Frequency estimates of seismic traces

Local estimates of amplitude, frequency, and phase have been used in the past to characterize seismic data. In particular, these attributes have sometimes been successfully related to well attributes at the reservoir scale (net pay thickness, sand fraction, etc.). This paper introduces a method called SINFIT for computing local amplitude, frequency, and phase estimates of seismic traces over short-time windows. The SINFIT method uses a sine-curve fitting approach. The method is shown to give more accurate and robust frequency estimates than four other common methods on a set of test traces where the true frequency components are known. The four methods compared with SINFIT are instantaneous frequency, zero-crossings, short-time Fourier analysis, and a more recent time-frequency method called AOK. In a field case with fluvial sands, an average frequency over a 30-ms time window of seismic data correlates with estimated shale volume from well logs. The SINFIT method gives an average frequency attribute that more strongly correlates with shale volume than corresponding attributes from any of the other four methods.