Detailed geomorphological maps , aerial and satellite images (Google Earth and Bing databases), and DTM analysis (cell size: 5 m, vertical accuracy: 30 cm; http://idt.regione.veneto.it/app/metacatalog/, last access: 17 October 2019) were used to obtain topographic profiles and to estimate the area of the Masiere di Vedana deposit. The areal distribution of lithologies in the deposit was gauged by observation of boulders in the field and verified by thin-section analysis (Sect. S1 in the Supplement). Six boreholes, up to 3 m deep, were taken in the fine-grained sediments of the Torbe area (Fig. S2a) with a hand auger (Edelman combination type, EjikelkampTM), which allows the extraction of 10 cm wide cylindrical cores. Orientations of bedrock discontinuities, such as bedding, foliation, joints, fractures and faults, were measured in the southern wall of Mt. Peron.

Twelve different boulders located in topographically high positions with respect to the surroundings within the deposits were sampled for dating with cosmogenic 36Cl. For boulders VB13 (VB13a, VB13b) and VB14 (VB5 same boulder as VB14) two samples were taken. Samples were taken to cover the full extent of the deposit, from right near the source area to the distal sector.

For 36Cl sample preparation we used the method of isotope dilution as described by . Total Cl and 36Cl were determined at the ETH AMS facility of the Laboratory for Ion Beam Physics (LIP) with the 6 MV tandem accelerator. The 36Cl/Cl ratios of the samples were normalised to the ETH internal standard K382/4N with a value of 36Cl/Cl=17.36×10-12, which is calibrated against the primary 36Cl standard KNSTD5000 . Full process chemistry blanks (3.4×10-15) were subtracted from measured sample ratios. All 14 rock samples were processed. Only seven were measured successfully due to too high 36S, also in relation to the very low 36Cl concentrations in these samples. All measured data are presented here. Major and trace element concentrations were determined with X-ray fluorescence (XRF) (Sect. S3) and inductively coupled plasma mass spectrometry (ICP-MS) (Sect. S4), respectively. We calculated surface exposure ages with the LIP ETH in-house MATLAB code based on the parameters presented in detail in and references therein. A production rate of 54.0±3.5 36Cl atoms (g Ca)-1 yr-1, which encompasses a muon contribution at the rock surface of 9.6 %, and a value of 760±150 neutrons (g air)-1 yr-1 were found. These values are in excellent agreement with production rates recently published by . Production from all major elements and through low-energy neutron capture in light of the trace elements (Table S4a) was fully considered. Production rates were scaled to the latitude, longitude and altitude of the sites based on . No correction was made for karst weathering of the boulder surfaces see. The extent of karst dissolution on the boulder surfaces varies significantly from boulder to boulder. Implementing a rate of 5 mm ka-1 would change the ages by less than 4 %, which does not affect any of the conclusions drawn here. Stated errors of the exposure ages (Table ) include both analytical uncertainties and those of the production rates . Two different surfaces of boulder 13 were analysed (VB13a, 1.45±0.12; VB13b, 1.45±0.12 ka); the weighted mean of 1.45±0.08 ka is used for further discussion.

The Mt. Peron scarp is 700 m wide and 600 m high, S-to-SW facing, and partially circular. No secondary scarps are present. Numerous faults and fractures are visible on the wall (Fig. b) and are grouped into five main discontinuity sets (Fig. c). These comprise (1) bedding, (2) WSW–ENE-directed frontal thrust planes, (3) NW-verging backthrust-related planes, (4) NW–SE-aligned local conjugate fracture plane sets, and (5) persistent N–S-oriented planes interpreted as reactivated Jurassic faults . Bedding is nearly vertical, and its orientation ranges from 146/78 to 170/80 (dip direction/dip angle). The Belluno thrust, average orientation 337/64 , crops out at the base of the steep wall, whilst other Belluno thrust planes (295/53 to 340/67) were measured higher up along the wall (Fig. b). The NW-verging planes related to the backthrust are 111/15 to 175/54 and steepen to 80∘ at higher elevations along the wall. The NW–SE-aligned fractures are 209/16 to 245/32 with an associated conjugate set, from 20/44 to 62/30, and nearly vertical fractures with a dip direction of 219 to 255. The N–S-striking fracture planes dip both to the east (75/40 to 83/75) and to the west (240/71 to 299/18). Today a myriad of large and small individual rock prisms bounded by these discontinuities are present in the upper part of the release area.

