We thank P.-Y. Galibert and S. Rutard for their helpful assistance during the GPR surveys on the Sachette rock glacier. D. Goodfellow and J. Havskov provided helpful assistance in improving the language. We also thank R. Delaloye and three anonymous reviewers who provided constructive remarks on the paper. The chief editor of Geomorphologie: Relief, Processus, Environnement made additional and helpful corrections to the manuscript.
1Rock glaciers are remarkable high alpine mountain landforms characterized by (1) their striking appearance - long (102-104 m), voluminous (104-107 m3) and well-delineated debris tongues with lava-like surface morphology; (2) their internal structure – apart from relict landforms - permafrost in which the ice may originate from periglacial processes or be inherited from glacier remnants (Washburn, 1979); and (3) their motion – apart from inactive landforms – downslope creep of permafrost due to gravitational forces, in the form of longitudinal displacements of a few cm to a few m.yr-1 (Haeberli et al., 2006). The internal structure has always been the most delicate, controversial, fascinating, and still unresolved issue in rock glacier studies. Indeed, because direct observation (see Haeberli et al., 2006) requires fresh and large exposures - which are very sparse - or boreholes - which are dramatically difficult and expensive - most information about the inside of rock glaciers has been acquired by indirect and therefore interpretative means, for that matter geophysical methods. Geophysics-based studies of rock glaciers have mainly focused on the detection and the characterization of ground ice using seismic techniques (e.g., Hausmann et al., 2007), gravimetry (e.g. Vonder Mühll et al., 2002), transient electromagnetic techniques (Bucki et al., 2004), and especially electrical resistivity (e.g., Hauck et al., 2003; Delaloye, 2004; Ribolini and Fabre, 2006; Bodin, 2007). However, these methods cannot provide accurate visualisation of the internal architecture of rock glaciers, especially stratigraphic features, which are critical in the understanding of rock glacier deposition. The recent and successful use of ground-penetrating radar (GPR) to high mountain debris slopes has provided insights into this topic. Several GPR surveys carried out on rock glaciers in Colorado (Degenhardt et al., 2003), in Svalbard (Berthling et al., 2000; Farbrot et al., 2005) and in the Alps (Otto and Sass, 2006; Hausmann et al., 2007; Monnier et al., 2008) proved to be efficient in revealing the contact between the rock glacier and the underlying till or bedrock floor and pointed to complicated layered geometries in internal structure. Moreover, on a so-called glacier-derived rock glacier in James Ross Island, Antarctica, Fukui et al. (2008) gathered data that revealed debris-bands dipping up in predominant massive glacier ice.
2Within the framework of collaboration between geographers and geophysicists, we have started to investigate rock glaciers in the Vanoise Massif, northern French Alps, with GPR. Until now, we have chosen to concentrate on stratigraphy even if it means failing to detect permafrost. Taking into account recent work demonstrating that rock glaciers can originate from complex and multi-phase evolutions (Guglielmin et al., 2001; Degenhardt et al., 2003), the starting hypothesis of our research is that rock glaciers are not simple one-unit objects. They may be composed of imbrications of spatial and stratigraphic units of varying complexity. By integrating subsurface information from GPR surveys and surface information from aerial photos and field-based geomorphological analyses, the aim of this research is to understand the architecture of rock glaciers in the study area. On the one hand, the data acquired from surface investigations are expected to highlight, or not, major geomorphological boundaries. On the other hand, the data acquired from GPR surveys are expected to depict the stratigraphic features, to highlight, or not, major internal boundaries, and to enable us to examine the possible correlation between internal and surface boundaries. Finally, interpretation should provide insights into the composition or even the way of deposition of the rock glaciers. This paper presents the main principles, steps, problems and limits of our technical approach using two examples of longitudinal GPR profiles.
