The virtual source method has recently been proposed to image and monitor below complex and time-varying overburden. The method requires surface shooting recorded at downhole receivers placed below the distorting or changing part of the overburden. Redatuming with the measured Green's function allows the reconstruction of a complete downhole survey as if the sources were also buried at the receiver locations. There are still some challenges that need to be addressed in the virtual source method, such as limited acquisition aperture and energy coming from the overburden. We demonstrate that up-down wavefield separation can substantially improve the quality of virtual source data. First, it allows us to eliminate artifacts associated with the limited acquisition aperture typically used in practice. Second, it allows us to reconstruct a new optimized response in the absence of downgoing reflections and multiples from the overburden. These improvements are illustrated on a synthetic data set of a complex layered model modeled after the Fahud field in Oman, and on ocean-bottom seismic data acquired in the Mars field in the deepwater Gulf of Mexico.

Seismic interferometry and the virtual-source method are related approaches for extracting the Green's function that accounts for wave propagation between receivers by making suitable combinations of the waves recorded at these two receivers. These waves can either be excited by active, controlled, sources, or by natural incoherent sources. We compare this technique with the deconvolution of the wave field recorded at different receivers. We show that the deconvolved wave field is a solution of the same wave equation as that for the physical system, but that the deconvolved wave forms may satisfy different boundary conditions than those of the original system. We apply this deconvolution approach to the wave motion recorded at various levels in a building after an earthquake, and show how to extract the building response for different boundary conditions. Extracting the response of the system with different boundary conditions can be used to enhance, or suppress, the normal-mode response. In seismic exploration this principle can be used for the suppression of surface-related multiples.

Using model and field data, this article reviews the virtual-source method and its acquisition geometry requirements. Before we go into the details of the acquisition geometry requirements, let us briefly review the basic concept and the advantages of the virtual-source method. A typical surface seismic experiment has sources on the surface to excite waves that propagate through the subsurface. Surface receivers record the reflected waves. In order to image the subsurface, we migrate the reflected wavefield recorded by the receivers, using an estimate of the subsurface velocity model. However, the near surface is usually complex, and the velocity is difficult to estimate. These nearsurface inhomogeneities, if not represented in the migration velocity model, defocus the deeper image. In order to avoid the estimation of the near-surface velocity model, Bakulin and Calvert (2006) proposed the virtual-source method, a technique that uses cross-correlation of the wavefield recorded by a given pair of receivers to estimate the response between them.Virtual-source method. Application of the virtual-source method for imaging and time-lapse monitoring below the near-surface overburden requires sources on the surface and receivers below the near-surface, time-varying overburden. Figure 1 depicts the application of the virtual-source method. For the geometry depicted in Figure 1a (sources S1, S2, S3 on the surface and receivers R1 and R2 in the subsurface), Figures 1b and 1c show the receiver gathers corresponding to R1 and R2, respectively. The trace depicted by S1R1 is the wavefield recorded by receiver R1 excited by source S1. A similar notation holds for other traces.The two receiver gathers when cross-correlated pairwise, create a correlation gather (Figure 1d). Horizontally summing the correlation gathers creates a new trace ( Figure 1e) that approximates a signal that would have been recorded at receiver location R2 as if a source were excited at receiver location R1. Figure 1f depicts this virtual source location as VS. Since there is no physical source at the location of R1, the method is known as the virtual-source method. Hence, by cross-correlation and summing, we redatum the data down to the receiver locations without knowledge of the overburden velocities. Figure 1 shows generation of a virtual-source trace by summing the correlation gather over three sources. Theory, however, states that cross-correlation provides the true response between the virtual source and a receiver provided that a continuous distribution of physical sources surround the receivers, as shown in Figure 2a. The triangles A and B depict the receivers, and the dots show the sources. Figure 2b shows the virtual-source data generated after cross-correlation and summing (similar to Figure 1e). For a homogeneous model with velocity c and receivers A and B separated by distance d, cross-correlation and summing results in a nonzero signal for positive and negative times, and, hence, the virtual-source data contain a causal and an acausal pulse...

In the Gulf of Mexico, fault zones are linked with a complex and dynamic system of plumbing in the Earth's subsurface. Here we use time-lapse seismic-reflection imaging to reveal a pulse of fluid ascending rapidly inside one of these fault zones. Such intermittent fault 'burping' is likely to be an important factor in the migration of subsurface hydrocarbons.