This study of the 2018–2019 eruptive activity at Shiveluch is based on remote sensing data acquired by optical sensors complemented by SAR and thermal sensors. While optical sensors provide a high spatial resolution and permit the three-dimensional mapping, SAR sensors possess the high temporal resolution required to detect and measure the processes involved. Accordingly, the methods section describes the techniques used to process and analyze the optical, radar, and thermal sensor data. The limitations of these methods are discussed in the Supplementary Material Limitations section.

An initial overview was obtained by utilizing satellite imagery acquired by the Dove satellites, the largest satellite constellation dedicated to imaging the Earth, and distributed by Planet Labs, Inc. (https://api.planet.com). We downloaded the available Planet data and imported the 3-m resolution imagery as background information in a geoinformatics framework using ArcGIS 10.2.1 to aid the interpretation of our higher-resolution photogrammetric dataset.

The photogrammetric dataset was acquired by processing high-resolution aerial and satellite optical imagery. Photogrammetry was conducted to extract the pre-eruption topography for July 18, 2018 and to monitor the co-eruption topography on August 22, 2019 and October 22, 2019. We also reprocessed the July 12, 2012 data described in (Shevchenko et al., 2015) with new techniques to achieve a better resolution and used the data in this work to continue the topographic chronology recorded for 2001–2012 (Shevchenko et al., 2015).

The optical satellite data were tri-stereo panchromatic 1-m resolution imagery acquired on July 18, 2018 with the Pleiades satellite PHR1B sensor. We processed the data in Erdas Imagine 2015 v15.1 similar to (Bagnardi et al., 2016; Shevchenko et al., 2020). To determine the relative orientation of the images, 37 tie points were calculated automatically with a manual correction, while a rational polynomial coefficient block adjustment, which is a transformation from pixel information to latitude, longitude, and height, was automatically employed for the interior and exterior orientations. After constraining the image orientation, we obtained a photogrammetric model with a total root mean square error (RMSE) of 0.2 m. By using the Enhanced Automatic Terrain Extraction (eATE) module with the normalized cross-correlation algorithm implemented in Erdas Imagine, we were able to extract a 2-m resolution point cloud (PC) referenced to the World Geodetic System 1984 (WGS84) coordinate system (UTM zone 57). This PC was filtered with the CloudCompare v2.9.1 noise filter and then manually cleaned with the CloudCompare segmentation tool. As strong volcanic steam emissions caused a large gap in the PC in the NE part of the dome, we used a 5-m resolution DEM constructed from TanDEM-X data (see SAR Data and Monitoring the Material Redistribution (d)) to fill the gap and obtain the missing topography.

Helicopter surveys allowed us to acquire nadir and oblique aerial images during overflights on July 12, 2012 with a conventional Canon EOS 20D digital camera (focal length: 14.183 mm, resolution: 3,504 px × 2,336 px) and on August 22, 2019 and October 22, 2019 with a PhaseOne IXA 160 digital aerial camera (focal length: 28 mm, resolution: 8,984 px × 6,732 px). The average flying height was 4,000 m a.s.l. for the nadir survey and 3,200 m a.s.l. for the oblique survey. We processed the images in Agisoft Metashape 1.5.2. To ascertain the interior orientation, we set the camera parameters (focal length and sensor size), and the relative orientation was performed automatically by aligning the images and calculating tie points. To perform the exterior orientation and assign ground control points (GCPs), we used coordinates taken from stable, prominent topographic peaks identified in the 1979 photogrammetric model (Dvigalo, 1988; Dvigalo, 2000), which is referenced to the USSR State Geodetic Network coordinate system (Chumachenko, 1966); the same technique was employed to georeference the previous photogrammetric dataset of Shiveluch volcano (Shevchenko et al., 2015). The total RMSEs of the aerial models orientation vary from 1.5 to 2 m (Supplementary Table S1). As a result of processing, we obtained three aerial PCs with an average resolution of 2 m; these PCs were then filtered and cleaned in the same way described above. The gaps caused by volcanic steam emissions and by atmospheric clouds were filled in by collecting manual points using the anaglyph stereo mode of Photomod 5, which was performed by placing a floating mark on the visible surface and storing the XYZ coordinates of each point similar to (Schilling et al., 2008; Shevchenko et al., 2020).

