Improving locations of earthquakes along the Central Alborz, Iran, using waveform cross-correlation-based time delays

In this study, a complete waveform database of Alborz and its surroundings was processed so as to ameliorate the locations of the earthquakes and obtain an enhanced picture of the past decade’s seismicity distribution.
In the first step of this study, P- and S-wave arrival times were manually re-picked at 41 stations extending from 33° N to 37° N and from 48° E to 54° E. Our initial locations, including 4152 events, were implemented using Jackknife resampling method, normally employed for statistical inference. This up-to-date technique reliably estimates hypocentral errors by deleting one observation at a time. In order to ensure that the relocation would provide valid results, only events that met certain criteria were selected. The selection criteria were (1) largest primary azimuthal gap between stations less than 210°, (2) arrival time residuals less than 1 s, (3) number of recording stations no less than 6, and (4) initial event uncertainty in epicenter and depth of less than 10 km.
The second step of this study focused on improving the arrival time pickings of event pairs utilizing P-wave cross-correlation-based time delays. Correlated events are those occurring within a few kilometers of one another to generate similar waveforms. All event pairs with separation distances less than 10 kilometers were processed. The differential times of event pairs with corresponding travel time residuals for all observations were combined into a system of linear equations and weighed based on the quality of arrival time picks. We computed a total of more than 280000 P-wave differential times and selected waveform pairs with coefficients of 0.7 or larger.
In the third step, to minimize the effect of inaccurate velocity structure, we applied the double-difference location approach. The algorithm, hypoDD, determines relative locations within clusters of closely spaced events using double-difference method developed by Waldhauser and Ellsworth (2000). By relocating merely closely spaced events, this algorithm ameliorates relative location accuracy along with reducing the effects of unmodeled velocity structure. The nearest neighbor approach was applied so as to link events using a maximum search radius of 10 km and a minimum number of 8 links. Event linkage strongly controls how the dataset is broken into clusters for relative relocation in hypoDD. For example, a single link between two closely spaced events, but perhaps occurring along different faults, causes all linked events to collapse into a single cluster rather than forming two clusters. Because of the relatively small number of stations recording each event and due to the closely spaced known faults in Alborz region, we, instead, visually prescribed cluster identification. In this way, we used such essential documentary sources as seismotectonic maps, the hypocenter locations of seismic events in the initial locating procedure, and the expansion of the major faults.
The distribution of 2409 relocated events delineated more coherent features, and in general, the relative relocations increased the agreement with major active faults. The absolute and relative relocations discussed in the present research are an improvement because of either the carefully re-picked P- and S-wave arrival times or the applied appropriate waveform phase-picking algorithm.