Low-frequencies that are absent in seismic data impair accurate estimation of acoustic impedance. There are several methods for restoring these missing frequencies, they can be estimated by model inversion, borrowed from well logs, predicted using the available frequencies in the spectra, or recorded in the field. This paper uses the Hussar low-frequency experiment to investigate the low-frequencies available in the data recorded by three types of receivers and four sources. The receivers used were 10Hz geophones, 4.5Hz geophones and Vectorseis MEM accelerometers. The sources used were 2 kg of dynamite at 15 m, low-dwell sweep done by an INOVA 364 vibrator (a low frequency source), a linear sweep done by the INOVA 364 and a low-dwell sweep done with a Failing vibroseis source. We compare stacked data, created with a flow that included Gabor decon, to the sonic and density logs in well 12-27. From each of 12 stacks, we constructed average traces at the well location by summing over a 230m wide spatial window centered on the well. To compare the different types of receivers to each other the geophones were correlated and phase shifted to match the accelerometer data for each source. A synthetic seismogram was created using the well logs and convolved with a wavelet designed with the average smoothed amplitude spectrum of the different receivers and with a phase to match the accelerometer data, Figure 1A. The amplitude spectra of the traces were compared and were found to have a 20 decibel roll-off in the low frequencies. A similar roll-off was seen in reflection coefficients constructed from the well logs. The low frequency portion of the stacked traces was compared to the low frequency portion of the well reflection coefficients and they were found to be roughly comparable down to about 2 Hz, Figure 1B. This is an encouraging result as it suggests that impedance inversion can be done using only a very minimal frequency band from nonseismic (i.e. well) data.
View full article as PDF (2.21 Mb)