6.1 Pre-failure situation and gliding surface
It has long been known that the geologic structure of the source area plays an important role in the triggering of rock avalanches. Important parameters include the mechanical properties of the rocks and pre-existing deformation features like joints or foliations, and the presence of a potential detachment horizon. The Flims rock avalanche developed in predominantly massive limestone on a steep valley flank carved by glaciers. By the time the rock avalanche was mobilized the glaciers had melted away for at least 5000 years. Thus, it is unlikely that a direct link between the disappearance of the support of the ice on the mountain flank and the triggering of the rock avalanche exists. However, the disappearance of the glaciers had an isostatic effect, which initiated faults raising the valley-side compartments relative to mountain-side compartments (Ustaszewski et al., 2008). In addition, substantial temperature fluctuations occurred in the time interval between the LGM and the time of the rock avalanche events (Blümel, 2002). For example, the warm period of the Bölling interstadial was followed by the cold Younger Dryas, which in turn was followed by a longer warm period with ice-free Alps (the so-called Climate Optimum). Both, the Tamins and Flims rock avalanche events occurred in the early phase of this warm period. It can be argued that the preceding temperature changes with freezing and thawing permafrost had fractured and weakened the bedrock. The effect of water percolating in ice-free tension cracks, along joints, foliation planes and bedding planes may affect stability at greater depth (Fischer et al., 2010; Gischig et al., 2011a, b; Kenner & Phillips, 2017). With climate warming at the onset of the Climate Optimum, permafrost became warm and at − 1 to 0 °C tensile strength and frictional strength between formerly ice-bound bedrock pieces are greatly reduced (Davies et al., 2001; Haeberli & Hohmann, 2008) leading to destabilizing slopes.
The paleo-landsurface is interpreted to have a steep slope just north of the Plaun fault where bedrock consists of «Upper Quinten limestone», a poorly bedded massive limestone which tends to form high cliffs throughout the Alps. As discussed by Wolters and Müller (2008) such abrupt slope angle changes give rise to gravity-induced high shear stress concentrations activating a glide surface emanating from the break in slope. The cross-sections showing the paleo-landsurface and the glide surface of the Flims rock avalanche (Fig. 7A) highlight the lateral continuation of the break in slope.
Pre-existing structural weaknesses were likely to have influenced the instability. In the Flims rock avalanche, both lateral scarps follow a fault system. New mapping has revealed that the eastern fault runs along the valley of Bargis down to Trin Mulin, the western from Plaun Segnas Sut down towards Flims. Both faults are parallel to the main regional jointing system (Pfiffner, 1977) and allowed some independent deformation within the nappe pile on either side. As discussed by Volken (2015) and Aaron et al. (2020) the lateral scarp located on the western flank of Flimserstein is also parallel to the main jointing system and associated faults. Additional vertical fault zones and joints enhanced the effect of percolating water and air venting in destabilizing the rock mass and ultimately controlled the orientation and location of the lateral scarps.
Remains the question if the basal gliding surface of the rock avalanches is also structurally controlled. Recent mapping (Pfiffner & Wyss, 2023, and work in progress) combined with data from the road tunnel in Flims have provided crucial information. The gliding surface is sub-parallel to bedding in the lower reaches (see Fig. 19) and cuts up section towards the head scarp. The gliding surface has a flat segment near Punt Desch from where it cuts up section to the north. Volken (2015) and Aaron et al. (2020) speculated that the «Mergelband» acted as detachment layer. However, construction of the Flims tunnel accompanied by several boreholes (Geologengemeinschaft Umfahrung Flims 2007) and recent mapping revealed that the «Mergelband» is located below the top-bedrock surface beneath the rock avalanche deposit as shown in Fig. 19. Moreover the «Mergelband» is not a marl, but a 20–30 m thick sequence of thin bedded limestone in which 2–5 cm thick limestone beds alternate with 2–3 cm thick slightly argillaceous limestone beds (Pfiffner, 1972a, b).
