UNDERSTANDING TALUS SLOPE GEOMORPHOLOGY FROM PERMAFROST DISTRIBUTION USING GEOPHYSICAL METHODS AND DETAILED GEOMORPHOLOGICAL MAPPING COL DU SANETSCH,

Word count: 25817

Joke Laporte Stamnummer: 01411341

Promotor: dr. Amaury Frankl – UGent – physical geography Copromotor: Prof. dr. Reynald Delaloye – Université du Freibourgh Begeleider: drs. Hanne Hendrickx – UGent – physical geography

Masterproef voorgelegd voor het behalen van de graad master in de richting Geografie

Academiejaar: 2018 - 2019

PREFACE

This thesis describes the results from a study about permafrost in the European . Because of my passion for nature, mountains and my concerns about the current climate change I wanted to get involved and contribute to this scientific fields. My interest helped me when I had hard times and finishing this dissertation shows me that even though something looks impossible, if I work hard I’m able to finish it. However it wouldn’t be possible without the help and support from others. Therefore I want to thank everyone who supported me along this process. First of all my supervisor Hanne Hendrickx and promotor Amaury Frankl. It was never a problem to pass by when I was struggling and they provided me with valuable tips, insights and feedback. Furthermore, my field research wouldn’t be possible without them, my fellow student Ewout and the other visitors who passed by in Switzerland. Geophysical measurements and drone campaigns can’t be done on your own and are quit hard in a challenging environment such as the Alps. So thank you for all your help and support!

I

POPULARIZING TEXT

Climate change is a hot topic, global temperatures are rising, precipitation patterns are changing. In the Alpine environment, permafrost, ground that remains at or below 0°C for at least two consecutive years, is present but degrading. As a result the environment will experience important modifications. Because of the causal relationship between the thermal regime and geohazards in mountain permafrost regions, scientific attention raises. This research forms a case study about the influence of permafrost on the geomorphology of a talus slope, how are they related, which trends can be found? Based on drone imaginary a geomorphological map of the study area was made. By combining temperature data, information about the horizontal displacement of the surface layer, measurements of permafrost probability and the geomorphological map, the interrelationships are explored and linked to the existing literature and theories.

POPULARISERENDE TEKST

Vandaag de dag staat klimaatsverandering hoog op de agenda. De temperatuur neemt toe en neerslag patronen veranderen. In Alpiene gebieden resulteert dit in een degradatie van permafrost, grond dat gedurende minimaal twee jaar een temperatuur lager dan 0°C heeft. Aangezien er een causaal verband is tussen permafrost degradatie en de toename aan geohazards in periglaciale gebieden is het belangrijk deze processen goed in kaart te brengen en te begrijpen. Dit onderzoek is een case studie waarin nagegaan wordt wat de invloed is van permafrost op de geomorfologie van een talus slope. Door het combineren van geomorfologische data, permafrost distributie, meteorologische data en horizontale oppervlakte snelheden, worden de relaties bestudeerd en gerelateerd met bestaande literatuur en theorieën.

ABSTRACT

Within this research we try to understand the influence of permafrost on talus slope geomorphology. Based on temperature data, measurements about the surface velocity and geophysical transects the permafrost distribution on the talus slope Col du Sanetsch is explored and explained. This information is used to interpret the geomorphology of the talus slope and its interrelationship with permafrost existence. A geomorphological map was developed by the combination of a high resolution DEM, resulting from an UAV campaign in 2018, and field observations. We concluded that the talus slope on Col du Sanetsch is a good example to explain the evolution, concepts and processes which influence the geomorphology of talus slopes. Furthermore, this map suggest permafrost by the existence of a rock glacier and protalus rampart, gelifluction and stress related landforms, such as transverse ridges. Nivation processes are related to the long lasting snow in this periglacial environment. According to our geophysical measurements a possible or probable existence of permafrost is present in a coarse grained area on the talus slope, the rock glacier and the landslide. Temperatures from springs at the foot of the talus slope and the Winter Equilibrium Temperature (WEqT) can be used as indicators for permafrost distribution. However, when using the WEqT as an indicator it is important to incorporate the context. When you only have data from one year, the intra-site differences of WEqT are equally important as the absolute value to indicate permafrost. Furthermore, the presence of unfrozen water

II can have a positive effect on the WEqT due to the latent heat released while freezing. The permafrost distribution is influenced by the temperature. The Mean Annual Ground Surface Temperature (MAGST) is lowest in areas were permafrost exist. A long lasting snow layer and a coarse grained surface negatively influence this temperature and will favour permafrost existence.

One of the main landforms on the talus slope is a landslide. Due to gelifluction, this landform is slowly moving downwards. A network of 28 points is measured every summer since 2011 to monitor the displacement rates. High horizontal surface velocities are measured within areas and years with high WEqT. The intra-site variations mainly depend on the difference in moisture content and the presence of an impermeable layer resulting from permafrost or seasonal frost. The several transverse ridges which can be found on the landslide are proves of this displacement.

In the summer of 2018, a webcam was installed to observe the evolution of the snow layer through the year. This will partly close the knowledge gap resulting from the absence of precipitation and snow data and make it possible to further explain and investigate the observed patterns.

SAMENVATTING

Deze studie tracht de invloed van permafrost op de geomorfologie van de talus slope op Col du Sanetsch te begrijpen. Een geomorfologische kaart is ontworpen door de combinatie van een hoog resolutie DEM en veldwerk. Het DEM resulteert uit luchtfoto’s verkregen tijdens een UAV campagne in de zomer van 2018. De permafrost distributie op Col du Sanetsch werd in kaart gebracht met behulp van geofysische metingen en indicatoren zoals de Winter Equilibrium Temperature (WEQT) en de temperatuur van de verschillende bronnen aan de voet van de talus slope. Aan de hand van meteorologische en topografische data wordt getracht de distributie te begrijpen en uit te leggen. Deze informatie wordt op zijn beurt gebruikt om de geomorfologie en de relatie met de permafrost distributie te interpreteren.

De talus slope op Col du Sanetsch is een goed voorbeeld om de evolutie, concepten en processen die aan de basis liggen van de geomorfologie van een talus slope uit te leggen. Daarnaast wordt de aanwezigheid van permafrost gesuggereerd door het voorkomen van een rotsgletsjer, gelifluctie en landvormen, zoals transversale ruggen, die gerelateerd worden met stress en interne deformatie. Verder zijn ook nivatie processen aanwezig in deze periglaciale omgeving. Geofysische metingen bevestigen de mogelijke en waarschijnlijke aanwezigheid van permafrost in de rotsgletsjer, landslide en de gebied op de talus slope dat bedekt wordt door grotere rots blokken. De distributie van permafrost wordt beïnvloed door de temperatuur, de Mean Annual Ground Surface Temperature (MAGST) is het laagst in gebieden waar permafrost voorkomt. Een sneeuwlaag die lang blijft liggen en de aanwezigheid van een oppervlakte laag met grote rots fragmenten beïnvloeden de grond temperatuur negatief en hebben op deze manier invloed op de distributie van permafrost.

III

Een opvallende landvorm op de talus slope is de landslide. Deze verplaatst zich langzaam richting de vallei via gelifluctie. Op deze landvorm werden 28 punten gemarkeerd die sinds 2011 ieder jaar opnieuw opgemeten worden. Dit maakt het mogelijk om de evolutie en ruimtelijke variatie in deze verplaatsing te analyseren. Een exponentiële relatie werd gevonden tussen de WEqT en de horizontale snelheid. De intra-site variaties worden voornamelijk gelinkt met de verschillende vochtigheidsgraad van de bodem en de aanwezigheid van een ondoordringbare laag, als resultaat van permafrost of seizoensgebonden vorst. De vele transversale ruggen die aanwezig zijn op de landslide getuigen van deze verplaatsing.

In de zomer van 2018 werd een webcam geïnstalleerd. Deze maakt het mogelijk om de evolutie van de sneeuwlaag doorheen het jaar te observeren. Op deze manier kan dieper ingegaan worden op de rol van de sneeuwlaag in de verschillende processen en trends.

IV

TABLE OF CONTENT

1. Introduction ...... 1 2. High Alpine talus slopes and their interaction with permafrost distribution ...... 2

2.1 Geomorphology of talus slopes ...... 2

2.1.1 Formation of talus slopes ...... 2

2.1.2 The cross section of a talus slope ...... 3

2.1.3 Geomorphological transport processes and their resulting landforms on talus slopes ...... 3

2.2 Distribution and evolution of mountain permafrost ...... 4

2.2.1 Definition of mountain permafrost ...... 4

2.2.2 Understanding permafrost distribution on mountain slopes ...... 5

2.2.3 Mountain permafrost maps and modelling ...... 9

2.2.4 Permafrost degradation ...... 9

2.3 Geomorphological processes and dynamics on talus slopes as impacted by permafrost degradation ...... 11

3. Study objectives ...... 12 4. Study site ...... 12 5. Material and methods ...... 15

5.1 Topographic survey based on field observations and UAV ...... 16

5.1.1 Field observations ...... 17

5.1.2 Data acquisition with UAV ...... 17

5.1.3 Real-Time Kinematic GNSS (RTK-GNSS)...... 17

5.1.4 Data processing in Agisoft PhotoScan ...... 18

5.1.5 Geomorphological mapping based on DEM interpretation ...... 20

5.1.6 Defining topographic roughness ...... 20

5.2 Surface velocity measurements on the landslide ...... 22

5.3 Temperature measurements...... 24

5.4 Permafrost mapping based on geophysical measurements ...... 25

5.4.1 VES ...... 27

5.4.2 ERT ...... 28

5.4.3 Transects ...... 31

5.4.4 Defining permafrost probability classes ...... 34

V

5.5 Statistical analyses ...... 34

5.5.1 Analysis of the meteorological trends ...... 35

5.5.2 Analysing temporal and spatial differences in surface velocity and their relation with temperature, permafrost distribution and geomorphology ...... 35

5.5.3 Analysing the distribution of surface roughness and the interrelation with permafrost distribution ...... 36

5.5.4 Comparing the measured permafrost distribution and permafrost probability maps ...... 36

6. Results ...... 37

6.1 Geomorphological map ...... 37

6.1.1 Gravitational processes ...... 37

6.1.2 Snow, frost and (peri)glacial landforms ...... 37

6.2 Temperature measurements...... 42

6.3 Geophysical survey ...... 44

6.3.1 VES ...... 44

6.3.2 ERT transects ...... 46

6.4 Annual surface velocity of the landslide ...... 49

6.5 Relations between geomorphological characteristics, permafrost distribution, meteorological parameters and surface velocity on the landslide...... 50

6.5.1 Surface velocity related to meteorological factors, permafrost distribution and topographical characteristics ...... 50

6.5.2 Surface roughness as an explanatory factor of permafrost distribution ...... 53

7. Discussion ...... 55

7.1 Geomorphological mapping ...... 55

7.1.1 Gravitational processes ...... 55

7.1.2 Snow, frost and (peri)glacial landforms ...... 56

7.2 Meteorological analysis ...... 59

7.3 Permafrost distribution indicated by geophysical and temperature measurements ...... 61

7.3.1 Permafrost distribution on the talus slope ...... 61

7.3.2 Permafrost distribution on the rock glacier ...... 62

7.3.3 Permafrost distribution on the landslide ...... 64

7.3.4 Differences in apparent resistivity between landslide and talus slope ...... 65

VI

7.3.5 Permafrost probability maps ...... 65

7.4 Annual surface velocity ...... 69

7.5 Possibilities for future research ...... 70

8. Conclusions ...... 70 9. References ...... 71

Literature ...... 71

Maps and data ...... 80

10. Attachments ...... 81

10.1 Attachment 1: VES field template: Wenner array ...... 81

10.2 Attachment 2: VES field template: Schlumberger array ...... 82

LIST OF FIGURES

Figure 1: Surface temperature anomalies, relative to the period 1961 – 1990, in the Swiss Alps (Säntis, Lugano and Zürich) compared to the global anomalies...... 1

Figure 2: resistivity of soil, rock and minerals ...... 7

Figure 3: The chimney effect ( ...... 8

Figure 4: Permafrost degradation ...... 10

Figure 5: General meteorological context: long-term means of monthly mean temperature, monthly maximum and minimum temperature as well as monthly precipitation sums in Sion ...... 13

Figure 6: Installation and measurement of GCPs. a) a fixed GCP, b) a ‘cloth’ GCP ...... 18

Figure 7: a) Workflow in Agisoft to build a DEM based on Hendrickx et al. (2019). b) A model after the first alignment. In the upper zone, different gaps are visual, here we added more photographs...... 19

Figure 8: a) Flow chart to calculate SDrestopo. b) Example of the residual topography, the difference between the LiDAR DTM and mean DTM results in the residual topography. c) The influence of the moving-window size on the surface roughness...... 21

Figure 9: Topographic roughness classes based on photographs ...... 22

Figure 10: Overview data points: temperature and surface velocity (source: Google earth, 2016) ...... 23

Figure 11: Principles of ERT prospecting. a) Relation between the resistivity and the temperature. b) An example of the electrical field created by inserting I into the quadripole. c) The most used ERT electrical arrays ...... 27

Figure 12: Relative position of the electrodes in a Wenner (a) and Schlumberger array (b), VES ...... 28

Figure 13: concept of the Wenner-Schlumberger array ...... 29

VII

Figure 14: Relative differences in apparent resistivity for transect CdSLS1a. Bad data points (green) can be selected and removed...... 30

Figure 15: RMS error statistics of CdSLS1a after a preliminary inversion. By moving the green line, bad data points can be selected and removed...... 31

Figure 16: Overview geophysical measurements ...... 33

Figure 17: Debris flow deposits and gullies ...... 40

Figure 18: Landslide ...... 40

Figure 19: Panorama of the talus slope ...... 40

Figure 20: Rock glacier (dotted lines = ridges) ...... 41

Figure 21: Nivation hollow (dotted lines = ridges) ...... 41

Figure 22: Nivation zone on the talus slope, small depressions followed by an asymmetric ridge ...... 41

Figure 23: Ridges in the nivation zones visualised using the DEM-derivative ‘aspect’ on a fine resolution DEM ...... 41

Figure 24: Lateral arcuate ridge on the rock glacier, as seen from the UAV ...... 41

Figure 25: Climatological graphs based on the output of 3 t-loggers located on the landslide at Col du Sanetsch (August 2014 – August 2018)...... 43

Figure 26: Results from the VES measurement (Schlumberger array) on landslide position 1 (t-logger 017 – down) (transect 1) ...... 45

Figure 27: Results from the VES measurement (Schlumberger array) on landslide position 2 (t-logger 010 – mid) (transect 4) ...... 45

Figure 28: Results from the VES measurement (Schlumberger array) on the talus slope, (t-logger AT – GST – 1, down) (transect 10) ...... 45

Figure 29: Results from the VES measurement (Schlumberger array) on the talus slope, (t-logger AT – GST – 2, up) (transect 11) ...... 45

Figure 30: Results from the VES measurement (Wenner array) on landslide position 2 (t-logger 010 mid). 0 = t-logger, ‘-20’ means the position 20 m from the t-logger, direction NE. The red line indicates the location of the ERT profile. (transect 7) ...... 46

Figure 31: ERT measurements on landslide position 1 (t-logger 017 – down) ...... 47

Figure 32: ERT measurements on the landslide position 2 (t-logger 010 - mid) ...... 47

Figure 33: ERT measurements above the landslide, position 3 (t-logger 005 - up)...... 48

Figure 34: Lateral ERT profile rock glacier ...... 48

Figure 35: Vertical ERT profile rock glacier ...... 49

VIII

Figure 36: Annual horizontal surface velocity on the landslide. Mean of a set of points selected in several sections of the moving landform ...... 49

Figure 37: Exponential relation between the annual horizontal displacement and the WEqT...... 50

Figure 38: Annual and winter precipitation (mm) Sion and the annual horizontal displacement on Col du Sanetsch (m) ...... 51

Figure 39: Relation between the displacement rate and slope (light blue = area above the landslide, dark blue = on the landslide) ...... 52

Figure 40: Grain size distribution on the talus slope ...... 53

Figure 41: Grain size distribution on the landslide ...... 53

Figure 42: Landslide on Col du Sanetsch ...... 57

Figure 43: Representation of an idealized landslide ...... 57

Figure 44: Rooting zone of the landslide ...... 58

Figure 45: Gully filled with materials from the landslide ...... 59

Figure 46: Mean MAGST and average zero curtain period for the different t-loggers ...... 60

Figure 47: Borehole temperatures at approximately 20 m depth compared to the WEqT and MAGST of Arp - 010 - mid ...... 61

Figure 48: T-logger located below the front of the rock glacier (GST – 13 – RG) ...... 64

Figure 49: Temperature of t-loggers GST-9-RG and GST-13-RG in 2016 - 2017 ...... 64

Figure 50: Permafrost distribution: Geophysical measurements compared to the Swiss Potential Permafrost Distribution Map ...... 67

Figure 51: Permafrost distribution: Geophysical measurements compared to the APIM ...... 68

LIST OF TABLES

Table 1: overview of some typical geophysical outcomes ...... 6

Table 2: Possible influences of permafrost degradation on geomorphological processes ...... 12

Table 3: Overview data ...... 16

Table 4: Technical table UAV survey ...... 20

Table 5: Topographic roughness classes ...... 22

Table 6: Technical specifics of the t-loggers ...... 24

Table 7: Overview of the t-loggers on Col du Sanetsch ...... 25

Table 8: Metadata for the different transects ...... 32

IX

Table 9: Permafrost probability classes ...... 34

Table 10: Mean WEqT (Bolt = permafrost existence probable) ...... 42

Table 11: Colour legend of permafrost classes, ERT profiles on the landslide ...... 48

Table 12: Permafrost distribution from geophysical and temperature measurements related to topographic factors on the talus slope ...... 62

Table 13: Permafrost distribution from geophysical and temperature measurements related to topographic factors on the rock glacier ...... 63

Table 14: Permafrost distribution from geophysical and temperature measurements related to topographic factors on the landslide ...... 65

LIST OF MAPS

Map 1: Location study site: Col du Sanetsch ...... 14

Map 2: Location talus slope and different landforms ...... 14

Map 3: Geomorphological map Col du Sanetsch ...... 39

Map 4: Thematic map - slope, permafrost distribution and annual surface velocity (1 = slow moving frontal zone, 2 = fast moving median zone) ...... 52

Map 5: Thematic map surface roughness and permafrost distribution ...... 54

LIST OF ABBREVIATIONS

APIM = Alpine Permafrost Index Map BTS = Bottom Temperature of Snow DEM = Digital Elevation Model ERT = Electrical Resistivity Tomography GCP = Ground Control Point GPR = Ground Penetrating Radar GST = Ground Surface Temperature MAAT = Mean Annual Air Temperature MAGST = Mean Annual Ground Surface Temperature PERMOS = Swiss Permafrost Monitoring Network

SDrestopo = Standard Deviation of Residual Topography SRT = Seismic Refraction Tomography UAV = Unmanned Aerial Vehicle VES = Vertical Electrical Sounding WEqT = Winter Equilibrium Temperature

X

1. INTRODUCTION

Since the late 19th century, the climate has changed rapidly because of anthropogenic impacts (Pachauri & Meyer, 2014). Global temperatures are rising, precipitation patterns are changing. There have always been periods with increasing or decreasing temperatures, but since the 1970s this warming is particularly marked (Stocker et al., 2013, Haeberli & Beniston, 1998). Following the IPCC (Pachauri & Meyer, 2014) it is 95 % certain that this current global warming is caused by humans. As can be seen in Figure 1, this temperature increase is more marked in the Alps, where the average temperature has risen twice as fast as the global average (Gobiet et al., 2014; Beniston, 2006). As this warming will continue during the 21st century (Stocker et al, 2013), the Alpine cryosphere will experience important modifications (Deluigi et al., 2017). Permafrost, ground that remains at or below 0°C for at least two consecutive years (Dobinski, 2011), degrade due to this climate change (Etzelmüller & Frauenfelder, 2009). Because of the causal relations between the thermal regime and geohazards in mountain regions, mountain permafrost has been receiving an increased scientific attention (Etzelmüller, 2013). Examples of research topics related to permafrost warming are the influence on rock fall activity (e.g. Gruber and Haeberli, 2007; Ravanel et al., 2010), rock glacier acceleration (e.g. Kääb et al., 2007; Roer et al., 2008; Delaloye et al., 2010) and sediment transfer (e.g. Lane et al., 2007; Kobierska et al., 2011). The aim of this study is to understand the influence of permafrost distribution on talus slope geomorphology. The first section gives the state-of-the-art on the current knowledge and advances related to the scientific research on permafrost degradation and talus slope geomorphology. The second part is the core of the study.

