Satellite Ocean Research Articles

An Ensemble Machine Learning Approach for Sea Ice Monitoring Using CFOSAT/SCAT Data

Sea ice is a crucial component of the global climate system. The China–French Ocean Satellite Scatterometer (CFOSAT/SCAT, CSCAT) employs an innovative rotating fan beam system. This study applied principal component analysis (PCA) to extract classification features and developed an ensemble machine learning approach for sea ice detection. PCA identified key features from CSCAT’s backscatter information, representing outer and sweet swath observations. The ensemble model’s performances (OA and Kappa) for the Northern and Southern Hemispheres were 0.930, 0.899, and 0.844, 0.747, respectively. CSCAT achieved an accuracy of over 0.9 for close ice and open water but less than 0.3 for open ice, with misclassification of open ice as closed ice. The sea ice extent discrepancy between CSCAT and the National Snow and Ice Data Center (NSIDC) was −0.06 ± 0.36 million km2 in the Northern Hemisphere and −0.03 ± 0.48 million km2 in the Southern Hemisphere. CSCAT’s sea ice closely matched synthetic aperture radar (SAR) imagery, indicating effective sea ice and open water differentiation. CSCAT accurately distinguished sea ice from open water but struggled with open ice classification, with misclassifications in the Arctic’s Greenland Sea and Hudson Bay, and the Antarctic’s sea ice–water boundary.

Enhancing the Resolution of Satellite Ocean Data Using Discretized Satellite Gridding Neural Networks

Ocean satellite data are often impeded by intrinsic limitations in resolution and accuracy. However, conventional data reconstruction approaches encounter substantial challenges when facing the nonlinear oceanic system and high-resolution fusion of variables. This research presents a Discrete Satellite Gridding Neural Network (DSGNN), a new machine learning method that processes satellite data within a discrete grid framework. By transforming the positional information of grid elements into a standardized vector format, the DSGNN significantly elevates the accuracy and resolution of data fusion through a neural network model. This method’s innovative aspect lies in its discretization and fusion technique, which not only enhances the spatial resolution of oceanic data but also, through the integration of multi-element datasets, better reflects the true physical state of the ocean. A comprehensive analysis of the reconstructed datasets indicates the DSGNN’s consistency and reliability across different seasons and oceanic regions, especially in its adept handling of complex nonlinear interactions and small-scale oceanic features. The DSGNN method has demonstrated exceptional competence in reconstructing global ocean datasets, maintaining small error variance, and achieving high congruence with in situ observations, which is almost equivalent to 1/12° hybrid coordinate ocean model (HYCOM) data. This study offers a novel and potent strategy for the high-resolution reconstruction and fusion of ocean satellite datasets.

Study on the Typical Environmental Factors in the Middle Part of Zhoushan Fishery Based on HY-1C/D and Other Multi-Source Data

This study utilizes satellite data, including HY-1C/D, along with reanalysis data, to unveil the typical environmental characteristics of the sea surface in the middle of Zhoushan fishery. The article addresses three main issues. The first one is the development of an ocean primary productivity (OPP) inversion algorithm model. The second one is the study of chlorophyll-α (Chl-α) concentration and OPP distribution characteristics in Zhoushan fishery using China’s domestically produced ocean satellite HY-1C/D CZI data. The last one is the revelation of the characteristics of typical environmental factors on the sea surface at Zhoushan fishery by combining HY-1C/D with multi-source data. The results show the following: (1) The middle part of Zhoushan fishery exhibits significant seasonal and regional variations in Chl-α concentration and OPP. Chl-α concentration ranges mainly between 0.2 and 2.9 µg/L, with higher concentrations in spring and summer and lower Chl-α concentrations in autumn and winter. Spatially, Chl-α concentration gradually decreases from west to east. The OPP in the study area ranges from 100 mg·m−2d−1 to 1000 mg·m−2d−1, with high OPP values distributed on the western side, ranging from 400 mg·m−2d−1 to 1000 mg·m−2d−1, and gradually decreasing seaward. The highest OPP occurs in summer and the lowest in winter. (2) The correlation analysis between Chl-α concentration and OPP revealed a strong positive relationship. Consequently, this study developed an empirical model for estimating OPP based on Chl-α concentration and validated its feasibility. The model applies to areas with Chl-α concentrations ranging from 0.2 to 4 µg/L. (3) The convergence of freshwater injection, multiple ocean currents, and seasonal upwelling in the study area brings about a rich supply of nutrients. Additionally, the region is characterized by suitable conditions, including optimal Chl-α concentrations, OPP, SST, salinity, currents, and geological water depths. The synergistic effect of these factors together contributed to the formation of Zhoushan fishery.

