Automatic Processing System For Seismic Event And Phase Detection

Since the 1980s, the advance of the continental lithospheric seismic array, which isrepresented by the American PASSCAL seismic array, offered a technological base forpromoting the high-resolution imaging theory and methods of continental lithosphericstructure. The movable broadband seismic array has became one of the important methodsused for geodynamics and tectonophysics. For a large-size dense seismic array, the spacingbetween stations can reach to the kilometer's order. In terms of such high-resolution seismicarray observations can be improved greatly the results of earthquakes location, focalmechanism, source process as well as seismic tomography of the lithosphere structure.Seismic array observations also open up a new wide field for the seismic waveform study.The American PASSSCAL has built a seismic array composed of 550 sets of broadband movable seismographs. China Earthquake Administration will build a mega movableseismic array consisted of 600 seismographs during the "Tenth-fives-Year-Plan". The rapidgrowing of seismic array and a series of large seismic array observation plans make seismicarray data accumulate rapidly.Based on the statistic by IRIS Data Management Center(DMC),the earthquake data acquired from the movable seismic array had taked more than1/3 of the whole data-base and will increase by 2.1TB per year,becoming the quickestincrease of data resources.Institute of Geology,China Earthquake Administration,hasaccumulated over 1.5TB data in the past ten years.Now the movable array data are managed with the "Distributing Mode", the data areassembled to a seismic net work center. This fashion is only suitable for a small movableseismic array and difficult for a mega array system.A mega seismic array needs building a"Concentrating Mode" data management system, which requires the data processedconcentratively and offer uniform data format,detecting precision and concentrated storage.The seismic event detection and onset phase picking are the main contents of concentratedprocessing. The seismic event detecting asks for the least missing of seismic events and theonset phase picking asks for the accurate picking time based on the event seismic detecting.Because such a processing is aimed at seismic array recordings ,favorable factors such assmall distance between stations and coherence between arrival times should be taken intoaccount in seismic event detection and phase picking procedures. This can reduce the leakof seismic events and improve the accuracy of phase picking.In this study are included two parts: In the first part is discussed the methods of seismicevent detection.Different kinds of the methods for seismic event detection are discussed.TheShort-Term to Long-Term Average(STA/LTA) arithmetic is improved and the artificialneural network(ANN) arithmetic programs is tested with data recorded by the Tian Shanmovable seismic array .The following results are obtained:1)The method of STA/LTA after band-pass filter and multiple-station correlation havethe advantages of simplicity ,explicity and a high detecting rate. The multiple-stationcorrelation implies that an event must be detected by a certain number of stations in a shorttime window,which is in favour of reducing the error detecting.Our results show that 90%earthquakes with only 21% non-earthquakes can be detected successfully.For the movableseismic array observations, this result should be accepted,if considering the influence of thestation environment .2)The seismic event detecting based on the ANN is a pattern-analysis method. If thesignal spectra are used as the input parameters,the ANN classifies the power-spectra ofseismic waveform and noise intellectually.Because the power spectra of seismic waveformshave a very complex pattern,which is influenced by many factors such as the sourcemechanisms,wave path and set effects.In addition,due to a movable seismic arrayobservation at the same place,it can not accumulate enough samples only have a short-term.This method is not suitable for movable seismic array observations.In the second part is discussed the method of the onset phase picking. After analysisand comparison of all available methods,the single-station fractal phase pickingalgorithm ,single-station autoregression AIC arithmetic, multi-channel cross-correlation andthe least squares arithmetic program are developed and tested from data recorded by theTian Shan movable seismic array.The following results are obtained:1)The two available methods of estimating the phase picking based on fractal dimensiondo not adapt to the complex earthquake waveforms. The first method is the "ConstantMethod", in which the noise before the onset makes the change of the fractal dimensioncause errors of the onset phase picking. The second method is not suitable for the onsetphase picking,because this method requires a repeat computation of waveforms.This causescomplex process.The method of the fractal dimension for the onset phase picking is onlysuitable for active seismic events,instead of passive seismic observation.2)The noise of the signal is filtered,before the single autoregression AIC detecting andmulti-channel cross correlation with the least square .When the window time of AIC isselected to be 5 seconds, the AIC method can get the best result.The phase picking errorrelative to the manual picking is less than 0.3s for the local earthquakes and 0.1s forteleseismic earthquakes.3)Because some noise can not be filtered ,the single station AIC method is not able topick the onset phase well.For the teleseismic records,the multi-channel correlation and leastsquare processing can be used.For high noisy teleseismic records, this method can improvethe reliability and precision of the onset phase picking. It also can correct the errors ofmanual picking.To complete the multi-channel correlation detecting arithmetic,the MySQL as thedatabase management system(DMS) is also described,which interface was developed byusing the C++ language.In the DMS are four data tables using for storing the informationabout phase and earthquake catalog.The DMS is divided into five modules,which serve asseismic event detection,the multiple-stations correlation proving,data extracting andstorage,single phase picking and multiple-station correlation phase picking,respectively.Based on the DMS,an array seismic event detecting and phase picking have beencompleteed.The raw data will be standardized.Most of processing and management of thedata are finished automatically by the computer.This reduces manual intervention andimcreases the efficiency of management and safeguard the data.