Abstract
Understanding subsurface temperature-transition structures is essential for interpreting upper-ocean stratification; however, most existing methods rely on two-dimensional profiles and fail to resolve the full three-dimensional geometry of temperature anomalies. This study proposes the Three-Dimensional Ocean Temperature Structure Extraction method (3D-OTSE), a flexible data-driven framework that identifies coherent three-dimensional thermal-transition features directly from multi-depth ocean temperature fields. The method defines a Temperature-Contrast Index (TCI) based on local three-dimensional temperature differences, determines an adaptive threshold from the curvature of the TCI distribution, and employs 3D DBSCAN to extract volumetric structures. Rather than assuming a thermocline, 3D-OTSE detects a wide range of vertical temperature anomalies—including thermoclines, inverse thermoclines, and multilayer transitions—according to their spatial organization in the data. Applying this method to the South China Sea Basin (SCS) can reconstruct thermocline-like structures that conform to large-scale regional patterns and can also capture complex lateral variations that are difficult to detect by traditional profile diagnosis methods. The region-adaptive threshold enables this framework to adapt to inhomogeneous formation states and spatio-temporal scales. In general, 3D-OTSE provides a universal, parameter-adaptive tool for finding three-dimensional underground temperature anomaly layers, supplements perspectives for traditional methods, and lays the foundation for future multivariate and time-varying applications.