Introduction

1Landslides are known to be a major process of landscape evolution (Brunsden and Thornes, 1979; Astrade et al., 1998; Brunsden, 2001; Thomas, 2001; Chang and Slaymaker, 2002; Dapples, 2002; Schmidt and Beyer, 2002; Martino et al., 2004; Dominguez-Cuesta et al., 2007; Moeyersons et al., 2008). Any study of the geomorphological evolution of a mountainous area must therefore take this process into account and try to describe it precisely. The Tertiary basin of Puy-en-Velay is known for its contribution to our knowledge of the Villafranchien period (Bout, 1960), the epigenesis of the Loire (Defive, 1996; Defive et al., 2005; Defive, 2008) and the colonisation of pre-historic man relative to climate evolution (Raynal, 1986, 1988 a and b; Raynal et al., 2001). Few studies have focused on hillslope evolution in the basin, outside those of G. Kieffer (1962), P. Bout (1974), B. Valadas (1984) and A. Le Griel (1990), who worked jointly on a large scale (Massif Central). G. Kieffer (1962) highlighted the major role of landslides and the importance of morphostructure in the back-tilting of basaltic escarpments. P. Bout (1974) and B. Valadas (1984) showed the major impact of climatic variation on the geomorphological processes, in particular the fundamental role of periglacial climate in the transit of sediment from the dismantlement of volcanic edifices by frost splitting and the extensive area covered by periglacial superficial formations. A. Le Griel (1990) made a thorough literature review on the general geomorphological evolution of the Massif Central since the Palaeozoic, pointing out the major contribution of the tectonic crisis of the Miocene and Pleistocene in explaining the present-day general morphostructure and ensuing hillslope dynamics, especially in the south-eastern Massif Central. However, the issue of the geomorphological significance and morphology of landslides in the Tertiary basin of Puy-en-Velay has never been fully addressed. The geological mapping of the southern part of the basin (national programme by BRGM, national geology research centre) and the inventory of landslides over the entire basin to map landslide proneness (Poiraud, 2007) provide new data that help to achieve a better understanding of the importance of landslides in the geomorphological evolution of the basin. The main objectives of the present work were as follows: (i) identifying the principal factors that determine the spatial pattern of relict landslides; (ii) describing the main morphology of relict landslides affecting the hillslope of the basin; (iii) providing information on the ensuing superficial formations; (iv) defining hypotheses on the role of landslides on the evolution of valleys.

Geomorphology of the basin of Puy-en-Velay

2The basin is located at approximately 45°N and 3.85°E (fig. 1). Two major volcanic areas enclose the basin: (i) the Velay massif formed by Mio-Pliocene volcanic activity (basalt and phonolite) to the southeast, and (ii) the Deves massif with Plio-Pleistocene basaltic lava flow to the west. The basin is crossed centrally from south to north by the Loire River, which controls part of the overall erosion by its incision trend. The basin corresponds to the southern extension of the West-European Rift and is directly linked to the Alpine orogeny (Michon, 2000). G. Kiefer (1962) and A. Le Griel (1990) identified the topography of the area as a peneplain of the end of the Miocene. This ante-topography was dislocated by the Pontian tectonic crisis (transition between the Miocene and Pliocene), which set the limits of the different tectonic compartments and fixed the general hydrographic network. The entire area was subject to a general uplift during the Villafranchian crisis (transition between Pliocene and Pleistocene). This major tectonic event, coupled with the cooling of the climate (first glacial period of the old Quaternary) caused the epigenesis of the Loire and corresponded to the beginning of the landscape differentiation (Bout, 1960; Kieffer, 1962; Defive, 1996; Defive et al., 2005; Defive, 2008). During the Quaternary period, the incision of the hydrographic network was coupled with an intense hillslope dynamics connected to the climatic oscillations. The periglacial conditions and the increasing of slope gradients by stream incision caused numerous landslides, shaping the current landscape (Bout, 1948; Kieffer, 1962; Bout, 1974; Valadas, 1984; Van-Vliet-Lanoë and Hallégouët, 1998). The result is a marked relief inversion, with the volcanic structure at the top and an extensive hillslope development in the sedimentary formations. This type of morphostructure is prone to landslide activity. Numerous inherited large landslides border the main valleys, probably connected with the incision timing; others are located under the basaltic escarpment at the top of hillslopes, resulting from volcanic border dismantling.

