Spatial correlation of ground motions in estimating seismic hazards to civil infrastructure - Seismic Risk Analysis and Management of Civil Infrastructure Systems - page 63

Correlation
distance 25 km

Correlation
distance 5 km

No site-to-site
correlation

90

Residuals
1. 2
1. 0
0.8
0.6
0.4
0.2

80

70

60

50

0.0

40

-0.2
-0.4
-0.6
-0.8
-1.0
-1.2

30

20

10

10 20 30 10 20 30 10 20 30

Distance, km Distance, km Distance, km

3.1 Examples of distribution of strong-motion residuals (PGA, ln g,
σ T 0.5) generated for various correlation distances (25 and 5 km) and

for the case of spatially uncorrelated ground motion (all the variability
is within-earthquakes).

=

structures (lifelines). Therefore an important part of the estimations is
selection of proper correlation model.
Several modern ground-motion attenuation equations (e.g. New Genera-
tion Attenuation (NGA) models; Power et al. , 2008) allow the recognition
of the between-earthquake correlation, because the equations include spec-
ifi cation of the between-earthquake and within-earthquake components of
variability (Boore et al. , 1997; Douglas, 2003, 2006; Tsai et al. , 2006; Abraha-
mson et al. , 2008). Usually, within-earthquake standard deviation is found
to be larger than that for between-earthquake standard deviation, although
the relation depends on the vibration period (e.g. Bommer et al. , 2003;
Bommer and Crowley, 2006). Atkinson (2006) and Morikawa et al. (2008)
showed that the total standard deviation of ground-motion prediction may
be reduced when using a single site-specifi c model or a region-specifi c cor-
rection factor, which is determined by grouping ground-motion data at
specifi c stations of a dense strong-motion array. Tsai et al. (2006) showed
that a reduction of the total standard deviation could be achieved if the
path effect could be specifi ed. This reduction, obviously, is related to the