|
|
|
| Anisotropic tomography with rock physics constraints | |
|
Next: Acknowledgement
Up: Li et al.: RP
Previous: Posterior Uncertainty Analysis
In this paper, we have proposed a new formulation to incorporate rock physics prior information with the anisotropic tomography. Two models were analyzed using this method, and the inversion results demonstrate the trade-off among the parameters and the instability due to the huge null-space when no prior information is included. Any estimation of the local cross-parameter distribution (column weighting, diagonal covariance and full covariance) is helpful to stabilize the inversion and leads to a better representation of the subsurface. However, we should be careful in using too tight of a prior distribution when the lithology is uncertain, especially for areas where parameters are not well-constrained by the data. The posterior distribution analysis shows that by adding the rock physics prior information, we will obtain a better estimation of the true prior statistics in the inversion and a smaller uncertainty in the posterior statistics.
The experiment of rock physics constrained tomography suggests to us a new workflow in anisotropic model building.
- First, Build an initial model using the deterministic rock physics modeling and obtain the initial image.
- Second, Build the point-by-point cross-parameter covariance according to the stochastic rock physics modeling.
Build the spatial covariance using geology information and/or the initial image.
- Third, run the rock physics constrained joint tomography with surface seismic data and borehole data.
Repeat the workflow if necessary. Up to now, it is possible to use all the information: surface reflection seismic, borehole data, geological estimation and the rock physics covariance in the tomography to produce a unique earth model that explains the seismic data and satisfies the geological and rock physics theory at the same time.
|
|
|
| Anisotropic tomography with rock physics constraints | |
|
Next: Acknowledgement
Up: Li et al.: RP
Previous: Posterior Uncertainty Analysis
2011-05-24