In shallow seismic survey source generated coherent noises have a wide range of spectral overlap with seismic signals. Consequently, conventional band-pass filtering could not suppress them. Their elimination requires the implementation of some special techniques that may not commonly used for this purpose. The K-L decomposition technique treats each trace as a data vector, and computes a set of uncorrelated principal component traces from an eigenvalue decomposition of the matrix of zerlag cross-covariance of the given multi-trace input data set. The principal component traces are arranged in order of decreasing energy content, i. e. the signal with the largest variance will appear as first principal component and so on. Subsequently, the input record is reconstructed utilizing only the information contained in a specified selection of the principal component traces, those associated with large eigenvalues. This amounts to reconstruction of the coherent energy present in the input data set. In this study the K-L transformation technique was applied on 96-trace shot records of 3-D seismic survey flattened on the air wave, then the first five of the 96 principal components were used to estimate the air wave components of the records. The technique was found an effective way of eliminating air waves from engineering-scale 3-D seismic shot records.
In shallow seismic survey source generated coherent noises have a wide range of spectral overlap with seismic signals. Consequently, conventional band-pass filtering could not suppress them. Their elimination requires the implementation of some special techniques that may not commonly used for this purpose. The K-L decomposition technique treats each trace as a data vector, and computes a set of uncorrelated principal component traces from an eigenvalue decomposition of the matrix of zerlag cross-covariance of the given multi-trace input data set. The principal component traces are arranged in order of decreasing energy content, i. e. the signal with the largest variance will appear as first principal component and so on. Subsequently, the input record is reconstructed utilizing only the information contained in a specified selection of the principal component traces, those associated with large eigenvalues. This amounts to reconstruction of the coherent energy present in the input data set. In this study the K-L transformation technique was applied on 96-trace shot records of 3-D seismic survey flattened on the air wave, then the first five of the 96 principal components were used to estimate the air wave components of the records. The technique was found an effective way of eliminating air waves from engineering-scale 3-D seismic shot records.