The deposit covers an area of ∼9 km2 from the base of Mt. Peron southwards to Roe Basse (∼5 km) and westwards to Mis River (∼3 km; Fig. ). By means of open sections, the thickness of the deposit is estimated ∼10 m in the proximal area (near boulder VB2, Fig. ), ∼15 m in the central sector near Torbe, >30 m near the boundary between the Vedana and Masiere sectors (Ponte Mas section, Fig. ), ∼5 m in the Masiere central sector, and ∼15 m in the southern (distal) sector (Suppiei section, Fig. ). Using a mean thickness of 20 m, a rough estimation of the total debris volume of ∼170 Mm3 is obtained. Such a volume corresponds to a released rock mass of about 130 Mm3 bulking coefficient of 25 %; see. With a vertical drop (H) of about 1150 m (from the top of Mt. Peron: 1486 m a.s.l., to Roe Basse area: ∼340 m a.s.l.) and a travel distance (L) of about 5900 m (Fig. ), we calculate an H/L ratio of ∼0.2 (Fahrböschung angle of 11∘). Based on spatial pattern, boulder lithology and morphological character, we distinguish five sectors of the deposit: Peron, Vedana, Torbe, Masiere and Roe (Roe Alte and Roe Basse).

The Peron sector includes the talus apron deposits at the foot of Mt. Peron, the rock avalanche deposits on the east side of the river and the terrace of the town of Peron (at about 380 m a.s.l.). Boulders at the foot of the slope range up to 20 m in diameter (Fig. a). Three boulders in the Peron sector were dated with cosmogenic 36Cl (Fig. ; Table ): VB3a (Rosso Ammonitico; 1.83±0.28 ka), VB3c (Fonzaso Fm.; 3.62±0.41 ka) and VB14a (Calcari Grigi Group; 2.39±0.39 ka). Based on the trend of all obtained ages, the age of VB3c is interpreted as an outlier, its age possibly reflecting the presence of inherited 36Cl due to pre-exposure. The town of Peron lies on a terrace made of rounded gravel layers with rare sand lenses that are interfingered with talus deposits .

In the Vedana sector, the rock avalanche deposit displays an irregular forested topography with relief on the order of tens of metres (Fig. b). Huge blocks hundreds of cubic metres in size, mostly made of Calcari Grigi, dominate the carapace. This covers the main body of the deposit made of shattered rocks, which is comprised of very angular clasts (up to tens of centimetres in diameter) in a matrix of silty sand. The VB2 boulder (Calcari Grigi) gave an age of 1.49±0.26 ka; it lies on top of an ∼10 m thick sequence of rock avalanche deposits. In the Ponte Mas quarry (Figs.  and g), an open section showed glacial till (up to 3 m thick) incorporated into the base of the rock avalanche deposit; its original bedding is completely obliterated. This sediment is >30 m thick and is composed of sub-rounded clasts (up to 20 cm in length), some of them striated, supported by a silty clay matrix. Clasts are sedimentary and volcanic, reflecting the catchment of the Cordevole palaeoglacier see. Several ENE–WSW-trending incisions cut through the Vedana and Torbe sectors (main ones highlighted in Fig. ). Irregular patches of sandy-silty and fine gravel sediments are found in the Vedana low-lying areas between the blocky reliefs.

