Sunday 26 January 2020

Trying to understand the cause of the 22 December 2018 Anak Krakatau volcanic tsunami.

During the 1883 Krakatau eruption and tsunami, which is estimated to have killed over 36 000 people, 12 km³ of dense rock equivalent was erupted; a caldera collapse occurred as a result, leaving only small and steep subaerial remnants of the former volcano edifice along the rim of a 7 km wide deep-water caldera basin. Volcanism continued after the 1883 events, eventually producing Anak Krakatau ('the son of Krakatau'), where several additional smaller tsunamis were triggered at this site by processes such as underwater explosions. Anak Krakatau may be preconditioned for landslide triggered tsunamis, as it is situated on a steep morphological cliff. This edifice probably first breached the sea surface in 1928 and gradually formed a 150 m high tuff ring by 1959. A gradual shift in activity then occurred toward the southwest, resulting in further growth of the edifice over the cliff and toward the deep submarine caldera basin. This lead to recent concerns about a possible landslide from the southwestern island flank and the corresponding generation of a tsunami. These concerns and scientific assessments turned into reality following an intense (but unidentified) increase in precursor activity. Although flank motion was identified, the hazard was not systematically monitored. On 22 December 2018, this volcanic centre once again became the source of a tsunami that struck the highly vulnerable Sumatran and Java coasts. According to the Indonesian National Disaster Management Authority, the 22 December 2018 tsunami caused over 430 fatalities, injured 14 000 people, and displaced 33 000 more along the Sunda Strait. The tsunami risk of this area is particularly high as the coast is very popular with both locals and tourists and is home to over 20 million people within a 100 km distance from the volcano.

In a paper published in the journal Nature Comunications on 1 October 2019, Thomas Walter of the German Research Centre for Geosciences, Mahmud Haghshenas Haghighi, also of the German Research Centre for Geosciences, and of the Institute of Photogrammetry and GeoInformation at Leibniz University Hannover, Felix Schneider, again of the German Research Centre for Geosciences, Diego Coppola of the Dipartimento di Scienze della Terra at the Università di Torino, Mahdi Motagh, again of the German Research Centre for Geosciences, and the Institute of Photogrammetry and GeoInformation at Leibniz University Hannover, Joachim Saul, Andrey Babeyko, and Torsten Dahm, again of the German Research Centre for Geosciences, Valentin Troll of the Department of Earth Sciences at Uppsala University, and the Faculty of Geological Engineering at Universitas Padjajaran, Frederik Tilmann, again of the German Research Centre for Geosciences, and of the Institute for Geological Sciences at the Freie Universität Berlin, Sebastian Heimann, again of the German Research Centre for Geosciences,  Sébastien Valade, once again of the German Research Centre for Geosciences, and of the Department of Computer Vision & Remote Sensing at the Technische Universität Berlin, Rahmat Triyono of the Earthquake and Tsunami Center at the Indonesian Agency for Meteorology, Climatology and Geophysics, Rokhis Khomarudin of the LAPAN Remote Sensing Application Center, Nugraha Kartadinata of the Volcano Research and Monitoring Division of the Geological Agency of Indonesia, Marco Laiolo, again of the Dipartimento di Scienze della Terra at the Università di Torino, Francesco Massimetti, once again of the Dipartimento di Scienze della Terra at the Università di Torino, and of the Dipartimento di Scienze della Terra at the Università di Firenze, and Peter Gaebler of the Federal Institute for Geosciences and Natural Resources, discuss the causes of the 22 December 2018 Anak Krakatau volcanic tsunami, and the predictability of future similar events.

Volcanic islands are typically fast-growing edifices that rest on a complex morphology and weak substrata, and they are frequently made up of highly fragmented, mechanically unstable material. Therefore, many volcanic islands rise rapidly but also erode swiftly via volcano flank instability, leading to irregular shapes and embayments owing to sector collapses. Geomorphic amphitheatres are common subaerial remnants of fast lateral landslides; these events often recur at the same location. They may result in distal run-out submarine deposits, which demonstrate the intense dynamics of tsunamigenic mass movements in the oceans. In fact, data collected from around the world reveal that historical volcano-induced tsunamis have caused significant damage and loss; about 130 events have been recorded from 80 different source volcanoes since 1600 AD. These events have been caused by the entry of pyroclastic flows into the ocean and their submarine continuation, by caldera collapse, and by landslides entering the ocean, or combinations thereof. Among the progenitors of these events, seventeen historically identified source volcanoes are located in Southeast Asia. Volcano-induced tsunamis have probably led to the demise of ancient civilisations and are responsible for approximately one quarter of all fatalities attributed to volcanic activity.

