We use high-resolution visible range imagery, which we acquired during a series of UAS surveys in August 2019 over the Eastern Askja caldera for the purpose of mapping different geomaterials by means of RGB-color distinction and compare these to the results obtained from very high resolution (VHR) satellite images. To better characterize the different rock types and to validate our classification of different rock units from drone imagery, these surveys were further complemented by ground-based hyperspectral measurements on representative rock samples in the field. In the following we describe 1) the VHR satellite data and then in detail how we acquired 2) our imagery data from the drone camera and 3) the narrow-band spectra collected by means of the field spectroradiometer.

We applied processing steps to prepare the VHR satellite images and our high-resolution drone imagery for PCA analysis, and the latter data additionally were used in order to obtain a classification map distinguishing between different surface materials found in the field (Figure 2).

VHR satellite scenes acquired over the Askja edifice reveal a talus-covered inner Eastern caldera flank. Images acquired prior to and following the 2014 landslide (Figures 3A–C) allow to reconstruct the major changes in consequence of the event. The 2014 landslide covered a large portion of the talus-covered flank by mostly light-gray material. These are redeposited pyroclastic deposits of the 1875 eruption, which are particularly widespread in the NE section of the caldera (compare areal extents in Figure 1C and Figures 3A–C). Our field visit reveals that most of these pyroclastic deposits consist of unusually vesicular and crystal-poor white rhyolitic pumices, which in the field near the source appear colorized yellow to orange, while the remainder consists of pale to dark-gray pumice, which further have been distinguished into three different types according to their composition: (1) dark steel-gray rhyolite pumice, (2) pale to dark-gray brownish pumice consisting of varying amounts of rhyolite and basalt glass, and (3) pale gray crystal-rich basaltic andesite pumice, mainly found as enclaves within the white pumice. Since the pyroclastic deposits are commonly heterogeneously colored and individual pumice clasts are max. 1 m large, the colors of these deposits as they are resolved by satellite merge into different shades of bright-gray and brown-gray, however.

The two small 1920s basaltic lava flows north and south of the landslide (see outlines in Figure 3A) appear in dark-gray to black colors in the VHR satellite image. This Virtually resembles the talus of a third 1920s basaltic lava flow that is situated N of the study area, which covers much of the lower caldera terraces at the N margin of the study area. In addition, the RGB data resolve the rhyolite lava body of unknown age situated in the SE corner of the caldera (outlined in Figure 3A). The body located right within the source region of the landslide is dominated by yellow-brownish rhyolites showing signs of intense hydrothermal alteration (Figure 3A). This body is attached to and conformably overlain by the subglacial deposits (Graettinger, 2012; Figure 1C) forming the part of the caldera wall that later collapsed, which mainly consisted of similarly yellow-brownish colored hydrothermally affected hyaloclastite and pillow lavas of the Austurfjöll massif. A pinkish-white spot on the northern slopes of that dome indicates the location of a hydrothermal replacement deposit (clays and sulfates) is visible in satellite images that formed at a likely still active fumarolic vent. Notably, we find several eruptive vents and active hydrothermal vent sites within and surrounding the landslide area (Figure 3A), and a large cluster of fumarolic vent sites, which is appearing in the satellite image as reddish-brown spots with bleached margins. The fumarole sites are located on the NE slopes of the caldera wall (Figure 3A). The chemically altered and reworked material of alluvial fans found in morphological depressions and drainage channels, e.g., south of the northern rhyolite lava body, and the talus deposited along the steep cliff in the NE part of the caldera is characterized by similar red-brown to yellowish-brown hues, implying active or inactive hydrothermally active sites.