A penetrative Alpine foliation is present in the limestones above and beneath the gliding surface, which is oriented sub-parallel to the gliding surface and may represent a potential weakening fabric. Flattened calcite grains and occasional sheet silicates oriented parallel to the long axes of the calicite grains define this foliation (see Fig. 20). This anisotropy may have controlled the orientation of fracturing.
The structural level of the gliding surface is possibly controlled by the pre-failure topography (Figs. 7A and 19) and the associated high tensile stresses at the break in slope at the foot of a steep limestone cliff (Savage, 1993; Wolters & Muller, 2008).
As mentioned earlier, the gliding surface contains steps and lateral ramps that are up to 50 m high. The lateral ramps are parallel to the lateral scarps and possibly also structurally controlled. More interesting are the steps where the basal gliding surface jumps to a lower level. The limestone forming the step are highly shattered and fragmented into blocks such that one is often not sure whether to consider it part of the rock avalanche or the in situ collapsing substrate. The situation is presented in Fig. 21 with two blocks swimming in a crushed matrix of the rock avalanche. Block A was detached but remained coherent such that bedding is preserved. Block B is shattered with steeply dipping joints cutting across a more brecciated part. It seems that the blocks were torn away from the bedrock at the step and gradually incorporated into the rock avalanche, reflecting the transition from incipient collapse of the substrate to comminution in the moving rock avalanche.
6.2 Disintegration of rock avalanche deposit
There are two ways to look at the internal structure of a rock avalanche. One is to consider the lithological composition in relation to the source area, the other the deformation state reflecting the mechanical behavior of the rocks. In case of the Flims rock avalanche the lithologies in the rock avalanche deposit show a distribution surprisingly akin to the one in the source area. Flimserstein has a simple structure (see Fig. 19), which most likely also existed in the broken away mass to the south and west: a 600 m thick succession of Late Jurassic limestones, i.e. Quinten and Tros Limestone overlain by about 100 m of Early Cretaceous limestones. These in turn are overlain in thrust contact by klippen of Permian volcaniclastics («Verrucano»). The thickness of these volcaniclastics is at most 100 m on Flimserstein but could have been higher in the broken away parts. Mapping the lithologies in the rock avalanche (Gsell, 1917/18 and own work) reveals that the lower part exposed in the gorge of Ruinaulta is exclusively Quinten Limestone at the base of the cliffs and Tros Limestone in their upper part. The forested area above the cliffs contains Cretaceous limestones and locally «Verrucano» (Fig. 5). Although the thick forest in this area conceals a lot of information, Cretacous limestones and «Verrucano» blocks outcrop in distinct areas. Only the boundary between areas with blocks of Tros limestone and Cretaceous limestones is sometimes difficult to trace because of some mixing. It thus follows again that the rock avalanche moved as a more or less coherent (internally broken) mass with no sign of turbulent flow.
Regarding the deformation state within the rock avalanche deposit, the observation is in concert with rock avalanches world wide: a carapace at the top of a fragmented body and a basal mixed zone (see e.g. McSaveney et al., 2006, Strom, 2006; Dufresne et al., 2016b or Zhang et al., 2019). The distinction between differently fragmented body parts is unlike the Tschirgant rock avalanche (Dufresne et al. 2016a) simpler in the Flims rock avalanche due to its homogeneous composition. This structural distinction is shown in the cross-section of Fig. 8. The blocky layer of the carapace travels along passively (Strom, 2006) but individual blocks, which can escape upward in the rock avalanche, are free to rotate. This is supported by the observation that these blocks display chaotic orientations of bedding and foliation.
As Strom (2006) and Zhang et al. (2019) discuss, the body of rock avalanches beneath the carapace is highly shattered and thus has a lower strength facilitating flow. Dynamic fragmentation comminutes the rocks and leads to the formation of matrix-supported breccias (McSaveney et al. 2006). This process permits granular flow to happen and is according to these authors a prerequisite for a rock avalanche to spread.
