Correlation of waveforms from closely spaced regional events

Abstract

Abstract We studied the similarity of waveforms recorded at the high frequency element of the NORESS array from 137 events (ML = 0.8 to 2.1) in or near the mining districts of Central Sweden. Waveform correlations based on the covariance matrices of three component recordings and on maximum amplitudes were calculated from traces filtered between 2.5 and 4.0 Hz, in which band the signal-to-noise ratio consistently peaked for the waveforms. A cutoff value for the correlations to separate waveforms with poor and good correlation values could be defined from the statistical uncertainty in back azimuth estimates of NORESS. More than 80 per cent of the events could be grouped with hierarchical clustering into one large group of 98 events and five smaller groups. The NORESS epicenters of the events in the large group were scattered over an area of 20 × 75 km. An exponential decay of the waveform correlation with event separation (d in km) could be fitted to the data of this group (exp(−d/8), taking NORESS epicenter uncertainties into account. A procedure for location of close events based on correlation values is defined. When applied to the large event group it limits the event epicenters to be within an area of about 4 km. For one small event group, high correlation values were obtained above 15 Hz. These frequencies are significantly higher than those reported in other studies, and the so called quarter wavelength argument constrains the event epicenters to within 0.1 km.