National-level long-term eruption forecasts by expert elicitation

Abstract

Volcanic hazard estimation is becoming increasingly quantitative, creating the potential for land-use decisions and engineering design to use volcanic information in an analogous manner to seismic codes. The initial requirement is to characterize the possible hazard sources, quantifying the likely timing, magnitude and location of the next eruption in each case. This is complicated by the extremely different driving processes at individual volcanoes, and incomplete and uneven records of past activity at various volcanoes. To address these issues, we carried out an expert elicitation approach to estimate future eruption potential for 12 volcanoes of interest in New Zealand. A total of 28 New Zealand experts provided estimates that were combined using Cooke’s classical method to arrive at a hazard estimate. In 11 of the 12 cases, the elicited eruption duration increased with VEI, and was correlated with expected repose, differing little between volcanoes. Most of the andesitic volcanoes had very similar elicited distributions for the VEI of a future eruption, except that Taranaki was expected to produce a larger eruption, due to the current long repose. Elicited future vent locations for Tongariro and Okataina reflect strongly the most recent eruptions. In the poorly studied Bay of Islands volcanic field, the estimated vent location distribution was centred on the centroid of the previous vent locations, while in the Auckland field, it was focused on regions within the field without past eruptions. The elicited median dates for the next eruptions ranged from AD2022 (Whakaari/White Island) to AD4390 (Tuhua/Mayor Island).