As discussed earlier, fragmentation within the Flims rock avalanche deposit is inhomogeneous. Large compartments 200–300 m in diameter float within a highly fragmented and comminuted matrix and are obviously able to travel as semi-coherent compartments, semi-coherent meaning that they contain original bedding structure but are heavily jointed (see Fig. 22). The outcrops above Isla Davos (coord. 2745000/1185600) contain large compartments within the rock avalanche, which retained the original stratification in the Quinten limestone. The outcrop 300 m east of Isla Davos shows a tilted compartment with a diameter of 200 m (Fig. 22A). In the outcrop 300 m SW of Isla Davos a slightly tilted compartment is embedded in a cataclastic limestone, which develops scree slopes (Fig. 22B). The compartment contains a 15 m thick light layer, which might represent a dolomitic passage within the limestone. Dolomitic passages are typical for the «Upper Quinten limestone» (Pfiffner, 1972b). As discussed earlier, the Flims rock avalanche deposit is also cut by km-sized faults (see Fig. 9C and 11A) that express themselves morphologically.
The comminuted matrix corresponds to a cataclasite resp. matrix-supported breccia with a “block-in-powder” texture (see Fig. 9C). The process of cataclasis is illustrated in the outcrops shown in Fig. 23A, B and C. The outcrop at Isla Casti (coord. 2743300/1185770) is an erosional scarp of the Vorderrhein. In the larger scale view of Fig. 23A a subvertical fault cuts across the brecciated limestone. It is topped by gravel deposited in the course of river incision into the rock avalanche deposit. A 30 cm wide fault gouge made of comminuted limestone outlines the fault. Close inspection of the outcrop reveals that thinner fault gouges accompany the major one on either side, and within the brecciated limestone some anastomosing veins are aligned parallel to the fault. This suggests that brecciation and comminution are associated with the fault gradually invading the neighboring limestone. The detailed view in Fig. 23B shows a brecciated dark gray (Quinten) limestone containing several generations of light colored and very fine-grained veins. The veins are either short and straight or long and anastomosing. Vein material is a comminuted very fine-grained limestone powder. In several places, the anastomosing veins become thick and contain small angular fragments of dark gray limestone. This is particularly true for old anastomosing veins, which are cut by younger and thinner veins. All in all, it seems that multiple veining increases the volume of comminuted (crushed) limestone gradually. It is however impossible to interpret the veining sequence in a simple scenario of shearing.
In the quarry Carnifels (coord. 743050/1184240) at the southern border of the Flims rock avalanche a S-dipping fault zone may be observed (see Fig. 23C). A 10 m thick layer containing many limestone blocks up to 2 m in diameter cuts across a body of comminuted limestone containing fewer and smaller limestone blocks. A thin fault gouge of comminuted limestone marks the contact on either side of the layer. This outcrop suggests that comminution and faulting at km scale occurred simultaneously in successive steps.
Schneider et al. (1999) and Pollet and Schneider (2004) who studied microstructures in the Flims rock avalanche claim that major discontinuities of 10 cm thickness constituted of comminuted rock developed from bedding planes; they are supposedly sub-horizontal and accompanied by limestone fragments bound by joints and displaced in a book-shelf manner. The present study could not confirm this interpretation.
Summing up it seems that the Flims rock avalanche was disintegrated internally as it moved. Disintegration occurred at all scales including km-scale planar faults, preservation of large semi-coherent compartments and pervasive comminution throughout the body of the rock avalanche. The planar faults within the rock avalanche debris cannot be correlated to pre-existing faults in the source area; on the other hand, a reactivation of faults or heavy jointing in the the source area cannot be ruled out. The process of fragmentation is illustrated in Fig. 24. In a first stage comminution occurs along fractures cutting across the limestone and results in veins made of comminuted limestone powder. As deformation continues multiple cross-cutting veins appear. At the intersections of veins pieces of limestone embedded in comminuted limestone powder persist. Ongoing deformation then results in bodies of comminuted limestone powder containing smaller or larger angular fragments of limestone. Local stress conditions within the moving rock avalanche determine to which stage fragmentation develops, leaving high-grade and low-grade fragmented volumes side by side.
