南海上层海洋温度垂向结构的反演

Inversion of Ocean Temperature Structure in the South China Sea

摘要:

南海上层海洋温度垂向结构对海洋气候研究及海洋防灾减灾具有重要意义，然而由于现场观测数据有限，很难获取高时空分辨率的网格化数据。基于 2007–2021 年的 Argo 剖面数据、海面高度异常数据和月平均气候态数据，评估了两层动力模型和多层回归模型在南海海区反演海洋温度结构的性能。两层动力模型反演得到的 26 ℃(D26)与 20 ℃(D20)等温线深度的均方根误差分别 13.25 m 和 21.12 m，多层回归模型的 D26、D20 均方根误差分别 11.55 m 和 14.32 m。通过对比 2 种模型的结果：多层回归模型在时间与空间上反演结果性能更佳。2 种模型反演的南海上层海洋热含量空间分布较为一致，均能应用于台风“威马逊”的强度评估；然而，在南海特殊的强内潮的背景下，2 种模型得到的 D20 性能都有所降低。

Abstract:

The vertical structure of ocean temperature in the South China Sea is important for climate research and marine disaster prevention and mitigation. Due to the limited in-situ observation data，it is difficult to obtain the vertical temperature structure in the South China Sea with high spatiotemporal resolution. This study evaluates the performance of the two-layer dynamic model and the multi-layer regression model in ocean temperature structure inversion based on the Argo data，the sea level anomaly(SLA)data and the World Ocean Atlas 2018(WOA18) data from 2007 to 2021. After comparing the two models，we find the Root Mean Square Error(RMSE)of D26 and D20 of the two-layer dynamic model are 13.25 m and 21.12 m，while RMSE of D26 and D20 of the multi-layer regression model are 11.55 m and 14.32 m. Overall，the multi-layer regression model has smaller error and better performance in time and space than the two-layer model. However，further analysis indicates that the spatial distribution of tropical cyclone heat potential in the South China Sea inverted by the two models is relatively consistent，and both of them can react to the intensity assessment of Typhoon Rammasun(2014). Additionally， under the influence of the special strong internal tides in the South China Sea，the performance of D20 obtained by both models is reduced.