A Statistical Investigation on Regional Dependency of Strong Ground-Motions in Iran and California

Iran is one of the most seismically active regions in the world with many historical and instrumental recorded earthquakes. Despite such high seismicity, the number of strong ground motion records available for main tectonic regions in Iran is not sufficient for reliable regional-dependent ground motion prediction equations (GMPEs). Moreover, it is generally acknowledged that sparse sampling and poor distribution of data result in limited applicability of GMPEs with regard to source-to-site distance and magnitude ranges. In order to cope with the lack of data, supplementary earthquake ground motions (GM) recorded in certain regions of the world with similar tectonic regimes can be combined into one dataset. Therefore, we aimed to explore the possibility of combining Iran and California datasets through assessment of the regional dependency of strong GMs. For this purpose, the statistical technique known as Analysis of Variance (ANOVA) is applied to both large sets of strong GMs from Iran and California. Analysis of variance does not rely on deriving GMPEs which makes it a strong and reliable technique for understanding of regional differences. California dataset comprises of 5902 records obtained from 218 earthquakes which are extracted from NGA-West2 database. The Iranian database used in the present study was obtained from Iran Strong Motion Network (ISMN) in unprocessed format. Data are processed using modified wavelet de-noising approach by which large numbers of noisy acceleration records that are usually discarded from datasets, can be corrected. The database comprises of 1943 pairs of horizontal acceleration time series obtained from 426 earthquakes with moment magnitudes ranging from 4.5 to 7.4 and distances up to 200 km. As source-to-site distance measure, closest distance to the fault rupture plane (denoted as Rrup) is computed for all records. Due to paucity of normal-faulting earthquakes in Iran, we excluded such events from our database.
The large number of strong motion datasets from both regions available in present research enables us to perform the analysis of variance technique to many intervals with overlapping data from both regions. Consequently, more reliable results can be concluded compared to previous researches. We applied ANOVA to 5%-damped horizontal component of elastic response spectral acceleration at six different vibration periods (0.01, 0.1, 0.3, 0.5, 1, 2 and 5 seconds) and at intervals of to 5 km and 0.2 Mw. In order to combine two horizontal components, a single period-independent rotation measure (GMRotI50) is used.
To provide more reliable results, site response effect on GM amplitudes is reduced using averaged horizontal-to-vertical spectral ratio (HVSR) for stations grouped by Iranian building design code based on time-averaged shear-wave velocity in the top 30 m ( VS30). For stations with unknown VS30 values, the VS30-prediction equation proposed by Darzi et al. (2019) is applied to Iranian strong-motion stations having fundamental frequency greater than 1 Hz. Note that due to small number of stations in class IV, they are combined with class III.
This study first investigates the possibility of classifying the Iranian database into three tectonic regions of Northern Iran, East-Central Iran and Zagros. The ANOVA technique is applied to Mw-Rrup intervals with sufficient data from these regions two by two; however, the results did not show any strong regional-dependency, indicating that the nation-wide data could be considered as a single dataset. Then, the regional differences between GM amplitudes from Iran and California are evaluated using ANOVA technique. High evidence of regional dependency was revealed in some magnitude and distance ranges, supporting regional classification of dataset to prevent biased predictions. Based on these analyses, there is little evidence of regional difference for reverse-faulting events except for Mw = 5.8-6 and Rrup = 25-30 km across all periods and for Mw = 6.2-6.4 and Rrup = 30-35 km at short periods. For strike-slip events, there seems to be three intervals showing high regional dependence of GMs. To conclude, authors believe that supplementary data is required to make a definitive statement in the whole magnitude and distance range. In case of negligible regional differences, such investigation would be beneficial for development of future GMPEs in Iran to overcome incompleteness of data and existence of gaps in distance-magnitude intervals.