Reconstruction of Wide Swath Significant Wave Height From Quasi‐Synchronous Observations of Multisource Satellite Sensors

AbstractGlobal sea wave monitoring is of utmost importance for tasks such as analysis of ocean climate change, offshore fisheries, and early warning of marine disasters. Significant wave height (SWH) is one of the most vital and widely used metrics for measuring sea waves in marine research. Hence, obtaining high precision and extensive coverage measurements of SWH is of great significance for comprehensive sea wave studies. A data set is constructed by combining wave spectrometer and scatterometer data from China‐French Ocean Satellite, synthetic aperture radar (SAR) wave mode data from Sentinel‐1, and altimeter data from Jason‐3 and HY‐2B through space‐time matching method. The multi‐sources integrated data set is used to complete the reconstruction of the wide swath SWH. A model based on stacked autoencoder and deep neural network (SAE‐DNN) is developed. The SWH reconstructed by the model is evaluated with the training‐independent test set. The results demonstrate that the accuracy of the SAE‐DNN model is significantly improved by incorporating SAR joint quasi‐synchronous observations and the root mean square error can reach 0.217 m, which is comparable to the SWH measured by altimeter and highlights the effectiveness and reliability of the model in accurately reconstructing SWH. We further examine and analysis the distance variations between SWH reconstruction sites and Surface Wave Investigation and Monitoring observations, the influence of different SAR features, and the influence of sea state, highlighting the benefits of incorporating SAR data into SAE‐DNN model.

A Systematic Review of the Application of the Geostationary Ocean Color Imager to the Water Quality Monitoring of Inland and Coastal Waters

A Systematic Review of the Application of the Geostationary Ocean Color Imager to the Water Quality Monitoring of Inland and Coastal Waters

Global Sea Surface Temperature Analysis Based on Domestic Ocean Satellite Data. Part II: Comparative Analysis of Multiple Sets of Global Sea Surface Temperature Products in 2022

With more and more global gap-free fusion products of the sea surface temperature, understanding the consistency and discrepancy of the different SST fusion products will not only help data providers to improve their algorithms, but also help them to select the one that may better suit their applications. In this article, we have compared and analysed 10 sets of global gap-free fusion products of sea surface temperature in 2022, with different fusion techniques and related configurations. It is found that each SST analysis product has the same spatial distribution, with the minimum NMEFC mean value (20.11) and the maximum MGDSST mean value (20.31). Compared with ARGO in-situ data, the RMSE ranges are from 0.3233 (OSTIA) to 0.5180 (MGDSST). The RMSE between NMEFC fusion products and Argo in-situ data is 0.3861, ranked fifth out of 10 sets of fusion products. Compared with GMPE analysis among 9 sets of fusion products, the RMSE ranges are from 0.1579 (CMC) to 0.3199 (K10), and the NMEFC fusion product has a RMSE of 0.3040, which is at the intermediate level, ranked sixth out of 9 fusion products.

