In controlled-source seismic interferometry, waves from a surrounding boundary of sources recorded at two receivers are crosscorrelated and summed to synthesize the interreceiver Green's function. Deviations of physically realistic source and receiver geometries from those required by theory result in errors in the Green's function estimate. These errors are manifested as apparent energy that could not have propagated between receiver locations -so-called nonphysical energy. We have developed a novel method of velocity analysis that uses both the physical and nonphysical wavefield energy in the crosscorrelated data generated between receiver pairs. This method is used to constrain the root-mean-square ͑rms͒ velocity and layer thickness of a locally 1D medium. These estimates are used to compute the piecewise constant interval velocity. Instead of suppressing multiple energy as in conventional common midpoint velocity analysis, the method uses the multiply reflected wavefield to further constrain the rms velocity and layer-thickness estimates. In particular, we determined that the nonphysical energy contains useful physical information. By using the nonphysical energy associated with the truncation of the source boundary and the crosscorrelation of reflected waves, a better-defined estimate of the rms velocity and layer thickness is achieved. Because this energy is excited far from the receiver pair, the technique may be ideally suited to long-offset seismic reflection data. We found that interferometric velocity analysis works best to characterize the first few layers beneath a receiver array. We have considered an acquisition configuration that can be used in a marine seismic setting.

Deconvolution imaging conditions offer improved resolution over standard, crosscorrelation-based imaging conditions. Additionally, these imaging conditions produce a result more directly related to a reflection coefficient than do crosscorrelation-based imaging conditions. In simple analytical cases, deconvolution imaging conditions also offer the possibility of eliminating crosstalk ͑i.e., energy in the image due to reflected energy arriving at a location at the same time as incident energy that did not cause the reflected energy͒ when the full up-and down-going wavefields are used. This means that in such cases, surface-related multiples can be eliminated from the image, or that multiple shots could potentially be fired simultaneously without degrading the image. However, this cross-talk-suppression property is not observed in most situations. We show that this is due to a number of issues: the correct order of deconvolution must be used, stabilization causes imperfect deconvolution, finite apertures lead to some of the signal being lost, and an assumption of horizontal stratification is often not being met. Further, imperfect knowledge of the incident and reflected field due to such factors as anisotropy, poorly estimated velocity fields, and measurement noise can also lead to imperfect deconvolution. Thus, deconvolution imaging conditions should not be counted on to completely eliminate crosstalk from images.

This paper introduces a perturbative inversion algorithm for determining sea floor acoustic properties, which uses modal amplitudes as input data. Perturbative inverse methods have been used in the past to estimate bottom acoustic properties in sediments, but up to this point these methods have used only the modal eigenvalues as input data. As with previous perturbative inversion methods, the one developed in this paper solves the nonlinear inverse problem using a series of approximate, linear steps. Examples of the method applied to synthetic and experimental data are provided to demonstrate the method's feasibility. Finally, it is shown that modal eigenvalue and amplitude perturbation can be combined into a single inversion algorithm that uses all of the potentially available modal data.

When the frequency of a cw source is known precisely, it is possible to relate the time rate-of-change of the measured phase to the time rate-of-change of the distance between source and receiver. However, when the source frequency is not known precisely, or when the source frequency is unstable, an unknown time dependence will remain after demodulation, causing the phase to appear to vary rapidly. Thus, the phase cannot be used to find the rate of separation of the source and receiver. By examining the difference between the phase measured by two independently moving receivers, it is possible to eliminate this unknown time dependence from the phase. In this talk, a source localization method based on this idea is presented and applied to data from a frequency-unstable source (near 20 Hz) in a shallow-water, low-frequency acoustics experiment. [Work supported by ONR and the WHOI Education Office.]

No abstract