The aerial PCs had the same spatial scale as the Pleiades PC but were geopositionally shifted due to the different coordinate systems. Hence, to compare the PCs, we aligned the aerial PCs to the Pleiades PC with several points along the rim of the amphitheater using the CloudCompare alignment tool. The RMSEs of the alignment vary from 2.3 to 3.1 m (Supplementary Table S1). Thus, we obtained four stacked PCs in WGS84 UTM57 and were able to calculate the differences between them. We outlined the areas of the lava dome, including the talus, separately for each date and calculated the volumes within these specific areas between two consecutive PCs. The total volumes of the dome were calculated relative to the 1979 DEM with the topography of the amphitheater floor prior to the dome growth (Dvigalo, 1988; Shevchenko et al., 2015). The time-averaged growth rates were obtained by dividing each volume difference by the time interval between the two corresponding consecutive acquisitions.

To estimate the uncertainties in the calculated volumes, we distributed the RMSE for each acquisition date (Supplementary Table S1) over the corresponding area of the dome (Table 1). For the volume derived from the Pleiades DEM, we used the RMSE of the photogrammetric model orientation; for the volumes derived from the aerial DEMs, we used the total RMSEs, which includes the orientation errors and the errors of alignment to the Pleiades PC. The maximum volume error is 12 × 10⁶ m³ on August 22, 2019, which is 1.6% of the total dome volume (Table 1). The uncertainties in the growth rates were calculated by dividing the volume errors by the time interval between two neighboring dates. The maximum growth rate error amounts to 31,000 m³/day (0.36 m³/s) or 8.5% for the August 22, 2019 data (Table 1; Supplementary Table S1).

Tasking the TerraSAR-X (TSX) satellite allows us to select a set of 6 ascending and 26 descending amplitude images of Shiveluch volcano from December 5, 2018 to December 25, 2019. The data were acquired in spotlight mode, which allows the redistribution of the dome material to be monitored at a ∼1-m spatial resolution with a temporal resolution reaching 11 days (for descending data). The SAR data were then analyzed in the following ways:

(a) First, we visually described the amplitude imagery used for interpretation in radar coordinates similar to the method previously applied to other volcanoes (e.g., Arnold et al., 2019). For this purpose, we mainly used the descending data, as their acquisition was more regular and the viewing geometry was more favorable with fewer foreshortening and shadowing effects than the ascending data. We loaded all available data into a coregistered stack using ENVI SARScape v. 5.5.3 and applied a 3 × 3 kernel filter to reduce the speckle effect and visualize the reflectance changes observed in the imagery.

Satellite-acquired thermal infrared data were also used in this study and were validated during our fieldwork overflights using handheld cameras.

(a) The satellite images used for this study were acquired by different infrared satellite platforms, including Moderate Resolution Imaging Spectroradiometer (MODIS) images from the Terra and Aqua satellites, Multispectral Instrument (MSI) images from Sentinel-2 (S2), and Landsat 8 (L8) Operational Land Imager (OLI) images (Coppola et al., 2020; Plank et al., 2021). These products are complementary, as, for instance, MODIS acquires images with a near-daily temporal resolution at a low spatial resolution of 1 km, whereas the S2 and L8 sensors provide data at spatial resolutions of 20 and 30 m, respectively, although they have temporal resolutions of only 2–5 days and 8–16 days, respectively (Ramsey and Harris, 2013; Coppola et al., 2016). The MODIS data were processed by the MIROVA system in the mid-infrared (3.9 µm) region, providing the VRP (in watts) emitted by volcanic activity at magmatic temperatures (T > 500 K) with an average error of ±30% (Coppola et al., 2016). MODIS infrared data have a 1 km spatial resolution and a high revisit frequency (up to four times daily) for volcanic monitoring purposes. High-spatial resolution thermal satellite datasets, including S2 and L8, have been investigated by applying a novel hotspot detection algorithm based on fixed ratios in the shortwave infrared (SWIR) regions with a contextual threshold derived from a statistical distribution of hotspot pixel clusters (Massimetti et al., 2020), as SWIR signals, in particular, record almost purely thermal emissions produced by hot emitting bodies (Blackett, 2017). Images were analyzed considering the SWIR top-of-atmosphere (TOA) reflectances in the ρ12 (2.19 µm), ρ11 (1.61 µm), and ρ8A (0.86 µm) bands for the S2 MSI and the ρ7 (2.11–2.29 µm), ρ6 (1.57–1.65 µm), and ρ5 (0.85–0.88 µm) bands for the L8 OLI. The algorithm works on both imagery datasets and detects the number of “hot” pixels (S2-L8 number of pixels), where a hotter area is superficially exposed, with an overall estimate of 2–4% false alerts detected (Massimetti et al., 2020). The reliability of the applied algorithm has already been successfully verified on a variety of different volcanological thermal-emitting phenomena worldwide (Valade et al., 2019; Massimetti et al., 2020). The combined thermal datasets allow major hot volcanic thermal processes and their effects to be tracked, particularly those in a complex volcanic context such as Shiveluch volcano with a variety of volcanic products and styles.

