Time-frequency Methods And Application To Crustal Deformation Data Analysis

As an important tool of modern signal processing, time-frequency analysis is a hotspot fortime-varying and non-stationary signal research. Crustal deformation observation data are akind of typical non-stationary and time-varying signals. Using time-frequency analysis method,we can extract important information from joint distribution of signals in time and frequencydomains, understanding crustal deformation in a new idea. In this paper, we compared therespective merits, scope of application and areas for improvement of these three methods,introducing the theories of wavelet transform, S transform and Hilbert-Huang transform whichare all nonlinear analysis methods, designing simulation signals which are similar to thedeformation signals and used to test the methods. We estimated the effect of time-frequencymethod for investigating anomalies, analyzing one-second sampling, one-minute sampling andone-hour sampling crustal deformation data.Through analyzing meteorological disturbances that may affect crustal deformationobservation data quality, we find that crustal deformation is very different on responses todifferent meteorological phenomena in frequency domain. Wherein, the effect of barometricpressure is obvious on the data observed in one-minute sampling and its predominant frequencyis lower than0.05Hz which means its predominant period greater than20minutes, with highsignal power spectra responding to apparently barometric changes. Typhoon has a greaterimpact on the data observed with one-second sampling, of which the dominant frequencyranges from0.15Hz to0.3Hz. The impact of precipitation is obvious on the one-hour samplingdata, which concentrates on the part of the lower frequencies. Finally using Wavelet and Stransforms, we analyze crustal deformation observation data during earthquakes withmagnitude7and over occurred in Wenchuan&Yushu&Lushan of China, Burma, Pakistan andJapan, and find that there is a phenomenon of signal power increase in two or three frequenciesin the crustal deformation data observed with one-second sampling before the events, that wewould like call it as double/triple-frequency increase phenomena (d/t-fip) on dynamic spectra ofthe data.