It has been over 135 years since the infamous volcano-induced Krakatau tsunami occurred on 27 August 1883. A common problem of such events is that they are rare, and thus, although volcanic islands introduce numerous recognisable threats such as instability, sector collapse and tsunamis, little is known about their precursor activity and possible strategies to mitigate the associated risks. Moreover, the preparation and initiation of sector collapses are complex, insomuch that they could possibly be associated with faulting, slumping, and pyroclastic flows or combinations thereof. Consequently, at present, no consensus exists regarding what constitutes a reliable precursor signal for sector collapse on a volcanic island.

Location of Anak Krakatau. Walter et al. (2019).

By combining different ground and satellite data Walter et al. outline the details of the complex hazard cascade leading up to the events on the 22nd December 2018. Our study reveals that Anak Krakatau showed clear signs of flank motion and elevated volcanic activity prior to sector collapse which triggered the destructive tsunami.

Satellite monitoring and ground observations reveal that Anak Krakatau was in an elevated stage of activity throughout 2018. An analysis of infrared data recorded by the thermal sensors of the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument on the Terra satellite indicates that a new intense eruptive phase initiated at Anak Krakatau on 30 June 2018. This eruptive phase was the most intense recorded since systematic data acquisition began in 2000 and was characterised by a mean volcanic radiative power of 146 MW, which is roughly 100 times the long-term thermal emission average (1.6 MW) recorded between 2000 and June 2018. This thermal activity was associated with persistent Strombolian to Vulcanian activity (ejection of incandescent cinders driven by the bursting of gas bubbles and dense clouds of ash-laden gas exploding from the crater) and the emplacement of eruptive deposits along the centre and the western and southern flanks of the volcano. This eruptive phase continued for 175 days until 22 December 2018, when the activity suddenly evolved into a sector collapse.

Thermal emissions recorded by satellite data. Visual summary of the total extent of thermal anomalies, as measured from the Sentinel 2 images showing the emplacement of several hot and new material depositions on the southwestern flank of Anak Krakatau during July-December 2018. Whether these hot materials are due to lava flows or due to clastic deposition cannot deduced from this data. The approximate extension of land inundated by new lava flows and/or covered by hot ejecta (about 0.85 km²), is shown on the lower right panel. Walter et al. (2019).

An estimate of the erupted volume derived from thermal data indicates that the eruption phase produced approximately 25 500 km³ of deposits, implying a mean output rate of 1.7 m³ per second from June 2018 to just prior to the collapse event. Thus, the load acting on the summit and especially the southern flanks of the island progressively increased over this time by about 54 million tons (assuming a mean rock density of 2110 kg per cubic metre).The 2018 eruptive period was punctuated by 11 pulses with time averaged discharge rates higher than 3 m³ per second. The occurrence of these effusive pulses peaked between September and October 2018, with the three highest time averaged discharge rates of 10.5, 33.4, and 50.9  m³ per second, on 9, 15, and 22 September 2018, respectively. Starting in October 2018, the rate of these pulses declined, both in intensity and in frequency, except for a period in mid-November with a peak time averaged discharge rate of 23.2 3 m³ per second. The general decrease in activity after mid-October is also suggested by the trend in the development of the cumulative volume of erupted materials.

According to satellite images from the European Sentinel 2 mission, at least 0.85 km² of the island (28% of the total area) was covered with abundant hot ejecta and new deposits. Many of these entered the sea, adding 0.1 km² of land surface to the southern shore of the island (the island area increased from 2.93 to 3.03 km²) by early December 2018, as assessed by shoreline edge detection analysis. A compositional analysis of ash sampled during the intense eruption phase on 22 July 2018 indicates a basaltic andesite composition, which overlaps with the typical compositional spectrum displayed by Anak Krakatau in recent decades.

