Signal-to-noise study in low-frequency passive seismic survey
Published in
72nd EAGE Conference & Exhibition
Date of publication
14 June 2010
Abstract
Spectral attributes of the low-frequency (LF) ambient wave field have been found to correlate with hydrocarbon (HC) reservoirs throughout the world. A major challenge today for using the LF attributes as a HC detection tool is surface wave noise. Knowing the signal-to-noise ratio (SNR) can provide insight for a better understanding of the LF phenomenon and as a guide for the noise tolerance of LF attribute based HC detection. We analyse a noisy dataset acquired in an urban area in Germany over a known oil reservoir with constrained synthetic noise models to address the SNR question in the area. Seven SNR scenarios were modeled and these synthetic datasets were used to train neural networks for HC detection. The performance of these predictors on the real dataset was used as an objective measure to estimate the SNR present in the actual data attributes. We estimate the SNR for the field data to be greater than 0.06 but less than 1.01. Based on the synthetic data alone, we estimate the minimum SNR allowable for reliable HC detection to be 0.06 for this survey. To our knowledge, this work represents a first attempt at quantitatively describing the noise tolerance of HC detection based on spectral LF features.