For a base map, we used a photogrammetric dataset that was acquired on July 12, 2012 and has already been described earlier (Shevchenko et al., 2015). This base map shows the collapse amphitheater resulting from the 1964 collapse, the central dome that grew inside the amphitheater, and two smaller scars that formed in 2005 and 2010 (highlighted in Figure 2A). Two lava lobes can be identified on the dome: the northern lobe that partially overlaps the NW rim of the 2010 scar and the SSE lobe with a crater on top (see Figure 3 and the details in Shevchenko et al., 2015). Reprocessing the data at a higher resolution allowed us to identify a SW–NE-striking lineament with an azimuth of 027°N (dashed line in Figure 3A). This lineament is also traceable through complex topography; because it notably bisects the lava lobes, the crater, and the talus in the NE sector of the amphitheater and is visible in both the optical and the topographic data, this lineament may represent an important structural feature (Figures 2A, 3A).

The DEM that we constructed from the July 18, 2018 Pleiades dataset (Figures 2B, 3B) shows the morphology of the volcano several months prior to the beginning of the studied extrusive eruption (which commenced in December 2018). The western to SW sector of the dome currently accommodates a new collapse scar (800 m × 2,000 m) originating from the September 18, 2016 partial collapse. This scar outcrops the core parts of the previously formed lava lobes and narrows towards the SSW into an avalanche chute within the borders of the 2005 scar. Another new, smaller scar (370 m × 700 m) can be observed on the NE flank of the dome, where it narrows into a chute following the eastern rim of the amphitheater. At the summit of the dome, a central crater with a diameter of approximately 350 m is identified (red outline in Figure 2B). The 2012 lava lobe on the SSE flank persists without any major morphological changes; it was destroyed only at the very summit due to the formation of the scar and crater. The main accumulation of material is detected in the northern sector of the dome near the former amphitheater rim (Figures 3B, 4A). A peak-like remnant of the northern part of the dome, which was destroyed by the two collapses (blue outlines in Figures 2B, 3B), had an elevation of 2,757 m and almost reached the highest point of Young Shiveluch, the Fourth Summit (Chetvertaya Vershina), which is 2,782 m in height. The abovementioned SW–NE-striking structural lineament is also observed here, where it bisects the peak-like remnant and the northern half of the central crater (Figure 3B). The relative height of the dome was 613 m. By July 2018, the northern and NE rims of the amphitheater had been almost completely buried under new unconsolidated material. The 2010 collapse scar at the SE foot of the dome was also filled with new material. By comparing the DEMs from 2012 to 2018, we estimate that the dome volume increased by 16 × 10⁶ m³, with a total dome volume of approximately 646 × 10⁶ m³ (Figure 4A). Hence, the average growth rate over the 2012–2018 period was 7,200 m³/day (0.08 m³/s), and further details are provided in Table 1.