3The Vanoise Massif (fig. 1) is a mid-latitude (43°30’N) high mountainous area (summits ca. 3500 m) with a concentration of both glaciers and rock glaciers, which is remarkable in the French Alps. Three sites were investigated in this area: the Plan du Lac rock glaciers (Monnier et al., 2008), the Planette rock glacier, and the Sachette rock glacier. The investigation methods described below were used for the Planette and Sachette rock glaciers, two single and pristine rock glaciers. The Planette rock glacier (NNW aspect, 2520-2450 m, fig. 2) is located in the Mont Thabor Massif, a southern appendix to the Vanoise Massif. The rock glacier originates from a NW-facing corrie surrounded by rock walls made of limestones, dolomites and, to a lesser extent, mineralized breccia, so-called cargneules (that characterize a main thrust). The Sachette rock glacier (N aspect, 2630-2480 m) is located in the north-eastern part of the Vanoise Massif, south of the Mont Pourri summit (3779 m), on the northern face of a calcareous and quartzitic crest at an elevation of ca. 2900-3000 m a.s.l. The Planette and Sachette rock glaciers are quite short (~300 m) but are nevertheless conspicuous landmarks: they exhibit archetypal and dense ridge-and-furrow surface features, and have well delineated, steep (ca. 40°) high margins, more than 40 m in the case of Sachette, ca. 25 m in the case of Planette. Both rock glaciers show indications of the probable existence of permafrost. There are no springs at the front of the Planette rock glacier to enable measurement of the water temperature. Nevertheless, according to perfectly interpretable 1970’s map-, high resolution 2001 orthophoto- and field (GPS with ± 4 m accuracy, 2007)-based surveys, the bottom of the rock glacier front has moved a distance of up to 25 m therefore exceeding the total margin of error. Consequently the rock glacier can be considered as active. Equivalent surveys did not show that the Sachette rock glacier was active; however, repeated measurements of water temperature in springs at the front of the rock glacier gave values close to 0°C.
Fig. 1 – General overview of the Vanoise Massif and location of the rock glaciers surveyed.
Fig. 1 – Localisation des glaciers rocheux étudiés au sein du massif de la Vanoise.
1: glaciated areas. 2: rock glaciers (position of the front). 3: southern and eastern border of the study area.
1 : zones englacées. 2 : glaciers rocheux (position du front). 3 : limites sud et est de la zone d’étude.
Fig. 2 – The Planette rock glacier site.
Fig. 2 – Le site du glacier rocheux de la Planette.
The South is in the direction of the background. The picture was taken in November 2007.
Le sud est en direction de l’arrière-plan. Photo prise en novembre 2007.
4For both sites, several sets of air photos were examined and field surveys were made to characterize and map the surficial deposits and features of the rock glacier as well as of the upper slopes. We especially aimed at depicting geomorphological boundaries. Contrary to other rock glaciers previously investigated in the same area (Monnier et al., 2008), both rock glaciers appeared to be typically compact and simple specimens; however, closer studies revealed that the Planette and Sachette rock glaciers in fact have a complex spatial organisation.
5Firstly, regarding the rooting zone where the rock glacier starts in the landscape, and according to the terminology of D. Barsch (1992, 1996), the Planette rock glacier is a debris rock glacier. The scree slopes, which are located in the upper part of the site between 2990 and 2700/2600 m a.s.l., are disconnected from the rock glacier. Upstream, the latter shows clear glacial deposits and features between 2700 and 2500 m a.s.l. (see fig. 2). Thus, the material in the rooting zone is typically till – heterometric, with a relatively fine matrix, and boulders measuring up to several metres. The morphology consists of a system of lateral and median and longitudinal moraine crests. The two largest of them are 20-30 m high and connect to the lateral margins of the rock glacier. The northernmost crest is partly covered by rock fall debris. On the other hand, the Sachette rock glacier does not appear as a typical talus rock glacier. The rooting zone (2650-2600 m) appears as a gently-sloping extension of the upper scree slopes (2780-2650 m) with distinct coloured stripes according to the lithology of the crest outcrops. Although the material does not appear as characteristic till, the presence of two 10-15 m high, rounded lateral crests that connect to the lateral margins of the rock glacier calls into question the possible glacial genesis of the rooting zone (e.g., Whalley et al., 1995).