The 2014 landslide mobilized large parts of the rhyolitic lava body in the caldera wall together with the overlying basaltic hyaloclastites and pillow lavas. The landslide has emplaced blocks of up to 80 m in diameter on the slopes of the unroofed dome, split up into several different streams further downslope, and near the lakeshore additionally partly buried two of the 1920s basaltic lava flows, Kvíslahraun to the North and Suðurbotnahraun to the South, before it finally reached the Öskjuvatn lake (Figure 3B). We measure the onshore portion of the landslide deposition that covers an area of ∼0.8 km². Previous to the landslide event we measure that 27% of the onshore depositional area were covered by bluish black basaltic lava flows and their talus, 33% by yellow-brown altered rocks, most of which originate from the rhyolite lava body, and 40% by dark-brown to bluish black and greenish gray undifferentiated mostly basaltic talus material of the Austurfjöll (compare Figures 3A, B, features compiled in Supplementary Figure S3).

The central material stream of the slide was emplaced on the rather flat surface of the Suðurbotnahraun lava flow, where it built up an ∼250 m wide horn-shaped coastal block. Which initially extended ∼100 m into the lake, at the same time eroding the coastline by some 10 m forming embayments at its northern and southern margins (indicated by white arrows Figure 3B). The Northern half of the subaerial landslide deposition area is dominated by a bilobate central blocky stream of yellow-brown hydrothermally altered rhyolite and hyaloclastite material, from which ∼50 m wide secondary branches of finer yellow-brown debris material split off, spreading towards North. In the southern half, in contrast, we find less altered and less blocky materials (light-gray rhyolitic lava, and greenish-gray hyaloclastite and pillow lavas), mostly in the upper half of the slope, while at the coast merely little debris was deposited (also compare with Eliasson and Saemundson, 2021) and thus the black basaltic lavas of the Suðurbotnahraun flow shine through the debris cover in large parts of that area. The material stream in the center of the landslide area, which formed above the mentioned block of the coastline, seems to have been the preferential path of unaltered light-gray rhyolitic lava, which was exhumed at the southern margin of the hydrothermally affected rhyolite dome.

Comparing coastlines in the satellite images of 2012 and 2014 shows that the shoreline in our study area significantly retreated as a consequence of the sliding event and tsunami resulting in loss of 32,800 m² land along this ∼3.5 km long portion of the coast. Satellite images acquired in the months following the landslide further show a pronounced turbidity and turquoise colorization of the usually dark-blue and clear water in the caldera lake, which lasted for several months following the landslide event.

In August 2019, the month of our field campaign, the waterbody had regained its typical dark-blue color and was completely clear, enabling us to see the submerged parts of the steep coastal cliff (Figure 3C). Furthermore, the shoreline advanced again following the landslide event in 2014, adding an area of 29,074 m² between 2014 and 2019, i.e., almost as much land surface was regained as was previously lost in the study area. Particularly within the landslide area significant amounts of additional debris were deposited along the coast leading to infilling of previously formed embayments and further protrusion of the coastline up to ∼10 m, thus hinting at intense erosion of the landslide surface and formation of downwash.

Features can be highlighted in aerial or satellite imagery by applying a PCA to the bands of interest. Here, we first will take a look at the results of a PCA performed on the RGB bands of the three natural color VHR satellite images displayed in Figure 2. In the RGB composite (Figure 4) we can observe different features in false-color that can be associated with fumarolic areas, snow (in green), and a variety of different volcanic rocks. It has to be noted that the false-colors of some of these PCA features vary slightly in between the scenes, partly due to the redeposition or alteration of materials and partly due to the different light conditions that were encountered during the acquisition of the three scenes. The yellow-brownish to red-brownish colors associated with altered materials of the landslide and debris fans in the natural color satellite image, however, are commonly highlighted by PCA in false-color magenta to purple colors and light-gray to whitish materials commonly appear in a false-color dark blue. The whitish hydrothermal replacement deposit in the fumarolic area in the northern part of the study area appears in PCA in blue hues surrounded by a mixture of light bluish and pinkish colors. Light-green colored areas are associated with snow deposits except for in the scene of September 2014, in which additionally a stream of very light-gray rhyolite extending from the central part of the landslide to the lake has this PCA color. In the scene of August 2019, the same rhyolite stream appears in rather dark blue colors, despite having the same light-gray color in the satellite image, and the same is true for the dirty-gray snow cover, e.g., in the landslide area. Furthermore, shadowed areas are similarly colored in orange hues as the basaltic lava flows thus posing a potential problem for distinction by color. Especially the Southern limit of the landslide area and its source region in satellite images commonly is obscured by a large shadowed area extending along the southern caldera wall which is particularly pronounced in the scenes of 2014 and 2019 (Figures 4B, C).