Eruptions and island perimeter growth map. (a) Several selected Sentinel 2 images (band combination of 12, 11, 4) showing the emplacement of hot and new material (red-yellow) on the southern flank of Anak Krakatau during the increased eruptive activity in 2018. (b) Island perimeter maps derived from satellite radar amplitude images show little variation from January 2018 to June 2018, followed by southward growth from June 2018 to December 2018 prior to the sector collapse (light grey). The black lines indicate the new scarps formed by the sector collapse. The outer outline indicates the post collapse island perimeter in January 2019 (dark grey). (c) Land area change based on monthly island perimeter analysis. The area changed gradually prior to the flank collapse, but more rapidly after the collapse event. Walter et al. (2019).

Interferometric synthetic aperture radar analysis and time-series analysis show that the southwestern and southern flanks of Krakatau were already slowly subsiding and moving westward at the beginning of the analysis window in January 2018, despite the absence of significant thermal anomalies. The deformation that occurred in the subsequently collapsed sector was advancing at an approximately constant rate with a peak of almost 20 mm (or about 4mm per month) in the satellites’ line-of-sight direction until the volcanic activity increased in late June 2018, at which point the deformation markedly accelerated (to about 10mm per month). In addition, short-term eruptive pulses in Sepember–October 2018 resulted in a minor step change. Therefore, the data show that increased eruption rates coincide with increases in flank movement. An analysis of the deformation field pattern reveals that it affected over one-third of the island, exhibiting a moderate gradient on the west side and a well-identified gradient on the southeast side. The cumulative deformation pattern indicates a progressively sliding flank that can be explained by a deep décollement plane, simulated as a rectangular dislocation, with a dip (steepest angle of descent of a tilted bed or feature) of 35°, a strike of 163° (quadrant compass bearing in terms of east or west of true north), and a slip of 3.36m. Notably, deformation also affected the island summit, and therefore potentially shearing its main magmatic and hydrothermal-plumbing systems.

Island perimeter monitoring. (a) Selected Sentinel 1 GRD images before the collapse, on the day of the collapse (the collapse was at 13.55 GMT, the GRD image at 22.33 GMT) and afterwards. Arising from higher radar reflectivity, steep cliffs of landslide amphitheatre can be partly identified in the 22 December 2018 image (dashed black line). On 25 December 2018, the southwest sector blurs due to eruption plume affecting the radar path1,2. The image on 27 December 2018 reveals the complete amphitheatre geometry (partially infilled in the northwest) and the new coastline (shifted by up to 260 m) in the northeast. Dashed black lines in second and fourth image represent approximate outlines of satellite headwalls, which might indicate two outlines, enclosing approximately 0.63km² and 0.84km², which is 44% and 58% of the deformation area identified in interferometric synthetic aperture radar time series, respectively. (b) An island perimeter detection algorithm was used to determine the growth of Anak Krakatau from 1 January 2018 to 31 January 2019. Walter et al. utilised the Google Earth Engine cloud computing environment³ by first computing monthly stacks from the Sentinel 1 ground range detected (GRD) scenes to reduce the speckle and then segmenting the stacked images using an adaptive threshold to separate land from water bodies. The figure shows the examples of December 2018 and January 2019, for which the stacking has been done to overcome the speckle effect at the cost of averaging short-term variations in the coastline. Each stack considers all Sentinel 1A/B GRD images, ascending and descending, acquired from the first to the last days of a month, except for December 2018, for which GRD images only until December 21 (before the sector collapse) were considered. The enlarged views of the backscatter channel of the GRD images in December 2018 and January 2019 and the average images used to delineate the outlines indicated by red polygons. (c) Monthly evolution of the coastline of Anak Krakatau. Note that largest changes on the southern flank occurred during June-August 2018. Walter et al. (2019).

The dynamics of the moving flank were relatively slow; as a consequence, seismic stations installed on the mainland for tsunami early warning were hardly able to record this type of movement. Then, conditions started to change shortly before the sector collapse event. Satellite thermal data show a pulse (5.6 m³ per second) on 22 December 2018 at 06.50 GMT, just a few hours before the onset of the collapse. Compared with the thermal pulses recorded earlier in 2018, this eruption was relatively small. Infrasound records show the release of continuous high-frequency energy (0.5–5 Hz) from Krakatau, indicating high levels of volcanic activity in the hours prior to the collapse followed by a brief period of quiescence. Both the intense activity earlier in the day and the quiet period were further confirmed by eyewitness accounts. Seismic stations then suddenly recorded a high-frequency event (2–8 Hz4b), just ~115 s before the flank collapsed on 22 December 2018, representing the last and most immediate precursor, or even trigger, of the main sector collapse. The seismic signal originated at Anak Krakatau and was associated with either an earthquake or an explosion with seismic amplitudes that exceeded even those of the sector collapse in the 4-8 Hz frequency band and was even recorded by infrasound stations at large distances. The coda (1–8 Hz) of this event is unusually long compared with those of tectonic earthquakes of comparable magnitude; in fact, the coda is still discernible when the onset signal of the catastrophic sector collapse becomes identifiable.

