1. Combining ERA5 data and CYGNSS observations for the joint retrieval of global significant wave height of ocean swell and wind wave: a deep convolutional neural network approach.
- Author
-
Bu, Jinwei, Yu, Kegen, Ni, Jun, and Huang, Weimin
- Abstract
As an emerging remote sensing technology, GNSS reflectometry (GNSS-R) has been widely investigated for retrieving ocean parameters including ocean significant wave height (SWH). Ocean SWH consists of contribution from swell and wind waves, which are commonly modeled separately in the field of marine science and engineering to facilitate practical application. In this study, we present a deep convolutional neural network (DCNN) model for retrieving swell and wind wave SWHs. The DCNN model makes use of auxiliary data and effective DDM features extracted in the convolution layer, and it is trained by using the ERA5 data and CYGNSS observations. The proposed DCNN model and seven existing models [i.e., random forest, extremely randomized trees, bagging tree (BT), decision tree, support vector machine (SVM), artificial neural network, and convolutional neural network] were extensively tested using the ERA5 and WaveWatch III (WW3) data. The results show that when ERA5 is used as reference data, the proposed DCNN model performs best among the eight models, with the root mean square errors (RMSEs) of retrieving swell and wind wave SWH being better than 0.394 m and 0.397 m, respectively, and the correlation coefficient (R) being 0.90. Compared with the SVM model, RMSEs are improved by 28.82% and 31.92%, respectively. When WW3 is employed as reference, the RMSEs of retrieving swell and wind wave SWH are better than 0.497 m and 0.502 m, respectively, with R of 0.89 and 0.90. Compared with the BT model, RMSEs are improved by 26.74% and 27.41%, respectively. The research also found that the auxiliary variables are important for swell and wind wave SWH retrieval. Furthermore, the retrieval of SWH for swells and wind waves using spaceborne GNSS-R technology is affected by rainfall, resulting in about 6% increase in RMSE. This method provides a new idea for studying global ocean swell and wind waves using CYGNSS data. [ABSTRACT FROM AUTHOR]
- Published
- 2023
- Full Text
- View/download PDF