Search

Since deep low‐frequency tremors are considered to be associated with large earthquakes that occur adjacently on the same subducting plate interface, it is important to investigate tremors that occurred before the establishment of modern seismograph networks such as the High Sensitivity Seismograph Network (Hi‐net). We propose a deep‐learning solution to detect evidence of tremors in scanned images of paper seismogram records from over 50 years ago. In this study, we fine‐tuned a convolutional neural network (CNN) based on the Residual Network, which was pre‐trained using images of synthetic waveforms from our previous study, using a data set comprised of images generated from real seismic data recorded digitally by Hi‐net to facilitate a supervised analysis. The fine‐tuned CNN was able to predict the presence or absence of tremors in the Hi‐net images with an accuracy of 98.64%. Gradient‐weighted Class Activation Mapping heatmaps created to visualize model predictions indicated that the CNN's ability to detect tremors is not degraded by the presence of teleseisms. Once validated using the Hi‐net images, the CNN was applied to paper seismograms recorded from 1966 to 1977 at the Kumano observatory in southwest Japan, operated by Earthquake Research Institute, The University of Tokyo. The CNN showed potential for detecting tremors in scanned images of paper seismogram records from the past, facilitating downstream tasks such as the creation of new tremor catalogs. However, further training using an augmented data set to control for variables such as inconsistent plotting pen thickness is required to develop a universally applicable model. In 2002, deep low‐frequency tremors, which periodically occur with much smaller amplitudes and longer durations than common earthquakes, were discovered owing to the establishment of dense seismic arrays in Japan. These tremors are considered to be associated with megathrust earthquakes because they occur in deeper regions of plate boundaries compared with common earthquakes. Investigating tremors that occurred before the establishment of dense seismic arrays is important since megathrust earthquakes recur on intervals of 100–200 years, and digital records from modern seismic arrays cover only a fraction of that time. In this study, we developed a convolutional neural network based on the Residual Network (ResNet) that extracts tremors from images of waveforms recorded more than 50 years ago, when seismometers drew waveforms directly on drum‐rolled paper using a pen. The proposed ResNet, which was trained on seismogram images generated from synthetic waveforms and real data recorded digitally by modern seismic arrays, was applied to scanned images of paper seismograms recorded from 1966 to 1977 at an observatory in southwest Japan. A list of tremors was successfully obtained; however, further training using data that accounts for variables such as inconsistent plotting pen thickness is required to develop a universally applicable model. Convolutional neural network model for detection of deep low‐frequency tremors from seismogram images is proposedThe model trained with seismogram images converted from real seismic data successfully detects tremorsThe detection performances of the trained model for the paper records at the Kumano observatory in southwest Japan are discussed Convolutional neural network model for detection of deep low‐frequency tremors from seismogram images is proposed The model trained with seismogram images converted from real seismic data successfully detects tremors The detection performances of the trained model for the paper records at the Kumano observatory in southwest Japan are discussed

The geomagnetic jerk amplitudes, which are defined as abruptness of changes in the trends of geomagnetic time series, are investigated with geomagnetic monthly means computed from hourly mean values at each local time. A statistical time series model in which the trend component is expressed by a second order spline function with variable knots is constructed for each time series. The optimum parameter values of the model including positions of knots are estimated by the maximum likelihood method, and the optimum number of parameters including the number of knots are determined based on the Akaike Information Criterion (AIC). The jerks are detected objectively and automatically by regarding the optimized positions of knots as the occurrence epochs. This analysis reveals that the spatial distributions of jerk amplitudes essentially do not depend on the local time, which indicates that the jerks cannot be explained by abrupt changes in intensities of latitudinally flowing external currents such as the field-aligned currents. Longitudinally flowing currents, on the other hand, such as the ring current could explain the distributions. The abrupt changes of the ring current intensity are estimated from the distributions of jerk amplitudes in the eastward component in 1969, 1978, and 1991 supposing that an abrupt change in the ring current intensity causes a jerk. However those estimated changes cannot consistently explain the distributions of the jerks in the northward and downward components. Therefore it is plausible that the jerks which occurred in 1969, 1978, and 1991 are not caused by external sources but internal ones. It is also confirmed that the occurrence epochs of jerks in the southern hemisphere are a few years after those of the 1969 and 1978 jerks in the northern hemisphere, and it is also found that the jerk in the southern hemisphere occurred a few years after the occurrence of the 1991 jerk in Europe. Taking these time lags in occurrence epochs into account, it can be said that the 1969, 1978, and 1991 jerks are global phenomena.

Graphical Abstract:

Not Available