Research On Digital Seismic Waveform Anomalies Recorded Before Mid-strong Earthquakes

The plenty of seimic data recorded by the China Digtal Seismograph Network(CDSN) provides richer data services for the earthquake prediction, earth science,national economic construction and the social public. Under general circumstances,there are usually some seismic waveforms of small events recorded by the stations inthe epicenter area before the strong earthquake. Research on the waveformcharacteristics of these events can provide useful information for earthquakeprediction. On the other hand, there is no obvious foreshock activity, but a lot ofcontinuous waveform recorded before the earthquake. Extracting useful information,such as slow seismic events, can also be a part of the research from these continuouswaveform data.According to the presence or absence of foreshocks before a strong earthquake,the study of digital seimic waveform anomalies can be divided into two directions:For the presence of foreshocks, the source parameters can be used to analyse the stressfield variation. For the absence of foreshocks, slow earthquake signal can be extractedfrom continuous seismic waveform several months before the mainshock.So far, the general and effective method for analyzing the stress field variation canbe classified into2categories: the one is using source parameters (stress drop) toevaluate the level of the stress. The other is utilizing the focal mechanism to deducethe variation of stress direction. Combining the source parameters and the focalmechanism solution to study the variable process of stress field with time can be aneffective method for earthquake prediction, it also can be provide a new method andthe physic base of strong aftershock prediction.For the presence of foreshocks, this paper does mainly three researches:(1) The source parameters and spectral amplitude correlation coefficients arecalculated using the spectrum analysis methods and the Brune’s source model.(2)The events with similar focal mechanisms can be grouped using clusteringmethod. The differences in each group focal mechanism solutions can be analyzed by comparing with the obtained focal mechanisms.(3) The stress level and stress direction can be analyzed by combining the sourceparameters and the mean focal mechanism solutions of each group.Based on the digital waveform data of the Yingjiang earthquake in2008and2011recorded by Yunnan Digital Seismic Network, some conclusions as follows areobtained by studying the source parameters and the focal mechanism type：（1）Overall, the stress drop increases with magnitude,while it doesn’t vary withmagnitude for Ms≥4of the Yingjiang earthquake sequence in2008and Ms≥3.5of theYingjiang earthquake sequence in2011.(2) The earthquakes with small magnitude and high stress drop and with bigmagnitude and low stress drop may be a sign of the activity of the earthquakesequence.（3）There are good relations with the trend of P axes in each group focalmechanism, and if there are higher correlation coefficients, the trend of P axes arecloser.(4) The focal mechanism solutions converse at a certain type, and the reflectedstress field is dentical or close to the regional stress filed. These must be thecharacters of foreshock stage。(5) In the aftershock stage, the focal mechanisms are mainly scattered. The typesof most of earthquakes are the main shock type and the local stress field type. And therest of focal mechanisms type is divergence and random in small amount.(6) The change and transformation of types of focal mechanism solutions atdifferent stages reflects the temporal characteristics of the regional stress field. And atype of earthquake focal mechanism solutions concentrated in a time period andswitching to regional stress field may be a sign of strong earthquakes. When the typesof focal mechanism solutions diverge over time, it may be the sign of the activitydecreasing of the earthquake sequence.(7) There is some relationship between the stress drop and the type of focalmechanism, the closer the direction of the focal mechanism in a group close to theregion stress field, the higher the stress drops increase. The noises in the various frequency bands are available before earthquakes, suchas the slow earthquakes which release the energy slowly in various forms. Theseevents are very useful for study the anomalies before the earthquake occurred. Thedetection of slow earthquakes is the most basic problem in the study of slowearthquake. Based on the current research, the waveforms of the slow earthquakes aredivided into2frequency bands. The one is very low frequency events with10-20sspectrum range. The other is low frequence events with1-10Hz spectrum range.For the absence of foreshocks, an algorithm for detection of low-frequencyseismic events is developed, it contains three steps:(1)The continuous vertical-component waveforms of some broad band stations ina few months surrounding epicenter area have been processed by applying a bandpassfilter in2–8Hz for removing the influence of body and surface waves of far-distancestrong earthquakes and other low frequncy noises (as typhoon).(2) Enveloping the filtered waveform with a smoothing time of10s window.(3) Applying a median filter with a20min window for the envelopes waveform tomaintain the event lasting for more than10minutes. It can filter small earthquakesand local noises.Using the continues waveforms of some broad band stations in a few monthssurrounding Wenchuan earthquake and Lushan earthquake, the following conclusionsare by the above process,(1) The filtered amplitude is obviously larger than that of noise and last for adozen of minutes to several hours during a few days for a small number of stations:this phenomenon appears one month before Wenchuan earthquake, one and halfmonth before Lushan earthquake.(2) The waveform and envelope are similar to that of non-volcanic tremor (NVT),we think they may be a kind of NVT.(3) The preliminary application demonstrates that this algorithm is potentiallyuseful for extracting NVT signal from continuous seismic waveform.The seismogenic process and occurrence of the earthquake is a complex process.We use some methods that include both studying the variation of stress field and detecting the slow earthquake to analyze the abnormity features before theearthquakes. This task is very significant for enhancing the understanding ofearthquake occurrence, and is of an important practice meaning on applying digitalearthquake data amply.