Preprocessing of earthquake tremor data for tremor location

Sammanfattning: When analysing volcanic seismicity and volcanic events, it is often interesting to focus on tremor, which is a signal of long duration. This is because not much is known about tremor. Unlike earthquakes, sources of tremor are difficult to locate due to them not having any clear onsets in the seismic signal. Different location methods can be used to attempt to locate tremor sources, one of which incorporates correlation of seismic signals between two different recording stations. This method is sensitive to the frequency content in the signal, and will not operate correctly if the signal consists of non-tremor signals, e.g. earthquakes. Earthquakes generally contain higher frequencies than tremor, and if these are present in the signals being correlated, this will affect the location method negatively. Therefore, it is desirable to try to detect these events containing other frequencies, to be able to remove them from the data and improve the location method. Constructing this method is the objective of the study, and is done using the programming language Matlab. The method uses correlation between the amplitude spectra of short adjacent time intervals of the bigger time series, in an attempt to detect when the spectral signature of the time series radically changes. An outlier is deemed to exist when the correlation coefficient from the comparison falls beneath a pragmatically chosen multiple of standard deviation from the mean of coefficient values. The objective of the study has been achieved, by constructing a method using correlation between time intervals to detect sudden fluctuations in the spectral shape of the time series. The method does seem promising, however, the accuracy of the method now needs to be rigorously tested. Time has not allowed for this to be done in this study, and is left for future research.