We have developed a new method for estimating azimuth in local three- component seismograms through wavelet analysis. The proposed process proceeds in three stages. First, the seismogram is filtered through a wavelet packet approach. Second, the first arrival is determined through a P picker similar to a short-term- average long-term-average scheme that is applied on the wavelet domain. Finally, an adaptive-length window around the pick is selected and used for determining the azimuth, using the property of linear polarization of the first arrival.
The proposed method has been applied to three-component short-period seismograms for local earthquakes recorded by the seismic network of Alicante province in southeastern Spain. The locations of these events were previously obtained through the software HYPO71PC (Lee and Valdes, 1989) and the seismograms recorded by four analog stations of vertical component distributed within the province of Alicante. The results obtained by the wavelet-based algorithm have been compared with the azimuth angles obtained through the results from the location software. The comparison indicates that the proposed algorithm can determine the azimuth of the analyzed events to within a mean bias of 4.5°.