Geometry-Driven Prediction of Microplastic Transport in Saturated Sediments: Fast and Memory-Efficient Pore-Scale Modeling

Microplastics are now widely detected across sediments systems, from soil and riverbeds to estuarine deposits and groundwater-connected aquifers. In soils and saturated sediments, accumulated particles can change pore structure and water-flow pathways, modify aggregation, and persist for decades because of low degradability. Field observations reveal that soils can act as long-term reservoirs that are intermittently remobilized by infiltration events and changing hydraulic gradients. Besides, microplastics can sorb and co-transport hydrophobic organic contaminants, pesticides, and trace metals, and can also leach polymer additives. These risks motivate a hydrological question central to contaminant fate in soils and sediments: how fast and how far can microplastic particles migrate through realistic pore geometries under saturated flow? Answering it requires models that connect pore structure to transport behaviours such as breakthrough curves over representative sediment volumes, while remaining computationally feasible for heterogeneous natural media.

We develop a fast, memory-efficient, purely geometry-driven pore-network framework that predicts microplastic transport in water-saturated sediments using only reconstructed pore-space geometry, without solving the Stokes field or performing particle-resolved tracking. At the pore scale, we derive a flux-weighted transit-time distribution for one-in-one-out pore cells and obtain a near-universal decay close to t^-3. For pores with multiple inlets and outlets, we partition each pore into one-in–one-out subdomains using an optimal-transport allocation that minimizes viscous energy dissipation, yielding physically consistent weights and a mechanistic pore-scale transit-time model. We then propagate these statistics through the network via flux-weighted random walks and compute macroscopic breakthrough curves by convolving the inlet signal with the predicted transit-time distribution.

Benchmarks against direct NS equations simulation of lattice Boltzmann method and immersed boundary methods on identical micro-CT geometries and against microplstic trnasport through quartz-sediment column experiments show that the model captures arrival times, tailing, and non-Fickian spreading, while reducing runtime and memory demands by orders of magnitude compared with direct simulations. By requiring only geometry, the approach scales to representative sediment volumes relevant to hyporheic zones and shallow aquifers, providing a practical tool to predict microplastic migration and associated contaminant risks in soil-sediment environments.

Fig. Stokes flow through a quartz column (left) and a comparison of breakthrough curves of direct simulation and our PNM method (right).