Although dikes are found of all magma compositions, by far the most common and volumetrically significant dikes are basaltic in composition (Jerram and Bryan, 2015). Polygenetic mafic systems are commonly fed by arrangements of radial or sub-parallel mafic dikes (Figures 7A,B), frequently sourcing monogenetic flank eruptions or fissure events, but also central-vent eruptions (Figure 2C; Hildreth et al., 1998; Keating et al., 2008; Rose et al., 2010; Gudmundsson, 2012; Gudmundsson et al., 2014; Harp and Valentine, 2018). Magmatic intrusions, especially dikes, are an important source of instability, since they produce deformation, seismicity and release of mechanical (i.e., expansive) and thermal (i.e., heating) energy. This can increase pore fluid pressures, acting on potential basal failure planes and decreasing strength (Figure 6A; Elsworth and Voight, 1996; Norini and Acocella, 2011; Hacker et al., 2017). Despite thermal pore fluid pressure changes acting faster than their mechanical counterparts, the latter have a more significant impact on the strength drop (Elsworth and Voight, 1996). Earthquakes may be induced by pore fluid pressure variations caused by dike intrusions (Elsworth and Voight, 1995). In this sense, dike intrusion during modern eruptions of basaltic volcanoes has led to moderate magnitude (Mw 4–6) seismicity (Liu et al., 2018; Williams et al., 2019; Ágústsdóttir et al., 2019), but occasionally these earthquakes may reach Mw > 7 (Chen et al., 2019). During the December 2018 eruption of Etna, a dike intruding in the eastern flank caused spreading (see section “External factors” for spreading; Bonforte et al., 2019). Magma intruded within the sedimentary cover on top of the broad intrusive mesh, generating a transient velocity increase consistent with stress propagation on the unstable flank, and culminating, a few days later, with a magnitude MW = 4.9 earthquake on the Fiandaca–Pennisi Fault (10 km SE from the eruption fissure; Giampiccolo et al., 2020). Thus, increased magma supply rate, preferential intrusion orientation parallel to the maximum regional stress or even intrusion through pre-existing fault planes may promote flank failure (Siebert, 1984; Tibaldi, 2004; Famin and Michon, 2010; Giampiccolo et al., 2020; Hickey et al., 2020). In Cumbre Vieja volcano (Canary islands), dikes of a meter in thickness, but with significant length (>1 km) may cause flank destabilization (Elsworth and Day, 1999).

The temporary development of shallow intrusions (such as shallow sheet-like intrusions within a few hundreds of meters in depth, Figure 7C) is also able to produce flank deformation and collapse (Tormey et al., 1995; Solaro et al., 2010). Deep magma intrusions may also produce edifice deformation and enhanced sliding, as observed during eruptive cycles in Etna, Stromboli (both in Italy), and Cumbre Vieja (Canary Islands; Walter et al., 2005; Bonaccorso et al., 2011; Alparone et al., 2013; Nolesini et al., 2013; González and Palano, 2014). Events of cryptodome formation, such as that observed during the 1980 Mt. Saint Helens collapse (Reid et al., 2010) are uncommon in mafic volcanoes given the comparatively lower viscosity of basaltic or basaltic andesite magmas [see Lesher and Spera (2015) for additional physical and physico-chemical properties of basaltic magmas], but cannot be discarded as in some volcanoes highly crystalline basalts are found (Kushendratno et al., 2012; Fox, 2019). The exact same factors related to deep magma intrusion and edifice deformation may act at mafic volcanoes in the same way as in volcanoes with wider compositional range.