Sea Ice Extent Retrieval Using CSCAT 12.5 km Sampling Data

Polar sea ice extent exhibits a highly dynamic nature. This paper investigates the sea ice extent retrieval on a fine (6.25 km) grid based on the 12.5 km sampling data from the China France Ocean Satellite Scatterometer (CSCAT), which is generated by an adapted Bayesian sea ice detection algorithm. The CSCAT 12.5 km sampling data are analyzed, a corresponding sea ice GMF model is established, and the important calculation procedures and parameter settings of the adapted Bayesian algorithm for CSCAT 12.5 km sampling data are elaborated on. The evolution of the sea ice edge and extent based on CSCAT 12.5 km sampling data from 2020 to 2022 is introduced and quantitatively compared with sea ice extent products of Advanced Microwave Scanning Radiometer 2 (AMSR2) and the Advanced Scatterometer onboard MetOp-C (ASCAT-C). The results suggest the sea ice extent of CSCAT 12.5 km sampling data has good consistency with AMSR2 at 15% sea ice concentration. The sea ice edge accuracy between them is about 7 km and 10 km for the Arctic and Antarctic regions, and their sea ice extent difference is 0.25 million km2 in 2020 and 0.5 million km2 in 2021 and 2022. Compared to ASCAT-C 12.5 km sampling data, the sea ice edge Euclidean distance (ED) of CSCAT 12.5 km data is 14 km (2020 and 2021) and 12.5 km (2022) for the Arctic region and 14 km for the Antarctic region. The sea ice extent difference between them is small except for January to May 2020 and 2021 for the Arctic region. There are significant deviations in the sea ice extents of CSCAT 12.5 km and 25 km sampling data, and their sea ice extent difference is 0.3–1.0 million km2.

Networking Observation and Applications of Chinese Ocean Satellites

Networking Observation and Applications of Chinese Ocean Satellites

RESEARCH ON COOPERATION STRATEGY BASED ON SATELLITE REMOTE SENSING DATA SERVICE AND TECHNOLOGY APPLICATION BETWEEN CHINA AND ASEAN

Abstract. Remote sensing (RS) and earth observing technology are flourished with the development of a series of high-resolution earthobservation satellites. As the improvement of China’s earth observation data acquisition capability, one critical issue is put on the agenda that is what kind of models and techniques can promote the future data processing into a new level in terms of service model, massive data processing, development methods, business models, resource sharing, and system sustainability (1). In order to embed domestic satellite advantages into global world and provide increasingly more Chinese wisdom and solution to support the ASEAN regional development, China highlights the propositions on ASEAN-China remote sensing cooperation. On November 1, 2022, the ASEAN-China Satellite Remote Sensing Application Centre (hereinafter referred to as the ACSAC) was officially inaugurated in Beijing. The ACSAC will mainly focus on the establishment of a system and mechanism for a series of substantive operation, promoting the comprehensive sharing of China's land and ocean satellite data, comprehensively rich the international application scenarios of Chinese satellites, promoting applications in multiple fields and carrying out satellite remote sensing application promotion and typical demonstration, continuing to carry out technical exchanges and training with mutual demands and development. At present, ACSAC has stepped into the substantive construction stage. It will give full play to the advantages of China's natural resources land and ocean satellite remote sensing resources, to form a multi-scale, full coverage data resource support, jointly carry out the construction of the ASEAN-China satellite remote sensing data network platform based on the cloud environment, and to jointly build an integrated service open portal for land and ocean satellite data products to ASEAN region. Focusing on the core construction content of ACSAC, this paper systematically collects, summarizes and analyse the data and application needs of satellite remote sensing in ASEAN countries, finds out the current needs to meet the gap, and puts forward some thoughts and ideas on cooperation mechanism, data sharing, customized products and application demonstration of advanced products, aiming to lay a solid foundation for the substantial construction of ACSAC.

