This directory contains the following files which present the results of the GSN background noise PSD study: sur_psd.ps <- SUR.II.SHZ background noise PSD estimate. GSN_composite_Z-comp_LNM.ps <- Composite GSN Low Noise PSD Model. GSN_station_median_PSD_ranking <- Tabular file of PSD results. README <- This file. pxy_data_file.01.ps <- Bargraph presentation of PSD results. pxy_data_file.02.ps pxy_data_file.03.ps pxy_data_file.04.ps pxy_data_file.05.ps pxy_data_file.06.ps pxy_data_file.07.ps pxy_data_file.08.ps pxy_data_file.09.ps pxy_data_file.10.ps pxy_data_file.11.ps pxy_data_file.12.ps pxy_data_file.13.ps pxy_data_file.14.ps pxy_data_file.15.ps pxy_data_file.16.ps pxy_data_file.17.ps pxy_data_file.18.ps pxy_data_file.19.ps pxy_data_file.20.ps pxy_data_file.21.ps pxy_data_file.22.ps pxy_data_file.23.ps pxy_data_file.24.ps pxy_data_file.25.ps pxy_data_file.26.ps pxy_data_file.27.ps pxy_data_file.28.ps pxy_data_file.29.ps pxy_data_file.30.ps pxy_data_file.31.ps The gsn_psd_results.10oct01.tar.gz file is a gzipped tar file which contains all of the above files. Notes: 1) The GSN Composite Low-Noise PSD Model "GSN_composite_Z-comp_LNM.ps" plot is a composite derived from 780k+ PSD estimates from 04/1999-03/2000 data recorded at 116 GSN stations. The solid lines represent the fraction of ranked PSD estimates in percent which exceed a given threshold (5, 10, 25, 50, 75, 90, 95, 96, 97, 98, and 99%). Prior to ranking, the PSD data are binned into logarithmic decrement bins with 20 bins per decade. The PSD estimates for periods longer than 10 seconds are derived from LH data, the PSD estimates for periods from 0.25 seconds to 10 seconds are derived from BH data, and the PSD estimates for periods shorter than 0.25 seconds are derived from EH, SH, or HH data. The estimates derived from BHZ and LHZ are stable to the 99th percentile while the estimates derived from HHZ, SHZ, and EHZ (T < 0.25 sec) are not stable below the 98th percentile owing to the much smaller high-sampling-rate data set (the LH data typically have 4000+ PSD estimates in each bin while the EH, SH, or HH data may have as few as a couple hundred PSD estimates in each bin). Each time series is parsed into eight contiguous sub-series which are individually analyzed and of the eight PSD estimates, the one with the lowest integral value is taken as representative of the low-noise PSD for the time series. This procedure is used so that the PSD algorithm can produce robust results when processing large amounts of data in an automated fashion. The volume of data processed in the analysis of the GSN data is so large that it is not possible to inspect each individual time series for evidence of instrumental problems or contamination by seismic signals. The PSD algorithm uses a windowing, detrending, and demeaning algorithms designed to provide a stationary time series to the Fourier transformation algorithm. Care is taken to reduce biases caused by signals that violate the stationarity assumption inherent in Fourier transformation theory. Tests indicate that even insidiously small violations of the stationarity assumption can result in a bias of the estimated PSD that is high by several dB at long periods so that the windowing, detrending, and demeaning details are very important. The GSN derived LMN is significantly below the USGS LNM at periods longer than ~30 seconds and it dips to -191.6 dB at a period of ~400 seconds (~5.1 dB below the USGS LNM). Also the GSN derived LMN is up to ~1.2 dB below the USGS LNM in the 0.3-0.6 sec (1.6-3.3 Hz) band with an average PSD level of -168.7 dB. 2) The GSN_station_median_PSD_ranking provides a tabular synopsis of the data presented in the 31 bargraph plots. The columns entries are: station code, network, location, component (HZ = vertical and HH = horizontal), period band, 1st quartile PSD value, median PSD value, 3rd quartile PSD value, PSD spread (3rd quartile - 1st quartile), and number of number of PSD estimates. The entries are ranked in order of increasing median PSD value. The best stations typically have both a low median PSD value as well as a small spread. The data are separated by lines such as "Ranking for Borehole HZ.3" which indicate the sensor type (Borehole), component (HZ = vertical), and period band (3), for example. The period bands are: 1 -> 0.0625 - 0.091 sec 2 -> 0.091 - 0.15 sec 3 -> 0.15 - 1.0 sec 4 -> 0.5 - 5 sec 5 -> 5 - 15 sec 6 -> 15 - 60 sec 7 -> 60 - 200 sec 8 -> 200 - 500 sec 2) The 31 pxy_data_file.xx.ps files present the PSD results for: four different sensor types (Borehole, Guralp, STS-1, and STS-2), eight period bands (0.0625-0.091 sec, 0.091-0.15 sec, 0.15-1 sec, 0.5-5 sec, 5-15 sec, 15-60 sec, 60-200 sec, and 200-500 sec), and two different components (vertical and horizontal). Only those sensor/band/ components for which there are data available are plotted. The three columns on the left are the station code, the network.location code and the median PSD value. The data are presented in order of increasing median PSD value to facilitate interpretation of the data and the vertical lines in the bargraphs represent the individual PSD estimates that were determined from data recorded between April 1999 and March 2000. PSD estimates below -196 dB and above -90 dB are not included in the bargraphs. Likewise, when there are more than 20 PSD estimates available for a particular bargraph, the lower 10% (to minimize potential bias by a malfunctioning sensor) and upper 10% (to minimize potential bias due to a seismic event) were discarded. Bob Uhrhammer Berkeley Seismological Laboratory 10 October 2001