Relict landslide patterns

3The inventory of relict landslides was produced by aerial photography analysis and a field survey to validate the interpretation of aerial photographs. A literature review provided scant but relevant information. Twenty-nine relict landslides were mapped (7 in the Borne sector, 22 in the Laussonne sector). Fig. 2 shows the location of relict landslides in their geological context. Two significant zones appear, one along the Borne River, west of the town of Le Puy, and the other near the town of Monastier, southeast of the basin:

4- At the overall scale, landslides follow a southeast/northwest line that crosses the basin centrally. This axis corresponds to the main Hercynian fracture orientation, which determines the morphology of the basin (Scanvic et al., 1991; Michon, 2000) and the general orientation of secondary valleys (NW/SE), which are the major axes of erosion.

5- The 'Borne sector' corresponds to the border of the Plio-Pleistocene plateau of the Deves. The incision is recent (less than 2 million years old) and the border erosion of the plateau is essentially concentrated in deep, closed valleys. All the landslides arise at the border of the basaltic plateau, along the streams that incise it, and affect sectors with a height difference greater than 100 m (fig. 3). Lastly, most of the landslides affect the Quaternary formations corresponding to the palaeo-channel of the Loire and/or sediments related to volcanic dams (Defive, 1996; Defive et al., 2005). These formations are made up of sand and clay and are prone to liquefaction.

6- The 'Laussonne sector' is located at the edge of the Velay plateau. The incision is older than the preceding sector (> 7 million years old) and the valleys are deeper and wider. All the landslides arise at the border of basaltic mesa or plateaus. The altitude difference of affected hillslopes is greater than 150 m. This sector is covered by the Oligocene formation of 'Sables de la Laussonne', known for their poor mechanical properties (low cohesion and shear angle) and is prone to landsliding. We also observe that the landslides cover nearly all the hillslopes in this Tertiary sediment.

7- The centre part of the basin is devoid of relict landslides. The geological formations are essentially illitic (prone not to landsliding but to gullying) and the hillslopes are generally concave with a regular topography. This observation is particularly relevant between St-Paulien and Le Puy, where a large plateau forms a vast interfluve between the Loire and the Borne and is disconnected from the incision dynamics of both rivers. Run-off (linear or diffuse) seems to be the main process of landscape evolution.

8As principal control factors determining the pattern of relict landslide we can retain: (i) erosion level of basaltic plateau borders: this factor is determined by the age of the volcanic formations and the distance from the principal hydrographic axes of the basin; (ii) mechanical properties of the sedimentary rocks: the Quaternary formation of the palaeo-Loire River or volcanic dam and the sandy-clay formation of the Oligocene are weak and determine the vulnerability of hillslope to landsliding.

Morphology of relict landslide

9Morphometric measurements of relict landslides are heterogeneous. However, we highlight a difference between the two sectors. In the Borne sector, landslides cover an average surface area of 0.29 km² (s = 0.2) and has an average slope of 13.4° (s = 2.6). In the southeast sector of the Laussonne, these values are respectively 0.61 km² (s = 0.52) and 10.3° (s = 1.9). The elevation extension discussed above ranges from 114 m (s = 17.8) for the Borne sector to 155 m (s = 48.6) for the Laussonne sector. The length is between 514 m (s = 103.1) for the Borne sector and 856 m (s = 303.2) for the Laussonne sector. These two last parameters differentiate landslides in the two sectors (fig. 4). To compare landforms between the two sectors, we analysed two specific representative examples (fig. 2). The first (St-Vidal landslide, Example 1) were prospected by field observation and review of literature and engineering studies. The second (Example 2, Monastier landslides) were investigated by field observation, geological boreholes and geophysical analysis.

The St-Vidal landslide

10This landslide is located on the right bank of the Borne River. It covers an area of approximately 0.73 km². The basaltic plateau forms a large embayment at its border corresponding to the major scarp of the landslide (fig. 5A). Along the entire cross section AB (fig. 5B), from top to bottom, we discern three morphological areas within the landslide:

11(i) a basaltic cornice, covered by a talus scree with a slope > 30°. This is the main scarp of the landslide;

12(ii) a disturbed area with many large outcropped basaltic boulders (length >100 m) and anti-scarps. Two mounds are present upslope the anti-scarp. We observe some outcrops (point with SVF annotation in fig. 5A) characterised by a mixture of sedimentary material (grey sandy clay) and volcanic material (white tuff). The layers of formation of point SVF’ are clearly dipped to upslope, evidencing a rotational movement. This material is reworked and constitutes the matrix of the superficial formations. The general topographic profile is hummocky;

13(iii) a convex area at the bottom of the landslide. Forms are smoother than the previous area and the outcropping of basaltic boulders are less visible. Numerous convexities caused by solifluction flows characterise this part of the landslide. Along the Borne River, erosion outcrops a large weathered basaltic boulder 25 m in height. This boulder seems to be “floating” in the earth matrix and is evidence of slow flowage processes.

