is protected by a large dike on both sides which is 2-3 meters high and 10-15 meters wide. The terrain material is mainly bedrock with a thin to medium soil cover, where the soil cover is thicker further downslope. The soil is also high in block/gravel content. The ground is in generally harder here than further west in Tveit A. On the basis of these observations it seems sensible to use a terrain type close to bedrock hardness on the top part, as in seen in green on figure 25. The lower section seen in purple (Tveittwoscree), consists of a thicker layer of soil, or scree deposits from the actual rockfall and/or previous smaller rockfalls. Outside of the rock fall path, the area can still be described as softer than above, and blocky in nature. Below this is ground which is even softer, and starts abruptly where the slope slacks off completely. This is still quite compact soil, with parts of it having thick moss and water saturated soil. There is also a forest road winding its way close to the slope of Tveitfjellet. North of this again is the RV9 road, and another patch of forested area with soft soil (no color) south of the river , Otra in blue (figure 25).
Figure 25: Showing the drawn polygons over terrain types in the Tveit B location. The coverage of the lidar survey from which the hillshade map derives stops about 50 meters north of the outlined working area. The release area is small, at just 8,4 meters across along its diagonal and is therefore not visible in the map.
5.3.3 Ruafjødd
Figure 26: A pretty overview image from the middle of the runout path at Ruafjødd.
The Ruafjødd locality is situated 35 kilometers further NNW upvalley from the church at Tveit across the road from the Tveit A work area. The mountain at which the rock fall occurred is colled Ruafjødd, and its peak is at 1055 m.a.s.l. The very top of the release area is situated at 800 m.a.s.l, while the furthest travelling block came to a halt at 380 m.a.s.l, while the road lies at 372 m.a.s.l. Overall, this is the locality with the largest volume rockfall, that also came quite close to the road (about 20 meters horisontal distance). In terms of steepness, it is the steepest of the four locations, with a inclination from top of the staring area to the closest part of the road section at 38.5 degrees. Given that there is still 200 vertical meters to the top in less than 100 meters horisontally, the inclination could have been even more dramatic if the release area was even higher. In terms of terrain materials it consists predominantly of bedrock, partially covered with scree and/or soil in certain parts. The trend is also here towards softer materials nearer the base of the slope. The top section (green color in figure 29) has mostly been defined as bedrock or partially scree covered bedrock in most final simulations. Below is the area marked with yellow color in figure 29 which has a thick and coarse layer of scree material, with an abundance of rocks larger than 2 meters along their longest axis. Below, in pink color is a softer ground with a higher percentage of soil overall, and likely moraine material below the new scree deposits and soil in the lower parts. This is based on the observation of more rounded blocks in the soil matrix. The road is marked in grey color and black outlines between the river and the pink terrain polygon. The vegetation on the site consists of large but widely spaced spruce and birch trees, as well as in places dense undergrowth of young trees and bushes.
Figure 27: Three images from fieldwork at Ruafjødd. A. shows an overview photo of the rock fall from a distance, B is an image taken upwards mid-slope. While C is an image of the farthest traveling rock.
Figure 28: An image showing the marking of a rock and georeferencing on a handheld GPS device, from the Ruafjødd location.
Figure 29: Showing the terrain material boundaries used for Ruafjødd. Polygons overlay a hillshaded relief map derived from SVV’s lidar survey.
Due to good accessibility on this site, measurements of fractures were possible and carried out here during fieldwork. The stereoplot with these fracture measurements is shown in figure 30. There was a fracture set which were near vertical, that goes parallel to the cliff face with a dip-dip of (85/180). There are also two low angled fracture sets that intersects the vertical fracture set, and cut the sheets of rock formed by the vertical set into slabs that are prone to topple. The kinematic analysis shows exactly this, where toppling is predicted in the southern direction (figure 30).
Figure 30: This figure shows a stereoplot and a kinematic analysis done on the fracture sets.
Stereoplot and analysis was performed in Dips by Rocscience.