The Torbe sector encompasses the distal northern lobe of the rock avalanche, characterised by 10 to 20 m high isolated hills and hummocks (Fig. d) that emerge from a flat topography. They are roughly aligned ENE–WSW, are circular at the base and have slope angles of 35–40∘. They are made of very angular Calcari Grigi boulders and clasts, with many jigsaw puzzle structures in a sandy, gravelly matrix (Fig. e). Such morphological structures are “toma” . They are found in association with some large rockslides and are mainly made of landslide material, in many cases showing a gradation from very comminuted fragments in the outer part to less fractured material at the core see. Six cores taken in the flat area between the hills (see Fig. 3 and Sect. S2) show up to 2 m of silty sand above the rock avalanche becoming more fine upwards. Torbe is crossed by the largest incision of the whole Masiere di Vedana (Fig. ), ENE–WSW trending, ∼50 m wide and up to 20 m deep with respect to the mean topographic surface. Cenozoic lithologies crop out at the base of this incision, with the rock avalanche deposit being ∼15 m thick. A shallower incision, a few metres deep, is located at the base of the Piz Vedana slope, still conveying a small amount of water coming from the Vedana Lake.

The Masiere sector is strikingly different from Vedana and Torbe. It is a bleak, vegetation-free, desert-like sea of limestone blocks, mainly of dolomitised Vajont Limestone (Fig. c). Boulders up to 3 m in diameter and abundant angular and sub-angular clasts, with almost no matrix, are present in the surficial part. In the southern part of Masiere, numerous 2–3 m high and up to 150 m long ridges, aligned roughly E–W, are present (Fig. ). The contact of the rock avalanche with the underlying Pleistocene conglomerate crops out near the southern boundary of the Masiere (white asterisks in Fig. ), where the deposit is only ∼5 m thick. Next to the northern boundary of Masiere, the Cordevole River flows upon the Cenozoic rocks of the Castel Cuch ridge, covered elsewhere by the rock avalanche. An ENE–WSW-trending shallow incision (Fig. ) with associated well-sorted, medium-to coarse-grained sandy deposits is present, almost parallel to Castel Cuch. A roughly N–S trending incision, a few metres deep, ∼10 m wide and about 500 m long, is located just to the west of the Cordevole River. The town of Mas is built upon a flat terrace, mainly made of sorted rounded gravels with a sandy matrix, that bounds the Masiere to the east.

Roe Alte and Roe Basse comprise the distal sector of the Masiere di Vedana to the south (Fig. ), where there are angular clasts up to 20 cm in diameter and very few boulders immersed in a sandy matrix are scattered in the meadows. Boulders belong to the Fonzaso Formation, Rosso Ammonitico and Maiolica, with very few of them made of Vajont Limestone. Roe Basse is made of silty/sandy alluvial sediments deposited by the Gresal and other minor streams coming from the east, mantling the rock avalanche deposit on the northwestern side. The Suppiei section, 100 m long and 25 m high, on the left flank of the Cordevole River (Figs.  and h; Sect. 5), shows the Bolago Marl unconformably overlain by 0.5–2 m of glacial till. This is in turn covered, with a sharp and undulated contact, by up to 20 m of rock avalanche debris decimetric in size, with boulders (∼1 m diameter) on top. On the Roe Alte rocky upland, the rock avalanche is at most 2 m thick, with rare boulders (1–2 m diameter). Two boulders made of Vajont Limestone (Fig. ) have been dated with 36Cl (VB12, 2.35±0.21 ka; VB13, 1.45±0.08 ka; Table ). South of the town of Mas, the Cordevole River flows into narrow meanders entrenched ∼20 m into rock avalanche deposits, alluvial material, glacial sediments and bedrock. The terrace of the Vignole village is almost totally made of rock avalanche debris, despite being remarkably flat.

