Seismic load estimates of distant subduction earthquakes affecting Singapore

This paper introduces the use of a simple stochastic model for predicting elastic response spectra of 5% damping for structures founded on rock sites in Singapore based on earthquake scenarios of moment magnitude Mw=9−9.5Mw=9−9.5 generated from the Sunda-Arc subduction source at a closest distance of 600 km. Structures founded directly on rock are predicted to be subject to low response spectral accelerations of up to 1.3% gravitational acceleration. The maximum velocity demand on a structure is estimated to be in the order of 50 mm/s. The drift demand on a structure is estimated to be generally low but can be up to some 80 mm on single-degree-of-freedom systems possessing a high natural period of 10 s. Evidence of significant drift demand amplification on flexible soil sites due to the phenomenon of resonance is also presented. The long period ground shaking, when amplified, can be potentially hazardous to certain vulnerable structures such as buildings with a soft-storey and precast beams resting on supports of limited width. The stochastic model was originally developed by the authors based on observations from the magnitude 8 event of 4 June 2000 generated by the Sunda Arc subduction source as reported in an earlier publication by Balendra and others in this journal in 2002. Response spectra simulated by the same model based on the same set of parameters were found to be very consistent with those recorded in Singapore from the M9.3 Aceh earthquake of 26 December 2004, the M8.6 Nias earthquake of 28 March 2005 and the M8.4 earthquake of 12 September 2007 in southern Sumatra. In summary, response spectra generated by these major events in recent times could be simulated by the same model. The scale of events observed on this subduction source and the exceptionally long distance of wave travel (to Singapore) considered in this study are of a unique category. Developing an attenuation model by regressing recorded data is not a viable approach given the scarcity of data recorded from such events and hence stochastic modelling was used.