Figure 1: Surface temperature anomalies, relative to the period 1961 – 1990, in the Swiss Alps (Säntis, Lugano and Zürich) compared to the global anomalies. (Source: Beniston, 2006)

1

2. HIGH ALPINE TALUS SLOPES AND THEIR INTERACTION WITH PERMAFROST DISTRIBUTION

In order to analyse and understand the relations between talus slope geomorphology and permafrost distribution, it is important to have a theoretical background. Providing this theoretical background is the purpose of the first chapter. We first focus on the talus slope geomorphology and more specific the formation and transport processes. The second part elaborates on the distribution and evolution of mountain permafrost. In the last part, these topics are combined and we discuss possible influences of permafrost on high alpine talus slope geomorphology.

2.1 Geomorphology of talus slopes

Talus slopes, or scree slopes, are formed by unconsolidated clasts of different sizes accumulating at the foot of a rock cliff (Müller et al., 2013; Gutiérrez & Gutiérrez, 2016). They are one of the dominant types of debris storage and has a high potential of debris release and mass movements (Phillips et al., 2009). On an idealized periglacial mountain slope, a sequence of characteristic landforms can be found: an upper convex segment, with the headwall, a talus slope, which normally has a constant slope angle, and a concave segment at the base. Depending on the actual situations, some of these components may not be present or negligible (Ritter et al., 1995; Huggett, 2017; Müller et al., 2014).

2.1.1 Formation of talus slopes

Talus slopes are the result of active sedimentation of materials destabilized on the headwall (Schoeneich et al., 2011; Müller et al., 2014). The destabilization of these materials depends on the shear stress and shear strength. Stress is formed by any force that tends to move the materials, for instance, gravity. Shear strength are the properties of the matter that resist the stresses generated by gravitation. Many different factors influence the shear strength and shear stress, for example: freeze- thaw cycles, the slope, roughness of the plane, size and shape of the particles and the cohesion between the particles. When the shear stress exceeds the shear strength the slope is unstable, and a small trigger may initiate a mass movement (Huggett, 2017; Ritter et al., 1995). The travelling distance of a boulder is a balance between the energy gain and loss. As a boulder moves downslope, he will gain energy due to his movement in the gravity field. Collisions and friction will result in a loss of energy. When shear strength increases, by for instance an increase of surface roughness, or the shear stress decreases because of a less steep slope, this may result in a higher loss/ smaller gain of energy and the boulder may halt (De Blasio & Sæter, 2015; Ritter et al., 1995).

Two important processes for the development of a talus slope are frost weathering and paraglacial adjustment (Schoeneich et al., 2011; Ballantyne, 2002). These processes decrease the shear strength of the headwall and so they contribute to the production of debris. Frost weathering, or frost shattering, breaks off small grains and large boulders from the bedrock. It depends on freeze-thaw cycles, and is a function of freeze-thaw frequency, moisture content and the tensile strength of the bedrock. More information about this process can be found in other literature (e.g. Ritter et al., 1995; Hugget, 2017). Another process that can be related to the origin of a talus slope is paraglacial adjustment. When a bedrock is covered by ice, this overlying ice induces internal stress levels that are much higher than

2 those that might be expected from the loading alone. Glacier downwastage result in a relaxation of tensile stresses within this rock mass. This stress release results in propagation of the internal joint network, a loss of cohesion and a reduction of internal locking stresses. This may lead to immediate or delayed rock-slope failure, deformation or rock fall, depending on the lithology and structure (e.g. joint density and weaknesses) of the rock mass (Ballantyne, 2002). The block and grains that breaks off, will move downslope by bouncing, rolling or sliding (Gutiérrez & Gutiérrez, 2016; Huggett, 2017; Ritter et al, 1995).

2.1.2 The cross section of a talus slope

Talus slope typically has a straight longitudinal profile, with a slope angle of 30° - 40° and a basal concavity (Gutiérrez & Gutiérrez, 2016; Luckman, 2013). The slope angle is determined by the friction angle, the angle at which a block will begin to slide, of the cohesionless debris (Gutiérrez & Gutiérrez, 2016; Ritter et al., 1995). On most talus slopes, an increasing grain size can be found downslope (Lambiel & Pieracci, 2008; Ritter et al., 1995; De Blasio & Sæter, 2015; Gómez et al., 2003). This process, known as fall sorting, is most marked at the bases of slopes. It can be explained by a combination of different mechanisms. As a large boulder has a greater momentum, it has the possibility to travel a greater distance. Furthermore, the frictional resistance of the surface depends on the relationship between the size of a moving boulder and the irregularities of the surface. In that way, boulders will only come to rest in areas with boulders of more or less the same size (Luckman, 2013). De Blasio & Sæter (2015) did an experimental study of the behaviour of a single grain falling and travelling on a homogeneous granular bed. Small grains, which fall on larger grains, will stay close to the point of falling. Large grains who fall on a bed of smaller grains may behave in two manners. Or they stop immediately, in a self-created crater, or they roll down slope. Their abundance peak around the fall zone and further downslope. The degree of sorting depends on different aspects, length of the slope, cliff height and the size and shape of the dominant particles (Luckman, 2013).

2.1.3 Geomorphological transport processes and their resulting landforms on talus slopes

In addition to this ‘primary’ processes, which are responsible for the formation of the talus slope, different other geomorphological processes can be found. They will commonly rework the initial structure of a talus slope. In periglacial conditions processes related to frost, thaw and snow are highly active (Huggett, 2017). On the talus slope, snow avalanches and permafrost creep, for instance rock glacier creep, dominate (Müller et al., 2014). Other processes that occur are debris flows, solifluction, landslides and dry ravel (Ritter et al., 1995; Huggett, 2017).

Solifluction is a slow transport process related to freeze-thaw actions. Depending on the temperature of the soil, different solifluction processes may occur. A plug-like flow, which result from thawing ice lenses produced by an upward frost penetration, will only occur in cold permafrost regions (Matsuoka, 2001). Other solifluction processes, such as diurnal frost creep, annual frost creep and gelifluction may influence slopes underlain by both cold or warm permafrost or even non-permafrost zones. All of those, slow, mass wasting processes result from freeze-thaw actions in fine-textured soils. They are controlled

3 by freeze-thaw cycles and the depth and thickness of ice lenses (Matsuoka et al., 2001). The specific form, frost creep, is a result of the expansion of the soil during freezing and a contraction during the thawing. This expansion is perpendicular on the slope, the contraction is vertical, because of the gravity. The result is a net downslope moment of the materials (Huggett, 2017). Gelifluction is a very slow downslope movement of a water-saturated soil over the frozen ground during the summer months (Huggett, 2017). In this season ground become saturated because of the restricted drainage associated with a seasonally frozen water table or permafrost and the moisture delivered by thawing. This increased moisture may also intensify other mass movement processes (Hugget, 2017; Ritter et al., 1995). As the moving soil overturned the superficial soil, due to a reduced velocity, solifluction lobes, large tonguelike masses of surface debris, may be produced. Those are usually poorly sorted, and the form of these features depend on the texture, gradient and soil moisture. A fine-grained soil layer overriding a coarse sediment, will also produces solifluction lobes and sheets (Matsuoka, 2001; Huggett, 2017).

Fast acting processes that rework the structure of a talus slopes, are for example debris flows, landslides and dry ravel. Debris flows consist of a fast-moving body of sediment particles (70 – 90%) with water that moves downslope. An abundant source of moisture, fine grained sediments and relatively steep slopes are required conditions. They typically form in small gullies, where run off can be concentrated (Luckman, 2013). Landslides are rapid mass movements along clear-cut shear planes. Dry ravel is the movement of individual particles by rolling, bouncing of sliding (Huggett, 2017). All those processes will influence the form and profile of the talus slope. For instance, debris flows may gully the upper slope and produce debris cones where the transported materials accumulate (Luckman, 2013).

Snow and related processes may also influence talus slopes. As the frictional properties of snow- covered talus are quite different, large boulders can be trapped or slide over the surface. This result in another distribution of grain sizes (Luckman, 2013). Nivation hollows may also be formed. This are depressions, associated with late-lying or perennial snow patches, where frost action and intensified weathering may increase erosion (Gutiérrez & Gutiérrez, 2016). Nivation hollows can be associated with protalus ramparts. These ridges, at the downslope margin of a snow patch, consist of unsorted gravel and boulders. Rockfall debris from the cliff, rolls and slides down over the snow and accumulate at the foot, were it builds a ridge (Gutiérrez & Gutiérrez, 2016; Shakesby 2004). However, some other authors see protalus ramparts as embryonic rock glaciers (e.g. Scapozza et al., 2011; Haeberli, 1985). We decided to use the morphological definition. Furthermore, snow avalanches may erode loose, fine materials and transport them downslope, where they are deposited on coarser materials.

2.2 Distribution and evolution of mountain permafrost

2.2.1 Definition of mountain permafrost

The definition of a permafrost body is based on the following thermal criteria: ‘ground that remains at or below 0°C for at least two consecutive years’ (Dobinski, 2011; Gruber & Haeberli, 2009, Harris et al., 2009). It refers to a thermal state of the lithosphere, so it is not a material phenomenon (Dobinski, 2011). Permafrost results from a negative energy balance and depends on incoming solar radiation and

4 sensible heat flux (Schoeneich et al., 2011). Mountain permafrost is found in mountain areas and will be influenced by the characteristics of topography, such as altitude and aspect (Gruber & Haeberli, 2009). This results in a complex, discontinuous permafrost distribution in the European Alps (Pieracci et al., 2008; Deluigi et al., 2017; Lambiel & Pieracci, 2008). Most of the mountain permafrost in the European Alps is temperate or warm, which means it will have a temperature close to 0° C (Etzelmüller & Frauenfelder, 2009).

2.2.2 Understanding permafrost distribution on mountain slopes

Permafrost is invisible, which makes it hard to know its exact distribution on mountain slopes. Theoretically, the lower limit of permafrost in the Alps is 2500 m.a.s.l. (Deluigi et al., 2017). In reality, this is influenced by many topoclimatic factors, such as altitude, slope, aspect, Mean Annual Air Temperature (MAAT), snow and solar radiation (Harris et al., 2009; Deluigi et al., 2017). All those factors can give a first indication were permafrost may be present. Furthermore, the presence of mountain permafrost can be indicated by thermal data, geomorphological phenomena and geophysical characteristics.

2.2.2.1 Thermal indicators

As permafrost is a thermal state, a first, but not absolute, hypothesis of the existence of permafrost can be made based on the temperature (Scapozza et al., 2011). A MAAT of -3°C is the theoretical threshold value that can be used to identify altitudinal bands in the European Alps that contain a significant amount of permafrost (Gruber & Haeberli, 2009). The MAAT is influenced by both the altitude and aspect of the slope. As a result, permafrost on northern slopes will occurs at lower altitudes than on southern slopes (Schoeneich et al., 2011). A commonly used thermal indicator for permafrost distribution is the Winter Equilibrium Temperature (WEqT). The WEqT is the mean temperature over a period of 30 days during which the snow layer is stable. This period ends 14 days before the start of the melting period. As snow is a good insulator, the WEqT depend on the ground thermal regime, and thus the presence or absence of permafrost (Schoeneich, 2011). The snow cover must have a depth of 80 – 100 cm depth and be established since the beginning of the winter, in order to eliminate short term influences of the air and surface (Hoelzle et al., 1999; Schoeneich, 2011). With WEqT lower than -3°C, permafrost is probable, permafrost is unlikely if the temperature is higher than -2°C. Between -2°C and -3°C only marginal permafrost in thick active layers may eventually occur (Hoelzle et al., 1999; Schoeneich, 2011). During the melting period, the so called ‘zero curtain period’, the temperature is constant around 0°C.

2.2.2.2 Geomorphological indicators

Not only temperature data may indicate permafrost. Rock glaciers are commonly used as geomorphological indicator of permafrost presence (e.g. Deluigi et al., 2017). If active or inactive rock glaciers are present, permafrost can be expected at the slope. The presence of relict rock glaciers suggests the absence and degradation of permafrost conditions (Deluigi et al., 2017). Another possible indicator of permafrost is solifluction. However, this one is not so clear as the presence of rock glaciers. Only if plug-like flow exists, the presence of permafrost can certainly be expected. The characteristics

5 of the other kinds of solifluction will differ between cold, warm or non-permafrost zones. For instance, in areas with cold permafrost, the surface velocity will rarely exceed 5 cm year-1. This velocity may rise and reach a maximum (> 10 cm year-1) towards regions underlain by warmer permafrost or those lacking permafrost. Furthermore, the maximum depth of the movement, which is reflected by the minimum frontal height of the solifluction lobe, may decrease toward warmer permafrost. In non-permafrost areas, this depth ranges from a few centimetres up to 50 cm (Matsuoka, 2001). However, there seems to be insignificant difference in solifluction between zones of warm permafrost and seasonal frost (Matsuoka, 2001).

2.2.2.3 Geophysical measurements as indicator

Last, but not least, as geophysical properties of the soil alter significantly with a phase change of water, those properties can also be used as an indicator for permafrost presence (Harris et al., 2009). Based on characteristics as the electrical resistance, propagation speed of georadar waves and velocity of seismic waves, frozen ground can be distinguished from unfrozen ground (Harris et al., 2009). Commonly used methods are Vertical Electrical Soundings (VES) (e.g. Lambiel & Pieracci, 2008), Electrical Resistivity Tomography (ERT) (e.g. Kenner et al., 2017; Hauck et al., 2003; Kneisel et al., 2000, Scapozza et al., 2011), seismic refraction tomography (SRT) (e.g. Kenner et al., 2017; Hausmann et al., 2007) and Ground Penetrating Radar (GPR) (e.g. Hausmann et al., 2007). SRT may be used complementary to resistivity techniques, to divide air-filled porous layers from ice-filled porous layers (Kenner et al., 2017). GPR is commonly used in high latitude regions, for instance Alaska, but can also be used to determine the internal structure of a rock glacier (Hausmann et al., 2007). Table 1 gives an overview with some typical values for every material and method. The values are based on the earlier mentioned researches. It is important to keep in mind, that the exact values may differ depending on the host materials, ice content, temperature of the ice and impurities (Hauck & Vonder Mühll, 2003). In Figure 2 the resistivity of different rocks, soils and minerals are listed (Loke, 2002).

VES ERT SRT GPR Debris / open-work 6 – 300 kΩ.m 10 – 100 kΩ.m 950 m.s-1 surface layer 0,14 – 0,15 Ice rich permafrost 8-300 kΩ.m > 50 kΩ.m 3300 m.s-1 m.ns-1 Depend on the material e.g. Bedrock < 10 kΩ.m Dolomites: < 8 kΩ.m 4100 m.s-1 Serpentinite 18 – 25 kΩ.m Unfrozen materials/ 1000 – 2000 1 – 20 kΩ.m < 10 kΩ.m low ice content m.s-1

Table 1: overview of some typical geophysical outcomes (Based on: Lambiel & Pieracci, 2008 (VES); Hauck et al., 2003 (ERT); Scapozza et al., 2011 (ERT); Kenner et al., 2017 (ERT, SRT); Hausmann et al., 2007 (GPR))

6

Figure 2: resistivity of soil, rock and minerals (source: Loke, 2002)

2.2.2.4 Distribution of mountain permafrost on talus slopes

Within the belt of discontinuous permafrost, which is above the 2200 – 2300 m.a.s.l, it is common to measure highest resistivity and coldest temperatures in the lower part of the debris accumulation. In the higher parts, temperatures increase and resistivity decreases (Lambiel & Pieracci, 2008). This suggest a higher probability of permafrost in the lower parts and a warming or absence of permafrost in the higher parts of a talus slope. Different case studies confirm this unexpected contrast (e.g. Pieracci et al., 2008; Scapozza et al., 2011; Delaloye & Lambiel, 2005). The indicated theoretical distribution of permafrost in a talus slope is the result of three main controlling factors: grain size, presence of snow and in some cases a chimney effect (Lambiel & Pieracci, 2008; Pieracci et al., 2008; Scapozza et al., 2011). These are complex processes, so the exact impact of every single process is difficult to disentangle.

In coarse, blocky materials, which dominate at the foot of the talus slope, a negative temperature anomaly exists (e.g. Harris & Pederson, 1998; Lambiel & Pieracci, 2008). There are different theories to explain this temperature anomaly: the Balch effect, the water retention capacity, the continuous air exchange with the atmosphere and the chimney effect. The Balch effect is based on the fact that cold air is denser and displace warmer air. This process can only occur where there are large connecting

7 spaces between the blocks (Harris & Pedersen, 1998). Additionally, a continuous air exchange, which is possible where there is no continuous snow cover, will favour ground cooling in blocky areas (Lambiel & Pieracci, 2008). The difference in water retention capacity is also a possible explanation. Coarse grained soils have a lower water retention capacity, which result in less water. Since there is a smaller amount of water, less latent heat will be released while freezing and there will be a faster cool down during autumn (Kenner et al., 2017).

The chimney effect, a mechanism of air circulation, is another explanation. This process does not influence all talus slopes. The efficiency depends on the macro-porosity of the soil, the temperature contrast between the inside and outside air, and the snow depth (Phillips et al., 2009; Pieracci et al., 2008). In small grained formations, or formations completely sealed with ice, circulation of air is far more difficult, and the chimney effect may be less important, or non-existent (Pieracci et al., 2008). In Figure 3 the chimney effect is visualized. It exists of two different phases: winter and summer. In winter, the air inside the slope is warmer, and thus lighter, than outside. This air rises inside the slope and expel at the top. In the lower parts, this causes an aspiration of cold air into the talus. In summer the process reverses and there will be a gravitational discharge of cold, heavy air at the base of the talus slope. This process causes an overcooling at the base of the slope and a positive temperature anomaly in the upper parts. In winter, the ascent of warm air, can create a basal melting of the snow cover and eventually snowmelt windows are formed (Delaloye & Lambiel, 2005).

Figure 3: The chimney effect (Source: Wicky & Hauck, 2016)

A last factor which may explain the existence of permafrost in the lower parts of the talus slope is the snow distribution. Usually as a result of avalanches or the chimney effect, snow will be redistributed. Long lasting snow deposits can be found at the bases of talus slopes (Pieracci et al., 2008; Kenner et al., 2017; Lambiel & Pieracci, 2008; Millar et al., 2014). This will result in a delayed snowmelt and, because of the insulating properties of snow, a relative cooling of the ground. Kenner et al. (2017) hypothesize that this ‘insulation theory’ has a minor influence on the presence of permafrost, and that the thermal characteristics of the ground are more important.

8

2.2.3 Mountain permafrost maps and modelling

Since permafrost is invisible, this makes it very difficult to get to know the exact distribution. Boreholes and geophysical data can tell us if permafrost is present, but only at the site itself. To get an area covering map, models are developed (e.g. Deluigi et al., 2017). This models are based on topoclimatic factors and case studies. Mostly they are calibrated for a specific site or regions and do not cover the whole Alpine region. The Alpine Permafrost Index Map (APIM) is the first map that covers the whole European Alps. This map is created by Boeckli et al. (2012) and can be found as a kml-file or downloaded as a georeferenced png-format. The index is an indicator for the probability for permafrost occurrence, but cannot allocate the extent and thickness of the permafrost as this depend on various local and regional processes. A gradation between ‘permafrost in nearly all conditions’ and ‘permafrost only in very favourable conditions’ is made. These conditions refer to the topographical and ground characteristics. For instance: permafrost will not be expected in very fine materials or solid rocks (Boeckli et al., 2012). The map of potential permafrost distribution is another map, which concentrate on the distribution in Switzerland. It incorporates different parameters such as, altitude and solar radiation (www.bafu.admin.ch, 11/5/2018).

2.2.4 Permafrost degradation

Permafrost is a thermal state and will therefore be influenced by the climate and climate change. The thermal evolution of permafrost depends on air temperature and snow cover. An increase of sensible heat flux can result in a general warming of ground temperatures, which result in a degradation of permafrost (Dobinski, 2011; Deluigi et al., 2017). Not only a decrease in the extent of permafrost, but also an increase in the temperatures, will occur (Dobinski, 2011). Figure 4 shows the degradation processes in a permafrost body.

