Using Filter Banks to Identify Phases for Early Warning

In earthquake early warning, it is important to rapidly distinguish between different types of arrivals. In order to generate a warning in time, only the first few seconds of data can be used. This can cause misidentifications that would not occur if the full record was available. For example, the S-wave of a teleseismic event could be mistaken for the P-wave of a local event. One way to discriminate between these types of events is to look at the frequency content, which will depend on both the magnitude and distance of the event. We propose to accomplish this by running the data through a filter bank made up of several different band-pass filters. We then take the ratio of the short term average to the long term average in each band to distinguish the arrival time from any background noise. By examining the relative and absolute amplitudes in each band, we can determine the type of arrival, whether it is a P or an S-wave and whether it is from a local, regional, or teleseismic event.