Site Effects at Tarzana Implied from the Northridge Earthquakes

Yuehua Zeng, Feng Su and John G. Anderson
Seismological Laboratory, University of Nevada - Reno

The January 17, 1994 Northridge earthquake (Mw=6.7) generated one of the 
highest peak accelerations ever recorded on a strong motion instrument
at Tarzana, California, operated by CSMIP.  Located about 6 km south of
the epicenter on a crest of Tarzana Hill, this site has recorded
repeated acceleration over 1g for 7 to 8 seconds, with a peak
horizontal acceleration of about 1.8 g (Shakal et al., 1994).  The same
site also recorded a factor of 10 times higher peak acceleration of the
1987 Whittier-Narrows earthquake, than that were observed at other
sites at similar epicentral distances (Shakal et al., 1988).  This
unusually high peak ground acceleration at Tarzana attracted wide
attention among seismologists and engineers.  Using aftershock data
from a temporary array on the hill, Spudich et al. (1996) studied the
relative site amplifications, topographic effects and possible
polarization effects on the ground motion caused by the hill.  They
found Tarzana Hill has dominant site amplification perpendicular to the
strike of the hill.  While the dominant site response is nearly
north-south, surprisingly, the largest peak ground acceleration during
the Northridge earthquake appeared on the east-west component.  Using a
composite source model, Su et al. (1995, 1996) have simulated the
ground motion at Tarzana and found motions in the east-west direction
are nearly nodal to the source radiation.  By applying the weak motion
site amplification to the synthetic simulation, they found the motions
along the east-west direction are about comparable to the observation.
However, the north-south component has a peak ground acceleration
several times higher than the observation.  We resolve it to a
nonlinear response of the site to the input ground motion.  We assumed
a nonlinear shear modulus reduction relation with strain, which we
believe is suitable to nonlinear rock behavior (Stokoe et al., 1997),
for the shallow surface layer beneath the station, and computed the
nonlinear response of the model to the input ground motion.  This
nonlinear response reduces the acceleration of the north-south
component to a level comparable to the observation.  The same nonlinear
effect has much smaller influence to the east-west component because of
the relative lower level of ground motion.  Thus a model invoking
nonlinear rock behavior in concert with otherwise high site effects
provides a simple explanation of ground accelerations that is
consistent with the well-documented source mechanism.  Drilling
experiment at the Tarzana site found that rock below the surface is
highly fractured (Nigbor et al., 1997), which is consistent with the
strong nonlinear reduction in site effects for the Northridge
earthquake ground motion at Tarzana Hill.