Low-Frequency Seismic Noise Characteristics From the Analysis of Co-Located Seismic and Pressure Data

Abstract

We analyze the characteristics of seismic noise generated by surface pressure changes, using co-located seismic and pressure data from the USArray Transportable Array for the entire year of 2014. We compare pressure and seismic signals at frequencies of 0.01 and 0.02 Hz and model their main characteristics. At these frequencies, vertical and horizontal seismic noise can be explained by the formulas SZ = A * Sp + NZ and SH = B * Sp + NH, where SZ and SH are the power spectral densities of vertical and horizontal ground velocities (sum of two horizontal components), Sp is the power spectral density of pressure, and NZ and NH are constant terms. We first determine A and B from the high pressure range (Sp > 10 Pa2/Hz) in which NZ and NH are small. From A and B, we then determine the near-surface rigidity and pressure wave speed. This rigidity is modified from true rigidity by a factor (λ + μ)/(λ + 2 μ); this factor is about 0.7 for typical crustal rocks but approaches 1 near the Earth's surface. To estimate true rigidity, empirical formulas that relate density and seismic velocities are needed. Once A and B are determined, NZ and NH are determined. NZ and rigidity of the subsurface determine a threshold pressure in vertical data. NH is generally small but at some stations it changes seasonally and creates separate branches in plots of pressure versus seismic velocity. This may explain why seismic-noise reduction approaches using co-located pressure data often fail to work for horizontal data but usually work well for vertical data.