Eddy-induced underwater acoustic field reconstruction and computation based on sound speed classification and B-spline surface fitting

In polar regions, the unique conditions created by sea ice coverage pose challenges for both remote sensing and in-situ observation methods. As a result, underwater acoustic detection has emerged as an effective approach for observing complex oceanic physical phenomena in these environments. Focusing on anomalies in local seawater acoustic properties caused by eddies, we propose a method for reconstructing the three-dimensional structure of eddies. An ice-edge eddy observed during the ACOBAR project serves as a case study to demonstrate the implementation of this approach. By combining an unsupervised learning strategy with B-spline surface fitting, the proposed approach reconstructs the eddy structure without relying on highly idealized axisymmetric assumptions. Using a sound speed anomaly threshold of -3 m/s to define the eddy boundaries, the reconstruction achieves an accuracy of 74%. To further assess the method’s effectiveness, the reconstructed eddy is used to simulate the eddy-induced underwater acoustic field through finite element method (FEM) modeling. The results show that this approach reduces computational time and resource consumption by more than 30%, while maintaining a mean transmission loss error of only 1.2 dB over a 20 km range. This work represents an effective integration of acoustic sensing, machine learning, and FEM simulation in oceanographic research, offering a practical and efficient solution for studying subsurface phenomena in ice-covered regions.