- Ground-motion model is: where y is in m∕s2, θ1 = 2.4393, θ2 = -0.1375, θ3 = -2.7372, θ4 = 0.3348, θ5 = 7.6, θ6 = 0.0145,
θ7 = 0.0663, θ8 = -0.0780, θ9 = 0.0517, θ10 = -0.0405, θ11 = -0.0141 and σ = 0.2552 for AAK;
θ1 = 1.6123, θ2 = -0.1329, θ3 = -1.8494, θ4 = 0.2043, θ5 = 7.6, θ6 = 0.1427, θ7 = 0.0489, θ8 = -0.0864,
θ9 = 0.0466, θ10 = -0.0358, θ11 = 0 and σ = 0.2891 for CEI; and θ1 = 2.0872, θ2 = -0.1335,
θ3 = -2.5370, θ4 = 0.3173, θ5 = 7.6, θ6 = 0.0770, θ7 = 0.0578, θ8 = -0.0884, θ9 = 0.0329, θ10 = -0.0283,
θ11 = -0.0082 and σ = 0.2836 for ZM.
- Use 3 site classes:
- SS = SA = 0.
- SA = 1 and SS = 0.
- SS = 1 and SA = 0.
- Use 3 faulting mechanisms:
- AAK: 5 records, CEI: 4 records, ZM: 3 records. FN = 1 and FT = FO = 0.
- AAK: 48 records, CEI: 31 records, ZM: 76 records. FT = 1 and FN = FO = 0.
- AAK: 20 records, CEI: 60 records, ZM: 42 records. FO = 1 and FN = FT = 0.
- Focal depths from 3 to 59km.
- Use data from Iranian Strong-Motion Network (195 different stations) of the Building and Housing Research
Center from 1976 to 2012.
- Select data from earthquakes with Mw > 5 and R < 100km.
Also only select data from stations with known site classification and Mws from Global CMT.
- Data from earthquakes with fault rupture mainly at depths < 15km.
- High-cut filter with roll-offs of 23 (analogue instruments) and 50Hz (digital) and cut-offs of 25 (analogue)
and 100Hz (digital). Baseline correct and low-cut filter to exclude data where signal-to-noise ratio < 2.
Note that predictions for PGA and PSA< 0.1s may be affected by filters.
- Update model of Ambraseys et al. (2005a) using the Bayesian approach of Wang and Takada (2009). This
model was chosen as prior because: previous work ranked this model highly for use in Iran and the data for
evaluation of this model are available (e.g. site classes rather than V s,30). Find Ambraseys et al. (2005a)
overestimates motions in different areas.
- Split Iran into 3 regions: Azarbayejan-Alborz-Kopeh Dagh (AAK, 73 records), Zagros-Makran (ZM, 121
records) and central-east Iran (CEI, 95 records) and derive models for each.
- Note that little data for most important M-R ranges.
- Include Mw2 term to correct for bias outside Mw range of database.
- Do not update θ5 from Ambraseys et al. (2005a) because approach works for linear coefficients.
- Plot normalised residuals w.r.t. Mw and find no trends and almost zero bias.
- Compare predictions and observations for records from earthquakes with 5.5 ≤ Mw ≤ 6.5. Find better
match than for original model.
- Apply approach of Scherbaum et al. (2004) to check fit of model to observations. Find a rank of ‘B’.