The Haihe River is a typical sluice-controlled river in the north of China. The construction and operation of sluice dams change the flow and other hydrological factors of rivers, which have adverse effects on water, making it difficult to study the characteristics of water quality change and water environment control in northern rivers. In recent years, remote sensing has been widely used in water quality monitoring. However, due to the low signal-to-noise ratio (SNR) and the limitation of instrument resolution, satellite remote sensing is still a challenge to inland water quality monitoring. Ground-based hyperspectral remote sensing has a high temporal-spatial resolution and can be simply fixed in the water edge to achieve real-time continuous detection. A combination of hyperspectral remote sensing devices and BP neural networks is used in the current research to invert water quality parameters. The measured values and remote sensing reflectance of eight water quality parameters (chlorophyll-a (Chl-a), phycocyanin (PC), total suspended sediments (TSS), total nitrogen (TN), total phosphorus (TP), ammonia nitrogen (NH₄-N), nitrate-nitrogen (NO₃-N), and pH) were modeled and verified. The results show that the performance R² of the training model is above 80%, and the performance R² of the verification model is above 70%. In the training model, the highest fitting degree is TN (R² = 1, RMSE = 0.0012 mg/L), and the lowest fitting degree is PC (R² = 0.87, RMSE = 0.0011 mg/L). Therefore, the application of hyperspectral remote sensing technology to water quality detection in the Haihe River is a feasible method. The model built in the hyperspectral remote sensing equipment can help decision-makers to easily understand the real-time changes of water quality parameters.

Numerous aquaculture ponds are intensively distributed around inland natural lakes and mixed with cropland, especially in areas with high population density in Asia. Information about the distribution of aquaculture ponds is essential for monitoring the impact of human activities on inland lakes. Accurate and efficient mapping of inland aquaculture ponds using high-spatial-resolution remote-sensing images is a challenging task because aquaculture ponds are mingled with other land cover types. Considering that aquaculture ponds have intertwining regular embankments and that these salient features are prominent at different scales, a Row-wise and Column-wise Self-Attention (RCSA) mechanism that adaptively exploits the identical directional dependency among pixels is proposed. Then a fully convolutional network (FCN) combined with the RCSA mechanism (RCSANet) is proposed for large-scale extraction of aquaculture ponds from high-spatial-resolution remote-sensing imagery. In addition, a fusion strategy is implemented using a water index and the RCSANet prediction to further improve extraction quality. Experiments on high-spatial-resolution images using pansharpened multispectral and 2 m panchromatic images show that the proposed methods gain at least 2–4% overall accuracy over other state-of-the-art methods regardless of regions and achieve an overall accuracy of 85% at Lake Hong region and 83% at Lake Liangzi region in aquaculture pond extraction.

Satellite remote sensing is a highly useful tool for monitoring chlorophyll-a concentration (Chl-a) in water bodies. Remote sensing algorithms based on near-infrared-red (NIR-red) wavelengths have demonstrated great potential for retrieving Chl-a in inland waters. This study tested the performance of a recently developed NIR-red based algorithm, SAMO-LUT (Semi-Analytical Model Optimizing and Look-Up Tables), using an extensive dataset collected from five Asian lakes. Results demonstrated that Chl-a retrieved by the SAMO-LUT algorithm was strongly correlated with measured Chl-a (R² = 0.94), and the root-mean-square error (RMSE) and normalized root-mean-square error (NRMS) were 8.9 mg∙m⁻³ and 72.6%, respectively. However, the SAMO-LUT algorithm yielded large errors for sites where Chl-a was less than 10 mg∙m⁻³ (RMSE = 1.8 mg∙m⁻³ and NRMS = 217.9%). This was because differences in water-leaving radiances at the NIR-red wavelengths (i.e., 665 nm, 705 nm and 754 nm) used in the SAMO-LUT were too small due to low concentrations of water constituents. Using a blue-green algorithm (OC4E) instead of the SAMO-LUT for the waters with low constituent concentrations would have reduced the RMSE and NRMS to 1.0 mg∙m⁻³ and 16.0%, respectively. This indicates (1) the NIR-red algorithm does not work well when water constituent concentrations are relatively low; (2) different algorithms should be used in light of water constituent concentration; and thus (3) it is necessary to develop a classification method for selecting the appropriate algorithm.