Statistical characterization and time-series modeling of seismic noise

Developing statistical models for seismic noise is an exercise of high value in seismic data analysis since these models play a critical role in detecting the onset of seismic events. A majority of these models are usually built on certain critical assumptions, namely, stationarity, linearity, and Gaussianity. Despite their criticality, very little reported literature exists on validating these assumptions on real seismic data. The objectives of this work are (i) to critically study these long-held assumptions and (ii) to propose a systematic procedure for developing appropriate time-series models. A rigorous statistical analysis reveals that these standard assumptions do not hold for most of the data sets under study; rather they exhibit additional special features such as heteroskedasticity and integrating effects. Resting on these novel discoveries, ARIMA-GARCH models are developed for seismic noise. Studies are carried out on 185 real-time data sets over different time intervals to study the daily and seasonal variations in noise characteristics and model structure. Nearly all datasets tested positive for first-order non-stationarity, heteroskedasticity, and Gaussianity, while 19% tested negative for linearity. Analysis of the structural uniformity of the developed models with respect to daily and seasonal variations is also presented.