An alternative approach to the ground motion prediction problem by a non-parametric adaptive regression method

Abstract

Ground Motion Prediction Equations (GMPEs) are empirical relationships which are used for determining the peak ground response at a particular distance from an earthquake source. They relate the peak ground responses as a function of earthquake source type, distance from the source, local site conditions where the data are recorded and finally the depth and magnitude of the earthquake. In this article, a new prediction algorithm, called Conic Multivariate Adaptive Regression Splines (CMARS), is employed on an available dataset for deriving a new GMPE. CMARS is based on a special continuous optimization technique, conic quadratic programming. These convex optimization problems are very well-structured, resembling linear programs and, hence, permitting the use of interior point methods. The CMARS method is performed on the strong ground motion database of Turkey. Results are compared with three other GMPEs. CMARS is found to be effective for ground motion prediction purposes.

Keywords:

• conic multivariate adaptive regression splines
• ground motion prediction equation
• non-parametric regression
• continuous optimization
• engineering seismology

Acknowledgements

Fatma Yerlikaya-Özkurt is supported by the TUBITAK Domestic Doctoral Scholarship Program. The authors would like to thank Dr Zeynep Gülerce for the processed version of the ground motion dataset.