5.3.4 Kjønnesvikvatn
Kjønnesvikvatn is a lake situated at an elevation of 800 m.a.s.l and 15 km NW of Ruafjødd, along RV 45 going from Setesdalen to Dalen in Tokke county. The hillside from which the rock fall occurred was a part of Berehomsfjellet, which has its peak at 1223 m.a.s.l. The rock fall started at 924 meters elevation, and traveled right down to the ditch on the side of the road at 818 meters elevation. The slopes inclination is at 21.2 degrees from the starting point to the roads closest point. Meanwhile, the terrain material is predomainantly scree cover with some exposed areas of clean bedrock. Also here, there is more scree material further downslope.
therefore, the slope is divided into two separate terrain types. The harder scree-bedrock area in orange on figure 32, and the softer, more scree dominated type in green below. There is also the road and lake as separate areas. There is little vegetation, with the exception of a few scattered bushes and small trees. None of which will make any difference to a falling rock of any size.
Figure 31: Three images from the rock fall at Kjønnesvikvatn. Image A. is taken by the author, while credits for B and C goes to Jone Strømsv˚ag at SVV. A shows an overview image taken from across the alke, B shows the rock in-situ after the rock fall, while C. indicates the base of the release area.
Figure 32: Showing the terrain material boundaries for Kjønnesvikvatn. Polygons overlay a hillshaded relief map derived from SVV’s lidar survey.
6 Runout simulations
In this section the results from simulations will be presented location by location, and results from all four softwares will be shown for each location. The position of the actual furthest runout block measured in the field, will be marked on the simulation outputs for comparison.
6.1 Simulation workflow and results
Figure 33: The table indicates the various parameters used for the simulations shown below.
Values are listed from top to bottom of the terrain material sections shown in figures showing the drawn terrain material polygons. (Tveit 1 = Tveit A ; Tveit 2 = Tveit B)
6.1.1 RAMMS Rockfall
All four rock fall locations have been simulated using RAMMS, with several different input settings. The overall results from the simulations with the most sensible settings derived from field observations and the RAMMS manual showed very sensible results for both Tveit locations, and Kjønnesvikvatn. Perhaps with a slight underestimation of runout lengths compared to the real world event in both Tveit locations. In contrast both Kjønnesvikvatn and Ruafjødd generally seems to show longer runout lengths overall than the real event, even if a number of
rocks do stop at approximately the same runout length, a large percentage do travel significantly farther. This is true for both Ruafjødd and Kjønnesvikvatn. At both locations there is a water body directly on the other side of the road that crosses the runout path. This very effectively stops the rocks, as the water body is also put into the model calculation. Still it is evident that the rocks entering the water body still has much kinetic energy left. This will be shown in further detail in the results section. Runout length alone compared to the actual runout distance is not sufficient because of the effect of the water body in the simulations. To bypass the problem runout length in combination with rock velocity at intersection with the road could be shown. The most frequently altered settings is the terrain types defined through polygon shapefiles and a choice of 7 predefined hardnesses which correspond to certain terrains, which is defined in the user manual for RAMMS. There are also polygons going into the model over forested area and lake/moors. Forest is defined as three types, either open, medium or dense forest. All three were tested for Ruafjødd, and here it was found that it did not alter the results greatly, but here a large number of rocks had a lot of energy even when crossing the road. It was found that rocks had somewhat lower kinetic energy when crossing the road with dense tree area, as would be expected. Forested areas go into the model as a friction coefficient that acts as drag on the rock as long as it is within the forested area defined, and within the tree height defined by the model. When rocks have high energies as in Ruafjødd and large rock size, trees are not expected to have a large impact on the runout length. That would likely be different on a smaller slope like Tveit number one, but even here the rock is large enough to likely plough through anyway. What was found here was that dense forest gave approx. 10 meters shorter runout distance than open forest in RAMMS, of a total of 110 m for the longest runout in the two simulations. Other settings were kept identical for that comparison.
