24 September 2019

A new analysis of the deadly Anak Krakatau flank collapse

Posted by Dave Petley

A new analysis of the deadly Anak Krakatau flank collapse

Schematic diagram of the Anak Krakatau failure

Schematic diagram of the sequence of events in the Anak Krakatau flank collapse. From Williams et al. (2019).

Almost a year ago, the collapse of the flank of Anak Krakatau in Indonesia generated a tsunami that killed over 400 people.  The flank collapse occurred without warning, although there had been concerns for a long time that sch an event could occur (as is the case for other volcanic flanks of course).  The journal Geology has recently published a paper (Williams et al. 2019) that provides a first detailed analysis of this event.  The findings are in some ways surprising.  I previewed the article when it was published on Earth ArXiv; the full paper has now been published.  I should also note that the authors acknowledge my earlier comments in the paper.

Williams et al. (2019) provide some interesting information about the consequences of the event – for example, the tsunami that caused the losses had a maximum height of 1.40 m; it killed 431 people and injured a further 7,200; it destroyed 1,778 houses; and it damaged 434 boats and ships.  But the focus of the work is the use of satellite data to reconstruct the chronology of events and, most importantly, to analyse the magnitude and dynamics of the flank collapse itself.  The key conclusion about the sequence of events is contained in the schematic diagram to the left.

In essence, the volcano failed in a large rotational landslide that removed the flank of the volcano.  The toe of the slide (and thus the main mass of the landslide) was below sea level.  This triggered the partial failure of a second portion of the flank of the volcano (see diagram B in the schematic illustration), but this section did not proceed to full failure at this point.  Subsequently, the volcano replumbed to generate a new vent through the basal surface of the landslide (see schematic diagram C).  The eruption through this vent involved the ingress of sea water, generating a violent phreatomagmatic eruption.  This eruptive event removed the remainder of the flank of the volcano, including the partially slipped landslide mass.

The subsequent eruptive events on the volcano then led to the generation of a new, more stable morphology, allowing the activity to stabilise (see schematic diagram D).

The surprising element of this analysis is the scale of the main Anak Krakatau flank collapse.  The satellite imagery allows the construction of detailed cross sections, although clearly some assumptions need to be made about the underwater configuration given the lack of bathymetric data.  But this analysis yields a volume of about 4 million m³ for the subaerial (i.e. above water) component and 100 million m³ for the below sea level component of the landslide.  Whilst this is a very large landslide, it is remarkably small for a flank collapse, and it is also remarkably small for a landslide to have generated such a large tsunami.  Williams et al. (2019) compare their findings with a previous study that modelled the generation of a tsunami from an Anak Krakatau flank collapse, but which assumed a failure volume of 280 million m³.  The observed waves generated by the actual flank collapse were on a similar scale to those modelled for the much larger event.  Interestingly, they also moved through the water much faster than the model suggested.  This suggests that the model is under-predicting tsunamis generated by large landslides.

So, overall, this is a really interesting analysis of the Anak Krakatau flank collapse.  The results have some profound implications for localised but highly destructive tsunami generation from these events, and the study implies that we will need to look again at the ways in which tsunamis are modelled.


Rebecca Williams, Pete Rowley, Matthew C. Garthwaite. 2019. Reconstructing the Anak Krakatau flank collapse that caused the December 2018 Indonesian tsunami. Geology. https://doi.org/10.1130/G46517.1