Abstract Volcanic lateral collapses represent major hazards through associated eruptive activity, landslide inundation and, in island or coastal settings, the generation of tsunamis. Forecasting the timing, precursory indicators, and magnitude of lateral collapses still remains a significant gap in volcanic risk management, a challenge exemplified by the 2018 lateral collapse of Anak Krakatau, Indonesia. The volcano’s southwestern flank collapsed without recognised warning, resulting in a devastating tsunami which inundated > 300 km of regional shorelines. Since then, the edifice has rapidly regrown, prompting a necessary assessment of its future stability. Here we analyse Anak Krakatau’s uniquely detailed archival growth records, in combination with satellite and drone datasets, to reconstruct historical edifice development from 1919 to 2023, using this to then project and contextualise its future growth trajectories and assess potential instability conditions. Our time series of 3D models reveals that post-collapse regrowth has mimicked historical development trends, but on a considerably accelerated timescale. If future growth follows the long-term pre-collapse (1960–2018) average, then a subaerial edifice morphology equivalent to 2018, but with larger overall dimensions, would be met by around the year 2100. If future development maintains the much higher 2019–2023 growth trends, then a morphology comparable to 2018 could instead be reached by the 2030s. We consider the continuation of such high growth rates unlikely, but this highlights the importance of monitoring edifice growth and flank instability signals over the coming decades, following the methodology provided here. Additionally, the submarine SW flank now has a considerably reduced gradient relative to 2018, which is likely to promote stability further into the future. To comprehensively assess future stability, the role of alteration, hydrothermal activity and the structure of the submarine flank should also be evaluated.
Meredew et al. (Thu,) studied this question.