A Preliminary Analysis on Empirical Attenuation of Absolute Velocity Response Spectra (1 to 10s) in Japan

The Mw 9.1 Tohoku-oki earthquake caused strong shakings of super high rise and high rise buildings constructed on deep sedimentary basins in Japan. Many people felt difficulty in moving inside the high rise buildings even on the Osaka basin located at distances as far as 800 km from the epicentral area. Several empirical equations are proposed to estimate the peak ground motions and absolute acceleration response spectra applicable mainly within 300 to 500km from the source area. On the other hand, Japan Meteorological Agency has recently proposed four classes of absolute velocity response spectra as suitable indices to qualitatively describe the intensity of long-period ground motions based on the observed earthquake records, human experiences, and actual damages that occurred in the high rise and super high rise buildings. The empirical prediction equations have been used in disaster mitigation planning as well as earthquake early warning. In this study, we discuss the results of our preliminary analysis on attenuation relation of absolute velocity response spectra calculated from the observed strong motion records including those from the Mw 9.1 Tohoku-oki earthquake using simple regression models with various model parameters. We used earthquakes, having Mw 6.5 or greater, and focal depths shallower than 50km, which occurred in and around Japanese archipelago. We selected those earthquakes for which the good quality records are available over 50 observation sites combined from K-NET and KiK-net. After a visual inspection on approximately 21,000 three component records from 36 earthquakes, we used about 15,000 good quality records in the period range of 1 to 10s within the hypocentral distance (R) of 800km. We performed regression analyses assuming the following five regression models. (1) log10Y (T) = c+ aMw - log10R - bR (2) log10Y (T) = c+ aMw - log10R - bR +gS (3) log10Y (T) = c+ aMw - log10R - bR + hD (4) log10Y (T) = c+ aMw - log10R - bR +gS +hD (5) log10Y (T) = c+ aMw - log10R - bR +∑gS +hD where Y (T) is the 5% damped peak vector response in cm/s derived from two horizontal component records for a natural period T in second; in (2) S is a dummy variable which is one if a site is located inside a sedimentary basin, otherwise zero. In (3), D is depth to the top of layer having a particular S-wave velocity. We used the deep underground S-wave velocity model available from Japan Seismic Hazard Information Station (J-SHIS). In (5), sites are classified to various sedimentary basins. Analyses show that the standard deviations decrease in the order of the models listed and the all coefficients are significant. Interestingly, coefficients g are found to be different from basin to basin at most periods, and the depth to the top of layer having S-wave velocity of 1.7km/s gives the smallest standard deviation of 0.31 at T=4.4s in (5). This study shows the possibility of describing the observed peak absolute velocity response values by using simple model parameters like site location and sedimentary depth soon after the location and magnitude of an earthquake are known.