Help: noise-psd v.1

Description

The noise-psd web service returns Power Spectral Density estimates for seismic channels.

Summary

As the first of a two-step process for generating probability density function (PDF) plots of power spectral density for comparison with the Peterson (1993) noise models, these power spectral densities describe time series prior to instrument response removal. This intermediate storage step was introduced to minimize recalculation whenever instrument responses change. Power is in units of decibels (dB). The algorithm for this metrics follows NcNamara and Boaz (2005).

McNamara, D.E and Boaz, R.I., 2005, Seismic Noise Analysis System Using Power Spectral Density Probability Density Functions – A Stand-Alone Software Package, U.S.G.S. OFR 2005-1428.

Peterson, J, 1993, Observations and Modeling of Seismic Background Noise, U.S.G.S. OFR-93-322

Uses

The primary purpose of storing PSDs in this form is so that PDF generation can be performed using the latest metadata with a minimum amount of recalculation. The PDFs, in turn, can be used to evaluate the general noise characteristics of a channel, providing data quality information as a function of frequency.

Data Analyzed

Traces – one N.S.L.C (Network.Station.Location.Channel) per measurement
Window – 1 hour for sample rates >= 10 Hz; 2 hours for sample rates between 1 and 10 Hz; 3 hours for 1 Hz sample rates
Data SourceIRIS SEED archive

SEED Channel Types – ?H?, ?L?, ?N?, ?G?, ?P? | High Gain, Low Gain, Accelerometer, Gravimeter, Geophone | excluding very long period channels

Algorithm

  • Request 24 hours of data for the current N.S.L.C.
  • Divide the trace into windows having 50% overlap where the window length is
    • 1 hour for sample rates >= 10 Hz,
    • 2 hours for sample rates between 1 and 10 Hz,
    • 3 hours for 1 Hz sample rates.
  • For each window,
    • Truncate the time series to the nearest power of 2 samples.,
    • Smooth and average the PSD to reduce variance by
      • Dividing the window into 13 segments having 75% overlap.
      • For each segment,
        • Removing the trend and mean,
        • Apply a 10% sine taper,
        • FFT,
        • Calculate the normalized PSD.
      • Average the 13 PSDs & scale to compensate for tapering.
      • Frequency-smooth the averaged PSD over 1-octave intervals at 1/8-octave increments,
      • Convert power to decibels.

Metric Values Returned

target – the trace analyzed, labeled as N.S.L.C.Q (Network.Station.Location.Channel.Quality)
start – beginning of averaged PSD window in UTC
end – end of averaged PSD window in UTC
freq, power – frequency in Hz, power in dB

Notes

During smoothing, there is a tradeoff between enhancing frequency resolution and reducing variance of the PSDs (i.e. increasing repeatability). These PSDs were smoothed to reduce variance. This choice gives a clearer general picture of station noise by reducing power smear, it is consistent with steps used to generate the Peterson new high and low noise models to which they are being compared and it optimizes data storage. Consequently, features requiring detailed frequency resolution will appear smeared in these PSDs.

Features requiring detailed frequency resolution will appear smeared in these PSDs.

Contact

dmc_qa@iris.washington.edu

See Also

dead_channel_exp ,
dead_channel_lin ,
pct_above_nhnm , pct_below_nlnm , transfer_function

Updated

2018-03-05