9

Figure 4: Permafrost degradation (Source: Dobinski, 2011)

As a result of this degradation, there will be more temperate permafrost areas (Gruber & Haeberli, 2009) and the altitudinal distribution of frost types and permafrost will shift towards higher regions. Sporadic permafrost disappears from the middle altitudes and the lower permafrost limit shift towards higher altitudes, where continuous permafrost is replaced by discontinuous permafrost (Schoeneich et al., 2011). The thermal regime of permafrost also depends on the snow cover, which act as an insulator. Most climate models predict an increase of winter precipitation (Schoeneich et al., 2011). Depending of the trend in timing and the thickness of the snowpack, this will result in warming or rather a cooling. An early or thick snow cover prevent cooling during the winter, but a late melt prevents warming in the spring. Snow cover will only impact in areas of moderated slopes, where the snow cover can develop and stay (Schoeneich et al., 2011). The last decades, there has been an increasing interest in permafrost research. This resulted in an augmentation of monitoring sites and research. An example is the Swiss Permafrost Monitoring Network (PERMOS), which has as main goal to document the state and temporal variations of permafrost in the Swiss Alps on a long-term basis (Vonder Mühll et al., 2008). All the observed elements (e.g. MAGST, borehole temperatures and ERT) showed a warming trend since the year 2009. This trend is likely to be a cumulative effect of continuously warm climate conditions (Nöztzli et al., 2016).

10

2.3 Geomorphological processes and dynamics on talus slopes as impacted by permafrost degradation

This paragraph provides a non-exclusive overview of recent research about the impacts of permafrost degradation on geomorphological processes. The focus will be on geomorphological processes that can occur on talus slopes. Harris et al. (2008) identifies the change in active layer thickness as a key response of permafrost on climate change. The thickness of this layer depends on interannual variations in temperature. This is also stated by other researcher and case studies (e.g. Guo & Wang, 2017; Deline et al., 2015). The rate of increase, and in general the influence of a changing temperature on the geomorphology in periglacial environments, depend on the ice content. In ice-rich permafrost, the latent heat released during thawing, will reduce the thermal conductivity. As a result, there may be a delayed warming. Therefore, one could say that ice-rich permafrost will be less sensitive to climate change, from a thermal point of view (Deline et al., 2015; Gruber & Haeberli, 2005; Schoeneich et al., 2011; Phillips et al., 2009). Furthermore, the altitudinal range, with the highest frequency of freeze-thaw cycles is likely to move to higher altitudes. As a result, the geomorphological activities in the middle altitudes may decrease. Contrary, an increase is expected at the lower limit of the continuous permafrost belt (Schoeneich et al., 2011). Most geomorphological processes with permafrost occurrence are connected to freeze-thaw cycles and active layer thickness. Therefore, the changes mentioned above, together with the increasing temperatures of the permafrost, may influence the nature, intensity and frequency of these processes (Harris et al., 2008; Dobinski, 2011; Deluigi et al., 2017).

Table 2 provides an overview of possible geomorphic reactions on climate change, related to permafrost degradation. One of the possible geomorphological processes reactions is thermokarst. The loss of ice can lead to thermokarst phenomena (Schoeneich et al., 2011). Thermokarst phenomena may tell something about the history and formation of the permafrost. Kenner et al. (2017) identified, for instance, a thermokarst depression in a lobate structure on the Flüelapass. These observations were used to reconstruct the history and origin of the geomorphology and the presented permafrost. Another possible influence is the change in solifluction rate. If there is no frozen layer left, the soil drains and the solifluction will stop (Schoeneich et al., 2011). However, there was no empirical evidence of this phenomenon on the different case studies in action 5.3 from permaNET (Lieb & Kellerer-Pirklbauer, 2011). Rock glaciers has different reaction on permafrost degradation. In general, an increased creep rate can be observed. However, if most of the ice is melted out, the creep rate could decrease, because of the increased friction and eventually they may collapse (Harris et al., 2009). Other permafrost creep phenomena will also show a non-linear increase in velocity because of the decrease in viscosity of ice (Deline et al., 2015). At altitudes with an increased frequency of freeze-thaw cycles, the frost weathering and talus production may intensify. An increased instability of steep bedrocks, and following rock falls, may be the result. Those rock falls may impact lower situated talus slopes (Schoeneich et al., 2011). Furthermore, the talus production provides a higher amount of available, erodible material for debris flows. Together with the changing precipitation regimes, this may lead to an increasing frequency of debris flows (Marchi et al., 2009). However, if there is no permafrost less, the ground drains and the threshold value of precipitation to start a debris flow may increase (van Steijn, 1996)

11

Geomorphological process Examples of studies

Kenner et al., 2017 Changes in frost weathering and talus production Stoffel & Huggel, 2012

Thermokarst Kenner et al., 2017

Ikeda & Matsuoka, 2002 Changes in the rate of rock glacier creep Bodin et al., 2017 Giaccone et al., 2016

Changes in solifluction rates Matsuoka et al., 2005 Changes in the frequency and magnitude of mass Stoffel & Huggel, 2012 (rock fall) movement events Marchi et al., 2009 (debris flow)

Stoffel & Huggel, 2012 Changes in the volume of unstable materials Ravanel & Deline, 2011

Table 2: Possible influences of permafrost degradation on geomorphological processes

3. STUDY OBJECTIVES

The purpose of this study is to understand the influence of permafrost distribution on talus slope geomorphology. The key question of this research is ‘How does permafrost distribution influence the talus slope geomorphology on the Col du Sanetsch, Switzerland?’. To accomplish this, we defined some more specific objectives: First of all, a geomorphological mapping is made. In addition to this, measurements of talus slope dynamics, ground surface temperature and the current permafrost distribution are done. Afterwards, we will try to relate those factors to explain the permafrost distribution and related processes. In the last phase the geomorphological and permafrost mapping are combined to answer the research question. Out of the literature review and the research objects, some hypothesis are set to be tested. First off all permafrost is expected in the rock glacier, the landslide and eventually in blocky surface layers along the talus slope. Furthermore, we assume that the displacement of the landslide will be related to the distribution of permafrost and the related factors such as GST and moisture content.

4. STUDY SITE

The case study that will be examined is Col du Sanetsch (46°20’22.60” N, 7°18’41.02” E), located in the western Swiss Alps, at the northern side of the Rhone valley, nearby Sion (Commune Savièse – VS) (Map 1). The slope is oriented towards the northwest with an elevation ranging from 2340 to 2700 m.a.s.l. The talus slope, which mainly consist of limestones (Federal Office of Topography, SGTK, 2012), is located at the base of the rock wall between the Arpelistock (3036 m.a.s.l) and Arpelihore (2921 m.a.s.l) in the Sanetsch – Wildhorn Massif. In the south, the border of the talus slope is defined by the highly erodible badlands. Below the talus slope Late – Pleistocene, Holocene moraine deposits can be found (Federal Office of Topography, 2017). The average inclination of the talus slope is

12 approximately 30 – 40 °, with a basal concavity. In the upper region slopes up to 45° can be found, while at the base the inclination approaches 25°. The 0°C isotherm in this area is situated around 2300 m.a.s.l (Lambiel, 2006). In this context, the talus slope is located within the belt of discontinues permafrost (Scapozza et al. 2011). A series of interesting landforms, such as a landslide, rock glacier and debris flows can be found on the site (Map 2). The tong-shaped rock glacier is formed by two convex ridges and covered with a mix of medium sized to coarse grained boulders. Its present suggest the existence of discontinues permafrost. The landslide is covered by a mix of soil and small to large boulders. To determine the surface velocity and deformation, a network of measuring points was installed in 2011. Since 2014, continuous ground surface temperature (GST) data is available. However, there is no weather station where long term temperature and precipitation data can be obtained. The weather station ‘Les ’ is located at 6 km, but in a different context, at 2964 m.a.s.l. on a glacier, and without long term data. The weather station in Sion, 14 km to the south, is located in the valley at 482 m.a.s.l. and is therefore also not completely representative for the study area. At this station the mean annual precipitation from the period 1981 – 2010 is 603 mm (MeteoSwiss, 2016). The mean annual temperature is 10,1°C. Figure 5 visualise the general meteorological context.

Figure 5: General meteorological context: long-term means of monthly mean temperature, monthly maximum and minimum temperature as well as monthly precipitation sums in Sion (MeteoSwiss, 2016)

13

Map 1: Location study site: Col du Sanetsch

46°21’00” N

46°20’34” N

46°20’08” N

7°17’36” E 7°18’03” E 17°18’29” E 7°18’55” E

Map 2: Location talus slope and different landforms

5. MATERIAL AND METHODS

The purpose of this research is to understand talus slope geomorphology, and more specific the geomorphology of the landslide in relation to the permafrost distribution. This chapters provides an overview of the data and the applied methods. Most of the data was collected during a field work campaign in August 2018. The gathered data can be divide into three groups: geomorphological, geophysical and meteorological data. The geomorphological data exist of UAV (Unmanned Arial Vehicle) photographs used to build a Digital Elevation Model (DEM), field observations and a geodetic network to measure the surface velocity. The temperature data exist of hourly measurements of the ground temperature. Those different data sources (Table 3) are brought together to analyse eventual relationships and explain the talus slope geomorphology.

Description Period

DEM

LiDAR – 2m – CH1903+/ LV95 2013

UAV imagery – 6,6 cm – CH1903+/ LV03 Summer 2018

UAV imagery – 10 cm – CH1903+/ LV03 Summer 2017

Differential GNSS

27 points + 4 control points on landslide – CH1903+/ LV03 2011 – 2018 (every year)

Meteorological data

Tloggers_DebrisFlow 01/09/2016 – 12/08/2017 (every hour)

Tloggers_RockGlacier 01/09/2016 – 23/08/2018 (every hour)

Tloggers_TalusSlope 12/08/2017 – 1/07/2018 (every hour)

Tloggers_Landslide (PERMOS) 18/08/2014 - 18/08/2017 (every hour)

Temperature and precipitation data Sion 01/2011 – 11/2018 (every month)

Geophysical data

Lateral and vertical ERT profiles on rock glacier August 2016

Lateral and vertical ERT profiles on three locations along August 2018 landslide

Two Schlumberger VES profiles in blocky surface layer on August 2018 the talus slope

Schlumberger and Wenner VES profile on landslide August 2018

Potential permafrost distribution maps

Alpine Permafrost Index Map (APIM) Boeckli et al., 2012 2012

Map of Potential permafrost distribution (FOEN, 2005) 2005

Table 3: Overview data

5.1 Topographic survey based on field observations and UAV

The first important purpose is the creation of a geomorphological map (Map 3). The different features were mapped in Qgis 3.4.5. The visualisation is done in Arcmap 10.6.1. The mapping itself is a combination of two different methods: Field observations and a DEM created by photographs from an UAV. Articles of Hendrickx et al. (2019) and Westoby et al. (2012) provide more detailed information about the principles of this so-called, ‘Structure-from-Motion’ photogrammetry.

16

5.1.1 Field observations

Equipped with a handhold GPS (Garmin eTrex 20), the landslide and talus slope where explored. During this field campaign, waypoints of interesting features, e.g. ridges, snow patches, proofs of sorting etc. where mapped. A description was written down and additionally pictures and sketches were made. In order to make it possible to compare pictures and get an idea of the grain sizes, the same pen is visible on every picture. This information is complementary to the DEM and provides information to interpret or clarify observations on the DEM.

5.1.2 Data acquisition with UAV

5.1.2.1 UAV

To obtain a high resolution DEM of the study site, we did an UAV survey. The photographs were taken using a 16MP Panasonic Lumix DMC-GM5 with a 12-32 mm F3.5 – 5.6 lens. The focal length was fixed on 20 mm, but there was also a flight on 12 mm. The shutter speed varied between 1/400 and 1/800 second, the ISO between 400 and 800 depending on the light conditions. The camera was fixed on a custom-made Hexacopter DJI F550 with a Pixhawk flight controller. The flight speed was 4 – 6 m.s-1. To establish sufficient overlap between the pictures, the acquisition interval was set on 1 second. The ground sampling density was 1,64 cm.pixels-1. Eleven flights were flown, from which eight were used to create the DEM.

The preparation of this survey consist of two different tasks. The first one to be done was the flight planning. Therefore, flight lines were drawn in QGIS. These flight lines are parallel to the contour lines of the area. In this way, the flight height, which is approximately 90 m, is adjusted to the topography. The flight lines were translated into a waypoint-file, with flight direction, using a Python script. In the field, these files are imported in the program ‘APM Mission Planner’ and written to the UAV. The other preparation task is done in the field. Six ground control points (GCP’s) are distributed over the talus slope and landslide. They are needed to reference the model to a real-world system. It is important that these GCP’s are easy to identify on the pictures, so they need a strong contrast and clear centre (Figure 6) (Hendrickx et al., 2019). On the lower places we used plates (Figure 6a). These ground control points are fixed on an iron pin, which is drilled on a stable rock. The clear centre of these plates is an advantage, but the weight of 1 kg makes it inefficient to use them on higher locations. Furthermore, in the upper part of the talus slope, there are no stable rocks where we could install the plates. As a result we used cloths (Figure 6b). The centre of these GCPs is less clear, but we can carry them in our backpacks and they weigh less. Since the accuracy of the GCP’s defines the accuracy of the final DEM, it is important that the GCP’s are measured with a high precision. We used a RTK-GNSS to perform these measurements.

5.1.3 Real-Time Kinematic GNSS (RTK-GNSS)

An RTK-GNSS improves the accuracy of the measurements by using two different receivers. One fixed receiver, which is used as reference, and assumed to be motionless and a second receiver, the rover,

17 which is mobile. The first receiver determines the difference between the received position and its’ known position. These differences will be used to correct the measurements from the second receiver. The rover, second receiver, measures the position of the different points on the landslide. These receivers communicated permanently and the reference receiver sent the correction immediately to the rover (Wee, 2018). To control the accuracy of the measurements, the average of five to ten measurements is retained (Wee, 2018). This method allows to reach a precision up to 3 cm. However, these final measurements need to be corrected. The first correction converts the altimetry data from the Bessel ellipsoidal reference to the geoidal reference. This is done by the Swisstopo platform (https://www.swisstopo.admin.ch/fr/cartes-donnees-en-ligne/calculation-services/reframe.html). In addition, by measuring a geodetic point in the area, we found out that the location of the base station, the first receiver, is not exact. Therefore, a second correction is needed. To get the correct absolute coordinates following corrections are done: easting (-1,438), northing (-1,117) and the altitude (+2,976).

b)

a)

Figure 6: Installation and measurement of GCPs. a) a fixed GCP, b) a ‘cloth’ GCP

5.1.4 Data processing in Agisoft PhotoScan

The obtained UAV photographs were imported into the program Agisoft PhotoScan 1.2.6 to compose the final DEM. This program uses the principles of the Structure-from-Motion photogrammetry, a technic which builds 3D structures from a series of overlapping images. Contrary to the conventional photogrammetry, there is no need to specify a priori network of targets with known 3D-positions. To estimate the camera positions and object coordinates, matching features on multiple images are tracked. More information about this technic can be found in the article of Westoby et al. (2012). The workflow followed to create the DEM can be found in Figure 7a. On the first try, one out of four photographs were imported. Afterwards, we added more photos in the zones were there where gaps

18

(Figure 7b) or where the alignment did not work. We only managed to align eight of the eleven flights. As a result, the upper part of the talus slope is not covered. Furthermore, there is also a small data gap in the southern part of the landslide. Several runs were executed to investigated the problem. These runs varied in the amount of photographs they included or the settings used. There were tries in which, all photographs were imported within one single chunks and runs with one chunk for every flight or even flight line. Even though these problems could not be encountered. As we used the same methodology on the whole area, the question rise what the difference is between the upper region and the other areas. Between the flight lines, there is not only a horizontal displacement, but the UAV also ascends. We presume that the overlap problem can be related to this ascent and has its origin in the steepness of the area. Too much ascending, resulting from the steepness in the area, between the flight lines can cause an insufficient overlap. Adding some flight lines is something which can be tried during the next field work.

Following Hendrickx et al. (2019) we run everything on medium quality. In this article, the performance of ten medium and high quality run are compared. The differences were very small, but the variability for the medium quality run is more predictable. Furthermore, the variations in DEM and curvature also show a better performance. Considering those facts and the computational effort calculation time, a medium quality run is more efficient. In the final model the GCPs have a RMS error of 10,9 cm. The used coordinate system is CH1903+/ LV95 (EPSG: 2056). The resolution of the achieved DEM is 6,6 cm. In Table 4 the technical details of our UAV survey and the resulting DEM can be found. a) b)

Figure 7: a) Workflow in Agisoft to build a DEM based on Hendrickx et al. (2019). b) A model after the first alignment. In the upper zone, different gaps are visual, here we added more photographs.

19

Number of flights 8 Number of points 1 779 247 Flying altitude 90 m Ground sampling density 1,64 cm.pixel-1 Number of images 890 RMS of GCPs 10,9 cm Covered area 0,408 km² RMS of check points 84,9 cm Coordinate system CH1903+/LV95 Resolution DEM 6,6 cm (EPSG: 2056)

Table 4: Technical table UAV survey

5.1.5 Geomorphological mapping based on DEM interpretation

The geomorphological map is based upon the DEM. As we did not want to visualise every single small rock, we first applied a moving window (15x15) on the DEM. In the resulting ‘smoothed DEM’, every cell represents the mean height of a square of 99 x 99 cm (Arcmap > focal statistics). From this layer we derived aspect, slope, surface roughness and tangential curvature (Qgis > Saga). Each of these layers give some extra insight in the geomorphology of the landslide and talus slope. By combining the layers, different features were highlighted and identified. The northern part of the talus slope was not covered by our DEM, in this area the DEM from 2017, made by Hanne Hendrickx was used. In the area above the landslide, where there is no DEM from 2017 or 2018, LiDAR data was used to close the data gap. It is important to keep this in mind as the parts mapped on the LiDAR data are less detailed e.g. on the rock glacier it was not possible to map the small scale geomorphology. A hillshade and contour lines derived from LiDAR data provide the background of the map. The focus of the geomorphological map lies within the explanation of the different processes and displacement. This formed the starting question to analyse the data and map all the different landforms. Source areas, accumulation zones and proves of (earlier) displacement are visualised.

5.1.6 Defining topographic roughness

Topographic roughness is calculated as it can be used as a proxy for grain size (Otto et al., 2012). Grohmann et al. (2011) compares six different methods to define topographic roughness within the field of geomorphology. In this research, we choose the standard deviation of residual topography (SDrestopo). This method is also applied in other researches in mountain areas (e.g. Otto et al., 2012; Haneberg et al., 2005) and filters out the large-scale topography, which is an advantage. Furthermore, it is derived from the DEM itself, so small errors in the DEM will not be enhanced, as can be the case if we use a derivative such as the standard deviation of slope (Grohmann et al., 2011; Haneberg et al., 2005).

Finally, it is also easy to produce the SDrestopo using raster GIS algebra (Haneberg et al., 2005).

Calculating the roughness index is done in three different steps (Figure 8a). In the first step, a moving- window of 15 x 15 is applied on the high resolution DEM. This results in a ‘smoothed DEM’ in which every cell gives the average height over approximately 1 m². This resolution was chosen as a compromise. If we used a smaller resolution, more big boulders would be seen as normal topography and only the edges would be highlighted. By using a higher resolution small depressions and ridges could be smoothed, which would result in a topographic roughness value higher than the effective

20 roughness. It is important to keep this in mind, when analysing the map. Afterwards the residual topography is calculated by subtraction of this ‘smoothed’ DEM from the real DEM (Figure 8b). In the third step we used another moving-window (15 x 15) to calculate the standard deviation.

a) b)

c)

Figure 8: a) Flow chart to calculate SDrestopo (source: Cavalli et al., 2008). b) Example of the residual topography, the difference between the LiDAR DTM and mean DTM results in the residual topography (source: Cavalli et al., 2008). c) The influence of the moving-window size on the surface roughness (source: Grohmann et al., 2011).

For some of the statistics, surface roughness was divided into three classes (Table 5). The borders of these classes depend on literature and own classification using photographs. Otto et al. (2012) defined a topographic roughness from 0,01 m – 0,02 m as fine grained, 0,03 m – 1 m as coarse grained. To subdivide the coarse grained, one representative photograph for every class was selected (Figure 9). Afterwards, other photographs were given a code (1 = fine grained/ soil, 2 = medium grainsize, 3 = coarse grained) based on the comparison with the example photographs. With the aid of the coordinates, taken with a handhold GPS (Garmin eTrex 200), the photographs were located and the

SDrestopo value was distracted. As these coordinates are not exact, another moving-window was applied on the SDrestopo raster. We took a look at the maximum value in the environment (1 m²). From all the gathered information, the borders of the three categories were set.

