Prediction of pressure-generated earth motion using optimum filters

abstract The performance of long-period seismographs is often seriously degraded by atmospheric pressure variation; the problem is particularly severe at periods greater than 20 sec. The pressure variations associated with wind-generated turbulence and acoustic waves are sufficient to deform the surface of the Earth, thus adding to the background noise level recorded by the seismometer. If microbarographs are operated together with the seismograph system, a large percentage of the atmospherically generated noise can be eliminated by the use of optimum filters. The filters are designed based on the least-mean-squares criterion, with the seismograph time trace as the desired output and the microbarographs as the inputs. Single-channel filters, using only one microbarograph, located at the seismometer vault are used to attenuate wind-generated noise. In order to attenuate the noise on windless days from other pressure sources, multichannel filtering is usually necessary and therefore an array of microbarographs is required. The filters used to predict the wind-generated noise are shown to be stable despite the complicated source. The performance of the multichannel varies widely depending on the structure of pressure variations predominating in the atmosphere.