6Both the Planette and Sachette rock glacier bodies are massive, domed and limited by steep fronts and lateral margins. In both cases, the apparent thickness of the rock glacier decreases from the median part in the downslope direction (from 25 m to 15 m on the Planette rock glacier, from 50 m to 40 m on the Sachette rock glacier). The materials that make up the surface of the rock glaciers are mainly <30 cm blocks. Most often surface of the rock glaciers exhibits mixed petrography (e.g., limestones and dolomites, limestones and quartzites). Most of the morphology of the rock glaciers’ surface consists of convex-profiled, quasi-concentric and arcuate ridges, whose height ranges between 1 and 2.50 m, separated by narrow, V-shaped furrows. In both cases, many arcuate ridges are connected to the downslope end of the lateral crests of the rooting zone.
7Topographical, appearance and sedimentological boundaries are frequent and often correlated. Apart from the lateral and frontal margins, topographical boundaries are defined by break in slope such as particularly high ridges that can turn into front-like slopes, as observed on the Sachette rock glacier. Topographical boundaries can also be related to dissymmetric ridges. Dissymmetric ridges have a convex downslope face and a subvertical upslope face that separate the ridge crest from a depressed area. The Planette and the Sachette rock glaciers exhibit one and three such ridges, respectively. This kind of ridge is known from other cases in the Posets Massif, Pyrenées (Lugon et al., 2004) and in the Thabor Massif, northern French Alps (Monnier, 2006) and is assumed to originate from differential ice ablation and downwasting. Appearance boundaries are related to the occurrence of vegetation. They are especially noticeable on the Sachette rock glacier: blocks in the distal part are significantly covered with lichens, and a distinct decayed rock glacier appendix covered with vegetation is visible in the NE corner of the landform. Sedimentological boundaries are defined by surface variations in petrography or grain-size distributions, which are detected on the surface or within the lateral and frontal steep margins area. For instance, on the SW margin of the Planette rock glacier, in the slope at the limit of equilibrium above the talus cone, fresh and well-exposed outcrops of the material forming the internal rock glacier are visible (fig. 3). In these relatively fine-grained (sands, pebbles) exposed sediments, continuous lines of blocks represent major granulometric boundaries. Finally, the main and continuous geomorphological boundaries can be mapped on the surface of the rock glaciers (fig. 4 and fig. 5).
Fig. 3 Continuous lines of blocks on the SW lateral margin of the Planette rock glacier.
Fig. 3 – Lignes continues de blocs dans le rebord SW du glacier rocheux de la Planette.
1: surface boundaries. 2: lines of blocks on the lateral margin. 3: talus cone. The height of the lines of blocks (h) and the height of the margin (H) are indicated.
1 : discontinuités à la surface du glacier rocheux. 2 : lignes de blocs visibles dans le rebord. 3 : cônes d’éboulis. La hauteur des lignes de blocs (h) est indiquée de même que la hauteur du rebord (H).
Fig. 4 – Simplified sketch of the Planette rock glacier survey based on a 2001 aerial orthophoto (French National Geographical Institute).
Fig. 4 – Croquis simplifié des investigations réalisées sur le glacier-rocheux de la Planette, à partir d’une orthophoto aérienne (Institut Géographique National, 2001).
1: summits of the steep margins. 2: surface boundaries. 3: GPR longitudinal profile layout. 4: intersections between the internal boundaries and the surface.
1 : sommets des rebords raides du glacier rocheux. 2 : discontinuités de surface. 3 : tracé du profil radar longitudinal. 4 : intersections entre les discontinuités internes et la surface.
Fig. 5 – Simplified sketch of the Sachette rock glacier survey based on a 2001 aerial orthophoto (French National Geographical Institute).
Fig. 5 – Croquis simplifié des investigations réalisées sur le glacier rocheux de la Sachette, à partir d’une orthophoto aérienne (Institut Géographique National, 2001).
1: summits of the steep margins. 2: surface boundaries. 3: GPR longitudinal profile layout. 4: intersections between the internal boundaries and the surface. R : rooting zone. D : decayed rock glacier appendix covered with vegetation.
1 : sommets des rebords raides du glacier rocheux. 2 : discontinuités de surface. 3 : tracé du profil radar longitudinal. 4 : intersections entre les discontinuités internes et la surface. R : racine. D : annexe dégradée couverte par la végétation.