More important perhaps than forested area is the terrain types. These were inferred in field and their extent set by GPS-points collected in the field. The polygons were later drawn in arc-gis. Each location has a simulation with the input settings closest to what was observed in the field, and several others were done to find the models sensitivity to other inputs. The first run location was Ruafjødd, where the most inputs were tested, both because it is a very interesting site in terms of rock fall dynamics, and just because it was first in line and my scientific curiosity could roam free. It is probably the largest one of the rock falls in terms of sheer volume, its only competitor being Tveit B. It also had a sharp transition in steepness, which collected a lot of the rock debris. This was smoothed out by the debris itself, and subsequently in the lidar-derived DEM used for modelling. As I do not have the means of accurately restoring the terrain model to its original state before the rock fall, the model will run on a terrain that is significantly different than what the real life rock fall ran across. The same is true for the tree input polygon, which is drawn as closely as possible to where there was likely to be trees also before the rock fall. The latter is not likely to have a large implication in
the simulation, as it seems like small changes in the tree friction layer is not the most important parameter as the rocks used are large and friction forces from vegetation are not very significant in comparison to the total kinetic energy of the rock in these simulations. The rocks used in the RAMMS simulations are constructed in the Rock Builder tool integrated in RAMMS, on the basis of observations in the field. Rock sizes used range from 2-12 cubic meters. The smallest for Kjønnesvikvatn, and largest for Ruafjødd.
Another few things that were tested was with using half the rock size, and using a lower and higher part of the release area in Ruafjødd for separate simulations. Both of these cases gave small changes to the overall results in this location.
Tveit A Results from the Tveit A location are presented below.
Figure 34: This table shows the most important inputs and the associated results from each run on the Tveit 1 location. (Tveit 1 = Tveit A ; Tveit 2 = Tveit B)
Figure 35: A hillshaded relief map over the Tveit B locality, showing the simulated deposited rocks from the RAMMS simulations, as well as the marked points from fieldwork. These include the furthest travelling block from the rock fall event.
Tveit B Results from the Tveit B location are presented below.
Figure 36: Shows the stopping points of the simulated rockfall by RAMMS and with identical inputs except for the DEM resolution. DEM with 2 m resolution shown in blue, while 0.5 meter resolution is portrayed in blue colors. The furthest travelling rock of the actual rockfall is shown by a green square. (Tveit 1 = Tveit A ; Tveit 2 = Tveit B)
Figure 37: This table shows the most important inputs and the associated results from each run on the Tveit B location. (Tveit 1 = Tveit A ; Tveit 2 = Tveit B)
Figure 38: A hillshaded relief map over the Tveit B locality, showing the simulated deposited rocks from the RAMMS simulations, as well as the marked points from fieldwork. These include the furthest travelling block from the rock fall event. (Tveit 1 = Tveit A ; Tveit 2 = Tveit B)
Ruafjødd Results from the Ruafjødd location are presented below.
Figure 39: This table shows the most important inputs and the associated results from each run on the Ruafjødd location.
Figure 40: A hillshaded relief map over the Ruafjødd locality showing the simulated deposited rocks from the RAMMS simulations, as well as the marked points from fieldwork. These include the furthest travelling block from the rock fall event.
Kjønnesvikvatn Results from the Kjønnesvikvatn location are presented below.
Figure 41: This table shows the most important inputs and the associated results from each run on the Kjønnesvikvatn location.
Figure 42: A hillshaded relief map over the Kjønnesvikvatn locality, showing the simulated deposited rocks from the RAMMS simulations, as well as the marked points from fieldwork.
These include the furthest travelling block from the rock fall event.
