It is well established that sedimentary basins can significantly amplify ea
rthquake ground motion. However, the amplification at any given site can va
ry with earthquake location. To account for basin response in probabilistic
seismic hazard analysis, therefore, we need to know the average amplificat
ion and intrinsic variability (standard deviation) at each site, given all
earthquakes of concern in the region. Due to a dearth of empirical ground-m
otion observations, theoretical simulations constitute our best hope of add
ressing this issue. Here, 0-0.5 Hz finite-difference, finite-fault simulati
ons are used to estimate the three-dimensional (3D) response of the Los Ang
eles basin to nine different earthquake scenarios. Amplification is quantif
ied as the peak velocity obtained from the 3D simulation divided by that pr
edicted using a regional one-dimensional (1D) crustal model. Average amplif
ication factors are up to a factor of 4? with the values from individual sc
enarios typically differing by as much as a factor of 2.5. The average ampl
ification correlates with basin depth, with values near unity at sites abov
e sediments with thickness less than 2 km, and up to factors near 6 above t
he deepest ( approximate to 9 km) and steepest-dipping parts of the basin.
There is also some indication that amplification factors are greater for ev
ents located farther from the basin edge. If the 3D amplification factors a
re divided by the 1D vertical SH-wave amplification below each site, they a
re lowered by up to a factor of 1.7. The duration of shaking in the 3D mode
l is found to be longer, by up to more than 60 seconds, relative to the 1D
basin response. The simulation of the 1994 Northridge earthquake reproduces
recorded 0-0.5 Hz particle velocities relatively well, in particular at ne
ar-source stations. The synthetic and observed peak velocities agree within
a factor of two and the log standard deviation of the residuals is 0.36. T
his is a reduction of 54% and 51% compared to the values obtained for the r
egional 1D model and a ID model defined by the velocity and density profile
below a site in the middle of the basin (DOW), respectively. This result s
uggests that long-period ground-motion estimation can be improved considera
bly by including the 3D basin structure. However, there are uncertainties c
oncerning accuracy of the basin model, model resolution, the omission of ma
terial with shear velocities lower than 1 km/s, and the fact that only nine
scenarios have been considered. Therefore, the amplification factors repor
ted here should be used with caution until they can be further tested again
st observations. However, the results do serve as a guide to what should be
expected, particularly with respect to increased amplification factors at
sites located above the deeper parts of the basin.