Modes of sector collapse of volcanic cones: Insights from analogue experiments

Authors


Abstract

[1] Many volcanic edifices are subject to sector collapse. Analogue models are used to better understand specific modes of collapse. Four sets of experiments have been performed with a cone (volcano analogue) of dry sand (87% in volume) and flour (13%) simulating collapse through the following: set A, basal failure due to the horizontal sliding of a basal plate; set B, unbuttressing due to the horizontal movement of a lateral wall; set C, summit growth due to the addition of sand and flour on the top of the cone; set D, injection due to the intrusion of silicone (magma analogue). Sets A and B show common results, highlighting the influence of topography on the dip and strike of the collapse surfaces. In set C, the location and direction of collapse are controlled by the cone shape and occurrence of preexisting collapses. In set D, if the silicone is injected eccentrically, it induces collapse on the nearest slope; this process leads to repeated collapse on the side of a preexisting scarp, possibly triggering a feedback mechanism between magmatic activity and topography. In general, sets A and B induce deeper and wider collapse, whereas shallower and narrower collapse is observed in set C; set D shows intermediate geometries. Comparison with natural examples shows an overall similar distribution in the geometry of the sector collapses and their known triggering mechanisms. Similarities and discrepancies between each experimental set and corresponding cases in nature are discussed.

1. Introduction

[2] Volcanic edifices are the result of the repeated emplacement, usually within time spans of <105 years, of magmatic products in a limited area. As a consequence of this relatively rapid construction, any volcanic edifice with significant height (on the order of 103 m) can become unable to support its own load. This lack of support may result in the collapse of a sector of the volcano. A sector collapse is here defined, in its broadest meaning, as a gravity-driven movement of a portion of a volcano, independently from its size, origin and type. Sector collapse is observed at many volcanoes, independently of their composition (mafic and silicic), shape (stratovolcanoes, calderas and shield volcanoes) and geodynamic setting (divergent and convergent margins, hot spots).

[3] Sector collapses are characterized by very different velocities of the mobilized mass, ranging from creep-like movements (velocity of 10−9–10−10 m s−1 [Froger et al., 2001]), to catastrophic fast moving landslides (velocity ∼102 m s−1 [Voight et al., 1981]). Collapse may occur suddenly [van Wyk de Vries et al., 2000; Cervelli et al., 2002] or consist of accelerated movements within prolonged periods of creeping of the volcano flank [Neri et al., 2004].

[4] Mobilized volumes vary enormously, even at the same volcano (Figure 1); from a few cubic meters [Calvari and Pinkerton, 2002] to huge flank movements (1012 m3), involving the sedimentary substratum [Neri et al., 2004]. Significant collapses commonly mobilize volumes of 108–1011 m3 [Carracedo et al., 1999; Day et al., 1999a; Hall et al., 1999; Masson et al., 2002]. The larger the volume of the mobilized mass, the lower is the frequency of collapse [McGuire, 1996].

Figure 1.

Different examples of sector collapses of volcanic edifices: (a) Sciara del Fuoco, Stromboli Island, Italy (based on data of Tibaldi [2001]); (b) collapsing blocks at Etna, Italy (based on data of Neri et al. [2004]); (c) failures observed during the 1980 eruption at Mount S. Helens (based on data of Donnadieu and Merle [2001]); and (d) sector collapse of Tungurahua volcano, Ecuador (based on data of Hall et al. [1999]).

[5] The direction of collapse can be controlled by the morphology of the volcano, which is partly the result of its structural setting [Tibaldi, 1995]. Also, it can be the result of levels of weakness within the edifice, such as unwelded and altered pyroclastic rocks, or the contact with older volcanoes [Vallance et al., 1995].

[6] Most sector collapses extend to a distance roughly equal to the diameter of the volcanic edifice; nevertheless, in some cases, mud and debris flows produced by sector collapse can travel for significant (up to 105 m) distances [Reid et al., 2001]. Rapid falls of volcanic masses along coastal areas may also trigger tsunamis [Ando, 1979; Moore and Moore, 1984].

[7] Because of their sudden occurrence, extent and association with tsunamis, sector collapses can pose a significant risk to populations living nearby. About 20,000 people are estimated to have been killed by historic volcano flank collapses [Siebert et al., 1987].

[8] Several factors have been proposed to trigger the collapse of volcanoes. These may act independently, if the collapse is the product of a predominant factor, or, more commonly, simultaneously, when the collapse results from some hybrid mechanism.

[9] Magma emplacement is possibly the most common triggering factor. Many collapses are the consequence of various processes related to magmatic activity: dike emplacement [Dietrich, 1988; Delaney et al., 1998; Elsworth and Day, 1999; Tibaldi, 2001], volcanic activity [Capra et al., 2002], caldera formation [Marti et al., 1997], the intrusion of viscous magma [Voight et al., 1981; Belousov et al., 1999; Richards and Villeneuve, 2001]. Intrusions can in turn induce mechanical or thermal straining of the rocks, generating excess pore pressures and reducing the sliding resistance on faults within the edifice [Voight and Elsworth, 1997]. Sector collapse can also initiate volcanic activity [Alvarado and Soto, 2002; Acocella et al., 2003], generating a feedback mechanism between gravity failures and magmatic activity [McGuire et al., 1990].

[10] Another common trigger is the failure of a volcano related to fault activity [Hall et al., 1999; Ventura et al., 1999; van Wyk de Vries et al., 2003]. Earthquakes may also trigger sector collapse [Ando, 1979; Acocella et al., 2003].

[11] A weak basement or layer on which the volcano has been built (commonly clays) has also been proposed as a trigger of volcano collapse and spreading, in particular for larger volcanoes [Borgia et al., 1992; van Wyk de Vries and Francis, 1997; Morgan et al., 2000; van Wyk de Vries et al., 2001; Borgia and van Wyk de Vries, 2003].

[12] Another supposed trigger is weakening of the volcanic edifice by weathering [Day et al., 1999b; Hurlimann et al., 1999; Kerle and van Wyk de Vries, 2001], possibly coupled with climatic processes, or by hydrothermal activity with hyperacidic brines [Lopez and Williams, 1993; Kempter and Rowe, 2000; van Wyk de Vries et al., 2000; Reid et al., 2001].

[13] Several attempts have been made to model sector collapse. Numerical models investigated how the load of a volcano may be controlled by, or exerts control on, magmatic activity. In the first case, the stability of a volcano slope is evaluated under an increase of internal magmatic pressure [Dietrich, 1988; Russo et al., 1997], excess pore pressures due to intrusion [Voight and Elsworth, 1997; Elsworth and Day, 1999] and during cryptodome emplacement [Donnadieu et al., 2001]. In the second case, the effect of volcano loading on dike propagation has been studied [Muller et al., 2001]. Numerical models show that even a magmatically inactive volcano with a significant mass, if resting over a weak substratum, is capable of inducing slope failure [Borgia, 1994; van Wyk de Vries and Matela, 1998]. Moreover, numerical models have been applied to reveal the importance of topography on the stability of volcanic edifices [Reid et al., 2000] or to simulate the emplacement of landslides resulting from caldera collapse [Hurlimann et al., 2000].

[14] Analogue models have also been widely used to simulate sector collapses of volcanoes. Experiments of volcanic spreading were performed to predict whether or not spreading takes place in a volcano as a function of its height and the brittle-ductile ratio of the substratum [Merle and Borgia, 1996], in extensional [van Wyk de Vries and Merle, 1996] and strike-slip [van Wyk de Vries et al., 2003] settings. The role of a ductile horizon at the base of a hydrothermally altered volcanic pile has also been investigated [Merle and Lenat, 2003].