Deformation maps and island perimeter changes. (a) Interferometric synthetic aperture radar time series showing the movement of the ground in the satellites’ line-of-sight (LOS) direction, generated for the period between 1 January and 22 December 2018 for one ascending and two descending tracks. The different viewing geometries reveal the deformation of the southwestern flank of Anak Krakatau. (b) Average vertical (colours, red is down) and E-W (black vector symbols) movement computed from the different interferometric synthetic aperture radar viewing geometries. (c) Deformation trends for the different tracks (average pixel values in the inner sector collapse scarp outline) showing pronounced trend changes in June–September 2018, labeled I and II, and identifying the 22 December 2018 collapse event. Walter et al. (2019).

Local, regional, and even some teleseismic seismic stations show clear signatures of the tsunami-triggering event. The abrupt onset of a short-period seismic signal is followed by about 5 minutes of strong emissions at 0.1–4 Hz, approximately coinciding with a long-period signal (0.01–0.03 Hz) occurring over a shorter duration (~90 s) that Walter et al. interpret as the seismic signature of the main mass movement of the landslide. The onset times of the short-period signals at stations in Sumatra and Java are consistent with the location of the volcano and an origin time of 13:55:49 GMT. The inversion of low-pass filtered (0.01–0.03 Hz) surface waves reveals an event with a moment magnitude of 5.3. A significant non-double-couple component is retrieved from the inversion of low-pass filtered seismograms, indicating a linear vector dipole oriented to the southwest at 222° and a dip angle of 12° (or alternatively representing tensile opening mixed with a shear rupture dipping about 61° to the southwest). As these parameters are close to those of the pre-eruptive décollement plane derived from interferometric synthetic aperture radar data (northwest–southeast strike and southwest dip), we conjecture that it was this plane that constituted the failure plane during the sector collapse. Following the main event, a nearly continuous tremor-like signal exhibiting a slowly decreasing intensity with frequencies 0.7–4Hz was recorded at all nearby stations and was attributed to strong volcanic explosions.

Interferometric synthetic aperture radar deformation modelling. (a. Unwrapped mean line of sight (LOS) velocity data for the investigated tracks and viewing geometries, used for modelling. Landslides may include a plastic flow rheology, but may be modelled using elastic dislocations especially for slow-moving landslides, in order to assess the geometric and dynamic parameters of the landslide detachment plane. (b) Walter et al. use the rectangular dislocation (RD13) to search for the source of deformation. The RD is based on the angular dislocation in a half-space and provides geometrical parameters allowing to assess shear and tensile faults in an elastic medium. Parameters that determine the orientation of the RD plane are described by the strike angle (α) and the dip angle (δ), while the ‘plunge angle’ (θ) is the angle between the upper edge of the RD and the intersection of the free surface with the extended RD plane. In this work, the deformation modelling considers the 2018 mean displacement velocity prior to the sector collapse. As the tracks provide different viewing geometries (ascending and descending), the method is sensitive to both vertical and horizontal displacements. No further weighting was applied. Walter et al. implement the RD in a non-linear inversion scheme based on the genetic algorithm (GA), where the objective function (OF) that is minimised is the L2-norm of the model residuals. The solution space chosen was initially wide (dimension was free, plunge 0 degrees, dip from 0 to 90 degrees, strike from 0 to 360 degrees, rake from -100 to -80 degrees). Walter et al. define a GA population size of 60, a mutation rate of 0.15, and 1000 iterations, with a Poisson ratio of 0.25. (c) Residuals created by calculating the difference of the data and the model. Largest absolute residual values exceed 10 cm in those regions of young material depositions in the southern sector of the island. Walters et al. examined this feature and conjecture that there some signal may arise from compaction or cooling of young volcanoclastic material. Independent tests have been made where we were masking out the southern sector, but found generally comparable results for the RD source, implying that the new material addition in the southern sector of the island had minor effect only on the inversion. Walter et al. (2019).