14The geological cross section AB (fig. 5B) was built from a field survey and analysis of neighbouring boreholes. This interpretation assumes a rotational component for the large boulders at the top, corresponding to the presumed last phases of basaltic cornice dismantling (III and II), and processes of flowage with “floating” boulders at the bottom. This hypothesis is supported by the difference in basaltic boulder weathering (oxidation spots, important friability, brightening of basaltic matrix) and size between the top part of the landslide (relatively fresh basalt and large boulder) and the bottom part (weathered basalt and dismantled boulder). This observation suggests a progressive back-tilting of the cornice rather than a single paroxysmal event. The boreholes located on the left border of the landslide (fig. 5A) show less than 10 m of reworked material. The sliding material is constituted by the Quaternary sandy clay sediments interlocked in the Tertiary illitic clays. Unfortunately, we have no data on the local hydrogeology.

The Monastier landslides

15These landslides are located at the north face of the basaltic plateau of Monastier, which makes up part of the western border of the volcanic massif of Velay. The basaltic scarp presents a succession of embayments with benches at its foot (fig. 6). This scarp is dismantled with numerous topples of basaltic boulders. The rest of the hillslope presents a hummocky topography, disseminated basaltic boulders, reverse slopes and clearly marked convexities at its foot, which are considered as the bottom of the landslides. Three destructive boreholes were made and seismic refraction profiles constructed along the cross section AB (Poiraud et al., 2008) and morphometric measurements of 130 basaltic outcropped boulders. Earthworks revealed the geological section 'a', located along the cross section CD. This geological section was described and we cut samples for post-requisite dating. Boreholes and seismic profiles along cross section AB are consistent and show a typical profile with three layers:

16- The first one is 2-m thick and is composed of dark silt matrix material with numerous clasts. Seismic velocities are low, less than 1000 m/s. The thickness decreases from P1 to P2 and is less than 1 m at the downslope of the second borehole (P2). It disappears between P2 and P3. This layer is interpreted as a colluvial formation.

17- The second one appears at the downslope of P1. It is constituted by reworked material with clay and sand mixed with basaltic pebbles or small blocks (0-500 mm), or weathered basaltic boulders. Seismic velocities are high, greater than 1500 m/s. The thickness of this layer increases downslope, from 2 m (P1) to 15 m (P3). This is interpreted as the sliding mass. Unfortunately, the destructive boreholes do not allow the structure or microstructure of this mass to be observed.

18- The third one outcrops at the top of the cross section (P1) and is then located below the sliding mass. It is composed of alternating sandy or clayey beds. Basaltic elements are absent from this layer. Seismic velocities are medium, between 1000 m/s and 1500 m/s. The thickness is more than 30 m (interruption of borehole). This layer corresponds to the Oligocene substrate.

19The seismic profile P1 (fig. 7) shows two high velocity zones interpreted as clastic formation or weathered basaltic boulders. We can clearly see the deformation that affects the formation between these two zones with the presence of an ancient depression and reverse slope, both of which were filled by subsequent sedimentation. This deformation is caused by a rotational movement of the upslope clastic formation.

20Concerning the boulders, we observe a strong variation in the volume, number and orientation from top to bottom of the slope affected by landslides (fig. 8). The boulders near the basaltic scarp are not fragmented (30% of all the 130 blocks are located closer than 150 m from the cornice and make up 70% of the apparent volume). Their position is subvertical or flat on the ground, but with an orientation consistent with the overall orientation of the scarp (N-NW). The orientation values are not so widely dispersed. Beyond 150 m, the number of basaltic boulders increases with the decrease in their respective volume (70% of the blocks make up 30% of the apparent volume). The orientation values are more widely dispersed and all the blocks are flat on the ground.

