Reconstrução multidimensional de dados sísmicos 2D usando métodos de Fourier e empilhamento por superfície de reflexão comum

Abstract

One of the great challenges in land seismic imaging is in data acquisition, as these are often sampled irregularly with many shots and missing traces, due to logistical constraints in the distribution of sources and receivers. This causes a serious problem of coverage of complex geological structures of interest in the subsurface, in addition to making it difficult to apply processes to treat the data and obtain reliable images for a better interpretation of these structures. Thus, regularization and interpolation of seismic data is an essential step to obtain better results and facilitate the subsequent stages of processing these data. Techniques based on the Fourier transform are cost-effective (quality/computational cost) due to the ability to capture both correct amplitude and continuity of events. Some of the main algorithms are Minimum Weighted Norm Interpolation (MWNI), Matching Pursuit Fourier Interpolation (MPFI) and Anti-Leakage Fourier Transform (ALFT). Despite being widely used in the industry, the literature does not include a comparison between these three algorithms. We propose a comparative study between the three algorithms mentioned. Numerical examples using 1D and 2D harmonic functions exemplify the operation and show the ability of methods based on the Fourier transform to deal with irregular sampling. Application in the synthetic section shows the ability to deal with gaps between seismic traces. The application to the synthetic Marmousi data, missing almost 80% of the data and more than half of the shots, shows the ability of these algorithms to reconstruct seismic data with poor spatial sampling and coming from complex geological environments. The application to the Tacutu data shows that MPFI and ALFT have similar and better results than MWNI. However, MWNI is still the most efficient. It is well known that wave equation methods/algorithms, based on the Kirchhoff integral operator or similar, have the best results. One of the algorithms that provides high quality results is interpolation based on the Common Reflection Surface (CRS) stacking operator. Its main disadvantages are its excessive computational cost and the presence of artifacts, due to the presence of gaps or sparse sampling. We propose two new workflows (strategies) to solve these CRS issues. We use Fourier algorithms (MWNI and MPFI) to precondition the original seismic data before searching the kinematic wavefield attributes (CRS attributes). We then apply interpolation based on the CRS partial stacking operator with the enhanced attributes to the original data (strategy 1) and to the preconditioned data with Fourier algorithms (strategy 2). We applied the new strategies to 2D land seismic data from the Tacutu and Parna´ıba basins. The results showed that both strategies had satisfactory results, significantly attenuating artifacts and improving the quality of shallow events, especially in relation to the algorithms used separately (MWNI, MPFI and CRS).