Primary production is the most fundamental ecological process in surface waters. Bio-optical models allow for the estimation of aquatic primary production from optical remote sensing data, providing a cost-effective alternative to incubation measurements. Existing bio-optical models use different parameters for common model components, namely: (1) the vertical resolution of chlorophyll- a , (2) the planar-to-scalar irradiance conversion factor, and (3) the photosynthesis-irradiance relationship. Using a unified modelling framework, we evaluate the impact of these three model components for optically complex inland and transitional waters by testing 16 model configurations against a comprehensive in situ dataset spanning oligotrophic to eutrophic conditions in Lake Geneva, three Estonian lakes, and three stations in the Baltic Sea. Our results demonstrate that the choice of photosynthesis-irradiance function is the most critical component, causing a 23.2 percentage range in median magnitude error and a 0.26 range in median vertical shape correlation. In contrast, using depth-resolved chlorophyll- a profiles shows a modest impact on primary production estimates, while the choice of irradiance conversion factor has a negligible impact. Time-series analysis further reveals that while the models successfully reproduce seasonal dynamics, their ability to resolve short-term variability is limited. Our findings underscore that model accuracy is constrained by the parameterisation of model components, uncertainties in empirical constants, and the in situ validation data itself. This highlights the need to prioritise the refinement of photosynthesis-irradiance relationships alongside community efforts to develop standardised validation datasets that include concurrent measurements of carbon isotope incubations and inherent optical properties.

The COVID-19 pandemic inadvertently offered a condition to evaluate how abrupt human-activity reductions affect freshwater ecosystems, particularly water turbidity. Using satellite-derived data from 774 lakes worldwide (2017-2022), here we show turbidity declined significantly in highly turbid zones of lakes following COVID-19 containment, with minor effects elsewhere. Globally, average peak turbidity decreased 7.0% in 2020 relative to 2019; 5.9% was directly attributable to containment measures, independent of climate. Without these measures, peak turbidity would have been similar to 5% higher (0.45 Nephelometric Turbidity Units) during 2020-2022. Lakes in regions with stricter containment and higher anthropogenic footprint exhibited larger declines and faster rebounds post-restriction. Among individual lakes, 75.2% experienced average peak turbidity reductions. For 168 lakes, containment-driven improvements averaged -18.6%, strongly correlated with reduced night-time light as a proxy for anthropogenic inputs. By disentangling human and climatic influences, our study provides globally relevant, actionable insights for targeted lake restoration strategies.

Measurements of water colour and clarity are used to track the environmental status of lakes, estuaries and oceans. The oldest standardised methods for measuring water clarity and colour are the Secchi disk and Forel-Ule colour scale. Both techniques were developed in the 19th century and require use of the human eye. Despite the advent of optoelectronic-based sensing, these visual methods are still used today, owing to their sensitivity, affordability, simplicity and long history of use. Recently, a hand-held device was developed for measuring the Secchi depth and Forel-Ule colour (presented in two formats, named the mini- and midi-Secchi disk). Designed to be small, light and convenient-to-use, it is well suited for participatory science projects that involve monitoring water colour and clarity. To date, over 900 mini- and midi-Secchi disks have been distributed to citizens and scientists, primarily through six projects, with data mostly transferred via mobile phone applications to data servers and dashboards. In this paper, we describe the methods used in the projects and show some characteristics of the datasets collected so far. We showcase how the device can be used for scientific applications, such as verifying satellite data products, gaining new scientific insights, and supporting public engagement and education. Finally, we provide suggestions for methodological improvements and future developments.