Hydrothermal alteration is widely known as an instability factor for the lateral collapse of volcanoes (Delmelle et al., 2015). Studies carried out in andesitic arc stratovolcanoes point to acid-sulfate alteration (i.e., smectite, alunite, kaolinite ± pyrite, and gypsum) as a main indicator of structural weakening (Zimbelman et al., 2005 and references therein). Similar processes are likely to occur in mafic volcanic edifices (Navarre-Sitchler et al., 2009). According to Moon and Jayawardane (2004), the early stages of basalt weathering (fresh to slightly weathered) consist of a rapid reduction in both CaO and MgO concentrations without major mineralogical change, but with a dramatic loss of strength (-43% of bearing capacity). Weathering also involves loss of cations and disruption of mineral lattice structures, and with more extensive weathering, secondary clay minerals form. The loss of shear strength will be a function of the type of alteration (related to T and pH) rather than the amount or degree of alteration (del Potro and Hürlimann, 2009), and in basaltic lava and pyroclasts it may vary between -45 and -85% (Pola et al., 2014). Low-T hydrothermal alteration products of basalts (i.e., <300°C) include smectites, zeolites, calcium silicates, calcite, pyrite, and quartz (Kristmannsdóttir, 1979): mineral assemblages that would result in structural weakening of the affected rocks. Geophysical surveys have imaged hydrothermal systems at large mafic arc stratovolcanoes such as Fuji (Japan), Llaima or Osorno (Chile) at shallow levels (∼1 km), with kilometric extension inside or below the edifice (Aizawa et al., 2005; Franco et al., 2019; Díaz et al., 2020), indicating the possible presence of hydrothermally altered cores inside these edifices. These cores may deform under gravitational deformation producing a series of surface landslides before, during and after the main flank failure event (Cecchi et al., 2004). Thus, there is good evidence that a very similar range of hydrothermal alteration and weakening processes can apply in mafic edifices, in the same way that occurs in edifices dominated by other compositions. Indeed, the rates of such weathering may commonly be high in mafic edifices, particularly those characterized by a vigorous hydrothermal system.

Volcanoes that are particularly unstable are those with weak zones represented by intense silica-clay alteration, active acid lakes and/or solfatara fields (Figure 7D; van Wyk de Vries et al., 2000; Cecchi et al., 2004; Delmelle et al., 2015) commonly related to hydrovolcanic and glaciovolcanic interactions. Mafic ice-capped or tropical volcanoes, such as Copahue, Planchón (Both in Chile-Argentina) or Poás (Costa Rica) can develop persistent intracrater acid and clay alteration (Rodríguez and van Bergen, 2017; Báez et al., 2020; Romero et al., 2020b) in this way. However, other mafic volcanoes with persistent magmatic influx such as Villarrica (Chile), Etna or Stromboli (Italy) have only developed a weak or peripheral hydrothermal system, which respond to magmatic intrusions and seasonal meteoric waters (Finizola et al., 2003; Ortiz et al., 2003; Liotta et al., 2010). Such systems are less prone to being extensively affected by hydrothermal alterations, and it is not therefore simply the case that a high magmatic flux and elevated thermal gradient within an edifice is associated with more active or more extensive hydrothermal alteration.

Rapid edifice building, typical of mafic stratovolcanoes in arc settings, normally means the generation of significant height differences above the surroundings (2–4 km), constructional processes that outpace erosion, and the development of typically steep slopes (30–40° for stratovolcanoes and cinder cones). These volcanoes may be active during hundreds of years (e.g., Sangay erupting since at least 1628 AD or Villarrica since ∼1400 AD; Monzier et al., 1999; van Daele et al., 2014). As consequence of persistent eruptions at mafic, they constitute many of the largest continental volcanic edifices (e.g., Semeru and Slamet, Indonesia; Fuji, Japan; Etna, Italy; Arenal, Costa Rica; and Llaima and Villarrica in Chile, etc.). These all act to increase flank instability (Ponomareva et al., 2006), and are factors promoted by the elevated magmatic flux and frequent eruptions that characterize many mafic arc systems.