We note that already in July 2012 (Figure 4A), prior to the occurrence of the landslide the upper slope of the area is covered with hydrothermally affected materials appearing in purple hues side by side with less altered volcanic materials appearing in orange to orange-green. Furthermore, we can note the redistribution and areal increase of such purple materials that are partly mingled with less altered surrounding materials in the two satellite scenes acquired after the landslide (Figures 4B, C). Moreover, in the scene of 2019, we observe an increase in intensity of such purple features also along the steep slopes despite there is hardly any morphologic change visible in the northern part of the study area.

Figure 4 shows that the spatial distribution of different surface materials has drastically changed at the location of the landslide before and after the event. Outside the landslide, a similar spatial distribution of materials is observed in the 3 images: light green areas that are interpreted to be snow covers, and purple areas that are spatially associated with altered materials. On the other hand, in the Southern part of the images, the distribution of materials has changed before and after the event. Outcrops of altered materials that were not exposed before the event became exposed after the event in the landslide areas. The different PCA colors, therefore represent the spatial distribution of materials that are similar.

In order to obtain more detailed information on the study site, we analyzed the high-resolution imagery recorded during the UAS overflights of 28 August 2019 (Figure 5A). The orthophoto mosaic and the DEM that we obtained from our drone imagery (Figures 5A, B) provide highly detailed information on the spatial distribution of different types of materials (rocks, snow, vegetation, waterbody) at the time of our survey, and of different morphological structures (cliffs, drainage gullies, and tectonic structures), which occur in the area. We resolve some of the different materials small-scale surface structures, such as flow-textures of lava and debris flows, or fracture networks and layering in bedrock and sedimentary deposits. Moreover, the data are almost free of large deeply shadowed areas as they appear in the VHR satellite images, and only some minor patchy and blurry shadows from clouds are visible due to the heterogeneously overcast skies we encountered during their acquisition. Some general large scale features can be observed, such as: the deep-blue water body and current shoreline of Öskjuvatn, a widespread light-gray pumice which covers large parts of the Northern portion of the eastern inner caldera flank, several blackish basaltic lava flows from previous eruptions and the entire subaerial portion of the landslide area with its central blocky rockslide and a number of secondary debris flows. Furthermore, we can see details such as the pinkish-to-whitish hydrothermal deposits and reddish-brown coated (probably Fe-oxide) rocks of two major actively degassing fumarole areas that are located on the inner slopes of the NE and SE caldera rim, the latter of which is located near the landslides source region (see Figure 5). Small batches of minor vegetation such as moss can be observed growing on the clay-rich substrates close to the hydrothermally active areas. Note, that the checkerboard pattern of interchangeably dark and bright blue on the water surface of Öskjuvatn is the result of an image (compression) artifact.

A slope map generated from the DEM of the surveyed area (Figure 5B) illustrates the flat surfaces of lava flows and of pumice filled depressions (green colors) and several oversteepened slopes (orange-red colors) along the shoreline of the lake and morphological edges demarking the Southeastern caldera rim and drainage channels within the caldera. A detailed view into the landslide area reveals a chaotic pattern of smooth plains, e.g., in the depletion zone and the coastal areas of the landslide, between rugged surfaces of debris flows and steep slopes with irregularly distributed and differently sized blocky materials. Particularly large blocks, which are up to 100 m in diameter, have been emplaced in the downslope portion of the collapsed rhyolite dome, which is situated in the zone of accumulation of the landslide.

