Probabilistic assessment of tephra fall hazards in Japan using a tephra fall distribution database

Abstract Tephra falls can disrupt critical infrastructure, including transportation and electricity networks. Probabilistic assessments of tephra fall hazards have been performed using computational techniques, but it is also important to integrate long-term, regional geological records. To assess tephra fall load hazards in Japan, we re-digitized an existing database of 551 tephra distribution maps. We used the re-digitized datasets to produce hazard curves for a range of tephra loads for various localities. We calculated annual exceedance probabilities (AEPs) and constructed hazard curves from the most complete part of the geological record. We used records of tephra fall events with a Volcanic Explosivity Index (VEI) of 4–7 (based on survivor functions) that occurred over the last 150 ka, as the database contains a very high percentage (around 90%) of VEI 4–7 events for this period. We fitted the data for this period using a Poisson distribution function. Hazard curves were constructed for the tephra fall load at 47 prefectural offices throughout Japan, and four broad regions were defined (NE–W, NE–E, W, and SW Japan). AEPs were relatively high, exceeding 1 × 10 −4 for loads greater than 0 kg/m 2 on the eastern (down-wind) side of the volcanic front in the NE–E region. In much of the W and SW regions, maximum loads were heavier, but AEPs were lower (<10 −4 ). Tephras from large (VEI ≥ 6) events are the predominant hazard in every region. A parametric analysis was applied to investigate regional variability using AEP diagrams and slope shape parameters via curve fitting with exponential and double-exponential decay functions. Two major differences were recognized between the hazard curves from borehole data and those from the digitized tephra database. The first is a significant underestimation of AEP for frequent events using the tephra database, by one to two orders of magnitude. This is explained in terms of the lack of records for smaller tephra fall events in the database. The second is an overestimation of the heaviest tephra load events, which differ by a factor of two to three. This difference might be due to the tephra fall distribution contour interpolation methodology used to generate the original database. The hazard curve for Tokyo developed in this study differs from those that have been generated previously using computational techniques. For the Tokyo region, the probabilities and tephra loads produced by computational methods are at least one order of magnitude greater than those generated during the present study. These discrepancies are inferred to have been caused by initial parameter settings in the computational simulations, including their incorporation of large-scale eruptions of up to VEI = 7 for all large stratovolcanoes, regardless of their eruptive histories. To improve the precision of the digital database, we plan to incorporate recent (since 2003) tephra distributions, revise questionable isopach maps, and develop an improved interpolation method for digitizing tephra fall distributions.