6.1.2 Rocfall
Arcmap was a frequently used tool in setting up the simulations in Rocfall. As RAMMS was the first simulation software that was run, many of the needed steps in Arcmap were already completed. The terrain polygons and release areas/points are already drawn. All that needs to be done is to make a 2D representation of the slope. To accomplish that there is a useful tool in Arcmap that derives the steepest path downslope from a point. This would have been more
useful on a DEM of lower resolution, but when applied to a 0,5-meter resolution DEM, the path stops every time the next cell has a slightly higher elevation. Instead of using this tool, the path was drawn manually perpendicular to contour lines derived from the DEM and in accordance with the rock fall paths from the RAMMS simulations. The drawn path is used to make a profile with elevation values retrieved from the DEM, which is exported as an ASCII text file to be used in Rocfall. The only change needed for Rocfall to read it, is to replace all of the files commas with dots. Further, the file is imported and viewed as a 2d slope in Rocfall. Now the slope is viewed with very high detail, although the software is not designed to handle such a high resolution in its calculations. To fix this issue there is a tool in Rocfall that simplifies the slope according to the user’s specifications. In this case it was desirable to keep a high resolution as possible, and the slope was simplified by removing vertexes that were less than 0.01 meters from an automatically simplified slope-polyline. After the bare bone slope profile is done, the terrain materials are defined. This is done by identifying the elevation of the points where the profile path crosses the terrain polygons used in RAMMS. these elevations are used to define boundaries. Over to the actual material properties, the parameters that define the material is restitution values of both the tangential and normal components, and the friction angle of the material. In the preliminary simulations, standard values found in a report from SVV “Sikring av veier mot steinskred” for the closest match of material type found during fieldwork. The table in the report only describes values of restitution, not friction angles.
Other notable preparations include placing a seeder from which the rocks will be released.
The seeder can either be a line or a point. For the largest rock falls, namely Ruafjødd and Tveit 2, a line seeder was used. This was done as the source area of the rock fall is fairly large, and has a vertical distribution that does likely affect the final runout distance. They also had very steep release areas. This will also make the rock fall easily start moving in the software simulations, as the rocks quickly gain velocity. For this reason, the seeder was placed directly on the slope profile. For the other two locations, the seeders were defined as points. This is both because the release areas were small and easy to define, and as it is easier to define exact coordinates.
For both of these, the seeder was placed 0,5 meters above the slope profile, and was released with a vertical velocity of 0,5 m/s. The model used here is the lumped mass model, which does not take rock shape into account. There is a rigid body model also included in the Rocfall 5.0 software, but this has not been used in these simulations. Rock shape has not been included, but the mass of each rock estimated through the rock builder in RAMMS and field observations was used, as well as a density of 2700 kg/m3. In the project settings, velocity scaling was used on the restitution parameters. As Rocfall is based heavily on statistics, there are options to use several different statistical distributions on most input parameters. In these calculations the two restitution parameters are normally distributed, while all other inputs are kept at constant values.
Tveit A Results from the Tveit A location will be presented below.
Figure 43: Above is the slope profile constructed in Rocfall for the Tveit 1 locality, with the different slope materials indicated as well as the simulated rock paths in purple. Along the axis are vertical and horisontal distance in meters.
Figure 44: This plot shows the endpoint location of the simulated rocks in Rocfall for the Tveit A locality with horisontal position along the X-axis in meters, and the number of rocks and vertical position along the Y-axis.
Tveit B Results from the Tveit B location will be presented below.
Figure 45: Above is the slope profile constructed in Rocfall for the Tveit 2 locality, with the different slope materials indicated as well as the simulated rock paths in purple. Along the axis are vertical and horisontal distance in meters.
Figure 46: This plot shows the endpoint location of the simulated rocks in Rocfall for the Tveit B locality with horisontal position along the X-axis in meters, and the number of rocks and vertical position along the Y-axis.
Ruafjødd Results from the Ruafjødd location will be presented below.
Figure 47: Above is the slope profile constructed in Rocfall for the Ruafjødd locality, with the different slope materials indicated as well as the simulated rock paths in purple. Along the axis are vertical and horisontal distance in meters.
Figure 48: This plot shows the endpoint location of the simulated rocks in Rocfall for the Ruafjødd locality, with horisontal position along the X-axis in meters, and the number of rocks and vertical position along the Y-axis.
Kjønnesvikvatn Results from the Kjønnesvikvatn location will be presented below.
Figure 49: Above is the slope profile constructed in Rocfall for the Kjønnesvi locality, with the different slope materials indicated as well as the simulated rock paths in purple. Along the axis are vertical and horisontal distance in meters.