Phytoplankton play a central role in aquatic ecosystems, influencing biogeochemical cycles, food web dynamics, and overall water quality. Monitoring their composition is essential for assessing water ecosystem health and detecting environmental changes. Chlorophyll-a concentration is widely used as a proxy for phytoplankton abundance in inland waters. Together with colored dissolved organic matter and total suspended matter, these parameters can be retrieved from remote sensing reflectance data. However, identifying the detailed taxonomic composition of phytoplankton in lakes remains a major challenge. Spectral matching algorithms offer promising solutions to overcome this limitation. In this study, we investigated the potential of retrieving phytoplankton taxa composition from high-resolution in situ spectroscopy measurements by applying radiative transfer inversion and validating the results against phytoplankton abundance data obtained from an imaging microscope. First, we assessed the performance of our approach in retrieving four phytoplankton taxa under cloud-free conditions. Then, we extended the analysis to two seasons, covering multiple consecutive blooms using data acquired independently of cloudiness. The high agreement between the imaging microscopy results and those obtained from in situ spectroscopy indicates that remote sensing with radiative transfer inversions can track the evolution of phytoplankton blooms. The results suggest that low phytoplankton concentrations and the lack of unique spectral features for some taxa may prevent the accurate identification of phytoplankton composition through spectroscopy. In addition, the natural variability in cell size, along with physiological changes such as fluctuations in intracellular chlorophyll-a content, impacts the empirical conversion from cell cross section to intracellular chlorophyll-a content.

Emerging evidence suggests that temperature increases due to climate change not only differ strongly between regions but also across seasons. As a rule of thumb, one could argue that colder seasons (e.g., winter) tend to warm up faster than warmer seasons, although there are notable exceptions to this rule (e.g., due to changes in the polar vortex). The implications of such seasonal differences in warming trends for plant phenology, i.e., the timing of key events during the plant seasonal cycle, however remain poorly understood. A gap in knowledge that arises, in part, because we lack a global overview of the period(s) of the year during which changing temperatures impact on the phenological cycle of plants the most.Here, we provide a global analysis of the interrelationships between seasonal temperature changes and global land surface phenology using satellite data from the period 2001-2019. More specifically, we determined the annual period of highest correlation between temperature fluctuations and the onset of different phenological stages within a 100km radius around 10.000 point locations. We found that, across most of the Northern Hemisphere’s mid and high latitudes, a wide range of these stages, i.e., from the onset of ‘greenup’ to ‘greendown’, correlate strongly with temperature fluctuations during roughly the same period of the year, i.e., up until a few weeks before or after the onset of greenup. We found that warming rates during this period were roughly 1.5-2.5 times faster than regional mean annual temperature increases, which, in turn, were roughly 1.5-2.0 times faster than the increase in global mean annual temperature (which includes air above the oceans).When assessing the impact of global mean annual temperature changes on global land surface phenology, it is thus crucial to consider seasonal differences in warming. These differences are likely to affect not only plant phenology but also many other key processes related to plant growth and development.

The rapid worldwide formation and expansion of glacial lakes has increased the likelihood of glacial lake outburst floods, threatening lives and infrastructure, particularly in vulnerable mountain communities. Given the rapid increase in the popularity of artificial intelligence methods for remote sensing of glacial lakes, a comprehensive review is essential. We survey a decade (2015-2024) of research on glacial lake monitoring from space, with a focus on classical machine learning and deep learning approaches. We identify key trends, research gaps, and best practices for future studies. Most studies rely on optical imagery, especially Landsat-8 and Sentinel-2, while Sentinel-1 serves as a complementary radar source. However, monitoring glacial lakes in mountainous regions remains a challenge on cloudy days due to the limitations of radar and the unusability of optical data. Deep learning, particularly U-Net and DeepLab derivatives, dominates learning-based glacial lake studies but remains computationally demanding. Critical challenges involve balancing performance gains against trade-offs in data availability, computational cost, and model transferability. Geographic and methodological gaps, especially in regions experiencing rapid lake growth, underscore the need for broader spatial coverage and improved spatiotemporal model generalization. Moreover, transitioning from a focus on static seasonal mapping to frequent multi-temporal monitoring is beneficial for understanding glacial lake evolution and outburst flood hazards. Adapting emerging deep learning architectures to integrate multispectral, hyperspectral, and radar data could enhance glacial lake detection capabilities. Furthermore, thorough inter-method comparisons, benchmarking with rigorous evaluation metrics, and open-sourcing datasets and code would facilitate robust, large-scale glacial lake monitoring efforts.