Improving blank ocean satellite data through machine learning: Case study and application in the Bohai Sea, China

Ocean satellites provide accurate and precise data across various scales, making them a vital tool for investigating the association between global change and ocean processes. However, low-quality data creates unavoidable gaps in satellite data, diminishing its usefulness and continuity. These deficiencies can be resolved by implementing machine learning techniques as valuable tools. This paper details a new satellite data prediction tool titled “SatelliteFixer”. The SatelliteFixer model, utilizing a custom-built random forest structure, can generate dependable data with enhanced temporal-spatial continuity. This model has demonstrated feasibility with diverse satellite data sources and light bands, and outperforms the basic machine learning approach. The juxtaposition of model data with in-situ cruise sampling results allows for widespread analysis of the movement and dispersion of suspended sediment. The above entails the inversion of long-term events and the observation of short-term events, which enables accurate seasonal analysis using continuous data without the influence of uneven data volume distribution and outliers, and is also the first-time satellite data has tracked the entire process of pulsed artificial flooding. SatelliteFixer provides a fresh outlook for detailing the varying trends on consecutive timescales and successional spaces among ocean processes.

Global ocean observations and applications by China’s ocean satellite constellation

Satellite remote sensing data form the basis of ocean observation and applications. China has established a satellite network platform comprising ocean color satellite constellations, ocean dynamic environment satellite constellations, and ocean observation and monitoring satellite constellations. This platform provides consistent and reliable ocean observation data crucial for marine scientific research, economic development, and early warning and forecasting. This paper comprehensively describes the development process and plans for China’s ocean satellites from their inception. It offers detailed technical specifications of ocean satellites and outlines the current applications of ocean water color satellites (HY-1), ocean dynamics and environment satellites (HY-2), and ocean surveillance and monitoring satellites (GF-3) in ocean parameter inversion, target identification and detection, and early warning and forecasting. In the future, to enhance the level of industrialization in ocean remote sensing in China, it is imperative to leverage the diversity and timeliness of ocean remote sensing data. Additionally, emerging technologies such as cloud computing and artificial intelligence should be harnessed, and the application potential of various satellite data resources should be explored.

In Situ Validation of Altimetry and CFOSAT SWIM Measurements in a High Wave Environment

Abstract While satellite altimeters have revolutionized ocean science, validation measurements in high wave environments are rare. Using geodetic Global Navigation Satellite System (GNSS) data collected from the Southern Ocean Flux Station (SOFS; −47°S, 142°E) since 2019, as part of the Southern Ocean Time Series (SOTS), we present a validation of satellite missions in this energetic region. Here we show that high rate GNSS observations at SOFS can successfully measure waves in the extreme conditions of the Southern Ocean and obtain robust measurements in all wave regimes [significant wave height (SWH) ranging from 1.5 to 12.6 m]. We find good agreement between the in situ and nadir altimetry SWH (RMSE = 0.16 m, mean bias = 0.04 m, and n = 60). Directional comparisons with the Chinese–French Ocean Satellite (CFOSAT) Surface Waves Investigation and Monitoring (SWIM) instrument also show good agreement, with dominant directions having an RMSE of 9.1° (n = 22), and correlation coefficients between the directional spectra ranging between 0.57 and 0.79. Initial sea level anomaly (SLA) estimates capture eddies propagating through the region. Comparisons show good agreement with daily gridded SLA products (RMSE = 0.03 m, and n = 205), with scope for future improvement. These results demonstrate the utility of high rate geodetic GNSS observations on moored surface platforms in highly energetic regions of the ocean. Such observations are important to maximize the geophysical interpretation from altimeter missions. In particular, the ability to provide collocated directional wave observations and SLA estimates will be useful for the validation of the recently launched Surface Water and Ocean Topography (SWOT) mission where understanding the interactions between sea state and sea surface height poses a major challenge.

Assimilation of AMSU-A Surface-Sensitive Channels in CMA_GFS 4D-Var System over Land