36Cl surface exposure ages from boulders all across the deposit range from 1.45±0.08 to 2.39±0.38 ka (Table ). All ages show a good overlap within uncertainties. A single sample (VB3c) gave a result markedly different from the others: 3.62±0.41 ka. Although this age may point to pre-exposure of the sampled boulder surface see, as for example seen at Lavini di Marco , the possibility exists that this boulder is part of a partially buried older deposit located right at the foot of Mt. Peron. The poorly developed karst dissolution features (0.5–1 cm deep karren) on the tops of many boulders suggest as well that the deposit is relatively young. The average of 36Cl ages, excluding VB3c as an outlier, is 1.90±0.45 ka. The uncertainty of the mean is based on the cumulative probability of uncertainties for all samples based on a Gaussian probability distribution (one sigma level). Such a value indicates that the rock avalanche, considering the error range, occurred during historical times, between 340 BCE and 560 CE. These results are in stark contrast to previous reconstructions, which pointed to a late glacial age . The dates 1113, 1114, and 1117 CE proposed for the main landslide event suggested by some authors may be associated with the Verona earthquake at 1117 CE. That event was clearly felt in the Belluno area, inducing numerous landslides , but its age is not consistent with the cosmogenic dates on the Masiere di Vedana. To search for independent constraints for the age of the main landslide event, detailed research in numerous archives and chronicles was undertaken. This area during Roman time was largely and uniformly inhabited by “incolae” for agricultural aims, being the area located next to the Claudia Augusta Altinate road connecting Feltre with Belluno (Fig. ) . The presence of a Roman bridge crossing the Cordevole River north of Mt. Peron indicates there was a connection to the main Claudia Augusta Altinate road. While numerous Neolithic and Roman archeological sites are reported around the Masiere di Vedana , no Roman or pre-Roman archaeological remains have been found within the rock avalanche deposits (Fig. ). If the Masiere di Vedana deposit was settled after Roman times, previous settlements eventually located in the area would have been buried by the event. The oldest record for the post-event human presence is a hospice built in the 12th century CE 1155 CE; on the fluvial terrace near the village of S. Gottardo (Fig. ). Therefore, historical data indicate a time frame between late Roman times and the early Middle Ages.

The age uncertainties do not allow us to directly determine if the Masiere di Vedana deposit was due to a single failure or multiple events. The distribution of the available dates has no spatial pattern across the deposit, no physical boundaries occur and no buried soil layers have been found within the deposits. Moreover, the volume and H/L ratio of the landslide (see Sect. 5.2 for further discussion), together with the morphology of the scarp and the absence of secondary scarps, indicate a single huge catastrophic event. Therefore, a single rock avalanche occurred in historical time, contradicting previous interpretations e.g.. Such a young age implies that glacial unloading was not responsible for the destabilisation of the southern side of Mt. Peron and makes a re-evaluation of the involved driving factors and potential triggers necessary.

A schematic reconstruction of the reach of the Cordevole Valley involved in the rock avalanche can be depicted before, during and after the event. Before the rock avalanche, the Cordevole River flowed through a gentle rolling landscape along the foot of Piz Vedana and through a breach cutting the Castel Cuch bedrock ridge just north of the village of Mas (Figs.  and a). Topographic highs, like Castel Cuch and Roe Alte, were at that time mantled with glacial sediments attributable to the last glaciation.

The rock avalanche involved the detachment of about 130 Mm3 from the southern face of Mt. Peron. Initial movement was sliding along the NW-verging backthrust-related planes (Fig. ). En bloc movement may have been only briefly sustained as the pervasive network of fractures favoured a massive collapse. The rock mass immediately evolved into a rock avalanche whose volume increased by fragmentation up to 170 Mm3 and spread out onto the flat plain below. The H/L ratio of ∼0.2 (apparent friction angle of 11∘) and comparison with empirical and modelling plots of H/L vs. volume e.g. mark the Mt. Peron event as extremely mobile. As a basis for comparison, the Fernpass rock avalanche has a H/L of 0.9, a volume of 1 km3 and a significantly longer runout distance of 15.5 km . It may be possible to glean information about the failure style from the distribution of boulder lithologies, which follows the stratigraphic order of the bedrock exposed in the source area. This has also been noted at the Tschirgant rock avalanche deposits in Austria and the Frank slide in Canada . In the Mt. Peron bedrock, the lithologic sequence from west to east is Calcari Grigi Group, Vajont Limestone, Fonzaso Fm., Rosso Ammonitico and Maiolica. This pattern is mirrored in the deposits (Figs.  and ): Vedana and Torbe are dominated by Calcari Grigi Group, Masiere by Vajont Limestone, and Roe by Fonzaso Fm., Rosso Ammonitico and Maiolica. Experiments and modelling suggest that this kind of zonation is likely to occur when the sliding mass propagates as a flexible sheet, with laminar flow .