Changing environmental conditions caused by climate change, eutrophication, and other anthropogenic factors affect the timing, duration, and surface extent of lake algae blooms across the globe. It remains, however, challenging to quantify the relative impacts of different environmental changes on the timing and characteristics of lake algae blooms, and to detect phenological trends over time, as these blooms vary considerably from year to year. Global data sets that may allow us to study algae-bloom properties along a wide range of environmental conditions and years are needed to address these challenges. For this study, we developed such a data set using satellite remote sensing. We analyze the phytoplankton phenology of 2025 lakes across a wide range of climate zones over a period of approximately 20 years. More specifically, we used daily lake chlorophyll estimates derived from MERIS and OLCI data to extract phenology metrics (e.g. the onset and decline of peaks in chlorophyll concentration) for individual pixels within each of the 2025 lakes. Through a newly developed method, we determined the timing of blooms, i.e. clusters of peaks in different pixels occurring within the same lake during the same period of the year, and, subsequently, studied the change in the timing, duration, and size of those blooms across years. This will, ultimately, help us to get a better overview of the extent to which lake algae blooms have changed across the globe, to attribute those changes to anthropogenic drivers, and to develop effective environmental policies to combat those changes where needed.

Turbidity is a key indicator of water quality and has significant impacts on underwater light availability of lakes. But the spatiotemporal variability of turbidity, which is important for understanding comprehensive changes in the water quality and status of aquatic ecosystems, remains unclear on a global scale. In this study, the spatial distribution pattern, seasonal variability, spatiotemporal variability, and influencing factors of turbidity in 774 lakes worldwide have been investigated using the turbidity product of Copernicus Global Land Service (CGLS) derived from Sentinel-3 OLCI. We found that 63.4% of lakes show low turbidity (≤ 5 Nephelometric Turbidity Units). The ranking of turbidity by climate zone is as follows: arid climate > tropical climate > temperate climate ∼ polar climate > cold climate. Turbidity decreased significantly in 40% of studied lakes, and increased significantly in 32% lakes. The lake with low turbidity has less seasonal variation, and there is a large seasonal variation in lake turbidity in the tropical and polar climate zones of Northern Hemisphere. Positive covariates to turbidity of global lakes include wind speed of lake, slope, surface runoff, and population in the catchment. Conversely, negative covariates include lake area, volume, discharge, inflow of lake, and GDP. Abundant water volume, favorable flow conditions, and more financial investments in lake management can help to reduce turbidity. These findings highlight the spatiotemporal changes of global lake turbidity and underlying mechanisms in controlling the variability, providing valuable insights for future lake water quality management.

Machine learning models have steadily improved in estimating inherent optical properties (IOPs) from remote sensing observations. Yet, their generalization ability when applied to new water bodies, beyond those they were trained on, is not well understood. We present a novel approach for assessing model generalization across various scenarios, including interpolation within in situ observation datasets, extrapolation beyond the training scope, and application to hyperspectral observations from the PRecursore IperSpettrale della Missione Applicativa (PRISMA) satellite involving atmospheric correction. We evaluate five probabilistic neural networks (PNNs), including novel architectures like recurrent neural networks, for their ability to estimate absorption at 443 and 675 nm from hyperspectral reflectance. The median symmetric accuracy (MdSA) worsens from >= 25% in interpolation scenarios to >= 50% in extrapolation scenarios, and reaches >= 80% when applied to PRISMA satellite imagery. Across all scenarios, models produce uncertainty estimates exceeding 40%, often reflecting systematic underconfidence. PNNs show better calibration during extrapolation, suggesting an intrinsic awareness of retrieval constraints. To address this miscalibration, we introduce an uncertainty recalibration method that only withholds 10% of the training dataset, but improves model calibration in 86% of PRISMA evaluations with minimal accuracy trade-offs. Resulting well-calibrated uncertainty estimates enable reliable uncertainty propagation for downstream applications. IOP retrieval uncertainty is predominantly aleatoric (inherent to the observations). Therefore, increasing the number of measurements from the same distribution or selecting a different neural network architecture trained on the same dataset does not enhance model accuracy. Our findings indicate that we have reached a predictability limit in retrieving IOPs using purely data-driven approaches. We therefore advocate embedding physical principles of IOPs into model architectures, creating physics-informed neural networks capable of surpassing current limitations.

