Preliminary observations of noise spectra at the SRO and ASRO stations

The seismic noise spectra presented in this report were derived from SRO and ASRO station data for the purpose of evaluating the performance of the seismic instruments. They are also useful for constructing a spectral estimate of earth noise at a quiet site based on noise samples obtained from a network of globally distributed sites. It is hoped that the spectra will be usefull for other purposes as well.

The term 'noise' is used here to describe the ambient signals recorded during a quiet period when earthquake signals have not been detected by visual inspectino of the analog seismogram. The total recorded noise is the sum of instrumental noise, environmental noise (such as effects of temperature, pressure, wind), earth background noise from both natural and cultural sources, and very possibly low-level signals from earthquakes that cannot be visually identified. It is not possible to separate and quantify the signals generated by these independent noise sources using a single sample of station data, although instrumental problems may be indicated by gross changes of noise levels, if the changes are not in the microseismic bands. Since seismic data at the SRO and ASRO stations are recorded in a digital format, spectral computations can be automated so that station noise levels can be monitored as part of data-review procedures. The noise spectra presented in this study are intended to serve as an initial baseline against which relative changes in noise levels can be measured.

Total noise power was computed separately for the short- and long-period bands, which are recorded separately at the stations. Power spectral densities were derived by averaging the spectral estimates of a number of contiguous dat segments. The mean value and slope were removed from each segment, cosine-tapered windows were applied, and the estimates were obtained using a fast Fourier transform. In the short-period analyses 16 segments were used, each segment being 1024 samples in length. Because the sampling interval is .05 seconds, the total record length is nearly 13.7 minutes. Normally, the short-period SRO and ASRO data are recorded in an event-only mode. However, several days of continuous short-period data were acquired from most stations for the purpose of this study. Where there was appreciable diurnal variation in short-period noise, spectral data were computed for both day and night intervals. In most cases the long-period spectral densities were obtained by averagin the estimates from 16 data segments, each segment having a length of 2048 samples. Since the long-period sampling interval in 1 second, the total record length used was nearly 9.1 hours. In a few instances, a smaller number of segments was averaged. Spectral data were computed from the vertical-component short-period signals and all three components of long-period signals. All of the spectral plots have been corrected for known instrument response and presented in units of earth displacement.

With a few exceptions, the samples of noise data used were acquired during the early months of 1980, winter at some of the stations and summer at others. The starting times for the intervals analyzed are listed in Table 1. A seasonal variation of noise levels in microseismic bands is to be expected. However, none of the stations were experiencing a noticeably high level of microseisms during the intervals analyzed. Weltman and others (1979) have studied and reported daily and seasonal RMS (root-mean-square) noise trends at the SRO and ASRO stations.