The investigations carried out in the Petit Mont Rouge and Lapires talus slopes were funded by the Swiss National Science Foundation (SNF) (project No. 200021-119854) in the framework of the Ph.D. thesis of Cristian Scapozza performed at the University of Lausanne. The investigations carried out in the Swiss Rhône Valley are part of the Ph.D. thesis of Laetitia Laigre at the University of Lausanne, under the supervision of Prof. Emmanuel Reynard and Prof. G. Arnaud-Fassetta. The ERT data for the Cimadera landslide are reported with the permission of the Ufficio dei pericoli naturali, degli incendi e dei progetti of the public administration of the Canton of Ticino. A special thanks to all those who helped in the fieldwork, to Prof. Emmanuel Reynard for the comments on a first version of the manuscript, and to the anonymous reviewers.
1For the last three decades, geophysical prospecting remarkably increased in geomorphological researches (Schrott and Sass, 2008). For a long time, studies of geomorphological landforms were carried out from direct observation of internal structure visible in outcrops or obtained by diggings and corings, thus in a one-dimension. With the technological improvements like light-weighted materials, a change in the practice of geomorphological investigations was noted thanks to the integration of the geophysical approach (Schrott et al., 2003). The use of continuous measurements along a profile added the horizontal dimension to the vertical one, and consequently the classical one-dimensional approach became bi-dimensional. Moreover, the increase of computer power also contributed both to accelerate and to improve data processing and finally to improve the viewing of geomorphological landforms (Gilbert, 1999; Hecht, 2003).
2Electrical Resistivity Tomography (ERT) is a method based on the capacity of the subsurface to resist to an electrical current. Based upon the electrical properties of materials in ground, it highlights the heterogeneities of both nature and structure of elements in two different dimensions: vertical and horizontal (Marescot, 2006). The potential of this method is higher when the contrast in resistivity between the objects is important (Baines et al., 2002). The wide spectrum of possible resistivity values explains that ERT is used for different geomorphological fields as diverse as permafrost distribution in periglacial environments (e.g., Marescot et al., 2003; Hilbich et al., 2009; Scapozza et al., 2011) or for the analysis of alluvial fans and landslides internal structures (e.g., Godio and Bottino, 2001; Bichler et al., 2004; Sass et al., 2008). In fluvial environments, several studies aimed to evaluate the quality of water in groundwater table (e.g., Decloistres et al., 2008; Coscia et al., 2011; Gourry et al., 2003).
3This paper aims to present the contribution of the ERT method for geomorphological studies in Alpine environments on the base of geo-electrical prospecting carried out in the Swiss Alps in the Cantons of Valais and Ticino (fig. 1). A large part of the paper is devoted to the description of the method and the evaluation of both theoretical potentialities and limits of the modelling during data processing. Then, three selected case studies of the Swiss Alps are presented and several types of contribution of ERT are illustrated: the first one concerns the glacial and periglacial environments with a focus on the study of permafrost distribution and the contribution of the 2.5-D imaging; the second one aims to highlight the interest of geo-electrical study in fluvial geomorphology through the identification of former 2-D river channels geometry; the third one presents the characterisation of the internal structure of a complex landslide. Based on the results of the case studies, several recommendations are finally presented.
Fig. 1 – Localisation of the four sites where ERT prospecting was carried out.
Fig. 1 – Localisation des quatre sites où les prospections géoélectriques ont été réalisées.
1: Les Lapires talus slope, Nendaz area; 2: Petit Mont Rouge talus slope, Arolla Area; 3: Rhône River floodplain near Fully and Charrat, between Martigny and Sion; 4: Cimadera landslide, Lugano area.
1 : Eboulis des Lapires, région de Nendaz; 2 : Eboulis du Petit Mont Rouge, région d’Arolla; 3 : Plaine alluviale du Rhône entre Fully et Charrat, entre Martigny et Sion ; 4 : Glissement de terrain de Cimadera, région de Lugano.
4The electrical resistivity ρ (in Ωm) is defined as the resistance R (in Ω) of a unity cube (1 m3) for an electric current of intensity I (in A) and of potential U (in V) flowing between two opposed faces of the cube. The resistivity of a solid matter (that can be a ground) of section s (in m2) and length l (in m) can be expressed as a function of the resistance as follows:
ρ = R * (s / l) (1)
5The resistivity of a formation depends on the resistivity of the imbibition fluid (which can be ice in frozen grounds), on the porosity and on the degree of saturation, as expressed by the Archie’s law, which is an empirical law derived from laboratory experiments (Archie, 1942):
ρr = a * ρw * Φ-m * S-n (2)
with: ρr = resistivity of the formation (in Ωm); a = coefficient close to 1; ρw = resistivity of the imbibitions fluid (in Ωm); Φ = porosity (from 0 to 1); m = coefficient close to 2; S-n = saturation (from 0 to 1, with n that is often 2). In function of a different degree of saturation and/or of an imbibition fluid presence, the resistivity varies although the rock or the unconsolidated deposits have the same nature. When several layers compose the subsurface, the measured resistivity is not a specific resistivity but becomes an apparent resistivity (ρa), which integrates the resistivities of all the layers crossed by the electric field. The specific resistivities are then calculated by an inverse modelling of the measured apparent resistivities.