Abstract The assimilation of two surface-sensitive channels of the AMSU-A instruments on board the NOAA-15/NOAA-18/NOAA-19 and MetOp-A/MetOp-B satellites over land was achieved in the China Meteorological Administration Global Forecast System (CMA_GFS). The land surface emissivity was calculated by 1) the window channel retrieval method and 2) the Tool to Estimate Land Surface Emissivities at Microwave frequencies (TELSEM2). Quality controls for these satellite microwave observations over land were conducted. The predictors and regression coefficients used for oceanic satellite data were retained during the bias correction over land and found to perform well. Three batch experiments were implemented in CMA_GFS with 4D-Var: 1) assimilating only the default data, and adding the above data over land with land surface emissivity obtained from 2) TELSEM2 and 3) the window channel retrieval method. The results indicated that the window channel retrieval method can better reduce the departure between the observed and simulated brightness temperature. Over most land types, the positive impacts of this method exceed those of TELSEM2. Both TELSEM2 and the window channel retrieval method improve the humidity analysis near the ground, as well as the forecast capability globally, particularly in those regions where the land coverage is greater, such as in the Northern Hemisphere. The data utilization of the two surface-sensitive channels increase by 6% and 12%, respectively, and the additional data every 6 h can cover most land, where there was no surface-sensitive data assimilated before. This study marks the beginning of near-surface channel assimilation over land in CMA_GFS and represents a breakthrough in the assimilation of other surface-sensitive channels in other satellite instruments. Significance Statement Surface-sensitive microwave channels are difficult to assimilate in NWP due to the lack of both direct measurement and appropriate modeling for instantaneous land surface emissivity. This paper discusses a method that improves the surface emissivity estimates, which has allowed the utilization of surface-sensitive microwave channels in CMA_GFS. Those capabilities have resulted in better data utilization, improved forecasts of temperature, geopotential height, and winds in the Northern Hemisphere at 3–7 days, and represent an incremental and important improvement to CMA_GFS.

Wave Groups and Small Scale Variability of Wave Heights Observed by Altimeters

AbstractRecent satellite altimeter retracking and filtering methods have considerably reduced the noise level in estimates of the significant wave height (Hs), allowing to study processes with smaller spatial scales. In particular, previous studies have shown that wave‐current interactions may explain most of the variability of Hs at scales 20–100 km. As the spatial scale of the measurement is reduced, random fluctuations emerge that should be associated to wave groups. Here we quantify the magnitude of this effect, and the contribution of wave groups to the uncertainty in Hs measurements by altimeters, with a particular focus on extreme extra‐tropical storms. We take advantage of the low orbit altitude of the China‐France Ocean Satellite (CFOSAT), and the low noise level of the nadir beam of the SWIM instrument. Our estimate of wave group effects uses directional wave spectra measured by off‐nadir beams on SWIM, and signal processing theory that gives statistical properties of the wave envelope, and thus the local wave heights, from the shape of the wave spectrum. We find that the standard deviation of Hs associated to wave groups is a function of satellite altitude, wave height and spectral peakedness. For CFOSAT these fluctuations generally account for about 25% of the variance measured over a 80 km distance. This fraction is largest in storms and in the presence of long swells. When the estimated effect of wave groups is subtracted from the variance of Hs measurements, the remaining variability is higher in regions of strong currents.

Drastic hydrodynamic changes in the western Bay of Bengal caused by tropical cyclone Nada

To understand the interaction between tropical cyclones (TCs) and the upper ocean in the Bay of Bengal (BoB), TC Nada in 2016 is studied by analyzing ocean reanalysis data, ocean satellite remote sensing data and Argo buoy data. The results show that Nada influenced the anticyclonic eddy (ACE) present at 83.8°E,9.8°N, which propagated westward and formed a new ACE east of Sri Lanka, leading to the EICC bifurcation. An Eastern Sri Lankan Jet (ESLJ) was produced. After the ACE disappeared, the boundary current velocity increased from 1.4 m·s−1 to 1.6 m·s−1. Meanwhile, the main axis of the boundary current is shifted westward. The barrier layer thickness (BLT) was thickened by 33 m at the ACE and thinned by 10 m at the cyclonic eddy (approximately 11°N, 82°E). The mixed layer depth (MLD) became shallower within both the CE and the ACE, which can explain why there were no large-scale phytoplankton blooms in the eddies. Additionally, a part of the sea boundary current intruded into the Palk Strait affected by Nada. The chlorophyll-a (Chl-a) concentration in the Palk Strait increases from 0.08 to 0.67 mg·m−3, causing a phytoplankton bloom. This study helps to estimate the ecological impact of TCs in the western BoB.

