It is a common experience that undesired noise is invariably recorded during seismic data acquisition. The real signal is masked by the noise that may be coherent or random in nature. Coherent noise is generally identifiable by their apparent velocities. The aim of seismic data processing is to identify the noise, attenuate it and improve signal to noise ratio. In this paper, our aim is to remove coherent noise in shot domain before any subsequent data processing is taken up. Although f-k fan filters are commonly used to remove coherent noise by defining fan-like region in f-k spectrum but these are not applied on NMO corrected shot gathers. In our method, NMO correction is applied on shot gathers using reasonable velocities in such a way that primaries are either perfectly corrected or over-corrected and coherent noise remains under-corrected. Since events having different dips in t-x domain occupy separate zones in f-k spectrum, NMO corrected shot gathers are f-k filtered to remove under-corrected events that represent the coherent noise. Data thus f-k filtered is passed through inverse NMO process to remove any changes made to the input shot records during application of NMO correction. Subsequent processing of data is carried out as usual. Significant improvements are seen in data processed with this technique.
The salient features of this technique are two folds. The first is that the NMO correction before f-k filtering widens the separation between real signal and the noise and the second is that the foldage in shot domain remains double of that in CDP domain thus no spatial aliasing takes place.
Noise has always been the most troublesome problem in seismic prospecting. In spite of best efforts, the unwanted noise is always recorded on field seismograms during data acquisition that masks the real signal. The noise may be coherent or ambient in nature. Two primary approaches are generally used for suppression of noise to enhance desired signal. One is to setup a recording arrangement that will cancel the unwanted signals by using arrays of shots or geophones where the noise has the wavelengths that are not appreciably greater than the lengths of the arrays themselves. The other is to process the data after they are recorded. Since the length of most of the arrays is limited to a distance that is not substantially greater than the separation between group centers along the spread, the noise with wavelengths more than a few tens of meters cannot be handled by conventional arrays. The CDP technique is designed to cancel noise of larger wavelengths, regardless of its origin. The technique involves stacking of data traces from common reflection points after applying proper time corrections. The output can be improved further if the noise is eliminated from data before stacking. As the output quality and also success of seismic imaging depends on reliability of velocity estimates, it is of utmost importance to appropriately refine shot records to have high signal to noise ratio. This can be achieved to a large extent during data processing.
Seismic data processing is purely a mathematical process and provides a link between data acquisition and interpretation. Its main aims are two folds. The first is to identify, extract and enhance the signal or identify the noise, attenuate it and improve signal to noise ratio and the second is to display the data in an appropriate form so that the interpreter can draw maximum advantage out of it about earth subsurface. The focus in this paper is to remove coherent noise from seismic data during processing. This has been achieved by applying NMO correction on data in shot domain in such a way that real signals are well separated from the noise. Then the noise has been eliminated effectively using f-k filters without affecting the primaries. F-k filtered data is converted back to its original shape by applying inverse NMO correction for subsequent processing.
Theory and Principle
The principle of attenuating coherent noise from seismic data in shot domain is based on the following observations:
Shot related noise generally remains more coherent in shot domain.