The global seismology group at Harvard University operates the Harvard Waveform Quality Center (WQC), a node of the IRIS Data Management System. Because of our routine analysis of GSN data for the determination of earthquake focal mechanisms for the Harvard CMT Project, we are in a good position to provide routine checks on the quality of IRIS GSN data. The WQC reports data and software problems to the DMC and to the DCCs. The WQC also provides a testbed for various real-time and near real-time data distribution methods, including SPYDER®, SOD, LISS, NRTS, COMSERVE and, in the near future, Antelope.
Waveform Quality Control
All of the data from the GSN which are available within approximately 60 days of an earthquake are used in the standard Harvard CMT analysis. We keep track of which data must be discarded during the analysis, and for what reasons. Errors easily identified by comparison of observed and model seismograms include polarity reversals, timing errors larger than about 10 seconds, amplitude errors due to erroneous transfer functions, and generally poor data quality. The primary means by which we report data quality problems is through the filing of an IRIS Data Problem Report (DPR). These reports are filed as soon as we have identified a problem, whether during regular monthly CMT processing or during Quick CMT processing. Specific problems are documented with figures which are linked with our DPRs.
The performance of the network is analyzed statistically each month, and we identify the stations which consistently provide the highest quality data for CMT analysis. We also identify those stations which have experienced large changes in their utility as compared with earlier months. These data are presented in graphs that are included in the WQC’s Quarterly Reports to the DMS (available from the DMC).
Apart from timing errors and station malfunction, most problems with GSN data are related to errors in the seismogram header information; for example, instrument polarity, station location, instrument orientation, and instrument transfer functions and gain are frequently incorrectly described. An additional activity of the WQC is to analyze the historical header information provided by the DCCs in the form of dataless SEED volumes for consistency and accuracy.
Real-Time Network Performance
Quality control is most useful when it identifies station problems as early as possible. We are now developing various tools to allow us to incorporate as much of the available real-time GSN data as possible into our Quick CMT analysis. We are also working to provide continuous monitoring of GSN station performance in terms of signal and noise levels.
Over the last few months, we have developed the software necessary to robustly retrieve and archive all of the continuous long-period GSN data available in real time via the internet, using the LISS, NRTS and COMSERVE protocols (Availability of Real Time Data). We now routinely have continuous data from more than 50 GSN stations available for Quick CMT analysis. Following larger events, we also obtain data from a number of stations via the SPYDER® system, by automatically retrieving SEED volumes from the DMC. As end-users of the real-time data, we are in a position to help improve these and other new data distribution mechanisms. Though both the NRTS and LISS protocols work quite well, there are several aspects of both systems which could be improved, and we continue to provide feedback as we monitor the performance of this software.
In addition to utilizing the real-time data from a significant fraction of GSN stations in our Quick CMT analysis, we have developed tools which continuously measure the vertical signal and noise levels at all available stations (Signal levels of the GSN). The signal power in narrow frequency bands is calculated hourly from the rms amplitude of the LHZ channel, and this power is converted to a deviation (in decibels) from Peterson’s (1993) new low noise model. The resulting values are plotted in graphs that also indicate the times of recent earthquakes. These graphs, and the underlying data, give an up-to-date view of global seismic noise levels, and occasionally also show long-period signals not associated with any earthquake reported by the NEIC.
Peterson, J, 1993. Observations and Modeling of Seismic Background Noise, USGS Open-File Report 93-322, U. S. Geological Survey, Albuquerque, New Mexico.
by Göran Ekström (Harvard University)