An Exploratory Verification Method for Validation of Sea Surface Radiance of HY-1C Satellite UVI Payload Based on SOA Algorithm

To support the application of ocean surface radiance data from the ultraviolet imager (UVI) payload of the HY-1C oceanographic satellite and to improve the quantification level of ocean observation technology, the authenticity check study of ocean surface radiance data from the UVI payload was conducted to provide a basis for the quantification application of data products. The UVI load makes up for the lack of detection capabilities of modern ocean remote sensing satellites in the ultraviolet band. The UVDRAMS (Ultra-Violet Dual-band RadiAnce Measurement System) was used to verify the surface radiance data collected at 16 stations in the study area and the pupil radiance data collected by the UVI payload to establish an effective radiative transfer model and to identify the model parameters using the seeker optimization algorithm (SOA). The study of the UVDRAMS measurement system based on the SOA algorithm and the validation of the sea surface radiance of the UVI payload of the HY-1C satellite shows that 97.2% of the incident pupil radiance of the UVI payload is contributed by the atmospheric reflected radiance, and only 2.8% is from the real radiation of the water surface, while the high signal-to-noise ratio of the UVI payload of the HY-1C ocean satellite can effectively distinguish the reflectance of the water body. The high signal-to-noise ratio of the UVI payload of the HY-1C ocean satellite can effectively distinguish the amount of standard deviation in the on-satellite radiation variation, which meets the observation requirements and provides a new way of thinking and technology for further quantitative research in the future.

Atmospheric correction under cloud edge effects for Geostationary Ocean Color Imager through deep learning

The Geostationary Ocean Color Imager (GOCI) is widely employed in tracking diurnal dynamics of oceanic conditions. However, the current atmospheric correction (AC) algorithms for GOCI often mask pixels around cloud edges to exclude pixels contaminated by cloud edge effects (CEEs; including stray light, cloud shadows, and cloud adjacent effects (AEs)), which results in massive data loss. In this paper, we propose a novel AC algorithm to correct these CEE-affected pixels based on deep learning (namely, DLACC) to achieve a comparable quality level to those pixels far away from cloud edges. We used the standard near-infrared iterative AC algorithm in SeaDAS (namely, NIR) to obtain Rayleigh-corrected reflectance (Rrc) and remote sensing reflectance (Rrs). Then, high-quality Rrs values were extracted using a quality assurance (QA) algorithm. Next, we obtained 2,821,668 matchups by matching CEE-free noontime Rrs values with CEE-affected Rrc values in a time window of 1 h. These matchups were used to develop DLACC. Validations using in situ data from Aerosol Robotic Network-Ocean Color (AERONET-OC) stations showed that more matchups were obtained with the DLACC model than with the NIR algorithm and Korea Ocean Satellite Center standard AC algorithm in GDPS 2.0 (KOSC), and the accuracies were similar. More importantly, the DLACC algorithm is more tolerant of AEs and reduces AEs in a 2-pixel distance from cloud edges by 10% in the blue bands and by 50% in the green band. As a result, more valid observations are obtained with the DLACC algorithm, with the daily percentage of valid observations (DPVOs) increasing by 71% and 62% over those with the NIR algorithm and the KOSC algorithm, respectively. These increases in valid observations could further result in more consistent patterns in terms of space and time. Our DLACC algorithm can not only be used to process GOCI images but also provide an open framework to develop corresponding deep-learning models for other geostationary satellites.

