Detection of P-wave on broadband seismometer using discrete wavelet denoising

Parametric filters are widely used to determine arrival time of P wave. The disadvantage of its method is phase shifting occurred on filtered seismic signal. Consequently, an arrival time of P wave became less accurate. In this paper, two processing steps are proposed. On the First step, spectral density(PSD) estimation of seismic signal is used for early detection of earthquakes and to detect possible duration of information carrier P-wave. On the second step, on the specific duration of seismic signal that detected carrier P-wave is filtered using discrete wavelet denoising in order to minimize phase shift. To evaluated performance of wavelet denoising, the seismogram that recording a local earthquake in Tasikmalaya 2 September 2009 occurred at sea was used. Next, the seismic signal was segmented with data length in about 25.6 s, resulting 20 data segments. Using first step processing, the segment signal that carrier P-wave is detected on 4th data segment. Next, wavelet basis function of Daubechies was used to decomposition seismic signal into 9th level. The P wave arrival time is detected on wavelet decomposition on detail 2 (d2) components at 2.5-5 Hz frequency range. From this data it is found that P wave arrival time is 85.5 seconds (07.55.25.5 UTC). There is a 0.9 s difference earlier by the method of discrete wavelet filter compared to the Meteorological, Climatological and Geophysical Agency (BMKG) reports. The same process was applied on 10 recordings of aftershocks earthquake with magnitude 3.6 until 5.6 SR and resulting 0.25-0.9 s time range difference earlier than BMKG phase report.