Marine acquisition: Moving beyond the signal-to-noise ratio?

Andrew Long, PGS Technology (Perth), offers a challenge to conventional thinking on signal-to-noise ratio in the context of 3D multi-streamer acquisition surveys arguing for more attention to all the factors contributing to ‘noise’ in seismic images. The definition of exactly what comprises ‘signal’ and ‘noise’ on seismic data is ill-defined, but it can be said that noise is the unwanted component of the target frequency spectra, which is not directly related to, or correlated with, the primary reflection energy. If signal and noise could be separated into two distinct amplitude spectra, then the maxima of the spectra may be closely aligned (less easily separable), or more distinct (more easily separable). Irrespective of the comparative spectra, the signal- to-noise ratio (S/N ratio) is typically defined as being simply the logarithmic ratio of the maximum amplitudes from each of the signal and noise spectra. This is obviously too vague to be usefully descriptive of data quality, however, the S/N ratio is a universally ascribed term in seismic data analysis. Quite commonly, the term is used as a more qualitative or colloquial description of general data coherency and resolution. Improvements in S/N quality are generally attributed to the fold of stack. By the well-known square root relationship, increasing the stack fold suppresses random (not coherent) noise. However, many other factors contribute to the ‘noise’ contaminating seismic images, as discussed below. Historical efforts have been made to quantify the nature of the S/N ratio, and its relationship to data quality. Junger (1964) observed that ‘For a signal-to-noise ratio greater than two, the signal predominates visually, and only a slight improvement in quality can be obtained with additional improvements in the signal-to-noise ratio’. Hence, it is observed that once the random noise component is suppressed below a certain threshold, other factors than mere fold are clearly contributing to the quality of the seismic image. It is quite poorly established how more complicated acquisition parameters, such as multi-streamer spread dimensions and shooting templates, influence the ‘S/N ratio’ of seismic data, particularly after the application of multichannel pre-stack processing algorithms, notably pre-stack migration. In the sections below, I describe how ‘noise’ is manifested both during acquisition and processing, in the context of 3D marine multi-streamer acquisition parameters. I demonstrate how misleading and inappropriate it is to only consider data quality in terms of fold and simple S/N ratio measurements.