Prediction of large-scale earthquakes with precursor earthquakes: a regression study

Earthquake prediction is a extremely challenging problem in seismology. Despite ongoing efforts understanding the dynamical mechanisms of earthquake occurrence, there is no reliable prediction method. Previous studies found that large earthquake (triggered earthquake, TDE) are associated with several precursor earthquakes (triggering earthquakes; TGEs). However, no quantitative relation was proposed so far. In this study, four quadratic regression equations between the magnitude and distance between TGEs and TDE were acquired; three linear regression equations between the magnitude and time interval between TGEs and TDE were calculated. Based on the regression equations, a prediction method was developed. We verified 87 large earthquakes over the past 100 years with magnitudes greater than 6.6, including the M 7.8 Turkey earthquake in 2023 and the M 7.7Myanmar earthquake in 2025. Although we have been able to predict only 89% of large-scale earthquake epicenters and 63% of occurrence times, earthquakes of M ≥ 9.0 and earthquakes that have caused significant casualties had all met our model with a prediction accuracy of 100 km and 3 days, which is better than most other prediction strategies.Based on this model, we predict a future earthquake in Japan (September 26-27, 2025 in the Southwest of Tokyo).