Abstract: Submesoscale oceanic processes exhibit complex dynamics and three-dimensional structures, yet their direct observation remains challenging due to limitations in existing observational techniques, which hinders further understanding of submesoscale dynamics. In this study, high-resolution towed observations collected across an oceanic eddy-front system, together with numerical model data, are used to investigate the feasibility of reconstructing the three-dimensional structure of frontal submesoscale processes using a three-dimensional variational data assimilation method (3D-Var) and optimal interpolation (OI). Based on model-simulated datasets, the performance of the two methods in reproducing frontal structures and temperature gradients is systematically evaluated. The results indicate that the 3D-Var method exhibits clear advantages in regions characterized by strong temperature gradients, yielding more accurate reconstructions with significantly lower root-mean-square errors than those obtained by OI. In contrast, the OI method performs relatively better in stably stratified layers but shows limitations in representing complex nonlinear structures. Furthermore, reconstructions based on shipborne towed observations demonstrate that the 3D-Var approach is capable of recovering the three-dimensional structure of submesoscale frontal processes from limited sectional observations. These results suggest that three-dimensional variational assimilation provides an effective framework for investigating oceanic submesoscale processes using high-resolution towed measurements.

Key words: Submesoscale processes, Ocean fronts, Three-dimensional reconstruction