We applied the PCA in order to reduce the dimensionality of the data and highlight the main spectral features that can be extracted from the image. Since the vibrant blue water body was masked out previous to analysis, our PCA focused on the reduced color palette of the less colorful volcanic materials in the subaerial portion of the orthomosaic.

Having a look at the grayscale image reveals that the values of PC1-PC3 are largely uncorrelated, and that different landscape features preferentially have been captured by one of the different components. PC1 preferentially captured the bright features of the orthomosaic image such as snowfields and hydrothermal mineral deposits, which in PC1 occur as sharply outlined dark features (Figure 6A). PC2 is dominated by intense dark features, which correspond to talus deposits and the redeposited material in the landslide area (Figure 6B). Furthermore, the hydrothermally active areas and eruptive vents appear as dark-gray spots in PC2, while basalts and snow patches are bright. PC3 generally looks rather noisy, but closer visual inspection reveals that the brightest spots as well demark the locations of hydrothermally active venting areas and rock surfaces with intense iron oxide coatings (Figure 6C).

The RGB composite of our PCA output (Figure 7A) shows a variety of color patches that correspond to different types of materials. Especially the mixed material of the landslide area, and the loose material of other areas that were affected by mass movements such as the talus accumulation along the cliff and the steep inner caldera wall in the NE part of the caldera, is strongly outlined in dark red to purple colors. Moreover, some color differences between pumice-covered areas (olive) in morphological lows, snowfields (bright green), and several lava flows (bright yellowish-green) which are distributed along the eastern lake shoreline are clearly visible. Particularly, the snow bodies seem to be captured perfectly. Furthermore, the two main fumarole areas in the NE and the SE of the caldera appear in blue-violet color and are easily detected with a simple PCA analysis. However, the landslide outline and full extent of the redeposited materials particularly along the southern landslide margin are not fully and precisely detected and still, there might be some other features to extract which so far after applying PCA, still are not well distinguished. Therefore, we conducted classification for further analysis of the PCA results in order to better visualize different features. Next to the PCA analysis, we see the classification output (Figure 7B) for detecting different materials in more detail, which we will explain and analyze in the next section.

We applied classification methods to the RGB composite of the PCA output obtained from our orthomosaic. We defined 7 general classes to be classified by our maximum-likelihood classifier, 6 of which are related to distinguished geological materials and structures. Defined classes and their color-coded visualizations are illustrated in Figure 7B, and they are as follows: 1. Whitish materials like snow (White class), 2. Yellow-brown Fe-oxide coated materials as found in the hydrothermally altered rhyolite dome, yellow-orange pumice and materials with a characteristic ochre stain including loose reworked debris depositions along steep slopes and materials covering the landslide area, but also minor light-green vegetation which preferentially occurs in these reworked areas (Red class), 3. Black to dark-gray mafic materials like, e.g., the basaltic lava flows, dark brown basaltic pumice (Dark Green class), 4. Light-gray to whitish materials as found in hydrothermally affected areas (bleached volcanic rocks, clay- and sulfate-rich hydrothermal replacement deposits and solfataric mineral coatings) or dirty snow with slight irregularities in color (Light Green class), 5. Brownish dark-gray materials, basalts, and rhyolites with minor Fe-oxide coating (Yellow class), 6. Red-brown materials, mostly oxidized basaltic lavas with reddish Fe-oxide coatings as found at volcanic vents, and in oxidation zones of hydrothermal vents (Light Blue class), 7. Light-gray to medium-gray materials, mostly white and pale-gray pumice, but also very dirty snow (Dark Blue class).

