Single-station and Small Source Regions GMPE for Site-specific PSHA

Abstract:

Modern ground motion prediction equations (GMPEs) are based on datasets of ground-motion parameters recorded at multiple stations and different earthquakes in various source regions. This would cause excessive ground-motion variability and lead to overestimate of exceedance probability in probabilistic seismic hazard analysis (PSHA). Recent efforts on decomposition for aleatory and epistemic variabilities and moving of the epistemic uncertainty to logic tree reveal that hazard level remains unchanged. To reduce the epistemic uncertainty so that the total variability is reduced in a GMPE becomes critical in the PSHA.

In the present study, a total of 30,602 strong-motion records from TSMIP in Taiwan are selected and used to accomplish the regression analysis of a regional GMPE at the first step. Then, 9 stations in different part of Taiwan, each contains at least 55 records, are selected to complete a single-station GMPE, respectively, at the second step. The results reveal that the sigma of the regional GMPE is 0.626 in ln unit and the sigma of single-station GMPE is ranging 0.416 to 0.567. The single-station sigma is about 9% to 33% reduction from the regional one. At last, source zones are further divided in GMPEs, the sigma of single-station and small source regions GMPEs can be reduced 16% to 36% relative to the regional one.

When PSHA is performed by adopting the single-station sigma, the hazard is 11% to 48% smaller then that regional sigma is used. The result of a PSHA may be further reduced, if this single-station to small source region attenuation relationship is used.