21

Class Description SDrestopo (m)

1 Soil/ fine grained Soil and/or vegetation visible, mixed with small boulders < 0,025 (< 20 cm)

2 Medium grained Soil and/or vegetation not visible, boulders (< 1m) 0,025 – 0,1

3 coarse grained Big boulders (+ 1m) > 0,1

Table 5: Topographic roughness classes

a) Soil, fine grained b) Medium grained c) Coarse grained

Figure 9: Topographic roughness classes based on photographs

5.2 Surface velocity measurements on the landslide

The presence of permafrost can result in the internal displacement and downslope movement of landforms. In order to get more insight in these processes, the members of the alpine geomorphology research group at Fribourg university designed a network of 32 points to monitor the deformation and surface velocity of the landslide. These points are marked with paint. 28 of the points are located on the landslide (Figure 10). The four other points act as control points, and are located on nearly immobile locations in the environment of the landslide. Since 2011 there is an annual measuring campaign at the end of August. During this campaign, the position of each point is measured using a RTK-GNSS. To calculate the effective displacement rate of every point triangulation is used. The result is a dataset with the annual displacement rate from August 2011 – August 2018.

22

¯

Figure 10: Overview data points: temperature and surface velocity (source: Google earth, 2016)

5.3 Temperature measurements

As temperature is one of the factors which influences permafrost distribution, it is important to assemble this data. Hourly temperature measurements are gathered using t-loggers HOBO U23-001, HOBO

U23-003 and UTL-3 (Table 6). The t-loggers are installed just below the surface. Consequently, they measure the ground surface temperature. From the raw data, several metrics were derived: the Mean Annual Ground Surface Temperature (MAGST), the monthly means, the WEqT and the melting period or zero curtain.

Table 7 and Figure 10 give an overview of the different t-loggers, their location, measuring period and the responsible party. Within this research the focus lies on the landslide. Therefore, we mainly use data collected by t-loggers arp-005, arp-010 and arp-017. The t-logger arp-005 is in fact located above the landslide. The data from the t-loggers on the other landforms mainly serve to make the comparison. However, the t-loggers on the talus slope have two other purposes. They are located in a way that we can investigate the influence of grain size and the possible existence of air flow. The influence of grain size will be examined by comparing the two t-loggers located in the coarse grained surface layer (AT – GST – 1, AT – GST – 2) with the one in the fine grained surface layer (AT – GST – 3). We expect to find colder temperatures in the coarse grained surface layer. The contrast between the upper t-logger (AT – GST – 1) and lower t-logger (AT – GST – 2) will make it possible to explore the possibility of an air flow in the coarse grained surface layer.

UTL – 3 HOBO U23 – 001 HOBO U23 - 003

Temperature range -30°C – 40°C -40°C – 70°C -40°C – 100°C

0,1°C at +/- 20°C 0,21°C from 0°C – 50°C 0,21°C from 0°C – 50°C Accuracy 0,2°C at +/- 40°C

Data resolution < 0,1°C 0,02°C at 25°C 0,02°C at 25°C

Table 6: Technical specifics of the t-loggers

Additional to this data, the temperature of the different springs at the foot of the talus slope were measured. On 17 August 2018, the springs were measured using an HOBO U23-003 logger. The temperature was logged every five minutes, as a result the loggers were put for at least six minutes in the springs.

t-logger Latitude Longitude Landform Altitude Period Responsible

A Arp-005 46°20’14.98” N 7°18’40.14” E Landslide 2641 m 18/08/14 – 24/08/18 1

B Arp-010 46°20’16.87” N 7°18’29.91” E Landslide 2519 m 18/08/14 – 24/08/18 1

C Arp-017 46°20’20.64” N 7°18’24.87” E Landslide 2448 m 18/08/14 – 24/08/18 1

D AT-GST-1 46°20’59.22” N 7°17’48.78” E Talus slope 2462 m 12/8/17 – 1/7/18 2

E AT-GST-2 46°21’0.39” N 7°17’47.54” E Talus slope 2400 m 12/8/17 – 1/7/18 2

F AT-GST-3 46°20’57.56” N 7°17’51.52” E Talus slope 2507 m 12/8/17 – 1/7/18 2

G GST-9-RG 46°20’23.12” N 7°18’42.96” E Rock glacier 2639 m 1/09/16 – 28/07/18 2

H GST-13-RG 46°20’22.96” N 7°18’38.34” E Rock glacier 2595 m 1/09/16 – 28/07/18 2

I GST-15-RG 46°20’21.00” N 7°18’41.55” E Rock glacier 2639 m 1/09/16 – 28/07/18 2

J GST-16-RG 46°20’22.03” N 7°18’42.65” E Rock glacier 2643 m 1/09/16 – 12/08/17 2

K GST-7-DF 46°20'55.05"N 7°17'55.34"E Debris flow 2211 m 1/09/16 – 12/08/017 2

L GST-3-DF 46°20'59.22"N 7°17'48.78"E Debris flow 2142 m 1/09/16 – 12/08/17 2

M GST-14-DF 46°21'0.39"N 7°17'47.54"E Debris flow 2126 m 1/09/16 – 12/08/17 2

Table 7: Overview of the t-loggers on Col du Sanetsch (1 = Fribourgh University, 2 = Ghent University)

5.4 Permafrost mapping based on geophysical measurements

The last important group of data that we need acquire in order to answer our research question is the distribution of permafrost. As permafrost is invisible, it is hard to know the exact spatial distribution. In this thesis the geophysical characteristics of the ground are used as an indicator. Two different techniques, complimentary to each other, are used: VES and ERT. In the next paragraphs, the different techniques and the data processing steps are explained. In addition we will clarify the chosen transect locations.

Both ERT and VES are based on the difference in electrical resistivity of the ground. As Archie’s law expresses, this does not only depend of the nature of the rock, but also on the resistivity of the imbibition fluid, the porosity and the degree of saturation (Scapozza & Laigre, 2014). The subsurface exist of different sublayers. Therefore, the measured resistivity will become the apparent resistivity. This is an integration of the resistivity of all the layers crossed by the electrical field (Scapozza & Laigre, 2014). A

25 decrease in temperature, results in a rise of electrical resistivity. This follows a linear relationship above 0°C. Below 0°C it becomes exponential (Figure 11a). An increase in the amount of ice content will also result in a rising resistivity (Scott et al., 1990).

The main advantage of the 1D VES technique is the weight of the materials and the ease to process and interpret the results. Contrary to the ERT, the VES equipment is lighter and no specific software is needed to process the data, only excel. On the other hand, ERT is a 2D technic, in this way it gives more complete information. To work as efficient as possible with the ERT, VES transects where carried out to explore the site and decide the spacing of the electrodes for ERT.

For both methods, the main principle is the same. There are two current electrodes (A & B) and two potential electrodes (M & N), those four electrodes form a quadripole. An electric current, I, is generated between A and B. This creates a semi-spherical electrical field (Figure 11b). The volume of this electrical field depends on the distance between A and B. By adding the two potential electrodes, the difference in potential, ∆푉, can be measured. The use of the resulting quadripole, allows to calculate the ground apparent resistivity 휌푎 (Scapozza & Laigre, 2014):

∆푉 퐴푀 . 퐴푁 휌 = ( ) 푘, with the geometric factor 푘 = [ ] 휋. 푎 퐼 푀푁

Since k builds upon the distances between the different electrodes, it depends on the used electrode array (Figure 11c) (Scapozza & Laigre, 2014). An electrode array defines the relative distance between the electrodes of a quadripole. There are different possible electrode arrays, e.g. Wenner, Schlumberger, dipole-dipole etc. Within a Wenner array, the electrodes will be placed (VES) or selected (ERT) in a way that the distance between every electrode is the same. When using a Schlumberger array, the distance between AN and MB is a multiple of the distance between the potential electrodes, N and M (Figure 12).

26

Figure 11: Principles of ERT prospecting. a) Relation between the resistivity and the temperature. b) An example of the electrical field created by inserting I into the quadripole. c) The most used ERT electrical arrays. (Source: Scapozza & Laigre, 2014)

5.4.1 VES

The VES measurements were executed with a McOhm resistivity meter, model-2115. Both the Schlumberger and the Wenner array were used.

The Schlumberger array gives the opportunity to measure the resistivity underneath the centre on different depths. The four electrodes are placed in line around the centre (Figure 12). The current electrodes, A and B, are moved outward every measurement. Meanwhile, the potential electrodes, M and N, stay in the same position. When the observed voltage becomes too small to measure, they are also moved outward (Sharma, 1997). The Wenner array, on the other hand, is used to determine the resistivity of the soil on a fixed depth, along a transect (Sharma, 1997). This depth depends on the distance between the electrodes. In the Wenner array, the electrodes are spaced equidistant from each other (Figure 12). As a rule of the thumb, the investigated depth equals a * 0,75 with a=AM=MN=NB (Wee, J. personal communication, August 2018). However, the effective depth also depends on the subsurface layering. A conductive surface layer will reduce the investigation depth (Samouëlian et al., 2005). To decide the spacing, we first executed different Schlumberger arrays. The results suggested permafrost around 5 m depth. These observations led to the decision to take a spacing of 6 m for the VES measurements using a Wenner array. The resulting investigation depth equals 4,5 m.

27

Figure 12: Relative position of the electrodes in a Wenner (a) and Schlumberger array (b), VES (Source: Omasanya et al., 2014)

To carry out the measurements, the four cables are connected to the ohm-meter. Using steel stakes, around 30 cm long, the current was injected into the ground. If needed, a sponge soaked in salt water was used to improve the galvanic coupling between the electrode and the surface (Marescot et al., 2003). On the terrain, metadata, such as weather circumstances, coordinates of the centre, the grain size around the central point and the direction of the transect were recorded. Complementary to this metadata, photographs were taken. The measured values were noted on a template (attachment 1 and 2). To get accurate information, every measurement was carried out two times. In the Schlumberger array, a current electrode C was planted 200 m away from the centre, perpendicular on the transect.

Since, ∆푉퐴퐵 = ∆푉퐴퐶 + ∆푉퐵퐶, the measurements AC and BC can be used to control the measured ∆푉퐴퐵. To visualise and analyse the assembled data, graphics are made using excel.

5.4.2 ERT

5.4.2.1 Fieldwork

To perform the ERT measurement, a SYSCAL R1+ Switch 48 was used. On the field, the different coordinates from the centre, first and last electrode were taken with a handhold GPS (Garmin, eTrex 20). The topography was mapped using an inclinometer. Furthermore, descriptions of grain size and materials along the transect were written down, enhanced by photographs.

The ERT measurements were carried out using a Wenner-Schlumberger array (Figure 11c). This array has a slightly lower investigation depth, but a better horizontal resolution of the subsurface structures, then the Wenner array. The dipole-dipole array has an even better horizontal resolution, but the lowest investigation depth (Scapozza & Laigre, 2014). We chose the Wenner-Schlumberger array, as it is the best compromise on depth, resolution and running time. For all the possible quadripoles within the Wenner-Schlumberger array, the apparent resistivity is automatically measured. With 48 electrodes, this results in 529 measurements. Those are divide over 23 data levels. Figure 13 visualize the concept of

28 the different measurements and data levels for a Wenner-Schlumberger array with 33 electrodes. On the Col du Sanetsch, two electrode cables were used, each of them having place to connect 24 electrodes. In total 48 electrodes were placed on every transect. These were the same steel stakes as used to execute the VES measurements. Before executing the measurements, the connection of the electrodes and the ground is checked by using the ‘RS-check’ function. When needed the ground around the stakes is wetted with salt water, or the stakes are replaced by sponges soaked in salt water. This to improve the connection. Since the VES measurements indicated 5 m as the most interesting depth to find permafrost (Figure 27), the spacing was 2 m. This gives a maximum penetration depth of approximately 15 m and a good resolution in lower depths.

Figure 13: concept of the Wenner-Schlumberger array (Source: Ekinci et al., 2013)

5.4.2.2 Processing data

The obtained apparent resistivity gives us a rough idea of the resistivity distribution. To improve this and calculate the specific resistivity, an inversion of the data is needed. More information about this inversion and why it is needed can be found in different articles (e.g. Loke, 2002; Scapozza & Laigre, 2014). The software used to execute the inversion was Res2DInv. There are different methods, but all have the purpose to minimize the differences between the measured and modelled data (Scapozza & Laigre, 2014; Barandun & Hilbich, 2013). Scapozza & Laigre (2014) compared the different methods. The robust method creates a model with clear and linear borders. This allows the best resolution for regions with steep transitions of resistivity such as Alpine periglacial landforms. The method starts with a first homogeneous model of the underground. Then, the model is improved by comparing the model and the field data. This is done several times and each of these iterations will decrease the RMS error exponential. After some iterations the difference in RMS error becomes really small and no further improvement can be expected. As a rule of the thumb, this should be after 3 to 5 iterations. In a terrain with a high level of noise, as a periglacial talus slope, a RMS error under 10 % is useful, under 5 % is optimal (Barandun & Hilbich, 2013; Scapozza & Laigre, 2014).

29

To transform the apparent resistivity into the effective resistivity, different step are taken. First the topographical data needs to be added to the data file. This can be done using excel and WordPad. In excel, the height of every electrode is calculated using the data from the inclinometer and triangulation. The result is a list with every electrode and the corresponding height. This list can be added to the data (.dat) file. Afterwards, the data is imported into the software Res2DInv. The next step is to control and eventually improve the quality of the data. Topographical data can be verified visual. To check the apparent resistivity, the relative variation are visualized (Figure 14). Bad data points will be shown as unusual high or low values (Loke, 2002). They can be removed, but caution is needed. At some places, for example the surface, more variation is normal. (Barandun & Hilbich, 2013). The bad data points can be caused by different kinds of failure during the survey, e.g. breaks in a cable, poor ground contact, cables which are connected in the wrong way etc. (Loke, 2002). Bad data points can also be identified in the RMS error statics window (Figure 15) after a first preliminary inversion. Points with significant higher errors can be bad data points. For every profile, both methods were used to control the data. In Table 8 the final amount of data points can be found. A transect, where no data points where removed has 529 measurements.

Figure 14: Relative differences in apparent resistivity for transect CdSLS1a. Bad data points (green) can be selected and removed.

30

Figure 15: RMS error statistics of CdSLS1a after a preliminary inversion. By moving the green line, bad data points can be selected and removed.

After cleaning the data points, the inversion can be performed. Res2DInv is a program created to ask minimal input from the user. There are several settings, e.g. damping parameters, that can be used to improve the outcome, but in most cases the default parameters give a reasonable result (Loke, 2002). More information about all different options can be found in the course notes from Loke (2002). In this research the default settings were used. Every transect is visualised in two different ways: once with an automatically defined colour scheme (logarithmic colour intervals) and once with user defined contour intervals. The latter makes it possible to easily compare the different transects.

5.4.3 Transects

Table 8 gives an overview of the performed transects, and their metadata. The location can be found on Figure 16. The locations of the transects are chosen based on different arguments. In the first place, the purpose was to have a good distribution over the different parts of the landslide. Secondly, they are located in the proximity of a t-logger. This makes it possible to combine the datasets. Normally the VES measurements with a Wenner array were meant to explore the site and identify interesting places to perform an ERT transect. However, because of some logistical problems, among which two broken cables, only one transect was executed. In the end, the exact location of the ERT transects depended on some practical issues, such as big boulders. To avoid this boulders, the transects were moved one or two meters. Because of the limited amount of time, we only did two VES transects on the talus slope. The focus of this research lies within the landslide, but they can be used to compare the talus slope and landslide. Finally, there are also two ERT transects from the rock glacier, these were measured during the field campaign in summer 2016, by Adrian Emmert (Würzburg University).

31

No. of RMS No. Name VES/ERT Spacing Length Array data error points (%)

Landslide location 1 – t-logger 017 – down

1 CdSLS1_VES_S VES 160 m Schlumberger 19

Wenner- 2 CdSLS1a ERT 2 94 m 508 11,45 Schlumberger

Wenner- 3 CdSLS1b ERT 2 94 m 524 6,29 Schlumberger

Landslide location 2 – t-logger 010 – mid

4 CdSS2_VES_S VES 400 m Schlumberger 23

Wenner- 5 CdSLS2a ERT 2 94 m 529 4,26 Schlumberger

Wenner- 6 CdSLS2b ERT 2 94 m 529 7,58 Schlumberger

7 CdSLS2_VES_W VES 6 120 m Wenner 23

Landslide location 3 – t-logger 005 – up

Wenner- 8 CdSLS3a ERT 2 94 m 527 4,83 Schlumberger

Wenner- 9 CdSLS3b ERT 2 94 m 529 8,11 Schlumberger

Talus slope

10 CdSTS1_VES_S VES 300 m Schlumberger 22

11 CdSTS2_VES_S VES 250 m Schlumberger 21

Rock Glacier

Wenner- 12 CdSRG_a ERT 2 70 m 4,1 Schlumberger

Wenner- 13 CdSRG_b ERT 2 70 m 5,0 Schlumberger

Table 8: Metadata for the different transects

32

¯

Figure 16: Overview geophysical measurements (Source: Google earth, 2016)

5.4.4 Defining permafrost probability classes

To analyse the data, every electrode was given a code, from 1 to 3, which represent the permafrost distribution at 5 m depth (Table 9).

Permafrost description Resistivity values class

1 No permafrost < 1500 Ω.m

2 Possible permafrost 1500 Ω.m – 5000 Ω.m

3 Probable permafrost > 5000 Ω.m

Table 9: Permafrost probability classes (based on Loke, 2002; Lambiel & Pieracci, 2008; Hilbich et al., 2008; Russil, 2000 and field campaign)

The three different categories are based upon literature and field observations. Most researchers report permafrost when resistivities are at least 8 kΩ.m (e.g. Scapozza et al., 2009; Kneisel, 2004; Lambiel & Pieracci, 2008, Otto et al., 2012). However, the exact borders depend strongly on the site: lithology, ice content and ground temperature will all influence the resistivity values (Lambiel & Pieracci, 2008). For instance, Hilbich et al. (2008) reported permafrost at the Schilthorn from resistivities around 2000 Ω.m. Russil (2000) executed some VES measurements in the environment of our study site Col du Sanetsch. He identified very humid, frozen soil with values of 600 Ω.m (Delaloye, 2004). Lithological, the talus slope exists mainly of limestones (Federal Office of Topography, SGTK, 2012). Those have a theoretical resistivity range of 90 – 8000 Ω.m (Figure 2). This theoretical resistivity corresponds to the resistivity of the bedrock measured by VES: Around 25 – 30 m depth, the resistivity values become more or less constant, namely 100 – 300 Ω.m (Figure 27). As a result, the hypothesis is set, that the bedrock is located at 25 – 30 m depth. Furthermore, following Lambiel and Pieracci (2008), high resistivity values will be found at blocky surfaces on the talus slope. Underneath, the resistivity decreases. However, sometimes another increase of resistivity can be measured at depth. There are two possible explanations for this increase: The presence of air or permafrost. As a porous layer does not fit with the stratigraphic observations of other talus slopes (Lambiel & Pieracci, 2008), the increased resistivity suggests the presence of permafrost. In this way, the resistivity measured in those layers give an indication of the resistivity values of permafrost.

5.5 Statistical analyses

The purpose of this research is to understand the geomorphology of the talus slope, the existence of permafrost and the interrelationships between the geomorphology and the permafrost distribution. To pursue this, relationships within and between the different datasets were analysed, using the program SPSS Statistics 24. In some cases visualisations are made by graphs in excel or maps in Arcmap. In the following chapter, the executed analyses are listed and linked to the hypotheses. Furthermore, the construction of the different databases is explained.

5.5.1 Analysis of the meteorological trends

The analysis of the meteorological trends is a first exploration of the meteorological circumstances on the talus slope. The temperature analyses are built upon the data collected with the t-loggers. Over the years, a warming trend is expected, however, the dataset from the t-loggers has a too short time window to perform valuable and significant statistical analysis. Therefore, we mainly use visual analyses performed on graphs made in excel. Four different graphs are constructed: MAGST, WEqT, duration of the zero curtain period and the start and end of the zero curtain period. There are no precipitation or snow measurements on the talus slope. To get this information, data from MeteoSwiss is obtained (Federal Office of Meteorology and Climatology MeteoSwiss, 2018). There are different weather stations in the environment: Tsanfleuron, Les Diablerets, and Sion. Only the data from the weather station in the valley of Sion is available. The annual precipitation and winter precipitation are transformed into a graph. We defined the precipitation from November till April as winter precipitation. In this period the air temperature on Col du Sanetsch is below 0°C.

5.5.2 Analysing temporal and spatial differences in surface velocity and their relation with temperature, permafrost distribution and geomorphology

We set the hypothesis that the surface velocity depends from permafrost existence and in this way the thermal regime of the soil and the moisture content. To investigate this hypothesis, different statistics will be executed. In the first place two one sample t-test were performed to explore the temporal and spatial differences in surface velocity. The temporal differences were based on a dataset with the mean displacement over all points per year. To analyse the spatial differences, the dataset exists of the mean surface velocity for every location over the years. The resulting datasets have respectively, 7 and 28 data points.

