Regularization in 3-D DC resistivity tomography
- Massachusetts Inst. of Technology, Cambridge, MA (United States)
We demonstrate that the solution of 3-D DC resistivity tomography has strong numerical artifacts if the inverse problem is not properly regularized. With only few data points but a large number of model parameters (unknowns), the nonlinear inverse problem is ill-posed. Many studies have shown that some kind of model correlation must be constructed to stabilize the inversion. Among those, Tikhonov regularization takes a more explicit approach by damping spatial derivatives of the model function as opposed to applying ad hoc smoothing. However, we show evidence that not all smoothness criteria in the class of Tikhonov methods are well-posed for 3-D DC resistivity inversion. In fact, only under the second- or higher-order derivative regularization, 3-D DC resistivity tomography can produce a physically meaningful solution which has no dependence on the model discretization. In adopting effective smoothness criteria, the solution approximates a continuous function with no more structure than is necessary to fit the data. Further, we demonstrate that using Tikhonov method to regularize the model stepsize rather than the model itself does not improve the ill-posedness of the inverse problem. As the result, only the data misfit has been minimized and model correlation is not constrained. Finally, we apply our tomography approach to model real data collected at the Mojave Generating Station in Laughlin, Nevada. For different model discretization, our approach produces similar subsurface image.
- OSTI ID:
- 381506
- Report Number(s):
- CONF-960477--
- Country of Publication:
- United States
- Language:
- English
Similar Records
Use of regularization method in the determination of ring parameters and orbit correction
Spatially-Variant Tikhonov Regularization for Double-Difference Waveform Inversion