6In a frozen ground, the electrical resistivity rises with the decrease of temperature below 0 °C and the increasing of the ice content, following a relationship that is linear for temperatures above 0 °C and becomes exponential for temperatures below 0 °C (Scott et al., 1990):
ρt = ρ18 / [1 + α * (t – 18)] (3)
with: ρt = resistivity at the temperature t (in Ωm); ρ18 = resistivity at 18 °C (in Ωm); t = temperature (in °C); α = coefficient next to 0.025. For example, the chloritic-albitic gneiss of the Métailler Formation (Mont-Fort Nappe; Sartori et al., 2006), in the Penninic Alps of Valais (Verbier area), has an electrical resistivity of 5-8.5 kΩm, typical of gneiss (Hoekstra and McNeill, 1973). At 0 °C, the mean electrical resistivity is 12.3 kΩm, which corresponds to a linear gradient of -0.28 kΩm/°C (fig. 2A). Below 0 °C, the resistivity is 22.5 kΩm at -10 °C, and 135 kΩm at -20 °C.
Fig. 2 – Principles of ERT prospecting.
Fig. 2 – Principes de la prospection TRE.
A: Relationship between resistivity and temperature for the chloritic-albitic gneisses of the Métailler Formation (Mont-Fort Nappe) calculated with the equation 3. Modified from Scapozza (2013). B: Distribution of the current streams and of the equipotentials in the subsurface and disposition of a quadripole at the ground surface. Modified from Meyer de Stadelhofen (1991). C: The electrode arrays most used in the geo-electrical prospecting and the corresponding geometric factor. Modified from Marescot (2006). D: Data acquisition of a pseudo-section in apparent resistivity with the Wenner array. Modified from Loke (2004) and Marescot (2006).
A : Relation entre la résistivité et la température pour les gneiss chlorito-albitiques de la Formation du Métailler (Nappe du Mont Fort) calculée à partir de l’équation 3. Modifié d’après Scapozza (2013). B : Répartition des filets de courant et des équipotentielles dans le sous-sol et disposition d’un quadripôle à la surface du sol. Modifié d’après Meyer de Stadelhofen (1991). C : Les dispositifs utilisés le plus communément dans la prospection géoélectrique et le facteur géométrique correspondant. Modifié d’après Marescot (2006). D : Procédure d’acquisition avec le dispositif de Wenner d’une pseudo-section en résistivité apparente. Modifié d’après Loke (2004) et Marescot (2006).
7The ERT was developed at the end of the 1970s (e.g., Edwards, 1977; Barker, 1981; Dahlin, 2001) and rapidly substituted the “classic” one-dimensional geo-electric (Vertical Electrical Sounding and Electrical Profiling) after 2000. The base principle of the ERT is very simple: when a continuous electric current of intensity I (in mA) is generated between two current electrodes A and B – where A is the injection electrode (positive), and B is the reception electrode (negative) – placed at the surface of a ground, which is theoretically considered as homogeneous and isotropic, a semi-spherical electrical field is created (half-space; fig. 2B), and its volume is function of the distance between A and B. The more the two electrodes are spaced the more the spatial extent of electrical field is. If we add at this two current electrodes two potential’s electrodes, M and N, allowing the measurement of the difference of potential ΔV (in mV) due to the join action of A and B, the resulting quadripole allows measuring the ground apparent resistivity ρa (in Ωm), which is calculated in the following way (Kunetz, 1996):
ρa = (ΔV / I) k (4)
8The geometric factor k (in m) depends on the geometry of the electrode array and the topography, and for each quadripole is:
k = [(AM x AN) / MN] π (5)
9The arrangements of the quadripole at the ground surface differ in function of the electrode array. Theoretically, it exists many electrode arrays (Reynolds, 1997; Marescot et al., 2006); practically, the most used in environmental geophysics are the Wenner, Wenner-Schlumberger and Dipole-Dipole arrays (fig. 2C).
10The accuracy of the three arrays was tested by a pseudo-acquisition on a conceptual model that reflects a known physical reality (fig. 3). For the Wenner array, the electrodes are placed at a constant inter-electrodes spacing (a). This array has a good investigation depth and a good vertical resolution of the subsurface structures (see also Barker, 1979). The Wenner-Schlumberger array is a combination of the Wenner and Schlumberger arrays. In this latter array, the MN distance is normally lower than 1/5 of the AB distance, whereas the AM and NB distance is the same. The investigation depth of the Wenner-Schlumberger array is lower in respect to the Wenner array, but it has a better horizontal resolution of the subsurface structures. The dipole-dipole array, finally, is constituted by two dipoles (AB and MN). It has the best horizontal resolution by comparison with two other studied arrays, but the investigation depth is the lowest (Marescot, 2006; Kneisel and Hauck, 2008).
