Spatial Prediction of Aftershocks Triggered by a Major Earthquake: A Binary Machine Learning Perspective

Abstract

Small earthquakes following a large event in the same area are typically aftershocks, which are usually less destructive than mainshocks. These aftershocks are considered mainshocks if they are larger than the previous mainshock. In this study, records of aftershocks (M > 2.5) of the Kermanshah Earthquake (M 7.3) in Iran were collected from the first second following the event to the end of September 2018. Different machine learning (ML) algorithms, including naive Bayes, k-nearest neighbors, a support vector machine, and random forests were used in conjunction with the slip distribution, Coulomb stress change on the source fault (deduced from synthetic aperture radar imagery), and orientations of neighboring active faults to predict the aftershock patterns. Seventy percent of the aftershocks were used for training based on a binary (“yes” or “no”) logic to predict locations of all aftershocks. While untested on independent datasets, receiver operating characteristic results of the same dataset indicate ML methods outperform routine Coulomb maps regarding the spatial prediction of aftershock patterns, especially when details of neighboring active faults are available. Logistic regression results, however, do not show significant differences with ML methods, as hidden information is likely better discovered using logistic regression analysis.

Application of Artificial Intelligence in Predicting Earthquakes: State-of-the-Art and Future Challenges

“…Acc. was 90% (decision tree regressor) [111] RF, ANN, SVM EMSC and ISC-GEM P 0 , P 1 , Sn, Sp, MCC RF and SVM achieved the best result [112] DT, RF, ROT Forest USGS F-Measure, Sn, Sp F-Measure was 92.8% (ROT tree) [113] RF, NB, KNN, SVM Acc. Acc.…”

“…[111] RF, ANN, SVM Sixty different seismic features. [112] DT, RF, ROT Boost, ROT Forest 51 features from GR law, and energy release [113] RF, NB, KNN, SVM cscsf., onaf., slip distribution [114] LR, AR, DT, Rep Tree, MLP dpt., local mg., lo., la., local time, date the earthquake catalog. They collected earthquake catalog of 1932 to 2016 for the southern Californian region and calculated features for a range of half months.…”

“…Karimzadeh et al [113] used ML classifiers along with Coulomb stress change, stress distribution, and fault map to predict the aftershock of the Kermanshah earthquake. They treated this problem as a binary classification problem and predicted the location of aftershock with magnitude greater than 2.5.…”

“…The ANN-based methods [80], [81] achieved relative accuracy of 16.24% and 0% respectively. The ANN-SVM model [95] and the RF-based method [113] provided 6.59% and 12.33% relative accuracy, respectively. These models are reported as low performing methods based on relative accuracy.…”

Earthquake Prediction Using Expert Systems: A Systematic Mapping Study

“…Reference [36] studied the thermal anomalies that happened before the earthquake occurred in Imphal, India, in 2016 and investigated multiple seismic facts through satellite data using machine learning algorithms for an earthquake. Reference [37] collected records of aftershocks of the Kermanshah (Iran) Earthquake and applied different machine learning (ML) algorithms, including Naive Bayes, k-nearest neighbors, a support vector machine, and random forests to predict future earthquakes by observing aftershock patterns. Reference [38] exercised neural networks for earthquake signal detection.…”