Config dictionary for local, shallow events

[Cthuulhaa]

Hi,

I am testing the GaMMA associator to associate picks of shallow seismic events, including quarry blasts following the examples provided e.g. in SeisBench. This are the parameters in the config dictionary:

 config = {'use_amplitude': False, 'dims': ['x(km)', 'y(km)', 'z(km)'], 'vel': {'p': 5.6, 's': 3.9},
          'use_dbscan': False, 'dbscan_min_samples': 3, 'dbscan_eps': 30, 'min_picks_per_eq': 3, 
          'oversample_factor': 10, 'x(km)': (1e-3 * ev_east - 111.11, 1e-3 * ev_east + 111.11),
          'y(km)': (1e-3 * ev_nort - 111.11, 1e-3 * ev_nort + 111.11), 'z(km)': (0, 50), 'max_sigma11': 2.0,
          'max_sigma22': 1.0, 'max_sigma12': 1.0, 'method': 'BGMM'}

ev_east and ev_north are the event easting and northing, which I know beforehand.
The parameter selection does not seem to be optimal for the picks I am trying to associate, as many picks which clearly correspond to my event of interest are ignored, as you can see in this plot:

Do you have any suggestions?

Cheers!

[SquirrelKnight]

Have you tried with dbscan off, just to see if more picks get associated?

[Cthuulhaa]

@SquirrelKnight Hi, is that not the same as 'use_dbscan': False?

[zhuwq0]

@Cthuulhaa This is a bit strange. Could you set these parameters, 'max_sigma11': 2.0, 'max_sigma22': 1.0, 'max_sigma12': 1.0, to very large values, basically turn off the filtering. The filtering is independent of the association step, just to filter out bad picks. Because GaMMA is a clustering method, basically every pick will be assigned to some clusters. The clusters that do not fit the filtering criteria like min_picks_per_eq: 3 will be removed. You can check your region range is correctly set, the associated event locations are not at the edge of your domain.