A statistical strategy for ambient seismic wavefield analysis: investigating correlations to a hydrocarbon reservoir

Theoretical work and modelling studies have led to the hypothesis that the ambient seismic wave field on the surface can be affected by hydrocarbon reservoirs (>800 m depth). Several field studies have linked spectral features on the vertical component between 1 and 10 Hz to reservoir locations. However, such evidence has been criticized due to concerns that surface recordings typically contain a large amount of surface wave noise and correlations to hydrocarbon targets could be caused by non-hydrocarbon variables such as topography or weathering layer thickness. In this paper, we suggest a two-step analysis strategy to address such issues. First, spectral power is only averaged over time periods and frequencies where the distribution of polarization attributes show no obvious dominance of a few surface wave sources. An interferometric test reveals differences in the wave field composition between the filtered and unfiltered data. Second, the residual seismic power is correlated to hydrocarbon as well as non-hydrocarbon targets. The correlations are quantitatively compared using rank correlation and bootstrap confidence intervals. The method is illustrated on a passive seismic data set acquired with three-component, broad-band seismometers at the tight-gas Jonah field in Wyoming, USA. We find evidence that the wave field was dominated by a small number of surface sources in all of the data except for the quietest time periods in the low-frequency range 1.5-3.0 Hz. Seismic power within this subset significantly correlates to a published reservoir map but not with a digital elevation model and less so with an infrastructure density map. The investigated hypothesis can thus not be rejected with this data.