Figure 50: This plot shows the endpoint location of the simulated rocks in Rocfall for the Kjønnesvik locality, with horisontal position along the X-axis in meters, and the number of rocks and vertical position along the Y-axis.
Key observations from the Rocfall simulations shown below:
• Long runout distances in Ruafjødd and Tveit B, comparable or slightly longer than the event. Within reasonable distances. For Ruafjødd 300 rocks stopped between 450 and 470 meters. While for Tveit 2 less than 250 passed the 400 meter mark. These locations are nearly exactly where the longest travelling rocks stopped in the real event.
• Short runout distances in Tveit 1 and Kjønnesvikvatn, especially for Tveit.
This is also the slope with the definitively slackest slope angle. Here, the runout distance was barely 15 meters. A significant number of rocks reached 80 meters in Kjønnesvikvatn. For both of these locations the actual rock traveled further than 110 meters in the profile.
For the first set of simulations, the default friction angle of 30 degrees was used. The angle decides the point of which a steeper angle will make the rock roll/slide faster while a slacker angle will slow it down. This is in reality both dependent on the rock shape and the terrain.
As the first set of simulations showed very short runout distances for Tveit 1 and also for Kjonnesvikvatn, the friction angles were lowered to 5 degrees for all materials on the slope. It can be argued that this is reasonable on the basis of the field observations. For Tveit 1, the rock itself was a large spherical rock of about 2m in diameter. In addition it was rolling in a fully water saturated moraine material littered with boulders. It would likely give very favorable rolling conditions. For Kjønnesvikvatn the rock was disc shaped, and was approx. 2 m3 in volume. This would either give it very good rolling properties as a wheel, or likely slide well on
hard surfaces. Based on this, it is my opinion that a low friction angle is reasonable for both locations.
This gave very different results for both locations:
• Tveit 1 now had all 1000 rocks stopping at 65 meters. This seems artificial, and perhaps indicates that the newly defined friction angle of 5 degrees has a much larger impact than all other parameters, as friction angle is not normally distributed.
• Kjønnesvikvatn showed an even larger change of runouts with the newly defined friction angle. This is likely caused by the much steeper slope of this location. Here all rocks stopped at and within a couple of meters of 120 meters.
6.1.3 CRSP
After completing simulations on the four locations in Rocfall, the included features in the Rocfall software of simplifying the slope profile was used to make the slope profile usable in CRSP.
CRSP is an old software, first developed in 1988, with the latest update made in 1997. Naturally a 20-year-old software will not be optimized for large, detailed datasets. the resolution of the profile used in Rocfall is far to detailed to be imported to CRSP. The maximum number of cells to be used in a single slope is 99. The “simplify slope” tool seems to do a good job of maintaining the most important features of the original slope, while reducing the number of vertices to any desired number. For our case, the slopes have vastly different lengths, ranging between 100 to more than 600 meters long. This will lead to the small slopes being more detailed, while the large slopes will be smoother. This will likely have impacts on the resulting runout lengths. Likely a bias towards longer runouts on large slopes. To get the desired number of cells, which will be as close to 99 as possible for all slopes, the new slope was defined either as a polyline where vertices closer than a set distance from the polyline will be removed. This was used on Kjønnesvikvatn and Tveit 1. On the large slopes, only a defined percentage of the original vertices were kept, this was down to 15 percent on Tveit 2. The remainder of the required input file consists of a column, following the slope profile, with surface roughness, tangential restitution and normal restitution. These parameters are defined through a number of predefined material types linked to associated values of tangential and normal restitution.
The ranges of values for each material type is wide, and deciding on a representative value can be difficult without extensive calibration and field efforts. The surface roughness also has a table with typical material types and ranges of values. Rocfall has values associated with materials, not ranges of values. That makes the process of deciding on material properties simpler, although this may be an oversimplification. In addition to this, a line at the top of the text file, which defines the unit type (Metric) along with three analysis points and the top and bottom of the starting zone. Other required inputs are made in the GUI of the program. This
includes rock density (set to 2700 for all simulations), rock shape (circular) and diameter. As well as number of rocks used for the simulations, which was generally set to 1000.
