According to the data of four voyages of zooplankton survey in the southwest of Bohai Sea from August 2020 to May 2021, the community structure and seasonal variation of small medusae were analyzed, and the effects of environmental factors on the abundance of small medusae were discussed. There were seasonal variations in species composition and abundance distribution of small medusae in the southwest of Bohai Sea, a total of 13 species of small medusae were found throughout the year, including 11 species of hydromedusae and 2 species of ctenophores. The number of medusae species in spring, summer, autumn and winter were 4, 9, 7 and 2 species respectively. The mean abundance values were 30.74 ind./m³, 30.78 ind./m³, 12.08 ind./m³ and 0.57 ind./m³ respectively. The dominant species were Sugiura chengshanense, Rathkea octopunctata, Eirene ceylonensis, Clytia hemisphaerica, Pleurobrachia globosa. The average seasonal replacement rate of dominant species was 91.67%, showing obvious seasonal succession. Water temperature and salinity were main environmental factor affecting the seasonal variation of total abundance of small medusae in the southwest of Bohai Sea. The increase of water temperature and nutrients in spring promote the growth and reproduction of small medusae. In summer, copepods provided rich bait for small medusae to promote its growth, the community was mainly affected by salinity in autumn. According to the risk species of small medusae in the southwest of Bohai Sea, it was speculated that the peak of risk species in summer and autumn was a risk to the planned nuclear power.

Acanthaster planci, one of the predators of reef-building corals, has attracted much attention for its catastrophic damage to coral reef ecosystems. However, the spatial and temporal distribution characteristics of A. planci are still unclear in the coral reef ecosystem of the South China Sea. In this study, using environmental DNA and real-time quantitative PCR techniques, we analyzed the concentration variation of the mitochondrial cytochrome-c-oxidase subunits I (COTS-mtCOI) fragment of A. planci in the surface seawater of the Xisha Islands in September 2020, April 2021 and January 2022, and the correlations between the concentration variation with environmental factors such as seawater temperature, salinity, pH, chlorophyll content, nutrients content and other environmental factors. The results showed that COTS-mtCOI fragment concentration in the Xisha Islands varied from 0 copies/m³ to 4.13×10⁷ copies/m³ during 2020−2022, and there were always higher concentrations in the Yongle Atoll. For Huaguang Reef, Jinqing Islands, Lingyang Reef, Quanfu Island and Zhaoshu Island, the average concentration of COTS-mtCOI fragment in September 2020 was significantly (p<0.05) higher than those in April 2021 and January 2022. In addition, COTS-mtCOI fragment concentration was significantly (p<0.05) positively correlated with surface seawater temperature. These results suggest that the population of A. planci is widely distributed in the seawater of Xisha Islands, and higher density of A. planci could appear in Yongle Atoll. Moreover, ocean warming may accelerate the outbreak of A. planci. This study is helpful to understand the population distribution characteristics of A. planci in the coral reef ecosystems of the South China Sea, and can provide a theoretical basis for the early warning and forecast of the A. planci outbreak.

Bacteria play an important role in the aquatic ecosystem. In our work, bacterial community structure and assembly mechanisms were studied in Sansha Bay in Fujian Province based on DNA and RNA high-throughput sequencing. Our results revealed that: (1) A total of 1476 operational taxonomic units (OTUs) were detected, and γ-proteobacteria, α-proteobacteria, cyanobacteriia and bacteroidia were the most diverse bacterial groups. (2) γ-proteobacteria, α-proteobacteria, cyanobacteriia and bacteroidia were also the most abundant groups in Sansha Bay both in the DNA and RNA high-throughput sequencing results. The metabolic activities of these four dominant groups were different, and their metabolic activities were mainly regulated by salinity, total nitrogen, nitrite nitrogen and inorganic phosphate concentrations. (3) The bacterial community structure in Sansha Bay was different in spatial scale, and the closer the geographical location was, the more similar the bacterial community structure was. Neutral process was the main assembly mechanism affecting the construction of bacterial community in Sansha Bay. Our results were useful for the understanding the bacterial community structure and assembly mechanism in Sansha Bay, Fujian.