Several landforms within the Masiere di Vedana provide further clues on the processes of propagation and emplacement. The tomas in the Torbe sector suggest differential velocities in the moving mass propagating on a water-saturated substrate . Tomas, with likely similar origin, are present in the distal deposits at Fernpass in Tyrol and at Flims . Recently, proposed for the toma hills at Fernpass an alternative origin from internal erosion by suffusion. However, at Masiere di Vedana the fluvial deposition above the rock avalanche suggests that the suffusion process can be ruled out. In contrast to the increased mobility seen in the Torbe sector, in the central Masiere area, landforms indicative of stalling are present (Fig. ). The stacked subparallel transverse ridges, much like those noted at Tschirgant , with slight overrunning of the ridges in front by those behind, indicate slowing down of the moving mass due to longitudinal compression . Outcrop relationships (Fig. ) suggest that the Pleistocene conglomerate inhibited the rock mass flow, in combination with the slight uphill gradient. The ridges at Masiere were previously interpreted as neotectonic lineaments by .

After the event, the Cordevole River changed its channel several times. The rock avalanche blocked the river, creating accommodation space to the north, where possibly a temporary lake formed. The river was then forced to flow westward across the deposit as indicated by the palaeochannels in the Vedana and Torbe sectors (black arrows, Fig. ), taking different paths at different times. Low-lying areas were progressively filled, as shown by the sequence recorded in core TB1 becoming more fine upwards (Fig. S2a). The Torbe, Vedana and Peron terraces are flat surfaces at ∼380 m a.s.l. (Fig. b). Afterwards, a further sedimentation was hindered by the trenching in Torbe. The Cordevole River finally breached the landslide deposit to the southeast, through the Castel Cuch ridge made of Cenozoic rocks (Fig. ). The river initially flowed from Mas (about 375 m a.s.l.; green line in Fig. a) to the southern flank of Castel Cuch as suggested by the still recognisable palaeochannel (Fig. ) filled with well-sorted, medium- to coarse-grained sand . Subsequently, the Cordevole moved to the eastern side of the Pleistocene conglomerate cliff (Fig. ). The final diversion of the river formed the Peron, Mas and Vignole terraces and currently flows some metres below with upstream migration of the knickpoint (Fig. ).

In the Cordevole and Piave valleys many landslides have been recorded (Fig. ) and have caused a great deal of damage and casualties . Rock avalanches such as Masiere di Vedana are difficult to predict and may be very destructive due to their huge volume and extreme runout . In light of the results we obtained, the search for the drivers of the Masiere di Vedana rock avalanche is both timely and imperative. Even if what determines the moment of failure may be difficult to pinpoint, increased pore pressure and seismic ground shaking are the primary candidates in such cases . However, rock avalanches may start without a definite external trigger, the progressive accumulation of rock fatigue being enough to overcome the resisting forces of the rock mass. This is the case, for example, of the Tsatichhu landslide (10 September 2003) in Bhutan and the several Randa events (total of 30 Mm3) in 1991 in Switzerland .

The Belluno area has high mean annual rainfall (1643 mm in the time interval 1994–2018 at https://www.arpa.veneto.it/dati-ambientali/open-data, last access: September 2019) and is prone to extreme rainfall events e.g. >300 mm of rain at Sospirolo during a single event: 27 October–1 November 2018;. Moreover, at the time of the Masiere di Vedana rock avalanche, between late Roman times and the early Middle Ages, the southeastern Alps and northeastern Italy were affected by various periods of climate degradation during which several extreme meteorological events occurred . One of these extreme events had an impact all over Europe between 50 and 250 CE, with marked intensity and widespread flooding recognisable in the stratigraphic records . This period of severe rainfall could possibly have been the trigger for the Masiere di Vedana rock avalanche or at least may have acted as a driving and destabilising factor. Likewise, the 1987 Val Pola rock avalanche in the central Alps was triggered by a period of exceptional rainfall .