Sun-induced fluorescence (SIF) from phytoplankton has historically been used as a proxy for chlorophyll-a (chl-a) concentration estimates in water bodies using optical earth observation data. However, the relationship is often affected by spectral features caused by elastic scattering, and by the shifting incidence of different fluorescence quenching mechanisms. This study found that disentangling photochemical quenching (PQ) and nonphotochemical quenching (NPQ) cases improves SIF-based chl-a estimates. Furthermore, we defined strategies that can distinguish the two quenching mechanisms. We assembled a unique dataset collected between 2018 and 2022 by an autonomous profiler in Lake Geneva (Western Europe). We used NPQ-influenced chl-a estimates from the fluorometer and NPQ-corrected chl-a estimates to distinguish between PQ and NPQ cases. The correlation between SIF yield and chl-a is weak when considering the entire dataset (R-2 = 0.37 and median absolute percentage difference (MAPD) = 74%). It increases strongly when comparing PQ (R-2 = 0.72 and MAPD = 49%) and NPQ cases (R-2 = 0.48 and MAPD = 68%) separately. Analyzing a subset of in situ measurements acquired around Sentinel-3 overpasses (+/- 3 h) improved the performance metrics for both PQ (R-2 = 0.82 and MAPD = 35%) and NPQ cases (R-2 = 0.43 and MAPD = 61%). However, when applying the same approach to Sentinel-3 Ocean and Land Color Instrument data, we found that the errors in remote sensing reflectance products disable such an adaptation. We conclude that enhanced atmospheric correction in the red-to-near-infrared region for oligo-mesotrophic lakes is needed to demonstrate the upscaling of our in-situ-based results. This will enhance satellite-based SIF yield retrievals and, subsequently, obtain SIF-related phytoplankton physiology products.

Accurate forecasting of algal blooms in lakes can support effective freshwater management. However, observational datasets for calibrating and validating algal bloom forecasting models such as the General Lake Model - Aquatic Eco Dynamics (GLM-AED) are often scarce, which impedes robust model calibration and forecasting ability. Satellite remote sensing can help fill these gaps by offering high-frequency, large-scale measurements of phytoplankton chlorophyll-a concentration (mg m-3), but satellite chl-a products often carry high uncertainty. Here we introduce a novel approach to quantify uncertainty in satellite chl-a based on conformal prediction, with the aim of integrating robust chlorophyll-a products into GLM-AED. Using Sentinel-2 imagery from two eutrophic lakes in the UK, Esthwaite Water and Loch Leven, we obtain remotely sensed chlorophyll-a with low systematic signed percentage bias (-1.22 % and 0.38) and moderate median symmetric accuracy (15.87 and 43.02 %) using Polymer atmospheric correction. We effectively flag potentially uncertain chlorophyll-a estimates (coverage factor: 75.6 - 81 %). Integrating the screened remotely sensed chlorophyll-a estimates improved GLM-AED algal bloom forecasts by 50 % in Loch Leven and 13 % in Esthwaite Water, with the greater improvement in Loch Leven attributed to its higher initial model errors. In contrast, incorporating unscreened chlorophyll-a estimates into GLM-AED increases validation errors on average by 32 %.Our findings show that process-based model predictions can substantially benefit from incorporating additional satellite-derived chlorophyll-a estimates. At the same time, they highlight a crucial need for robust uncertainty quantification to support downstream applications such as algorithm validation, biological monitoring in data-scarce regions, and water management decision-making.Moreover, because conformal prediction is model-agnostic and satellite-derived chlorophyll-a products are globally accessible, our study paves the way for large-scale, well-calibrated bloom forecasting through process-based models.