11The global resolution of each array depends on the number of possible measurements in the same volume of terrain (density of measurements), which has a direct influence on the data acquisition run-time. For a 48 electrodes array, the number of measurements for the Wenner, Wenner-Schlumberger and Dipole-Dipole arrays is, respectively, of 384, 551 and 982. If we consider the global resolution, the data acquisition run-time and the investigation depth, the Wenner-Schlumberger array constitutes the best compromise between the Wenner and the Dipole-Dipole arrays (Marescot et al., 2003; Kneisel and Hauck, 2008). The investigation depth (d) varies in function of the electrode array, the distances a and na (fig. 2C) and the maximal distance between the two current electrodes (L). For the Wenner array where L = 3a, d = 0.173L (Loke, 2004). For the example shown in fig. 3 (48 electrodes all spaced of 4 m), d = 0.173x188 = 32.5 m.
Fig. 3 – Inversion of apparent resistivity data created by a pseudo-acquisition with several electrode arrays on a conceptual model (physical reality) of a periglacial talus slope.
Fig. 3 – Inversion de données de résistivité apparente créées par pseudo-acquisition à l’aide de différents dispositifs sur un modèle conceptuel (réalité physique) d’un éboulis périglaciaire.
12The data acquisition allows obtaining a pseudo-section in apparent resistivity. Along the acquisition line, the measurement of the apparent resistivity is automatically repeated for every possible combination of the quadripole for the selected array (fig. 2D). In the ERT prospecting presented here, the surveys were carried out using an automatic multi-electrode system (piloted by a computer) connected to a resistivity meter, a switching unit and, depending on the length of the prospected structures, one to four electrode cables (with 24 electrodes per cable). The electric current was injected into the ground using 30 to 60 cm long steel stakes, placed as deep as possible. Where large boulders were present at the surface like in glacial or periglacial environments, sponges soaked in salt water were used in order to improve galvanic coupling (Marescot et al., 2003). Finally, for most of the profiles measured in complex topography, the coordinates of each electrode were measured with a differential GPS system (DGPS) in order to know the exact spatial position of each ERT profile and to correct the topographical effects. Where the difference in elevation was low, like in fluvial environments, the topographical correction was neglected. The measurements were carried out using a SYSCAL R1+ Switch 48 or a SYSCAL Pro Switch 96 system (Iris Instruments). As in bi-dimensional geoelectrical imaging resistivity is supposed to be invariable perpendicularly to the measured profile (Marescot, 2006), the ERT profiles were placed preferentially perpendicular to the geomorphological landforms prospected. Several lateral profiles were also measured in order to determine the extension of the prospected formations and for assessing the lateral effects on the perpendicular ERT profiles.
13The pseudo-section obtained in apparent resistivity presents only a broad approximation of the electrical resistivity distribution in the ground, because it is distorted by this distribution and by the electrode array used for the data acquisition (Marescot, 2006). In order to get an electrical resistivity distribution close to the physical reality and then to calculate the specific electrical resistivities, it is necessary to do an inversion of data, which is the inverse processing of the forward modelling.
14The forward modelling consists in the reproduction of the observed data (dsyn) from a model (g) in which we could change the model’s parameters (m). The relationship between the conceptual model and the data is: m → g → dsyn. This purpose is presented in figure 4, where apparent resistivity data were created by a simulation from a conceptual model of the electrical resistivity distribution in a floodplain with a palaeochannel of a former braided river. This model was created by the construction of a network of rectangular blocks of homogeneous resistivity. Then, it was possible to modify the network by affecting specific electrical resistivities at each block of the model for building a physical model in which the “real” geometry of the subsurface is known (fig. 4A). In this particular example, the pseudo-section in apparent resistivities only partially reflects the structure of the floodplain (fig. 4B). The conceptual model and the apparent resistivity data simulation were realised with the software Red2Dmod, allowing to make forward modelling with finite-differences or finite-elements methods (Loke, 2002, 2004). To approximate the real fieldwork conditions, a Gaussian random noise of 2% was added to the simulated data (Hilbich et al., 2009).
15In order to obtain a physically correct model that reproduces the conceptual model, it is necessary to do an inverse modelling of the measured or simulated data (fig. 4C). The inversion modelling consists in a quantitative interpretation of the observed data (dobs), which, by an automatic iterative procedure, allows obtaining the best fit between a model (g-1) and the observed data by the variation of the estimated parameters (mest) of the model. The relationship between the data and the model is: dobs → g-1 → mest. In other words, the inverse modelling in geophysics consists in the construction of a physical model of the subsurface allowing to find the parameters of the ground modelling (inverted resistivities, which must be close to the “real” resistivities), which better corresponds to the data measured on the field (apparent resistivities).