[15] Attempts have been made to study the collapse of a cone as a consequence of a viscous intrusion. In particular, the modeling of the cryptodome intrusion at Mount St. Helens was used to interpret the deformation observed prior to the 18 May 1980 collapse and eruption and to infer the possible path of rise of the intrusion [Donnadieu and Merle, 1998, 2001].

[16] Analogue models have been recently used to study the instability of a volcanic cone due to a basal active fault. A vertical fault with a dip-slip movement at the base of a cone may produce distinct types of flank collapse, recognizable even at dormant volcanoes [Vidal and Merle, 2000; Merle et al., 2001]. Similarly, the activity of strike-slip [van Wyk de Vries and Merle, 1998; Lagmay et al., 2000] and oblique [Wooller et al., 2003] faults on volcanic cones permits prediction of the location of the associated flank failure. Other recent experiments examine the failures of domes and stratocones after caldera collapse [Belousov et al., 2003].

[17] This work aims (1) to define the modes of collapse of conical volcanic edifices (stratovolcanoes and large spatter cones), considering specific triggering mechanisms that have not been previously studied, and (2) to provide a preliminary experimental database on the geometry of sector collapse, defining and comparing the types of instability at volcanic cones. For these purposes, four sets of analogue experiments have been carried out, which are associated with specific conditions triggering collapse. In these experiments, the collapse of a cone was studied simulating (1) the basal failure of a volcano, (2) the unbuttressing of a volcano flank, (3) the growth of the volcano summit, and (4) the inflation of a volcano due to magma injection.

[18] The results show how different types of sector collapse, varying in location, depth, length, and width, develop. The experiments are compared with cases occurring in nature, highlighting similarities and discrepancies.

2. Experimental Apparatus

[19] This work aims at investigating those mechanisms of sector collapse that have not been studied, or have been previously studied in part. Four sets of experiments have been performed, consisting of: A) the basal failure of the central part of a cone, simulating the activity of an extensional structure (normal fault and extensional fracture) at the base of a volcano; this set investigates how tectonic failures affect a volcanic edifice, developing collapses; B) the unbuttressing of a cone, simulating collapses resulting from intense and focused erosion on the volcano flank; C) the addition of mass on the summit of a cone, to simulate volcano growth; D) the injection of a viscous indenter within a cone, to simulate flank instabilities during volcano intrusion.

[20] The four sets of experiments use different apparatuses to deform a cone, which represents the volcano analogue. The cone has a height varying between 5 and 15 cm and a slope angle of ∼35°, similar to the one (∼30°) associated with most collapses in nature [Voight and Elsworth, 1997].

[21] In set A the failure of a cone occurs through a mobile basal plate, sliding outward and inducing a linear velocity discontinuity (VD) with regard to the basal fixed plate; the VD propagates the deformation toward the upper part of the cone, simulating a deep-seated failure (Figure 2a). This setup, where the VD slides horizontally, differs from the set up used in previous experiments, where the VD slides vertically [Vidal and Merle, 2000; Merle et al., 2001].

Figure 2.

Schematic section views of the experimental apparatus used in the four sets of experiments. D is distance of velocity discontinuity (VD) from center of cone, h is height of cone, z is vertical distance between the silicone and the cone summit (or slope surface in subset D2), T is vertical distance between the silicone and the cone surface (or slope surface in D2), and L is lateral distance of the silicone from the center of the cone (or the onset of slope in D2). Subset D2 shows the section view of a slope of sand and flour.

[22] In set B the unbuttressing is achieved through the outward sliding of a vertical wall at the side of the cone; this induces a VD along the base of the wall, which propagates the deformation upward, simulating a deep-seated failure on the cone (Figure 2b).

[23] Considering the mean life of a volcano (∼105 years) undergoing ordinary tectonic slip rates of ∼1 cm yr−1, the resulting total deformation is ∼103 m. This is consistent with the total amount of imposed deformation in set A and B, which is ≤1 cm, corresponding to 102–103 m in nature (length scaling in next section).

[24] In set C the summit growth of the cone occurs through the addition of loose material (volumes of dry sand and flour 87% and 13%, respectively; see next section) on its summit; this induces the steepening and collapse of the upper slope (Figure 2c).

[25] In set D the deformation of the cone occurs through the injection of silicone, which triggers collapses (Figure 2d). This set aims at understanding a general mechanism, with a setup similar to that applied to simulate the 1980 Mount St. Helens collapse and eruption [Donnadieu and Merle, 1998, 2001].

3. Scaling and Materials

[26] Models have to be geometrically, kinematically, and dynamically scaled, following the principles discussed by Ramberg [1981]. We chose a length ratio between model and nature L* = 2.5 × 10−5 (1 cm in the model corresponds to 400 m in nature; Tables 1a and 1b). The densities of natural volcanic rocks (2000–2700 kg m−3) and of the commercially available experimental materials (900–1800 kg m−3) impose a density ratio ρ* ∼ 0.5. Since the models were run at 1 g, the gravity ratio g* = 1. These ratios imply that the stress ratio between model and nature is ρ*g*L* ∼ 1.2 × 10−5 (Table 1a).

Table 1a. Scaling Procedure Adopted in the Experiments for the Brittle Volcanic Edifice
 Value
Length ratioL* = Lmod/Lnat = 2.5 × 10−5
Gravity ratiog* = 1
Density ratioρ* = 0.5
Stress ratioσ* = ρ*g*L* = 1.2 × 10−5
Table 1b. Scaling Procedure Adopted in the Experiments for the Ductile Magma
Natural Viscosity, Pa sViscosity Ratio μmod = 104 Pa sStrain Rate Ratio e* = σ*/μ*Time Ratio t* = 1/e*Velocity Ratio v* = e*L*Natural Velocity vnat = m h−1 vmod = 12 cm h−1
10411.2 × 10−58.3 × 1043 × 10−104 × 108
10710−31.2 × 10−28.3 × 1013 × 10−74 × 105
101110−71.2 × 1028.3 × 10−33 × 10−34 × 101
101510−111.2 × 1068.3 × 10−73 × 1014 × 10−2
101810−141.2 × 1098.3 × 10−103 × 1044 × 10−5

[27] We assumed a Mohr-Coulomb failure criterion for the rocks of the volcanic edifice, with an angle of internal friction ϕ = 35° and a mean cohesion c = 107 Pa. Cohesion, having the dimensions of a stress, must be scaled at ∼1.2 × 10−5 in the experiments; this requires the use of material with c ∼ 120 Pa to simulate the volcanic edifice. For this purpose, the cone is made up of a mixture of well sorted, round grain, dry quartzose sand (87% in volume) and flour (13% in volume). These proportions give the mixture an angle of internal friction ϕ ∼ 35° and cohesion c = 100–200 Pa. Similar mixtures have previously been used to model volcanic cones with similar length ratios L* [Donnadieu and Merle, 1998; van Wyk de Vries and Merle, 1998; Vidal and Merle, 2000; Merle et al., 2001; Belousov et al., 2003; Merle and Lenat, 2003]. The mixture was released from a funnel at fixed height, building the cone with a natural repose angle. No stratifications between different materials were used to simulate internal variations of the volcanic edifice.

[28] Newtonian silicone, with a viscosity of ∼104 Pa s, was used as a magma analogue in set D to intrude into the cone. The vertical injection velocity of silicone within the cone was 12 cm h−1. The use of silicone to simulate magma has to take the viscosities, the related strain rates and timescales into account [Merle and Vendeville, 1995]. Magma viscosities vary widely, even within a limited range of temperatures [Talbot, 1999]. The possible viscosity ratios between model and nature are listed in Table 1b; for a given viscosity and stress ratio, the related strain rate, timescale, and velocity ratios are reported. Table 1b therefore shows the possible ratios between model and nature to be simulated by imposing the silicone viscosity, the injection velocity and the stress ratio. In order to understand general mechanisms of deformation, valid for a wide range of natural cases, the experiments did not aim to simulate specific natural cases. Also, the extent to which the experimental results are applicable is limited by current knowledge of natural parameters such as viscosities, timescales and strain rates. This knowledge permits the definition of precise ratios between model and nature and application of the experiments to specific cases.

