- Ground-motion model is:
where PGA is in g; c1 = -3.279, c2 = 1.035, c3 = -1.651, c4 = 0.152, c5 = 0.623 and σ = 0.651 for
hanging wall and rock sites; c1 = -3.232, c2 = 1.047, c3 = -1.662, c4 = 0.192, c5 = 0.630 and σ = 0.652
for footwall and rock sites; c1 = -3.248, c2 = 0.943, c3 = -1.471, c4 = 0.100, c5 = 0.648 and σ = 0.628
for hanging wall and soil sites; and c1 = -3.218, c2 = 0.935, c3 = -1.464, c4 = 0.125, c5 = 0.650 and
σ = 0.630 for footwall and soil sites.
- Use 2 site classes:
- Site classes B and C of Lee et al. (2001)
- Site classes D and E of Lee et al. (2001)
- Visually inspect baseline-corrected records and exclude poor-quality and questionable data and those where
the instrument resolution is too low.
- Classify records into:
- Defined as overlying side of a dip-slip fault, which assumed to extend to 30km from and beyond the
end of the fault line within 30∘ from normal to fault strike
- Defined as underlying side of dip-slip fault, which assumed to extend within 40km from and beyond
the end of the fault line within 30∘ from normal to fault strike
- Other sites
Only Chi-Chi and some of the foreign events include records from hanging wall.
- Lack of Taiwanese data from 6.5 ≤ Mw ≤ 7.6 and hence include near-source foreign data from similar
geotectonic environments (Iran, USA and Canada).
- Most earthquakes are reverse or reverse-oblique faulting and hence do not consider mechanism.
- Assume ‘total magnitude saturation’ and hence assume that c5 = -c2∕c3.
- Compare observations and predictions for two earthquakes (Mw6.05 and Mw7.6) and find good fit.
- Examine residuals with respect to distance and as histograms and find no trends and find that residuals
follow lognormal distribution.