8Ground-penetrating radar (GPR) is a non-destructive geophysical technique based on the propagation and reflection of high-frequency electromagnetic waves transmitted into the ground (e.g., Annan and Davies, 1976; Reynolds, 1997; Jol and Bristow, 2003). A GPR unit is composed of a couple of antennae tuned to a given peak frequency, a control unit and a laptop computer for visualization and data storage. GPR surveys are performed along marked tracks. GPR generates radargrams or GPR profiles: GPR profiles display, as functions of the two-way travel time, the arrival of successive waves (called traces) reflected off electromagnetic contrasts corresponding to boundaries between layers or heterogeneous media. Reflections are mainly caused by contrasts in dielectric permittivity and electrical conductivity. The relative dielectric permittivity controls, in the low-loss approximation, the propagation velocity of radar waves; relative permittivity values range between 1 (air) and 81 (demineralised water) and are strongly dependent on water content. The electrical conductivity controls the wave attenuation; the dielectric attenuation processes are considered to be negligible. Consequently, because of overall low conductivity, the GPR application is particularly suitable for rock glaciers, and the few attenuated areas are particularly informative.
9The central pulse frequency used in GPR surveys is chosen according to the nature of the investigation and the type of deposit or bedrock. For a low loss medium, which is generally the case in a rock glacier, several hundred MHz are used for high vertical resolution representations of shallow structures (a few metres in depth), while several dozen MHz are used for deeper structures but with a lower vertical resolution. The horizontal resolution is mainly determined by the offset between the antennas and the step size used along the profile.
10The GPR measurements on the Planette and Sachette rock glaciers were undertaken using a PulseEKKO 100 (Sensors and Software Inc.) unit, with 50 MHz antennae oriented in the broadside mode. The profiles shown as examples in this paper are the result of two longitudinal surveys (fig. 4 and fig. 5). The antenna separation was 2 m and the radar source was triggered every 50 cm along the profile layout. The instrument is quite light and easy to handle but rock glaciers are difficult to hike across, especially the front and side slopes. The survey of the Planette rock glacier during summer had to be stopped before the surface border. In contrast, on the Sachette rock glacier, which was investigated in winter, the field was softened by the snow cover and we were able to walk across the margins. Furthermore, the presence of a snow cover tends to smooth the undulating surface of rock glaciers and thus reduces the tilt of the antennae and the associated anomalies in transmitted and reflected wave angles.
11The common edition of the acquired data included first the merging of the records and the removal of the very low frequency components using a highpass filter (dewow). The two next steps were elevation static corrections and migration. Elevation static corrections are made in order to transform the data recorded on an irregular surface to the equivalent data recorded on a horizontal surface. Migration, which is commonly used in seismic processing, aims to correct dipping and arcuate reflectors distortions; its main visual effect is focusing diffraction hyperbolas. Thus, migration reconstructs the GPR profile in a form which is a better representation of the subsurface structure and then enhances the quality of the interpretation. Migration is a real challenge when applied to two-dimensional GPR data, because the wave velocity can vary strongly throughout the media investigated, and on account of the natural three-dimensional shape of occurring anomalies. Mostly, a common Stolt migration (Stolt, 1978) with a constant velocity is applied.
12Actually, the impact of migration depends to a great extent on the quality of the raw data which is related to the choice of a suitable spatial resolution: in our case 50 cm offset along each profile, which allows submetric resolution of any structure. In the case of extreme surface slope and/or reflector dip, Lehmann and Green (2000) showed that the correct evaluation of a diffraction hyperbola (used as a template for the migration) requires the knowledge of a topography-corrected ray path. The method used to achieve this correction upon acquisition of the surface profile is known as topographic migration. Synthetic simulations showed that GPR profiles are good candidates for topographic migration when: (i) the surface slope is slightly more than 10%; (ii) the error in the estimation of constant velocity is slightly less than 20%; (iii) the error on the surface coordinate is less than 10% of the dominant wavelength of the signal recorded. Theorically, GPR profiles that have been acquired on rock glaciers should be good candidates for topographic migration because of extreme topographic variations. From a practical point of view, this is not the case for the two following reasons. First, regarding the resolution of GPR acquisitions, the internal structuresof thePlanette and Sachette rock glaciers are too heterogeneous for a constant velocity to be reliable. Secondly, geodetic measurements are generally sparse (because they are time-consuming for most rock glacier investigations); it is therefore quite difficult to ensure accurate positioning in addition to fine sampling of topographical z-coordinates. Finally, when one or more of the three conditions required for migration are not fulfilled, commonly migrated data may be preferred to topographically corrected data, as in the case of Planette. Because of important anomalies, the GPR data of the Sachette rock glacier were not migrated at this stage.