Tveit A Results from the Tveit A location will be presented below.
Figure 51: This histogram displays the number of rocks that stopped within each defined cell in the CRSP simulation for the Tveit 1 locality
Tveit B Results from the Tveit B location will be presented below.
Figure 52: This histogram displays the number of rocks that stopped within each defined cell in the CRSP simulation for the Tveit 2 locality
Ruafjødd Results from the Ruafjødd location will be presented below.
Figure 53: This histogram displays the number of rocks that stopped within each defined cell in the CRSP simulation for the Ruafjødd locality
Kjønnesvikvatn Results from the Kjønnesvikvatn location will be presented below.
Figure 54: This histogram displays the number of rocks that stopped within each defined cell in the CRSP simulation for the Kjønnesvikvatn locality
6.1.4 Rockyfor3D
Working with Rockyfor3D is somewhat different from the other softwares, as most input parameters and must be imported as individual GRID files that must be constructed in a separate GIS-software. In total, a minimum of 10 individual input grids are required for a single simulation scenario in RF3D. These include the elevation model (DEM), surface material, surface roughness (rg10, rg20, rg70), starting cells, rock density, rock shape, rock dimensions (d1, d2, d3) There are several different possible GIS tools available to use for the purpose of exporting input grids to Rockyfor3d. In the RF3D manual (Dorren 2015) there is no specific recommendation on GIS software, the only requirement is that it is able to export grids in ESRI ASCII GRID format. This is true for most available softwares, including QGIS, SAGA-gis and ArcMap. In ArcMap there is the possibility of using third party toolboxes, and a specific toolbox for exporting input grids for Rockyfor3D has been developed by the Risk Analysis Group of the University of Lausanne. The toolbox is named PimpMyRockyfor, and is freely available as a public domain software under a beerware licence, meaning that no official support or updates are provided by the author (2014). As this toolbox was publicly available, and specifically designed for the job, the preferred GIS software for this thesis was ArcMap. The use of this toolbox is fairly straightforward once understood, and will be explained to some detail below. Firstly, a polygon over the entire workarea must be defined and assigned to a shapefile that is supplied with the folder containing the toolbox ”PimpMyRockyfor”, and the projection
of the shapefile must also be defined. In this work the ETRS 1989 UTM Zone 32N projection was the one assigned to all map elements. This polygon should further be split into sections that have different characteristics, using the ”cut polygon” tool in ArcMap. Polygons must be defined over each separate terrain type to be used in the simulations and release area. After all polygons are defined, the parameters for each are to be set in the attribute table. There are 10 columns in the attribute table, each corresponding to one exportable ASCII GRID file. The appropriate values for each attribute should be determined on the basis of field data or prior knowledge about the area, and in correspondence with the manual for the software (Dorren 2015). All rock attributes go under the same polygon, the release area. These include rock density (set to 2700 kg/m3), rock dimensions (d1, d2 and d3) and block shape (1 - 4). Block shape is defined as either rectangular (2), ellipsoidal (3) or disc shaped (4) s well as from the block dimensions entered. Meanwhile, the slope characteristics are defined through the other five attributes. Three of them define the slope roughness, as probability classes (rg70, rg20, rg10). The number after -rg corresponds to the probability of the block to encounter the specific obstacle height that is defined in the attribute table under each of the three probability classes.
An example is 0.5m entered under the rg70 attribute, which would imply that the block falling in any cell within the area has a 70 percent chance of hitting an obstacle of this height. This of course is the case for the two other probability classes also (Dorren 2015). The final attribute is the soil type, which has 8 different classes associated with it (0-7), see figure 55. As apparent from the figure 55, the type 0 represents a water body or a swamp, type 1 a fine soil material deeper than 1 meter, while type 2 is a fine soil shallower than 1 meter or deep and mixed with sand/gravel. Type 3 is scree with material finer than 10 cm in diameter, or a compact soil with small rock fragments which also could describe a forest road. Type 4 is used for talus slopes with material coarser than 10 cm in diameter, or a compact soil with large rock fragments.
