Volcano Monitoring System for Long-Term Eruption Forecasting Using Multiple Data Sources
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Accurately forecasting volcanic eruptions is challenging due to the complexity of precursory signals. Here, we develop a machine learning-based long-term eruption forecasting model for Mount Aso, Japan, by integrating multiple observational datasets—seismic tremors, magnetic field, crater wall temperature, thermal pool temperature and volume, tilt, and volcanic gas amount—at the characteristic temporal scales of the underlying physical phenomena. The temporal scales are aligned with the intrinsic dynamics captured by each dataset to enhance the model's predictive capability. We construct a theoretical framework to quantify the predictive performance improvement. Our proposed model significantly improves predictive performance, increasing the Matthews correlation coefficient by 0.65 compared to the conventional seismic-tremor-based model, and achieving a precision of >70% in predicting volcanic eruptions. Our findings demonstrate that an ensemble of multiple data sources over optimized temporal scales, underpinned by a theoretical ensemble framework, enables high-precision, interpretable eruption forecasts months in advance and makes effective disaster mitigation planning possible.