Constructing Initial Models for Full Waveform Inversion using Strip-off Controlled Directional Reception Velocity Analysis

Abstract

The controlled directional reception (CDR) method is a velocity analysis method using ray-tracing. It is one of the tomographic methods that use slope (or ray parameter), so it is often called the "slope tomography method". It does not require a pre-picking operation like traveltime tomography does. Auto-picked information from the local slant stack is regarded as more reliable than reflection traveltime picked directly from the seismic data. The method also provides more detailed information about the moveout than the imaging operator in migration-based velocity analysis (MVA). Therefore, we constructed a velocity macro-model using this strip-off CDR velocity analysis. When compared to the conventional CDR method, it increased the resolution of common receiver gathers (CRG) data and reduced computer storage space dramatically. Additionally, it improved the accuracy of the velocity model by using the migrated image as a background panel during the velocity analysis.

The results obtained by this method were applied to full waveform inversion (FWI) as the initial velocity model. In FWI, an exact initial model is important because it reduces instability and increases the probability of convergence to the global minimum. It is significant that the CDR model is first applied as the initial model of FWI. We confirmed good inverted results from two realistic synthetic data tests by comparison with the results obtained using the conventional Laplace inverted model and the linearly increasing model. In addition, the CDR macro-model has strength to seismic data with a weathered layer or short offset. Furthermore, it is possible to apply to multi-parameter inversion. Finally, we performed this method on real seismic data. Although we could not know the true velocity model for the real seismic data, we confirmed that the RMS error of the CDR model was lower than that of the Laplace inverted model. Additionally, the migrated images and common image gathers (CIGs) also proved that the inverted CDR model showed a good result.

Strip-off CDR velocity analysis has a disadvantage in computing time compared with Laplace domain inversion. However, even if the computation time is greater, the method has great value because of its high accuracy. In particular, it is expected to provide good results with difficult data, such as seismic data with a weathered zone or short offset, and so increase the accuracy compared with the conventional method. CDR velocity analysis also may be useful for work favoring higher accuracy over fast calculation time, such as some resource exploration and geological surveying. Furthermore, the good initial model can reduce computing time for inversion as well as increase the accuracy of the inverted results.

In summary, the macro-model obtained from strip-off CDR velocity analysis is suitable for frequency domain FWI. Three-dimensional exploration and exploration in complex terrains are being conducted more often, so seismic data with short offset or recording time are also encountered more frequently. This method will demonstrate strength with these seismic data. Additionally, it can be expected to be applicable to land seismic or ocean bottom seismic (OBS) data and also extend into other inversion fields such as multi-parameter inversion.