At the landslide area (Figure 7B) data reveal the transport paths of hydrothermally altered and Fe-oxide coated materials from the landslide source area. Several types of rock and material deposits can be seen in different areas all over the entire flank and specifically in steep areas and mainly in the landslide zone below the cliffs. The red class shows the spatial extent of Fe-oxide coated materials in debris fans and the landslide, at the northern margin of the landslide area, representing the landslide outline. The southern half of the landslide area is dominated by yellow class materials corresponding to brownish dark-gray materials, mainly basalts with minor alleged Fe-oxide coatings. The yellow class further is dissected by streaks of the light-blue class, which are elongated in sliding direction, and correspond to red-brown oxidized materials as found in surrounding eruptive fissures and vents.

We counted the number of pixels in each class and determined the approximate areas that each material covers in a) the entire subaerial part of the study area and b) the onshore portion of the landslide area (Table 1). Both areas are dominated by materials corresponding to classes 5 (undifferentiated brown gray talus) and 2 (hydrothermally affected ochre to red-brown talus), which in the entire area cover 37 and respectively 23%, and almost equal portions (each ∼36%) of unaltered and altered talus are found in the area of landslide deposition. The study area further contains significant portions of class 3 and 7 materials (17% and 13%) corresponding to black to dark-gray basaltic lava flows and medium gray rhyolite pumice, while these materials only constitute a minor fraction (4% and 6%) of the landslide area.

We closely compared the classes identified in our drone classification analyses to hyperspectral information (compare Figures 5, 7, 8). The rock samples were visually identified, and no chemical analysis was carried out. However, Van Der Meer (2004) describes that the position, shape, depth and width of the absorption features can be directly associated with the chemistry and structure of the rock sample. Thus, the hyperspectral information can provide selective, fast detailed information of specific minerals in rock samples without time-consuming chemical analysis.

We found that 6 of the 7 classes (class1=snow not considered here) were well sampled by hyperspectral measurements. The closer comparison shows that the information from classification analysis and from hyperspectral measurements are complimentary, rather than simply correlated.

The Askja landslide mass is best described by the classes 2 and 4, which we could sample with our hyperspectral device as well. The class 2 characterizes yellow-brown Fe-oxide coated materials in the hydrothermally altered rhyolite dome and yellow-orange pumice. The presence of iron oxide is also shown in the corresponding spectra for the class 2. Here, absorption features in the VNIR region are related to iron oxide due to electronic transition, bound on hydroxyl groups or other metal cations (Hunt, 1977; Gholizadeh et al., 2015). The absorption features around 1,400 nm are characteristic for the overtone of OH-groups in the hydroxyl-group or in clay minerals. In combination with the 1900 nm feature, the presence of water molecules is implied, which is likely due to wet sample surfaces (Hunt, 1977). The weak features around 2,200 nm are just shown in the rhyolitic pumice (D2-D3) and Fe-coated obsidian (D4) and are combination tones of the OH-group and the Al-OH bending mode found in clay minerals or mica (Hunt, 1977). In this view, the classification may indeed be a first order spatial characterization of hydrothermally altered rocks, which is relevant for slope stability measurements.

A closer look reveals, however, also important differences in the two data sets. Here the classification results show rather uniform classes 3 and 6. Class 3 is characterized by basaltic lava with a glassy crust or black pumice from basaltic lava flow or scoria. The spectra are rather featureless as is typical of volcanic glass (e.g., Henderson et al., 2021; Rader et al., 2022) with exception of weak spectral features around 500, 530, 630 and 900 nm in the VNIR region, which are related to iron oxides. The spectra in the SWIR region hardly show any absorption features, besides that of the pumiceous basalt. Here, additional weak absorption peaks are shown around 1,400 nm and 1900 nm due to the presence of hydrated silicate glass and water on the sampled surface. The class 6 is characterized by oxidized basaltic lava from flow or pillow breccia with reddish oxides coating. The reddish oxides coating is visible in the VNIR region (around 500 and 900 nm), which is related to the absorption features of iron oxides.

