Application of Large-Scale Inversion Algorithms to Hydraulic Tomography in an Alluvial Aquifer

Area of Science:

• Geosciences
• Hydrogeology
• Computational Methods

Background:

• Large-scale inversion methods are crucial for reducing computational demands in complex subsurface modeling.
• Hydraulic tomography is a key technique for characterizing aquifer properties.

Purpose of the Study:

• To apply a deterministic geostatistical inversion algorithm for 2D spatial transmissivity modeling in an alluvial aquifer.
• To evaluate the efficiency and accuracy of the Principal Component Geostatistical Approach (PCGA) for large-scale inversions.

Main Methods:

• Application of a quasi-Newton iterative Bayesian inversion algorithm.
• Comparison of sensitivity analysis methods: adjoint-state, finite-difference, and PCGA.
• Reconstruction of high-resolution transmissivity fields using PCGA and hydraulic tomography.

Main Results:

• High-resolution transmissivity fields (up to 25,600 cells) were reconstructed, showing good correlation between measured and computed hydraulic heads.
• The PCGA method, combined with hydraulic tomography, substantially reduced inversion computation time.
• PCGA yielded high-quality inversion results comparable to other sensitivity analysis methods.

Conclusions:

• The PCGA is an effective large-scale adapted method for geostatistical inversion, suitable for numerous parameters.
• Combining PCGA with hydraulic tomography offers a computationally efficient yet accurate approach for subsurface characterization.