Under gravitational load, volcanoes may also develop “spreading” or outward edifice displacement, and “sagging”, which implies inward displacement (Borgia, 1994; Merle and Borgia, 1996). Spreading promotes summit extension and basal constriction (Figure 6A), potentially forming incipient flank instabilities (Karstens et al., 2019), while sagging will produce inner thrusting, constricts the upper edifice and thus impedes magma rise. Within their “lifetime”, some volcanoes may experience a full cycle of deformation processes (Borgia, 1994) which may overlap, repeat, or be omitted: building, sagging, intruding (magmas trapped in the base of the edifice) and spreading. Such factors may be important in volcanoes of all types and compositions, but again the rapid construction characteristic of many mafic systems potentially exacerbates the influence of spreading and loading effects. Spreading is also promoted by external factors, such as a weak basement (substratum), and will be detailed in the next section.

External factors may not be expected to differ substantially between mafic edifices and those of other compositions, unless the factors are promoted in specific volcano-tectonic settings that are also characterized by mafic volcanism. The geometry and type of substratum controls basal edifice spreading in continental environments, and spreading is possible if (1) there is a weak layer in the substratum (Figure 6A; evaporates, marls, clays, hydrothermally altered rocks or shales, or poorly consolidated marine sediments), (2) the thickness of the brittle layer above the weak layer is relatively thin, and (3) the volcano is not buttressed on all sides (Merle and Borgia, 1996). Excellent examples of spreading are found in Mombacho (Nicaragua; van Wyk de Vries and Francis, 1997), Mt. Iriga (Philippines; Paguican et al., 2012), Mt Usu (Japan, Goto and Tomiya, 2019), and probably the volumetrically largest in Socompa (Chile, van Wyk de Vries et al., 2001) with 35 km³ of VDAD. There is limited evidence supporting spreading at mafic arc edifices, however, this process has been described in Ritter Island (Papua New Guinea; Karstens et al., 2019), and the lower slopes of Mt. Etna (Acocella et al., 2003; Lundgren et al., 2004; Solaro et al., 2010; Bonforte et al., 2011), as indicated by the growth of anticline structures, while the spreading is probably driven by deeper magma injection. In contrast, spreading has been widely reported in large-sized volcanic islands, where the underlying sediments form a low-viscosity decollement which results in the formation of rift zones, generally triangular, producing flanks prone to sliding (Walter and Troll, 2003; Münn et al., 2006). Possible explanations for the lack of documented spreading at mafic arc edifices are that (1) a weak substratum is not especially common in arc settings and (2) mafic arc volcanoes have comparatively smaller volumes than larger edifices with a broader compositional range (and smaller than intraplate volcanic islands), and as such, spreading is a less significant factor in many mafic settings.

Arc volcanism is associated with extensional, compressional, strike–slip or oblique motions; the structure of the arc does control the volcanic output through different processes, even though extension (also in strike–slip and compressional arcs) is directly related to output rate (Acocella and Funiciello, 2010). A review of volcanism in compressional settings (Tibaldi et al., 2009) demonstrates that active intra-arc fault systems commonly control the position of volcanic centers and contribute to a causal relationship between magmatism and thermal weakening of the upper crust (de Saint Blanquat et al., 1998). For instance, within arc fault systems, strike-slip domains favor the production of less hydrous magmas, limit crustal assimilation and shorten residence times, hence producing more mafic volcanism than occurs in contractional/transpressional settings, as can be found along the Southern Andean volcanic zone (SAVZ) or Eastern Anatolia (Tibaldi et al., 2009). The transtensional Liquiñe-Ofqui Fault System in the SAVZ hosts numerous stratovolcanoes dominated by basaltic and basaltic andesitic products (e.g., Planchón, Callaqui, Llaima, Antillanca, Osorno, Hornopirén, and Apagado volcanoes). Here, Andean transverse faults play a role in the distribution of basaltic lavas: Most late Pleistocene and Holocene volcanic centers defining NE-trending alignments, including both composite stratovolcanoes and minor eruptive centers, contain mainly basaltic to basaltic andesite lithologies (e.g., Osorno–Puntiagudo–Cordón Cenizos volcanic chain in the Southern Andes), while NW-trending fractures and faults have more evolved compositions, including rhyolites (Stern et al., 2007; Cembrano and Lara, 2009; Bucchi et al., 2015; Pérez-Flores et al., 2016; Sielfeld et al., 2019). This is in contrast to the Indonesian volcanic arc, where the Sumatra Fault has limited control on volcanic processes, distribution and size (Acocella et al., 2018). These crustal fault intersections represent highly permeable zones for fluid circulation and magmatic intrusion (Piquer et al., 2019; Norini et al., 2020; Pearce et al., 2020), but also transfer structural instabilities to the overlying edifice. The role of the regional tectonic regime and the collapse of stratovolcanoes has been studied for decades (Moriya, 1980; Francis and Wells, 1988; Tibaldi, 1995; Lagmay et al., 2000; Wooller et al., 2009; Mathieu et al., 2011; Paguican et al., 2012), and as is illustrated by the examples above, is a potentially significant control on both the construction and failure of mafic edifices, just as in edifices dominated by broader or more evolved compositional ranges.