During the December 2018 – August 2019 period, a new extrusive eruption emplaced a massive lava lobe (see Chronology Derived From the TSX Amplitude Images). By August 2019, this lava lobe occupied the entire central part of the dome and covered its western flank (Figures 2C, 3C) with dimensions of 700 m × 1,080 m and a volume (as determined by DEM differencing) of 51 × 10⁶ m³. The lava lobe might be the largest extrusive body that emerged on the dome over the current period of regrowth (since 2001). Furthermore, the data allow the recognition of two new craters: one in the NE sector of the lobe (approximately 180 m in diameter) and another along the northern rim of the amphitheater (130 m × 200 m). During the observation period (July 2018 – August 2019), the thickness of added loose material along the 2005 (western) collapse scar reached 90 and 100 m within the 2010 (eastern) scar. We also estimated the volumes of added material since July 2018 to be approximately 19 × 10⁶ m³ in the 2005 scar and 33 × 10⁶ m³ in the 2010 scar. The height of the dome remained almost the same at 611 m, as its highest northern part did not exhibit any morphological changes that can be identified in the DEMs. The dome volume increased by 146 × 10⁶ m³ after July 2018, which includes the talus and deposits within the two collapse scars (Figure 4B). Ultimately, the total dome volume on August 22, 2019 was 792 × 10⁶ m³. Thus, the average extrusion rate over 1 year was 365,000 m³/day (4.22 m³/s).

On August 29, 2019, a large explosive eruption occurred at Shiveluch, which led to a secondary partial dome collapse (see Chronology Derived From the TSX Amplitude Images). Analysis of the August 2019 (Figure 2C) and October 2019 (Figure 2D) aerial orthophotos allows us to describe details of the morphological changes that occurred in association with this eruption. A new large central summit crater is identified in the October 2019 dataset with dimensions of 430 m × 490 m, and a new extrusion is observed inside this crater with a blocky appearance and dimensions of 220 m × 250 m. At the northern part of the dome, another large crater 300 m in diameter can be recognized. We identify four new small side craters (Figure 3D) with diameters varying from 45 to 90 m. In the eastern and NE parts of the dome, a newly formed scar is visible, which narrows into an avalanche chute in the south; the size of this scar is 600 m × 1,100 m, and its depth varies from shallow (up to 50 m) in its southern sector to deep (up to 270 m) in its NE sector. The western (2005) and eastern (2010) scars also appear to have lost material (Figure 4C). The October 2019 dataset suggests that after the August 29, 2019 eruption, the western scar became deeper by 15 m, while the eastern scar deepened by 40 m on average. The volume within the western scar decreased by approximately 5 × 10⁶ m³, while that within the eastern scar decreased by 84 × 10⁶ m³. The height of the dome decreased by 26 m due to the partial destruction of the northern peak and the formation of the northern crater, yielding a new height of 585 m. The total volume of the dome decreased by 110 × 10⁶ m³ (including the scars); thus, the total dome volume became 682 × 10⁶ m³.

The August 29, 2019 eruption produced a PDC that traveled 12.3 km towards the SE, and the majority of the volume was deposited on top of the 2010 PDC deposits, albeit with a shorter run-out distance. A visual inspection and measurements of the Planet satellite image (Figure 1A) suggest that the August 29, 2019 PDC was deposited in several scattered fields along its path over a total area of 12 km². A small amount of material also descended along the western (2005) scar and was deposited at the SW foot over an area of 0.5 km², but this deposition was not captured during our aerial survey. To further analyze the main August 2019 deposition field, we inspected the aerial orthophoto and performed a comparison of the optical DEMs (Figure 5). Material deposition occurred within an area of 7.2 km² (Figure 5A), reaching a maximum thickness of almost 30 m and an average thickness of approximately 12 m (Figure 5B). We estimate that the volume of the main deposits was 87 × 10⁶ m³, which is 23 × 10⁶ m³ lower than the volume estimated to have been removed from the source edifice. The differences between the volumes of the source material and the deposits can be explained by the scattered regions of deposition in other areas not captured during the aerial survey, by the lack of dense rock equivalent (DRE) measurements, and by the dispersion of fine particles to more distant locations. A geometric analysis of the deposition area implies that this was the third-largest PDC that occurred during the current period (since 2001), after the 2005 and 2010 deposits, which extended over distances of 19 and 16 km, respectively, and spanned areas of 25 km² (Ramsey et al., 2012) and 27 km² (Dvigalo et al., 2011).