[29] The experiments simulate volcanic cones with a slope of ∼35°. Therefore their results should be applied to stratovolcanoes and large spatter cones. The processes in set A and B share similarities with failure mechanisms observed at shield volcanoes; even though the overall shape of shield volcanoes is different from the simulated one, the experiments may give qualitative insights on these mechanisms. The experiments of set C may also be applied to submarine or partly submerged arc volcanoes, with unconsolidated material along their slopes. Finally, it should be considered that the experiments usually simulate a history of events, while studies of natural cases mostly concern single events.

4. Experimental Results

4.1. Set A: Basal Failure

[30] This set consists of 11 experiments, in which the following parameters are varied (Table 2): the eccentricity E (E = Lmin/Lmax, where Lmin is the minor axis and Lmax is the major axis) of the base of the cone, varying between 1 (circular cones) and 0.8 (elongated cones); the distance D of the VD from the center of the base of the cone; the height h of the cone and, for elongated cones, the angle α between the directions of Lmax and the VD. Three representative experiments are described.

Table 2. Experiments and Related Parameters Considered in Set A (Basal Failure)a
ExperimentED, cmα, degh, cm
  • a

    E is eccentricity of the base of the cone, D is distance of velocity of discontinuity (VD) from center of cone (positive on the left side of the cone, negative on the right side; see Figure 2), α is angle between maximum elongation of cone and trend of VD, and h is height of cone.

COL 1717.508
COL 180.857.509
COL 190.87.5308
COL 200.865.5308.5
COL 210.847.5458
COL 220.83−7.508.5
COL 231−608
COL 261−508
COL 271−406
COL 390.837.5158
COL 400.857.5608

[31] Experiment COL 17 has E = 1, D = 7.5 cm, h = 8 cm and α = 0°. The map view of the undeformed model, at t = 0′ (minutes), is shown in Figure 3a. The outward sliding of the mobile plate (velocity = 2.7 cm h−1) induces a failure surface propagating from the VD.

Figure 3.

Set A, evolution of experiment COL 17. (a) Undeformed stage, map view. (b) Experiment at t = 2′. A deep failure (X) develops as a result of the lateral slide of the basal VD. (c) Experiment at t = 7′. A shallower failure develops on the cone summit.

[32] At t = 2′ the failure (X in Figure 3b) reaches the surface of the cone, with an arcuate shape in map view; this curvature corresponds to the intersection between a plane with a dip of ∼50° and the topography. A second failure (Y in Figure 3b), with opposite dip, is present. The formation of this “antithetic” failure is due to the fact that the VD generates conjugate sets of failures. The effect of the antithetic failure on the structure of the cone is negligible and its presence will not be considered further.

[33] The kinematics along failure X varies from purely dip-slip (near the cone summit) to left-lateral and right-lateral (near the cone base; Figure 3b). The capability of the failure to reach the summit of the cone depends on its dip, the distance of the original location of the VD from the summit and the height of the cone. In Figure 3b the failure does not reach the summit of the cone, because D is too large for its given dip and height of the cone.

[34] At t = 7′ the failure increases its dip-slip and strike-slip displacement (Figure 3c). As a result of the increase in the height of the central scarp, a third failure, parallel to the slope and with decollement depth ∼4 mm, propagates downward from the cone summit. Therefore a shallower failure (with hcol/hcon ∼ 0.05, where hcol is the depth of the collapse and hcon is the height of the cone) forms in the summit as a result of the activity of a deeper failure (hcol/hcon ∼ 0.3).

[35] Experiment COL 27 has E = 1, D = 4 cm, h = 6 cm and α = 0°; the major difference with the previous experiment is that the VD is located closer to the center of the cone. The map view of the undeformed model (t = 0′) is shown in Figure 4a. At t = 2′, the failure reaches the cone summit (Figure 4b). Its overall kinematics, characterized by dip-slip motion in the summit and strike-slip motion at the base, are similar to COL 17. However, this failure shows a marked dip variation (Figure 4c), from ∼55° (at the bottom and central portion) to ∼85° (in the uppermost portion). At t = 14′, as a result of the increased displacement, the crest of the cone partly collapses (block Z in Figure 4d). Moreover, the collapsed half crest of the cone shows fractures roughly following a concentric and radial pattern (Figure 4d, inset). These fractures are present only in experiments with summit failure and t > 10′, corresponding to a horizontal sliding >4.5 mm. At t = 60′ (horizontal slide of 2.7 cm) the collapsed part of the cone is covered by minor collapses originating from the scarp (Figure 4e), 30° to 60° wide and with 0.1 < hcol/hcon < 0.38.

Figure 4.

Set A, evolution of experiment COL 27. (a) Undeformed stage, map view. (b) Experiment at t = 2′. A deep failure develops. (c) Section view of the failure at t = 2′. The failure increases its dip while approaching the summit. (d) Experiment at t = 14′. Radial and concentric fractures (see inset) develop on the sliding part of the crest of the cone, characterized by minor collapses (Z). (e) Experiment at t = 60′. Several collapses develop as a result of the formation of a prominent scarp.

[36] The failures in all the experiments with E = 1 (Table 2) have a mean dip of 51° and standard deviation of 2.6°. Some failures steepen in the upper part: their dip variation (Δδ) as a function of D and h is shown in Figure 5 (solid circles). The failures with constant dip (dashed lines) do not reach the cone summit. Conversely, all the steepening failures (by up to 30°; solid lines) tend to reach the summit area (Figure 5). The dip variations of the failures (where Δδ ≠ 0°) are observed only for 1.3 < h/D < 1.65. Within this range, the best fit equation (R = 1) that describes the relationship between the dip variation (Δδ) and h/D is

display math
Figure 5.

Variation, in section view, of the dip δ of the failures (Δδ) as function of their position with regard to the horizontal distance from the summit (D) and the height of the cone (h). Solid circles refer to set A experiments (all the experiments with α = 0° in Table 2 have been included); open circles refer to set B experiments (see section 4.2). Error bars refer to the measurement uncertainty in a single experiment.

[37] Experiment COL 20 has E = 0.86, D = 5.5 cm, h = 8.5 cm and α = 30°; it therefore consists of a cone elongated obliquely with regard to the VD direction. The map view of the undeformed model (t = 0′) is shown in Figure 6a. At t = 4′ (Figure 6b) the failure is asymmetric in map view; this is quantified by the different values of L1 and L2 (corresponding to the length of the subrectilinear portions of the failure in map view; Figure 6b). The L1/L2 ratio therefore estimates the asymmetry of the failure. Such a ratio (∼1 in COL 17 and COL 27) is ∼0.7 in COL 20; the failure follows the elongated crest of the cone and its curvature (in map view) occurs at the far end of the crest (Figure 6b).

Figure 6.

Set A, evolution of experiment COL 20, characterized by an elongated cone. (a) Undeformed stage, map view (dashed line marks the elongated crest). (b) Experiment at t = 4′. An asymmetric failure (given by L1/L2 < 1) develops.

[38] The symmetry of the failures in map view for elongated cones with comparable boundary conditions (E ∼ 0.85, D = 7.5 cm; h ∼ 8.5 cm) as function of α is shown in Figure 7. The failures are symmetric (L1/L2 = 1) only when α = 0° or α = 90°. The maximum asymmetry (L1/L2 = 0.8) in this type of experiments is obtained for α ∼ 30°.

