Optimal initial errors related to the prediction of the vertical thermal structure and their application to targeted observation: A 3-day hindcast case study in the northern South China Sea

This 3-day hindcast case study focuses on the effects of the initial errors, a non-negligible factor in oceanic prediction, on VTS prediction in the NSCS and provides scientific guidance on observation design through tar-geted observation. Based on the Conditional Nonlinear Optimal Perturbation (CNOP) approach and the Regional Ocean Modeling System, the impact of optimal initial errors (OIEs) on the 3-day VTS prediction is estimated and their nonlinear evolution mechanism is clarified. The OIEs, horizontally distributed along the 1000 m isobath and vertically confined within the upper 50 m, experience non-localized evolution and cause a rapid increase in the root mean square errors by about 100 times during the prediction period. Eddy energetics analysis indicates that barotropic instabilities constitute a key factor in the evolution of the OIEs. In addition, sensitive areas are identified for targeted observation based on the vertically integrated total energy of the OIEs. Sensitive exper-iments demonstrate that the prediction errors caused by random initial errors in the sensitive areas are larger than those in other nearby areas. Observing system simulation experiments (OSSEs) further suggest that elimi-nating the initial errors in sensitive areas can improve VTS prediction by over 50%. This study reveals the patterns and growth dynamic mechanism of CNOP-type OIEs in 3-day VTS prediction in the NSCS and provides scientific guidance for operational observation design.

Automated summary

This 3-day hindcast case study focuses on the effects of initial errors on VTS prediction in the NSCS and provides scientific guidance on observation design. The study estimates the impact of optimal initial errors (OIEs) on 3-day VTS prediction and clarifies their nonlinear evolution mechanism. Sensitive areas are identified for targeted observation, and eliminating initial errors in these areas can significantly improve VTS prediction.