Particular collapse directions might be influenced by basement fault interactions, such as fault intersections and complex fault damage zone architectures. In general, for volcanoes lying over normal faults, the fault will propagate through the edifice and produce a shallow graben, which will constrain instability to be nearly parallel, not perpendicular to the fault strike (Figure 6B; Wooller et al., 2009) and the collapse is in favor of the downthrown block (Figure 6B; Vidal and Merle, 2000). Vertical and thrust faulting favors landslides normal to the fault strike (Figure 6B; van Wyk de Vries and Davies, 2015). Significant instability directed normal to thrust fault strike is most likely to develop when a fault is sited beneath a peripheral flank, where it is able to destabilize a large portion of the edifice (Wooller et al., 2009; Paguican et al., 2012). Volcanic edifices lying over strike-slip fault systems usually generate a pair of sigmoidal faults (one reverse and another normal; Figure 6B) with a “flower-like” structure (folded structures associated with strike-slip faults) inside the edifice which allows magma intrusion and promotes instability (Lagmay et al., 2000). Collapse areas are normal to the strike of the dike injections and affect the cone flanks subject to extension (Mathieu et al., 2011). In summary, the role of basement faults seems to be the same for volcanoes of different compositions, with the exception that some transtensional tectonic settings favor the occurrence of mafic volcanoes, as aforementioned. In transtensional tectonic regimes such as these typically controlling mafic volcanism in the SAVZ or Indonesia, volcanic edifices are prone to growth following elongated architectures (as result of the overall feeder dikes geometry), but these elongated edifices may increasingly lead to normal-to-the-volcano-elongation steeper flanks, triggering avalanches and flank erosion (e.g., Callaqui volcano in Chile, Figure 7E; Polanco and Naranjo, 2008; Sielfeld et al., 2017).

Climate and the related erosive forces may also play a role in volcanic instability (Figure 7F; Roberti et al., 2021), at both long and immediate time-scales. However, this is a field in progress and subject to active discussion, and it is not clear whether there are particular factors that would make these processes more or less significant in mafic versus other settings, except for the fact that the often relatively smaller dimensions and younger age of mafic edifices may reduce the overall importance of edifice erosion in influencing lateral collapse. Several works (Keating and McGuire, 2004; Capra, 2006; Quidelleur et al., 2008; Roverato et al., 2011, 2015; Tost and Cronin, 2016) suggest that major inter-eruptive lateral collapses were absent during glacial climax, but started immediately during glacial retreat and sea level rise. Some of this evidence (Capra, 2006) is subject to the new geochronologic constrains and the finding of other different triggering mechanisms at these case studies (van Wyk de Vries et al., 2001; Clavero et al., 2008; Clavero and Godoy, 2010; Jicha et al., 2015).

The pressure drop caused by sector collapses (Manconi et al., 2009) may trigger eruptions by either sudden decompression of the hydrothermal system, shallow magma or by initiation of renewed magma ascent (Watt, 2019), but also to suppress an ongoing eruption (Pinel and Albino, 2013). For instance, “Bandai-type” collapses are entirely non-magmatic in nature, but may be accompanied by phreatic explosions as hydrothermal system is decompressed; non-juvenile tephra fallouts (10⁶–10⁸ m³) are typically emitted, and then the activity can be followed by magmatic or phreatomagmatic eruptions. This may be the case of Antuco, where the VDAD does not contain juvenile material, but diluted PDCs overlie the collapse sequence (Moreno et al., 2000; Romero et al., 2021).