16The inversion of apparent resistivity data to reconstruct the subsurface structures was performed using the 2-D inversion program Res2DInv (Loke and Barker, 1996; Loke and Dahlin, 2002), applying a finite-element calculation for the forward model and including the topography in the data processing by the use of a distorted finite-element mesh (Loke, 2000). The topography was integrated by a Schwartz-Christoffel inverse transformation (Loke, 2000), which is probably the best method to build a model reflecting the reality in ERT profiles with important topographical variations (as in hillslopes) or with complex topography (as in a rockglacier or in a landslide; Loke, 2004). The inversion technique used was of the type Gauss-Newton which, despite the slowness in the calculation of inverse models, is more appropriate that the “classical” quasi-Newton technique for contrast in resistivities higher than 10:1 (Dahlin and Loke, 1998). In order to represent correctly the anomalies located just below the surface without important distortions in the lower parts of the inversion tomogram, a refinement was carried out by using a model where the size of the blocks corresponds to the half of the inter-electrodes distance. The result was to multiply by four the number of the model’s blocks, which allowed reducing of three-quarter the maximum misfit.
Fig. 4 – Forward and inverse modelling of virtual data based on a conceptual model of a unique palaeochannel (gravels and pebbles with an electrical resistivity of 400 Ωm) buried in a sedimentary homogeneous floodplain (saturated clayous silts with an electrical resistivity of 10 Ωm).
Fig. 4 – Modélisation directe et inverse de données virtuelles basées sur un modèle conceptuel d’un unique paléochenal (graviers et galets avec une résistivité électrique de 400 Ωm) enseveli dans une plaine alluviale homogène du point de vue sédimentaire (limons argileux saturés avec une résistivité électrique de 10 Ωm).
17The software Res2DInv offers several possible inversion methods always based on a least-squares inversion that allows the minimisation of the root mean square error (RMS error) between apparent resistivities measured on the field and specific resistivities calculated by inverse modelling. Three inversion methods were tested (fig. 4C and fig. 5A): the “normal” inversion, which is smoothness-constrained, the inversion including smoothing of model resistivity (“smooth” inversion) and the robust inversion. At the opposite of the other two methods, the robust inversion is not based on the minimisation of the root mean square error but on the minimisation of the absolute difference between measured and calculated apparent resistivities, by the calculation of an absolute error. If we look at the buried palaeochannel (fig. 4A), the smooth inversion gives a very rounded result, which does not correspond to the steep limits of the conceptual model. This inversion method is advised for resistivities of the subsurface that change in a smoothed manner, as in the case of pollution plumes. For steep transitions of resistivity, the best method is the robust inversion, which allows the production of inverted models of resistivities distribution with clear and linear borders. This inversion method allows also a good resolution of very high resistivity contrasts (>10 kΩm), that are typical of Alpine periglacial landforms (rockglaciers, protalus ramparts, push moraines) and of hillslope landforms (talus slopes, landslides, rockfall deposits).
18The power of the robust inversion in the resolution of very high resistivity contrasts is shown in a comparison between 10 ERT profiles realised on the Lapires talus slope (fig. 5A; location in fig. 1). After the first iteration, the robust inversion presents always the great error with respect to the “normal” and the “smooth” inversion. This difference is very important, as shown in ERT profiles Lap-5, Lap-6 and Lap-8. The situation after five iterations is completely different: in this case, the robust inversions give systematically the smallest error. This inversion method is then the most adapted for a good convergence between measured and calculated apparent resistivities when the subsurface presents important variations and steep transitions of resistivity.
Fig. 5 – Evolution of the error in function of the inversion method and of the number of iterations.
Fig. 5 – Evolution de l’erreur en fonction de la méthode d’inversion choisie et du nombre d’itérations.
A: Example of 10 ERT profiles measures on the Lapires talus slope (Scapozza, 2013). B: Evolution of the absolute error (dots) and of the maximal calculated apparent resistivity (dashed line) in function of the number of iterations in a robust inversion process. Example of the ERT profile Lap-5 measured on the Lapires talus slope. 1: robust inversion; 2: “normal” inversion; 3: inversion including smoothing.
A : Exemple des 10 profils TRE mesurés sur l’éboulis des Lapires (Scapozza, 2013). B : Evolution de l’erreur absolue (points) et de la résistivité apparente maximale calculée (en pointillés) en fonction du nombre d’itérations lors du processus d’inversion robuste. Exemple du profil géo-électrique Lap-5 réalisé sur l’éboulis des Lapires. 1 : inversion robuste ; 2 : inversion « normale » ; 3 : inversion avec lissage (« smooth »).