The effects of this event were recorded extensively. The Australian infrasound array located over 1150 km to the southwest of Anak Krakatau recorded a high-energy impulse at 15.01.09 GMT, which translates to a modelled origin time of 13:55:49 GMT at the Krakatau site. This timing is consistent with the origin time of the short-period seismic signal at Anak Krakatau (identified as the landslide signal). The duration of the impulse is broadly comparable to the long-period seismic signal and indicates that subaerial sliding lasted for about one minute only. Both the seismic records at local stations and the infrasound records from the Australian infrasound array continued to be dominated by coherent emissions from Anak Krakatau (presumably related to strong volcanic eruptive activity there) for at least several hours. The dominant frequency of the eruption signature in the infrasound signal shifted by nearly an order of magnitude (from about 0.8–4 Hz prior to the landslide to 0.1–0.7 Hz afterwards). Even the closest local stations did not pick up a clear signature of any pre-landslide eruption, but post-landslide eruptions dominated the seismograms of stations even a few hundred kilometers away. Together, these observations suggest a profound change in eruptive style following the landslide. Furthermore, on 23 December 2018 at 06.31 GMT, a large sulphur dioxide cloud was detected, likely resulting from the decapitated and degassing hydrothermal system. In contrast, no similarly strong degassing was detected in the weeks prior to the flank collapse event. 

 Bounce-point distribution from rays starting at Krakatau. The infrasound wave was clearly identified at the Australian infrasound array (southwestern corner of map), but an infrasound phase of the main event was also identified at seismic stations LWLI and MDSI. Networks GE and IA are indicated with red and green symbols, respectively. Walters et al. (2019).

Tsunami arrivals were recorded by four tide gauge stations on the Sumatra and Java coasts. Backtracing from these four stations, using the classic tsunami travel time approach, suggests that the source location corresponds to the southwestern part of Anak Krakatau and that the source origin time corresponds to that revealed by the broadband seismic analysis. Therefore, the backtracing simulation shows that the tsunami was triggered by the long-period landslide during the first minutes of the event and not by the following volcanic eruptions.

Tsunami travel time backtracing. The isochrons show all possible tsunami source points for each station at approximately 13.56 (the seismically determined onset time of the slope instability). The isochrons intersect at Anak Krakatau. It was necessary to shift the picked arrival times by 1-3 minutes to make all four isochrons overlap and agree with the seismic origin time (marked as DT+x in the figure). Given that a coarse global bathymetry was used, this small discrepancy is expected AT: arrival time, OT: (seismic) origin time. Walters et al. (2019).

The full extent of the sector collapse event initially remained hidden owing to intense post-collapse eruptive activity but became visible when the eruption intensity decreased again by 27 December 2018. As a result, a new and steep amphitheatre enclosing a deep valley became distinguishable on the southwestern sector of the island. The deposition of new material shifted the coastlines. The collapsed area is readily identified in satellite radar imagery and is located in the area that was subsiding and moving laterally outward prior to the collapse event. The area affected by landslides, however, is smaller than the area affected by precursory deformation; accordingly, we estimate that only 45–60% of the deforming subaerial flank actually failed. High-resolution camera drone records in January 2019 allow the partial derivation of a digital elevation model. By comparing the digital elevation models from before and after the event, we ascertain that the sector collapse reduced the height of the island from 320 to 120 m and removed the former edifice peak, thereby decapitating the main eruption conduit. Detailed volumetric estimates obtained upon differencing the two digital elevation models suggest an estimated volume loss of 102 000 000 m³, which is a minimum estimate, as it does not consider the volume gained by new eruptive deposits (which may exceed another 100 000 000 m³); furthermore, the submarine collapse volume is not included and necessitates forthcoming bathymetric surveys. Tephra deposition occurred immediately after the sector collapse (between 22 and 25 December 2018, as determined by satellite radar images), causing a shift in the perimeter of the island and overprinting the collapse scar geometry.

Morphological changes by sector collapse, material deposition, and subsequent erosion. (a)–(c) TerraSAR-X satellite radar images in high-resolution spotlight mode showing the extents of changes associated with sector collapse and with the formation and erosion of the new coastline, tuff ring, and crater lake before and after the 22 December 2018 collapse. (d), (e) Before and after comparison of the morphology deduced from TanDEM-X data (before, light-gray shading) and camera drone data (after, dark gray shading) revealing profound topographic changes, material loss and deposition of new materials. Erosion was carving gullies on the volcano flanks by early January 2018. Walters et al. (2019).