The basal zone of the Flims rock avalanche is not exposed in direct contact with the water saturated gravel of the former valley floor. However, the dark brecciated limestone outcropping in direct contact with the bedrock in Val da Mulin is identical to the lowermost dark colored brecciated limestone shown in Fig. 9A and B; more importantly, it does not differ in texture from the overlying gray brecciated limestone. Weatening by percolating water is responsible for its dark color and points to enhanced permeability at the base of the rock avalanche deposit. At several locations the dark limestone is offset by faults, which suggests that the higher permeability was attained in early stages of motion and disintegration of the rock avalanche.
Given that injections of silt, sand and gravel are observed in the lower part of the Ruinaulta gorge, a complex basal zone with mixture of rock avalanche deposit and substrate is to be expected (as shown in the cross-section of Fig. 8 and in Fig. 25).
6.3 Long-runout
The problem of how to explain the long-runout of rock avalanches has been addressed by many scientists proposing a variety of mechanisms. Davies and McSaveney (2012) give a summary of this conundrum. Explanations put forward include for example air-layer lubrication, fluidization by extruding air, dust dispersion, steam from frictional heating, gaseous pore pressure or melting of rock (see Davies & McSaveney, 2012 for references). Additional explanations (op. cit.) invoke high vibration or pressure wave propagation as well as loading of saturated substrates and mechanical fluidization.
The authors conclude that from an energetic point of view dynamic rock fragmentation is the most suitable process to explain long-runout. In particular they emphasize the effectiveness of the release of elastic energy of breaking grains. Apart from this, they note that fragmentation is universal in long-runout rock avalanches. Considering the Flims rock avalanche fragmentation is in fact ubiquitous.
Abele (1997) and Calhound and Clague (2018) consider that liquefaction of the substrate consisting of water saturated glacio-lacustrine and fluvial sediment played an important role for explaining the long-runout of the Flims rock avalanche. Liquefaction may be enhanced by entrainment of substrate pieces as Hungr and Evans (2004) have shown for the Eagle slide and the Momash River slide in British Columbia (Canada). Zeng et al. (2021) observed entrainment and injection of liquefied substrate from the Nixu rock avalanche in southern Tibet. In this case the rock avalanche moved on a basal shear zone and bulldozed and gradually overran the substrate at the front of the rock avalanche. The entrained substrate was subsequently liquefied and injected upward into the rock avalanche. Similarly, in case of the Tschirgant rock avalanche Dufresne et al. (2016a) report that the substrate was intricately involved in the internal deformation processes.
Aaron and McDougall (2019) studied the phenomenon of long-runout of rock avalanches as function of the substrate. They concluded that saturated substrate correlates with longer runouts in comparison to bedrock substrate. Their data are shown in Fig. 26, which is complemented with the H/L ratios of the Tamins and Flims rock avalanches. The Flims rock avalanche with a H/L ratio of 0.18 plots way beneath the rock avalanches with bedrock and unsaturated substrate, in the field of saturated substrate.
For the Flims rock avalanche there are several observations that need to be considered regarding long-runout. First of all, there are several places where the rock avalanche deposit is in steep contact with the Bonaduz Formation (north of Bonaduz at Bot Dagatg and east of Sagogn near Ruina Schiedberg). In addition, and maybe more importantly, the hills west of Bonaduz at the front of the rock avalanche are locally covered with Bonaduz Formation suggesting bulldozing; and the hills south of Crest’Aulta west of Bonaduz at the lateral limit of the rock avalanche (coord. 2746150/1185100) contain slate and sandstone (see Fig. 5) that have been scraped off from the southern valley flank (Penninic nappes).
