Converted wave imaging in anisotropic media using sea-floor seismic data
The aim of this work is to improve the practice of multicomponent data processing in the time domain. I present a detailed study carried out on a 2D multicomponent dataset acquired over the Lomond Field, North Sea. I show that this area is seismically anisotropic and that failure to account for the anisotropy leads to poor converted wave imaging results. Anisotropy is included in a complex model-building scheme prior to Pre-Stack Time Migration (PSTM). The basic parameters required in converted wave processing are the converted-wave stacking velocity based on non-hyperbolic moveout and different P-wave to S-wave velocity ratios. These parameters are extracted from analysis on asymptotically binned gathers, that is, gathers binned with a constant value of the velocity ratio vp/vs. I present results of a sensitivity analysis and I show that in areas affected by dip the stacking velocity is sensitive to changes in the initial vp/vs ratio. These small velocity errors are propagated as the square in the re-calculation of the depth-variant velocity ratio and cannot be ignored. I show that using imaging criteria to define the binning velocity ratio provides a valid and velocity-independent estimate in zones of complex geology. The vertical velocity ratio is derived conventionally by event matching in the P-wave and converted wave stacks. I present an attempt to use well-log derived velocity ratios to avoid this interpretative step. The velocity ratio derived from 4C seismic data is about 30% higher than that derived from well logs. I analyse three possible causes for this discrepancy: the effects of gas, polar anisotropy and frequency-dependent dispersion. Gas has little effect in the Lomond Field logs, while polar anisotropy lowers the well-log derived vp/vs ratio by about 15%. Frequency-dependent dispersion also lowers the well-log derived velocity ratio, but it is difficult to quantify. Residual errors in the seismic interpretation have also to be considered. Importantly, I prove that the ratio leading to the best image is the one derived from seismic data, which suggests that the use of the raw well-log derived velocity ratio in multicomponent processing should be avoided. I quantify anisotropy using an effective parameter, representing converted-wave anisotropy, ceff, which is a combination of P- and S-wave anisotropy. This parameter can be estimated from converted wave seismic data alone and I illustrate two different ways of extracting it. I present imaging results from a full anisotropic PSTM processing sequence. This flow requires careful model building and allows updating in the time-migrated domain. Comparing the values of the anisotropic parameter and of the binning velocity ratio before and after PSTM highlights the difference between the initial model and the updated model. Both parameters are in fact sensitive to the presence of dip. I show that the values of the anisotropic parameter change after PSTM, suggesting that part of the residual moveout attributed to anisotropy prior to PSTM was caused by dip. This consideration confirms the importance of defining the model in the time-migrated domain. The PSTM image matches with a high degree of accuracy the geological interpretation carried out by BG Group. PSTM tests show that the inclusion of anisotropy allows the use of the full range of offsets, which is important to produce the correct image of the target area. I compare this result with the image obtained from a flow based on isotropic Dip Moveout (DMO) and post stack migration. Differences in the position of the steep-dipping events and geological misties are evident in the post-stack migrated image. This mis-positioning is due to the isotropic approximation and to the limitation of the DMO and post-stack migration flow. I also present results of an integrated analysis of local geology, well logs and seismic data to confirm the presence of polar anisotropy in the Lomond Field. The sediments forming the overburden are mainly composed of finely laminated shales. The image I obtained from the full Pre-Stack Depth Migration on P-P data reveals a depth mismatch with the well markers. Since the pre-stack gathers show that the correct velocities are applied, this depth mismatch has to be attributed to the presence of anisotropy. Other clear evidences of anisotropy come from well logs. P-velocity angular dependency is evident in sonic logs. I show that a similar angular dependency also exists when comparing interval velocities and average velocities from seismic data and from vertical well logs and check shots. These considerations leave little doubt that the Lomond Field is seismically anisotropic.