Figure 7.

Set A, variation in the symmetry (expressed as L1/L2) of the failures of the elongated cones (in map view) as function of the angle α between the trend of the VD and the direction of elongation of the cone. Error bars refer to the measurement uncertainty in a single experiment.

4.2. Set B: Unbuttressing

[39] This set consists of seven experiments in which the following parameters have been varied (Table 3): the eccentricity E (0.7 < E < 1) of the base of the cone; the distance D of the vertical wall from the center of the cone; the height h of the cone. One representative experiment is described.

Table 3. Experiments and Related Parameters Considered in Set B (Unbuttressing)a
Experimenth, cmD, cmh/DE
  • a

    The parameter h is height of cone, D is distance of wall from the center of the cone, and E is eccentricity of the base of the cone.

COL 187.11.11
COL 2541.251
COL 4981.120.73
COL 563.81.580.7
COL 146321
COL 15871.141
COL 16531.671

[40] Experiment COL 5 has E = 0.7, D = 3.8 cm and h = 6 cm. The undeformed stage is shown in Figure 8a. The outward slide of the mobile vertical wall (velocity of 2.7 cm h−1) induces failure at the base of the elongated cone. Similarly to what was observed in set A, the failure propagates toward the cone summit. At t = 4′ the shape of the failure in map view is the result of the intersection of a surface with variable dip (45° in its bottom and central parts and 65° in its uppermost part) and the topography (Figure 8b). The failure has dip-slip motions toward the summit and predominant strike-slip motions at the base, similarly to what was observed in set A. At t = 7′ the left of the collapsed part (Figure 8c) is characterized by minor collapses, with width and depth of ∼45° and ∼3 mm, respectively; these are therefore shallower (hcol/hcon ∼0.05) collapses propagating from the crest of the scarp.

Figure 8.

Set B, evolution of experiment COL 5. (a) Undeformed stage, map view. (b) Experiment at t = 4′. A deep failure forms. (c) Experiment at t = 7′. The left part of the sliding cone has minor arcuate collapses (indicated by arrows).

[41] Other experiments of this set show similar failures from the base of the cone. Similarly to set A, only the failures with variable dip reach the summit area, like COL 5. The dip variation Δδ of the failures of set B as function of the h/D ratio is shown in Figure 5 (open circles). Introducing the Δδ and h/D values of COL 5 changes equation (1) into a very similar equation (equation (2)), with R2 = 0.94:

display math

Equation (2) is valid for 1.3 < h/D < 1.65, where Δδ ≠ 0°. The geometry of the failures of set B depends upon the h/D ratio in a very similar way as was observed in set A (Figure 5). Therefore, despite the different apparatus used in sets A and B, the overall deformation pattern observed in the two sets is nearly identical.

4.3. Set C: Summit Growth

[42] This set consists of three experiments, in which the following parameters have been tested (Table 4): the eccentricity E (circular or elongated cone) and the location of growth of the cone (centered or lateral). All the experiments ended after 50–60 collapses. Three experiments are described.

Table 4. Experiments and Related Parameters Considered in Set C (Summit Growth)a
Experimenth, cmES, cmGrowth
  • a

    The parameter h is height of cone, E is eccentricity of the base of the cone, and S is distance between the summit of the cone and the area of growth by addition of material.

COL 6810centered
COL 7813peripheral
COL 88.50.810centered

[43] Experiment COL 6 is characterized by h = 8 cm, E = 1 and by the addition (rate of 6.6 cm3 per minute, corresponding to ∼1 m3 min−1 in nature) of the sand/flour mixture over a constant area of 4 mm2 on the cone summit. The material fell from a funnel 1 cm above the cone summit, to minimize the impact effect due to gravity. The result of the addition of material (black mixture) on the cone (white mixture) summit is shown in Figure 9. The added material increases the steepness of the upper slope, triggering a collapse from the summit. Limited in width (∼10°) and depth (∼5 mm), this collapse is localized and shallow; it is characterized by a scar in the upper part and a deposit in the lower part. The direction of the onset of the first collapse is not predictable, as it may be triggered by initial minor instabilities on the topography of the cone or by an imperfect axis-symmetric addition of mass on the cone summit. Nevertheless, the direction of the following landslides is consistent; this is shown in Figure 10a, where the orientation of the collapses (horizontal bars) is reported as function of the number of collapses. The first collapses show a clustering (shaded areas) in orientation, implying that, after the first collapse, the following one insists on the same area or at its immediate borders. However, the higher the frequency of collapses, the more scattered is their orientation, as suggested by the shorter shaded areas in Figure 10a. The collapses observed in experiment COL 6 have a mean width of 40°, with a standard deviation of 26.3° (Figure 10b).

Figure 9.

Set C, oblique view of experiment COL 6, characterized by a collapse due to the addition of material (black mixture of sand and flour) on the summit of the cone (white mixture).

Figure 10.

Set C. (a) Variation of the orientation of the collapses (horizontal bars) during experiment COL 6; the shadings cluster the collapses with similar (≤45°) orientation. (b) Mean width (in degrees) of the collapses in experiment COL 6.

[44] Experiment COL 7 is characterized by h = 8 cm, E = 1 and the addition of material at 3 cm of horizontal distance (S) from the summit of the cone, corresponding, for a slope dip ∼35°, to ∼2 cm below the summit. As a result of the lateral addition of material, an eccentric cone forms; the added material (black mixture) forms the eccentric cone on the former cone (white mixture). The black mixture covers almost half of the former cone, as a result of several shallow (hcol/hcon ∼ 0.05) collapses radiating from the new cone. The orientation of the landslides as a function of the number of collapses is shown in Figure 11a. Differently from central growth (Figure 10a), the eccentric growth of COL 7 controls the side on which collapses occur, oriented on the opposite side (south) with regard to the former cone summit. Only when the height of the newly formed eccentric cone becomes >0.3S (corresponding to ∼1 cm above the former summit in Figure 11a), the direction of the collapses starts to widen. The collapses observed in experiment COL 7 have a mean width of 21.4°, with a standard deviation of 16.1° (Figure 11b). The width of the part of the cone characterized by the collapses increases with the height of the eccentric cone (h), as shown in Figure 11c). However, this increase is not constant: in the first portion of the curve, when h/S < 0.6, all the collapses are restricted to half of the cone; conversely, when h/S > 0.6, the width of the area affected by collapse increases more rapidly with regard to the height. When h/S ∼ 1.2, the collapses radiating from the new cone have reached 360° (Figure 11c) and the topography of the new cone completely dominates that of the former cone.

Figure 11.

Set C. (a) Variation of the orientation of the collapses during experiment COL 7. The collapses mostly cluster to the south of the cone. The numbers in the upper line (in degrees) indicate the cumulative width of the cone progressively undergoing collapse, whereas those in the lower line (in cm) indicate the difference in height between the former and the new summit. (b) Mean width of the collapses in experiment COL 7. (c) Cumulative width of the cone covered by the collapses as function of the height of the newly formed eccentric cone h with regard to its distance from the former summit S.

[45] Experiment COL 8 is characterized by h = 8.5 cm, E = 0.81 and the addition of material on the center of the elongated cone. The addition of material on the cone summit induces several collapses, with depth ∼3 mm and width <10°, as shown in Figure 12a. After an initial phase of collapses on the south side of the cone, the collapses gradually switch between the north and south sides. At the end of the experiment (after ∼ 60 collapses), the collapses are mostly subparallel to Lmin (Figure 12b). Their orientation as a function of the number of collapses is shown in Figure 13a). The collapses in experiment COL 8 have a mean width of 17.8°, with a standard deviation of 13.3° (Figure 13b).

