| 摘要: |
| 海表面温度(sea surface temperature, SST)是影响海洋和气候变化的重要因素之一,准确预测SST的变化对于海洋生态环境、气象和航行等至关重要。传统的SST预测方法通常依赖于数值模式,但是其计算成本较高。该文基于深度学习模型(3D U-Net),将SST、海表面高度异常(sea surface height anomalies, SSHA)以及海表面风(sea surface wind, SSW)作为输入变量成功构建了南海SST的快速化智能预报模型。 结果表明,与卷积长短时记忆(convolutional long short-term memory, ConvLSTM)模型相比,3D U-Net模型在所有预测时间中均显示出更高的准确度,其均方根误差(RMSE)为0.53℃,皮尔逊相关系数(R)达到0.96。在不同季节和南海不同区域,3D U-Net模型均表现出较小的预测误差,而且在季风盛行期间也具有较强的鲁棒性。此外,3D U-Net模型在预测2021年南海的海洋热浪(marine heatwave, MHW)事件时,大部分海域的准确率达到了80%以上。敏感性实验结果表明,SSHA和SSW对模型的预测性能有显著影响,并在不同的预报阶段中发挥着不同的作用。综上所述,结合多源海表数据的3D U-Net模型能够快速准确地预测出南海SST,并为预测MHW事件提供了新方法。 |
| 关键词: 海表面温度 3D U-Net模型 深度学习 南海 海洋热浪 |
| DOI： |
| 分类号: |
| 基金项目:中央引导地方科技发展资金项目“山东海洋装备设计与数值模拟研发公共服务平台”（YDZX2022022）；山东省自然科学基金项目“台湾以东黑潮次表层高营养盐水的来源、变异及入侵东中国海的研究”（ZR2021MD022） |
|
| INTELLIGENT PREDICTION OF SEA SURFACE TEMPERATURE IN THE SOUTH CHINA SEA BASED ON DEEP LEARNING |
|
Xie Bo-Wen1, Zhang Cong2, Yang Shu-Guo1, Feng Zhong-Kun1, Sun Gui-Min3
|
|
1.School of Mathematics and Physics,Qingdao University of Science and Technology;2.Marine Science Research Institute of Shandong Province;3.CAS Key Laboratory of Ocean Circulation and Waves,Institute of Oceanology,Chinese Academy ofSciences
|
| Abstract: |
| Sea surface temperature (SST) is one of the crucial factors influencing oceanic and climatic change, and accurately predicting SST variations is vital for marine ecological environments, meteorology, and navigation. Traditional SST prediction methods typically rely on numerical models, which have high computational costs. This paper presents a rapid and intelligent forecasting model for SST in the South China Sea (SCS) based on a deep learning model (3D U-Net), using SST, sea surface height anomalies (SSHA), and sea surface wind (SSW) as input variables. The results indicate that compared to the convolutional long short-term memory (ConvLSTM) model, the 3D U-Net model shows higher accuracy across all prediction times, with a root mean square error (RMSE) of 0.53°C and a Pearson correlation coefficient (R) of 0.96. In various seasons and regions of the SCS, the 3D U-Net model consistently exhibits smaller prediction errors and maintains robust performance during monsoon seasons. Moreover, in predicting marine heatwave (MHW) events in the SCS in 2021, the model achieved over 80% accuracy in most sea areas. Sensitivity experiments reveal that SSHA and SSW significantly influence the model"s predictive performance, playing different roles at various forecasting stages. In summary, the 3D U-Net model, combined with multi-source sea surface data, can predict SST in the SCS quickly and with relative accuracy, offering a new method for predicting MHW events. |
| Key words: sea surface temperature 3D U-Net model deep learning South China Sea marine heatwave |