13At the display and interpretation stage, the goal is to enhance GPR profiles. Usually, GPR signals decrease rapidly with depth and the signal needs amplifying using a gain function: without any gain, GPR profiles provide very little stratigraphic information. In our work, we use two types of gain: a constant gain and an automatic gain control (AGC). These gains have specific uses. The constant gain multiplies all data by a user-specified value and is useful for highlighting the overall reflectors and the media where radar waves are attenuated. The automatic gain control (AGC) emphasizes reflectors according to the signal amplitude and in a time window of application (AGC window) (fig. 6). AGC is useful for defining the continuity of reflection events in a radar profile. In this study, AGC was used with progressive modulation of the time window of application (between 20 and 200 ns), which is an appropriate method for identifying the internal boundaries of rock glaciers. Indeed, this display process efficiently discriminates reflection events: short AGC windows (20-60 ns) provide limited discrimination whereas large AGC windows highlight the most continuous and prominent events.
Fig. 6 – Effect of the Automatic Gain Control (AGC) on a single trace of a GPR profile.
Fig. 6 – Effet de la fonction de contrôle automatique de gain (CAG) sur une trace extraite d’un profil radar.
14The constant gain-derived longitudinal GPR profile of the Planette rock glacier (fig. 7) revealed complex stratigraphy including many undulating reflection events. The GPR profile can be divided into two main sections. (1) The upper section, which corresponds to the morainic rooting zone, is approximately 200 m long and has a quite steep regular slope. In this section, the amplified signal decreases abruptly from a depth of 15 m, before which there is a dense pattern of quite conform reflectors, parallel to the surface. (2) The lower section, which is 200 m long and corresponds to body of the rock glacier, is characterized by a gentle slope and a ridge-and-furrow profile. In this section, the amplified signal does not decrease before a depth exceeding 40 m. The profile shows a very dense and complex pattern of undulating conform to toplapping reflectors that rarely correlate with the surface. At depth, the reflection events tend to stratigraphic conformity and correlation with the surface. The constant gain-derived longitudinal GPR profile of the Sachette rock glacier (fig. 8) is very different from that of the Planette rock glacier. Apart from one gently bulging section, the surface slope throughout the profile is relatively constant, partly due to the snow cover. Above all, the stratified reflectors are generally straight and the stratigraphy never correlates with the surface topography. The GPR profile can be divided into three main sections. (1) The first 50-100 m-long upper section corresponds to the rooting zone where the amplified signal energy decreases from a depth of 5-10 m, then extends to depth of 60 m a third of the way along the profile. The boundary between the high-energy signal and the low-energy signal below is clearly marked by a continuous series of straight, 50° downslope-dipping, truncating reflectors. (2) In the second section, between ca. 50 m and ca. 275 m downslope, the maximum depth of the surficial deposit is not reached: the signal is still strong at a depth exceeding 60 m. This section corresponds to the high ridge-and-furrow density area of the rock glacier. The stratigraphy, which is very clear and constant, is composed of a dense pattern of smooth, synclinally curved, onlapping reflectors with a remarkably long ~45° upslope-dipping segment, which intersects the surface at a wide angle. From ca. 150 m downslope, these onlapping reflectors become straight lines. A vertical column of diffraction hyperbolas represents an important anomaly in the structure. It is quite similar to GPR features associated with boreholes in glaciers (Fountain et al., 2005) or glacier ice-cored rock glaciers (unpublished GPR profile of the Galena Creek rock glacier, Wyoming): it may be associated with a cavity in the rock glacier structure. (3) Finally, in the distal part of the profile, the stratigraphy is much less clear. However, it is possible to identify a straight, horizontal, truncating major reflector, which intersects the surface at the bottom of the front slope.