Topographic profiles help to further visualize the effects of the constructive and destructive processes at Shiveluch volcano over the studied period (Figure 6). The 2018 profiles show significant enlargement of the edifice in the northern part of the dome relative to 2012 but, in contrast, the loss of material in the central and western parts due to the 2016 collapse and formation of the summit crater (Figures 4A, 6B,C). By August 22, 2019, the edifice height had increased with respect to that in 2012 in the central part and notably surpassed the 2018 height in the eastern sector of the dome (Figures 4B, 6B,C). Because of the substantial loss of material due to the August 29, 2019 eruption, the central dome height decreased even in comparison with the 2012 height; relative to the 2019 pre-eruption morphology, this decrease in height exceeded 250 m in the eastern part of the dome (Figures 4C,D, 6B,C). The accumulation of loose material alternated with its removal within the western (2005) and eastern (2010) scars (Figures 6D,E). In 2018, there was a significant loss of material in the western scar due to the 2016 collapse; then, new material was added in 2019 and eroded again during the 2019 explosive eruption and collapse (Figures 4A–C, 6D). In the eastern scar, deposits accumulated significantly after 2012 but were subsequently removed due to the 2019 destructive events (Figures 4A–C, 6E).

One of the main distinctions between the 2019 and previous destructive events at Shiveluch is the preservation of the southern flank of the dome. The southern flank was partially destroyed during the 2005 eruption and completely destroyed during the 2010 collapse but appeared stable during the August 29, 2019 eruption. This preservation could be explained by the presence of the extensive lava lobe that formed in 2012, which remained stable on the flank for more than 7 years. We speculate that this lobe could have acted like armor blanketing the flank and preventing the collapse of unconsolidated material. For example, loading, gradual cooling, and degassing will close pore spaces and increase the compressive and tensile strengths of the lobe and dome rocks over time (Zorn et al., 2018). Additionally, hydrothermal alteration can significantly change the strength of volcanic rocks depending on the type of alteration (Heap et al., 2021). In most studied cases, hydrothermal alteration weakened volcanic rocks (Wyering et al., 2014; Ball et al., 2015; Mordensky et al., 2019) or even induced explosive behavior (Heap et al., 2019), mainly through changes in the mineral content, porosity and permeability of the rocks. However, volcanic rocks also exhibit increased strength under certain conditions through the deposition of minerals in hydrothermal areas (Heap et al., 2020). Here, it is not clear whether the dome or lobe were subject to either type of alteration, but since this lobe was relatively stationary and stable for several years, it would have likely experienced substantial cooling and degassing while also being subject to nearly constant hydrothermal alteration. Theoretically, these combined effects could have significantly increased the rock strength of the lobe. We also note similarities with Bezymianny volcano, on which the slopes of the lava dome were protected with lava flows that promoted stability and led to stratocone development (Shevchenko et al., 2020). Similarly, the western flank of Shiveluch, covered by the 2019 lava lobe, remained intact during the August 29, 2019 explosive event and partial destruction of the dome, which may indicate that the stabilization effect produced by lava lobes is significant for other episodes.

Another notable distinction from the previous activity is the never-before-seen formation of multiple small craters along the northern and NE peripheries of the dome, either at or near the rim of the former amphitheater (Figures 3C,D). The first of these craters appeared in mid-April 2019 (Figure 7H) at the northern rim of the amphitheater, which was covered by new material. The half-circular area of darkening in the region where this crater formed became noticeable in late January 2019, resembling the melting of perennial snow preserved by loose volcanic deposits coverage (see Chronology Derived From the TSX Amplitude Images and Figure 7E). The second peripheral crater appeared at the NE rim of the amphitheater after the August 29, 2019 eruption (Figure 7K) and was preceded by the appearance of a dark spot (Figure 7J). Furthermore, clusters of craters originated within the new 2019 collapse scar, probably due to the contact of the incandescent outcropped dome core with meteoric water. As these craters appeared suddenly and in association with only short eruptive/explosive activity, they may be considered monogenetic. Our careful analysis of available photogrammetric data does not reveal an extrusion or emission of large volumes of materials from these new crater sites. This finding, together with the observation of high thermal emissions, suggests that the origin of these craters could be phreatomagmatic.

