Two papers have been presented in 1995 and 1996 on this subject.

A common limitation to the analysis of low-amplitude, long- period (10-300s), horizontal seismic data is the prevalence of long-period tilt noise caused by barometric fluctuations. Reducing this noise will greatly facilitate the study of long- period data from moderate-size seismic events. This requires simultaneously recording the barometric pressure variations near the seismometers and then subtracting an appropriate function of the barometric pressure from the seismic data. I illustrate this process using the long period (where the problem is most severe) and the very long period data recorded at IRIS station CCM (Cathedral Cave), located in a cavern in the Ozark Mountains at Onondaga Cave State Park, Missouri.

To measure the barometric fluctuations, a high resolution hybrid micro-barometer system is recorded by auxiliary 24-bit channels of the IRIS-2 data system, with identical sample rates as the seismic data. The barometric data is converted to a rate- of-change value by using a finite-differencing algorithm. The resulting barometric pseudo-velocity series is appropriately scaled to match the pre-signal seismic noise with a correlation of better than 95%. Then the Fourier transform of it is subtracted from the transform of each of the horizontal seismometer data series, which have been appropriately time shifted to accommodate the delay of the differencing process.The effect is a reduction of the noise in the long- period databy 10 to 30 db in power spectral density, to levels comparableto the noise level of the vertical, with the greatest improvementso far in the 30 to 1000 second range.

A limitation to the analysis of low-amplitude, very- and ultra- long period (100-3000s), horizontal seismic data is the prevalence of VLP and ULP tilt noise caused by barometric fluctuations. Reducing this noise requires simultaneously recording the barometric pressure near the seismometers and then subtracting an appropriate function of the pressure from the seismic data. This has previously been demonstrated using the long-period data in the 30 to 100 second range, and now the process is applied to the very long period and ultra long period data recorded at IRIS station CCM (Cathedral Cave), located in a cavern in the Ozark Mountains of Missouri.

For the barometric data, a high resolution hybrid micro- barometer system is recorded by auxiliary 24-bit channels of the IRIS-2 data system, with identical sample rates as the seismic data. The barometric data is converted to a rate-of- change value by using a finite-differencing algorithm. The resulting barometric pseudo-velocity series is appropriately scaled to match the pre-signal seismic noise with a correlation of better than 95%. Then it is subtracted from each of the horizontal seismometer data series, which have been time shifted to accommodate the delay of the differencing process. The variance of the pre-event seismic data is reduced by 85 to 95%. The effect is a reduction of the noise in the VLP data by 10 to 30 db in power spectral density, to levels comparable to the noise level of the vertical, with the greatest improvement so far in the 100 to 1000 second range. Similar success has been achieved with the ULP data, where the method improves the resolution of earth tide harmonics in 10 to 30 day series.