Regional Adjustments to NGA-West2 Ground-Motion Models for Turkey

Abstract

This paper presents a ground-motion model updating (GMMU) framework to adjust NGA-West2 models for predicting a set of intensity measures in Turkey, including peak ground acceleration (PGA), peak ground velocity (PGV), and pseudo-spectral acceleration (PSA) at periods ranging from 0.01 to 10 s. The GMMU framework integrates bias identification and coefficient updating for the NGA-West2 models based on more than 9000 Turkish ground-motion records from over 500 earthquakes during 1977–2023. Only a portion of coefficients mainly related to the small-to-moderate magnitude scaling, distance scaling, and linear site response are adjusted using the regional data, whereas the other well-constrained aspects (e.g., hanging wall and nonlinear site response) are retained. The GMMU is based on a series of crossed-nested mixed-effects regressions incorporating between-event, site-to-site, and event-site-corrected residual terms. The results show that the global NGA-West2 models developed by using less than 70 Turkish records exhibit notable incompatibilities with the current large database, especially for short-to-moderate spectral periods within the earthquake magnitude range of 4.5–6.5. In contrast, the region-adjusted models provide improved estimates of both medians and standard deviations by incorporating the Turkish ground-motion attenuation characteristics. Additionally, a distance- and site-condition-dependent directionality model is developed for Turkey to estimate the maximum intensities over horizontal orientations (RotD100), mitigating the slight overestimation by the NGA-West2 model for periods longer than 0.2 s. The updated models can be applied for shallow crustal earthquakes with magnitudes ranging from 3.5 to 7.8 and distances shorter than 400 km.