- Ground-motion model is (based on Akkar and Çağnan (2010) and the site terms are from Sandikkaya
et al. (2013) with smoothed coefficients): where Y in in g, c1 = 6.75, Δc1 = 0.25, b1 = 1.74221, Δb1 = -0.21234, b2 = 0.193, Δb2 = -0.146,
b3 = -0.07049, Δb3 = -0.03826, b4 = -1.18164, Δb4 = 0.17210, b5 = 0.170, Δb5 = -0.120, b6 = 8.00,
Δb6 = 0.00, b7 = -0.354, Δb7 = 0.396, b8 = -0.01329, Δb8 = -0.11697, b9 = -0.09158, Δb9 = 0.0,
b10 = -0.00156, Δb10 = 0.00156, V REF = 750m∕s, c = 2.5, n = 3.2, V CON = 1000m∕s, sb1 = -0.41997 and
sb2 = -0.28846.
Model for σ is: where a1 = 0.570, Δa1 = 0.120, a2 = 0.450, Δa2 = 0.050, sd1 = 1.0521, Δsd1 = -0.0808, sd2 = 0.7203
and Δsd2 = -0.3250.
- Use V s,30 to characterise sites. Only used data with 150 ≤ V s,30 ≤ 1200m∕s to be consistent with site
terms of Sandikkaya et al. (2013). Turkish sites are mainly NEHRP class C and D (180–760m∕s) and
Iranian sites are mainly NEHRP class B and C (360–1500m).
- Use 3 styles of faulting:
- Reverse. FRV = 1 and FNM = 0.
- Strike-slip. Best represented for both countries. FRV = FNM = 0.
- Normal. Few records from Iran. FNM = 1 and FRV = 0.
- Focal depths ≤ 33m to exclude subcrustal events. Iranian events slightly deeper on average.
- Use data from the database compiled for the Earthquake Model of the Middle East Region (EMME).
Select data with highest waveform and metadata quality. 670 records, 175 earthquakes and 163 stations
from Turkey and 528 records, 138 earthquakes and 254 stations from Iran.
- Use approach of Akkar and Bommer (2006) to process data and choose maximum spectral period.
- Estimate rjb either from reported fault geometries or by using fault extent estimates from equations of
Wells and Coppersmith (1994).
- Exclude earthquakes without true Mw estimates.
- Vast majority of data from Mw ≤ 6.5 and from rjb > 30km. Distribution of data from Turkey is more
uniform than data from Iran, which has few records from Mw < 5 and rjb > 80km.
- Retain singly-recorded earthquakes (62 from Turkey and 50 from Iran).
- Before regression plot PGA scaled to reference rock and binned into distance bins for data from 5 ≤ Mw ≤ 6
and 6 ≤ Mw ≤ 7 against rjb for the two countries. Find evidence for differences.
- Use relatively simple functional form because of available metadata.
- Terms with Δ are to evaluate model for Iran.
- Use pure-error analysis (Draper and Smith, 1981; Douglas and Smit, 2001; Ambraseys et al., 2005a) to
derive weighting function (w) for regression by dividing data into 0.5Mw bins and fitting model within
each bin to find standard deviations. From these standard deviations w was found. Some bins give large
standard deviations, which are ignored when finding w.
- Coefficients b2, b3, b5, b6, b7, b8 and b9 are obtained from data with rjb ≤ 80km. b4 is obtained from all
data. b10 is obtained using only data with rjb ≥ 80km. b1 is found in final step of regression considering
all data. Smooth coefficients after each step to remove jagged variation in estimated response spectra.
- Use inter- and intra-event residual plots to explore potential differences between Turkish and Iranian
- Find that inter-event residuals from Turkey show more dispersion than those from Iran, which does not
disappear if only the best-recorded events are considered. Suggest that this is related to the complex nature
of some Turkish events.
- Find differences in ground motions between Turkey and Iran, which relate to differences in Q, κ,
near-surface V s profiles and focal depth.
- Re-run regressions using only data from Mw > 5 to check if differences in data distribution between
Turkey and Iran are affecting the results. Find that the new model predicts similar motions. Conclude
that findings are not being strongly affected by the differences in data distribution.