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.

Deep Learning Approach for Downscaling the Significant Wave Height Based on CBAM_CGAN

A Depth-Adaptive Waveform Decomposition Method For Airborne Lidar Bathymetry

GSDNet: A deep learning model for downscaling the significant wave height based on NAFNet

Deep learning for statistical downscaling of sea states

Research on High-Resolution Reconstruction of Marine Environmental Parameters Using Deep Learning Model

A deep learning super-resolution model to speed up computations of coastal sea states

Statistical Downscaling of Coastal Directional Wave Spectra Using Deep Learning

Wave Height and Period Estimation from X-Band Marine Radar Images Using Convolutional Neural Network

Data-driven rolling model for global wave height

Super-Resolution works for coastal simulations

Improving wave height prediction accuracy with deep learning

Advancing Spatio-temporal Storm Surge Prediction with Hierarchical Deep Neural Networks

Efficient downscaling of satellite oceanographic data with convolutional neural networks

A training strategy to improve the generalization capability of deep learning-based significant wave height prediction models in offshore China

High-Resolution Digital Bathymetric Model (DBM) With Uncertainty Reconstructed by Deep Learning With Dropout

Enhancing Significant Wave Height Retrieval with FY-3E GNSS-R Data: A Comparative Analysis of Deep Learning Models

Machine Learning‐Based Wave Model With High Spatial Resolution in Chesapeake Bay

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

Application of Deep Learning-based Interpolation Methods to Nearshore Bathymetry

Estimation of Significant Wave Heights from ASCAT Scatterometer Data via Deep Learning Network

A Deep Learning-Based Bias Correction Method for Predicting Ocean Surface Waves in the Northwest Pacific Ocean