SAGE Facility MUSTANG metrics Web Service Documentation

orientation_check Channel orientation check

Summary

This metric takes advantage of Rayleigh waves’ polarized retrograde elliptical particle motion to empirically estimate event back azimuths and compare them to values calculated from the metadata. Radial and vertical components have similar motions but differ in phase by 90 degrees.

For events having Ms or mb >= 7.0 and depth < 100 km, a station’s horizontal components are rotated incrementally through 360 degrees to find the bearing from north that maximizes cross-correlation of this “trial radial component” with the Hilbert transform of the vertical component. (The Hilbert transform introduces a 90-degree phase shift.) Two peak correlation coefficients are calculated – they differ slightly in their normalization. One coefficient is maximized to find the estimated back azimuth (radial direction); the other characterizes the quality of the cross-correlation results.

The observed (empirical) channel orientations are reported for channels “X” and “Y” where the sensor is assumed to have two horizontal channels that differ by 90 degrees. Channel “X” is oriented 90 degrees clockwise from channel “Y”, just as the X-axis is oriented 90 degrees clockwise from the Y-axis in a Cartesian coordinate system.

Stachnik, J.C., Sheehan, A.F., Zietlow, D.W., Yang, Z, Collins, J. and Ferris, A, 2012, Determination of New Zealand Ocean Bottom Seismometer Orientation via Rayleigh-Wave Polarization, Seismological Research Letters, v. 83, no. 4, p 704-712. https://doi.org/10.1785/0220110128

Uses

Channel orientation estimates having large correlation coefficients from many events and back azimuths can be averaged to give an empirical orientation estimate. This estimate can be used in the metadata when the channel orientation is unknown (e.g. for ocean-bottom seismometers) or can verify metadata orientations reported from the field. Effects such as multi-pathing make it unreliable to estimate orientation using a statistically small sample of measurements. (See the reference above for details on how to clean and average these measurements. Note that there is an apparent contradiction in the paper about which correlation coefficient to use for data cleaning; max_Czr should be > 0.4)

Because orientation_check measurements require statistical processing before they can be interpreted, we do not recommend basing rrds requests on this metric.

Data Analyzed

Events – Ms or mb >= 7.0 and depth < 100 km.
Traces – three N.S.L.C (Network.Station.Location.Channel) per measurement where the Channels are three components of the same sensor and sample rate.
Window – 620 seconds beginning 20 seconds before and ending 600 seconds after the predicted Rayleigh wave arrival of the event.
Data SourceIRIS miniSEED archive or IRIS PH5 archive, and USGS event service

SEED Channel Types – [BCDFHLM]H? | High Gain

Algorithm

  • For events with Ms or mb >= 7.0 and depth < 100 km,
    • Get the event latitude, longitude, depth and origin time,
    • For each broadband network.station.location having three components,
      • Find the metadata orientation of ??N or ??1,
      • Find the angular distance and event back azimuth for the current station,
      • Find the surface distance between the event epicenter and station,
      • Predict the arrival time of a Rayleigh wave traveling 4.0 km/s along the surface,
      • Request 3-channel instrument-corrected data starting 20 seconds before and ending 600s after the predicted Rayleigh wave arrival,
      • If the vertical channel metadata dip already describes the channel as phase reversed, multiply channel amplitudes by -1,
      • Apply a 10% cosine taper to each time series,
      • Bandpass filter from 0.02 to 0.04 Hz (50-25s),
      • Take the Hilbert transform of the vertical channel (H{Z}),
      • For trial empirical ??N/??1 channel azimuths from Y = 0 to 360 degrees from North,
        • Rotate the two horizontal channels to find the radial component R,
        • Calculate the cross-correlation of R and H{Z} at zero lag:
          Szr = sum(i): [R[t(i)] * H{Z[t(i)]}] where i = 1,...,N samples
          
        • Calculate auto-correlations of R and H{Z} at zero lag:
          Srr = sum(i): [R[t(i)]^2] where i = 1,...,N samples
          Szz = sum(i): [H{Z[t(i)]}^2] where i = 1,...,N samples
          
        • Calculate the two correlation coefficients:
          Czr = Szr / sqrt(Szz*Srr)
          C_zr = Szr / Srr
          
      • Find the orientation Y at which the correlation coefficient C_zr is maximized.
      • Report:
        • the empirical orientation Y and X=Y+90,
        • metadata azimuths for horizontal channels:
          azimuth_Y = backAzimuth – azimuth_R
          azimuth_X = azimuth_Y + 90
          
        • the correlation coefficients,
        • the start and end time of the data window,
        • the magnitude of the event.

Metric Values Returned

azimuth_R = rotation angle from N at which correlation coefficient max_C_zr is maximized
backAzimuth = calculated back azimuth between the event and station
azimuth_Y_obs = empirical bearing from N of channel Y (see Summary)
azimuth_X_obs = empirical bearing from N of channel X (see Summary)
azimuth_Y_meta = bearing from N of channel Y reported in the metadata
azimuth_X_meta = bearing from N of channel X reported in the metadata
max_Czr = maximum correlation coefficient that best characterizes the quality of the cross-correlation
max_C_zr = maximum correlation coefficient used to find the radial component orientation
magnitude = event magnitude
target – the traces analyzed, labeled as N.S.L.??Z.Q (Network.Station.Location.Channel.Quality)
start – 20 seconds (UTC) before the predicted Rayleigh wave arrival of a suitable event.
end – 600 seconds (UTC) after the predicted Rayleigh wave arrival of a suitable event.
lddate – date/time the measurement was made and loaded into the MUSTANG database (UTC)

Contact

Updated