The Impacts of the Application of the Ensemble Optimal Interpolation Method in Global Ocean Wave Data Assimilation

The ensemble optimal interpolation method was used in this study to conduct an examination of the assimilations of significant wave height (SWH) data from HY-2A satellite altimeter based on the WAVEWATCH III global ocean wave model. The results suggested that the ensemble optimal interpolation method using HY-2A SWH data played a positive role in enhancing the accuracy of the global ocean wave simulations and could effectively improve the deviations of SWH in the simulation processes. The root mean square errors of the NDBC buoy inspections were improved by 7 to 44% after the assimilation, and those of China’s offshore buoy inspections were improved by 3 to 11% after the assimilation. It was observed that the farther the buoys were from the shore, the better the effects of the assimilation improvements. The root mean square errors of the Jason-2 satellite data validations were improved by 17% after the assimilation, with monthly improvements of 8–25%. The improvements occurred in most of the global oceans, particularly in the Southern Ocean, the Eastern Pacific Ocean and the Indian Ocean. The results obtained in this research can be used as a reference for the operational applications of China’s ocean satellite data in ocean wave data assimilation and prediction.

Modulation Effects of Mesoscale Eddies on Sea Surface Wave Fields in the South China Sea Derived From a Wave Spectrometer Onboard the China‐France Ocean Satellite

AbstractThis study aims to examine the modulation effects of mesoscale eddies on sea surface wave fields in the South China Sea using data measured by a wave spectrometer named Surface Wave Investigation and Monitoring onboard the China‐France Ocean Satellite from July 2019 to August 2021. The statistical results show that the significant wave heights (SWHs) decrease (increase) from inside to outside of warm (cold) eddies, while the wavelengths have minimum values at the eddy edge in all cases. In cases of strong warm eddies, the maximum variations in SWH, wavelength, and wave direction caused by mesoscale eddy modulation reach up to 18%, 24%, and 28.4°, respectively. The eddy modulation on wave parameters is stronger at low sea states. The correlation analysis reveals that the wave parameter variations across eddies result from the eddy current shear and the direction difference between the eddy current and the waves. The dynamic analysis indicates that the deformation term of the eddy current is a dominant term affecting the SWH at the eddy edge, and the eddy current divergence and its product with wave direction determine the wavelength variation. The eddy vorticity deflects the wave direction by approximately 2°–3°, and the wave‐eddy angle and eddy‐modulated wind variation also contribute to the wave direction variation. The background current‐wave interaction is negligible in the eddy cases studied.

Significant Wave Height Prediction in the South China Sea Based on the ConvLSTM Algorithm

Deep learning methods have excellent prospects for application in wave forecasting research. This study employed the convolutional LSTM (ConvLSTM) algorithm to predict the South China Sea (SCS) significant wave height (SWH). Three prediction models were established to investigate the influences of setting different parameters and using multiple training data on the forecasting effects. Compared with the SWH data from the China–France Ocean Satellite (CFOSAT), the SWH of WAVEWATCH III (WWIII) from the pacific islands ocean observing system are accurate enough to be used as training data for the ConvLSTM-based SWH prediction model. Model A was preliminarily established by only using the SWH from WWIII as the training data, and 20 sensitivity experiments were carried out to investigate the influences of different parameter settings on the forecasting effect of Model A. The experimental results showed that Model A has the best forecasting effect when using three years of training data and three hourly input data. With the same parameter settings as the best prediction performance Model A, Model B and C were also established by using more different training data. Model B used the wind shear velocity and SWH as training and input data. When making a 24-h SWH forecast, compared with Model A, the root mean square error (RMSE) of Model B is decreased by 17.6%, the correlation coefficient (CC) is increased by 2.90%, and the mean absolute percentage error (MAPE) is reduced by 12.2%. Model C used the SWH, wind shear velocity, wind and wave direction as training and input data. When making a 24-h SWH forecast, compared with Model A, the RMSE of Model C decreased by 19.0%, the CC increased by 2.65%, and the MAPE decreased by 14.8%. As the performance of the ConvLSTM-based prediction model mainly rely on the SWH training data. All the ConvLSTM-based prediction models show a greater RMSE in the nearshore area than that in the deep area of SCS and also show a greater RMSE during the period of typhoon transit than that without typhoon. Considering the wind shear velocity, wind, and wave direction also used as training data will improve the performance of SWH prediction.