The interrelationship between temperature and surface velocity was investigated using a one – sample Kolmogorov – Smirnov test. We choose this test over a one sample t-test as we expect a non-linear relationship. To construct the dataset, temperature data from the t-loggers were grouped with the differential GNSS points. We took following clustering rules into account: 1) The distance between the t-logger and the differential GNSS point is less than 100 m. 2) The t-logger and the differential GNSS point has the same category of surface roughness (Table 5) Rule one guarantees that, besides the distance, the difference in height is less than 50 m. This is important as the height will also influence the temperature. The second rule ensure that a different micro- topography has a minimal influence on the results. This is necessary because the ground temperatures are influenced by air circulation and thus grain size. For every location four years of displacement and temperature measurements are connected. This result in a data set with 32 points.

Another executed analysis investigates the interrelationship between the permafrost distribution and the spatial differences in surface velocity. Previous studies set that the ground temporal regime is influenced by permafrost distribution. As a result, this analysis is closely related to the previous analysis between

35 the thermal regime and the surface velocity. However, as we found out that the relation between the temperature and permafrost distribution on Col du Sanetsch is not as straightforward as we expected, this analysis is useful to give more insight. The permafrost distribution on the landslide is discontinuous. In addition, most of the dGNSS points are not located in the nearby environment of the ERT-transects. As a result, it is challenging to link the dGNSS points with the geophysical output. This is why we performed a visual analysis instead of a statistical analysis. The basis of this analysis is a map on which the surface velocity and the permafrost existence are visualized with colours.

In the end we also investigated the relationship between slope and displacement rate. We expect a positive correlation, with steeper slopes related to higher horizontal surface velocities. A first visual analyses is performed on a map, where both the slope and the displacement rates are visualised. Furthermore, a linear regression analyses is done in SPSS.

5.5.3 Analysing the distribution of surface roughness and the interrelation with permafrost distribution

According to Lambiel and Pieracci (2008), an increased grain size can be found downslope a talus slope. Furthermore, creep processes will influence the distribution. In this way, we expect a more mixed distribution on the landslide. This hypothesis is investigated using a one sample t-test between the surface roughness and height. The data was extracted from a 10 x 10 m grid on the DEM on the one hand and the topographical roughness layer on the other hand. The functions ‘create fishnet’ and ‘create points along lines’ in Arcmap made it possible to create the grid. To extracted the data we used the function ‘sample’. There were 2 different datasets, one for the talus slope and one for the landslide. To get these datasets, the former created grid was clipped on a shapefile from the talus slope or landslide.

Furthermore, a blocky surface layer has a negative influence on the thermal regime of the ground. As a result, it is more likely to find permafrost in coarse grained areas. Therefore, the relationship between permafrost and surface roughness was analysed. In the first place, we did a visual analysis based on the field observations and the visualisation of surface roughness on a map. The categories on the map, were based upon the division between smooth and rough surfaces, from Otto et al. (2012). The results from the visual analysis were checked using statistics. We chose to focus on the ERT profiles at location two, to ensure that the height would not influence our outcome. For every electrode, except from the three outer ones, the permafrost category and the surface roughness were determined. As it is not possible to located the electrodes very precisely, the maximum surface roughness within an area of 90 x 90 cm was taken. The resulting dataset contained 84 points. The three electrodes at every site of the transects were not included, as there are no resistivity measurements on a depth of 5 m underneath these electrodes.

5.5.4 Comparing the measured permafrost distribution and permafrost probability maps

In the last analysis we use the outcomes of the geophysical measurements to verify the permafrost probability maps, APIM and the map of potential permafrost distribution in Swiss, on the site. This comparison is done visually by the creation of two maps, with exactly the same area. On the Swiss

36 geoplatform (map.geo.admin.ch), the WMS layer of the APIM and the transects were imported. The transects were coloured using the same colour scheme as the maps, which makes it easy to compare them. The result are two maps with the transects and each containing one of the permafrost maps as a background.

6. RESULTS

6.1 Geomorphological map

The geomorphological map (Map 3) is based upon the DEM derived from the drone campaign in August 2018. To discuss the map, we focus on the system and the relation between the landforms. We divide the landforms in those resulting from gravitational processes and those influenced by snow, frost and (peri)glacial processes.

6.1.1 Gravitational processes

Rockfall is the main gravitational process and is responsible for the general formation of the talus slope. Rock fragments, roll, fall or slide down from the rock wall, which can be identified by the steep slope. On the talus slope they come to rest or move further downward. The main part of this supply is concentrated into different rock channels. On the transition between the rock wall and the talus slope, the channels form a point source of material, and influence the geomorphology of the talus slope underneath. In the first place, the materials do not form a straight sheet talus slope, but several talus cones are identified. Some of these talus cones are clear, for others the border cannot be identified in the lower parts. The upper part of these cones are dark grey (Figure 19), indicating wetter materials

Additional to rock fall, debris flows originating from the rock channels will also rework the talus slope. Several gullies and levees can be found. The most southern debris flow runs down into the vegetation under the talus slope. In the downward part of the depositions, a sequence of gullies and associated levees is observed (Figure 17). Only during heavy rain showers, water is running through parts of this gully. On all the other places on the talus slope, water infiltrates at the border between the rock wall and the talus slope, runs underneath and reappears in the different springs at the foot of the talus slope. These springs are located on the elevation were the grain size decreases and soil and vegetation appears.

6.1.2 Snow, frost and (peri)glacial landforms

Besides the gravitational processes, the talus slope geomorphology is also influenced by processes related to snow, frost and permafrost. Along the talus slope, several small depressions followed by an asymmetric ridge can be found. These are probably influenced by nivation processes (Figure 22). The asymmetric ridges have a steep back slope, which is built by coarse grained materials which fall, roll, or slide down the slope. The front slope exists mostly out of finer grained materials. The ridges can be defined using the DEM-derivative ‘aspect’ from the high resolution DEM (Figure 23). However, to keep

37 the geomorphological map clear, we did not indicate every single ridge. Only the areas, nivation areas, where they occur were identified.

Another landform resulting from periglacial processes is the rock glacier (Figure 20). The rock glacier is located underneath the most outer part of the rock wall. Bellow the rock wall, a depression with a perennial snow patch is found. On the downward part, the depression is bounded by an asymmetric ridge. Nivation processes enhance the depression above the ridge. The rock glacier itself exist of a coarse grained debris layer, with two convex ridges. Furthermore, one lateral arcuate ridge is observed in the southern margin of the rock glacier. Along every side of the ridge, a V-shaped furrow from 1 to 2 m depth is observed (Figure 24). Due to the low resolution of the LiDAR data, it is not possible to map other small scale features on the rock glacier. The presence of the active rock glacier suggest that discontinuous permafrost is represent on the slope.

On the talus slope there is also a landslide present (Figure 18). The landslide can be found on the most southern part of the talus slope, underneath an outcrop of bedrock. On the landslide two different kind of slopes are found. The downward part is lobe shaped, with convex ridges, which is typical for landslides. Contrary, the southern edge is formed by a steep, straight slope. This can be related to the materials of the badlands, which are more prone to erosion. Over time, melting snow and rain incised a deep channel. Along this edge, there are also different springs represent. The springs on the lowest locations were represent during the whole field period. Higher up they only appeared during heavy rain showers. The existence of these springs suggest a less permeable layer in the underground. The landslide has a high surface roughness, certainly when comparing with the talus slope (Map 5). Only in the downward part, where vegetation grows and the soil is visible, the surface roughness is lower (Figure 18). As a result, it is more easy to identify the several transversal and longitudinal ridges. In the higher zones, they are also represent, but due to the high surface roughness these are more difficult to identify. The transversal ridges can be found in two different zones and are proves of previous activity and stress on the landslide.

Finally, between the rock glacier and landslide, two ridges can be identified (Figure 21). We called them structural ridges as they are partly formed by bedrock outcrops. Between them a depression can be found. The snow in the depression, which was still present in July, melted away by the end of August. The hypothesis is set that the long lasting snow enhance the depression by nivation.

38

Map 3: Geomorphological map Col du Sanetsch Incised channel and talus cone

Debris flow Rock glacier Bedrock outcrop

Landslide Badlands

Springs

Figure 19: Panorama of the talus slope

Figure 17: Debris flow deposits and gullies Figure 18: Landslide

Perennial snow patch Bedrock

Front

Figure 20: Rock glacier (dotted lines = ridges) Figure 21: Nivation hollow (dotted lines = ridges)

Figure 22: Nivation zone on the talus slope, small depressions Figure 23: Ridges in the nivation zones Figure 24: Lateral arcuate ridge on the followed by an asymmetric ridge visualised using the DEM-derivative ‘aspect’ rock glacier, as seen from the UAV on a fine resolution DEM

41

6.2 Temperature measurements

In this part the output of the temperature data is discussed. In the first part we focus on the WEqT at different places along the talus slope. Together with the temperature measured in the different springs this can be used as an indicator toward permafrost existence. In the last part, we concentrated on the t- loggers on the landslide. Here we will do a both a zonal and temporal analysis. The location of the t- loggers can be found on Figure 10.

Table 10 gives an overview of the mean WEqT at the different t-loggers. Depending on the location, one, two or four years are included. According to this temperatures, it is probable to find permafrost on two places (RG - 13 - Front and RG - 9 - Up). Two other places have a possibility to find permafrost (RG - 9 - Up and AT - GST - 2). On the landslide temperature data do not suggest permafrost existence. This outcome is enhanced by the temperature of the different springs. Temperatures of springs below 2°C indicate a possibility to find permafrost (Frauenfelder et al., 1998). All the springs had temperatures warmer than 6°C, except from the one underneath AT - GST – 1 and 2, where a temperature of 2,3°C was measured (Map 3). These measurements where done on a rainy day. As a result, the temperature data will not only depend from the melt water and the thermal state of the underground, but also the temperature of the rain water. In this way 2,3°C can indicate permafrost existence. On the talus slope, the highest located t-logger (AT – GST – 3, 2507 m.a.s.l) has the highest temperature. The other two t- loggers (AT – GST – 1 and AT – GST – 2) are situated lower, at 2462 m.a.s.l and 2400 m.a.s.l, but lie within a coarse grained area. As mentioned earlier (2.2.2.4), this favours cooling and the existence of permafrost.

T-logger Landform WEqT (°C) Period A Arp - 5 Up Landslide -1,715 2014 – 2018 B Arp - 10 mid Landslide -1,475 2014 – 2018 C Arp - 17 Down Landslide -1,1075 2014 – 2018

G RG - 9 - Up Rock glacier -1,98 2016 – 2017 H RG - 13 - Front Rock glacier -3,79002 2016 –2018 I RG - 15 - Right Rock glacier -0,4283 2016 –2018 J RG - 16 - Left Rock glacier -0,80041 2016 –2018

D AT - GST - 1 Talus slope -3,12518 2017 – 2018 E AT - GST - 2 Talus slope -2,80057 2017 – 2018 F AT - GST - 3 Talus slope -0,23216 2017 – 2018 Table 10: Mean WEqT (Bolt = permafrost existence probable)

WEqT (°C) MAGST (°C) 0 2,5 -0,5 2 -1 1,5 -1,5 1 -2 0,5 -2,5 0 2014-2015 2015-2016 2016-2017 2017-2018 2014-2015 2015-2016 2016-2017 2017-2018

Arp - 5 Up Arp - 10 mid Arp - 5 Up Arp - 10 mid Arp - 17 Down Mean Arp - 17 Down Mean

Duration zero curtain (days) Start and end data zero curtain 100 29/aug 80 10/jul 60 40 21/mei 20 0 1/apr 2014-2015 2015-2016 2016-2017 2017-2018 2014 - 2015 2015 - 2016 2016 - 2017 2017 - 2018

Arp - 5 Up Arp - 10 mid start Arp - 5 Up Arp - 10 mid Arp - 17 Down Arp - 17 Down Mean end Arp - 5 Up Arp - 10 mid Arp - 17 Down

Figure 25: Climatological graphs based on the output of 3 t-loggers located on the landslide at Col du Sanetsch (August 2014 – August 2018).

The t-loggers on and above the landslide provide data over a period of four years. This makes it possible to analyse both temporal and zonal differences. Figure 25 visualise the WEqT, MAGST and the duration, start and end date of the zero curtain on the different locations. Temporal, the means of every metric have an upward or downward peak in 2015 – 2016. The WEqT is higher than the other periods, in contrary the MAGST is on average lower in this period. The melting period, differs from place to place and year to year.

Zonal differences are also observed on the landslide. The WEqT is in accordance with the height. The highest t-logger (Arp – 5 – up), has the lowest mean WEqT. This is not the case for the MAGST. Here the location in the middle of the landslide (Arp – 010 – mid) has the lowest temperature. As Arp – 010 – mid is always longest covered by snow, this will prevent the soil from warming. As a result, a lower MAGST is measured. The last observation we can make is about the duration of the zero curtain period. Except from 2015 – 2016, Arp – 010 – mid has the longest melting period. The difference in 2015 – 2016 is due to the late start of the melting period compared to the other locations.

43

6.3 Geophysical survey

Based on temperature measurements and geophysical measurements, it is possible to get an idea of the permafrost distribution on the study site. In following chapter we discuss the outcome of the geophysical field campaign. Three different VES measurements were carried out on the landslide, two on the talus slope (Figure 26 until Figure 29). Figure 19, until Figure 35 give the results of the ERT measurements performed. We will first discuss the VES results. Afterwards we will interpret the ERT measurements and compare them with the VES analysis.

6.3.1 VES

The resistivity values measured by the VES suggest permafrost on landslide position two (mid – 2519 m.a.s.l) (Figure 27). An increased resistivity around the depth of 5 m (OA * 0.5) can be observed. At landslide position one (down – 2448 m.a.s.l) (Figure 26), there is also a clear increase at this depth. However, different things indicate that this is due to a measuring error instead of a real increase of resistivity. First of all, landslide position one (down – 2448 m.a.s.l) is located below the theoretical lower limit of permafrost in the Alps, namely 2500 m.a.s.l (Deluigi et al., 2017). Furthermore, it is only one measurement which stands out. The values just above and below have a far lower resistivity value. Besides, we investigated this depth two times, ones with distance MN = 1 m and one time with MN = 4 m. This resulted in resistivity values of respectively 956 Ω.m and a much lower 363 Ω.m. Last but not least, the ERT measurement also do not suggest permafrost at that location (Figure 31). This error might be caused by for example a bad connection between the ground and the electrodes during the first measurement. This is also observed on the VES profile on the lower position of the talus slope (Figure 28).

A second observation is the resistivity at depth. On both transects we find a resistivity around the 200 Ω.m – 300 Ω.m at depth. We observe them from depths around 30 m (landslide position two – mid – 2519 m.a.s.l) and 8 m (landslide position one – down – 2448 m.a.s.l). This resistivity corresponds to the theoretical resistivity of limestone (Figure 2). As a result, we suggest that the bedrock of the site is located at 25 – 30 m depth. At position one (down – 2448 m.a.s.l), these resistivities can be found less deep, around 8 m. The bedrock could be found at this depth, but this is not probable. The ERT profiles at this location, also show a body with a lower resistivity, but around 15 – 20 m another, small increase is observed (Figure 31b). Other causes such as water could also cause this low resistivity values.

By using the Wenner array, the area around the second position was investigated. From just above the centre until 35 m further down, the resistivity values exceeds 2000 Ω.m (Figure 30). We suggest that permafrost can be found in this area. The ERT survey at this location, confirms this result (Figure 32).

On the talus slope, the VES measurements suggest permafrost at the highest location, again at the depth of 5 m (Figure 29). On both the VES transects, the apparent resistivity values are higher than on the landslide. In the upper layers, this can be related to the blocky surface. Transects on the talus slope and position two (mid) of the landslide, have a coarse grained surface layer compared to the lowest

44

transect on the landslide (position one). However, at depth the apparent resistivity of the talus slope is approximately 1000 Ω.m, comparing to the resistivity of 200 – 300 Ω.m of the landslide.

VES Schlumberger: Landslide position 1 VES Schlumberger: Landslide position 2

10 10

.m)

.m) Ω

1 Ω 1

ra ra (k ra ra (k

0,1 0,1 1 10 100 1 10 100 OA (m)) OA (m)

Figure 26: Results from the VES measurement Figure 27: Results from the VES measurement (Schlumberger array) on landslide position 1 (t-logger (Schlumberger array) on landslide position 2 (t-logger 017 – down) (transect 1) 010 – mid) (transect 4)

VES Schlumberger: Talus slope down VES Schlumberger: Talus slope up

10 10

)

.m)

.m Ω

1 Ω 1

ra(k ra(k

0,1 0,1 1 10 100 1 10 100 OA (m) OA (m)

Figure 28: Results from the VES measurement Figure 29: Results from the VES measurement (Schlumberger array) on the talus slope, (t-logger AT (Schlumberger array) on the talus slope, (t-logger AT – –GST – 1, down) (transect 10) GST – 2, up) (transect 11)

45

VES Wenner: landslide position 2 10

1 NE SW

0 -100 -80 -60 -40 -20 0 20 40 (m)

Figure 30: Results from the VES measurement (Wenner array) on landslide position 2 (t-logger 010 mid). 0 = t-logger, ‘-20’ means the position 20 m from the t-logger, direction NE. The red line indicates the location of the ERT profile. (transect 7)

6.3.2 ERT transects

Six different ERT transects were executed during the field campaign in 2018. Table 11 relate the different colours of the profiles with the previously determined permafrost classes. However, high resistivity values do not necessarily point towards permafrost. In a blocky surface layer, high resistivity values result from the air between the boulders. The ERT transects on the rock glacier, visualised in Figure 34 and Figure 35, have another colour scale.

The most interesting location in terms of permafrost existence is location two (Figure 32). Here the resistivity values suggest permafrost. At location one, down the landslide, there will probably be no permafrost (Figure 31). The resistivity values along the profiles of location three (Figure 33) mainly suggest that there is no permafrost. However, it is possible to find permafrost in the north western and the upper part of the slope. In the environment of the middle t-logger, permafrost is possible around the centre and probable 20 m below. In the north western part of the transect CdSLS2b the presence of permafrost is also probable. These findings confirm the earlier described results from the VES measurements. On the ERT profiles the bedrock could not be detected, which is in line with the conclusions on the VES profiles, where the bedrock was identified on 25 – 30 m depth.

Both transects on the rock glacier suggest permafrost. On the lateral profile (Figure 34) permafrost is probable, except for the most outer parts. In the centre, around a depth of 8 – 10 m, resistivity values up to 15 000 Ω.m are measured. These indicate an area with more ice rich permafrost. The vertical profile (Figure 35), also contains permafrost, but the resistivity values are lower. In the central part, permafrost is possible. In the upper part and the area above the front, it is probable to have permafrost. In the front itself, no permafrost is present.

46

CdSLS1a

NE SW

CdSLS1b

SE NW

Figure 31: ERT measurements on landslide position 1 (t-logger 017 – down)

CdSLS2a

NE SW CdSLS2b

NW SE

Figure 32: ERT measurements on the landslide position 2 (t-logger 010 - mid)

47

CdSLS3a

NE SW

CdSLS3b

NW SE

Figure 33: ERT measurements above the landslide, position 3 (t-logger 005 - up)

Colours Permafrost class Resistivity values

No permafrost < 1500 Ω.m Permafrost possible 1500 – 5000 Ω.m Permafrost probable > 5000 Ω.m

Table 11: Colour legend of permafrost classes, ERT profiles on the landslide

Figure 34: Lateral ERT profile rock glacier (Source: Emmert, 2016)

48

Figure 35: Vertical ERT profile rock glacier (Source: Emmert, 2016)

6.4 Annual surface velocity of the landslide

The landslide is slowly moving downward, with annual velocity depends from year to year and region to region. In following chapter, the temporal and zonal difference in displacement rate are described.

Figure 36: Annual horizontal surface velocity on the landslide. Mean of a set of points selected in several sections of the moving landform (Source: University of Fribourg, 2018)

Over the period of seven years, the annual surface velocity of 2012 – 2013 and 2015 – 2016 stands out (Figure 36). Based on a one sample T-test the mean surface velocity in those years is significantly higher. Within the landslide itself, there are also differences in surface velocity. The landslide can be divided in different regions. The part above the landslide and the front are moving slower than average. In the middle there is one part where the surface velocity approaches the average surface velocity of the landslide. The median zone, just above the frontal zone has the highest displacement rate. Figure 36 compares the velocities of the frontal zone and median zone (Map 4). Another striking difference is

49 the variance between the years. All the points which are measured on the landslide have an increased velocity in 2012 – 2013 and 2015 – 2016. In these years the measured displacements are at least doubled compared to the previous and following years. Contrary the points measured above the landslide always have more or less the same values (between 4 and 13 cm).

