Time series analysis and stochastic processes modeling using explosive eruption data from Japan’s Sakurajima volcano

Abstract

This study examines the statistical properties of explosive eruptions at Japan’s Sakurajima volcano from 1955 to 2023. The investigation revealed that the distribution of eruption intervals conforms to a power-law distribution. Additionally, a Hurst exponent of approximately 0.81 was computed for the monthly average daily frequency of explosive eruptions. Based on these findings, it was determined that the occurrence of explosive eruptions of Sakurajima volcano over the period of study deviates from a Poisson process and exhibits characteristics of a non-stationary Poisson process influenced by prior explosive events. Subsequent modeling using the autoregressive integrated moving average (ARIMA) technique showed that the ARIMA(1,1,0) model provides an excellent fit to the observed data.