Additionally, the locations of the new craters, either elongated along the SW–NE structural trend or along the former amphitheater rim (Figures 3C,D, 7H,I,K), argue for a structural control, as further discussed below.

Craters that form during an explosion (Figures 3C, 7G) due to degassing or an interaction with meteoric water are typical on Shiveluch and have been previously recognized on lava lobes (Shevchenko et al., 2015) but never along or near the amphitheater rim. The central crater caused by the August 29, 2019 eruption is similar to the 2005 crater (∼750 m in diameter) but with smaller dimensions (430 m × 490 m), probably due to the smaller magnitude of the explosion.

We find that some of the explosion craters are thermally well expressed, while others appear cold. Some craters, such as the central summit crater, are characterized by thermal anomalies around the rim only but not on the crater floor. A cold crater floor might have different explanations, both volcanological and observational: the extrusion of cold crystalline material (Mania et al., 2019), the surface crust of cool lava masking hotter material beneath (Sahetapy-Engel and Harris, 2008), the occurrence of a phreatic explosion not followed by the extrusion of lava (Barberi et al., 1992), low thermal activity due to hydrothermal sealing or the tectonic blocking of fluid paths (Matthews et al., 1997; Laiolo et al., 2017), or even satellite techniques that are too inaccurate to detect anomalies with excessively inclined observation geometries or sensors that have an inadequate spatial resolution (Coppola et al., 2016; Blackett, 2017). One more reason that cannot be entirely excluded in this context is that the thermal infrared signal was partially covered and thus biased by the steam of whitish vapor (Hashimoto et al., 2018); however, these craters did not show any strong visible steaming. As we observed cold crater floors in multiple S2 acquisitions and our helicopter overflight, we are confident that the low temperatures in these craters are real. Therefore, we can also assume that the extrusion of new blocky material observed in this crater that started in September shortly after the eruption could represent the extrusion of cold debris. Indeed, the crater that developed off-center and farther to the north exhibited one of the highest thermal anomalies in the study area during the period of observation (Figures 10J, 11B).

The growth rate calculated for the period from July 2018 to August 2019 was 365,000 m³/day (4.22 m³/s), which is significantly higher than the long-term average growth rate over the 2001–2012 period (225,000  m³/day or 2.6 m³/s) but comparable to that over the 2001–2003 period (320,000 m³/day or 3.7 m³/s) (Shevchenko et al., 2015). These growth rates are also higher than the average growth rates at other dome-building volcanoes: 26,400  m³/day (0.31 m³/s) at Bezymianny in 1956–2017 (Shevchenko et al., 2020), 37,000 m³/day (0.43 m³/s) at Santa Maria, Guatemala, in 1922–2000 (Ebmeer et al., 2012), and 173,000  m³/day (2 m³/s) at Mount St. Helens in 2004–2005 (Schilling et al., 2008). At Shiveluch, we further identified particularly strong activity during the first month of eruptive activity coincident with the extrusion of the lava lobe. While no discrete growth rates for the whole dome could be measured, only loose material accumulated within the scars during the first month of eruptive activity, with the deposition rate reaching 670,000 m³/day (7.7 m³/s). Thus, the growth rates during the construction period significantly exceed the average rates presented here, illustrating that this volcano is currently very productive compared to its own previous episodes and other volcanoes.

The SW–NE lineament is inferred from 1) a pronounced linear kink in the slope during 2012, 2) structural features bisecting the peak-like remnant in 2018, 3) aligned explosion (monogenetic) craters that evolved during 2018–2019, and 4) thermal anisotropy. Moreover, we conjecture that the collapses (and especially the collapse directions) are largely governed by this structural feature. However, whether this feature has deeper roots and/or is the surface expression of a fault remains speculative. Accordingly, we hypothesize that this SW–NE lineament (Figure 12) could play an important role in the development of the dome and in its construction and destruction. The same SW–NE structural trend can be identified in optical and topographic data as early as 2012 (see Figures 2A, 3A) and was hypothesized in a previous study based on monogenetic cones located in the surrounding areas and lower apron of the Shiveluch edifice (Koulakov et al., 2020). Nevertheless, this is the first evidence of the presence and relevance of the structural architecture at the summit of Shiveluch volcano governed by a SW–NE-trending structure and of the alleged intersection between the curved plane and amphitheater of the 1964 sector collapse.

