Chasing hotspots: Assessing usability of reduced-order submesoscale-permitting models to improve surface plastic clustering predictions in the Great Pacific Garbage Patch

Abstract

Predicting where floating plastic concentrates within the Great Pacific Garbage Patch is difficult because operational circulation models resolve the broad accumulation region, but not necessarily the smaller flow structures that form local hotspots. Reduced-order two-dimensional models offer an attractive way to reach finer horizontal resolution at lower computational cost, but their usefulness depends on whether they correctly retain the near-surface dynamics that control particle clustering. This thesis assesses under which dynamical conditions, and to what extent, a high-resolution 2D MITgcm framework can reproduce the surface-flow and surface-particle clustering statistics of matched 3D MITgcm benchmark simulations in a semi-idealized GPGP setting. The comparison combines Eulerian diagnostics of vorticity, strain, divergence, spectral content and submesoscale motion with Lagrangian particle-clustering metrics based on Voronoï tessellations. The results show that the 2D model can retain part of the broad mesoscale organization and surface-vorticity structure, especially under weaker energetic conditions. However, it does not reliably reproduce the connected high-strain structures, high-strain-high-vorticity asymmetry and compressive surface organization present in the 3D benchmarks. These differences limit its ability to reproduce finite-time particle clustering and local concentration extremes. The unaugmented 2D framework is therefore more defensible for qualitative broad-transport characterisation than for quantitative hotspot prediction. Future hotspot-oriented use should focus on controlled augmentation of the particle-advection kernel, using proxies for persistent filamentary attraction and unresolved surface convergence.