While type 5 represents bedrock with a thin layer of weathered material or soil and 6 represents pure bedrock. Type 7 is used for asphalt roads. In addition to the soil types and roughness, it is also possible to run simulations with vegetation. This can be accomplished in one out of two ways. It is possible to construct a treefile.txt by using the FINT software, designed to identify single trees from a lidar derived terrain dataset. It used a DEM and a DTM constructed from the first returns from the lidar survey to identify each tree. This would be a good option to use for hazard evaluation purposes, but in this case the rock falls are in two of the cases very large and has demolished all the forest in its path. therefore this approach would be nonsensical to use, as there would be no trees in the most likely path of the simulated rockfalls.
The other approach is to define polygons in the work area that has forest with a specific dbh (breast height diameter), height and tree density (nr. stems per hactare), as well as a percentage value of coniferous trees and standard deviation of the DBH. This is done in much of the same way as the previously described approach for the slope characteristics by defining the
aforementioned set of attributes to drawn polygons in the work area. This process is described in more detail in the manual for Rockyfor3D (Dorren 2015).
To export the needed grid files for RF3D, a high resolution DEM is required. The stated ideal spatial resolution by the manual is 2 meters (Dorren 2015). As the spatial resolution of the available DEMs are higher, 0.5 meters, this resolution is what is used for both RAMMS and RF3D. This is done to have the results more comparable between the softwares, and test the capabilities of the softwares to use an even higher spatial resolution than proposed. Simulations have been run with a lower spatial resolution of 2 meters to test for sensitivities of the software to spatial resolution.
Figure 55: Table over all the 8 different soil types available in Rockyfor3D, retrieved from (Dorren 2015).
Tveit A Results from the Tveit A location will be presented below.
Figure 56: A hillshaded relief map over the Tveit A locality, showing the simulated deposited rocks from the RockyFor3D simulations, as well as the marked points from fieldwork. These include the furthest travelling block from the rock fall event.
Tveit B Results from the Tveit B location will be presented below.
Figure 57: A hillshaded relief map over the Tveit B locality, showing the simulated deposited rocks from the RockyFor3D simulations, as well as the marked points from fieldwork. These include the furthest travelling block from the rock fall event.
Ruafjødd Results from the Ruafjødd location will be presented below.
Figure 58: A hillshaded relief map over the Ruafjødd locality from the ROckyFor3d simulations showing the simulated deposited rocks from the Rockyfor3D simulations, as well as the marked points from fieldwork. These include the furthest travelling block from the rock fall event.
Kjønnesvikvatn Results from the Kjønnesvikvatn location will be presented below.
Figure 59: A hillshaded relief map over the Kjønnesvikvatn locality showing the simulated deposited rocks from the RockyFor3d simulations, as well as the marked points from fieldwork.
These include the furthest travelling block from the rock fall event.
7 Discussion
In this section the results from the simulations will be discussed and critically evaluated to assess the relative strengths and weaknesses of the different runout models.
During the early stages of the thesis work, many simulations were run with different input scenarios with forest in RAMMS. This had little effect, perhaps since both the actual rockfalls and the simulated rockfall had large volumes. Volume is thought to be an important factor in determining the effect of the friction layer, both in simulations and likely the actual rock falls (A Høydal and Sandersen 2013). This was not completed in the other softwares, as some do not
have the possibility of forest input, but merely by altering ground properties.
Results from the two 3D based softwares overall shows longer runout distances than the resulting block distribution from the actual rock fall events. This proves true in all simulations where the majority of simulated blocks gained sufficient kinetic energy during the release of blocks from the starting area, see figure 56. As mentioned, many more simulations were run per site, with varying terrain material properties. Most of these also shows overestimation of runout distances compared to the real events. See figures 39, 37, 34 and 41 for other RAMMS simulations.
In some scenarios the simulations stopped short, this is particularly the RockyFor3D simulations at Tveit B and Tveit A. For Tveit B, the input parameters are probably the cause of this along with the higher than recommended resolution of the terrain model. For Tveit A the slope did not have a steep enough inclination, and/ or the drop height of the rocks were not great enough for the rocks to start moving in the simulations.