The sea surface temperature (SST) products, obtained from the Chinese Ocean Color and Temperature Scanner (COCTS) on the two haiyang satellites (HY1C and HY1D), play an important role in oceanic and atmospheric researches. It is important to know whether they are consistent with products from other satellites such as the Moderate Resolution Imaging Spectroradiometer (MODIS) on Terra and Aqua satellites. The data quality of SST global images from COCTS on HY1C/1D is evaluated by the average bias, absolute difference, root mean square error, and correlation coefficient based on in-situ SST measurements and the MODIS products. The results show that the spatial patterns of the daily and monthly global SST of HY1C/1D are similar to those of Terra/Aqua. The average bias, absolute difference, root mean square error and correlation coefficient of the global daily SST/HY1C products at daytime, based on the SST values of Terra on June 2020, are 0.04℃, 0.60℃, 0.78℃ and 0.98, respectively, and that of the nighttime products are −0.16℃, 0.78℃, 0.95℃ and 0.86, respectively. Similarly, the values of the daytime SST products of HY1D comparing with the SST of Aqua on June 2020 are 0.02℃, 0.59℃, 0.79℃ and 0.98, and that of the nighttime products are −0.09℃, 0.61℃, 0.82℃ and 0.96, respectively. The evaluation on other seasons indicates that the SST products from COCTS are very stable. However, the data quality control scheme and inhomogeneity correction still need to be developed to improve the performance of SST products of HY1C/1D. As a whole, the SST products of HY1C/1D can be used in some applications, similar to the Terra/Aqua products.

Mangroves forests, as a coastal zone ecosystem dominated by mangrove plants in the tropics and subtropics, are one of the important coastal wetland types. In this paper, multi-source and multi-phase satellite data were used to form a data atlas of shoreline, reclamation, aquaculture area, mangrove distribution in the Guangdong-Hong Kong-Macao Greater Bay Area from 1969 to 2020, and the time series analysis of the evolution of mangroves in the Greater Bay Area was obtained by using the combine mangrove recognition index (CMRI). The results show that the existing mangrove forests data set can be obtained by interpreting the multi-source remote sensing data, and the CMRI time series data can establish the history of the existing mangrove forest change, and then effectively estimate the mangrove forest age. The temporal and spatial distribution of mangroves in the Guangdong-Hong Kong-Macao Greater Bay Area has undergone obvious changes, with the existing mangroves being about 3 316 hm², and the existing forest age in various regions in the Greater Bay Area is quite different, and the overall average forest age is 20 a. In the past 50 years, the shoreline as a whole has moved towards the sea, and the changes in shoreline, reclamation, and breeding areas have significantly affected the area, spatial distribution, and age of mangroves. Artificial cultivation has been the main reason for the restoration of mangroves in the past 20 years.

Mangroves are important for maintaining biodiversity as well as ecological balance. Therefore, it is necessary to extract mangrove vegetation information efficiently and accurately and to monitor it in real time. A deep learning method for pixel-level accurate extraction of mangroves from high-resolution remote sensing images is presented in this paper. For the problem of low accuracy of mangrove remote sensing classification, CU-Net model for mangrove identification is constructed by introducing CLoss loss function by strengthening image center information and weakening edge information, and adding Dropout and Batch Normalization layers. And a new prediction model is constructed by sliding overlap splicing method, which effectively solves the problem of insufficient edge information and splicing traces in the prediction results. The recognition results of the proposed method are compared with the prediction results of U-Net, SegNet and DenseNet models as well as the traditional SVM and RF methods. The results show that the proposed model has stronger generalization ability and better recognition effect compared with other deep learning models. In the two test areas, the average OA and MIoU reach 94.43% and 88.12%, respectively. The average F1-score in mangrove and ordinary trees reach 95.96% and 90.49%, respectively. The accuracy is significantly higher than that of traditional SVM and RF methods, as well as several other neural networks. The effectiveness of the model in the field of mangrove recognition is verified, which can provide a new idea for the field of high resolution remote sensing mangrove recognition.

Satellite derived bathymetric using multispectral imagery is an effective means to obtain shallow water depth information. However, its validity is limited to optical shallow water areas, but presents a “pseudo-shallow sea” distortion phenomenon in deep water areas. Therefore, accurately identifying the valid region of satellite derived bathymetry (SDB) data is crucial for its wide application. Based on high-spatial resolution remote sensing image, a data-driven method for evaluating the validity of SDB based on analysis of the differences in the statistical distribution of radiance in deep/shallow water regions is proposed in this paper. This method uses the local standard deviation of the radiance information of satellite images as a feature, optimizes the statistical characteristics of the optical deep water area based on the K-S test method, and uses the hypothesis test method to identify the SDB corresponding to the deep water invalid area. The experimental results in Ganquan Island region show that the method can effectively identify the invalid SDB associated with the optical deep water area by dividing the boundary between optical shallow and deep water area. After removing the invalid data, the mean absolute error (MAE) of SDB in the optical shallow region is 1.01, and the root mean square error (RMSE) is 1.52. The experimental results show that the proposed method can accurately identify the optical shallow region of SDB result, which benefits the interpretation and application of SDB results.