This study reports the outcomes of the third Atmospheric Correction Intercomparison Exercise (ACIX-III Aqua), which evaluated the performance of atmospheric correction (AC) methods for hyperspectral PRISMA satellite data over inland and coastal waters. The exercise included five AC processors (ACOLITE, hGRS, iCOR, MIP, and POLYMER), the standard PRISMA Level 2C product, and an adjacency correction tool (T-Mart) tested with ACOLITE. A total of 239 cloud-free PRISMA scenes from 2019 to 2024 were compared with in situ data of remote sensing reflectance, gathered from both hyperspectral and multispectral radiometers across eight distinct optical water types (OWTs). The accuracy of each AC method varied with spectral band, but all showed largest and lowest discrepancies with in situ data at 443 nm and 560 nm, respectively. All AC methods showed the best agreement with in situ data in greenish waters (OWT 4b) and highest uncertainties were yielded in humic-rich waters (OWT 7). Consistently with the previous ACIX-Aqua study focused on multispectral data, no single AC method outperformed the others across all OWTs. The study confirmed the ongoing challenges of AC over optically complex waters, yet the exercise allowed the community to advance in developing AC methods for hyperspectral satellite images and supporting the development of future operational hyperspectral missions, such as PRISMA Second Generation (PRISMA 2G) and CHIME.

Constructing a robust ocean color (OC) record(e.g., water transparency, phytoplankton absorption) for long-term assessments of coastal and inland water ecosystems from past, present, and future missions requires high-quality spectral remote sensing reflectance (R-rs) products. Using the GLORIA dataset (Lehmann et al., 2023), we evaluated the quality of R-rs products from the moderate resolution imaging spectroradiometer (MODIS on Terra and Aqua), medium resolution imaging spectrometer (MERIS), and visible infrared imaging radiometer suite (VIIRS) processed via the two-band heritage atmospheric correction method (a combination of near-infrared and shortwave infrared bands) available in the Sea WiFS Analysis Data Analysis System (SeaDAS). Overall, retrieval residuals are consistent within a few percentages among the four missions. Median residuals ranged from similar to 20% in the similar to 550-nm band to>60% in the similar to 412-nm bands. Spectrally averaged root mean squared differences for all the missions were similar to 0.0024 sr(-1)with one standard deviation of similar to 0.001 sr(-1). The corresponding(median) biases in the visible bands varied from-60% to-3%,with the largest biases identified in MERIS and VIIRS products. Despite the lower sensitivity of band-ratio algorithms to residuals in specific spectral regions [e.g., OC3 chlorophyll-a algorithm is less prone to residuals in R-rs(lambda>600 nm)], other algorithms or downstream products that leverage all the visible bands are highly compromised. We underscore the need to improve the quality of Rrsproducts, thereby enabling the reconstruction of baseline OC products of high caliber in global coastal and inland waters that are often near human activity.

