MNMakoto Naoi

Makoto Naoi

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Makoto Naoi has a total of 114 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    PhaseNeXt: Neural phase picker trained on 20-year records to process the JMA-unified data set

    by Makoto Naoi, Kengo Shimojo, Koji Tamaribuchi

    To develop a higher-quality seismic event catalog from Japan’s routine seismic observations, which have been continuously recorded at approximately 2000 stations, we trained deep-learning-based phase pickers using 20 years of arrival-time data read by the Japan Meteorological Agency (JMA). To enhance performance, we developed a new model, PhaseNeXt, by incorporating techniques proven effective in the field of computer vision—particularly in semantic segmentation—into PhaseNet, one of the most widely used neural phase pickers. The resulting model adopts an architecture inspired by DeepLab v3+, connecting parameter-efficient ConvNeXt blocks through residual connections, which mitigate vanishing gradients and allow scalable adaptation to larger training data sets. Furthermore, by including automatically read or briefly reviewed arrival-time readings that had not undergone detailed manual inspection in the training process, we demonstrated improved performance for small earthquakes. Using these insights, we trained three deep neural network models with different parameter sizes on 25 million waveforms associated with events listed in the 2002–2023 JMA unified catalog. When integrated into the current JMA workflow, the best-performing model detected approximately 3.5 times more events than those listed in the JMA catalog while using nearly the same number of arrival-time readings.

  • Open Access

    High-resolution monitoring of hydraulically induced acoustic emission activities using neural phase picking and matched filter analysis

    by Makoto Naoi, Shiro Hirano, Youqing Chen

    Monitoring the activities of very small seismic events or acoustic emissions (AEs) by estimating their hypocenters is useful in investigating fracturing processes in laboratory experiments. Here, we proposed an analysis procedure to develop high-quality AE event catalogs using deep learning and similar waveform searches from the continuous records of AE sensors. The proposed routine comprised the following five steps: 1) automatically developing catalogs using a conventional procedure, where the short-term average-to-long-term average ratio detects transient signals, and arrival times are identified using an autoregressive model and the Akaike information criterion; 2) training a deep learning model for arrival time reading (neural phase picker) using datasets based on the Step 1 catalog; 3) reproducing the AE catalog by applying the trained neural phase picker to continuous waveform records; 4) applying template matching to continuous waveform records based on the template events listed in the catalog in Step 3; and 5) determining the precise hypocenters of template events and newly detected events in Step 4 using a relative location method based on the cross-correlation travel time reading technique. We applied this procedure to continuous AE waveforms recorded at 10 MHz sampling during hydraulic fracturing experiments, resulting in a catalog with 10 times the number of events compared to the Step 1 catalog. This reproduced catalog revealed new aspects of the fracturing process, such as the propagating fracture front and tremor-like AE activity. The proposed procedure eliminates the need for manual labeling, thereby facilitating a fully automated analysis of the observed continuous records. This technique is expected to enhance our understanding of AE sensor records in laboratory experiments.


  • Open Access

    Neural phase picker trained on the Japan meteorological agency unified earthquake catalog

    by Makoto Naoi, Koji Tamaribuchi, Kengo Shimojo, Shinya Katoh, Shukei Ohyanagi

    As Japan is one of the most seismically active countries, seismic data from various institutions are shared in real time and made accessible via the Web to promote research. The Japan Meteorological Agency (JMA), in collaboration with the Ministry of Education, Culture, Sports, Science, and Technology, processes these data to compile a 'unified earthquake catalog' for use in the development of strategies for disaster prevention and public safety. Based on manual arrival-time measurements provided by the JMA, we retrained PhaseNet, the deep-learning phase picker, known as neural phase picker that has gained prominence in recent years, to promote the development of high-quality seismic catalogs in Japan. We utilized the PhaseNet architecture for our model and trained it using 6.1 million three-component seismic waveforms collected in 2014–2021. The performance of the original PhaseNet model, trained with data from California, was suboptimal when applied to routine Japanese data, particularly ocean-bottom seismometer records. Retraining the model with the JMA unified catalog and corresponding waveforms significantly enhanced its performance in picking the arrival times of regular and low-frequency earthquakes. Compared with the original PhaseNet, the dependency of the model on the type of seismic station was reduced by retraining and its performance for waveforms was improved even from stations not included in the training data set. The model performance varied with earthquake magnitude, highlighting the reliance on extensive data for small events in the training set. Compared with the conventional procedure, the model identified numerous events, particularly smaller ones with undetermined magnitudes when integrated into the routine automatic processing of the JMA. Furthermore, leveraging approximately ten times more training data than the California data set, we developed and trained PhaseNetWC, doubling the number of filter channels in each convolutional layer in comparison with those of the original PhaseNet. This modified phase picker surpassed the performance of its predecessor. The dissemination of these models is anticipated to enhance the analysis of routine observational data sets in Japan. Graphical Abstract


