Imaging multiples by data to data migration

In conventional migration methods, free surface related multiples are regarded as noise and only the primary wave is utilized in imaging. However, the multiples also contain information of subsurface structures and can offer extra illumination in migration procedure. There are already numerous methods to take advantage of multiples, but most of them need multiples prediction, which is time consuming. To avoid multiples prediction and wavelet estimation, this paper presents a new multiples migration method called data to data migration. Based on the migration method that uses primaries and multiples simultaneously, this paper presents one-way wave equation migration of data to data migration. The FFD (Fourier finite difference) method is employed in wave field extrapolation. For the data containing primaries and free-surface related multiples, the data to data migration replaces the source wavelet function and the primaries in conventional FFD migration with the recorded data. The same crosscorrelation imaging condition is used. In the FFD migration method, the field data is transformed into the frequency domain. Compared to RTM(reverse time migration), it has higher computation efficiency and less low frequency noise. The FFD is more suitable for the implementation of data to data migration. Compared to conventional FFD migration that uses primaries only, the images generated by data to data migration based on FFD operator have better illumination and higher resolution in the shallow zone. The images of data to data migration contain artifacts produced by the corsscorrelation of different seismic events. In the deep zone, the migration results are inferior to conventional migration. One reason is the influence of artifacts; another is that multiples need longer receiving time. In the dataset, higher order multiples may not yet be received completely. The results of the conventional migration and data to data migration have different polarities. Assumed the wavelet and the primaries have positive polarity, and then the conventional migration result has a positive polarity. The first order multiple is reflected by the free surface once, so it has negative polarity. It is crosscorrelated with the primary that has positive polarity and generates the result with negative polarity. The second order multiple that has positive polarity is crosscorrelated with the first order multiple that has negative polarity and generates the result with negative polarity. And the m+1 order multiple is crosscorrelated with the m order multiple which have different polarity and generates the result with negative polarity. Therefore the final result of data to data migration has negative polarity and it has different polarity with conventional migration. The computational speed of FFD is much faster than RTM which makes FFD appropriate for the multiples data with long record length. Recorded data without direct wave are utilized in data to data migration procedure. The numerical examples verify its effectiveness. The FFD operator improves the computational efficiency greatly. The proposed approach has three key advantages: (1) Compared to conventional migration method, data to data migration has wider illumination area and better resolution of scattering points. (2) It can generate subsurface image without multiples prediction, which is time consuming. (3) It can offer fairly good image of shallow reflectors. Another point is wavelet estimation. For the field data, how to extract wavelet function is intricate and prone to error. Data to data migration avoids this complicated process, which may be significant for imaging subsurface structures in the processing of field data. We should also notice that although most energy are imaging correctly, there are still artifacts in the final result.Least squares migration and wide azimuth acquisition technology will partly decrease the artifacts.