- Ground-motion model is:
Sa is in cm∕s2, a = -1.1682, b = 1.0821, c = -0.04182, d = -0.0052, e = 0.00236, S1 = 0, S2 = -0.18257,
S3 = -0.08882, S4 = -0.19547, S5 = -0.32286, S6 = -0.15611, S7 = -0.16570 and σ = 0.38438.
- Use 7 site classes defined based on predominant period (Tg) of horizontal-to-vertical response spectral ratios
(5% damping) following Zhao et al. (2006) and Fukushima et al. (2007c) with the addition of 3
- Tg < 0.2s. E.g. NCR. 22 stations, 155 records.
- 0.2 ≤ Tg < 0.4s. E.g. ASSI. 21 stations, 113 records.
- 0.4 ≤ Tg < 0.6s. E.g. CHT. 18 stations, 79 records.
- Tg ≥ 0.6s. E.g. RTI. 20 stations, 91 records.
- Tg not identifiable (flat H/V and amplitude < 2). E.g GSG. 11 stations, 73 records.
- Broad amplification/multiple peaks for Tg > 0.2s. E.g. MTL. 5 stations, 36 records.
- Unclassifiable. Tg not identifiable (multiple peaks over entire period range). E.g. MAJ. 14 stations,
δj = 1 for site class j and 0 otherwise.
- Data from 214 stations of Italian accelerometric network of Department of Civil Protection plus 4%
of records from broadband velocity sensors converted to acceleration, which helps constrain model for
r > 100km.
- Remove data with suspected soil-structure interaction. Examine time-histories and Fourier amplitude
spectra to confirm strongest part of shaking well recorded and that signal-to-noise ratio > 3 for
0.5 ≤ f ≤ 20Hz.
- Focal depths for largest events from 5km to 32km with one event at 46km (included since consistent with
shallower events). Some small events have shallower depths.
- About half of data from Friuli, Irpinia, Umbria-Marche and Molise sequences.
- Remove pre-event mean from record and then filter using 4th-order acausal high-pass Butterworth filter
after cosine tapering and zero-padding. Do not low-pass filter since sites do not have low κ and hence
short-period spectra not controlled by high frequencies.
- Most data have Mw < 5.5. Data distribution reasonable uniform w.r.t. site class.
- Find that applying M-R filter to remove distant data from small events has limited impact so do not apply
- Also consider subset of 111 stations that recorded > 1 event.
- Note that purpose of model is to check the impact of new site classification not to propose a model for
- Find decrease in σ w.r.t. rock (Eurocode 8 classes A, B and E, 91 stations, 492 records)/soil (Eurocode 8
classes C and D, 20 stations, 110 records) classification is small. Prefer proposed classification because it
recognises well-distinguished behaviour of the proposed classes.
- Examine match between predictions and observations for 2009 L’Aquila (Mw6.3) earthquake and find