Forecasting Eruptions at Poorly Known Volcanoes Using Analogs and Multivariate Renewal Processes
Ting Wang, Mark Bebbington, Shane J. Cronin, Joel Carman
Abstract
Abstract Forecasting future destructive eruptions from re‐awakening volcanoes remains a challenge, mainly due to a lack of previous event data. This sparks a search for similar volcanoes to provide additional information, especially those with better compiled and understood event records. However, we show that some of the most obviously geologically comparable volcanoes have differing statistical occurrence patterns. Using such matches produces large forecasting uncertainties. We created a statistical tool to identify and test the compatibility of potential analogue volcanoes based on repose‐time characteristics from world‐wide datasets. Selecting analogue volcanoes with compatible behavior for factors being forecast, such as repose time, significantly reduces forecasting uncertainties. Applying this tool to Tongariro volcano (NZ), there is a 5% probability for a Volcanic Explosivity Index (VEI) ≥ 3 explosive eruption in the next 50 years. Using pre‐historic geological records of a smaller available set of analogs, we forecast a 1% probability of a VEI ≥ 4 eruption in the next 50 years.