- Ground-motion model is:
- Response parameter is displacement for 5% damping.
- Use Eurocode 8 site classes to characterise sites:
- Most data from Swiss stations. SB = SC = SD = 0.
- Few data from Swiss stations. SB = 1.
- Few data from Swiss stations. SC = 1.
- Few data from Swiss stations. SD = 1.
Also derive model using V s,30 (available for 87% of records) with term bv log 10(V s,30∕V a) (and term where V a
constrained to 800m∕s to ensure consistency between approaches) replacing site terms using EC8 classes. Use
weighted regression on residuals to derive this term.
Also derive model using V s,QWL, where V s,QWL is the quarter-wavelength velocity, and term
bv,QWL log 10(V s,QWL∕V a,QWL), where V a,QWL is taken from generic Swiss velocity profile.
- Update of Faccioli et al. (2010) (see Section 2.327) by extending data below Mw5 by adding Swiss records.
Also add data from 2009 L’Aquila (Italy), 2010 Christchurch (New Zealand), 2002 Denali (USA) and 2008
Wenchuan (China) events. Majority of data from Mw < 4.5 is from Switzerland (with some from Italy
and Japan) and majority of data for higher Mw from elsewhere (mainly Japan). Gap in Swiss data for
3.8 ≤ Mw ≤ 4.3.
- All data from digital instruments (mainly 24-bit). Also use Swiss data from on-scale broadband velocity
sensors to improve M-R range of rock site data.
- Process records by removing pre-event offset from entire time-history or filter with high-pass 4th-order
acausal (fc = 0.05Hz) after cosine tapering and zero padding. For low-magnitude data main noise source is
microseismal peaks. Find for 2 < T < 20s and records not dominated by microseismic noise that spectral
displacements are independent of correction technique and that spectral displacements remain constant
after T > 2s. Hence uniformly filter Mw < 4.5 data with 4th-order acausal (fc = 0.5Hz) filter so that
number of records does not change with period.
- Find inclusion of r2Mw term has no impact on σ but required to stabilise regressions for T > 5s. Find
that when using only strong-motion part of database that had to assume r2 = 0 for T > 2s. Believe this
explained by increase in M dependency for weak-motion data because of, e.g., decrease in stress drop for
Swiss earthquakes for Mw < 4.
- Did not include anelastic term because of limited distance range and long periods of interest. Note residual
plots confirm this assumption a posteriori.
- Derive model using entire database and only part from Mw > 4.4. Find similar predictions. Moderate
changes only for rock sites and near-field predictions for class D sites.
- Find higher residuals for weak-motion contribution to database, especially at long periods.
- Compare predictions and observations for some records not used to derive model. Find some large
differences, which relate to site effects.
- Find that using V s,30 or V s,QWL results in similar predictions and σ.