| 基于融合U-Net及ConvLSTM的海面高度异常预报方法研究 |
| |
| 引用本文： | 周玮辰,韩震,张雪薇.基于融合U-Net及ConvLSTM的海面高度异常预报方法研究[J].海洋通报,2021,40(4). |
| |
| 作者姓名： | 周玮辰 韩震 张雪薇 |
| |
| 作者单位： | 上海海洋大学海洋科学学院,上海201306;上海海洋大学海洋科学学院,上海201306;上海河口海洋测绘工程技术研究中心,上海201306 |
| |
| 基金项目： | 电磁波信息科学教育部重点实验室开放基金 （EMW201909）；“全球变化与海气相互作用”专项资助 （GASI-02-PACIND-YGST03） |
| |
| 摘 要： | 海面高度异常是反映海洋环境状况的主要变量之一。本文使用1993—2019年的融合月均海面高度异常数据,建立了基于深度学习的海面高度异常预测神经网络模型,提出了基于融合U型网络(U-Net)和卷积长短记忆网络(ConvLSTM)的中长期海面高度异常预报模型。在研究海域0.25°×0.25°的空间分辨率下,模型测试集预报结果的均方根误差和平均绝对误差分别为0.039 m和0.027 m,均优于全连接LSTM预报模型和ConvLSTM+CNN预报模型,为大中尺度的海面高度异常预报提供了新的方法。
|
| 关 键 词： | 海面高度异常 卷积长短记忆 U-Net网络 深度学习 |
| 收稿时间： | 2020/12/9 0:00:00 |
| 修稿时间： | 2021/3/15 0:00:00 |
Research on sea level anomaly prediction based on U-Net structure and ConvLSTM layers |
| |
| ZHOU Weichen,HAN Zhen,ZHANG Xuewei.Research on sea level anomaly prediction based on U-Net structure and ConvLSTM layers[J].Marine Science Bulletin,2021,40(4). |
| |
| Authors: | ZHOU Weichen HAN Zhen ZHANG Xuewei |
| |
| Institution: | College of Marine Sciences, Shanghai Ocean University, Shanghai 201306, China;College of Marine Sciences, Shanghai Ocean University, Shanghai 201306, China;Shanghai Engineering Research Center of Estuarine and Oceanographic Mapping, Shanghai 201306, China |
| |
| Abstract: | The sea level anomaly is an important variable that reflects the state of the ocean. In this paper, the multi-satellite observation sea level anomaly data from 1993 to 2019 were used to build deep neural networks. A sea level anomaly prediction method based on U-Net structure and ConvLSTM layers was proposed. Under the spatial resolution of 0.25?~0.25?in the studied area, the RMSE and MAE of the test set were 0.039 m and 0.027 m, respectively, which were superior to the fully connected LSTM prediction model and ConvLSTM+CNN prediction model. This model provides a new way for the largescale sea level anomaly forecast. |
| |
| Keywords: | sea level anomaly ConvLSTM U-Net deep learning |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《海洋通报》浏览原始摘要信息 |
| 点击此处可从《海洋通报》下载免费的PDF全文 |
|
