The 2014 Lake Askja rockslide‐induced tsunami: Optimization of numerical tsunami model using observed data

Abstract

A large rockslide was released from the inner Askja caldera into Lake Askja, Iceland, on 21 July 2014. Upon entering the lake, it caused a large tsunami that traveled about ∼3 km across the lake and inundated the shore with vertical runup measuring up to 60–80 m. Following the event, comprehensive field data were collected, including GPS measurements of the inundation and multibeam echo soundings of the lake bathymetry. Using this exhaustive data set, numerical modeling of the tsunami has been conducted using both a nonlinear shallow water model and a Boussinesq‐type model that includes frequency dispersion. To constrain unknown landslide parameters, a global optimization algorithm, Differential Evolution, was employed, resulting in a parameter set that minimized the deviation from measured inundation. The tsunami model of Lake Askja is the first example where we have been able to utilize field data to show that frequency dispersion is needed to explain the tsunami wave radiation pattern and that shallow water theory falls short. We were able to fit the trend in tsunami runup observations around the entire lake using the Boussinesq model. In contrast, the shallow water model gave a different runup pattern and produced pronounced offsets in certain areas. The well‐documented Lake Askja tsunami thus provided a unique opportunity to explore and capture the essential physics of landslide tsunami generation and propagation through numerical modeling. Moreover, the study of the event is important because this dispersive nature is likely to occur for other subaerial impact tsunamis.

Short comment on: Comment on: Dynamics of the Askja caldera July 2014 landslide, Iceland, from seismic signal analysis: precursor, motion and aftermath

“…Assuming an average density of 2000 kg/m 3 , 35-80 million m 3 of landslide volume was obtained. This value overestimated the landslide volume compared to the 20 million m 3 reported by Gylfadóttir et al (2017). We attribute this discrepancy to (i) the underestimation of the runout distance used in the seismological determination of the landslide mass.…”

“…We also noticed that the U 0 value derived from the tsunami modelling of Gylfadóttir et al (2017) is very sensitive to the friction coefficient (µ), which ranges from 0.15 to 0.30 for the majority of rockslide configurations. With the fixed input parameters, such as µ, total deposit volume, drag coefficient (C d ), and add mass coefficient (C m ), U0 and the block thickness (d) are obtained through a grid-search scheme by fitting the observed water level of the lake.…”

“…We further attribute C2 the discrepancy of the landslide volumes to (ii) the limitation of applying a constant mass assumption in the waveform inversion, (iii) the effects of a rotational block slide, and (iv) uncertainties in the volume computation resulted from the poor constraint of the sliding surface. Nevertheless, we apologise for unintentionally not having quoted Gylfadóttir et al (2017)'s estimation of the landslide mass correctly.…”

“…Sliding velocity: Schöpa et al (2018) stated the fact that the waveform inversion gives the spatially and temporally averaged velocity of the whole sliding block on land, whereas the tsunami modelling is adopting the velocity (U 0 in Gylfadóttir et al, 2017) of the front of the slide entering the lake. Therefore, a comparison between these two velocity values is difficult to make.…”

Probabilistic Tsunami Hazard Analysis: Multiple Sources and Global Applications

“…There is however, still a long way to go in order to understand when the postfailure can be neglected in (probabilistic) modeling of tsunami generation. For subaerial landslide tsunamis, numerical models based on primitive equation sets ( Euler or Navier‐Stokes ) are generally needed to model tsunami generation and early propagation in detail (e.g., Abadie et al, ; Crosta et al, ), although the far‐field propagation may be treated by dispersive wave models (e.g., Gylfadottir et al, ). For a rigorous review of landslide tsunami models, see Yavari‐Ramshe and Ataie‐Ashtiani ().…”