Journal of Tropical Oceanography >
Application of convolutional neural network to sea surface temperature prediction in the coastal waters
Copy editor: SUN Cuici
Received date: 2023-03-23
Revised date: 2023-04-23
Online published: 2023-06-19
Supported by
Science and Technology Plan Projects of Guangdong Province(2021B1212050025)
Concerning the low sea surface temperature (SST) prediction accuracy of numerical modeling and empirical methods in near-shore stations, we consider sea surface temperature prediction as forecasting of multivariate time series data, construct the sea surface temperature time series model of near-shore stations by convolutional neural network (CNN) to predict the maximum, minimum and mean sea surface temperature for the next day, and compare CNN model with empirical forecast method and long short-term memory (LSTM) model through experiment. The experimental results show that compared with empirical forecast method, the mean absolute error (MSE) of CNN model on daily maximum SST forecast drops 0.14℃ to 0.36℃, root mean squared error (RMSE) drops 0.21℃ to 0.49℃, the MSE of CNN model on daily minimum SST forecast drops 0.17℃ to 0.36℃, RMSE drops 0.24℃ to 0.63℃, the MSE of CNN model on daily mean SST forecast is 0.30℃, RMSE is 0.47℃, its forecast performance is as good as LSTM model in the testing set. It shows that the application of CNN to SST modeling is feasible, improve the accuracy of sea surface temperature prediction which can compare favorably with LSTM model.
WENG Shaojia , CAI Jinhai , PANG Yunxi , LUO Rongzhen . Application of convolutional neural network to sea surface temperature prediction in the coastal waters[J]. Journal of Tropical Oceanography, 2024 , 43(1) : 40 -47 . DOI: 10.11978/2023037
图2 三种方法预报结果与实测日最高海温的比较Fig. 2 Comparison of predicted and observed values on the daily maximum SST |
图3 三种方法预报结果与实测日最低海温的比较Fig. 3 Comparison of predicted and observed values on the daily minimum SST |
表1 三种预报方法的SST预报效果Tab.1 Results of CNN forecast method, LSTM forecast method and empirical forecast method |
月份 | 日最高SST/ ℃ | 日最低SST/ ℃ | 日平均SST/ ℃ | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CNN预报 | LSTM预报 | 人工经验预报 | CNN预报 | LSTM预报 | 人工经验预报 | CNN预报 | LSTM预报 | |||||||||
MAE | RMSE | MAE | RMSE | MAE | RMSE | MAE | RMSE | MAE | RMSE | MAE | RMSE | MAE | RMSE | MAE | RMSE | |
