On computation of conditional mean spectrum in Eastern Canada

Abstract

This paper investigates the main ingredients required to compute Conditional Mean Spectra (CMS) in Eastern Canada and assesses their effects on the obtained CMS. We particularly address the influence of ground motion prediction equations (GMPEs) and correlations between spectral accelerations. CMS are computed using two approximate methods, and the results are illustrated for three locations with different seismic hazard and risk levels. It is found that selection of GMPEs considerably influences the CMS, particularly at shorter periods. A database of historical records from Eastern Canada is studied to obtain correlation coefficients. The results suggest higher spectral correlations than predicted by a model based on ground motions from Western North America (WNA). The sensitivity of correlation coefficients to magnitude and epicentral distance is also verified, revealing that magnitude has a more significant effect on these coefficients than distance. We also show that the effect of magnitude- or distance-based correlation coefficients on the CMS is (1) generally negligible at long periods and (2) significant at shorter periods particularly when the conditioning period is less than approximately 0.5 s. This work is the first study addressing in detail the ingredients and construction of CMS in Eastern Canada. The methodology and results discussed are expected to enhance the application of CMS in this region.