Profound changes continued to occur in the weeks following the catastrophic event. Numerous small slumps deposited material into the landslide amphitheatre and an explosion tuff ring formed inside the decapitated volcano conduit area. The eruption site now appears slightly shifted to the southwest, hosting a new 400 m wide water-filled crater. Thermal activity was detected after the collapse, possibly linked to ongoing eruptions. Although the collapse of the southwestern sector into the ocean was associated with a considerable volume loss, area calculations of the island reveal rapid regrowth (over 10%) from December 2018 to January 2019, which was mainly associated with the (re-)deposition of pyroclastic material.

Cascade of precursors leading up to the 22 December 2018 sector collapse event. (a) Precursors include flank motion (white arrow), eruptions (represented as eruption cloud), and increasing eruptive deposits (red shaded areas) as assessed by satellite data (thermal, interferometric synthetic aperture radar). A décollement (black line beneath the island) dips southwest, but faulting had not yet breached the surface. Approximately 2 minutes before the collapse, a seismic event was recorded (shown by seismic trace symbol). (b) The landslide collapse along a failure plane (black curved line beneath island) showed a 1–2-minute long low-frequency signal (seismic waveform). Infrasound instruments (speaker symbol) measured the collapse before the tsunami waves arrived. The collapse decapitated the island (grey shaded area). The tsunami (blue wave) caused damage and loss along the coast. (c) Post-collapse volcanic explosions occurred coincident with increased degassing (grey plume) caused by unloading (arrow symbol); the old topography is indicated (black dashed line). New eruptive deposits increased the island area (red shaded areas). Finally, rapid erosion deeply carved incisions into the fresh eruption deposits. Walters et al. (2019).

On the basis of the remotely sensed displacement data and thermal analysis, we conclude that the Krakatau volcano showed clear signs of flank motion and elevated volcanic activity prior to the 22 December 2018 sector collapse. The long-term hazard owing to Anak Krakatau’s steep volcanic edifice had already been described, and thus, the collapse event and subsequent tsunami were anticipated hazards. The month-scale precursors included the strongest thermal activity recorded at Anak Krakatau in about 20 years and an accelerated flank motion; these characteristics made Anak Krakatau one of the most rapidly deforming volcano flanks known on Earth prior to its collapse. In fact, deformation was already identified along the southwestern flank of Anak Krakatau in interferometric synthetic aperture radar time series over 10 years before December 2018. but this deformation was not interpreted to be a potential precursor of a larger sector collapse. Compared with other volcanoes that exhibited flank deformation prior to sector collapse, the movement at Anak Krakatau also corresponded with eruption pulses, possibly associated with pressure changes in the volcano interior, and therefore Anak Krakatau shares a similar behaviour with volcanoes elsewhere.

Destruction over the island as identified in Sentinel 2 data and as seen from drone photos. (a) Sentinel-2 bands 8, 4, and 3 in combination with the RGB channels show vegetation as red and loss of vegetation after eruption in dark gray. (b) The drone images reveal the profound destruction on the adjacent island of Kecil, most likely associated with devastation due to surge and gas plumes. The upper image shows coastline erosion on western Kecil. Lower image shows view over Kecil, looking to the north. Walters et al. (2019).

Walters et al. investigated whether changes in composition could explain the increase in magma production at Anak Krakatau prior to its collapse, but our analysis of syn-deformation tephra samples suggest that the material was not significantly different from the material erupted in past decades, implying that deep magmatic changes were likely not directly responsible for the observed dynamic changes at the surface. The orientation of the main sliding plane of the collapse event could be identified from the seismic records of the collapse, suggesting a steeply southwesterly dipping failure nodal plane. The strike and dip of this plane are geometrically in agreement with the amphitheatre morphology and also notably with the inferred dislocation plane of precursory creep motion. Therefore, we conclude that the landslide décollement had already developed before the collapse.

Furthermore, as a significant proportion of the island and its shallow plumbing system was removed by the flank collapse, unloading may affect (also future) post-collapse compositions. the structural evolution, magma pathways, and eruption locations. 

