Methods
In this work we use four different methods of data analysis, the order in which these are implemented are as follows:
1) Event Detection
2) Analyst Review
3) Template Generation
4) Clustering
5) Subspace Detection
Each method is described in more detail below.
Click to learn about:
1) Event Detection
We impliment aspectral domain detection method to detect local earthquakes within the footprint of the TA network [Linville et al., 2014]. In our initial exploratory testing we found, on average, ~2 detections per 3-hour time window. See "Analyst Review" page for examples.
2) Analyst Review
Analysts review the array images and waveforms and assign a confidence rating all detectiosn that are deemed earthquakes by the automated detection algorithm.
3) Template Generation
For the ~40,000 array-images we will process, we expect to generate a template catalog of ~200,000 template events. Our event template catalog will include the station name and start time for each template identified.
4) Clustering
With the aim to automate the identification of template events, we apply a clustering method to our detection database using a nearst neighbor data reduction technique.
5) Subspace Detection
Subspace detectors are the decomposition of a waveform, or a waveform collection, into vector components. To identify new events, linear combinations of these subspace detectors are compared to continuous data streams.