Spectral analysis of ambient ground-motion—Noise reduction techniques and a methodology for mapping horizontal inhomogeneity

Inhomogeneities in the subsurface can interact with and modify seismic waves. As a consequence, the quasi-stationary, ambient ground-motion response at the Earth's surface carries valuable information about underground inhomogeneities. Extracting and interpreting such information by analyzing low-frequency passive measurements is not an easy task because distinct events in the time traces are usually absent in ambient ground-motion recordings and the signal-to-noise ratio is low. This paper presents four processing techniques to reduce different types of anthropogenic noise effects and proposes a methodology for extracting and mapping passive seismic characteristics that help image horizontal inhomogeneity. The methodology uses Fourier amplitude spectra to calculate four independent spectral attributes that quantify signatures of interest in the data. The methodology is applied to a noisy data-set acquired at an oil and gas field in Austria. Two horizontally separated hydrocarbon reservoirs are considered as possible inhomogeneities in the subsurface. Results show that anthropogenic noise is successfully removed from the data, enhancing the signal-to-noise ratio. Attribute profiles are extracted and show stable spatial patterns that are apparently related to stationary site properties. Correlation between attribute patterns at the Earth's surface and the horizontal location of the reservoirs is discussed. Although this paper addresses an application for hydrocarbon reservoir characterization, the proposed methodology is suitable to quantify and map signatures related to any subsurface inhomogeneity of interest (e.g. underground cavities) that has a measurable impact on the ambient ground-motion at the Earth's surface.

► Discriminating signals of interest from anthropogenic noise in ambient ground-motion.
► Spectral attributes are introduced to laterally map signals of interest.
► Four processing methods to reduce noise in continuous ground-motion recordings.
► Cleaned data show site specific information, possibly related to the subsurface.
► Correlation of attribute patterns with hydrocarbon reservoirs is ambiguous.