The mobilization of the substrate manifests itself by the clastic dikes injected into the Flims rock avalanche (Fig. 12A and 24) and the blowout pipes (Fig. 11B and 24). The presence of lacustrine silt within the clastic dikes indicates that this mobilization was deep reaching. Bulldozing at the front of the rock avalanche near Bonaduz raised the ancestral valley floor from 500 to 650 m a.s.l. and produced the Bonaduz Formation with its somewhat chaotic content of fragments of rock avalanche deposits and lacustrine rip up clasts (Figs. 13C and D). Calhoun and Clague (2018) studied this deposit in detail and concluded that it moved as a hyperconcentrated flow. This mass movement carried rock avalanche fragments some 10 km up the Hinterrhein valley to Rodels (Fig. 1) and breached the Tamins rock avalanche deposit at Reichenau entraining fragments of this deposit 10 km down the valley to Domat/Ems and Chur (Fig. 14). Mapping the lithologies of the entrained rock avalanche fragments north and south of Bonaduz it is possible to differentiate between Flims and Tamins derived rock avalanche fragments (Cretaceous, resp. Permian and Triassic lithologies). Thus, it was possible to draw a line of contact passing east of Bonaduz in a N–S direction. Important to note is that the farthest travelled blocks (Rodels in Fig. 1) are derived from the Tamins rock avalanche.
6.4 Shortlived lake
The Flims rock avalanche dammed the Vorderrhein resulting in Lake Ilanz, which reached way upstream the main valley. Considering the mean annual discharges of the Vorderrhein and and its main tributary Glenner, von Poschinger et al. (2006) estimated that this lake was filled within 13 months up to the level of 820 m a.s.l. The ensuing erosion processes by overtopping could have involved three phases (Li et al. 2021): uniform bed erosion leading to a spillway and increasing outflow, backward erosion on the dam with increased outflow and enlargement of the spillway, and, finally, erosion of the outflow channel and rapid increase of discharge. The incipient stage with static overflow in natural dams is strongly influenced by the nature of the lithological composition of the dam (Mei et al., 2021). Fine-grained material is removed first and quickly. In case of the Flims rock avalanche the top of the dam contained large blocks of the carapace, which would likely have slowed down the formation of a spillway. Overflow occurred across a hummocky landscape with closed basins that may have collected and preserved the fine-grained material. Because of erosion along the breach the initial courses of the spillways can only be guessed at but were likely curved around small-scale hummocks. Costa and Schuster (1988) and Fan et al. (2020) report that most historic natural dams breached within less than a year. For Lake Ilanz there is evidence for outbursts and incision in the form of debris flow deposits. These occur on terraces of diminishing altitude downstream of the dam (Calhoun & Clague, 2018). The uppermost debris flow deposits pertain to the Ransun fan (Pfiffner & Wyss, 2023) and consist of gray limestone exclusively. This material can be derived from the Flims rock avalanche solely. At the locality Crestis west of Tamins archeological excavations yielded artefacts from within this fan. They are of late Mesolithic age and indicate fan activity 9500–9000 years ago. The apex of the Ransun fan is at 800 m a.s.l., i.e. about 20 m lower than the lake level which suggests that the Ransun fan is the first outburst of Lake Ilanz. This leaves a time interval of less than 480 years, and possibly only a few years for the existence of this lake. The Ransun fan can be followed over at least 10 km downstream to Domat/Ems. But the flood may correlate to a singularity in the lake deposits of Lake Constance located 80 km farther downstream (Schaller et al., 2022). These authors correlate a double-turbidite with the Flims rock avalanche (the Reichenau breach) and a overlying turbidite with the outburst of Lake Ilanz (Ransun fan).
The younger incision history indicated by the various river beds perched on the flanks of the Ruinaulta gorge and terraces that have not been dated yet.
There are two major tributaries of the Vorderrhein within the Flims rock avalanche upstream of the dam: Ual da Mulin east of Sagogn and Carrerabach between Valendas and Carrera. Both of them incised their delta that they had shed into Lake Ilanz once the dam was breached. This incision must have occurred rapidly: 120 and 100 m in 9000 years, resp. 133 and 111 mm/a and is still going on. According to an eye witness (Pfarrer Candrian) the erosional scarp of Ual da Mulin retreated up to 50 m between 1843 and 1880 (Heim 1883).