Fig. 7 – GPR profiles of the Planette rock glacier.
Fig. 7 – Profils radar du glacier rocheux de la Planette.
A: constant gain-derived GPR profile. 1: conform reflectors, correlated with the surface. 2: toplapping reflectors. R: rooting zone. G: meteorological-technical gap. B: 20 ns-window AGC-derived GPR profile. C: 200 ns-window AGC-derived profile. D: interpreted profile with prominent internal boundaries. The speed used for the time to depth conversion was 0.12 m.ns-1.
A : profil radar obtenu avec un gain constant.1 : réflecteurs concordants et parallèles à la surface. 2 : réflecteurs se terminant par un recouvrement au sommet des réflecteurs sous-jacents. R : racine. G : : lacune météorologique et technique. B : profil radar obtenu avec un CAG et une fenêtre d’application de 20 ns. C : profil radar obtenu avec un CAG et une fenêtre d’application de 200 ns. D : profil interprété indiquant les principales discontinuités. La vitesse utilisée pour la conversion des temps en profondeur est 0,12 m.ns-1.
15The AGC-derived longitudinal GPR profiles (fig. 7 and fig. 8) enhance the main divisions of the stratigraphy. Thus, in the case of the Planette rock glacier (fig. 7), eight main internal boundaries can be identified and most are visible over a long distance along the profile. Three internal boundaries in the upper section of the profile run parallel to the surface but turn into toplapping events at their lower end. The internal boundaries of the lower part of the profile can be divided into two groups: the internal boundaries at the top of the profile do not correlate with the topography, whereas those at the bottom do. The deepest boundary may correlate with the bottom of the frontal slope but, because the GPR profile stopped before the terminus of the rock glacier, this cannot be confirmed. The AGC-derived longitudinal GPR profiles of the Sachette rock glacier (fig. 8) are very different from those of the Planette rock glacier. The major reflection events are more numerous and generally shorter, and, following the general stratigraphy, all of them intersect the surface. The first effect of AGC was to enhance the continuous series of straight, 50° downslope-dipping, truncating reflectors, located 5 to 60 m beneath the surface in the upper part of the profile. This is the clearest major internal boundary of the investigated rock glacier. Furthermore, using AGC sharpened the continuity of the straight, horizontal, truncating reflector, which intersects the surface at the bottom of the rock glacier terminus. In this way, the Sachette rock glacier firstly appears as a clearly delimited stratigraphic body, and secondly shows several slices separated by upslope-dipping boundaries.
Fig. 8 – GPR profiles of the Sachette rock glacier.
Fig. 8 – Profils radar du glacier rocheux de la Sachette.
A: constant gain-derived GPR profile. 1: straight, truncating reflectors. 2: smooth, synclinally shaped, onlapping reflectors. 3: straight onlapping reflectors. 4: diffraction hyperbolas assembled in a vertical line. R: rooting zone. B: 20 ns-window AGC-derived GPR profile. C: 200 ns-window AGC-derived profile. D: interpreted profile with prominent internal boundaries. The speed used for the time to depth conversion was 0.16 m.ns-1.
A : profil radar obtenu avec un gain constant. 1 : réflecteurs discordants rectilignes. 2 : réflecteurs progradants disposés en synclinal. 3 : réflecteurs progradants, rectilignes. 4 : ligne verticale d’hyperboles de diffraction. R : racine. B : profil radar obtenu avec un CAG et une fenêtre d’application de 20 ns. C : profil radar obtenu avec un CAG et une fenêtre d’application de 200 ns. D : profil interprété indiquant les principales discontinuités. La vitesse utilisée pour la conversion des temps en profondeur est 0,16 m.ns-1.