19The multiplication of iterations has the effect to search a model of the subsurface always more complex with the goal to minimise the difference between measured and calculated apparent resistivities. As this model is never unique, the multiplication of the number of iterations does not increase inevitably the quality of the resulting model (with the consequence that the error has not inevitably decreased). In figure 5B, it is possible to observe the evolution of the absolute error in the case of real data robust inversion (measured on a periglacial talus slope). The error shows a major decreasing in the first six iterations (from 40% to 3%), and a very small decrease in the next seven iterations (with the minimum of the error function that is reached at iteration 13, with an absolute error of 2.34%). From iteration 14, the error increases again reaching 17% at the iteration 25. A higher number of iterations will have the tendency to overfit the data, with the creation of artefacts by the inversion of wrong data and by an amplification of high contrasts in resistivity (Hauck and Vonder Mühll, 2003; Kneisel and Hauck, 2008). This behaviour is visible in figure 5B, where the maximal calculated apparent resistivity increases in an important manner in function of the number of iterations for the first 10 iterations (from 21 kΩm for the iteration 1 to 34 kΩm for the iteration 10). At the opposite of the example reported by C. Hauck and D. Vonder Mühll (2003), the increasing of resistivity shown here (based on the inversion of measured data) is not linear.
20We conclude that in a terrain with a high level of noise and with important contrasts in resistivity (as it is the case, for example, in rockglaciers or periglacial talus slopes), the optimal model of the subsurface is found for an error lower than 5%, which is normally reached after 5 up to 7 iterations.
21In the Alpine periglacial domain, talus slopes represent the most common landforms, but a few numbers of studies has focused on the permafrost distribution and stratigraphy within these landforms. Several researches carried out in the last years, in particular in the Swiss Alps (e.g., Lütschg et al., 2004; Delaloye and Lambiel, 2005; Lambiel and Pieracci, 2008; Pieracci et al., 2008; Phillips et al., 2009; Scapozza et al., 2011), have shown that permafrost in talus slopes is often distributed in a discontinuous manner. In particular, direct investigations by C. Lambiel (2006) and C. Scapozza (2013) have shown that, in a talus slope, permafrost is present in the lower parts, whereas it is absent in the upper parts.
22From a geophysical point of view, talus slopes are an ideal field to test the limits and potentialities of ERT and to understand why they are often morphologically heterogeneous, they present variable ice content and they have a relatively complex structure. This is the case of the Petit Mont Rouge talus slope (Arolla area; fig. 1), where the presence/absence of permafrost in the northern part of the slope (characterised by a talus slope – protalus rampart complex, superimposed on a morphologically relict rockglacier; fig. 6A) was proved by borehole data (observations, stratigraphy and thermal profiles), and where permafrost is present in the protalus rampart and in the lower part of the talus slope, and absent in the upper part of the slope (Scapozza et al., 2011; Scapozza, 2013). Despite this (relative) simple permafrost distribution, the presence of permafrost in the rest of the slope and the ice content of the frozen sediments is very heterogeneous, as shown by indirect investigation (geoelectrical and seismic measurements and ground surface temperature monitoring; e.g., Lambiel, 2006; Scapozza, 2013). Seven ERT profiles were realised in summer 2009 (fig. 6B): three parallel ERT profiles were measured on an upslope-downslope transect (PMR-1, PMR-2, and PMR-3), whereas four other parallel ERT profiles cross totally or partially the upslope-downslope profiles (PMR-4, PMR-5, PMR-6, and PMR-7).
23According to the stratigraphy and the temperatures of the three boreholes (B0n; fig. 6), the presence of permafrost is probable for electrical resistivity higher than 50 kΩm, whereas it is improbable for values lower than 20 kΩm (Scapozza et al., 2011; Scapozza, 2013). For resistivities ranging between 20 kΩm and 50 kΩm, the presence of permafrost is possible: this signifies that it is very difficult to separate unfrozen porous sediments from frozen sediments with temperate ice. From this interpretation key, the permafrost distribution in the Petit Mont Rouge talus slope is considered very heterogeneous, and characterised by the presence of several (partially) isolated lenses of frozen sediments. This particular pattern in the permafrost distribution (in particular the geometry of the prospected structures) may be precisely visualised thanks to the four parallel ERT profiles perpendicular to the slope (fig. 6B). This is the case, for example, for the resistant body present in the middle-upper part of the slope (and visible in ERT profile PMR-2), which is probably an artefact due to a lateral effect produced by the position of two other resistant bodies located north and south of him, as shown in the ERT profiles PMR-4 and PMR-5 in these area (fig. 6B). In particular, the ERT profile PMR-4 presents a discontinuous body of frozen sediments with a resistivity higher than 50 kΩm and a maximal thickness of 10-15 m in his northern part. According to the resistivities shown in ERT profiles PMR-2 and PMR-4, where the profile PMR-4 crosses these two profiles, in fact, frozen sediments are probably absent. The lateral effect is due to the fact that the 2-D ERT data inversion processing supposes that there are no changes in resistivities perpendicular to the profile (Scapozza et al., 2011), that is obviously not the case at the profile PMR-2, where the inversion process probably considered the two resistive bodies situated north and south of them as continuous.
