Arvind K. Saibaba,Tania Bakhos,Peter K. Kitanidis
Abstract:We discuss efficient solutions to systems of shifted linear systems arising in computations for oscillatory hydraulic tomography (OHT). The reconstruction of hydrogeological parameters such as hydraulic conductivity and specific storage using limited discrete measurements of pressure (head) obtained from sequential oscillatory pumping tests, leads to a nonlinear inverse problem. We tackle this using the quasi-linear geostatistical approach \cite{kitanidis1995quasi}. This method requires repeated solution of the forward (and adjoint) problem for multiple frequencies, for which we use flexible preconditioned Krylov subspace solvers specifically designed for shifted systems based on ideas in \cite{gu2007flexible}. The solvers allow the preconditioner to change at each iteration. We analyze the convergence of the solver and perform an error analysis when an iterative solver is used for inverting the preconditioner matrices. Finally, we apply our algorithm to a challenging application taken from oscillatory hydraulic tomography to demonstrate the computational gains by using the resulting method.
What problem does this paper attempt to address?
This paper attempts to solve the computational efficiency problem when solving multiple frequency - shifted linear systems in Oscillatory Hydraulic Tomography (OHT). Specifically, the paper focuses on how to efficiently solve the frequency - shifted linear systems caused by different frequencies, which play a crucial role in the process of reconstructing hydrogeological parameters such as permeability and specific storage rate.
### Problem Background
Oscillatory Hydraulic Tomography (OHT) is a technique for characterizing underground aquifers. It estimates important aquifer parameters by applying oscillatory signals in wells and measuring pressure changes in receiving wells. One of the significant advantages of OHT is that it can extract additional information by using signals of different frequencies without drilling more wells. However, this method involves a large amount of computational burden, especially when multiple frequencies need to be processed.
### Core Problems of the Paper
1. **Non - linear Inverse Problem**: Reconstructing hydrogeological parameters (such as permeability and specific storage rate) from limited discrete pressure measurement data is a non - linear inverse problem.
2. **Solution of Frequency - Shifted Systems**: To solve this inverse problem, it is necessary to repeatedly solve the forward (and adjoint) problems, which involve frequency - shifted linear systems of multiple frequencies.
3. **Computational Efficiency**: When a large number of frequencies are used (for example, 200 frequencies), direct methods (such as sparse LU, Cholesky or LDLT decomposition) will become very expensive because the matrix needs to be re - decomposed every time the frequency changes. Therefore, a method needs to be developed so that the cost of solving multiple frequencies is not significantly higher than that of solving a single frequency.
### Solution
The paper proposes a flexible Krylov subspace method, which is specifically optimized for frequency - shifted systems. The main contributions of this method include:
1. **Extension of the Flexible Arnoldi Algorithm**: It can handle frequency - shifted systems in the form of \((K+\sigma_j M)x_j = b\) and use multiple pre - conditioners \((K +\tau M)\).
2. **Convergence Analysis**: It provides a convergence analysis of the solver.
3. **Error Analysis**: When an iterative solver is used to apply the pre - conditioner, an error analysis is derived to determine the stopping tolerance for convergence monitoring without constructing the full residual.
4. **Practical Application**: This method is applied to OHT, demonstrating its computational acceleration effect in practical problems.
### Conclusion
By introducing the flexible Krylov subspace method, the paper successfully improves the efficiency of solving multiple frequency - shifted linear systems, thus providing a more efficient solution for applications such as OHT. This not only reduces the computational time but also makes it possible to process more frequencies, thereby improving the image quality.