An outbreak of Ulva prolifera poses a massive threat to coastal ecology in the Southern Yellow Sea, China (SYS). It is a necessity to extract its area and monitor its development accurately. At present, Ulva prolifera monitoring by remote sensing imagery is mostly based on a fixed threshold or artificial visual interpretation for threshold selection, which has large errors. In this paper, an adaptive threshold model based on Google Earth Engine (GEE) is proposed and applied to extract U. prolifera in the SYS. The model first applies the Floating Algae Index (FAI) or Normalized Difference Vegetation Index (NDVI) algorithm on the preprocessed remote sensing images and then uses the Canny Edge Filter and Otsu threshold segmentation algorithm to extract the threshold automatically. The model is applied to Landsat8/OLI and Sentinel-2/MSI images, and the confusion matrix and cross-sensor comparison are used to evaluate the accuracy and applicability of the model. The verification results show that the model extraction of U. prolifera based on the FAI algorithm has higher accuracy (R² = 0.99, RMSE = 5.64) and better robustness. However, when the average cloud cover is more than 70% in the image (based on the statistical results of multi-year cloud cover information), the model based on the NDVI algorithm has better applicability and can extract the algae distributed at the edge of the cloud. When the model uses the FAI algorithm, it is named FAI-COM (model based on FAI, the Canny Edge Filter, and Otsu thresholding). And when the model uses the NDVI algorithm, it is named NDVI-COM (model based on NDVI, the Canny Edge Filter, and Otsu thresholding). Therefore, the final extraction results are generated by supplementing NDVI-COM results on the basis of FAI-COM extraction results in this paper. The F1-score of U. prolifera extracted results is above 0.85. The spatiotemporal distribution of U. prolifera in the South Yellow Sea from 2016 to 2020 is obtained through the model calculation. Overall, the coverage area of U. prolifera shows a decreasing trend over the five years. It is found that the delay in recovery time of Porphyra yezoensis culture facilities in the Northern Jiangsu Shoal and the manual salvage and cleaning-up of U. prolifera in May are among the reasons for the smaller interannual scale of algae in 2017 and 2018.

Although satellite remote sensing technology is intensively used for the monitoring of water quality, the inversion of coastal water bodies and non-optically active parameters is still a challenging issue. Few ongoing studies use remote sensing technology to analyze the driving forces of changes in water quality from multiple aspects based on inversion results. By the use of Landsat 5/8 imagery and measured in situ data of the total nitrogen (TN) and total phosphorus (TP) in the Shenzhen-Hong Kong Bay area from 1986 to 2020, this study evaluated the modeling effects of four machine learning methods named Tree Embedding (TE), Support Vector Regression (SVR), Gaussian Process Regression (GPR), and Back-propagation Neural Network (BPNN). The results show that the BPNN creates the most reliable and robust results. The values of the obtained correlation coefficients (r) are 0.83, 0.92, 0.84, and 0.90, and that of the coefficients of determination (R²) are 0.70, 0.84, 0.70, and 0.81. The calculated mean absolute errors (MAEs) are 0.41, 0.16, 0.06, and 0.02, while the root mean square errors (RMSEs) are 0.78, 0.29, 0.12, and 0.03. The concentrations of TN and TP (C_TN, C_TP) in the Shenzhen Bay, the Starling Inlet, and the Tolo Harbor were relatively high, fluctuated from 1986 to 2010, and decreased significantly after 2010. The C_TN and C_TP in the Mirs Bay kept continuously at a low level. We found that urbanization and polluted river discharges were the main drivers of spatial and inter-annual differences of C_TN and C_TP. Temperature, precipitation, and wind are further factors that influenced the intra-annual changes of C_TN and C_TP in the Shenzhen Bay, whilethe expansion of oyster rafts and mangroves had little effect. Our research confirms that machine learning algorithms are well suited for the inversion of non-optical activity parameters of coastal water bodies, and also shows the potential of remote sensing for large-scale, long-term monitoring of water quality and the subsequent comprehensive analysis of the driving forces.

Recurring green tides (also called Ulva prolifera ) cause significant damage to marine ecosystems in the Yellow Sea of China, especially in the settlement region. The settlement region is a critical area for the natural decay and decomposition of green tides, which obviously has ecological consequences. Recent studies in relation to this topic are mostly based on point observation data, which prevents to quantitatively analyze the ecological impact patterns of green tide biomass at large spatial scales. Therefore, we used remote sensing time-series of Geostationary Ocean Color Imager satellite data combined with in situ data to invert marine chlorophyll-a (Chl-a, main indicators representing phytoplankton biomass) concentrations by machine learning methods. Finally, based on a cross-satellite model, we quantified the green tide biomass that occurred in Haiyang City during 2015 and 2016. The main results found are as follows. First, the green tide biomass reveals negative correlations on Chl-a concentration in the settlement region. The Chl-a concentration showed a decreasing trend and remained at a low level (0.1–0.2 mg/m ³ ) when the biomass of the green tide increased. After the disappearance of the green tides, the Chl-a concentration accreted rapidly and then began to gradually decrease. Second, combined with the pixel-based statistics grid of the green tide and the average Chl-a concentration, the time cycle of green tides in the settlement region is about 30–35 d. Finally, some special cases (such as typhoon) can change the pattern and temporal cycle in the settlement region. This article provides support for marine ecosystem monitoring.