6.5 Relations between geomorphological characteristics, permafrost distribution, meteorological parameters and surface velocity on the landslide

6.5.1 Surface velocity related to meteorological factors, permafrost distribution and topographical characteristics

The WEqT and displacement rate are positive correlated. Based on a one – sample Kolmogorov – Smirnov test, we could concluded that this relation is exponential (Figure 37). All the fast-moving points are located in the environment of the t-logger with the highest mean WEqT (arp-017 – down – 2448 m.a.s.l). Moreover, for every t-logger the period 2015 – 2016, which had the highest annual surface velocity, had the highest WEqT. As there is no temperature data from 2011 – 2014, it is not possible to control if 2012 – 2013 also had a higher WEqT. Contrary to the WEqT, there is no correlation between the displacement rate and the MAGST. The WEqT mainly depends on the ground thermal regime, in this way, according to our results, the ground thermal regime will have an influence on the surface velocity. The surface deformation is probably the result from both gelifluction and frost creep.

2,5

2

1,5

1

0,5 Annualhorizontal displacement (m) 0 -2,5 -2 -1,5 -1 -0,5 0 WEqT (°C) y = 2,5538e1,948x R² = 0,5816

Figure 37: Exponential relation between the annual horizontal displacement and the WEqT.

Another climatic factor that influence the surface velocity is the moisture content (Matsuoka, 2001; Giaconne et al., 2016). This depends on precipitation, duration of the melting period, eventually melting water from permafrost and impermeable layers. As we only have indirect information about the snow layer and precipitation in the nearby environment, it is difficult to analyse this relationship. The precipitation data from Sion is visualised in Figure 38. It gives us a first idea, but we need to keep in mind that this station is located in the valley. Based on a one sample t-test, we could conclude that 2012 – 2013 and 2017 – 2018 were the only years where the precipitation significant differs from the mean

50 precipitation over the period 2011 – 2018. 2012 – 2013 is also the year with the highest surface velocity (mean value over all the points = 0,73 m). However, 2017 – 2018 had the lowest surface velocity (mean value over all the points = 0,09 m). Furthermore, the high amount of winter precipitation in 2017 – 2018 did not coincidence with a longer melting period. So there will probably be a higher amount of melt water, in a shorter period. Even though the surface velocity was lower, this is opposite to our expectation. On the other hand, the spatial variations in surface velocity are mainly affected by differences in soil moisture. The different springs and low resistivity values on the VES and ERT profiles, which suggest a high water content, can be found around the fast-moving area.

Precipitation Sion (mm) and annual horizontal displacement (m)

700 0,8 600 0,7 500 0,6 0,5 400 0,4 300 0,3

200 0,2 Precipitation(mm) 100 0,1 0 0

2011- 2012- 2013- 2014- 2015- 2016- 2017- Annualhorizontal displacement (m) 2012 2013 2014 2015 2016 2017 2018 Annual precipitation 571,9 639,2 623,5 542,1 553,4 506,8 637,1 (August - July) Winter precipitation 296,8 361,3 267,7 235,5 318,1 228,4 444,2 (November - April) displacement 0,1 0,73 0,25 0,25 0,58 0,11 0,09

Figure 38: Annual and winter precipitation (mm) Sion and the annual horizontal displacement on Col du Sanetsch (m)

To investigate the possible relationship between surface velocity, permafrost distribution and geomorphological characteristics, they are visualised together on (Map 4). Contrary to what we expected, no permafrost is found around the area with the highest velocity. Furthermore, a visual comparison between the slope and the displacement rate, suggest that the fast-moving areas mainly lie within normal to steep environments, whereas part of the slower moving areas are located on more levelled slopes. The measured points above the landslide do not follow this roughly division (Figure 39;Map 4). Here all the points are slower, but the slope is not necessarily more level. Additionally, the linear regression between both variables is also not significant (훼 = 0,1). The last geomorphological characteristic to relate with the displacement rate, are the several transversal ridges on the landslide. These landforms prove the internal deformation and displacement of the landslide. They are observed in the slow moving downward part of the landslide and around landslide location – 2, were permafrost is probable. In the fast-moving area they were not identified.

51

Relation displacement and slope 1,4 y = 0,0093x + 0,0993 1,2 R² = 0,0606 1

0,8

0,6 (m) 0,4

0,2

0 0 5 10 15 20 25 30 35 40 Meanannual horizontal displacement Slope (°)

Figure 39: Relation between the displacement rate and slope (light blue = area above the landslide, dark blue = on the landslide)

Map 4: Thematic map - slope, permafrost distribution and annual surface velocity (1 = slow moving frontal zone, 2 = fast moving median zone)

52

6.5.2 Surface roughness as an explanatory factor of permafrost distribution

6.5.2.1 Distribution of surface roughness along the talus slope and landslide

The grain size of the rock fragments is one of the factors which can influence the permafrost distribution on a talus slope. Map 5 displays the SDrestopo, a surrogate for grain size. Several authors report big boulders in two areas on a talus slope. Rocks which break of the head wall, can be trapped at the foot of the rock wall. However, most of the large boulders, will slide down, with an increase of mean grain size downwards the talus slope as a result (e.g. Luckman, 2013; Sanders et al. 2009). On the landslide, no downward sorting was expected, as other transport processes influence the geomorphology. Both the map (Map 5) and the field campaign (Figure 40, Figure 41) confirm this hypotheses. More specific for the landslide, we can identify an area with a high surface roughness in the middle of the landslide (Map 5). In the downward part of the landslide, a low surface roughness is the result of the presence of soil and mainly small rock fragments. Contrary, based on our proxy, surface roughness, the distribution of grainsize on a talus slope mentioned above does not count for our site. The relation between the height and surface roughness on the talus slope was very small (Pearson’s’ correlation coefficient = 0,239; 훼 = 0,000) and even positive. In other words, a higher location result in a higher grain size. On the landslide, the reported correlation is even less (Pearson’s’ correlation coefficient = 0,109; 훼 = 0,000), as we expected. The only conclusion we took from this statistical analysis, is the fact that the landslide is less sorted then the talus slope.

Figure 40: Grain size distribution on the talus Figure 41: Grain size distribution on the slope landslide

53

Map 5: Thematic map surface roughness and permafrost distribution

6.5.2.2 Permafrost and surface roughness

A blocky surface layer is one of the characteristics which promote the existence of permafrost above the theoretical border of permafrost. On the talus slope both the temperature and VES measurements suggest permafrost where the surface is covered with coarse grained materials. On the other hand, no permafrost is suggested by the t-logger located within the finer grained material. However, these are only two measurements, which makes it impossible to generalise. The statistical analysis on the landslide, did not show any significant relationship (훼 = 0,33) between the surface roughness and

54 permafrost existence. Although, on Map 5 the permafrost probability was placed upon the surface roughness layer. Here it seems that the area where permafrost is probable is located in a rougher surface, while the upper part of transect 2a, where no permafrost is suggested, is found in a smoother area.

7. DISCUSSION

Within following chapter, we approach our results critical and place them in the framework of international permafrost research. We try to link our results to theories and case studies in the European Alps. Do we find the same processes and relationships on other talus slopes and landslides in the European Alps? Furthermore, we will also focus on our methods and data. How do the results relate to the used methods, are their unexpected results which could be an indirect effect from our methodology? What do we need to keep in mind using the selected methods?

7.1 Geomorphological mapping

The geomorphological map (Map 3) makes it possible to investigate and discuss the formation and evolution of the talus slope. The usual concave talus slope profile is reworked, which result in morphological landforms such as a rock glacier, landslide, gullies, different kind of ridges etc. To explain these landforms, we combined our data and searched for other examples in international research. Here we elaborate this.

7.1.1 Gravitational processes

Based on our geomorphological map and field researches, we can conclude that the formation processes of a talus slope, as described in handbooks such as Luckmann et al. (2013) and Gutierrez & Gutierrez (2016) can be found on the site. The talus slope is formed by rock fall from the headwall and has an average inclination from approximately 30° – 40°C with a basal concavity. Although, not all of the characteristics can be recognised using statistical analyse. An increasing grain size at the base of the talus slope, for example, was not verified using the surrogate ‘surface roughness’ for grain size. However, based on the maps and field observations, it seems that this hypothesis could be valid. We argue that the used methodology possibly had an influence on this unexpected results. For example: We used a smoothed DEM from 15 x 15, which equals a resolution of 99 x 99 cm. For boulders which exceeds these values, only the sides will be clearly visualized. In this case, the surface roughness can point towards both coarse or fine grained soil, depending from the place where you subtract the data. When we only focus on this quantified data, without including the surroundings, the results will strongly depend upon the exact location of the selected data points. Furthermore, the big boulders at the foot of the talus slope are spread out. In our analyses we only took one point every 10 m. As a result, the chance exist that the majority of the boulders were not present in the dataset. To conclude, we need to be careful with the proxy surface roughness while working with exact data points, however when visualizing on a map, we can identify areas with large boulders or soil, but it is important to keep the field observations in mind.

55

7.1.2 Snow, frost and (peri)glacial landforms

The first periglacial landform we identified is the rock glacier. This landform is commonly found within the discontinues permafrost belt in the European Alps. A typical rock glacier topography is characterized by an unstructured upper part, eventually with longitudinal ridges and a pattern of transversal ridges and furrows in the tongue (Roer et al., 2008; Wahrhaftig & Cox, 1959). Furthermore, it has steep front and side slopes and the upper end is marked by a steep talus slope or a depression between the talus slope and rock glacier (Barsch, 1992). These characteristics can be recognized on the examined rock glacier. The front exceeds 40° (Map 4) and in the upper end a depression filled with perennial snow is found. The depression may indicate that buried ice has partially melted or a that the movement of the rock glacier greater is than the accumulation of rock debris (White, 1981). Furthermore, the rock glacier exists of two convex ridges, but clear transversal furrows were not observed. Additionally, a longitudinal ridge with at each side a V-shaped furrow, was reported in the upper zone. The transverse ridge on the upslope side of the rock glacier is defined as a protalus rampart, after the general morphological definition from Shakesby (2004). Hereby it is important to mention that other authors approach the term, protalus rampart, as an embryonic rock glacier (e.g. Scapozza et al., 2011; Haeberli, 1985). However, Scotti et al. (2013) argues that those two definitions lie close together in terms of ground thermal conditions. The second transverse ridge forms the transition towards the front, which is caused by the advancing rock glacier (Roer & Nyenhuis, 2007). To be able to clearly interpret origin of the geomorphological features, such as the depression, measurements about the differences in surface velocity on the rock glacier are needed. In the summer of 2018, ten points were marked by paint and measured with the RTK-GNSS.

Alongside the rock glacier, the landslide is found. Within this landslide geophysical measurements indicated the existence of permafrost. As a result, it is needed to discuss the terminology: Is this landform a landslide or could we identify it as a rock glacier? Barsch (1992) defines active rock glaciers as:

“Lobate or tongue-shaped bodies of perennially frozen unconsolidated material, supersaturated with interstitial ice and ice lenses that move downslope or downvalley by creep as a consequence of the deformation of ice contained in them and which are, thus, features of cohesive flow” (Barsch, 1992).

The landform on Col du Sanetsch has only permafrost in the middle region. Furthermore, the slope of the front is approximately 20 – 30° and covered by vegetation. In case of a rock glacier, this suggest that it is an inactive or even relict rock glacier. However, there are no morphological proves of relict rock glaciers, such as terminal ridges or depressions (Ikeda & Matsuoka, 2002). Besides, the highest surface velocity rates, are measured in regions were no permafrost is present, so they are related to other processes then permafrost creep (Map 4). Based on this definition and characteristic we conclude that the landform is a landslide, and not a rock glacier.

Another aspect of this landform we need to discuss is the rooting zone: ‘Does the landslide start below the bedrock outcrop, or is the area above also a part of it?’ The conclusion is based on the different

56

datasets and observations: On the geological map, the zone around the outcrop is identified as ‘tassement’ or settlement (Badoux et al., 1959; Figure 44). We argue that this is the upper part of the so-called ‘zone of depletion’ on the idealized landslide diagram, provided by the IAEG Commission on Landslides (1990) (Figure 43). Around this zone, the surface elevation decreases (Figure 42). Further down, the elevation increases, which corresponds to the zone of accumulation on the diagram. These findings suggest that the roots of the landslide lie bellow the outcrop. The hypothesis is strengthened by two other observations: First of all, the measurements of surface velocity indicate that the area above this zone has a displacement rate lower than average, while the data points below has an average surface velocity (Figure 44). Additionally, the displacement values above the landslide do not variate between the years, while the points measured on the landslide has at least a doubled displacement in 2012 – 2013 and 2015 – 2016. Together with the differences in permafrost probability, it suggests that the area above the bedrock outcrop forms another system then the area bellow the outcrop.

Zone of depletion

Figure 43: Representation of an idealized landslide (Source: B.C. Ministry of Forests, 1997)

Figure 42: Landslide on Col du Sanetsch

57

Figure 44: Rooting zone of the landslide (Source: Google earth, 2016)

On the landslide, different ridges, both transversal as well as longitudinal are observed. According to Parise (2003) the surface of active landslides are characterized with deformational features, a result of differential movement within the landslide. On the landslide on Col du Sanetsch, they are not present on the whole surface, but can be found from the area where permafrost is probable, toward the toe of the landslide. In the upper zone, below the bedrock outcrop, no ridges are visible. The lowest ridges are covered with vegetation. Vegetation could suggest inactivity, however, there is still a small displacement measured in the area. The reported mean velocity in this area is around 15 cm, with years with almost no movement (2 to 4 cm). Only during the fast moving years (2012 – 2013 and 2015 – 2016) velocities up to 50 cm are measured (Figure 36). At the southern part of the slope, around the two measurements points with the highest velocity values, no ridges are found (Figure 44). These points are located at the edge of the badlands were materials easily erode. This could be a reason why no ridges are found in this area. Besides the ridges, other geomorphological observations also prove displacement on the landslide. At some places, the landslide itself pushes the gully towards the badlands. Downhill, the gully is filled with both small grained materials, originating from the badlands and the landslide, and big boulders, from the landslide (Figure 45). The gully operates as a channel to remove the materials, which are pushed towards the badlands.

58

Figure 45: Gully filled with materials from the landslide

The last (peri)glacial and snow related landforms we should discuss are the so-called ‘nivation processes’. On the talus slope they are found on two different scales. Between the two structural ridges a depression of approximately 100 m long and a width of 40 m can be found. Snow will assemble in the depression and results in a long lasting snow layer. During the field period, we could observe snow until the beginning of August. Furthermore, on the steep back slope (approximately 35 – 45°) of the depression bedrock is visible (Figure 21). This indicates backwards erosion, a process typical for nivation hollows. In the upper part of the talus slope, several areas with transversal ridges and small depressions are observed. These ridges are grouped and on some places they form a long row up to 20 m (Figure 22). In literature we could not find papers were these landforms are investigated. Based on theoretical knowledge we assume that these landforms are formed and influenced by gravitational and nivation processes. On a talus slope, rock fragments from the headwall are rolling, sliding and bouncing downslope (Gutiérrez & Gutiérrez, 2016). The impact from these rocks can create small depressions. Because of the lower temperatures, snow will last longer in the upper region of the talus slope, with the longest persistence in the depressions. As a result, nivation processes could enhance these depressions. The ridges are asymmetric with a steep counter slope. In the depression and on the counter slope medium sized rock fragments can be found, while the front slope exists of small rock fragments with a clear matrix. We relate this pattern with gravitational processes, as the largest particles will slide down and come to rest against the counter slope of the ridge underneath.

7.2 Meteorological analysis

The explanatory variable temperature is very important within permafrost research. Our data is based on t-loggers installed since 2014. However, some t-loggers only have data from one year, so no generalisation can be made. Therefore, it will be useful to investigate if the observed results and trends continue the next years. If more data is present it will also be possible to do some statistical analyses. In this way, we would be able to statistical analyse the relationship between permafrost existence and WEqT or the influence of a long lasting snow cover on the MAGST. Based on our own findings we can conclude that the lower MAGST at Arp – 010 – mid is the result of the long lasting snow layer. On this location, the snow layer is always longest present (Figure 46). Consequently, the isolating characteristics of the snow will prevent the soil from warming, which can result in a lower MAGST and has a positive influence on permafrost existence. The longer persistence of snow is probably due to two

59 different factors: The high surface roughness indicates that the area around the t-logger is coarse grained. As a result, snow can be trapped and protected from direct solar radiation between the big boulders (Otto et al., 2012). Furthermore, snow falling on the steep slope in the rooting zone of the landslide, will likely slide down and assemble on the more level and rough surface underneath. This results in a thicker snow pack with a longer melting period and consequently a longer persistence of snow. However, the question arises why the duration of the melting period in 2015 – 2016 do not follow this trend. Comparing the start dates of melting period from the different t-loggers learns us that in 2015 – 2016 the melting period of the upper and lower t-logger starts more than one month earlier (respectively 5th of April and 22nd of April, comparing to the 22nd of May). The other years, the difference is maximum seven days. Further research is to find a possible explanation.

Figure 46: Mean MAGST and average zero curtain period for the different t-loggers

Another question which arises, is the striking difference between the different parameters in 2015 – 2016 and the other years. To analyse the trend, the WEqT and MAGST temperatures from Arp – 010 – mid are compared to the borehole temperature of ten permafrost sites (Figure 47). The difference in variability between our site and those from PERMOS are the result of the differences in measurement depth. However, as the WEqT depends on the thermal state of the underground they are related. In 2015 – 2016 the measured WEqT was not only high on our site, but all over Switzerland, record temperatures of permafrost are measured (PERMOS, 2017). Contrary, the MAGST decreases in 2015 – 2016. As mentioned before, we relate this to the insulation by the long-lasting snow cover. In 2016 – 2017 the warming permafrost trend was interrupted for the first time since 2009 (PERMOS, 2018). The cooling was the result from a long-lasting snow cover and the late arrival of the winter (PERMOS, 2018). This corresponds to the pattern of WEqT we measured on Col du Sanetsch. However, the duration of the snow cover did not significantly differ in 2016 – 2017 (Figure 25). Moreover, it was longer present in 2015 – 2016 then 2016 – 2017. Therefore, we presume that the arrival of snow plays a more important role on Col du Sanetsch. Unfortunately, we cannot be sure, as we do not have any information about the arrival of snow. The installation of the webcam in summer 2018 can counter this knowledge gap in

60 the future. In 2017 – 2018 the warming trend resumes (PERMOS, 2019), although temperatures at depth are still influenced by the cooling phase the year before Figure 47). At the surface, temperatures start to rise again. These observations are conform to our measurements. The WEqT, which is

influenced by temperatures at depth, does not rise significantly in 2017 – 2018.

BoreholeWEqTtemperature / (°C)

Figure 47: Borehole temperatures at approximately 20 m depth compared to the WEqT and MAGST of Arp - 010 - mid (Source: based on PERMOS, 2019)

7.3 Permafrost distribution indicated by geophysical and temperature measurements

Based on previous data about the surface temperature and surface deformation, we expected to find permafrost on the study site. The executed geophysical measurements confirmed this, and indicate permafrost on the rock glacier, landslide and talus slope. In this chapter we elaborate this distribution and discuss the influences from factors such as temperature, grain size and eventually air circulation. Furthermore, we compare the different indices, namely temperature and geophysical measurements. In the end our distribution will be used to evaluate the two permafrost maps on the study site.

7.3.1 Permafrost distribution on the talus slope

On the talus slope, t-loggers were installed both in a blocky surface layer and a finer grained surface layer. The purpose of these t-loggers was to investigate the influences of the grain size on temperature and permafrost existences. Based on studies from for example Harris & Pederson (1998) and Lambiel & Pieracci (2008) we predicted a lower temperature and possible existence of permafrost in the coarse grained surface layer. The temperature measurements and VES profiles confirm this (Table 12). The geophysical measurements were only executed in the coarse grained surface layer, but as the WEqT is only -0,2 °C in the fine grained layer (AT – GST – 3) we can be relatively sure that no permafrost is

61 present. However, it is important to mention that the WEqT is based on measurements from only one year (2017 – 2018). According to Schoeneich et al. (2011) WEqT can vary between 2 and 3 °C depending on the moment of arrival and thickness of the snow layer. To discriminate between permafrost and non-permafrost areas, the contrast in WEqT is more important. In this case, there is a well-marked contrast. So, despite we only have data from one year, we can conclude that permafrost is probable in the coarse grained surface layer and no permafrost is present in the fine grained layer.