21Concerning the geological section 'a', we observe three major facts (fig. 9): (i) evident deformation of the deposit formation, with a tilting of strata of 15/20° upslope, characteristic of a rotational movement; (ii) the deformation event occurred after the last organic deposition; (iii) the filling of the post-landsliding depression with clastic material. This section is the downslope extension of an actual bench with organic clay. There are no elements from Tertiary sediment in the different formations described. This section shows a similar geometry to that of the seismic profile P1, i.e. a deformation of the layers by a rotational movement from upslope. The absence of Tertiary sediment in the formations is evidence of movements that affect only the formation derived from the scarp dismantling. The two cross sections (fig. 10) were built on the above information. The cross section AB (fig. 10) is more precise for the nature and the thickness of reworked material along the entire hillslope (borehole and geophysical). The cross section CD is complementary, as it clearly shows the rotational component. The sliding mass is not precisely represented because of the lack of boreholes, but we assume that the thickness and shape are similar to the slide body of cross section AB.

22The dismantling of the basaltic scarp is caused by the deformation and flowage of the underlying sandy-clay substratum (fig. 7 and fig. 9). The basaltic panels stand from the scarp with a rotational component. This movement creates a depression under the scarp and a convexity downslope of this depression. The rest of the landslide is made up of a mass of reworked material with sandy-clay sediment mixed with fractured basaltic elements with no clear shear surface. The further away from the scarp, the more blocks are fragmented and mixed with Tertiary sandy-clay sediment.

Discussion

Pattern of landslides and controlling factors

23The pattern of landslide is correlated with the lithostructure and the main erosion axes constituted by the secondary valleys. Both the 'Sables de la Laussonne' formations and Quaternary sandy-clay sediment determines a sector highly prone to landsliding. Their mechanical properties are characterised by a low friction angle due to the presence of clay and low Atterberg limits. The 'Sables de la Laussonne' are locally known for their deformation and flowing capacities and their proneness to solifluction (Bouiller et al., 1978; Girod et al., 1979; Feybesse et al., 1998). This formation is rich in smectite and has a high water absorption capacity (Gabis, 1963; Larqué et Weber, 1978). The Quaternary formations are less well studied, except for the Villafranchian formation, locally called 'sables roux' and 'sables à mastodontes' (Bout, 1960; Ehrlich, 1967). This Villafranchian formation constitutes only one part of the Quaternary filling and does not concern the site of a landslide. E. Defive et al. (2005) give some information about the composition and stratification of one Quaternary sandy-clay sequence, but without granulometric precision or X-ray diffraction. Some technical reports of government geological services offer scattered values of mechanical properties, but not enough to characterise this material completely. However, from a statistical point of view, this sandy-clay formation is highly prone to landslide, like the 'Sables de la Laussonne' formation (Poiraud, 2007). Also, the sandy-clay Quaternary formation is interlocked in the Tertiary illitic sediment, which is totally impermeable. This configuration provides a potential slip plane at the contact between the two formations. The location of the clusters of both the Borne and Laussonne sectors indicates the importance of fluvial axes. These axes concentrate the Plio-Pleistocene uplift effect with a high incision of the valley and high slope values (Kieffer, 1962; Defive, 1996, 2008). These axes are the main control of dissection of both the basaltic plateau of Deves and the volcanic plateau of Velay and constitute incursions for erosion waves inside these plateaus. This process is well known in the Cévennes Massif, where the uplift effect was more expressive due to the proximity of the Rhône base level (Séranne et al., 2001). The example of the Coiron plateau is particularly remarkable (Bacconnier, 1924; Le-Griel, 1990).

