The strong motion accelerograrns recorded at two stations of SMART-1 array have been analyzed to investigate azimuth-dependent site amplification. Fifteen earthquakes which triggered both accelerographs at rock site E02 and soil site E01 were selected for this study. We measure the spectral ratio of E01 to E02 in a window which includes direct S waves. Since ground motions at the rock site are less modified by the site effects and E02 and E01 are only 2 km apart, the spectral ratio of the soil site to the rock site will cancel most of source and raypath effects and reflect the site amplification of the soil site.
These earthquakes e divided into 10 groups. The e quakes in the s e group have similar epicentral distance and backazimuth. Our studies show that different groups have different spectral ratios and no systematic changes have been found among the groups. In the same group, all earthquakes were expected to have the similar spectral ratio. However, similar spectral ratios were only found in some groups over some limited frequency ranges. These discrepancies are interpreted as the results of near-source d non-linear effects.