Height Grain size VES WEqT MAGST Temperature spring AT – GST – 1 2464 m Coarse Permafrost possible -3,1 °C 0,02 °C 2,3 °C AT – GST – 2 2400 m Coarse No permafrost -2,8 °C 0,90 °C 2,3 °C AT – GST – 3 2507 m Medium / -0,2 °C 2,05 °C 6,2 °C

Table 12: Permafrost distribution from geophysical and temperature measurements related to topographic factors on the talus slope

The predicted distribution from the temperature data can be evaluated using the outcome of the geophysical transects. Both the WEqT and temperature of the spring suggest permafrost in the coarse grained surface layer. On the upper location, VES measurement confirmed the presence of permafrost. Contrary, the lower location (AT – GST – 2) did not contain permafrost according to the VES measurement. We suggest that the WEqT in this case is influenced by the permafrost upslope. A geophysical transect will give more information about the extent of the permafrost in the blocky surface layer.

Another purpose of the t-loggers was to investigate if air ventilation is present in the talus slope. Therefore, the blocky surface layer contains two t-loggers, one at 2400 m.a.s.l and one at 2464 m.a.s.l. Delaloye & Lambiel (2005) describe different indices of air circulation, among which the reversed MAGST between the upper and lower regions of the slope. Since the higher t-logger (AT – GST – 1) has a lower MAGST compared to AT – GST – 2 (Table 12), no air circulation is present. As there is no air circulation, no negative anomaly in temperature is present and permafrost is not favoured in the lower area.

7.3.2 Permafrost distribution on the rock glacier

On the rock glacier, permafrost is also probable. Both the ERT profiles had resistivity values which denote permafrost probability. In the centre of the lateral profile one area with high resistivity values (>15 000 Ω.m) stands out. As reported by Scapozza & Laigre (2014) this can be related with a decrease in temperature or an increase in ice content. As in several other rock glaciers reported, we assume that the higher values are related to an increased ice content (e.g. Springmann et al., 2012). It can be interesting to follow this body in the coming years. How does it influence the geomorphology of the rock glacier? Is it melting away and how does this eventually degradation influence the geomorphology and dynamics? The transects do not lie within our DEM, as a result it was not possible to take conclusion

62 about the micro-topography in the environment of the ice rich body. Providing a high resolution DEM of this environment is the first step in analysing the influence of permafrost on the rock glacier morphology.

Height Grain size ERT WEqT MAGST GST – 9 – RG – up 2639 m Coarse Permafrost probable -1,98 °C -0,93 °C GST – 13 – RG – front 2595 m Medium/ fine No permafrost -3,79 °C 0,74 °C GST – 15 – RG – south 2636 m Medium No permafrost -0,43 °C 1,85 °C GST – 16 – RG - north 2643 m Medium No permafrost -0,80 °C 1,34 °C

Table 13: Permafrost distribution from geophysical and temperature measurements related to topographic factors on the rock glacier

Similar to the talus slope, the geophysical measurements can be used to verify the temperature indications of permafrost. The different temperature metrics and the outcome of the ERT profile can be found in Table 13. The lateral ERT profile do not suggest permafrost at the sides of the rock glacier. The WEqT of the t-loggers which are located along the rock glacier match this observation (Figure 50). The other t-loggers are located in the upper region of the rock glacier and below the front. In the upper region, the t-logger has a WEqT of -1,98° C. According to the literature (e.g. Schoeneich et al., 2011), this do not suggest permafrost. On the other hand, there is a clear difference with the nearby t-loggers located in non-permafrost areas, so the temperature measurement can indicate permafrost existence, which coincidence with the ERT measurements. In the area below the front the WEqT (-3,8°C) and the apparent resistivity values do not coincide. Delaloye & Lambiel (2005) observed the same trend on the Lapires talus slope (Swiss). Cold WEqT values below the permafrost area and warmer WEqT above the sector of high resistivities. In our case, the upper value does not approach 0°C, but the t-logger is located on the permafrost site and not above. These observations suggest the existence of an internal ventilation (Delaloye & Lambiel, 2005). However, three other observations contradict this. First of all, the MAGST of the t-logger up is warmer than the t-logger underneath the front. Secondly, the t-logger in the front is not placed within a coarse grained surface layer, one of the requirements to have a chimney effect (Figure 48). Thirdly, by plotting the temperatures of both t-loggers on a graph, we can observe that most of the time the temperature of the upper t-logger is colder than the lower one. We assume that both t- loggers are influenced by the permafrost, but the difference in WEqT is related to the differences in snow layer. According to Schoeneich et al. (2011) the WEqT can differ up to 3°C a year, depending on the arrival of snow. The upper t-logger is located in the depression above the rock glacier. In the summer of 2017 – 2018 there was still snow present above the t-logger. During the winter, snow from the steep slope above will easily assemble on the more levelled depression. Furthermore, the ridge from the rock glacier prevent the snow from moving further down. On the other hand, the t-logger bellow the front lies within a steeper area (> 30°) and there are no ridges which can prevent the snow from sliding down. As a result, it will take a longer time to build a sufficient snow layer in winter. We presume that due to the difference in isolation in the early winter, the area around the t-logger bellow the front had more time to cool, which leads to the lower WEqT.

63

Figure 48: T-logger located below the front of the rock glacier (GST – 13 – RG)

Temperature of t-loggers GST-9-RG and GST-13-RG in 2016 - 2017 25 20

15 C)

° 10

5 GST GST ( 0 -5

-10

05/20/17 09/18/16 09/26/16 10/14/16 10/23/16 10/31/16 11/18/16 11/26/16 12/14/16 12/23/16 12/31/16 01/18/17 01/26/17 02/13/17 02/22/17 03/20/17 03/28/17 04/15/17 04/23/17 05/28/17 06/15/17 06/23/17 07/20/17 07/28/17 08/15/17

5/02/2017 9/09/2016 1/09/2017 2/04/2017 3/02/2017 3/11/2017 4/06/2017 5/11/2017 6/06/2017 7/02/2017 7/11/2017 8/06/2017

10/05/2016 11/09/2016 12/05/2016

GST - 9 - RG (Up) GST - 13 - RG (Front)

Figure 49: Temperature of t-loggers GST-9-RG and GST-13-RG in 2016 - 2017

7.3.3 Permafrost distribution on the landslide

On the landslide the ERT measurements indicated permafrost probability around t-logger Arp – 010 – mid. Downslope no permafrost could be found and upslope only a few resistivity values indicate a possibility to find permafrost. On the other hand, none of the WEqT measurements indicate permafrost. The WEqT is everywhere higher than -2 °C and there are no clear differences between the t-loggers (Table 14). Romanovsky & Osterkamp (2000) investigated the influence of unfrozen water on heat processes in active layers and permafrost in Alaska. Due to the latent heat, released during the freeze up, unfrozen water will influence the temperature positively. In regions were warm, discontinuous

64 permafrost exist, this effect will last most of the winter period and can decrease the thermal offset. The different springs which are present on the landslide indicate that a significant amount of unfrozen water is represent which can influence the thermal regime of the ground.

After analysing the temperature indicators on all different landforms in our study area, we can conclude that the WEqT is a good predictor of permafrost if it is studied within its context.

Height ERT/ VES WEqT Temperature spring Arp – 005 - up 2641 m No permafrost -1,72 °C 6 – 9 °C Arp – 010 – mid 2519 m Permafrost probable -1,48 °C 6 – 9 °C Arp – 017 - down 2448 m No permafrost -1,11 °C 6 – 9 °C

Table 14: Permafrost distribution from geophysical and temperature measurements related to topographic factors on the landslide

On the landslide we also did a statistical analyse of the relation between permafrost probability and grain size. The visual analyse suggested a relation, but we could not prove it using a statistical test. Here we assume that our methodology had an influence on this unexpected result. To set up the dataset, we combined the permafrost probability and the surface roughness for every electrode of the transects CdSLS2a and CdSLS2b. As we used a handhold GPS to locate the transects, the location of the electrodes is not accurate and can have errors up to 5 m. In this way, the two variables will never be extracted from exactly the same location. Furthermore, as we mentioned before, the exact location from which we derive the roughness data will also have an influence. Within the small dataset (94 points, spread out over the three permafrost categories) to investigate the relation between surface roughness and permafrost, these differences may have a great influence on the result. As an attempt to overcome this problem, we took the maximum surface roughness in an area of 99 cm x 99 cm, but even so we need to be careful with the interpretation.

7.3.4 Differences in apparent resistivity between landslide and talus slope

The last question which arises regarding to the geophysical survey, is the about difference in apparent resistivity between the talus slope and landslide: “Why is the apparent resistivity at depth higher on the talus slope compared to the landslide?” On both the VES transects, the apparent resistivity values are higher than on the landslide. A possible explanation can be found in the materials: On the landslide, it is possible to find materials from the badlands (marl and glauconitic sandstone), which have resistivity values starting from approximately 5 Ω.m (Figure 2).

7.3.5 Permafrost probability maps

As mentioned earlier, there are two permafrost distribution models from the Swiss Alps. These models will give a first indication of permafrost existence. However, the real distribution depends on local factors such as for example a block surface layer. Our measurements of permafrost distribution can be used to evaluate the permafrost distribution models on Col du Sanetsch. On Figure 50 and Figure 51, the output

65 of the geophysical measurements, the WEqT and the spring temperatures are visualised on respectively the map of potential permafrost distribution, from Swiss and the Alpine Permafrost Index Map (APIM). In this way, we are able to compare and verify the different models. In both models, the height and aspect of the slopes plays an important role. On the APIM, the whole talus slope has a possibility to find permafrost. The Swiss map on the other hand, has a higher resolution and is more detailed. If we compare these maps with our own outcome, several observations can be made. In the first place, the area above the landslide do not contain permafrost in our measurements, while both maps indicate a high probability for permafrost. Secondly on both the maps and the geophysical measurements, the rock glacier has a probability to contain permafrost, with a decrease towards the front. A clear difference between both maps is the existence of permafrost on the landslide. The Swiss map has a high probability towards the northern part, independently from the height, on the APIM the probability mainly differs with height. Our own measurements point toward a high probability in the northern part of the talus slope. However, they are not located within the area with high probability on the Swiss map, so we cannot verify this pattern. We were not able to get the explanation of the Swiss model. As a result, it is not possible to explain the anomaly on the northern part of the landslide. It would be interesting to learn more about the parameters used in the model to try to explain and get more insight in this northern anomaly. To end, the measurements on the talus slope, differ from the Swiss map, but on the APIM, there is a probability to contain permafrost in very favourable conditions. The blocky surface layer will probably be one of the supporting conditions.

66

Figure 50: Permafrost distribution: Geophysical measurements compared to the Swiss Potential Permafrost Distribution Map (Source: FOEN, 2005 ; Federal office of Topography, 1998)

Figure 51: Permafrost distribution: Geophysical measurements compared to the APIM 68 (Source: Boeckli et al., 2012; Federal office of topography, 1998)

7.4 Annual surface velocity

The last results we need to discuss, are the annual surface velocities on the landslide. The zonal analysis learned us that there are two different systems. Above the landslide, there is only a small movement (4 till 13 cm.y-1) and no pronounced variation over time. On the other side, the points located on the landslide have an average surface displacement from 38 cm.y-1, which can go up to 250 cm.y-1, depending on the year and location. There is also a wider variation in the displacement of those points. In significant warmer years (2012 – 2013 and 2015 – 2016), the displacement rates are doubled.

Analysis of zonal and temporal differences in WEqT, permafrost distribution, moisture content of the soil and the slope rate gave us more insight in the process. An exponential relation between the WEqT and surface velocity was found. Furthermore, the intra-site variations in horizontal surface velocity were related to the higher moisture content in the fast-moving area. This is confirmed by Matsuoka et al. (2001) who identified moisture content as one of the most significant factors to determine gelifluction. However, the temporal differences in moisture content, based on the precipitation data of Sion, did not approve this. As a result, we suggest that other factors, such as WEqT are more important.

Slope is one of the other factors influencing gelifluction (Matsuoka et al., 2001). The effective dependence of movement on inclination is affected by the environment. In lower-latitude alpine areas spatial variations in other factors, such as freeze-thaw frequency and moisture distribution, will mask the influence of the inclination (Matsuoka et al., 2001). On the landslide on Col du Sanetsch it was indeed not possible to find a relationship between the slope and the surface displacement. One of those intra-site variables is the presence of an impermeable layer, resulting from permafrost. Based on the ERT measurements, we concluded that permafrost is present on location two, but not on location one. This impermeable layer encourages gelifluction rates as it serves as a moisture barrier, which improves the moisture content of the soil (Matsuoka et al. 2001). However, despite the absence of the permafrost layer on location one (down), this area is moving at a higher velocity than were permafrost is present. A possible explanation can also be found in the paper of Matsuoka et al. (2001). He argued that there is no significant difference between gelifluction rates in permafrost and seasonal frost areas. During spring, seasonal freezing will serve as an impermeable layer and also causes an intensified inflow of melt water.

From these analyses we can conclude that the horizontal surface displacement rate depends in the first place from WEqT. Years with low WEqT (e.g. 2017 – 2018, -1,7°C) have low surface velocities. The intra-site variations, such as moisture content and the presence of an impermeable layer, explain the zonal variations. Contrary to what we expected, slope do not have a significant influence.

Finally, it is interesting to compare the surface velocities from the landslide to other active landslide in the European Alps. In the measuring network of the university of Fribourg, five different landslides are monitored (Breithorn, Grabengrufer, Längenschnee, Mooshflush, Perroc). The only general evolution we could find is the decreased velocity in 2017 – 2018. In 2017 – 2018, all the examined landslides have

a lower horizontal surface displacement rate. This is in line with our case study. After a peak in 2015 – 2016 the surface velocities decrease. The maximum surface velocities on the other landslides were reached in 2015 – 2016 or 2016 – 2017.

7.5 Possibilities for future research

Within our study we analysed the geomorphology and the interrelations with meteorological factors, permafrost distribution and horizontal surface displacement. However, there are still some open questions. To answer these, some data gaps need to be closed. In the next paragraph we list some possibilities for further research.

First of all, by the continuation of the current measurements it will be possible to investigate if the trends persist and possibly generalise them. The installation of the webcam in the summer of 2018 will give more information about the snow layer. As a result, we will be able to elaborate the influence of the snow layer on the GST, the horizontal displacement and the distribution of permafrost. To get more knowledge about the origin and processes on the rock glacier, measurements of surface velocity will be useful to collect. To obtain this, a network of measurement points was created in the summer of 2018. Finally, an ERT measurement in the blocky surface layer on the talus slope can be used to explore the extent of permafrost in the talus slope. Is permafrost restricted to the coarse grained area, or is it also present in the nearby environment? How far does the permafrost reach downslope?

8. CONCLUSIONS

The purpose of this research was to investigate and explain the geomorphology, the permafrost distribution and their interrelationship on the talus slope of Col du Sanetsch. Based on a geomorphological mapping we concluded that the talus slope on Col du Sanetsch is a good example to explain the evolution, concepts and processes which influence the geomorphology of talus slopes. The primary, straight profile with a basal concavity is reworked by the different ways of transport. Underneath the rock channels in the head wall, rock fall and debris flows result in respectively talus cones and gullies with associated levees. There are several processes which can be related to frost, snow and periglacial processes. The presence of a rock glacier and protalus rampart, gelifluction and stress related landforms, such as transverse ridges, suggest the existence of permafrost on the talus slope and landslide. This was verified by our geophysical measurements. In the coarse blocky layer on the talus slope, the rock glacier and the landslide apparent resistivities which indicate permafrost are measured. Furthermore, nivation processes are related to the long lasting snow in this periglacial environment.

Data about temperature, topography, surface velocity and surface roughness were related to geophysical measurements in order to explain the permafrost distribution. One of the most important factor which influence the permafrost distribution is the temperature. The lowest Mean Annual Ground Surface Temperature (MAGST) are observed in the areas where permafrost is probable. One of the factors which influence the MAGST values, and thus the permafrost distribution, is the snow layer. A long lasting snow layer prevent the soil from warming and results in a lower MAGST. One of the most

70 used indicators of permafrost is the Winter Equilibrium Temperature (WEqT), the mean temperature over a period of 30 days during which the snow layer is stable. We found out that WEqT can indeed give indications about the permafrost distribution, if the context is included. The presence of unfrozen water in the soil will, for example, result in a higher WEqT due to the latent heat released during the freezing. Furthermore, the WEqT can differ depending on the arrival of snow. As a result, it is not only the WEqT, but also the differences in WEqT on the site which will indicate permafrost distribution. Additionally, the temperatures of springs at the feet of the talus slope can also be used as an indicator for permafrost existence. Another factor which influences the permafrost distribution on Col du Sanetsch is the grain size. Coarse grained soils have a negative temperature anomaly, which favour permafrost existence. Air ventilation is not observed in the talus slope.

One of the main landforms on the talus slope is a landslide, which is slowly moving forward. We defined several questions about this landform. First of all, we determined the rooting zone of the landslide. We asked ourselves the question: “Is the landslide starting just underneath the bedrock at the headwall or can we find the rooting zone underneath a bedrock outcrop further downslope?” Based on geomorphological observations, the horizontal surface velocity and the distribution of permafrost, we concluded that the area above the bedrock outcrop and the one bellow are two different systems. They have a different horizontal displacement and above the bedrock outcrop it is not probable to find permafrost, contrary to the area bellow.

The horizontal surface displacement of the landslide is the result from gelifluction. Horizontal surface velocities up to 250 cm.y-1 were measured depending on the location and year. The relationship between the horizontal surface velocity and the WEqT is exponential. The intra-site variations mostly depend on the difference in moisture content and the presence of an impermeable layer resulting from permafrost or seasonal frost. Contrary to what we expected the slope do not have an influence of the horizontal displacement rate. The several transverse ridges which can be found on the landslide are a prove of this displacement. The highest surface rates were measured in 2015 – 2016. Meteorologically, this was an exceptional year, not only on the Col du Sanetsch, but in the whole Alps. The warmest permafrost temperatures were measured, this coincidence with the high WEqT on our study site.

9. REFERENCES

Literature

Ballantyne, C. K. (2002). “Paraglacial geomorphology”. Quaternary Science Reviews, 21 (18-19), 1935- 2017.

Barandun, M., Hilbich, C. (2013). Tutorial zur Geoelektrik – ERT. Messung (SYSCAL) & Auswertung (Res2DInv).

Barsch, D. (2012). Rockglaciers: indicators for the present and former geoecology in high mountain environments. Heidelberg: Springer Science & Business Media.

71

B.C. Ministry of Forests (1997) Silviculture Prescriptions Field Methods Book.

Beniston, M. (2006). “Mountain weather and climate: a general overview and a focus on climatic change in the Alps”. Hydrobiologia, 562 (1), 3-16.

Bodin, X., Krysiecki, J. M., Schoeneich, P., Le Roux, O., Lorier, L., Echelard, T., ... & Walpersdorf, A. (2017). “The 2006 collapse of the Bérard rock glacier (Southern French Alps). Permafrost and Periglacial Processes”. 28 (1), 209-223.

Boeckli, L., Brenning, A., Gruber, S., & Noetzli, J. (2012). “Permafrost distribution in the European Alps: calculation and evaluation of an index map and summary statistics”. The Cryosphere, 6 (4), 807.

Carturan, L., Zuecco, G., Seppi, R., Zanoner, T., Borga, M., Carton, A., & Dalla Fontana, G. (2016). “Catchment‐scale permafrost mapping using spring water characteristics”. Permafrost and Periglacial Processes, 27 (3), 253-270.

Cavalli, M., Tarolli, P., Marchi, L., Dalla Fontana, G. (2008). “The effectiveness of airborne LiDAR data in the recognition of channel-bed morphology”. Catena, 73 (3), 249-260.

De Blasio, F. V., Sæter, M. B. (2015). “Dynamics of grains falling on a sloping granular medium: application to the evolution of a talus”. Earth Surface Processes and Landforms, 40 (5), 599-609.

Delaloye, R. (2004). Contribution à l'étude du pergélisol de montagne en zone marginale (Doctoral dissertation, Université de Fribourg).

Delaloye, R., Lambiel, C., Gärtner-Roer, I. (2010) “Overview of rock glacier kinematics research in the Swiss Alps”. Geographica Helvetica. 65 (2), 135-145.

Delaloye, R., & Lambiel, C. (2005). “Evidence of winter ascending air circulation throughout talus slopes and rock glaciers situated in the lower belt of alpine discontinuous permafrost (Swiss Alps)”. Norsk Geografisk Tidsskrift-Norwegian Journal of Geography, 59 (2), 194-203.

