Since ancient times, scientists have sought to identify precursory signals which might allow the prediction of impending earthquakes or volcanic eruptions. In recent decades, interest has focussed on electromagnetic disturbances, especially ultra-low-frequency (ULF) signals of less than one Hertz, as numerous ULF electromagnetic phenomena have been linked empirically to earthquakes from around the world. Even so, due to the difficulties and considerable costs in maintaining ULF magnetic observations, long-term monitoring of such signals in a seismic area is quite rare.
In a recent paper, LML External Fellow Jiancang Zhuang and colleagues examine data from a unique long-term continuous observation network which was installed in the Kanto region of Japan at the end of last century. One previous study of this data established a statistical correlation between ULF magnetic anomalies and local sizeable earthquakes, but did not clarify if these anomalies could be used to improve the forecasting of sizable earthquakes. To address this issue, Zhuang and colleagues chose to analyse geomagnetic data observed between 2000 and 2010 at two stations with the most complete data, these being the Seikoshi (SKS) and Kiyosumi (KYS) stations in Izu and Boso Peninsulas. The geomagnetic records consist of two orthogonal horizontal (N–S and E–W) and one vertical (Z) component, with a sampling rate of 1 Hz. To minimize the influences of artificial noises, the study considered data recorded only during 01:00–04:00 local time (LT) when the local train system was not in operation.
Their results demonstrate that earthquake predictions which exploit magnetic anomalies are significantly more accurate than random guessing, suggesting that the geomagnetic data recorded in the Kanto Region, Japan do contain useful precursory information for local sizable earthquakes. The authors caution, however, that this study has not explored the practical aspects of making such predictions. Many other case studies have also identified precursory signals, but most rest on retrospective data analysis, rather than real-time prediction of future events. Future work, the authors suggest, should explore the development of an optimal prediction model – specifying the details of time, location and magnitude – to test the possibility of using ULF anomaly data to make real predictions.
The paper is available at https://www.mdpi.com/1099-4300/22/8/859