1月 | 0.21 | 0.28 | 0.22 | 0.30 | 0.24 | 0.31 | 0.17 | 0.23 | 0.15 | 0.21 | 0.30 | 0.37 | 0.14 | 0.22 | 0.11 | 0.19 |
2月 | 0.31 | 0.51 | 0.30 | 0.50 | 0.36 | 0.59 | 0.19 | 0.26 | 0.17 | 0.25 | 0.29 | 0.37 | 0.18 | 0.24 | 0.15 | 0.21 |
3月 | 0.35 | 0.44 | 0.40 | 0.53 | 0.40 | 0.53 | 0.18 | 0.25 | 0.17 | 0.24 | 0.29 | 0.37 | 0.18 | 0.25 | 0.19 | 0.28 |
4月 | 0.35 | 0.45 | 0.33 | 0.43 | 0.40 | 0.57 | 0.27 | 0.38 | 0.26 | 0.36 | 0.31 | 0.46 | 0.26 | 0.37 | 0.27 | 0.38 |
5月 | 0.48 | 0.60 | 0.50 | 0.63 | 0.53 | 0.64 | 0.39 | 0.50 | 0.42 | 0.52 | 0.47 | 0.65 | 0.39 | 0.49 | 0.33 | 0.46 |
6月 | 0.54 | 0.67 | 0.49 | 0.62 | 0.91 | 1.15 | 0.79 | 1.13 | 0.74 | 1.15 | 1.28 | 1.78 | 0.54 | 0.74 | 0.54 | 0.71 |
7月 | 0.63 | 0.80 | 0.61 | 0.76 | 0.88 | 1.06 | 1.12 | 1.45 | 1.05 | 1.41 | 1.36 | 1.75 | 0.78 | 0.97 | 0.77 | 0.92 |
8月 | 0.41 | 0.52 | 0.41 | 0.51 | 0.68 | 0.83 | 0.45 | 0.72 | 0.44 | 0.71 | 0.62 | 0.85 | 0.32 | 0.41 | 0.34 | 0.44 |
9月 | 0.33 | 0.48 | 0.38 | 0.51 | 0.53 | 0.78 | 0.33 | 0.47 | 0.37 | 0.51 | 0.48 | 0.72 | 0.32 | 0.44 | 0.30 | 0.46 |
10月 | 0.22 | 0.29 | 0.21 | 0.26 | 0.46 | 0.56 | 0.08 | 0.13 | 0.10 | 0.15 | 0.20 | 0.30 | 0.12 | 0.16 | 0.18 | 0.22 |
11月 | 0.24 | 0.30 | 0.26 | 0.31 | 0.34 | 0.45 | 0.14 | 0.19 | 0.14 | 0.19 | 0.35 | 0.56 | 0.15 | 0.20 | 0.17 | 0.24 |
12月 | 0.19 | 0.23 | 0.20 | 0.26 | 0.30 | 0.40 | 0.21 | 0.29 | 0.24 | 0.33 | 0.38 | 0.46 | 0.17 | 0.25 | 0.18 | 0.24 |
全年 | 0.36 | 0.49 | 0.36 | 0.49 | 0.50 | 0.70 | 0.36 | 0.63 | 0.36 | 0.63 | 0.53 | 0.87 | 0.30 | 0.47 | 0.30 | 0.45 |
[1] |
匡晓迪, 王兆毅, 张苗茵, 等, 2016. 基于BP神经网络方法的近岸数值海温预报释用技术[J]. 海洋与湖沼, 47(6): 1107-1115.
|
[2] |
李燕, 张建华, 刘钦政, 等, 2007. 单站海温短期预报自动化[J]. 海洋预报, 2007, 24(4): 33-41.
|
[3] |
王兆毅, 李云, 王旭, 2020. 中国近岸海域基础预报单元海温预报指导产品研制[J]. 海洋预报, 37(4): 59-65.
|
[4] |
许金电, 蔡尚湛, 宣莉莉, 等, 2014. 粤东至闽南沿岸海域夏季上升流的调查研究[J]. 热带海洋学报, 33(2): 1-9.
|
[5] |
张建华, 2003. 海温预报知识讲座第一讲: 海水温度预报概况[J]. 海洋预报, 20(4): 81-85 (in Chinese).
|
[6] |
张建华, 2004. 海温预报知识讲座第二讲: 数理统计方法在海温预报中的应用[J]. 海洋预报, 21(1): 85-90.
|
[7] |
张雪薇, 韩震, 2022. 基于ConvGRU深度学习网络模型的海表面温度预测[J]. 大连海洋大学学报, 37(3): 531-538.
|
[8] |
张云翼, 江毓武, 2012. 汕尾外侧冷水跨陆架输送的形成机制[J]. 厦门大学学报(自然科学版), 51(4): 746-752.
|
[9] |
郑泽宇, 梁博文, 2017. TensorFlow: 实战Google深度学习框架[M]. 北京: 电子工业出版社: 6-9 (in Chinese).
|
[10] |
朱贵重, 胡松, 2019. 基于LSTM-RNN的海水表面温度模型研究[J]. 应用海洋学学报, 38(2): 191-197.
|
[11] |
庄桦, 2014. 沿岸海域海温研究进展[J]. 中国水运, 14(2): 194-195 (in Chinese).
|
[12] |
|
[13] |
|
[14] |
|
[15] |
|
[16] |
|
[17] |
|
[18] |
|
[19] |
|
[20] |
|
/
〈 | 〉 |