Freshwater algae exhibit complex dynamics, particularly in meso-oligotrophic lakes with sudden and dramatic increases in algal biomass following long periods of low background concentration. While the fundamental prerequisites for algal blooms, namely light and nutrient availability, are well-known, their specific causation involves an intricate chain of conditions. Here we examine a recent massive Uroglena bloom in Lake Geneva (Switzerland/France). We show that a certain sequence of meteorological conditions triggered this specific algal bloom event: heavy rainfall promoting excessive organic matter and nutrients loading, followed by wind-induced coastal upwelling, and a prolonged period of warm, calm weather. The combination of satellite remote sensing, in-situ measurements, ad-hoc biogeochemical analyses, and three-dimensional modeling proved invaluable in unraveling the complex dynamics of algal blooms highlighting the substantial role of littoral-pelagic connectivities in large low-nutrient lakes. These findings underscore the advantages of state-of-the-art multidisciplinary approaches for an improved understanding of dynamic systems as a whole. A large algal bloom in Lake Geneva in 2021 was triggered by a sequence of heavy rainfall followed by wind-induced coastal upwelling, and a prolonged period of warm, calm weather, according to a combination of satellite remote sensing, in-situ measurements and three-dimensional numerical modeling.

Given the use of machine learning-based tools for monitoring the Water Quality Indicators (WQIs) over lakes and coastal waters, understanding the properties of such models, including the uncertainties inherent in their predictions is essential. This has led to the development of two probabilistic NN-algorithms: Mixture Density Network (MDN) and Bayesian Neural Network via Monte Carlo Dropout (BNN-MCD). These NNs are complex, featuring thousands of trainable parameters and modifiable hyper-parameters, and have been independently trained and tested. The model uncertainty metric captures the uncertainty present in each prediction based on the properties of the model—namely, the model architecture and the training data distribution. We conduct an analysis of MDN and BNN-MCD under near-identical conditions of model architecture, training, and test sets, etc., to retrieve the concentration of chlorophyll-a pigments (Chl a), total suspended solids (TSS), and the absorption by colored dissolved organic matter at 440 nm (acdom (440)). The spectral resolutions considered correspond to the Hyperspectral Imager for the Coastal Ocean (HICO), PRecursore IperSpettrale della Missione Applicativa (PRISMA), Ocean Colour and Land Imager (OLCI), and MultiSpectral Instrument (MSI). The model performances are tested in terms of both predictive residuals and predictive uncertainty metric quality. We also compared the simultaneous WQI retrievals against a single-parameter retrieval framework (for Chla). Ultimately, the models’ real-world applicability was investigated using a MSI satellite-matchup dataset N=3,053) of Chla and TSS. Experiments show that both models exhibit comparable estimation performance. Specifically, the median symmetric accuracy (MdSA) on the test set for the different parameters in both algorithms range from 30% to 60%. The uncertainty estimates, on the other hand, differ strongly. MDN’s uncertainty estimate is ∼50%, encompassing estimation residuals for 75% of test samples, whereas BNN-MCD’s average uncertainty estimate is ∼25%, encompassing the residuals for 50% of samples. Our analysis also revealed that simultaneous estimation results in improvements in both predictive performance and uncertainty metric quality. Interestingly, the trends mentioned above hold across different sensor resolutions, as well as experimental regimes. This disparity calls for additional research to determine whether such trends in model uncertainty are inherent to specific models or can be more broadly generalized across different algorithms and sensor setups.

Climate change exerts a profound impact on lakes, eliciting responses that range from gradual to abrupt transitions. When reaching critical tipping points, the established lake dynamics stand to undergo substantial modifications, setting off a chain reaction that reverberates through the entire ecosystem. This lake shift ripples into related ecosystem services and even influences the well-being of human communities. Despite the importance of lake shifts, we lack a systematic overview of their occurrence, mainly due to the lack of systematic data at the global scale. We reviewed the literature focusing on climate-related lake shifts and assessed how satellite Earth Observation (EO) has contributed to the research topic, and what we can unlock from this novel data. Our results show that EO data are used in only 9% of studies on lake shifts, although this fraction has increased since 2012. EO data is most commonly used to assess shifts in surface extent, ice coverage, or phytoplankton phenology. These variables are directly observable and the spatio-temporal resolution of EO satellites is of great advantage. But lake shifts can also be identified indirectly from EO data, as in the example of the vertical mixing of lake water, which can be described on the basis of surface patterns. In all possible applications, we expect increasing use of EO satellites in the future, including the development of early warning systems that promise to provide timely alerts regarding impending lake shifts, thus serving as a vanguard against abrupt alterations that could ripple through interconnected ecosystem services.