The class 5 also describes the landslide area and is characterized by brecciated hyaloclastite with healed Calcite-filled fractures and hyaloclastite or tuff with greenish altered minerals. The class 5 summarizes four samples with nearly similar spectral shapes (Figures 8B2, B3, C3), except for sample C2 (Figure 8C2). All spectra show absorption features around 440, 700 and 900–1,000 nm, which indicate a presence of ferrous iron (∼900–1,000 nm) and of ferric iron (∼440 nm). The samples B2, B3 and C3 also show absorption features around 700 nm, related to ferrous ions (Hunt, 1977). The features around 1,400 and 1900 nm in the SWIR region are water-related absorption features, whereby the absorption feature at 1,400 nm is less distinct in sample C2 and C3. Sample B3 and C2 also indicate absorptions features around 2,200 nm and 2,350 nm, which normally appear in pairs with the 1,400 nm band, due to combination of the hydroxyl-group and the presence of aluminum (2,200 nm) and magnesium (2,350 nm) (Hunt, 1977). Here, both samples may belong to hydroxyl-bearing minerals, such as hydrothermal clays or sulfates (Van der Meer et al., 2012). Comparing the absorption features of sample B2 and B3 to the results of Van der Meer et al. (2012), shows that they are related to a volcanogenic massive sulfide formation.

The class 7 is described by the samples B4 and B5. The samples are characterized by rhyolitic pumice and to some extent with mafic enclaves (B4). Here, the reflectance of both samples is very low, even the sample color is bright. The low reflectance values are caused by the wet samples surface or due to the stored water in the pores.

The Askja landslide occurred on an unstable slope. Specifically, the current understanding is that the stability and instability of volcanic slopes is highly variable, changing in time and space. Conceptually, three main drivers of slope instability can be considered, i) a too steep slope, ii) changes to the system by extrinsic triggers such as earthquakes, iii) a progressive intrinsic strength reduction of rocks. This is on one side, because volcanoes are made up by very different materials, such as the Askja caldera rim, including products from effusive and explosive eruptions and endogenous growth. Consequently, these materials encompass highly variable physical and mechanical properties (Heap and Violay, 2021).

In a traditional way of thinking, unconsolidated and weak deposits may provide the main sliding horizons that govern dimensions, dynamics, and directions of mass wasting events. Depicting these sliding horizons is essential, as this contributes to our understanding of the collapse process, which may have high societal interest, as partial flank collapses at some 200 volcanoes globally have caused over 20,000 fatalities (Rosas-Carbajal et al., 2016). At Askja, we could demonstrate that different rock types were involved in the landslide mass, whereby hydrothermally altered materials are especially widespread and detectable by our drone remote sensing techniques throughout the landslide deposits. The finding of hydrothermally altered rocks is of relevance for (i) the destabilization of the slope, and (ii) the dynamics of the mass movement. The alteration-induced destabilization is in line with new research elsewhere that suggests that hydrothermal alteration is strongly affecting the strength of volcanic rocks. Indeed, studies suggest that hydrothermal alteration weakens a volcanic slope, which can increase with time to promote sudden collapse (van Wyk de Vries et al., 2000; Watters et al., 2000; Voight et al., 2002; Heap and Violay, 2021). Case studies are found globally, where hydrothermally altered materials are integrated in debris-avalanche deposits resulting from partial flank collapse. For instance, the 26 December 1997 rockslide at Soufrière Hills volcano (Montserrat, Eastern Caribbean) showed a mixed (“varicolored”) and hydrothermally altered material integrated into the avalanche deposits, implying that these materials could have contributed to the collapse (Voight et al., 2002). Following a strong storm, a flank collapsed at Casita volcano (Nicaragua), where detailed field mapping revealed a mixed material that formed by hydrothermally altered materials that contributed to the instability and the collapse that killed about 2,500 people (van Wyk de Vries et al., 2000). Similarly, deposits from La Soufrière de Guadeloupe (Eastern Caribbean, France) also entrained hydrothermally altered rocks (Salaün et al., 2011). The answer for most of these documented collapses can be found in the mechanical changes associated with hydrothermal alteration, causing a change in porosity and a dramatic reduction of compressive and tensile strength values (Heap and Violay, 2021).

