Time-domain Reverse Time Migration And Angle Gathers Construction

Reverse time migration (RTM) has been a standard imaging tool for complex structure forits characteristics such as high precision, being free from angle limitation and adapting tostrongly lateral velocity variation, so it has a great value in industry. RTM mostly usecross-correlation imaging condition. Because this imaging condition requires sourcewavefields and receiver wavefields could be available simultaneously, we need to input andoutput the whole process of source wavefield propagation or receiver wavefield propagation,which arouses significant storage and I/O problems. In this paper we use source wavefieldreconstruction to solve the problem. It begins with the modeling of source wavefields andsave the boundary condition and initial condition in this period. Then reconstruct the sourcewavefield in reverse time direction, together with the modeling of receiver wavefields.Another problem caused by cross-correlation imaging condition is the RTM artifacts. In thispaper, we propose up-down, left-right going wavefield decomposition to remove the artifacts,and for the aim of structure image, we could combine it with Laplace filter to make the imageclearer. Source wavefield reconstruction and wavefield decomposition imaging conditioncould arouse extra computational complexity, and for3D industry problem, the memoryrequirement is still large. Since RTM always requires the migration aperture to be wider thanthe target imaging region, the simulation area should be wider than the target area. Therefore,in this paper, we propose a scheme called source wavefield reconstruction in target imagingregion to solve the problem. We simply perform source wavefield reconstruction within thetarget imaging area that we mostly concern, which means only the boundaries of this regionneed to be stored, instead of the boundaries of the whole simulation region. Thereconstruction process and wavefield decomposition are also limited in this domain. RTMwith this scheme is verified to be effective. Compared to conventional scheme of sourcewavefield reconstruction, both the memory requirements and computing time are reducedwithout losing any accuracy.Angle gathers are also important outputs of RTM. Though the ray-based migration schemecould generate angle gathers naturally, they need to face the defects of themselves. Therefore,the RTM-based angle gathers attracts much concern. However, though methods those requirewavefield decomposition or extended imaging condition could produce high-quality gathers,their computational complexity is large in3D condition and considerable computational timeshould be spent. The scheme of Poynting vectors is an efficient way to output RTM anglegathers and very computationally attractive. However, it always suffers from instability. Oneimportant reason for this is that the Poynting vector is composed by the time derivative andthe space derivative. The absolute values of both the two kinds of derivatives would be verylow and near to zero when the values of wavefields are large. For the dominant wavefields domost contribution to the final image, this would result in instability. We propose a schemecalled time-lapse Poynting vector to solve the problem. In the method, the computation ofPoynting vectors would be several time steps later than the wavefields they describe, whichmeans we separate the time of computing Poynting vectors and imaging to ensure the absolute values of Poynting vectors corresponding to the dominant part of wavefield could be largeenough. This scheme could improve the stability and accuracy of Poynting vectors. Also,because the reflection angles computed by Poynting vectors are always not integers, weintroduce the Gaussian weighted function in the imaging process to solve the problem, andthis function could also help decrease the noise.Moreover, we combine the time-lapse Poynting vector with source wavefieldreconstruction in target imaging region. The Poynting vectors could be reconstructed well bythis method. The scheme in this paper could generate RTM angle gathers well and efficiently.