Figure 12.

Set C. (a) Detail of a collapse on the summit of the elongated cone in experiment COL 8. (b) Map view of the same experiment, showing how the area characterized by collapses (black mixture) is controlled by the former elongation of the cone.

Figure 13.

Set C. (a) Variation of the orientation of the collapses during experiment COL 8. The collapses cluster in the northern and the southern parts of the cone. (b) Mean width of the collapses in experiment COL 8.

4.4. Set D: Injection

[46] This set consists of 14 experiments, divided in two subsets, D1 and D2 (Figure 2d). D1 investigates how silicone injection affects the collapses within a cone; D2 investigates the effect of unbuttressing on the silicone rise, using a slope much longer than the width of the intrusion. The following parameters have been varied (Table 5 and Figure 2d): the eccentricity E of the base of the cone (only in D1); the height h of the cone (or the slope in D2); the initial vertical distance z of the silicone from the summit of the cone (summit of the slope in D2); the lateral distance L of the silicone nozzle from the center of the cone (or from the beginning of the slope in D2); the vertical distance T from the silicone intrusion to the surface of the cone (or of the slope in D2; here T = z) directly above the nozzle. This set is characterized by moderate values of T (< 4 cm; Table 5), implying shallow positions of the silicone intrusion.

Table 5. Experiments and Related Parameters Considered in Set D (Silicone Injection)a
ExperimentInjection Typeh, cmz, cmL, cmT, cmE
  • a

    The parameter h is height of cone (or of slope in D2), z is vertical distance between the silicone and the summit of the cone (or the slope surface in D2), T is vertical distance between the silicone and the surface of the cone, and L is lateral distance of the silicone from the center of the cone (or the onset of slope in D2). See also Figure 2.

Subset D1
COL 24cone-centered91011
COL 25cone-peripheral102−211
COL 28cone-centered52021
COL 29cone-peripheral11.52−210.86
COL 30cone-centered8.51.501.50.82
COL 31cone-centered9.52021
COL 32cone-peripheral83−121
COL 33cone-peripheral83−21.51
COL 34cone-peripheral73−0.52.51
COL 41cone-centered82021
 
Subset D2
COL 35slope7232 
COL 36slope2222 
COL 37slope2212 
COL 38slope4424 

[47] The experiments of subset D1 are illustrated first. Experiment COL 24 has a centered cylindrical intrusion of silicone (whose initial diameter is 1 cm), h = 9 cm, z = 1 cm and L = 0 cm. The undeformed experiment, in map view, is shown in Figure 14a. The intermediate stage of deformation (t = 45′) is characterized by several radial and concentric fractures (detail in Figure 14b). The final stage of deformation, when the silicone extrudes on the cone summit (t = 130′), is shown in Figure 14c; at this moment the silicone intrusion has reached a diameter ∼2 cm. The extrusion is immediately preceded by several collapses, with hcol/hcon ∼ 0.1, oriented radially with regard to the cone summit. The horizontal and vertical displacements (assumed axis-symmetric) induced by the intrusion below the cone summit (at an initial depth=1 cm), measured before the onset of the collapses, are shown in Figure 15. The cone summit is lowered and the sides are inflated and displaced outward (from R/T > 2, where R is the radial distance and T is the initial depth of the intrusion). This results in a local increase of the dip of the slope of ∼3° (from ∼34° to ∼37°). In a cone at its maximum repose angle, even such a moderate increase may cause instabilities, generating radial collapses.

Figure 14.

Set D, evolution of experiment COL 24, characterized by a centered intrusion of silicone. (a) Undeformed stage, map view. (b) Detail of the cone summit at t = 45′. Radial and concentric fractures form. (c) Experiment at t = 130′. The silicone extrudes on the summit, inducing, during its rise, radial collapses.

Figure 15.

Set D, horizontal and vertical displacements observed at surface (from map and section images in experiment COL 24) as function of the distance R from the center of the cone normalized to the initial depth T of the centered intrusion. Error bars refer to the measurement uncertainty in a single experiment.

[48] Experiment COL 25 has a peripheral intrusion of silicone (whose initial diameter is 1 cm), with h = 10 cm, z = 2 cm, L = 2 cm and T = 1 cm. The undeformed experiment, in map view, is shown in Figure 16a. Before silicone extrusion (t = 90′), the surface deformation above the intrusion consists of curved normal faults (in map view), with few milllimeters of displacement, and a bulge in the footwall, very similar to the findings of Donnadieu and Merle [1998]. The final stage of deformation, when the silicone extrudes (t = 120′), is shown in Figure 16b. The extrusion is accompanied by several collapses, with hcol/hcon ∼ 0.1, mainly at the front (Figure 16a) of the peripheral intrusion and, to a lesser extent, at the sides. The collapses at the front, radiating from the bulge, are narrower (30°–50°) compared to those at the sides (∼60°). Moreover, the final horizontal distance of the extruded silicone from the cone summit is 5 cm, larger than the initial horizontal distance (L = 2 cm). This shows that the path of the rising silicone is deflected toward the nearest slope. A similar deflection of silicone toward the nearest slope is observed in all experiments characterized by a peripheral location of the nozzle (Table 5). This suggests that the observed deflection is a result of the unbuttressing of the nearest slope; such a possibility was tested in the experiments of subset D2.

Figure 16.

Set D, evolution of experiment COL 25, characterized by a peripheral intrusion. (a) Undeformed stage, map view. White cross marks the location of the silicone nozzle (2 cm of horizontal distance from the summit). (b) Experiment at t = 120′. The silicone extrudes, inducing frontal and lateral collapses, at a larger horizontal distance (5 cm) from the summit.

[49] The lateral collapses observed in COL 25 do not occur if the cone is elongated. Experiment COL 29 consists of an elongated (E = 0.86) cone, with a peripheral intrusion of silicone (original diameter = 1 cm), with h = 11.5 cm, z = 2 cm, L = 2 cm and T = 1 cm. The intrusion is located along Lmin. As a result of the elongation, the lateral collapses are lacking and only the front of the experiment is characterized by shallow (hcol/hcon < 0.15) and localized (α < 30°) collapses. Similarly to COL 25, COL 29 displays a deflection (∼4 cm) of the extruded silicone from its initial position.

[50] The experiments of subset D2 are characterized by the intrusion of silicone within a sand/flour mixture, at variable distances from a slope (Table 5); the slope has a dip ∼35°, commonly observed along the flanks of stratovolcanoes, and a length much larger than the diameter of the nozzle, avoiding side effects in the along-slope direction. The aim of this subset is to evaluate the effect of a slope, possibly produced by previous collapses, on the propagation path of the rising silicone. The diagram in Figure 17 shows how the distance from the slope affects the deflection d of silicone during its rise. The deflection is measured as the horizontal distance between the nozzle and the top of the intruded silicone. The diagram includes data from subset D2 (where L > 0 and L < 0) and also from experiments of subset D1, where L is sufficiently large (L > 0.5 h) to avoid the possible unbuttressing effect of the opposite slope and is always considered negative (Figure 2d). If the upper portion of the slope is distant from the intrusion, the injected silicone rises vertically; the critical distance under which the injected silicone starts to deflect from its vertical path is 0.5 < L/z < 1 (Figure 17). For lower L/z ratios, the silicone deflects, migrating toward the slope. The deflection reaches a maximum, becoming twice the distance L, at L/z ∼ −0.12 (Figure 17); the negative value is due to the fact that the silicone was placed below the slope, where L < 0 (left nozzle position in Figure 17). For L/z < −0.12, the deflection d decreases, becoming eventually negligible with regard to L. The best fit (R2 = 0.91) equation of the points with d/L > 0 is

display math
Figure 17.

