The New K-Sigma Picker

The new K Sigma picker, named after the current picker CSN uses currently due to its similarity in approach is very simple yet effective statistical metric that has very successful in detecting Picks from the data stream. The working of this approach is simple in nature and can be understood from the diagram below:

The diagram above shows full procedure of picking an event using new k sigma algorithm. The wave in red (bottom) is the waveform that K Sigma sees, and we can see a spike when the earthquake arrives. By selecting a suitable threshold and by assuming if metric goes over that threshold at any instant as pick, we can identify when a seismic activity occurs. This new approach seems to be an improvement over current picker that CSN uses and reduces the false positives by multitudes. The figure below makes this argument more clear:

Figure 6. A plot comparing original picker reporting (Red line in top plot) and Neo Picker reporting (bottom plot) for the same data stream.

To find threshold ( k ), a cost minimization can be done over all the available acceleration streams, where the cost function is weighted sum of false positives ( fp(k) ), false negatives ( fn(k) ) and temporal inaccuracy of pick ( T(k) ). Declaring the cost function as:

Cost = α fp(k)+ β fn(k)+ γ T(k)

This function can be minimized over threshold (k) for appropriate weights (false negative should inflict greater cost than false positive, since we do not want to miss a quake.) to find out optimal value of k for which new picker gives best results.