Protected agriculture boosts the production of vegetables, berries and fruits, and it plays a pivotal role in guaranteeing food security globally in the face of climate change. Remote sensing is proven to be useful for identifying the presence of (low-tech) plastic greenhouses and plastic mulches. However, the classification accuracy notoriously decreases in the presence of small-scale farming, heterogeneous land cover and unaccounted seasonal management of protected agriculture. Here, we present the random forest-based pixel-level Open field and Protected Agriculture land cover Classifier (OPAC) developed using Sentinel-2 L2A data. OPAC is trained using tiles from Switzerland over 2 years and the Almeria region in Spain over 1 acquisition day. OPAC classifies eight land covers typical of open field and protected agriculture (plastic mulches, low-tech greenhouses and for the first time high-tech greenhouses). Finally, we assess (1) how the land covers in OPAC are labelled in the Sentinel-2 Scene Classification Layer (SCL) and (2) the correspondence between pixels classified as protected agriculture by OPAC and by the best performing Advanced Plastic Greenhouse Index (APGI). To reduce anthropogenic land covers, we constrain the classification task to agricultural areas retrieved from cadastral data or the Corine Land Cover map. The 5-fold cross-validation reveals an overall accuracy of 92% but other classification scores are moderate when keeping the separation among the three classes of protected agriculture. However, all scores substantially improve upon grouping the three classes into one (with an Intersection Over Union of 0.58 as an average among the scores of the three classes and of 0.98 for one single class). Given the recently acknowledged importance of Sentinel-2 Band 1 (central wavelength of 443 nm), the classification accuracy of OPAC for the Swiss small-scale farming is mostly limited by the band's reduced spatial accuracy (60 m). A careful visual assessment indicates that OPAC achieves satisfactory generalization capabilities also in North European (the Netherlands) and four Mediterranean areas (Spain, Italy, Crete and Turkey) without the need of adding location and temporal specific information. There is good agreement among natural land covers classified by OPAC and the SCL. However, the SCL does not have a class for protected agriculture, the latter being often classified as clouds. APGI achieved similar to lower classification accuracies than OPAC. Importantly, the APGI classification task depends on a user-defined space- and time-specific threshold, whereas OPAC does not. Therefore, OPAC paves the way for rapid mapping of protected agriculture at continental scale.

Billions of tons of hazardous mine waste are stored in thousands of tailings storage facilities around the world. These impoundments represent one of the most important environmental risk factors of industrial mining, since occasional tailings spills or dam failures cause devastating impacts on humans and ecosystems, specifically along river corridors. In this study, we developed a satellite remote sensing methodology to assess the impacts of tailings spills on water quality focusing on the controversial incident that occurred at the Catoca diamond mine in Angola in late July 2021. The spill allegedly caused important river pollution in neighbouring Democratic Republic of the Congo (DR Congo) and led to public health concerns including the loss of human lives – however the mining company denied any responsibility. We processed high resolution imagery acquired by ESA’s Sentinel-2 satellites using the Python package Acolite for atmospheric correction and turbidity retrieval, and applied a river skeletonizing algorithm to automatically extract turbidity values for the entire river system. This allowed tracking the propagation of the pollution front from the source at the Catoca mine through the Tshikapa- and the Kasaï River during more than one month and across 1400 km, until the pollution front finally dissipated after discharging into the Congo River. We further analyzed a 6-year time series of virtual stations in the Tshikapa River located up- and downstream of the effluent discharge to compare the impacts of the tailings spill to seasonal variabilities of water quality. Turbidity values caused by the spill largely exceeded the seasonal variability in the Tshikapa River in recent years. These findings confirm that the Catoca tailings spill has significantly affected water quality of the Tshikapa- and the Kasaï River with total suspended solids concentrations that were several 10-fold above drinking water standards in Lunda Norte Province, Angola, and Kasaï Province, DR Congo, making severe public health impacts for residents and fish kills highly probable. After investigating whether this methodology could be applied to other tailings dam failures that have occurred since the Sentinel-2 mission began in 2015, we recommend to apply it to four other incidents in Mexico, Myanmar, Peru and China, respectively. Overall, this Sentinel-2 workflow provides the opportunity to assess the large-scale impacts of pollution incidents in mining areas around the world in locations where hydrological- and water quality data are scarce and monitoring capacities are limited.