24In this case study, thanks to the ERT profiles perpendicular to the slope, which allow obtaining a pseudo-three-dimensional (2.5-D) image of the distribution of electrical resistivities in the whole talus slope, it was possible to determine the lateral extent of the frozen bodies prospected, in particular in the middle-lower part of the slope, to specify their three-dimensional structure and to discriminate lateral effects on the lower part of the ERT profile PMR-2. For this case, it is probable that “real” resistivities of this sector are lower than the specific resistivities obtained from the inversion of measured apparent resistivities because of the projection effect on the ERT profile PMR-2 of the high resistivities related to the presence, northern of him, of the protalus rampart’s ice-rich body and, southern of them, of the frozen body characterising the lower part of the talus slope.
Fig. 6 – ERT prospecting of the Petit Mont Rouge talus slope.
Fig. 6 – Prospection TRE de l’éboulis du Petit Mont Rouge.
A: Position of the measured ERT profiles measured (PMR-n). Stars indicate the location of the boreholes (B0n/09). B: Block-diagram of the realised ERT profiles.
A : Position des profils TRE mesurés (PMR-n). Les étoiles indiquent l’emplacement des trois forages (B0n/09). B : Bloc-diagramme des profils TRE réalisés.
25Most of Alpine rivers have been channelised from the second part of the 19th century and are now man-made rivers (e.g., Bravard, 1989, 1993) and the braided morphology that was the dominant pattern at the end of the Little Ice Age disappeared (e.g., Petts et al., 1989; Gautier, 1994; Miramont et al., 2002; Stäuble and Reynard, 2005; Laigre et al., 2009; Scapozza and Oppizzi, 2013; Siché and Arnaud-Fassetta, 2014).
26In Alpine hydrosystems, floodplains located in large valleys are situated at the direct output of torrents flowing from the glacial and periglacial domain. Channelbeds of large Alpine rivers undergo frequent variations of both discharges and sediment supply due to seasonal snow melting and rainfall. Depending of these, local erosive and depositional processes affect either local morphological change inside (lateral bank erosion, channel incision and bars formation) and outside (avulsion) the river channels. Large alluvial plains are ideal environments to observe the complexity and the spatial fitting of fluvial deposits, and to identify with ERT profiles the location of palaeochannels where the mean grain size is higher than in the rest of the floodplain. In fluvial environments, the contrast between the different sedimentological facies is not so high than in periglacial environments, but high enough to distinguish major grain-size classes of sediments. Resistivity values are admitted to vary from 10 Ωm to 1000 Ωm for fluvial formations (Marescot, 2006). A previous research carried out in the Swiss Rhône River floodplain correlating the electrical logs and corings allowed the definition of five major resistivity classes based upon the dominant facies (Laigre et al., 2012): 10-50 Ωm (dominant clayous silts); 50-120 Ωm (dominant sands); 120-300 Ωm (dominant mixed sands and gravels in a silty matrix); more than 300 Ωm (dominant mixed gravels and pebbles in a silty-sandy matrix); 800-1000 Ωm (unsaturated coarse grain size material). As the saturation of the ground may influence the resistivity value; therefore, the water table level has to be known before any investigation to avoid misinterpretation. Because of the seasonal variation of the water table, a test to localise it is previously required (Laigre et al., 2012).
27In this study case, two profiles crossing perpendicularly the Swiss Rhône River floodplain from the northwest to the southeast were measured in Charrat-Fully area (Valais Canton; fig. 1). The profiles were placed perpendicularly to some palaeolandforms previously identified from Digital Elevation Model (DEM) analyses, to evaluate the lateral variability of the channel. As the floodplain is highly urbanised, buildings, roads and railroad tracks have limited the tracing of a unique profile (fig. 7B). Despite this point, different depositional environments have been explored along this transects: the proximal floodplain (near the present Rhône channel; PF; fig. 7), the distal floodplain where a secondary channel was identified (DF; fig. 7), and a lacustrine depositional environments near this secondary channel, identified from historical maps analyses. This transect shows three major information. The northern part of the floodplain (PF and Ful-1 profile; fig. 8), near the present Rhône River channel, shows a range of resistivity values (10-800 Ωm) much higher than in the southern part of the floodplain (10-180 Ωm; DF and Cha-1 profile; fig. 8). On the ERT tomography Ful-1 (in PF environment), two sets of palaeochannels are well-defined: (i) A first palaeochannel is composed by three different coarse grain-size formations between 10 and 30 m of depth, in the middle of the floodplain (fig. 8A to C). The palaeochannels B and C are respectively 90 m and 40 m width and the mean resistivity is of about 250 Ωm, corresponding to a mixed formation (pebbles and sandy matrix). The relative same depth of these two palaeolandforms suggests both a similar time functioning and the existence of a braided pattern with several channels located southern from the present Rhône channel. The third resistive zone located in the southern part of the floodplain is a younger independent channel. (ii) The second palaeolandform is located from 0.8 to 4 m below the surface (fig. 8D). This highly resistive zone (until 800 Ωm) corresponds to the last position of the Rhône River channel before channelisation as attested by some historical maps (Dufour map of 1856; Laigre et al., 2012). This channel had a width of 130 m (against 40 m for the present one). In the present distal floodplain (DF), Cha-1 profile highlights three different depositional environments with, in the deeper part (10-30 m), several palaeochannels 40-50 m large and a relative thickness of 20 m (fig. 8E to G). Overlaying these typical coarse facies, a finer grain-size formation was found, corresponding to a distal floodplain depositional environment. This emphasises a modification of the channel pattern, from braided channels, that was progressively inactivated either by channel migration southward or by hydrosedimentary change of flow due to climate change. The present location of the “Older Rhône” (fig. 8H) supports the assumption of a migration.
