Title

Joint Inversion of Seismic Time Series and Permeability via Cross-Gradient Constraints for Enhanced Reservoir Fracture and Cavity Characterization

Abstract

Introduction and Problem StatementAccurately mapping fractures and localized cavities within heterogeneous, naturally fractured reservoirs is a foundational challenge for optimized fluid flow predictions, 
hydrocarbon recovery, and drilling risk mitigation. Conventional workflows typically invert seismic datasets and reservoir engineering data in isolated, sequential stages. 
This disconnected approach frequently generates structural inconsistencies, where the sharp boundaries resolved by geophysical imaging do not align with the transport property distributions derived from reservoir simulation. 
Because independent inversions fail to capture the shared structural frameworks inherent to the same subsurface volume, they result in smeared geological interfaces and poorly defined fracture networks. 
Consequently, small-scale structural features like cavities and macro-fractures remain obscured, compromising the predictive accuracy of the final static and dynamic reservoir models.

Methodology and Numerical Framework

To overcome these limitations, we introduce a novel, fully integrated Cross-Gradient Joint Inversion framework that enforces structural compliance between two highly disparate datasets: 
seismic time series (conventional seismic traces) and reservoir permeability. The core architecture of this joint inversion consists of two distinct numerical modeling engines operating 
simultaneously under a shared regularization constraint. First, a Least-Squares Reverse Time Migration (LSRTM) engine is deployed to handle the wavefield propagation. 
Crucially, this LSRTM formulation is specifically adapted to handle high-frequency sonic data, performing Full-Waveform Sonic (FWS) inversion to extract highly detailed near-borebore velocity 
variations and optimize subsurface reflectivity matrices at a fine structural scale. Second, a high-fidelity numerical reservoir simulator functions as the hydrodynamic modeling engine, evaluating fluid transport behavior to optimize the spatial distribution of reservoir permeability fields.
Rather than relying on rigid, empirical petrophysical regressions that often fail in complex facies, the coupling is achieved purely through a structural cross-gradient constraint embedded directly into the global objective function. The joint inversion mathematical engine solves a multi-objective 
optimization problem, concurrently minimizing the L2-norm data misfits of the seismic waveform residuals, the permeability production data residuals, and the cross-gradient function. The cross-gradient term computes the vector product of the spatial gradients of reflectivity and permeability. 
By driving this cross-product toward zero, the optimization framework forces the spatial gradients of both properties to become parallel or anti-parallel, effectively aligning their structural boundaries without enforcing a direct linear correlation between their absolute magnitudes.

Results, Validation, and Discussion

The developed joint inversion methodology was validated through rigorous numerical testing, demonstrating excellent convergence behavior for both the FWS-capable LSRTM and reservoir simulator components. By minimizing the cross-gradient constraint, the joint optimization successfully synchronizes the distinct structural boundaries of the high-resolution reflectivity profiles and permeability fields into tight spatial compliance. The joint model yields exceptionally well-defined geological interfaces and continuous, physically consistent rock property distributions across the reservoir volume. Most notably, subtle structural discontinuities, localized secondary porosity cavities, and complex, sub-seismic fracture networks become sharply resolved near and away from the wellbore. These features are frequently lost, smeared, or mathematically isolated when using conventional, uncoupled inversion techniques.

Conclusion and Significance

This innovative approach successfully bridges the long-standing operational gap between geophysical imaging and reservoir engineering. By embedding the cross-gradient joint inversion framework directly into the interpretation pipeline, asset teams can obtain highly accurate, structurally validated models of fractured reservoirs. The output models provide a mathematically rigorous tool for predicting directional fluid flow pathways, identifying sweet spots, and preventing hazardous cavity intersections during active drilling operations, perfectly aligning with the next generation of physics-guided, data-driven subsurface workflows.