Lakes are responding rapidly to climate change and one of the most tangible responses is the increase in lake surface water temperature. Such an increase can intensify thermal stratification and dampen the intensity of vertical mixing. In turn, surface warming has the potential to alter the mixing regime of lakes, potentially leading to abrupt shifts in ecosystem functioning. Reduced mixing between the surface and bottom waters can indeed decrease the upwelling of essential nutrients from deep water to the lake surface and the oxygen transport in the opposite direction. This can result in a decrease in lake productivity and can increase the risk of anoxia at depth, respectively.Despite the important consequences of such lake mixing anomalies, we lack a systematic overview of their occurrence, mainly due to the lack of systematic data to detect and analyze them worldwide. Remotely sensed lake surface water temperature available from ESA CCI (Climate Change Initiative) and similar sources represent spatial skin temperature gradients, but they do not resolve vertical gradients. They are hence often used to prove the lakes’ long-term warming in terms of spatial average. However, the horizontal gradients of such data could help us better understand the internal processes of lakes and the identification of lake mixing anomalies. Given that seasonal overturning often occurs at different times across the lake, the spatial character of remotely sensed data can reveal important processes in freshwater systems and can help assess the long-term variability in the overturning behaviour of large lakes in the context of climate change. Within our project, we use the spatial component of satellite Earth Observation data to reveal information about lake mixing and mixing anomalies. We apply a thermal front tracking method, a technique much more exploited in oceanography than limnology, to identify mixing anomalies in dimictic lakes worldwide.Our study suggests that the spatially distributed property of Earth Observation can be useful to spot lake mixing anomalies in dimictic lakes worldwide. Thus, we present the first global-scale assessment of lake mixing anomalies occurrence in the last 20 years. Earth Observation data can also be used to calculate how susceptible lakes are to undergo a mixing regime shift. Interestingly, we found that lakes experiencing more mixing anomalies are also those more susceptible to undergoing a mixing regime shift. Moreover, using Earth Observation, we detected mixing anomalies that have already been documented and, more interestingly, we spotted mixing anomalies occurring in unstudied remote lakes. Although further investigations would be needed to specifically assess the impact of climate change on these remote lakes, these cases highlight that remote sensing can be used as a first screening tool to spot lake mixing anomalies worldwide. Thus, Earth Observation and our methodology can be potentially used as an early warning system for lake mixing regime shifts.

The factors that govern the geographical distribution of nitrogen fixation are fundamental to providing accurate nitrogen budgets in aquatic environments. Model-based insights have demonstrated that regional hydrodynamics strongly impact nitrogen fixation. However, the mechanisms establishing this physical-biological coupling have yet to be constrained in field surveys. Here, we examine the distribution of nitrogen fixation in Lake Tanganyika - a model system with well-defined hydrodynamic regimes. We report that nitrogen fixation is five times higher under stratified than under upwelling conditions. Under stratified conditions, the limited resupply of inorganic nitrogen to surface waters, combined with greater light penetration, promotes the activity of bloom-forming photoautotrophic diazotrophs. In contrast, upwelling conditions support predominantly heterotrophic diazotrophs, which are uniquely suited to chemotactic foraging in a more dynamic nutrient landscape. We suggest that these hydrodynamic regimes (stratification versus mixing) play an important role in governing both the rates and the mode of nitrogen fixation.