Set D, deflection d (normalized to the horizontal distance of the initial location of the silicone, L, in absolute values) of the intruding silicone as a function of L normalized to the height of the layer above the silicone, z. The nearer the silicone to the intruding slope (lower L/z), the higher is its deflection. However, for very low L/z values the deflection of the silicone decreases. Error bars refer to the measurement uncertainty in a single experiment. The inferred deflection of the intrusions at Mount St. Helens in 1980 is also reported, after reconstructions from Donnadieu and Merle [2001] (1), Voight and Elsworth [1997] (2), and Reid et al. [2000] (3) and at Bezymianny in 1956 [Voight and Elsworth, 1997].

5. Discussion

5.1. Interpretation of the Experimental Sets

[51] The experiments of set A and B, despite the different boundary conditions and apparatus, display similar results. The collapses affect a significant portion of the cone and have similar depths (0.05 < hcol/hcon < 0.4). The geometry of the failures is partly controlled by topography, both in section and map view.

[52] In section view, the failures are always gentler (45°–55°) than ones (∼60°) obtained within horizontal layers, with a similar apparatus, in the absence of topographic gradients [Acocella et al., 1999]. This decrease in the dip of failures is interpreted as due to the inclined gravitational stresses within the middle and lower part of a cone. The cone in Figure 18 refers to the stress distribution of a section of a linear ridge [Dieterich, 1988]. This can qualitatively approximate the state of stress along a section passing through the summit of a cone, as both sections are subject to an outward stress produced by the topographic gradient [Borgia, 1994; van Wyk de Vries and Matela, 1998; Ventura et al., 1999]. The stresses in Figure 18 lower the dip of the failures at the base of the cone. The cone summit is affected by outward dipping gravitational stresses (Figure 18), responsible for tensile conditions; as a consequence, steeper failures develop. Therefore the variation in the dip of the failures (Figure 5) can be explained by the distribution of the gravitational stresses within a cone. In map view, the failures show asymmetries only when the cones are elongated, following their crest; the failure steepens along the crest, becoming longer and linear (in map view; Figure 6).

Figure 18.

Orientation of the maximum principal stress within a cone with a Poisson's ratio of 0.45 (modified after Dieterich [1988]).

[53] The kinematics of the failures is also consistent in both sets. The variation in the kinematics results from the intersection of an inclined surface, with uniform movement, and the cone topography. The parts of the failure which appear, in map view, subparallel to the VD (top of failure) show normal motions; the parts which appear oblique or perpendicular (base of failure) show predominant strike-slip motions (Figure 3). No significant geometric or kinematic similarities are observed between the experiments of set A and B and those performed with a vertically sliding VD [Vidal and Merle, 2000; Merle et al., 2001].

[54] The experiments of set C develop shallow (hcol/hcon < 0.1) collapses, with limited (very few tens of degrees) width. These are characterized by an elongated landslide scar in the upper part, associated with a landslide deposit, usually with a fan-like geometry, at the bottom. The tendency of the collapses to cluster in orientation in the earlier stages is due to the fact that, once a collapse develops, it produces a scar at its top. The scar is a preferred area to be filled by the forthcoming material that, after passing the repose angle, will slide again, preferentially reactivating the same basal detachment. The sides of the landslide scar usually provide topographic barriers for the forthcoming slide.

[55] The narrower collapses on elongated cones (Figures 12 and 13) are due to the decrease of its curvature, which better laterally constraints the sliding material. The collapses occur mostly on the sides perpendicular to the cone elongation, until the dip of the slope perpendicular to the elongation direction reaches the dip of the slope on the parallel sides; when this is achieved, the collapses propagate in all directions.

[56] The narrower collapses of an eccentrically growing cone (Figure 11) may also be explained by the initial topography of the cone. In fact, the propagation of the newly formed slides from the eccentric cone is hindered at the sides by the presence of the slope of the older cone (experiment COL 7).

[57] The experiments of set D show how intrusions trigger collapses and how slopes, possibly produced by previous collapses, affect the path of a rising intrusion. Before the collapse stage, the cone undergoes inflation and, in the upper part, subsidence (Figure 15); the deformation pattern is here given by concentric and radial fractures (Figure 14b). The former result from the tensional stresses on the crest of the cone; the latter accommodate the increase of circumference of the cone due to the intrusion [Acocella et al., 2001]. The inflation of the cone increases the dip (∼3°) of the upper slope where R/T > 2 (Figure 15), triggering shallow collapses (hcol/hcon ∼ 0.1). An increase of 3° in the slope angle can reduce its factor of safety by ∼15% [Reid et al., 2000], but in a cone already at its maximum repose angle, it generates collapses. Similar results highlighted, through numerical models, the relationships between reservoir overpressure and the shallow failures along a volcano slope [Russo et al., 1997].

[58] Thicker collapses (hcol/hcon ∼ 0.15) develop above eccentric intrusions, where the deformation due to injection focuses on one side of the cone. On elongated cones, collapses are hindered by the stronger buttressing along the major axis due to the lower dip of the upper slope (20°–25°).

[59] An inverse proportion between the distance from the slope and the deflection of silicone occurs for L/h < 1; for L < 0 the effect of the slope is widely distributed over the intrusion and the deflection decreases (Figure 17). These results suggest that the silicone deflection is due to the nearby slope or, more generally, to a topographic gradient. The outward distribution of the stresses within a cone (Figure 18) may explain the deflection of the eccentrically rising silicone.

5.2. General Features of the Collapses in the Experiments

[60] The performed experiments show how the collapses with different geometries (width, depth, length) develop on a cone depending on the boundary conditions (Figure 19). The width of the collapses in map view (angle β) with regard to the total width of the base of the cone (360°) and the length of the collapses in map view (Lcol) with regard to the length of the slope on the side of the collapse (Lcon) are shown in Figure 19a for the four sets of experiment. The collapses from set A and B are usually wider (0.2 < β/360° < 0.3) and of variable length (Lcol/Lcon > 0.3); conversely, the collapses from set C are mainly narrower (β/360° < 0.2) and mostly long (clustering of data for Lcol/Lcon > 0.8); the collapses of set D show an intermediate behavior, with medium width (0.1 < β/360° < 0.2) and moderate to high length (0.4 < Lcol/Lcon < 1).

Figure 19.

Distribution of the width β, the depth hcol, and the length Lcol (normalized to 360°, the depth hcon and length Lcon of the cone, respectively) of the experimental collapses in the four sets of experiments. (a) Width/length ratios of the collapses. (b) Width/depth ratios of the collapses.

[61] The β/360° ratio of the collapses and the depth of the collapses (hcol; measured from the mean point of the collapse, perpendicularly to the failure surface) with regard to the height of the cone (hcon) are shown in Figure 19b. The collapses from set A and B consist of deeper (0.1 < hcol/hcon < 0.4) and wider (β/360° > 0.2) failures. Conversely, the collapses of set C are shallower (hcol/hcon < 0.1) and with width β/360° < 0.3. The collapses of set D are slightly deeper (hcol/hcon ∼ 0.1) and with moderate width (less scattered: 0.1 < β/360° < 0.2). Therefore shallower and narrower collapses are related to summit growth (set C) or, to a lesser extent, to silicone intrusion (set D), simulating extrusive and intrusive activity, respectively. Differently, the deeper and wider collapses affecting the central portion of the cone are related to basal failure (set A) or unbuttressing (set B).

[62] Previously performed experiments also show that deeper and wider collapses are related to basal spreading [Merle and Borgia, 1996; Merle and Lenat, 2003], basal failures with vertical [Vidal and Merle, 2000; Merle et al., 2001], oblique [Wooller et al., 2003], or strike-slip [van Wyk de Vries and Merle, 1998; Lagmay et al., 2000] motions or all three [van Wyk de Vries and Merle, 1996, 1998].