A remaining question is whether the landslide of Anak Krakatau was triggered by volcanic or seismic activity. Our observations indicate that the climax of the eruptive phase was recorded in late September 2018, about 3 months prior to the flank collapse. Indeed, from September to December 2018, the volume of newly deposits followed a generally decreasing trend. In addition, sulphur dioxide gas emissions were low in the weeks prior to the collapse. Moreover, because the deformation rate remained almost constant throughout this period, Walters et al. suggest that only minor change, if any, was attributable to the accumulation of magma into the shallow portions of the edifice. However, the intense activity witnessed throughout the year likely increased the overall instability of the edifice owing to the rapid accumulation of new material. In fact, studies elsewhere show that slope instability at volcanoes is not always associated with eruptive phases. This relationship is observed because slope instability changes over time; in addition, fault planes and other zones of weakness are strongly affected by pore pressure changes, hydrothermal activity, and mechanical weakening by alteration, as well as by sea erosion and oversteepening. A similar but much smaller sequence recently occurred also at Mount Etna, where a short-term increase in the magma supply and eruption rate was accompanied by magmatic intrusion, leading to the collapse of an unstable cone. This example demonstrates that under such critical conditions, minor internal and external perturbations can potentially trigger a collapse and eruption, leading to a disaster. From this perspective, Walters et al.'s hypothesis that the seismic event identified two minutes before the Anak Krakatau landslide acted as an external trigger is plausible but remains to be tested further.

Volcano-induced tsunamis are thought to be rare and are therefore not commonly considered in tsunami early warning centres. Historic documents reveal, however, that Southeast Asia experiences volcano-induced tsunami hazards relatively frequently, with 17 events during the 20th century and at least 14 events during the 19th century, defining a recurrence rate of one event every 5–8 years. A volcano-induced tsunami from Anak Krakatau was anticipated, but accurate predictions were impossible owing to a lack of understanding of the processes involved. Hence, the study of the 22 December 2018 sector collapse at Anak Krakatau provides important new information about the precursors and processes that culminated in the disaster.

The tsunami reached the coastal towns of Jambu, Ciwandan, Agung, and Panjang within 31, 38, 39, and 57 minutes, respectively. The tsunami waves were overtaken by the faster seismic waves and the infrasound signals of the strong explosive eruption, that were associated with the landslide and decapitation of the hot interior of the volcano followed by steam-driven phreatomagmatic explosions.

In fact, the seismic records following the sector collapse event indicate that tremor activity continued for hours, resembling the volcanic tremors associated with steam-driven explosions elsewhere, although the large distance between the volcano and the seismic network may have blurred this interpretation. The eruptive style of the post-decapitation eruptions was different from that of the pre-decapitation eruptions. This is indicated by the different frequency contents in the seismic and infrasound data and the onset of significant sulphur dioxide emissions on 22–23 December 2018, in agreement with the period characterised by a reduction in radar amplitudes, the deposition of erupted material and a shift in the coastline, as seen in the ground-range detected images on 22–25 December 2018. The productivity of the eruptions also appeared to increase, as evidenced by the marked growth of the island area after the collapse. This suggests a major effect of unloading on the magmatic and hydrothermal interior of the volcano.

It appears that a perfect storm of magma-tectonic processes at Anak Krakatau culminated in the 22 December 2018 tsunami disaster. Leading up to the event, different sensors, and methods measured distinct anomalous behaviours, which in hindsight can be deemed precursory. However, at the time and when considered individually, none of the parameters, including the thermal anomalies, flank motion, anomalous degassing, seismicity, and infrasound data, were sufficiently conclusive to shed light on the events that were about to unfold.

Walters et al.'s study demonstrates that volcano sector collapses and the resulting tsunamis might be effectively anticipated by continuously monitoring the various preparation stages. Long-term flank motion, changes in thermal emission, and short-term seismic events precede the collapse, which itself was well monitored by low-frequency seismic waveforms and infrasound stations. Assessments of these parameters could be implemented in available early warning systems. Therefore, the next-generation tsunami early warning system must consider multiparametric observations, since our study reveals that a multitude of changes signified an unprecedented level of activity at Anak Krakatau prior to 22 December 2018. Walters et al. hence recommend a dedicated search for island volcanoes that are susceptible to flank collapses and those with a history of tsunamis, and we advise the development of appropriate monitoring programs to identify critical systems at these sites.

Because the Anak Krakatau sector collapse and tsunami are rare events, insights such as those reported by Walters et al. yield vital information on precursor processes and aid in refining existing monitoring and early warning technologies.

See also...
Follow Sciency Thoughts on Facebook.