Skip to content

Forecasting tephra deposition: the impact of input parameter uncertainty on tephra deposition accuracy

Emmy Scott, Melody Whitehead, Stuart Mead, Mark Bebbington, Jonathan Procter

Abstract

Accurate forecasts are critical to help mitigate the risks of volcanic hazards to society. While  post-event modelling, or hindcasting, allows for input parameters to be reasonably well constrained through observations (e.g., plume height), or post eruption analyses (e.g., TGSD),  forecasting these input parameter ranges prior to an eruption is much more uncertain. In these  cases, forecasts rely on probabilistic estimates from limited data (e.g., IVESPA database) and  expert judgement.   From an impact-based perspective, a forecasted ashfall of 3mm rather than 2mm may have  little impact, but forecasting 50mm when 200mm occurs could have major consequences and  cause significant risk to life. Tephra deposition, even during hindcasts where input parameters  are constrained through observation and post-eruptive studies, can deviate by up to a factor of  5 from actual ground deposits using current tephra dispersal models (e.g., Fall3D, Tephra2,  HAZMAP). Correctly quantifying this uncertainty is also crucial to hazard mitigation.   This research explores how input parameter ranges in tephra dispersion models Tephra2 and  Fall3D influence ash deposition forecasts in the context of the next eruption. Specifically, we want  to know how well input parameter ranges based on uninformed priors (i.e., we know nothing  about the next eruption) and informed priors (i.e., we know things such as volcano type and  previous eruption sizes and styles) can produce robust tephra deposition forecasts compared  to real deposit data. Using the example of the 17 June 1996 Mount Ruapehu eruption in  Aotearoa New Zealand, we evaluate how these priors impact forecast robustness and accuracy.