28Thus, this case study shows that ERT method presents a good potential to locate major buried palaeolandforms and to evaluate the lateral evolution of a channel and the different geometry of several successive palaeochannels.
Fig. 7 – Location of the two ERT-profiles in the Swiss Rhône River floodplain (Ful-1, Cha-1).
Fig. 7 – Localisation des deux profils TRE situés dans la plaine alluviale du Rhône Suisse (Ful-1, Cha-1).
PF = proximal floodplain, DF = distal floodplain of the Rhône River. COR-01 and COR-02 = boreholes.
PF = plaine proximale, DF = plaine distale du Rhône Suisse. COR-01 et COR-02 = forages.
Basemap source: Federal Office of Topography swisstopo.
Source du fond de carte : Office fédéral de topographie swisstopo
Fig. 8 – ERT profiles Ful-1 and Cha-1 emphasizing several generations of palaeochannels highlighted with high resistivity layers (in black and grey).
Fig. 8 – Profils géoélectriques Ful-1 et Cha-1 montrant plusieurs générations de paléochenaux, mis en évidence par la présence de niveaux à haute résistivité (en noir et grisé).
29The geophysical prospecting of landslides is often quite difficult because of the large heterogeneity of the subsurface (presence of cracks, fissures, trenches, shear zones, etc.) and the logistical problems due to topographical conditions (steep slopes, vegetation cover, etc.). These difficulties are often increased by the presence of anthropogenic infrastructures (e.g., roads, buildings, filled canalisations, pipelines for gas or electricity), which can noise the measured data and have an important effect on the interpretation of geophysical models. Despite these points, several geophysical methods were applied in the prospecting of landslides and several interpretation techniques were developed specifically for this kind of landforms (e.g., Bogoslovky and Ogilvy, 1977; Bichler et al., 2004; Jongmans and Garambois, 2007).
30The Cimadera landslide (Val Colla, southern Swiss Alps; fig. 1) presents all the field and logistic difficulties listed above. The structure of the mass movement is very heterogeneous because the large rotational landslide is included in a deep-seated gravitational slope deformation (DSGSD), which comprises the whole slope of Cimadera (Seno and Thüring, 2006; Ambrosi and Strozzi, 2008). In addition, the main landslide is composed in its lower part by several other active landslide bodies, with a thickness of some metres to some decametres, associated with zones of important regressive erosion.
31The natural geological and geomorphological framework is complicated by several anthropogenic structures, in particular an aqueduct in the upper part of the slope which contributes, in association with a fault related to the DSGSD, to saturate the ground in water and to transfer important quantities of water to the main active part of the landslide. In the lower part of the landslide, just above the several small landslide bodies, several springs associated with artificial water drainage supply in water some small streams that contribute to the regressive erosion of the main erosional escarpment on which is located the village of Cimadera.
32Seven ERT profiles were carried out in October 2012 within the perimeter of the Cimadera landslide. Five profiles (Cim-1, Cim-2, Cim-4, Cim-6 and Cim-7) were realised transversally to the slope, whereas two ERT profiles (Cim-3 and Cim-5; fig. 9) cross on an upslope-downslope transect the profiles Cim-1/2 and Cim-4 allowing to obtain a pseudo-three-dimensional geoelectrical image of the lower part of the Cimadera landslide and to discriminate eventual lateral effects produced by the heterogeneity of the prospected structures (see Case study 1). In order to provide a better viewing of the behaviour of electrical resistivities with the depth and to determine the depth of the subsurface structures (in particular the transition into the bedrock and the shear zones), one-dimensional virtual vertical profiles of resistivity were extracted from the tomograms in several selected points (fig. 10).
33Several superimposed landslide bodies are present in ERT profiles Cim-3 and Cim-5 (fig. 10). In particular, in profile Cim-3, it is possible to show a low resistivity anomaly (<0.2 kΩm), which ends suddenly about 40 m of distance, in correspondence of several trenches and scarps and of a spring. This anomaly can be related to the aquifer, which emerges in correspondence of the main scarp of the landslide body. Below this point, it is possible to find the higher resistivities typical of the fractured rock mass. The shear zone is evident on almost all the extension of the landslide body. The structure of this landslide body was determined thanks to the vertical profile of resistivity 07_17 (fig. 10), which presents: a layer 10-12 m thick, with decreasing resistivities, which corresponds to the fractured rock mass characterised by a decreasing porosity with the depth; a more conductive layer between 12 and 20-22 m deep, probably related to an important presence of water and friction gouge in correspondence of the shear zone; a deep layer with a resistivity of ca. 1 kΩm, value which is typical of the deepest part of almost all the vertical profiles of resistivity (fig. 10) and that characterises the unfractured bedrock.