Results from RAMMS at the Tveit 1 locations is the only result where simulated runout stops before or at the point of the furthest travelling block found during fieldwork. The figure shows an X at the final position of the rock, in the bus-stop at the side of the road 35. The reason for this simulation stopping short, is likely because of the slack inclination of the slope.
The slope is a moraine block field, and consists of rounded boulders. A heavy rainfall caused the soil beneath the block to become to soft to support the weight of the block, and is probably the trigger of the rock fall. This is not a typical rock fall site, with no sections steep enough for a normal rock fall to be triggered, where a block would topple or slide of a cliff face. Because of the untypical release mechanism of this rock fall, simulation results were not expected to be realistic with reasonable input parameters. RAMMS has the feature of automatically calculating the minimum necessary release height for the rocks to start moving downslope from the starting point. Release point shown in figure 34. RockyFor3D does not have this feature, and a high enough release height was not selected manually for the rockfall to commence, figure 56.
For the Ruafjødd rock fall site, a large percentage of rocks (up to 50 percent) stopped in the water body below the road for the RAMMS simulation, with significant kinetic energy left 40.
The three flags closest to the road are deposited blocks from the rock fall. No blocks intersected
the road in reality, but one bock stopped within 20 meters of the road. However, the main debris from the rock fall stopped further than 100 meters from the road. RockyFor3D had an even larger amount of rocks stopping in the water body (approx 75 percent) 58, overestimating the runout distance even further. This slide had large volume and was running out close to the road, and was perhaps the most significant hazard studied.
For the Kjønnesvikvatn locality and 3D simulations, see figures 42 and 59. In both simulations a significant amount of blocks stopped mid-slope, while a large number also crossed the road and entered the water body. Rocks stopping mid slope is likely because of the relatively small block used in the simulations, and the jagged, blocky nature of the slope. The highest available resolution of 0.5 meters was used for both simulations, also contributing to the early stopping of blocks because of roughness included in the terrain model.
At the Tveit B location, figure 57, the majority of blocks stops during the first 100 meters of the path. Where there is pure bedrock, and no significant amount of rocks stopped during the actual event. This seems unnatural, and is likely caused by the finer resolution than recommended for RF3D (Dorren 2015), and/or incorrectly defined roughness parameters during the set up of the simulations.
An important thing to consider is with these rock fall simulations, no particle interaction or breakage of rocks are included in the simulations, but is an important process in large rock falls like these. As mentioned in the background section, there is a positive relationship between rock fall volume and runout distance (De Blasio 2011). Both Tveit B and Ruafjødd had large volumes of rock in motion, up to or exceeding 100 000 cubic meters. Interestingly, the simulations at Ruafjødd had significantly longer runouts than the actual event. This is very likely to be caused by the change in slope profile after the rock fall. The slope was changed to having a smoother transition between steep and slacker portions, which is evidently important for conserving the kinetic energy of the falling rocks. At the site, the debris were concentrated in the break-point at the base of the steepest part of the slope. There was probably a sharp transition here that was smoothed out by the rock fall debris, and leading to the longer runouts of the simulated blocks.
This was not seen at the Tveit B location, where the simulated blocks of the rock fall stopped in the area of the actual furthest travelling block, see figure 38. The RF3D simulation results were artificially short because of less than ideal input parameters.
In terms of ease of use and intuitive input parameters, Rocfall and RAMMS have the upper hand. In RAMMS the terrain model is easily constructed with a DEM and polygons defining the terrain type boundaries. The terrain types are also intuitive. Rocfall is also intuitive and have parameters that can be measured directly at field sites (restitution values and friction angles). RockyFor3D is more demanding to use, and it is work-intensive to change parameters between different runs. CRSP is quite outdated, and has a maximum vertex count of 100 for
the slope. Input parameters also have wide ranges of values for each typical terrain type, where the appropriate value is difficult to establish without much experience with the software.