Applicability evaluation of Ground Motion Models (GMMs) for Korean Peninsula

Abstract

There are growing opinions that Korea is no longer safe from earthquakes after recent several earthquakes with magnitudes over 5.0 (e.g., moment magnitude 5.4 Gyeongju earthquakes occurred on September 12, 2016). Korean organizations and researchers have been conducting various studies regarding GMMs and selecting GMMs to reduce damage to earthquakes occurring in Korea, but it is necessary to re-evaluate which GMM is most suitable for Korea with recently developed GMMs.

In this study, we evaluate 8 GMMs of the active crustal region, 6 GMMs developed for the stable continental region, and 7 regional developed Korean GMMs. For several GMMs developed without consideration for site effect are applied the site amplification functions (Matsuoka and Midorikawa 1996, and Aaqib et al. 2021), so a total of 31 GMMs are considered. We compute Log-Likelihood differences (LLH; Scherbaum et al. 2009), the multivariate logarithmic score (mvLogS; Mak et al. 2017), the Euclidean distance-based ranking (EDR; Kale and Akkar 2013), the Euclidean metric distance (EMD, Cremen et al. 2020), the deviance information criterion (DIC; Kowsari et al. 2019), and a cumulative-distribution-based area metric (AM; Sunny et al. 2021) values for the response spectra of the 1,009 ground motions (M over 3.5) and GMMs for various periods (0.01, 0.05, 0.1, 0.2, 0.3, 0.5, 1, 2, 5, and 10 s). The un-normalized weight method is selected to consider all six ranking methods. For PGA and Sa, several GMMs developed for Korea, or the SCR region results in the top five models, respectively. Also, weighted GMMs are developed and evaluated to be suitable for use. The weighted GMMs and Eea15 rank at the top overwhelmingly.