Deluigi, N., Lambiel, C., Kanevski, M. (2017). “Data-driven mapping of the potential mountain permafrost distribution”. Science of The Total Environment, 590, 370-380.

Dobinski, W., (2011). “Permafrost”. Earth-Science Reviews, 108 (3–4), 158–169.

Ekinci, Y. L., Türkeş, M., Demirci, A., & Erginal, A. E. (2013). “Shallow and deep-seated regolith slides on deforested slopes in Çanakkale, NW Turkey”. Geomorphology, 201, 70-79.

72

Emmert, A. (2016) Col du Sanetsch.

Etzelmüller, B., (2013). “Recent advances in mountain permafrost research”. Permafrost and Periglacial Processes, 24 (2), 99–107.

Etzelmüller, B., Frauenfelder, R. (2009). “Factors controlling the distribution of mountain permafrost in the Northern Hemisphere and their influence on sediment transfer”. Arctic, Antarctic, and Alpine Research, 41 (1), 48-58.

Federal Office of Meteorology and Climatology MeteoSwiss. (2018). “Homogeneous data series since 1864. Sion”. https://www.meteoswiss.admin.ch/home/climate/swiss-climate-in-detail/homogeneous- data-series-since-1864.html?station=sio. 8/5/2019

Florentine, C., Skidmore, M., Speece, M., Link, C., Shaw, C. (2014). “Geophysical analysis of transverse ridges and internal structure at Lone Peak Rock Glacier, Big Sky, Montana, USA”. Journal of Glaciology, 60 (221), 453-462.

Frauenfelder, R., Allgöwer, B., Haeberli, W., & Hoelzle, M. (1998). “Permafrost investigations with GIS— a case study in the Fletschhorn area, Wallis, Swiss Alps”. Proceedings of the 7th International Conference on Permafrost, Yellowknife, Canada, Collection Nordicana, 57, 291-295.

Frehner, M., Ling, A. H. M., & Gärtner‐Roer, I. (2015). “Furrow‐and‐ridge morphology on rockglaciers explained by gravity‐driven buckle folding: A case study from the Murtèl Rockglacier (Switzerland)”. Permafrost and Periglacial Processes, 26 (1), 57-66.

Giaccone, E., Fratianni, S., Ambrosi, C., Antognini, M., Delaloye, C., Mari, S., Scapozza, C. (2016) “Recent evolution of mountain permafrost in the Southern Swiss Alps: results of 2006 - 2015 per iod”. International conference on permafrost 2016 - Book of Abstracts. Potsdam: Bibliothek Wissenschaftspark Albert Einstein, pp. 39-40.

Gobiet, A., Kotlarski, S., Beniston, M., Heinrich, G., Rajczak, J., Stoffel, M. (2014). “21st century climate change in the European Alps—a review”. Science of the Total Environment, 493, 1138-1151.

Gómez, A., Palacios, D., Luengo, E., Tanarro, L. M., Schulte, L., Ramos, M. (2003). “Talus instability in a recent deglaciation area and its relationship to buried ice and snow cover evolution (Picacho del Veleta, Sierra Nevada, Spain)”. Geografiska Annaler: Series A, Physical Geography, 85 (2), 165-182.

73

Grohmann, C. H., Smith, M. J., Riccomini, C. (2011). “Multiscale analysis of topographic surface roughness in the Midland Valley, Scotland”. IEEE Transactions on Geoscience and Remote Sensing, 49 (4), 1200-1213.

Gruber, S., Haeberli, W. (2007). “Permafrost in steep bedrock slopes and its temperature‐related destabilization following climate change”. Journal of Geophysical Research: Earth Surface, 112 (F2).

Gruber, S., Haeberli, W. (2009). “Mountain permafrost”. In Permafrost soils. Berlin: Springer, pp. 33-44.

Gruber, S., Hoelzle, M. (2008). “The cooling effect of coarse blocks revisited: a modeling study of a purely conductive mechanism”. In Proceedings of the 9th International Conference on Permafrost (29), 557-561.

Guo, D., Wang, H. (2017). “Simulated historical (1901–2010) changes in the permafrost extent and active layer thickness in the Northern Hemisphere”. Journal of Geophysical Research: Atmospheres, 122 (22).

Gutiérrez, F., Gutiérrez, M. (2016). Landforms of the Earth. Switzerland: Springer.

Haeberli, W. (1975). “Untersuchungen zur Verbreitung von Permafrost zweischen Flüelapass und Piz Grialetsch (Graubünden).” Mitteilungen der Versuchsantalt für Wasserbau, Hydrologie und Glaziologie der ETH Zürich, 77.

Haeberli, W., Beniston, M. (1998). “Climate change and its impacts on glaciers and permafrost in the Alps”. Ambio, 258-265.

Haeberli, W., Noetzli, J., Arenson, L., Delaloye, R., Gärtner-Roer, I., Gruber, S., ... & Phillips, M. (2010). “Mountain permafrost: development and challenges of a young research field”. Journal of Glaciology, 56 (200), 1043-1058.

Harris, S. A., Pedersen, D. E. (1998). “Thermal regimes beneath coarse blocky materials”. Permafrost and periglacial processes, 9 (2), 107-120.

Hauck, C., Vonder Mühll, D., Maurer, H. (2003). “Using DC resistivity tomography to detect and characterize mountain permafrost”. Geophysical prospecting, 51 (4), 273-284.

Hauck, C., Vonder Mühll, D. (2003). “Inversion and interpretation of two‐dimensional geoelectrical measurements for detecting permafrost in mountainous regions”. Permafrost and Periglacial Processes, 14 (4), 305-318.

74

Hausmann, H., Krainer, K., Brückl, E., Mostler, W. (2007) “Internal Structure and ice content of Reichenkar rock glacier (Stubai Alps, Austria) assessed by geophysical investigations”. Permafrost and Periglacial Processes. 18 (5), 351-367.

Hedding, D. W. (2011). “Pronival rampart and protalus rampart: a review of terminology”. Journal of Glaciology, 57 (206), 1179-1180.

Hendrickx, H., Vivero, S., De Cock, L., De Wit, B., De Maeyer, P., Lambiel, C., Delaloye, R., Nyssen, J., Frankl, A. (2019). “The reproducibility of SfM algorithms to produce detailed Digital Surfacce Models: The example of PhotoScan applied to a high-alpine rock glacier”. Remote Sensing Letter, 10 (1), 11-20.

Hilbich, C., Marescot, L., Hauck, C., Loke , M. H., Mäusbacher, R. (2009). “Applicability of electrical resistivity tomography monitoring to coarse blocky and ice‐rich permafrost landforms”. Permafrost and Periglacial Processes, 20 (3), 269-284.

Hoelzle, M., Wegmann, M., Krummenacher, B. (1999). “Miniature temperature dataloggers for mapping and monitoring of permafrost in high mountain areas: first experience from the Swiss Alps”. Permafrost and periglacial processes, 10 (2), 113-124.

Huggett, R. (2017). Fundamentals of geomorphology. Fourth Edition. Oxon: Routledge.

IAEG Commission on Landslides. (1990). Suggested Nomenclature for Landslides. Bulletin of the International Association of Engineering Geology, 41, 13-16.

Ikeda, A., Matsuoka, N. (2002). “Degradation of talus‐derived rock glaciers in the Upper Engadin, Swiss Alps”. Permafrost and Periglacial Processes, 13 (2), 145-161.

Kääb, A., Chiarle, M., Raup, B., Schneider, C. (2007). “Climate change impacts on mountain glaciers and permafrost”. Global and Planetary Change. 56 (1-2), 7-9.

Kääb, A., & Weber, M. (2004). “Development of transverse ridges on rock glaciers: field measurements and laboratory experiments”. Permafrost and Periglacial Processes, 15 (4), 379-391.

Kenner, R., Phillips, M., Hauck, C., Hilbich, C., Mulsow, C., Bühler, Y., Stoffel, A., Buchroithner, M., (2017). “New Insights on Permafrost Genesis and Conservation in Talus Slopes Based on Observations at Flüelapass, Eastern Switzerland”. Geomorphology. 290, 101–113.

Kneisel, C., Hauck, C., Vonder Mühll, D. (2000). “Permafrost below the timberline confirmed and characterized by geoelectrical resistivity measurements, Bever Valley, eastern Swiss Alps”. Permafrost and Periglacial Processes, 11 (4), 295-304.

75

Kneisel, C. (2004). “New insights into mountain permafrost occurrence and characteristics in glacier forefields at high altitude through the application of 2D resistivity imaging”. Permafrost and Periglacial Processes, 15 (3), 221-227.

Kobierska, F., Jonas, T., Magnusson, J., Zappa, M., Bavay, M., Bosshard, T., ... Bernasconi, S. M. (2011). “Climate change effects on snow melt and discharge of a partly glacierized watershed in Central Switzerland (SoilTrec Critical Zone Observatory)”. Applied geochemistry, 26, S60-S62.

Lambiel C. (2006) “Le pergélisol dans les terrains sédimentaires à forte déclivité : distribution, régime thermique et instabilités”. Thèse, faculté des géosciences et de l’environnement, université de Lausanne, Suisse, pp. 260

Lambiel, C., Pieracci, K. (2008). “Permafrost distribution in talus slopes located within the alpine periglacial belt, Swiss Alps”. Permafrost and Periglacial Processes, 19 (3), 293-304.

Lane, S. N., Tayefi, V., Reid, S. C., Yu, D., Hardy, R. J. (2007). “Interactions between sediment delivery, channel change, climate change and flood risk in a temperate upland environment”. Earth Surface Processes and Landforms, 32 (3), 429-446.

Lerjen, M., Kääb, A., Hoelzle, M., Haeberli, W., (2003). "Local distribution pattern of discontinuous mountain permafrost. A process study at Flüela Pass, Swiss Alps." In Proceedings of the 8th International Conference on Permafrost, 21-25.

Lieb, G.K., Kellerer-Pirklbauer, A. (2011). “Chapter 4: Synthesis of case studies”. In: Kellerer- PirklbauerA. Et al. (eds.): Thermal and Geomorphic Permafrost Respons to Present and Future Climate Change in the European Alps. PermaNET project, final report of Action 5.3., 170-177.

Loke, M. (2002). Tutorial : 2-D and 3-D electrical imaging surveys.

Luckman, B. H. (2013). “Talus Slopes”. In: Elias S.A. (ed.) The Encyclopedia of Quaternary Science. 3, 566-573. Amsterdam: Elsevier

Marchi, L., Chiarle, M., Mortara, G. (2008). “Climate changes and debris flows in periglacial areas in the Italian Alps”. In International conference on hydrological changes and management—From headwaters to the ocean: hydrological change and water management-hydrochange. 111-115.

Matsuoka, N., Ikeda, A., & Date, T. (2005). “Morphometric analysis of solifluction lobes and rock glaciers in the Swiss Alps”. Permafrost and Periglacial Processes, 16 (1), 99-113.

76

Millar, C. I., Westfall, D., Delany, D. L. (2014). “Thermal regimes and snowpack relations of periglacial talus slopes, Sierra Nevada, California, USA”. Arctic, antarctic, and alpine research, 46 (2), 483-504.

Müller, J., Gärtner-Roer, I., Kenner, R., Thee, P., Morche, D. (2014). “Sediment storage and transfer on a periglacial mountain slope (Corvatsch, Switzerland)”. Geomorphology, 218, 35-44.

Nötzli, J., Lüthi, R., Staub, B. (ed.) (2016). Permafrost in Switzerland 2010/2011 to 2013/2014. Glaciological Report (Permafrost) No. 12-15 of the Cryospheric Commission of the Swiss Academy of Sciences. Freibourg.

Omosanya, K. O., Akinmosin, A., & Balogun, J. (2014). “A review of stratigraphic surfaces generated from multiple electrical sounding and profiling”. RMZ Mater Geoenviron, 61 (1), 49-63.

Otto, J. C., Keuschnig, M., Götz, J., Marbach, M., Schrott, L. (2012). “Detection of mountain permafrost by combining high resolution surface and subsurface information–an example from the Glatzbach catchment, Austrian Alps”. Geografiska Annaler: Series A, Physical Geography, 94 (1), 43-57.

Pachauri, R.K., Meyer L.A. (eds.) (2014). “Climate Change 2014, Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.” Geneva: IPCC, 151.

PERMOS (2017). Permafrost in Switzerland 2015/2016. http://www.permos.ch/MM2017/permafrost2016.html. 7/05/2019

PERMOS (2018). Permafrost in Switzerland 2016/2017. http://www.permos.ch/MM2018/permafrost2017.html. 7/05/2019

PERMOS (2019). Permafrost in Switzerland 2017/2018. http://www.permos.ch/MM2019/permafrost2018.html. 10/3/2019.

Phillips, M., Mutter, E. Z., Kern‐Luetschg, M., Lehning, M. (2009). “Rapid degradation of ground ice in a ventilated talus slope: Flüela Pass, Swiss Alps”. Permafrost and Periglacial Processes, 20 (1), 1-14.

Pieracci, K., Lambiel, C., Reynard, E. (2008). “La répartition du pergélisol dans trois éboulis alpins du massif de la Dent de Morcles (, Alpes suisses)”. Géomorphologie: relief, processus, environnement, 14 (2), 87-97.

Ravanel, L., Allignol, F., Deline, P., Gruber, S., Ravello, M. (2010). “Rock falls in the Mont Blanc Massif in 2007 and 2008”. Landslides, 7 (4), 493-501.

77

Ravanel, L., Deline, P. (2011). “Climate influence on rockfalls in high-Alpine steep rockwalls: The north side of the Aiguilles de Chamonix (Mont Blanc massif) since the end of the ‘Little Ice Age’”. The Holocene, 21 (2), 357-365.

Ritter, D., Kochel, R., Miller, J. (1995) Process Geomorphology. Third Edition. Dubuque: Wm. C. Brown Publishers.

Roer, I., Haeberli, W., Avian, M., Kaufmann, V., Delaloye, R., Lambiel, C., Kääb, A. (2008). “Observations and considerations on destabilizing active rock glaciers in the European Alps”. In Ninth International Conference on Permafrost. 2, 1505-1510.

Roer, I., & Nyenhuis, M. (2007). “Rockglacier activity studies on a regional scale: comparison of geomorphological mapping and photogrammetric monitoring”. Earth Surface Processes and Landforms: The Journal of the British Geomorphological Research Group, 32 (12), 1747-1758.

Romanovsky, V., Osterkamp, T. (2000). “Effects of unfrozen water on heat and mass transport processes in the active layer and permafrost”. Permafrost and Periglacial Processes, 11 (3), 219-239.

Russill, J. (2000). “The application of DC resistivity tomography to permafrost prospecting and characterisation in the Swiss Alps”. Diss. project, Dept. Earth Sciences, Univ. Cardiff, Wales.

Samoulian, A., Cousin, I., Tabbagh, A., Bruand, A., Richard, G. (2005). “Electrical resistivity survey in soil science: a review.” Soil and Tillage research, 83 (2), 173 – 193.

Sanders, D., Ostermann, M., Kramers, J. (2009). “Quaternary carbonate-rocky talus slope successions (Eastern Alps, Austria): sedimentary facies and facies architecture”. Facies, 55 (3), 345.

Scapozza, C., Laigre, L. (2014). “The contribution of Electrical Resistivity Tomography (ERT) in Alpine dynamics geomorphology: case studies from the Swiss Alps”. Géomorphologie: relief, processus, environnement, 20 (1), 27 – 42.

Scapozza, C., Lambiel, C., Baron, L., Marescot, L., Reynard, E. (2011). “Internal structure and permafrost distribution in two alpine periglacial talus slopes, Valais, Swiss Alps”. Geomorphology, 132 (3-4), 208-221.

Scapozza, C., Lambiel, C., Reynard, E., Baron, L., Marescot, L. (2009). “Verification of geophysical models of the permafrost distribution within an Alpine talus slope using borehole information, Valais, Swiss Alps”. Geophysical Research Abstracts, 11.

78

Schoeneich, P., Lieb, G.K., Kellerer-Pirklbauer, A., Deline, P., Pogliotti, P. (2011). “Chapter 1: Permafrost Response to Climate Change”. In Kellerer-Pirklbauer A. et al.,(eds): Thermal and geomorphic permafrost response to present and future climate change in the European Alps. PermaNET project, final report of Action 5.3

Schoeneich, P. (2011) BTS: Bottom temperature of snow cover

Schneider, S., Hoelzle, M., Hauck, C. (2012). “Influence of surface and subsurface heterogeneity on observed borehole temperatures at a mountain permafrost site in the Upper Engadine, Swiss Alps”. The Cryosphere, 6 (2), 517.

Scotti, R., Brardinoni, F., Alberti, S., Frattini, P., & Crosta, G. B. (2013). “A regional inventory of rock glaciers and protalus ramparts in the central Italian Alps”. Geomorphology, 186, 136-149.

Scott, W., Sellmann, P., Hunter, J. (1990) “Geophysics in the study of permafrost”. Geotechnical and environmental geophysics, 355-384

Serrano, E., de Sanjosé, J., González‐Trueba, J. (2010). “Rock glacier dynamics in marginal periglacial environments”. Earth Surface Processes and Landforms, 35 (11), 1302-1314.

Sharma, P. V. (1997). “Environmental and engineering geophysics”. Cambridge university press.

Shakesby, R. A. (2004). “Protalus ramparts”. Encyclopedia of geomorphology, 1, 813-814.

Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.) (2013). “IPPC 2013: Summary for Policymakers.” In: Climate Change 2013, The Physical Science Basis. Working Group I Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press, 28

University of Fribourg (2018). Arpille (VS). https://www3.unifr.ch/geo/geomorphology/en/resources/study-sites/arpille.html. 7/03/2019.

University of Fribourg (2018). Grabengufer landslide (VS). https://www3.unifr.ch/geo/geomorphology/en/resources/study-sites/grabengufer-landslide.html. 9/05/2019 van Steijn, H. (1996). “Debris-flow magnitude—frequency relationships for mountainous regions of Central and Northwest Europe”. Geomorphology, 15 (3-4), 259-273.

79

Vonder Mühll, D., Noetzli, J., Roer, I. (2008). “PERMOS – A comprehensive Monitoring Network of Mountain Permafrost in the Swiss Alps”. In: Ninth International Conference on Permafrost. Pp. 1869- 1874.

Wee, J. (2018). Data Processing, GPS measurements.

Westoby, M., Brasington, J., Glasser, N., Hambrey, M., Reynolds, J. (2012). “ ’Structure-from-Motion’ photogrammetry: A low-cost, effective tool for geoscience applications”. Geomorphology, 179, 300-314.

White, S. E. (1981).” Alpine mass movement forms (noncatastrophic): classification, description, and significance”. Arctic and Alpine Research, 13 (2), 127-137.

Wicky, J., Hauck, C. (2016). “Numerical modelling of convective heat transport by air flow in permafrost- affected talus slopes”. The Cryosphere Discuss., 2016, 1-29.

Internetbronnen Carte de l’extension potentielle du pergelisol en Suisse (2005) https://www.bafu.admin.ch/bafu/fr/home/themes/dangers-naturels/info-specialistes/situation-de- danger-et-utilisation-du-territoire/donnees-de-base-sur-les-dangers/carte-de-l-extension-potentielle-du- pergelisol-en-suisse.html. 11/05/2018.

Federal office of Topography (s.d). map.geo.admin.ch

MeteoSwiss (2016) “Climate normal Sion, reference period 1981 – 2010.” www.meteoswiss.admin.ch/product/output/climate-data/climate-diagrams-normal-values-station- processing/SIO/climsheet_SIO_np8110_e.pdf. 14/5/2018.

Maps and data

Badoux, H., Bonnard, E., Burri, M., Visschier, A. (1959) Geological Atlas, GA25

Federal Office of Topography, SGTK (2012). Simplified map of the near-surface mineral resources of Switzerland 1 : 50 000.

Federal Office of Topography (1998) SWISSIMAGE: The Digital Orthophotomosaic of Switzerland Federal Office of Topography swisstopo (2017) Geological Vector Datasets GeoCover.

FOEN (2005). Map of potential permafrost distribution.

Google Earth (2016). Retrieved may 17, 2018.

80

University of Fribourg (2018) “Geodetic (GNSS) Col du Sanetsch”. https://www3.unifr.ch/geo/geomorphology/en/resources/study-sites/arpille.html

University of Fribourg (2018) “Ground surface temperature (GST) Col du Sanetsch.” https://www3.unifr.ch/geo/geomorphology/en/resources/study-sites/arpille.html

10. ATTACHMENTS

10.1 Attachment 1: VES field template: Wenner array

81

10.2 Attachment 2: VES field template: Schlumberger array

82