The observed SW–NE-trending feature is identified on the Shiveluch dome only. We could not find any evidence for its continuation farther north, such as crossing the amphitheater headwall, implying that this lineament is either a local volcano-tectonic feature or a structure that is amplified and strongly expressed on the Young Shiveluch edifice only. Similar local tectonic features have been suggested to impact the structural architecture of lava domes at other volcanoes, such as Merapi (Walter et al., 2015), supporting the idea that regional faults may indeed affect dome growth and collapse processes.

If we assume a SW–NE-trending, nearly vertical fracture zone to be present, then we may expect a curved intersection arc at the base of Young Shiveluch along the former décollement surface (Figure 13). We note, however, that although multiple lines of evidence support such a structural trend, a critical view regarding the emergence and locations of explosion vents is warranted. We note that many craters are consistent with this lineament, such as the crater on the upper surface of the 2012 lava lobe (Figures 3A, 12A), the summit crater in the center of the dome in 2018 (Figure 3B), and three craters (the central and northern craters and the side crater on the NE amphitheater rim) that originated during the period of study (Figures 3D, 12B), which is why we suggest that this structural feature can affect the formation of these craters. However, neither the radar observations nor the pixel offset method could confirm whether the fracture slides as a strike-slip or dip-slip fault or is an opening fracture. This lack of fault displacement evidence might indicate that the activity of this structure is below the detection threshold, that the fault kinematics simply did not activate over the observation period, that the lineament is a fracture zone without clear kinematics, or that it is possibly associated with shear fracturing, as identified in experimental studies of rock properties simulating shallow depths (Heap et al., 2015). Nevertheless, an assessment of the crater locations might provide evidence for an elongated zone or for several parallel alignments striking SW–NE, possibly associated with a fracture zone that is wide rather than focused along a single fault. The thermal anomaly locations (Figures 10E–H; Supplementary Figure S6) are also in agreement with this lineament. Moreover, this structure is parallel to the 1964 collapse direction, to the lineaments of the northern groups of volcanoes described in (Koulakov et al., 2020), to the regional tectonic structures (Kozhurin et al., 2006), and to the Kuril-Kamchatka volcanic arc in general. However, the recent partial dome collapse directions are approximately perpendicular (east-west) to the lineament in the upper sectors of the collapse scars and oblique (SE and SW) to the lineament in the lower sectors (Figure 12B). According to (Tibaldi et al., 2008), the tectonics of a volcano substrate influence the evolution of its edifice, and normal and strike-slip faults propagating through a volcano induce its instability in the directions perpendicular and oblique to the fault strike, respectively.

The complex interrelation and even bidirectional interaction of events responsible for constructing and destroying the volcano edifice have been identified for volcanoes worldwide, both on land (Germa et al., 2015) and in the ocean (Sibrant et al., 2014). The relevant observations range from local-scale lava dome growth and collapse (Kelfoun et al., 2021) to large-scale ocean islands, where collapse-related unloading even affects deep crustal magma reservoirs (Manconi et al., 2009). The details of such growth and collapse, however, may be strongly dependent on the local site conditions. At Merapi, it was recently identified that the mechanism of collapse is controlled by the basement and dome slopes, which also control the directions of destruction episodes (Kelfoun et al., 2021). At Shiveluch, when considering the outlines and directions of the western (originating in 2005) and eastern (originating in 2010) collapse scars, we find that these azimuthal directions have a certain longevity. These two scars serve as accumulation centers and erosion pathways; for example, they accumulate deposits from small explosive eruptions and gradual collapses or disintegration of the lava lobes and then repeatedly discharge materials during large destructive events (see Deposit Accumulation and Discharge Within the 2005 and 2010 Collapse Scars). These unconsolidated deposits are rather weak and unstable compared to the lava material composing the main dome edifice; thus, the repeated discharge of loose material from these scars, which can occur due to gravitational processes or earthquakes, can be hazardous even if the dome flanks remain stable.