34ERT prospecting of the landslide allowed the characterisation of the electrical stratigraphy of the deep layers, of the geometry and depth of the shear zones associated with several landslide bodies highlighted by detailed cartography, and of the hydrogeological conditions of the whole slope (in particular the transfer to the lower part of the Cimadera landslide of water that infiltrates in the ground in the upper part of the slope in relation with an aqueduct and/or a main fault related to the DSGSD). In a methodological point of view, the analysis of ERT profiles realised on the Cimadera landslide was prop up by the analysis of one-dimensional virtual vertical profiles of resistivity extracted from the tomograms, which allowed the definition of common characteristics of the subsurface helping the interpretation of the electrical stratigraphy.
Fig. 9 – ERT profiles Cim-3 and Cim-5 realised on the Cimadera landslide. Upper right, localisation of all the ERT profiles measured on the Cimadera landslide (with, in grey, ERT profiles not presented and discussed here) and of the one-dimensional geo-electrical profiles extracted from the 2-D tomograms and presented in fig. 10.
Fig. 9 – Profils Cim-3 et Cim-5 de TRE réalisés sur le glissement rocheux de Cimadera. En haut à droite, localisation de tous les profils de TRE mesurés sur le glissement rocheux de Cimadera (avec, en gris, les profils de TRE qui n’ont pas été présentés et discutés ici) et des profils géo-électriques unidimensionnels extraits des tomogrammes 2-D et présentés dans la fig. 10.
Fig. 10 – One-dimensional geo-electrical profiles extracted from the 2-D tomograms measured on the Cimadera landslide. For the localisation of the profiles, see fig. 9.
Fig. 10 – Profils géo-électriques unidimensionnels extraits des tomogrammes 2-D mesurés sur le glissement rocheux de Cimadera. Pour la localisation de ces profils, voir fig. 9.
35From the observations and ERT prospecting at the three study sites, it is possible to draw the followings recommendations:
In an environment that can be extremely heterogeneous, like an Alpine talus slope or a complex landslide, it is necessary to adopt a geo-electrical prospecting strategy based on the measurement of ERT profiles parallel and perpendicular between them, by the realisation of a network of profiles. On the one hand, this approach allows determining the three-dimensional extent of the prospected structures and, on the other hand, discriminating the real structure of the subsurface from those created by a lateral effect due to an anisotropic structure of the terrain.
At the opposite, in fluvial environments the geo-electrical prospecting strategy can be based on transversal profiles to the floodplain, cutting the main river streams perpendicularly and allowing to define the general geometry (width, depth) of the prospected palaeolandforms (palaeochannels, gravel banks, buried oxbow lake deposits, etc.).
The extraction of vertical profiles (virtual boreholes) from ERT tomograms based on specific resistivities allows a better understanding of the structure of the subsurface. This simple operation is particularly efficient in transition zones with a high gradient of resistivity, where a two-dimensional representation in a tomogram cannot represent correctly an abrupt change in resistivity, as it is the case in a one-dimensional vertical profile. These vertical profiles are interesting not only for the specific resistivity values, but also for determining an electrical stratigraphy of the subsurface. This is the case, for example, in an Alpine periglacial deposit, where the presence of a resistant body intercalated between two more conductive layers, can be interpreted as a permafrost lense (Scapozza and Lambiel, 2013). As presented in the Case study 3, vertical electrical profiles extracted from ERT profiles, allow a better definition of the depth of the shearing zone of the landslides.
When additional information about the structure of the subsurface are available (e.g., boreholes, other geophysical data, estimations of the ice content or of the depth of the bedrock, etc.), the direct modelling of the distribution of electrical resistivities allows a validation of the interpretation of ERT profiles, and this both for the measured values and for the geometry (extent, depth) of the detected structures. In absence of other techniques of evaluation of the quality of ERT profiles, as the Depth of Investigation Index (DOI; Marescot et al., 2003) or the Model Resolution Matrix (Hilbich et al., 2009), the direct and inverse modelling of conceptual models based on “real” resistivities is necessary for obtaining independent information on the quality of the realised ERT profiles (beyond information on the inversion errors, which depends on the quality of the measured data).
36Despite of an increasing use of Electrical Resistivity Tomography in geomorphology, several gaps are always present on the correct response of the method to the important spatial (and temporal) variations of resistivity that characterise some complex geomorphological landforms, as for example periglacial talus slopes, glaciers forefields or complex landslides. These landforms can constitute an ideal terrain to test geophysical methods applied in geomorphology, the evaluation techniques of their quality and their limits. In the medium term, this could allow an improvement of the interpretation of geo-electrical two-dimensional data in the optic of a better resolution of the spatio-temporal heterogeneities, which characterise some dynamic landforms typical of the Alpine environments.