Search

Lakes play a crucial role in the global biogeochemical cycles through the transport, storage, and transformation of different biogeochemical compounds. Their regulatory service appears to be disproportionately important relative to their small areal extent, necessitating continuous monitoring. This study leverages the potential of optical remote sensing sensors, specifically Sentinel-2 Multispectral Imagery (MSI), to monitor and predict water quality parameters in lakes. Optically active parameters, such as chlorophyll a (CHL), total suspended matter (TSM), and colored dissolved matter (CDOM), can be directly detected using optical remote sensing sensors. However, the challenge lies in detecting non-optically active substances, which lack direct spectral characteristics. The capabilities of artificial intelligence applications can be used in the identification of optically non-active compounds from remote sensing data. This study aims to employ a machine learning approach (combining the Genetic Algorithm (GA) and Extreme Gradient Boost (XGBoost)) and in situ and Sentinel-2 Multispectral Imagery data to construct inversion models for 16 physical and biogeochemical water quality parameters including CHL, CDOM, TSM, total nitrogen (TN), total phosphorus (TP), phosphate (PO4), sulphate, ammonium nitrogen, 5-day biochemical oxygen demand (BOD5), chemical oxygen demand (COD), and the biomasses of phytoplankton and cyanobacteria, pH, dissolved oxygen (O2), water temperature (WT) and transparency (SD). GA_XGBoost exhibited strong predictive capabilities and it was able to accurately predict 10 biogeochemical and 2 physical water quality parameters. Additionally, this study provides a practical demonstration of the developed inversion models, illustrating their applicability in estimating various water quality parameters simultaneously across multiple lakes on five different dates. The study highlights the need for ongoing research and refinement of machine learning methodologies in environmental monitoring, particularly in remote sensing applications for water quality assessment. Results emphasize the need for broader temporal scopes, longer-term datasets, and enhanced model selection strategies to improve the robustness and generalizability of these models. In general, the outcomes of this study provide the basis for a better understanding of the role of lakes in the biogeochemical cycle and will allow the formulation of reliable recommendations for various applications used in the studies of ecology, water quality, the climate, and the carbon cycle.

Nutrients are important elements in marine ecosystems and water quality, and have a major role in the eutrophication of water bodies. Monitoring nutrient loads is especially important for the Baltic Sea, which is especially sensitive to the eutrophication. Using optical remote sensing data in mapping total nitrogen (TN) and total phosphorus (TP) is challenging because these substances do not have a direct influence on the water optics that remote sensing sensors can detect. On the other hand, it would be very rewarding. In this study, more than 25,000 Sentinel-3 Ocean and Land Colour Instrument (OLCI) data algorithms were tested in order to detect the TN and TP concentrations in the Estonian marine waters between 2016–2021. The TN estimations were well derived for Estonian marine waters (R2 = 0.73, RMSE = 4.87 µmolN L−1, MAPE = 14%, n = 708), while the TP estimations were weaker (R2 = 0.38, RMSE = 0.23 µmolP L−1, MAPE = 24%, n = 730). The Estonian marine waters were divided into six geographic regions in order to study the effect of regional water quality on the TN and TP retrievals. The nutrient concentrations were derived in every region when spring and summer periods were treated separately. In this study, the detection of both nutrients was more successful in more closed areas with P deficiency, while in open sea areas it was more challenging. This study shows that it is possible to estimate nutrients, especially TN, from remote sensing data. Consequently, remote sensing could provide a reliable support to the conventional monitoring by covering large marine areas with high temporal and spatial resolution data.

Phytoplankton primary production (PP) in lakes play an important role in the global carbon cycle. However, monitoring the PP in lakes with traditional complicated and costly in situ sampling methods are impossible due to the large number of lakes worldwide (estimated to be 117 million lakes). In this study, bio-optical modelling and remote sensing data (Sentinel-3 Ocean and Land Colour Instrument) was combined to investigate the spatial and temporal variation of PP in four Baltic lakes during 2018. The model used has three input parameters: concentration of chlorophyll-a, the diffuse attenuation coefficient, and incident downwelling irradiance. The largest of our studied lakes, Võrtsjärv (270 km2), had the highest total yearly estimated production (61 Gg C y−1) compared to the smaller lakes Lubans (18 Gg C y−1) and Razna (7 Gg C y−1). However, the most productive was the smallest studied, Lake Burtnieks (40.2 km2); although the total yearly production was 13 Gg C y−1, the daily average areal production was 910 mg C m−2 d−1 in 2018. Even if lake size plays a significant role in the total PP of the lake, the abundance of small and medium-sized lakes would sum up to a significant contribution of carbon fixation. Our method is applicable to larger regions to monitor the spatial and temporal variability of lake PP.

Many lakes in boreal and arctic regions have high concentrations of CDOM (coloured dissolved organic matter). Remote sensing of such lakes is complicated due to very low water leaving signals. There are extreme (black) lakes where the water reflectance values are negligible in almost entire visible part of spectrum (400–700 nm) due to the absorption by CDOM. In these lakes, the only water-leaving signal detectable by remote sensing sensors occurs as two peaks—near 710 nm and 810 nm. The first peak has been widely used in remote sensing of eutrophic waters for more than two decades. We show on the example of field radiometry data collected in Estonian and Swedish lakes that the height of the 810 nm peak can also be used in retrieving water constituents from remote sensing data. This is important especially in black lakes where the height of the 710 nm peak is still affected by CDOM. We have shown that the 810 nm peak can be used also in remote sensing of a wide variety of lakes. The 810 nm peak is caused by combined effect of slight decrease in absorption by water molecules and backscattering from particulate material in the water. Phytoplankton was the dominant particulate material in most of the studied lakes. Therefore, the height of the 810 peak was in good correlation with all proxies of phytoplankton biomass—chlorophyll-a (R2 = 0.77), total suspended matter (R2 = 0.70), and suspended particulate organic matter (R2 = 0.68). There was no correlation between the peak height and the suspended particulate inorganic matter. Satellite sensors with sufficient spatial and radiometric resolution for mapping lake water quality (Landsat 8 OLI and Sentinel-2 MSI) were launched recently. In order to test whether these satellites can capture the 810 nm peak we simulated the spectral performance of these two satellites from field radiometry data. Actual satellite imagery from a black lake was also used to study whether these sensors can detect the peak despite their band configuration. Sentinel 2 MSI has a nearly perfectly positioned band at 705 nm to characterize the 700–720 nm peak. We found that the MSI 783 nm band can be used to detect the 810 nm peak despite the location of this band is not in perfect to capture the peak.