Novel nonparametric modeling of seismic attenuation and directivity relationship

Abstract

Seismic attenuation modeling is a core element in earthquake disaster assessment and earthquake early warning. In this paper, a novel nonparametric methodology, namely Bayesian Nonparametric General Regression (BNGR), is introduced for the modeling of peak ground acceleration attenuation relationship. In contrast to most existing methods, this method does not require a prescribed functional form of the attenuation relationship. Furthermore, it selects the proper set of variables necessary to model this relationship. Moreover, the proposed model does not only consider the epicentral distance but also the location of the measured station so it allows for the modeling of the attenuation directivity. The proposed method is demonstrated with a case study using a comprehensive database of ground motion records at 271 monitoring stations for the 2008 Ms 8.0 Wenchuan earthquake. Results show that the proposed methodology is capable to represent the seismic attenuation relationship. Furthermore, the contour plots of the peak ground acceleration provides valuable information for the study of the earthquake directivity characteristics.