Uncertainty Analysis of the Prediction of Massive Ash Fallout From a Large Explosive Eruption at Sakurajima Volcano

Abstract

AbstractVolcanic ash hazards present life‐threatening dangers to populations near volcanoes during large explosive eruptions. Vulnerable infrastructures demand a comprehensive disaster risk reduction strategy to protect residents from enormous ashfall accumulations. To prepare for the next large eruption of Sakurajima volcano, authorities in Kagoshima City are developing a countermeasure plan utilizing ash dispersal prediction 24 hr before an eruption. However, the absence of large eruptions in the last century makes it challenging to confirm the accuracy of predictions for hazard area designation. To ensure established protocols are effective, uncertainties in ashfall predictions must be addressed, enabling authorities to make appropriate responses. We simulated the previous large eruption of Sakurajima volcano using multiple wind forecast lead times as historical predictions from July 2018 to December 2022. The uncertainty in prediction results was evaluated by comparing simulation outputs with validation data from the ash dispersal data set. We examined uncertainty in hazard area assignment and its variation depending on risk items, wind field states, and the influences of seasonal patterns and events. Updating ashfall predictions with improved wind forecasts reduces uncertainty for all risk items and increases the precision of identified hazard zones. Furthermore, uncertainty levels are influenced by seasonality and can shift significantly with varying wind strengths controlling ash dispersal process. Providing uncertainty information is vital for decision‐making during ash fallout events, and it is recommended to update response decisions 12 hr after the initial prediction. This study's outcomes will aid in developing better disaster risk management strategies, focusing on volcanic ash hazards.