Long-Term Probabilistic Volcanic Hazard Assessment Using Open and Non-open Data: Observations and Current Issues

Pablo Tierz

doi:10.3389/feart.2020.00257

Abstract

Probabilistic volcanic hazard assessment (PVHA) has become the paradigm to quantify volcanic hazard over the last decades. Substantial aleatory and epistemic uncertainties in PVHA arise from complexity of physico-chemical processes, impossibility of their direct observation and, importantly, a severe scarcity of observables from past eruptions. One factor responsible for data scarcity is the infrequency of moderate/large eruptions; other factors include lack of discoverability and accessibility to volcanological data. Open-access databases can help alleviate data scarcity and have significantly contributed to long-term PVHA of eruption onset and size, while are less common for data required in other PVHA components (e.g. vent opening). Making datasets open is complicated by economical, technological, ethical and/or policy-related challenges. International synergies (e.g. Global Volcanism Program, WOVOdat, Global Volcano Model, EPOS) will be key to facilitate the creation and maintenance of open-access databases that support Next-Generation PVHA. Additionally, clarification of some misconceptions about PVHA can also help progress. Firstly, PVHA should be understood as an expansion of deterministic, scenario-based hazard assessments. Secondly, a successful PVHA should sometimes be evaluated by its ability to deliver useful and usable hazard-related messages that help mitigate volcanic risk. Thirdly, PVHA is not simply an end product but a driver for research: identifying the most relevant sources of epistemic uncertainty can guide future efforts to reduce the overall uncertainty. Broadening of the volcanological community expertise to statistics or engineering has already brought major breakthroughs in long-term PVHA. A vital next step is developing and maintaining more open-access datasets that support PVHA worldwide.

Highlights

300,000 people died due to volcanic activity from 1600 to 2010 AD (Auker et al, 2013)
One of the most infamous examples of volcano tragedy was the eruption of Mount Pelée, Martinique, in 1902, which claimed the life of around 30,000 people at St Pierre and Morne Rouge (Lacroix, 1904; Fisher et al, 1980)
This can be achieved through probabilistic analyses of the onset, size and location of volcanic eruptions, as well as of the spatiotemporal intensity of hazardous phenomena such as pyroclastic density currents (PDCs) or lahars

Summary

INTRODUCTION

300,000 people died due to volcanic activity from 1600 to 2010 AD (Auker et al, 2013). It is not necessarily statistically representative in terms of stratified random sampling, it serves to illustrate certain research trends in PVHA over the last two decades This collection, which I will denote as the “Sample” is analyzed (Figure 2; and Supplementary Material), focusing on the use of open, non-open and mixed (i.e., both open and nonopen) datasets (see Definitions) to assess different components of the long-term forecast: (1) eruption onset and size; (2) eruption vent location; (3) eruption impacts (by hazardous phenomena); and (4) combinations of any of the three, which I term integrated PVHA. With potential applicability to integrated PVHA, have designed: (a) community cyberinfrastructure platforms that host or link to other open-access databases and provide online tools to support hazard assessment (Vhub, Palma et al, 2014; Volcanic Hazards Assessment Support System by G-EVER, Takarada, 2017); or (b) methods to identify objective sets of analogue volcanoes from global databases (VOLCANS, Tierz et al, 2019)

DISCUSSION

DATA AVAILABILITY STATEMENT

Data Accessibility