Search

In this paper, a scalable end-to-end tree boosting system called XGBoost has been applied for predicting the magnitude of an earthquake from the early part of earthquake waveform data. This model uses the features extracted from the early P wave phase of the records as an input. The model's effectiveness has been verified by using data on earthquakes occurring in the Eurasian plate of Japan Islands from 1996 to 2021. Feature engineering has given 29 new features identified from the early P wave phase of the record, which show a high correlation with the magnitude of an earthquake. The comparison of predicted and actual magnitude shows that a trained XGboost model, which uses a single input record for magnitude prediction, gives an average prediction error of 0.004 ± 0.57 for earthquakes in the test dataset. In contrast, the average prediction error of –1.1 ± 0.80 and –0.65 ± 0.69 has been obtained for the magnitude estimated from conventional τc and Pd methods using the same test dataset. It is further seen that the average predicted magnitude of a single earthquake of magnitude 4.5 and 6.1 (MJMA) obtained by using multiple nearfield records using XGBoost model is 4.58 ± 0.33 and 6.32 ± 0.29, which is close to the actual magnitude of the earthquake. The results presented in this paper clearly show that the structured data can be effectively used by complex machine learning or deep learning models to predict earthquake magnitude from single or multiple records. [ABSTRACT FROM AUTHOR]

A cross-region prediction model named SeisEML (an acronym for Seismological Ensemble Machine Learning) has been developed in this paper to predict the peak ground acceleration (PGA) at a given site during an earthquake. The SeisEML model consists of hybridized models, kernel-based algorithms, tree regression algorithms, and regression algorithms. The model ablation study is conducted to examine the performance and the selection of meta-machine learning models in the SeisEML. The training and testing dataset consists of 20852 and 6256 accelerograms recorded by the Kyoshin Network, Japan. The mean absolute error (MAE) and root mean square error (RMSE) have been utilized to compare the predicted peak ground acceleration (PGA) for the test data. The SeisEML model yields approximately half the MAE and RMSE values obtained with conventional attenuation relations. The SeisEML model has been used to compute Japan's iso acceleration contour map of three earthquakes ( M JMA 7.4, 6.6, and 6.1). The qualitative comparison of iso acceleration contours obtained from actual and predicted PGA using SeisEML clearly shows that the model can reliably predict the PGA distribution during an earthquake compared to the regional ground motion prediction equation (GMPE). The cross-region prediction was performed on the datasets of the Iranian earthquakes using SeisEML. The comparison of predicted and observed peak ground acceleration in terms of MAE and RMSE shows that the machine learning model's performance is superior to the regional attenuation relation. The predictions of PGA from the developed ML model indicate that this trained model has the potential for predicting regional and global scenarios with similar tectonic setups. The ML model developed in this paper can considerably enhance the reliability of PGA prediction for seismic hazard mapping of any region and can serve as a reliable substitute for GMPEs. [ABSTRACT FROM AUTHOR]