Uncertainties in ground motion prediction in probabilistic seismic hazard analysis (PSHA) of civil infrastructure - Seismic Risk Analysis and Management of Civil Infrastructure Systems - page 32

age is reasonably dense, this is sometimes the result of combining data from
very different regions. This is a common practice.
For a hazard or risk analysis one ideally desires a model that provides
the distribution of ground motions that would be experienced at a particular
site from earthquake scenarios in the vicinity of this site. Aside from sce-
narios corresponding to very small earthquakes, it is generally not possible
to develop such a model. As a result, model developers invoke the ergodic
assumption in which the lack of recordings that have been made for a par-
ticular site is compensated for by assuming that recordings made in similar
regions elsewhere are representative. This ergodic assumption is often dis-
cussed in terms of replacing a lack of temporal information with spatial
information. The assumption hinges upon the equivalence of motions from
region to region.
Models for shallow active crustal regions refl ect motions from earth-
quakes occurring in the seismogenic portion of the crust (typically within
the uppermost 12-15 km of the crust). These models are also associated
with regions in which the active deformation associated with tectonic pro-
cesses causes the crust through which waves propagate to be highly
deformed. This deformation results in heterogeneities within the propaga-
tion medium that cause scattering and attenuation of relatively high-
frequency components of ground motion. The net result is that high-frequency
spectral amplitudes observed in active regions of shallow seismicity tend to
be lower than those in stable continental regions. Here, the explanation is
given in terms of differences in the propagating medium, but presently this
should be interpreted as relating to both the path attenuation and the near
surface attenuation (often regarded distinctly as a component of site
response).
Models for stable continental regions
Conceptually, the approach that one would adopt when developing ground-
motion models for stable continental regions is the same as that for shallow
crustal regions. However, the stability of these regions is linked to the fact
that they have lower rates of seismicity and hence smaller datasets of
strong-motion recordings associated with earthquake scenarios of engi-
neering interest. In fact, while the magnitude-distance distribution shown
in Fig. 2.1 features numerous regions where observations are sparse, the
situation is far worse in the case of stable continental regions. This issue is
illustrated in Fig. 2.2 in which the magnitude-distance distribution of record-
ings within the database of Atkinson & Boore (2006) is shown.
Additionally, it is clear from Figure 2.2 that there is a very clear bias
towards recordings of small magnitude events at relatively large source-to-
site distances. This refl ects both the fact that the locations of future events