Ocean is an important carbon sink in nature. The sea-air carbon dioxide flux is usually estimated by the difference of partial pressure of carbon dioxide (pCO₂) between the atmosphere and the sea surface. Due to the imbalance of observation data on temporal and spatial distribution and datasets used for prediction, there is still large room for improvement in spatial resolution for present reconstruction of pCO₂ on sea surface. In order to fit the temporal and spatial variability under high spatial resolution better, based on the sea surface fugacity of carbon dioxide (fCO₂) observations of the Surface Ocean CO₂ Atlas (SOCAT) and other multi-source data including remote sensing data, the nonlinear relationship between sea surface pCO₂ and physical, biological, optical factors was established by a XGBoost model and a weight model was built based on spatiotemporal frequency of samples. A 0.041 7°×0.041 7° monthly sea surface pCO₂ dataset in Atlantic from 2000 to 2018 was finally constructed with correlation coefficient of 0.966, mean squared error of 8.087 μatm and mean error of 4.012 μatm on prediction dataset. The reconstruction is highly consistent to other similar reconstruction results on temporal and spatial trend and also gains advantage in spatial resolution.

Separation of wind-wave and swell is the basis for studying the respective characteristics of wind-wave and swell. However, due to the lack of wave spectrum data, it is difficult to popularize and apply separation methods based on wave spectrums. An effective solution is to use wave observations that are easy to obtain, namely basic wave elements to separate wind-wave and swell. Existing methods cannot use basic wave elements to comprehensively calculate the proportions and characteristic parameters of wind-wave and swell. For this reason, this paper introduces machine learning into the separation of wind-wave and swell. Based on the multi-layer perceptron model, a method using wave elements and wind elements to accurately estimate wind-wave and swell parameters is proposed. This method requires each station to provide at least 466 training samples of wave data and 766 or more training samples are recommended. The method is suitable for 3 stations in the Taiwan Strait with its accuracy significantly better than traditional methods based on wave spectrums. The proposed method can provide alternative calculation schemes of wind-wave and swell for stations lacking wave spectrums in this sea area. It helps expand the source of measured data of wind-wave and swell, therefore strengthening the research on the characteristics and early warning and forecasting of wind-wave and swell.

The spatio-temporal variations of significant wave height and its extremum in the northern South China Sea from 1996 to 2015 are analyzed based on the wave hindcast simulation by the third-generation wave model WaveWatch III (WW3). Two extreme value distribution methods, Pearson-III and Gumbel methods, have been used to estimate the return period of the extreme significant wave height in the northern South China Sea. The results show that the seasonal variation and spatial distribution of significant wave height in the northern South China Sea are consistent with the monsoon field, similar to the results of reanalysis data. It is high in autumn and winter and low in spring and summer, and decreases from the west part of the Luzon Strait to the southwest. But the extremum of significant wave height is strongly affected by the temporal resolution. The higher resolution (such as hourly), the more typhoon wave characteristics are displayed. After deducting the seasonal cycle signal, both the significant wave height and its annual extremum show an intense linear trend, increasing by 7.7% and 31.6% in the last 20 years respectively. There are several large value regions of the return period wave height in this area. They are related to the typhoon's track and intensity directly, indicating that typhoons are the primary mechanism causing extreme waves in this region. Comparing the Pearson-III and Gumbel distribution, it is found that if the extreme significant wave height was relatively low and the frequency decreased slowly with the growth of the extremum, the two methods are both accurate, with little difference which could be ignored. However, when the extreme significant wave height was much higher in one year than that in other years, the estimated result of the Pearson-III would be much higher than that of the Gumbel method, and also closer to the actual value. In other words, the Pearson-III extreme value distribution behaves better in this situation. This study shows that when the extreme significant wave height caused by a super typhoon is much higher than that in other years, the estimation from different methods differs greatly, which will significantly affect the assessment of the return period. Besides, the strong increasing trend of the extreme significant wave height in the northern South China Sea will also bring a non-negligible impact on the calculation of return period wave height and marine engineering protection in the future.