  • Open Access

    Neural Phase Picker Trained on the Japan Meteorological Agency Unified Earthquake Catalog

    by Makoto Naoi, Koji Tamaribuchi, Kengo Shimojo, Shinya Katoh, Shukei Ohyanagi

    In Japan, one of the most seismically active nations, seismic data from various institutions are shared in real-time and made accessible via the web, which facilitates research by numerous scholars. The Japan Meteorological Agency (JMA), in collaboration with the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), processes these data to compile a “unified earthquake catalog.” This catalog is crucial for developing disaster prevention strategies and significantly enhances societal safety. To facilitate the efficient development of high-quality seismic catalogs from this data, we retrained a deep-learning phase picker, known as a neural phase picker, which has gained prominence in recent years. This retraining was based on manual arrival time measurements provided by the JMA. We utilized the PhaseNet architecture for our model and trained it using 6.1 million three-component seismic waveforms collected between 2014 and 2021. When the original PhaseNet model, which was trained with data from California, was applied to routine Japanese data, its performance was suboptimal, especially with ocean-bottom seismometer records. However, retraining the model with the JMA-unified catalogs and corresponding waveforms significantly enhanced its performance in reading arrival times of natural and low-frequency earthquakes. The retrained model reduced its dependency on the types of stations that were monitored when applying the original PhaseNet and displayed improved performance for waveforms even from stations not included in the training dataset. The performance of the model varied with earthquake magnitude, which highlights the reliance on extensive data on small events in the training set. When integrated into the automatic processing used in the routine operation of the JMA, the model identified a larger number of events, especially smaller ones with undetermined magnitudes, compared to the events recorded by the conventional procedure. Furthermore, leveraging approximately ten times more training data than the California study, we developed and trained PhaseNetWC, which is a model with greater expressiveness than the original PhaseNet. This new model surpassed the performance of its predecessor. The publication and dissemination of these models are anticipated to enhance the analysis of routine observational datasets in Japan.

  • Top Cited Papers
    Open Access

    Recent advances in earthquake seismology using machine learning

    by Hisahiko Kubo, Makoto Naoi, Masayuki Kano

    Given the recent developments in machine-learning technology, its application has rapidly progressed in various fields of earthquake seismology, achieving great success. Here, we review the recent advances, focusing on catalog development, seismicity analysis, ground-motion prediction, and crustal deformation analysis. First, we explore studies on the development of earthquake catalogs, including their elemental processes such as event detection/classification, arrival time picking, similar waveform searching, focal mechanism analysis, and paleoseismic record analysis. We then introduce studies related to earthquake risk evaluation and seismicity analysis. Additionally, we review studies on ground-motion prediction, which are categorized into four groups depending on whether the output is ground-motion intensity or ground-motion time series and the input is features (individual measurable properties) or time series. We discuss the effect of imbalanced ground-motion data on machine-learning models and the approaches taken to address the problem. Finally, we summarize the analysis of geodetic data related to crustal deformation, focusing on clustering analysis and detection of geodetic signals caused by seismic/aseismic phenomena.Graphical Abstract


  • Open Access

    Efficient similar waveform search using short binary codes obtained through a deep hashing technique

    by Makoto Naoi, Shiro Hirano

    SUMMARY: A similar waveform search plays a crucial role in seismology for detecting seismic events, such as small earthquakes and low-frequency events. However, the high computational costs associated with waveform cross-correlation calculations represent bottlenecks during the analysis of long, continuous records obtained from numerous stations. In this study, we developed a deep-learning network to obtain 64-bit hash codes containing information on seismic waveforms. Using this network, we performed a similar waveform search for ∼35 million moving windows developed for the 30 min waveforms recorded continuously at 10 MHz sampling rates using 16 acoustic emission transducers during a laboratory hydraulic fracturing experiment. The sampling points of each channel corresponded to those of the 5.8-yr records obtained from typical seismic observations at 100 Hz sampling rates. Of the 35 million windows, we searched for windows with small average Hamming distances among the hash codes of 16 channel waveforms against template hash codes of 6057 events that were catalogued using conventional autoprocessing techniques. The calculation of average Hamming distances is 1430–1530 times faster than that of the corresponding network correlation. This hashing-based template matching enabled the detection of 23 462 additional events. We also demonstrated the feasibility of the hashing-based autocorrelation analysis, where similar event pairs were extracted without templates, by calculating the average Hamming distances for all possible pairs of the ∼35 million windows. This calculation required only 15.5 h under 120 thread parallelization. This deep hashing approach significantly reduced the required memory compared with locality-sensitive hashing approaches based on random permutations, enabling similar waveform searching on a large-scale data set.


