Prediction and suppression of long-period nonpropagating seismic noise

Abstract

abstract Approximately half the noise observed by long-period seismometers at LASA is nonpropagating; that is, it is incoherent over distances greater than a few kilometers. However, because it is often strongly coherent with microbarograph data recorded at the same site, a large proportion of it can be predicted by convolving the microbarogram with some transfer function. The reduction in noise level using this technique can be as high as 5 db on the vertical seismometer and higher still on the horizontals. If the source of this noise on the vertical seismogram were predominantly buoyancy, the transfer function would be time-invariant. It is not. Buoyancy on the LASA long-period instruments is quite negligible. The noise is caused by atmospheric deformation of the ground and, since so much of it can be predicted from the output of a single nearby microbarograph, it must be of very local origin. The loading process may be adequately described by the static deformation of a flat-earth model; however, for the expectation of the noise to be finite, it is shown that the wave number spectrum of the pressure distribution must be band-limited. An expression for the expected noise power is derived which agrees very well with observations and predicts the correct attenuation with depth. It is apparent from the form of this expression why it is impossible to obtain a stable transfer function to predict the noise without an array of microbarographs and excessive data processing. The most effective way to suppress this kind of noise is to bury the seismometer: at 150 m the reduction in noise level would be about 10 db.