The Veneto region is prone to earthquake activity, and the study area is categorised as level 2 seismic hazard (“possible strong earthquakes” in Ordinanza del PCM n. 3519/2006), as the historical record testifies up to ∼Mw=6.5;. Continuous instrumental monitoring of the Belluno area dates back only to 1977 . Preceding catalogued major seismic events in the region are based on either historical chronicles, dated building damage and/or observed rockfalls . Within a radius of 30 km from Mt. Peron, eight earthquakes with Mw greater than 5.0, and one exceeding 6.0, were recorded (Fig. ). The strongest (Mw=6.3) occurred just 20 km to the east of Mt. Peron on 29 June 1873 and references therein. Severe damage to Belluno city was reported during the Asolo (25 February 1695 CE; Mw=6.4), Friuli (6 May 1976 CE; Mw=6.5) and Verona (3 January 1117 CE; Mw=6.5) earthquakes whose epicentres were located, respectively, 60, 65 and 140 km away . As for the time frame suggested by our chronology, historical records report an important seismic event in July 365 CE that resulted in damage to the city of Belluno . These data suggest that the Belluno area is sensitive to seismic shakings originating even hundreds of kilometres away. discuss evidence that active tectonics plays a key role in the reported intensification of slope instability registered in this area during the last 1500 years. The most important effect of the frequent seismic activity is the progressive increase in the rock fatigue, with the formation of failure surfaces and the removal of rock bridges and roughness on discontinuity planes . Earthquakes have been suggested as triggers for several Alpine rock avalanches e.g. and are known to have caused several rockfalls in the Belluno area e.g.. Moreover, they are considered to be a possible trigger for failure of some pillars located on the southern side of Mt. Peron .

The predisposition of the southwest Mt. Peron face to failure is attributable to the structural setting and the regional-scale framework. The entire Belluno Dolomites have experienced a long deformation history since the Miocene , related to regional-scale stress connected to the anticlockwise rotation of the Adria plate, indented with the Alpine orogeny . The thrusts and backthrusts (WSW–ENE oriented) that encase Mt. Peron relate to this phase of activity, also creating the two conjugate fracture sets (NW–SE oriented) and inducing the reactivation of the Jurassic faults (N–S oriented) and the overturning of the beds. The area between the Belluno thrust to the south, the Val Carpenada – Val di Vido – Val Madonuta backthrust to the north, the Caorame Valley to the west and the Piave Valley to the east (Fig. ) is structurally homogeneous and characterised by the same fracture sets identified at Mt. Peron . In limestones, where these intersect, rock dissolution and caves can form , further weakening the rock . The Mt. Peron southern wall is known locally as the “weeping rock” due to the numerous caves and karst springs along the steep rock face (Fig. b).

Our dating of the Masiere di Vedana landslide to late Roman times casts considerable doubt on the previous late glacial chronological assessment e.g.. Consequently, stress release following downwasting of the Last Glacial Maximum (LGM) Piave glacier, which had been previously considered the main driver of the rock avalanche, actually played no direct role in the process. Climatic and tectonic factors were much more important. In addition, the classification of the deposits as late glacial led to underplaying of the possible hazard at Mt. Peron. The situation at Mt. Peron, with steeply dipping-to-overturned bedding in a limestone massif crisscrossed by numerous faults and riddled with karst fissures, is presently the same that produced the massive failure of the Masiere di Vedana. Several-hundred-metre-high rock prisms along the top of the headscarp are partially detached along the discussed fracture systems and loom perilously over the inhabited valley below . Our results suggest the need for a reconsideration of the hazard related to not only the southern Mt. Peron face, as the rock slopes still present evident structural weakness, but also the whole area lying between the Belluno thrust and its backthrusts, as intense rainfall and earthquakes can occur at any time.