Item

Abstract

In this study, we developed a new empirical relation between various source and rupture parameters such as moment magnitude (M), surface rupture length (SRL), subsurface rupture length (RLD), rupture width (RW), rupture area (RA), and average (AD) and maximum slip (MD), based on an extensive updated database. A log-linear regression was used to study 402 earthquakes that occurred between 1857 and 2018, ranging in magnitude M~4.5-9.2, with various faulting mechanisms. All fault mechanisms correlate well with M-SRL, M-RLD, M-RW, and M-RA. In contrast, the M-AD and M-MD correlate moderately for reverse faulting, whereas the correlation is relatively better for other fault types. The results were compared with previous studies and found to be revised. On the other hand, the log-linear equations predict a single magnitude value as a function of a single fault parameter independent of the values of other fault parameters, which may lead to inconsistency. This limitation has been addressed by applying machine learning techniques to estimate the earthquake magnitude while simultaneously using all fault parameters to ensure consistency. We also investigated the performance and applicability of an artificial neural network (ANN) and a gradient-boosting machine (GBM) regression technique. Our analysis shows that GBM outperforms regression equations in estimating earthquake magnitude, but ANN outperforms both. The result of this study would be extremely useful for paleoseismic studies where reliable estimates of earthquake magnitudes and other source parameters are often unavailable from published literature.