[63] Overall, the experiments suggest that, to have a collapse of a consistent portion of a cone, the deformation of its base is required. This mechanism is usually not related to magmatic activity, as shallower and narrower collapses form in this case.

5.3. Comparison to Nature

[64] The comparison between each set of experiments and natural examples highlights aspects of the mechanisms of sector collapse. The variation in the geometry (dip and strike) of a failure plane within an experimental cone, observed in sets A and B, may have important implications in nature; for example, whether a failure or a collapse (both deep and shallow) reaches certain areas (magmatic conduit, rift zones, intrusions) of a volcano with a given topography.

[65] The section views of sector collapses at Vesuvio, Socompa and Tungurahua have been reconstructed from published maps [Ventura et al., 1999; van Wyk de Vries et al., 2001; Wooller et al., 2003], considering the altitude of the failures and assuming their constant dip along strike; the obtained depths of failure are therefore a minimum estimate. These volcanoes have been selected because they have, among the considered natural examples, wide and deep collapses, whose strike is parallel to nearby regional structures (Table 6). The collapses at Tungurahua and Vesuvio have been likely triggered by regional tectonics [Ventura et al., 1999; Wooller et al., 2003], while at Socompa by basal spreading [van Wyk de Vries et al., 2001]; therefore these cases are related to a basal type of failure. The reconstructed failures show a steepening (Δδ) of 10°–30° approaching the volcano summit, similarly to what observed in the experiments. However, their shallow (15°–35°) dip results in h/D ratios much lower than those experimentally observed (Figure 5). This discrepancy may be explained by the fact that the shallower dips in nature are controlled by mechanical anisotropies between the layers (lavas, scoriae, loose pyroclastic deposits, clays) constituting the volcano; also, these products are emplaced with similar dips (0° to 35°).

Table 6. Natural Examples of Collapsing Volcanoes, Related Geometric Features, and Referencesa
VolcanoLcol/Lconβcol/360°hcol/hconDirectionReferences
  • a

    Lcol/Lcon is ratio of length of collapse with regard to the length of the slope of the cone (if Lcol > Lcon, the resulting ratio has been considered as 1); βcol/360° is width of collapse with regard to the width of the cone; hcol/hcon is depth of collapse with regard to the height of the cone (if hcol > hcon, the resulting ratio has been considered as 1); and direction is direction of the collapse surface with regard to the trend of the regional structures.

1, Piton10.07 ± 0.03 obliqueLenat et al. [1989]
2, Vesuvio0.88 ± 0.120.19 ± 0.031parallelVentura et al. [1999]
3, Tenerife10.03 ± 0.04 parallelWatts and Masson [2001]
4, Fogo10.11 ± 0.030.33 ± 0.2 Day et al. [1999a]
5, Stromboli10.09 ± 0.030.18 ± 0.02parallelTibaldi [2001]
6, Shiveluch10.13 ± 0.040.2 ± 0.03perpendicularBelousov [1995]
7, Llullaillaco10.08 ± 0.04 parallelRichards and Villeneuve [2001]
8, La Palma10.08 ± 0.050.27 ± 0.15parallelCarracedo et al. [1999]
9, El Hierro10.08 ± 0.030.22 ± 0.05parallelMasson et al. [2002]
10, Pelèe10.08 ± 0.060.11 ± 0.04parallelLe Friant et al. [2003]
11, Tungurahua0.72 ± 0.280.27 ± 0.031parallelHall et al. [1999]
12, Agustine10.1 ± 0.04 obliqueSiebert et al. [1995]
13, Mombacho10.08 ± 0.030.1 ± 0.03parallelvan Wyk de Vries and Francis [1997]
14, Socompa10.15 ± 0.030.12 ± 0.03parallelvan Wyk de Vries et al. [2001]
15, Etna10.21 ± 0.031parallelRust and Neri [1996]
16, St. Helens10.11 ± 0.030.33 ± 0.05perpendicularVoight et al. [1981]
17, Kilauea10.41 ± 0.031parallelBorgia [1994]
18, Casita0.67 ± 0.20.08 ± 0.040.18 ± 0.03obliqueKerle and van Wyk de Vries [2001]
19, Rainier (1)0.6 ± 0.30.08 ± 0.030.23 ± 0.06parallelReid et al. [2001]
20, Rainier (2)0.9 ± 0.10.09 ± 0.060.19 ± 0.06parallelReid et al. [2001]
21, Iriga10.12 ± 0.03 perpendicularLagmay et al. [2000]
22, Jocototlan10.25 ± 0.04 parallelCapra et al. [2002]
23, Patamban10.06 ± 0.05 obliqueCapra et al. [2002]
24, Soufriere10.12 ± 0.06 parallelDeplus et al. [2001]
25, Bezymianny10.12 ± 0.020.33 ± 0.06parallelDonnadieu et al. [2001]

[66] The variation of the orientation of the failure surface in map view can be appreciated at Etna, which is N-S elongated, with an asymmetric failure that follows the ridge formed by the NE and the south rifts (Figure 1b). The proposed mechanism of collapse at Etna is basal spreading [Borgia et al., 1992; Merle and Borgia, 1996] due to the deposits of clay below its south and east flanks [Rust and Neri, 1996], rather than basal failure or unbuttressing. Nevertheless, the asymmetric and deep failure at Etna results from an asymmetric configuration of the edifice with regard to its basement (the underlying clay deposits), similarly to what is observed in set A.

[67] The experiments of set C permit to predict where the failure of a volcanic edifice may occur, during its life, as a result of an eruption. Many volcanoes are characterized by the persistence of collapse in the same sector [Siebert et al., 1995; Belousov et al., 1999; Cantagrel et al., 1999; Urgeles et al., 1999; Tibaldi, 2001; Le Friant et al., 2003], developing nested collapses, similarly to what is observed in the experiments. Nevertheless, some volcanoes are characterized by multiple collapses with different orientations [van Wyk de Vries and Francis, 1997; Reid et al., 2001]. The experiments of set C suggest that, independently from the overall shape of the volcano (central or elongated, with a central or eccentric crater; Figures 5, 10, and 11) several (at least 10) collapses should occur in the same direction before the direction of the collapses starts to change. When this happens, it usually occurs at the sides of the former collapses, as a result of the lateral enlargement of the cumulative collapsing area. The experiments of set C, possibly with the exception of COL 8 (elongated cone), did not simulate any tectonic control on the morphology and shape of the cone. However, the morphology and the shape of volcanoes in nature are influenced by regional tectonics [Tibaldi, 1995]. This is shown in Table 6, where ∼65% of the strike of the collapse surfaces is subparallel to the strike of the regional tectonic structures. These facts confirm that fractures promoted by regional tectonics play an important role in controlling collapses. Therefore the experimental results and the data of volcanoes from Table 6 show a persistency in the orientation of the collapses and their clustering subparallel to regional trends, respectively. This suggests that significant changes in the orientation of the collapses of a volcano should be mostly limited (somehow controlled by tectonics) and expected on mature edifices.