A previous study (Shevchenko et al., 2015) showed that the two major collapses at Shiveluch in 2005 and 2010 were preceded by periods of endogenous growth, while the dome remained stable during the extrusions of lava lobes due to the constant discharge of the new material. Collapses induced by intrusions and faulting activity at volcanoes have been debated based on experimental simulations (Acocella, 2005) and numerical modeling (Massaro et al., 2020), and evidence that intrusions such as plugs, dikes, and sills may ultimately cause complete or partial flank collapse is growing (Giampiccolo et al., 2020). Conversely, flank collapses may affect the locations and directions of intrusions (Maccaferri et al., 2017), so a complex interplay is likely (Lénat et al., 2012). This idea, together with the results of our remote sensing data analysis, led us to suggest the following conceptual model for the constructive and destructive events of Shiveluch (Figure 13). After intense extrusion (Figure 13A), a massive lava lobe covered the entire central part and SW flank of the dome. The lateral expansion and deformation of the eastern flank (see Lava Dome Deformations Revealed by PIV) started after the extrusion had slowed down, which could be evidence for blockage of the main vent by the massive solidified lava lobe on top of the dome that prevented the discharge of fresh eruptive material. As magma intruded the core of the dome, the flanks deformed and steepened, and several events of local instability occurred (slumping, hot avalanches, etc.), exposing hot materials and generating thermal anomalies. We hypothesize that, during this stage, when magma did not reach the surface, it began to propagate along the SW-NE fracture system, along the base of Young Shiveluch, or along the intersection between the fracture system and the decollement. As magma approached the surface to the north, close to the amphitheater rim, it could come into contact with meteoric water associated with rainfall and/or snowmelt (Figure 13B). Consequently, short-lived explosions formed craters aligned SW–NE. Further, the inability to discharge magma through extrusion led to a pressure buildup that could have caused the large August 29, 2019 explosive eruption and the formation of the central crater, which in turn led to the collapse of the eastern flank; another side crater appeared along the fault lineament after the eruption (Figure 13C).

The 2010 collapse was caused by dome instability due to gravitational processes and was followed by a relatively small decompression explosion. This is evidenced by the formation of the large and deep collapse scar that exposed the core of the dome, thereby revealing a small explosion crater at the top (Dvigalo et al., 2011; Shevchenko and Svirid, 2014; Shevchenko et al., 2015). In contrast, the 2005 collapse was secondary and was caused by a large explosive eruption that formed a 750-m diameter crater, which opened into the large collapse scar (Ramsey et al., 2012; Shevchenko and Svirid, 2014; Shevchenko et al., 2015). The formation of the large crater on top of the dome during the August 29, 2019 eruption suggests that the sequence of events (primary explosion and secondary collapse) is similar to the sequence of the 2005 destruction episode, though the collapse direction and deposition of the main part of the PDC coincide with the 2010 collapse.

The development of Shiveluch volcano is controlled through the interaction of constructive and destructive processes. Destruction follows construction when the dome height reaches a critical elevation and its flanks become oversteepened or when massive extrusive bodies block the vent and endogenous growth destabilizes its structure. Then, construction continues after destruction when the vent opens again, allowing new material to freely extrude onto the surface. Further modeling might help to better understand the details of these relationships, which are governed by stress changes, growing and changing masses and slopes, and material heterogeneities, such as fracture lines and former decollement surfaces. We speculate that if construction processes prevail at Shiveluch, they might lead to the formation of a new stratocone, as happened at Bezymianny (Shevchenko et al., 2020). The prerequisite for this phenomenon has already occurred, i.e., the 2012 lava lobe that stabilized the southern flank. Moreover, as was identified at Bezymianny, the establishment of a major and centralized summit crater, such as the Shiveluch central crater since 2018 (Figures 3B,D), might indicate the stabilization of a vent and the formation of a stratocone (Shevchenko et al., 2020). At Shiveluch, however, the relevance of the SW–NE structural trend and its intersection with the base of the edifice are of major importance and may govern future episodes of construction and destruction.