The effect of decompression in hydrous basaltic melts includes volatile (i.e., H₂O and CO₂) oversaturation, degassing, crystallization and cooling (Pichavant et al., 2013; Arzilli et al., 2015, 2019 and references therein). Isothermal (T of 1,025°C) basaltic andesite rapid magma decompression experiments (from 150 MPa to 100, 65, 42, 21, and 10 MPa) of Shea and Hammer (2013) show that the crystallizing textures are comparable to those observed by cooling. Crystallinity may reach ∼30% below 40 MPa, with an important effect on magma rheology, thus affecting degassing efficiency, and finally affecting explosivity (Lindoo et al., 2017; Arzilli et al., 2019). Apart from the crystal fraction, bubble nucleation also affects magma rheology (rigid spherical bubbles increase bulk viscosity while deformable bubbles reduce it; Giordano, 2019). One of the most immediate volcanic phenomena related to the rapid decompression of shallow magma during a sector collapse may be a blast (if a collapse occurs during an eruption or when magma is within the shallow conduit), which is caused by the rapid collapse of an inclined explosive column resulting from instantaneous decompression initiated by lateral collapse. Blasts form relatively dense, density-stratified and grain-size-stratified PDC (Belousov et al., 2007). As blasts require high viscosity magmas to be formed (i.e., andesitic to rhyolitic; Belousov et al., 2020) there are very few examples in mafic systems, and such processes may, in general, be more common in volcanic systems characterized by more evolved magmas. They may, however, occur in basaltic systems erupting high-crystallinity magma (Kitamura and Matías, 1995; Vallance et al., 1995; Natland and Atlas, 2003; Pallister et al., 2012).

Available examples of mafic syn-collapse magmatic eruptions include Secche di Lazzaro phreatomagmatic eruption (Stromboli) and its related PDCs during the Neostromboli sector collapse (6 ky BP) (Bertagnini and Landi, 1996; Giordano et al., 2008; Petrone et al., 2009), the 1888 collapse of Ritter Island followed by a powerful submarine explosive eruption triggered by collapse (Papua New Guinea; Watt et al., 2019), or the 23 December 2018 collapse of Anak Krakatau (Williams et al., 2019), which was immediately followed by intense surtseyan eruptions with an elevated magma flux and gas release relative to typical historical eruptions (Gouhier and Paris, 2019). Additional evidence of syn-collapse explosive activity, constrained by the existence of pyroclastic deposits directly overlying VDADs, is found in Sajaka and Kanaga (Aleutians), Oshima-Oshima island in Japan, Pacaya 1 ka (Guatemala), where an ongoing Strombolian eruption (unit Pc-t5a) became more intense after the lateral collapse (c. 0.6 km³) (Kitamura and Matías, 1995; Vallance et al., 1995; Coombs et al., 2007; Satake, 2007). Another potential case is the 7–4 ka sector collapse of Antuco volcano (Chile; 5 km³) which was apparently followed by a Plinian tephra fallout dispersed eastward, and westward stratigraphy includes massive PDCs and subsequent lava flows (Moreno et al., 2000; Clavero and Godoy, 2010; Romero et al., 2020c). However, classifying these eruptions as “syn-collapse” is a difficult task if the timing of these stratigraphic events is not well constrained.

The extrusion of lava domes or effusive eruptions, soon after a sector collapse without any syn-eruptive activity, are usually called “Unzen-type” or “Mayu-Yama” (Ui, 1989). Again, effusive eruptions of large volume may occur in mafic edifices, in a similar way to those observed in more evolved systems. A good example of this type of collapse may be that of the basaltic Planchón volcano (SAVZ; Chile-Argentina) about 90–20 ky BP, which produced a new volcano (Planchón II) fundamentally formed by basaltic lava flows, some of them up to 19-km-long, right on top the collapsed edifice (Naranjo et al., 1999; Naranjo and Haller, 2002).