[68] The typical shape of the collapses of set C (narrow, long, and shallow; Figure 19) is found at several volcanoes, such as Mombacho, Pelèe, Socompa and Stromboli. While the Pelèe collapses are due to the accumulation of volcanic products on the volcano slopes [Le Friant et al., 2003], the collapses at Mombacho and Socompa have been related to basal spreading [van Wyk de Vries et al., 2001; van Wyk de Vries and Francis, 1997] and, at Stromboli, mainly to diking [Tibaldi, 2001; Tibaldi et al., 2003]. Nevertheless, the shape of collapses (narrow, long, and shallow) suggests that an additional triggering factor to be considered at Mombacho, Socompa and Stromboli is the accumulation of volcanic material on their slopes; this is particularly evident comparing Figure 12a and the topography and bathymetry of Stromboli [Tibaldi, 2001]. A further possible application of set C experiments regards those volcanoes whose slopes are partly characterized by unconsolidated material, such as submarine and partly submerged arc volcanoes, notably the Canary Islands [Masson et al., 2002], the Lesser Antilles Arc [Deplus et al., 2001], or Tahiti Island [Clouard et al., 2001].

[69] The experiments of set D have shown how an intrusion may generate collapses. A close similarity exists with the Mount St. Helens eruption in 1980. The eruption followed a landslide induced by an earthquake and facilitated by instability due to the flank inflation [Voight et al., 1981]. Our overall results are consistent with previous models, which simulated the intrusion of viscous magma at Mount St. Helens and highlighted the role of the intrusion on the development of the sector collapse [Donnadieu and Merle, 1998, 2001].

[70] A novelty of the experiments of set D is the study of the deflection of the intrusion as a consequence of unbuttressing. The reconstruction of the events at Mount St. Helens has permitted the geometry of the intrusion that induced the 1980 eruption to be inferred; the slightly eccentric intrusion was deflected toward the north during its rise [Donnadieu and Merle, 2001; Voight and Elsworth, 1997; Reid et al., 2000]. Its geometric features have been included in the diagram in Figure 17, together with the reconstructed geometry of the Bezymianny intrusion in 1956 [Voight and Elsworth, 1997]; the resulting deflections, compared to the height of the cones and distances from its summit, are in agreement with the experimental results. These results suggest that cryptodome intrusions in nature are deflected, as a result of the unbuttressing of a near slope. Even though the experiments of set D did not specifically simulate the intrusion of dikes, similar processes may occur also in basaltic volcanoes, where the eruptions are mostly characterized by diking. In fact, the focusing of the intrusions along the scarp of the Valle del Bove at Etna [Acocella and Neri, 2003] and at the sides of the Sciara del Fuoco at Stromboli [Tibaldi, 2001] (Figure 1a) suggests that unbuttressing enhances fracturing and diking in the vicinity of scarps.

[71] It should also be noted that, in accord with the results of set A and B, collapse surfaces tend to cluster along rift zones of elongated volcanoes, creating asymmetric slopes in section view. The deflection observed in the experiments suggests that the development of rift zones within a volcano may be in turn enhanced by sector collapses, suggesting the possibility of a feedback mechanism between volcanic activity and topography. In this mechanism, the presence of a prominent scarp (due to previous collapses) on a volcano flank determines its unbuttressing, which induces the deflection of the rising magma, which, in turn can be responsible for further collapses at the head of the scarp.

[72] The overall geometric features of natural collapses (β/360°, Lcol/Lcon and hcol/hcon), as obtained from the works listed in Table 6, are displayed in Figure 20, in order to provide a comparison to the geometry of the experimental collapses (Figure 19). Sector collapses in nature similarly cluster at the bottom right of the β/360° versus Lcol/Lcon diagram (Figure 20a), indicating that they mostly reach the base of the volcano and have limited width. In particular, the collapses triggered by magmatic activity (both extrusion and intrusion of magma) are located at the bottom of the diagram (gray numbers in Figure 20a). The overall distribution is similar to the one observed in the experiments (Figure 19a); this is particularly evident for the data related to intrusive and extrusive activity, consistently with the experiments from sets C and D, simulating the effect of magmatic activity.

Figure 20.

Distribution of the width β, the depth hcol, and the length Lcol (normalized to 360°, the depth hvol and length Lvol of the volcano, respectively) of the natural collapses considered in Table 6. (a) Width/length ratios of the collapses. (b) Width/depth ratios of the collapses. Gray numbers refer to collapses related to intrusive or extrusive activity (from the references in Table 6).

[73] The sector collapses in nature also display hcol/hcon versus β/360° ratios similar to the experiments, mostly clustering at the bottom left of Figure 20b. Those resulting from extrusive and intrusive magmatic activity (Table 6; gray numbers in Figure 20b), display very limited width and depth. This distribution is consistent with the experimental results of Figure 19b, where the collapses at the bottom left are mostly related to the experiments from sets C and D, simulating the effect of magmatic activity. The deeper and wider collapses in the bottom left corner of Figure 20b (Tungurahua, Vesuvio) are related to basal failure associated or promoted by regional tectonics [Hall et al., 1999; Ventura et al., 1999], also confirmed by the parallelism between the failure surface and the regional tectonic structures (Table 6). This distribution is consistent with the experimental results of Figure 19b. The two collapses to the right of the diagram in Figure 20b (Kilauea, Etna) show deeper failures, related to basal spreading [Borgia, 1994].

[74] Therefore the comparison among the geometric features of the experimental and natural data shows many similarities, strengthened by the common distribution of the collapse mechanisms. However, minor discrepancies exist. These are mainly due to two factors. The first is the general lack of collapses with the lowermost Lcol/Lcon values in nature, differing from what is observed in the experiments. This fact is interpreted as due to a bias in the selection of the natural examples, aimed at the identification of the clearest and most visible (and thus longest) collapses; also, the lower Lcol/Lcon of a collapse, the higher is the probability to be obliterated by volcanic activity. The second discrepancy is due to the slightly higher hcol/hcon values of the collapses related to magmatic activity in Figure 20b. Of these, the known collapses (Mount S. Helens, Shiveluch, Bezymianny) with the highest hcol/hcon values occurred during the magma intrusion and were ultimately triggered by earthquakes [Voight et al., 1981; Belousov et al., 1999; Voight and Elsworth, 1997]. The experiments of silicone intrusion have slightly lower hcol/hcon values (Figure 19b). Such discrepancy may be explained by the shaking due to the earthquakes triggering the collapses, mobilizing a much larger mass than the one that may have been mobilized by an intrusion alone. This possibility is confirmed by theoretical studies [Donnadieu et al., 2001].

6. Conclusions

[75] 1. The deeper failures in sets A (basal failure) and B (unbuttressing) show variations in dip and strike, resulting from the distribution of the stresses within a cone. Similar variations are observed in nature.

[76] 2. In set C (summit growth), the persistent location and direction of the long, shallow, and narrow collapses are controlled by the cone shape and preexisting collapses. Many collapses in nature show these features and the accumulation of material along their slopes may be an additional triggering factor to be considered.

[77] 3. The silicone injected eccentrically in the cones of set D (silicone injection) is deflected as a result of unbuttressing due to nearby scarps. Similar deflections occur in nature (Mount St. Helens, Bezymianny) and, if persisting, may trigger repeated collapse, accounting for a feedback mechanism between topography and volcanic activity, at mafic and silicic volcanoes.

[78] 4. In general, collapses in sets A and B are deeper and wider, whereas thinner and narrower collapses are observed in set C; the collapses in set D have intermediate geometric features. The comparison with nature shows a similar distribution in the geometry of the sector collapses as a function of their mechanism.

Acknowledgments

[79] The author wishes to thank R. Funiciello for encouragement and helpful discussions, L. Minore for help in the set up of some experiments, B. Van Wyk de Vries for useful suggestions, and B. Behncke and M. Neri for a critical read of the manuscript. Careful reviews from William Chadwick, Doerte Mann, and an anonymous Associate Editor helped to significantly improve the work. Partly financed with GNV funds (“Emergenza Stromboli” special project, coordinated by A. Tibaldi). I thank the IGCP Project 455 for the possibility to discuss with several researchers on these topics.

Ancillary