Typology of landslides

24The typology of landslides gives information about the processes and mechanisms of the rupture (Varnes, 1984; Cruden and Varnes, 1996; Dikau et al., 1996). The St-Vidal landslide is a large landslide along the Borne valley. The anti-scarps, rotated floating panels, reverse slopes and upslope dip of displaced sediments are evidence of the rotational movements that affect the basaltic columns of the plateau border. The downslope part of the landslide presents characteristic forms of flowage. This landslide is fossil with superficial reactivations that affect the reworked material from the initial landslide phases. Concerning its dynamics, it is difficult to advance a firm hypothesis. The weathering difference between the downslope basaltic boulders and the upslope basaltic boulders could be evidence of successive rotational movements, with phases of retrogressive instability, but the morphologic profile of this landslide could assume the characteristics of a single rotational earthslide. This distinction needs further investigation. Similar morphological cases have been studied in the basin of Puy-en-Velay as the Ceyssac landslide (Bout, 1948) or in the Coiron Plateau, such as the Rochessauve landslide (Masurel, 1977), but with no clear definition of the type of landslide (a slumping type is proposed by both authors). The Collinabos landslide in Belgium (Van-den-Eeckhaut et al., 2007) presents similar morphology in a comparable morphostructural context. Researchers conclude this is a simple rotational earthslide with subsequent superficial deformation inside the landslide debris. The Monastier landslides have a different morphology, i.e. a relatively fresh scarp that delimits the border of the basaltic plateau, a bench zone located just below the scarp with reverse slopes and deformed basaltic clastic formations, and hummocky topography on the rest of the landslide with reworked material (basalt and sediment) and 'floating' boulders. The rotational movement is subsequent to the clastic formation deposition and is considered as a local reactivation of the retrogressive dynamics (scarp back-tilting). This clastic deposition, with rare silt matrix and no bedding disposition, is considered as a periglacial rubble-slope (Etlicher, 1977; Valadas, 1984; Etlicher, 1985; Bertran et al., 2004) and is disconnected from reworking processes that integrate Tertiary sediment. This rubble-slope is contemporaneous with or subsequent to the reworked sedimentary-basaltic formation. This last formation has a high seismic velocity and is disorganised. The facies of this formation and the H/L ratio of the landslides (0.15 to 0.21) are close to the values found by N. Vidal et al. (1996) in the Tertiary basin of Limagne (similar morphostructural context to the Tertiary basin of Puy-en-Velay). The authors conclude that there is a prevalence of post-eruptive debris avalanches that affect the volcanic relief of Limagne or slow-moving landslides that incorporate the volcanic and sedimentary material. Like for the St-Vidal landslide, we cannot draw any conclusion about the kinematics of these movements, but the typology of these landslides is similar to an earthflow or debris avalanche, i.e. movement with visco-plastic behaviour.

Geomorphological significance of landslides

25Landslides cover a large area and represent a large part of the hillslopes in the Tertiary basin of Puy-en-Velay. This proportion is greater inside catchments located on the right bank of the Loire (Miocene plateau of Velay) than catchments on the left bank (Plio-Pleistocene plateau of Deves). We observe a clear correlation between the widening of valleys, the mean slope of catchment and the density of old landslides. The valleys of the Miocene volcanic plateau of Velay are wide and have a more complex topography, with a mean slope between 8° and 9° and a standard deviation between 4° and 8°. All the hillslopes present symptoms of landsliding. By contrast, the valleys of the Pliocene plateau of Deves are narrow and have a more simple topography with a mean slope of 4° and a standard deviation of 4°. Old landslides are located only on particular sites. Dissection degrees of volcanic plateaus and landscape maturity are directly linked to the ages of those volcanic formations and are the expression of erosion work by the essential processes of mass wasting and stream development (Kieffer, 1962; Jefferson et al., 2010).

Conclusion

26The overall study of pattern and morphology of old landslides in the Tertiary basin of Puy-en-Velay allows the main controlling factor of location of those landslides to be determined, i.e. the lithology and the maturity of valleys. Landslides are linked to stream development (retrogressive dynamics) and are a major process of basaltic scarp back-tilting and valley widening. The study case of two sites provides information about the typology of landslides. Movements seem to have visco-plastic behaviour (H/L ratio) and present geometry like earthslides or earthflows with an upslope rotational component. Unfortunately, we cannot draw any conclusion about the kinematics or timing of these landslides, but slow and continuous movements seem to be a more reasonable assumption than rapid and paroxysmal events. Superficial formations derived from landslides cover a large area in the basin and are as important as pure periglacial formations ('head' type). These formations are heterometric, intensely reworked with strong incorporation of volcanic elements with Tertiary sediments and thicknesses ranging from 1-2 m to more than 15 m. They are subject to surficial reactivation today and are highly prone to landslide. This study is still incomplete and requires additional investigations to better characterise the kinematics of landslides and their involvement in the definition of basaltic scarp back-tilting pattern and, more generally, on landscape evolution. Palaeoenvironmental reconstruction is currently being engaged in the Laussonne valley with palynological analyses and dating of relict landslides. This study will allow the link between climate and hillslope instability and the pace of this phase of hillslope activity to be determined.

We thank the 'Natural Hazard Service' of BRGM for its earlier involvement in the study of the geophysical characteristics of the Monastier landslide. We thank David Jarman and two other anonymous referees for spending time reviewing the first draft of this article and for their helpful remarks.