Several processes have been recognized as indicators of a response of a magma plumbing system to decompression following lateral collapse (cf. Watt, 2019), including shifts toward more mafic erupted compositions, the eruption of anomalous magma compositions, changes in eruption rates and the migration of vents. Some of these factors may be expected to be less recognizable in mafic systems, given their more restrictive compositional ranges and that they are generally associated with a less well-developed shallow crustal storage and plumbing system (which is the part of a plumbing system likely most strongly influenced by surface loading and unloading associated with edifice growth and destruction; cf. Pinel and Jaupart, 2005; Pinel and Albino, 2013). Nevertheless, examples such as those at Ritter Island (Watt et al., 2019) and Stromboli (Petrone et al., 2009), cited above, indicate that mafic systems may be just as susceptible as more evolved systems to being modulated by surface loads.

The reorganization of the plumbing system after lateral collapse has been associated with main vent migration and a new arrangement of dikes. Even before collapse, a single outward-creeping flank is sufficient to modify the entire rift architecture of a volcano (Walter and Troll, 2003). In mafic volcanoes, new intrusions will tend to be oriented parallel to the collapse scar following the un-buttressed flank (Figure 9; Tibaldi, 2004; Acocella and Tibaldi, 2005); the dikes outside the ridge generally dip outward, while in the collapse scar and along the scar side, there is an important deviation from the ridge trend where dikes are found en échelon and strike oblique to the scar wall (Walter and Troll, 2003; Delcamp et al., 2012, 2018). The location of the main vent usually changes toward the collapse depression (Figure 9; Tibaldi et al., 2008; Maccaferri et al., 2017) by deflecting the pathways of magmatic intrusions underneath the volcano, which results in the formation of a new eruptive center within the collapse embayment. Evidence reported from Sajaka, Kanaga, Great Sinkin (all of them in the Aleutians), Planchón (SAVZ), Fuego (Guatemala), and Stromboli (Italy), shows that the shift in the location of the main vent after the sector collapse is commonly in the order of a few hundred meters to a few kilometers (1–3 km) (Supplementary Table 2). However, this pattern depends on the local tectonic stress, which may be large enough to mask the unloading effect of the collapse (Maccaferri et al., 2017). Particularly, the Neostromboli cone shows summit, flank and satellite vents and sheet intrusions with different geometry and location through time, reflecting the interplay of regional tectonics, precursory instability processes and volcano load, in addition to petrogenetic changes after its sector collapse (Vezzoli et al., 2014).

In terms of changes to erupted compositions, post-collapse shifts in activity may imply compositional changes driven by magmatic processes related to decompression (Supplementary table 2) which result from: (1) renewed replenishment of the reservoir (i.e., its initiation or increase in the supply rate) from a deeper source, or (2) the eruption of denser, and hence primitive, magma which otherwise would have stalled at shallow depth (Pinel and Jaupart, 2005). The effusion of compositionally anomalous lavas (e.g., Ritter Island or Antuco; Watt, 2019; Watt et al., 2019) is explained by mixing between pre-existing melts and renewed mafic inputs. Moreover, a rapid regeneration of the edifice has been reported (0.6–10 km³/k.y) in mafic volcanoes and is accompanied by more mafic products lasting 10³–10⁴ years, as in Stromboli, Antuco, Ritter Island, and Oshima-Oshima (Yamamoto, 1988; Tibaldi, 2004; Martínez et al., 2018; Watt, 2019; Watt et al., 2019). Other volcanoes do not show apparent compositional changes, as Planchon II, the products of which are very similar in composition compared to that of Planchón 1 (pre-collapse). As for other types of edifice, there is not an entirely consistent pattern that has emerged in terms of the magmatic response to collapse in mafic edifices. This response is likely dependent on the spatial organization of the plumbing system and the status of the system at the time of collapse (e.g., whether shallow, eruptible magma is present), although a similar array of behaviors has been observed across all volcano types.