Waveform Analysis

Waveforms are analyzed to decide if an event is suitable for use as a template in generating the database for detecting future template events. Each event is ranked on a scale of 1 to 5 (1 being most ideal) based on the appearance of the waveform and the timing of the pick (i.e. was the pick late, early, or correct)

1 == Likely a local event, include in the template catalog, but this is likely a larger event that may already be in the ANF catalog
2 == A local small event that might not be cleanly recorded on all 5 stations, include in template catalog
3 == Likely a blast, might consider later, but at this time not to be included in the template event catalog
4 == Teleseismic or off array event, not to be considered in the template event catalog
5 == false detection, auto-detected event that is not an earthquake of interest

In addition to a confidence ranking the analyst also determines if the STA/LTA time of arrival (gold pick in images below) is late (+), early (-), inappropriate (x) or correct (Z).

Confidence Rank: 1

These waveforms are characteristic of the recording of a local earthquake, with well defined P- and S-Waves. Confidence 1 ranking identifies an event as having clear template waveforms and are often those represented in the existing ANF catalog because these events are often larger (>M2.5). When they are reported by ANF they are a key part of clustering analysis because they help us understand which families of events are recoverable using existing catalogs, and which events represent new sources.

Confidence Rank: 2

Events that are assigned a confidence rank of 2 are key to this work. They signify events that are likely smaller than the completeness level of the ANF catalog and typically only show a clear signal on ~3 of the 5 closest stations.

Confidence Rank: 4

Regional seismic events that are not local to the array (from P-S separation) or teleseismic events from greater distances get a confidence rank of 4. When there is any ambiguity in the case of regional events, these events get a confidence of 2 and are futher analyzed with waveform clustering and cross-correlation. Confidence 4 events, although real seismic signals, are not considered in our study.

Confidence Rank: 3

Events that are likely a blast are assigned a confidence ranking of 3. Blasts can have different characteristics, but for our study region the most telling are the 'domino' effect blasts that have multiple pluses or protracted long period coda waves.

Confidence Rank: 5

We have tuned our algorithm to minimize false detections, but they still do occur. In general the number of false detections represent a small percentage of all detections (< 1%)