HR: 1340h
AN: S43B-1307    [Abstracts]
TI: Trained automatic pickers used to extract the full-recorded seismicity of the ANCORP network
AU: * Nippress, S E
EM: nippress@liverpool.ac.uk
AF: Univeristy of Liverpool, Department of Earth and Ocean Sciences, Jane Herdman Laboratories, Liverpool, L69 3GP, United Kingdom
AU: Rietbrock, A
EM: ariet@liverpool.ac.uk
AF: Univeristy of Liverpool, Department of Earth and Ocean Sciences, Jane Herdman Laboratories, Liverpool, L69 3GP, United Kingdom
AB: The developments in data acquisition, data storage and data transmission technologies in the last decade allow seismologist to collect new data at an unprecedented rate. On large datasets the traditional manual picking requires significant amounts of both man-power and time with sometimes variable results. In order to process these large recorded datasets reliable and accurate automatic event detection and phase picking has become essential. The ANCORP temporary seismic network (34 stations) ran continuously between November 1996 to March 1997 in northern Chile and southern Bolivia. Due to the high seismicity rate in this region only the largest events during the observation period were analyzed (~859 events), thus using only ~25% of the actual recorded seismicity. To fully exploit the full-recorded seismicity in this dataset we use a number of trained automatic picking algorithms. Manual picks are used to train a STA/LTA, a Tp and a statistical automatic picking algorithm at each station and establish the optimum picking parameters that produce the greatest accuracy for each algorithm at each station. The optimum picking parameters for each station show regional variability and have to be adjusted individually to get the best performance. We find that the STA window length is larger in the north (0.5sec compared to 0.2sec in the south) and the LTA window length is shorter to the east (27sec compared to 30sec in the west). The Ds and a parameters for the Tp picker are larger in the west (Ds is 10^-6 rather than 10^-5and a is 10sec rather than 8-9sec), while the trigger level is constant across the network. Using the optimum picking parameters the automatic pickers are set running on the continuous ANCORP continuous dataset. We locate the events initially using HYPO71 before refining the locations in a 3D velocity model. Additionally, local magnitudes for all events based on automatic determined peak-to-peak values are calculated. We find that after using the trained automatic picking and location procedure the number of events located increases from 118 to 408 (a factor of 4 increase) in just a 2 week period. We locate a large number of new events at 60-130km depth range as well as new events in the overlying continental crust and close to the trench. This new picture of the local seismicity at intermediate depths during the period that the ANCORP network was operational will help to us to better understand the dynamic and complex systems that generate this observed seismicity.
DE: 3270 Time series analysis (1872, 4277, 4475)
DE: 7230 Seismicity and tectonics (1207, 1217, 1240, 1242)
DE: 7240 Subduction zones (1207, 1219, 1240)
DE: 7290 Computational seismology
SC: Seismology [S]
MN: 2007 Fall Meeting