  • Open Access

    Efficient similar waveform search using short binary codes obtained through a deep hashing technique

    by Makoto Naoi, Shiro Hirano

    A similar waveform search plays a crucial role in seismology for detecting seismic events, such as small earthquakes and low-frequency events. However, the high computational costs associated with waveform cross-correlation calculations represent bottlenecks during the analysis of long, continuous records obtained from numerous stations. In this study, we developed a deep-learning network to obtain 64-bit hash codes containing information on seismic waveforms. Using this network, we performed a similar waveform search for ~35 million moving windows developed for the 30 min waveforms recorded continuously at 10 MHz sampling rates using 16 acoustic emission transducers during a laboratory hydraulic fracturing experiment. The sampling points of each channel corresponded to those of the 5.8-year records obtained from typical seismic observations at 100 Hz sampling rates. Of the 35 million windows, we searched for windows with small average Hamming distances among the hash codes of 16 channel waveforms against template hash codes of 6057 events that were catalogued using conventional autoprocessing techniques. The calculation of average Hamming distances is 1000 times faster than that of the corresponding network correlation. This hashing-based template matching enabled the detection of 23,462 additional events. We also demonstrated the feasibility of the hashing-based autocorrelation analysis, where similar event pairs were extracted without templates, by calculating the average Hamming distances for all possible pairs of the ~35 million windows. This calculation required only 15.5 h under 120 thread parallelisation. This deep hashing approach significantly reduced the required memory compared with locality-sensitive hashing approaches based on random permutations, enabling similar waveform searching on a large-scale dataset.

  • Source parameter estimation of acoustic emissions induced by hydraulic fracturing in the laboratory

    by Makoto Naoi, Keiichi Imakita, Youqing Chen, Kazune Yamamoto, Rui Tanaka, Hironori Kawakata, Tsuyoshi Ishida, Eiichi Fukuyama, Yutaro Arima

    The hydraulic fracturing technique is used for resource production, such as in shale gas/oil extraction and enhanced geothermal systems. The effects of fracturing are often monitored via induced earthquakes, and obtaining as much information as possible from those earthquakes is desirable. The stress drop—calculated from the seismic moment Mo and corner frequency fc—is an earthquake-related parameter that can help identify additional characteristics of the seismicity. To investigate the relationship between stress drops and hydraulically induced seismic events, we estimated the Mo and fc of acoustic emission (AE) events during hydraulic fracturing experiments performed in the laboratory in previous studies using 2 Eagle Ford shale and 10 Kurokami-jima granite samples. We estimated Mo by fitting the theoretical spectra to the observed spectra after correcting for the following effects: (1) frequency response of AE transducers under the installation method used in the fracturing experiment, including differences in sensitivity across every transducer used in each experiment; and (2) the difference in radiation pattern coefficients, which depends on the focal mechanisms of each AE event. This analysis used 46 857 focal mechanisms obtained from moment tensor solutions estimated using a deep learning technique. The range of the resultant Mo was found to be 2.8 × 10–5 ≤ Mo ≤ 4.5 × 10–1 [N·m], corresponding to −9.1 ≤ Mw ≤ −6.3, where Mw is the moment magnitude. We also estimated fc using the multiple-empirical Green's function method, reducing the influence of modelling errors in the AE sensor response and transfer function of the medium. Out of the 1053 events whose Mo and fc were estimated, 465 events (44.2 per cent)—regardless of their focal mechanisms—were found to have Mo and fc values consistent with the constant stress drop scaling of shear failure (i.e. shear failures have 0.1–100 MPa stress drops independent of their magnitude) that has been repeatedly confirmed in many previous studies. The remaining events showed lower fc values than those expected from the scaling law. This indicates that high pore pressure in a source region induced by fluid stimulation contributes to the occurrence of low-frequency earthquakes. Overall, we demonstrated that source parameter estimation was possible for laboratory AEs induced by hydraulic fracturing, which can improve our understanding of the characteristics of fluid-induced earthquakes.


  • Open Access

    Stress State in the Source Region of Mw2.2 Earthquake in a Deep Gold Mine in South Africa Determined from Borehole Cores

    by Yasuo Yabe, Shuhei Abe, Gerhard Hofmann, Dave Roberts, Halil Yilmaz, Hiroshi Ogasawara, Takatoshi Ito, Akio Funato, Masao Nakatani, Makoto Naoi, et al.

    In December 2007, an Mw2.2 earthquake occurred in a gabbroic dike at 3.3 km depth in a deep gold mine in South Africa. The fore- and aftershock activity was analyzed in an effort to understand the preparation and generation processes of earthquakes; these findings have already been published. The present paper focuses on the stress state in the source region of the mainshock. A 90-m-long borehole across the mainshock fault was drilled ~ 1.5 years after the mainshock and logged by an optical televiewer. The fault can be identified by severe damage to the borehole wall at the point where the borehole intersected the aftershock cluster. Except for a 10-m section in the hanging wall of the fault, borehole cores were fully recovered. Borehole breakout (BO) and core disking (CD) were found to occur. Two stress measurement techniques [Deformation Rate Analysis (DRA) and Diametrical Core Deformation Analysis (DCDA)] were applied to the borehole cores. By combining their results with occurrence criteria for BO and CD, the principal stress state in the source region of the mainshock was determined. The principal directions in the hanging wall of the fault were nearly identical to the virgin stress state, while it was significantly disturbed in the footwall. The vertical stresses were 106 MPa and 40 MPa in the hanging wall and footwall, respectively. The significant difference in the vertical stress between the two sides of the fault can be explained by the stress redistribution associated with a nonuniform slip on a nonplanar fault.