- Ground-motion model is:
_{1}= -1.8559, C_{2}= 0.2549, C_{3}= 0.0111, C_{4}= -0.0013, C_{5}= 0.3061, C_{6}= 0.0734, C_{7}= 0.3552, C_{8}= 1.5149, C_{9}= -0.103 and σ = 0.2137. - Use two site classes because of limited number of records:
- enumerate
- V
_{s,30}≥ 900m∕s, Rock Quality Designation ≥ 50% or compressive strength q_{u}≥ 10MPa. 25 records. Z = 0 - Otherwise. 92 records. Z = 1 following the Chilean seismic code.
- Focal depths (H) from 18 to 53km.
- Use data from 1985 to 2010 obtained from public databases and records from National Accelerograph Network.
- Most records from SMA-1s or similar analogue instruments.
- Most instruments on ground floor of 1-storey buildings (79 different stations).
- Only use data from large interface earthquakes (9 mainshocks and 4 aftershocks). 8 events in north (including two in Peru) and 5 in central area of Chile.
- Locations from Chilean Seismological Service and magnitudes from Global CMT.
- Data well-distributed w.r.t. M
_{w}and r_{rup}. - Two events: 3/3/1985 and 27/2/2010 contribute 27 and 31 records, respectively; about half the data.
- Estimate r
_{rup}using CMT solutions and aftershock distributions assuming that rupture area generally smaller than aftershock area. - Filter (using acausal 4th-order Butterworth) based on visual inspection of Fourier amplitude spectra in
log-log space. Generally use f
_{max}of 90Hz for Nyquist of 100Hz and 40Hz for Nyquist of 50Hz. Use iterative process to find f_{min}based on examination of displacements and displacement response spectra. Seek lowest f_{min}that preserves natural appearance in time domain and without obvious drift. - Coefficients C
_{6}, C_{7}, C_{8}and C_{9}derived using algorithm to minimize mean residual for PGA from r_{rup}≤ 80km (47 records). Fix these coefficients for other periods. - Compare predictions and observations for 8.3 ≤ M
_{w}≤ 9.3 and find good fit. - Plot total residuals w.r.t. distance and find no trends.