Reducing complexity in pore-scale reactive transport modeling

Pore-scale reactive transport simulations resolve biogeochemical processes within the complex geometry of porous media but are computationally expensive, which often leads researchers to use two-dimensional (2D) domains as proxies for three-dimensional (3D) systems. Using CompLaB3D, an open-source 3D lattice Boltzmann solver, we systematically compare 2D and 3D simulations across Fibrous, Gaussian, and Overlapping Spheres media at varying porosities to identify when 2D representations adequately capture 3D transport behavior and when they fail. Building on this, we develop machine learning surrogate models to reduce the computational cost of pore-scale reactive transport modeling while preserving accuracy, supporting broader and more accessible community use.