Overview
This corrigendum addresses a publication error in the article "PGTransNet: a physics-guided transformer network for 3D ocean temperature and salinity predicting in tropical Pacific." The correction ensures the mathematical integrity of the methodology section by removing redundant content that occurred during the final publication process.
Problem Addressed
The original research presents PGTransNet, a physics-guided transformer network designed to predict 3D ocean temperature and salinity in the tropical Pacific. Predicting these variables is a complex task because ocean dynamics are governed by intricate physical laws that standard machine learning models often fail to capture. To address this, the authors integrated physical constraints directly into the transformer architecture.
In the published version of the manuscript, a specific paragraph and its associated equation were accidentally repeated within the "Methodology" section. This duplication led to an inconsistency in the logical flow and the numbering of subsequent equations.
Correction Details
The correction involves the removal of the duplicated text and the adjustment of equation numbering to maintain consistency.
Location of Error: Section: Methodology, Subsection: Physics-Guided Information Integrating, Page 06, Paragraphs 4 and 5.
Corrective Actions: - The redundant paragraph and the repeated Equation 5 have been deleted. - Following the deletion, the subsequent equations (previously numbered 6, 7, and 8) have been renumbered to 5, 6, and 7.
Methodological Significance
The corrected section describes the application of the TEOS-10 (Thermodynamic Equation of Seawater - 2010) standard. This standard is used to determine seawater density ($\rho$) via the Gibbs function ($g$):
$$\rho = \rho(S_A, T, P) = (g_P)^{-1} = (\partial g / \partial P|_{S_A, T})^{-1}$$
where $S_A$ represents absolute salinity, $T$ is in situ temperature, and $P$ is pressure. The authors utilize TEOS-10 because it offers broader applicability than the older EOS-80 standard, which is vital for accurate 3D ocean state prediction across varying depths and conditions. The model operates within defined physical bounds: $0 \leq S_A \leq 120\,g/kg$, $-12^\circ\text{C} \leq T \leq 80^\circ\text{C}$, and $1\,Pa \leq P \leq 100\,MPa$.
Conclusion
The authors emphasize that these corrections are administrative and formatting-related. The removal of the redundant content and the renumbering of equations do not alter the scientific findings, experimental results, or the overall performance of the PGTransNet model. The integrity of the physics-informed approach remains central to the study's methodology.
Summary updated