Resolution and uncertainty in spectral decomposition

Over the last five years or so, spectral decomposition has become a mainstream seismic interpretation process. Most often applied qualitatively, for seismic geomorphologic analysis (e.g., Marfurt and Kirlin, 2001), it is increasingly being applied quantitatively, to compute stratal thickness (as described by Partyka et al., 1999). It has also been applied to direct hydrocarbon detection (Castagna et al., 2003). Broadly speaking, we also apply the principles of spectral decomposition any time we extract a wavelet or look at a spectrum. To date, most of this work has been done with the windowed (‘short-time’) Fourier transform, but other ways of computing spectra are joining our collective toolbox: the S transform (Stockwell et al., 1996), wavelet transforms, and matching pursuit to name a few (e.g., Castagna and Sun, 2006).