Dynamic Mode Decomposition of 4D imaging data to explore intermittent fluid connectivity in subsurface flows

The interaction of multiple fluids within a heterogeneous pore space gives rise to complex pore-scale flow dynamics, such as intermittent pathway flow. Synchrotron imaging has been employed to capture and analyze these dynamics. However, these imaging datasets are often extremely large (on the order of terabytes), and the spatial and temporal characteristics of the relevant flow phenomena are difficult to extract. As a result, identifying the locations of fluctuations that control fluid connectivity remains a significant challenge. In this work, a novel workflow is presented that uses Dynamic Mode Decomposition (DMD) to find critical spatio-temporal regions exhibiting intermittent flow dynamics. DMD is a data-driven algorithm that decomposes complex nonlinear systems into dominant spatio-temporal structures without relying on prior system assumptions.

The workflow is validated through three test cases, each examining the influence of viscosity ratio on flow dynamics while maintaining a constant capillary number. These scenarios demonstrate the capability of the DMD method to accurately capture underlying flow behavior and extract key intermittent structures from high-dimensional experimental data. DMD offers a powerful and computationally efficient approach for analyzing complex fluid dynamics in heterogeneous pore spaces. The proposed workflow enables rapid and objective identification of relevant time scales and spatial regions of interest. Given its speed and scalability, it holds strong potential as a diagnostic tool for the analysis of large synchrotron imaging datasets.