16The internal boundaries are no doubt true structural limits, associated with major variations in the nature, size and fabric of the material. But, given the absence of any borehole through the rock glaciers, their exact characteristic cannot be established. Moreover, the mapping of the identified surface boundaries in comparison with the internal boundaries-surface intersections revealed quite a good correlation between the two (fig. 4 and fig. 5). On the Planette rock glacier, four of the five transversal surface boundaries correlated with an internal boundary. On the Sachette rock glacier, three of the five transversal surface boundaries correlated with an internal boundary; in other cases, the internal boundary is located in the vicinity of the terrain boundary and is useful for enhancing the geomorphological interpretation.
17Applied to the examples described in this article, our approach raises important and fundamental questions related to the overall and detailed architecture of the rock glaciers. (1) GPR profiles make it possible to highlight the contact between the rock glacier and the underlying floor and therefore to estimate the thickness of the rock glacier. In both cases, the basal limit of the rooting zone is physically strong – marked by continuous prominent reflectors, below which the signal dramatically attenuates. On the Sachette rock glacier, the GPR profile crosses the entire terminus zone, where a continuous and truncating prominent reflector is considered as another basal limit. On the Planette rock glacier, the lack of data in the terminus zone prevented clear recognition of such a feature. Finally, the thickness of the Sachette rock glacier varies from a few metres in the rooting zone to 40 m in the terminus zone and ~60 m in the middle of the GPR profile. The thickness of the Planette rock glacier is ~20 m in the rooting zone and may reach 40 m in some parts of the landform, especially just downstream of the rooting zone. (2) As both Planette and Sachette rock glaciers exhibit several boundaries deduced by analysis, they are considered to be composed of several main units. This conclusion raises the possibility of multiphase evolution, as it is the case of other rock glaciers in the Vanoise Massif (Monnier et al., 2008) and elsewhere in the world (Degenhardt et al., 2003). This hypothesis is supported by several lines of evidence: the length and the geometry of the contacts between units, the decreasing thickness of the distal part of the rock glaciers, and the presence at the head of the rock glaciers of possible (Sachette) or obvious (Planette) morainic crests. The latter are connected to the arcuate ridges on the upper surface, and are related to the efficiency of small neoglacial glaciers in transporting rock fall material often to the same place.
18Especially in the case of Planette, there is an interesting link between the upper structure unit and the upper surface moraine unit. The latter was originally described as the rooting zone where the rock glacier starts in the landscape: in fact, it is only the place where the rock glacier landscape was completed by a final glacial deposition. Referring to dental terminology and architecture, instead of the root it is simply the crown. (3) The stratigraphic facies of the Planette and Sachette rock glaciers differ, especially at large and medium scales. The Planette rock glacier stratigraphy, with many undulating and toplapping reflectors, is quite similar to that revealed by GPR in other rock glaciers in the Rocky Mountains (Degenhardt et al., 2003) and in the Alps (Otto and Sass, 2006; Monnier et al., 2008). Furthermore, quite similar GPR layer patterns have also been observed in frontal push-moraine complexes (Bennett et al., 2004; Sadura et al., 2006) or in jökulhaup deposits (Cassidy et al., 2003). The Sachette rock glacier stratigraphy, with many smooth, onlapping and synclinally curved reflectors, is similar to common stratigraphic facies known from compression tectonics; the strong (~45°) upslope dipping of the GPR layers is a clear sign of thrusting. Finally, the Planette and Sachette rock glaciers can be considered as two different expressions of mechanisms with compressive stresses that produce thrusting. The difference in expression may be of rheological significance related with present or past ice content. The validation of this hypothesis requires further investigations.
19With the usual processing and specific image display enhancements, and completed by geomorphological analyses, GPR is an efficient tool for high resolution visualization of rock glacier architecture. Our study shows that: 1) GPR is an accurate method for delineating the sedimentary subsurface body of a rock glacier and consequently for estimating its thickness; 2) the stratigraphy of the rock glaciers studied here is dense and complicated, though the use of window-modulated AGC allowed identification of major internal boundaries; 3) the subsequently demonstrated multi-unit composition of the rock glaciers calls into question their possible multi-phase deposition; 4) rock glaciers can exhibit different stratigraphies and as a result, the exact mechanical processes involved in their genesis are likely to vary. These conclusions call for further work at local scale. We are currently working on other GPR survey modes, data processing and image interpretation techniques that may be useful in detecting permafrost.