1 Centre for Development of Advanced Computing (C-DAC), Pune University Campus, Pune, India. e-mail: kirank@cdac.in 10 th Biennial International Conference & Exposition P 028 Recent Developments in Spectral Decomposition of Seismic Data (Techniques and Applications): A Review Kiran Khonde* 1 , Richa Rastogi 1 Summary This paper presents a review of spectral decomposition of seismic data, since it's inception nearly in 1997. It discusses various techniques and applications of spectral decomposition in seismic data processing and interpretation. It is also known as time- frequency analysis consists of transforming non stationary signal in time/space from time/space domain to time/space vs frequency domain. The frequency domain representation illustrates many important features that are not apparent in time domain representation. Spectral decomposition is a non-unique process for which various techniques exists and newer modified techniques are being discovered. Over the years, spectral decomposition of seismic data has progressed from tool for stratigraphy analysis to direct hydrocarbon indicator (DHI) technique. This technique is mostly used by seismic interpreters and being DHI, it is a potential weapon for minimizing dry well drilling. In coming time, spectral decomposition may can play a significant role in analyzing time lapse seismic data. Keywords: Spectral decomposition, Time-frequency analysis, Direct hydrocarbon indicator, Matching pursuit decomposition. Introduction Spectral decomposition of seismic data is a mathematical tool of transforming seismic data from time domain to time vs frequency domain. It converts one dimensional seismic trace in time domain to two dimensional time vs frequency domain representation, thus provides information on variation of frequency with time. In a similar way, it converts two dimensional seismic section to three dimensional representation, third axis being frequency. The frequency domain representation of seismic data illustrates many features that are not apparent in time domain representation and hence spectral decomposition serves as an useful tool for seismic interpreters. Early work Early research work on spectral decomposition was performed by Greg Partyka. He introduced concept of spectral decomposition and was eventually awarded by SEG Virgil Kauffman Gold Medal in 2003 for the same (Cooper, 2004; Partyka, 2007). Gridley and Partyka (1997) used STFT based spectral decomposition to analyse seismic features as a function of frequency amplitude and phase which is helpful in interpretation. Partyka et al. (1999) used spectral decomposition for imaging and mapping temporal bed thickness and geologic discontinuities in 3D seismic data. Present status Currently, research in spectral decomposition can be broadly classified into two categories. In the first category (Category-A), researchers are trying to discover newer and accurate techniques for time- frequency analysis of seismic trace. In the second category (Category-B), researchers are trying to identify novel applications of spectral decomposition in studying different geological environment. Category-A: Research in developing advanced techniques of spectral decomposition Basic time-frequency analysis technique is Fourier transform which is an excellent tool to study stationary signals. For stationary as well as non-stationary signals, Fourier transform provides global picture of frequency in signal but incapable of providing local variation of frequency. The immediate solution that was put forward by Dennis Gabor for time-frequency analysis of non-