The hydrothermally altered rocks at Askja have also altered the dynamics of mass movement. Types of landslides are usually classified according to either the type of movement (falls, overturns, slides, spreads and flows) or the type of material (bedrock, coarse or fine soils), based on Varnes (1978); Cruden and Varnes (1996). On the steep slopes and sub-vertical cliffs we find large blocks, possibly rotated, indicating solid mass overturning and rockfalls. On the medium slopes towards the coast we find well mixed material, indicating unconsolidated material. Therefore, we find that the collapse involved characteristics of two types of materials, bedrock and weak material mass movements. The hydrothermally altered and weakened materials probably have an important role in this mobilization process, as they cause a high mobility and mixture of the heterogeneous rock types involved in the landslide. The finally deposited materials therefore bear complex characteristics resembling a rock slide avalanche (at higher levels) and also unconsolidated debris slides (at lower levels). Our remote sensing and classification work thus demonstrates that hydrothermal alteration materials have been incorporated in the landslide materials at Askja as well, developing a mixed (varicolored) material class. Therefore, by using our approach, a systematic search of hydrothermally active zones and regions becomes feasible, allowing (i) identification of those flanks that may be prone to alteration and mechanical weakening prior to collapse, and (ii) consideration of a higher mobility of the mixed materials capable of reaching coastal regions and triggering tsunami hazards. Also other regions at the Askja caldera wall, and at calderas elsewhere, display hydrothermal activity and may therefore lead to similar destabilization and mobilization processes. Future mechanical works are needed and might provide further insights into the degree of weakening of rocks (Heap et al., 2021) and the resulting mobilization of landslide materials.

We found that multicolored mixed materials in the landslide area are largely hydrothermally altered rocks. Furthermore, data suggest that they became mobilized during the slide process (Figures 9A, B). Our post-landslide drone survey suggests that a large number of steep sloped sites still host a similar material class, possible fumaroles, and sites of hydrothermal activity. We speculate that these may be prone to future collapses. Indeed a number of failures occurred after the main 2014 landslide, as was reported based on a local seismic network and high-amplitude short-period events that lasted up to 1 minute (Schöpa et al., 2018), interpreted by these authors due to secondary failures that followed the main slide.

Understanding the mobility of landslide masses is of high relevance for simulations and predictions. For instance, landslides can now be modeled by granular porous media, and their effect when entering water bodies can be studied (Rauter et al., 2022). It is well known that basal shear resistance and topography initially control the mobility of landslide masses, where during the movement the material property and the interaction of the debris with the underlying substrate strongly control mobility (Aaron and McDougall, 2019).

We also speculate that material changes associated with hydrothermal alterations may significantly contribute to the landslide hazard and mobility. During alteration typically a significant reduction in permeability occurs, possibly by four orders of magnitude (Heap et al., 2021). This will modify, direct or even inhibit an effective fluid circulation inside the volcano and at the caldera rim (Figure 9A), leading to zones of high pore fluid pressure that may strongly affect slope stability (Heap et al., 2021). Furthermore, a landslide initiating at a hydrothermally weakened slope may incorporate additional masses of unaltered materials. These so-called erosional landslides may be particularly common at altered volcano slopes, and hence strongly influence the related and possible tsunamigenic energy budget of a moving mass with high mobility (Pudasaini and Krautblatter, 2021).

We conclude that porosity-decreasing alteration, explored here, and porosity-increasing alteration can both promote volcano instability and collapse, but by different mechanisms. Hydrothermal alteration should therefore be monitored at volcanoes worldwide and incorporated into hazard assessments. Especially during periods of unrest, occurrence of local earthquakes, deformation and increased fumarole activity, close monitoring of the steep and altered materials may save lives. At the time of this work, Askja is inflating, therefore a close watch and awareness rising is vital.