One of the major sources of pollution affecting inland and coastal waters is related to poorly treated or untreated wastewater discharge, particularly in urbanized watersheds. The excess of nutrients, organic matter, and pathogens causes an overall deterioration of water quality and impairs valuable ecosystem services. The detection of wastewater pollution is essential for the sustainable management of inland and coastal waters, and remote sensing has the capability of monitoring wastewater contamination at extended spatial scales and repeated frequencies. This study employed satellite-derived water quality indicators and spatiotemporal analysis to assess the risk of wastewater contamination in Conceição Lagoon, a coastal lagoon in Southern Brazil. Using an analytical model, three water quality indicators were derived from Level 2A Sentinel-2 MSI images: the absorption coefficients of chlorophyll-a and detritus combined with coloured dissolved organic matter, and the backscattering coefficient of suspended solids. The temporal standardized anomalies were calculated for each water quality indicator for the period of 2019–2021, and their anomalies during a known outfall event were used to evaluate spatial variation modes. The spatial mode explaining most of the variability was used to estimate weights for the water quality indicators anomalies in a linear transformation that can indicate the risk of wastewater contamination. Results showed that the wastewater spatial mode for this region was characterized by positive anomalies of backscattering coefficient of particulate matter and absorption coefficient of detritus combined with coloured dissolved organic matter, each with a relative importance of 50%. The application of this spatiotemporal analysis was formulated as the Wastewater Contamination Index. With the aid of photographic records, and additional meteorological and water quality data, the results of the index were verified for wastewater outfall events in the study area. The methodology for constructing the proposed Wastewater Contamination Index applies to other locations and can be a valuable tool for operational monitoring of wastewater contamination.
Dissolved oxygen (DO) is crucial for the health of aquatic ecosystems, and plays an essential role in regulating biogeochemical processes in inland lakes. Traditional measurements of DO using the probe or analysis in a laboratory are time-consuming and cannot obtain data with high frequency and broad coverage. Satellites can provide daily/hourly observations within a broad scale and have been used as an important technique for aquatic environments monitoring. However, satellite-derived DO in waters is challenging due to its non-optically active property. Here, we developed a two-step model for retrieving DO concentration in Lake Taihu from Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua images. A machine learning model (eXtreme gradient boosting) was developed to estimate DO from field water temperature, water clarity, and chlorophyll-a (Chla) (root-mean-square error (RMSE) = 0.98 mg L−1, mean absolute percentage error (MAPE) = 7.9%) and subsequently was validated on MODIS-derived water temperature, water clarity, and Chla matchups with a satisfactory accuracy (RMSE = 1.28 mg L−1, MAPE = 9.9%). MODIS-derived DO in Lake Taihu from 2002 to 2021 demonstrated that DO ranged from 7.2 mg L−1 to 14.2 mg L−1, with a mean value of 9.3 mg L−1. DO in the northern region was higher than in the central and southern regions, and higher in winter than in summer. We revealed that DO in this decade (2010–2021) was considerably lower than that in the last decade (2002–2009). Meanwhile, annual mean of DO increased in 2002–2009 and decreased from 2010 to 2021. The spatial distribution of DO in Lake Taihu was related to Chla and water clarity, while seasonal and interannual variations in DO resulted from air temperature primarily. This research enhances the potential use of machine learning approaches in monitoring non-optically active constituents from satellite imagery and indicates the possibility of long-term and high-range variations in more water quality parameters in lakes.
Remote sensing is useful for quantifying water-quality parameters for managing inland water systems. However, the single water-quality retrieval model usually has poor applicability in large regions. To solve the issue of low retrieval accuracy of water-quality parameters in inland water, the study area herein is geographically divided into rural water and urban water according to the proportion of land-use types in the riparian zones. Furthermore, the machine-learning regression algorithms are used to construct the retrieval models suitable for the total nitrogen (TN) and total phosphorus (TP) concentrations based on the measured water-quality data and the simultaneous Sentinel-2 Multispectral Imager (MSI) images. Additionally, the optical retrieval models are applied to the MSI images acquired on different dates to analyze the variations of TN and TP concentrations in the water around Chaohu Lake of China. The results show that the three accuracy indices of determination coefficient (R2), mean square error (MSE), and mean absolute percentage error (MAPE) of the TN concentration retrieval models for rural water and urban water were 0.67, 0.37 mg/L, and 36.81%, and 0.78, 0.34 mg/L, and 8.34%, respectively, while those of the TP concentration retrieval model for rural water and urban water reached 0.46, 0.0034 mg/L, and 38.60%, and 0.58, 0.018 mg/L, and 37.57%, respectively. The accuracy of the TN and TP concentration retrieval model constructed using geographical division is significantly better than that which does not use geographical division. According to the retrieval results from MSI images, the TN and TP concentrations in urban water are higher than those in rural water. TN and TP concentrations in urban water are stable throughout the year and peak in December, while those of rural water are highest in March and lowest in November. The method proposed in this study can provide a new idea for improving the retrieval accuracy of water-quality parameters in different water bodies in a large-scale region, and the relevant conclusion can provide a theoretical basis for water pollution control and prevention strategies in agricultural basins.
Ultraviolet-visible spectroscopy is an effective tool for reagent-free qualitative analysis and quantitative detection of water parameters. Suspended particles in water cause turbidity that interferes with the ultraviolet-visible spectrum and ultimately affects the accuracy of water parameter calculations. This paper proposes a deep learning method to compensate for turbidity interference and obtain water parameters using a partial least squares regression approach. Compared with orthogonal signal correction and extended multiplicative signal correction methods, the deep learning method specifically utilizes an accurate one-dimensional U-shape neural network (1D U-Net) and represents the first method enabling turbidity compensation in sampling real river water of agricultural catchments. After turbidity compensation, the R2 between the predicted and true values increased from 0.918 to 0.965, and the RMSE (Root Mean Square Error) value decreased from 0.526 to 0.343 mg. Experimental analyses showed that the 1D U-Net is suitable for turbidity compensation and provides accurate results.
Long-term lake surface water temperature (LSWT) products are valuable for understanding the responses of lake ecosystems to climate warming and for proposing suitable policies to protect lake ecosystems. Here, using Landsat satellite data and various in situ data, we documented 36 years (1986–2021) of spatiotemporal variations in LSWT in Lake Qiandaohu, a subtropical deep-water lake in China, and explored the potential driving factors of these variations. We validated the performances of the practical single-channel (PSC) algorithm, the generalized single-channel algorithm and the Landsat Level 2 land surface temperature product on Lake Qiandaohu with long-term in situ buoy data. Overall, the PSC algorithm had the best performance, with a mean absolute percent error (MAPE) of 7.5% and root mean square difference (RMSE) of 1.7°C. With 36 years of Landsat data and the PSC algorithm, the spatiotemporal variations in LSWT were constructed. The Landsat-derived 36-year mean LSWT in Lake Qiandaohu ranged from 18.2 to 23.1°C, with a mean value of 20.2°C. The northeast and southwest subsegments had the minimum (19.7°C) and maximum (20.6°C) mean LSWT values, respectively. The spatial variations in LSWT could be explained in part by the water depth. From 1986 to 2021, a significant warming trend was observed in Lake Qiandaohu, with a warming rate of 0.07°C/year. The warming rate of Lake Qiandaohu was faster than that of the local air temperature (warming rate = 0.04°C/year). The LSWT warming in Lake Qiandaohu can mainly be attributed to the warming air temperatures. Lake warming has increased the thermal stability in Lake Qiandaohu and has had negative impact on the lake ecosystem. Our work highlights the importance of using satellite data to understand the responses of lake ecosystems to climate change.