In this work we show an application of the Sentinel 2 and Sentinel 3 satellite data for the implementation of a method that combines different Earth Observation data and in-situ measurements collected by stations, buoys and by a WaveGlider, for monitoring vulnerable coastal areas.
Due to the high temporal and spatial variability of marine parameters in coastal areas, a reliable coastal monitoring system needs to combine multiple sources of information to cope with the complexity of coastal phenomena of main interest and concern. In this work, selected biogeochemical and physical data derived from in situ and satellite sources are used to detect the conditions under which a potential damaging event for the environment is likely to occur, giving thus useful information to decision makers. The method can be particularly useful when applied to coastal areas of great environmental value, like Marine Protected Areas (MPA), especially if close to areas subject to strong anthropogenic pressures and recurrent issues, as for instance harmful algal blooms. This study describes an exploratory analysis of the method conducted in an area of the Tuscan coast (north west Mediterranean Sea), that was developed in the framework of Sympa (Satellite assets Integration for Marine Protected Areas), a demonstration project co-funded by the ESA Integrated Application Promotion (IAP) programme. In Sympa a system was set up to offer a technologically advanced support for the control of vulnerable coastal areas, in particular MPAs, based on a set of software applications for coastal environmental monitoring, built on Earth Observation (EO) observations, in situ and modelling data. Biogeochemical and physical data derived both from satellites and models are produced daily as an input to the Traffic Light service, a dedicated tool that gives information to MPAs managers about the risk level for the environment.
The project was demonstrated in some MPAs in the Mediterranean Sea. Many of the issues connected to satellite monitoring, as revisiting time and cloud coverage, can be dealt with by modelling: hydrodynamic and biogeochemical data from high-resolution models provide dynamic maps of physical and biogeochemical parameters. This is the reason for which a combined use of all those techniques including modelling, which will be described in a further work, was developed during the project. The modelling component of Sympa project is constituted by a chain of very high-resolution nested models, whose boundary and initial conditions are extracted from Copernicus Marine Service.
In this work, MODIS and VIIRS data were used for a first assessment of the method, in spite of their spatial resolution, too low for coastal observations, and of the atmospheric correction issues connected to coastal area characteristics. However, thanks to a strict level-2 flag masking applied to the coastal pixels, the selected data gave reliable observations, at least in terms of the observed space and time dynamics, also taking advantage of the almost daily overpass.
Here we show the results of the method implementation which was performed during the last period of the project. The method was then upgraded with the use of OLCI Sentinel 3 A and B data, together with MODIS and VIIRS ones, and also MSI Sentinel 2 data, as an additional exploratory test. MODIS and VIIRS data were always used because of Sentinel 3 and Sentinel 2 discontinuous time coverage, and in order to perform satellite vs satellite intercomparison. The dataset used for monitoring is thus composed by in situ data from meteorological coastal stations (namely wind speed, air temperature and precipitation), from buoys (sea surface temperature and wave height), and OLCI, MODIS and VIIRS satellite biogeochemical and physical products: trophic state assessed as chlorophyll surface concentration, water clarity assessed as Diffusion Attenuation Coefficient at 490 nm (KD490), and water temperature assessed as Sea Surface Temperature. The comparison of each parameter with specific thresholds defined from climatological analyses and scientific literature, specific for the target area, allowed to evaluate the conditions under which a potential critical event for the environment is likely to occur, with the aim of providing information profitable to the decision makers.
MSI Sentinel 2 sensor was developed for land observations, but, as it is turning out to be promising also for ocean observations, we included it in the method. Because of the high spatial resolution of the blue, green and red bands (10 m), this sensor is expected to perform well in complex coastal areas. We consequently tried to apply the NASA OCx algorithm and looked at the blue/green band ratio as a rough chlorophyll index (in fact chlorophyll ocean color algorithms calibrated on the study area are not available at the moment). Although we did not obtain validated chlorophyll maps, this allowed us to highlight coastal surface hydrodynamic patterns connected to coastal features (e.g. river mouths), and the presence of those patterns were confirmed by the in situ coastal oceanographic measurements achieved for the Sympa project. Such measurements were mainly from a waveglider, an unmanned surface vehicle powered by the wave motion and solar cells, capable to collect and transmit a mass of biogeochemical and physical ocean data, with no impact on the environment. Those preliminary results are encouraging also for proceeding with the development of MSI-based biogeochemical algorithms.

Area D - Board 371

Coastal ecosystems are exposed to significant environmental threats, including eutrophication, harmful algal bloom and habitat degradation, which can affect the whole marine communities and alter the ecosystem functioning and services. In this scenario, the European Union (EU) has set up the Marine Strategy Framework Directive (MSFD) where Member States are required to develop and implement adequate measures necessary to achieve or maintain “good environmental status” (GES) especially for transition waters, such as coastal waters and estuaries.
To this aim, Ocean colour (OC) remote sensing data offer potential advantages, providing near real-time and synoptic information as well as long-term archive data about optical in-water constituents such as Coloured Dissolved Organic Matter (CDOM), Non Algal Particulate (NAP) and phytoplankton. Among these, phytoplankton represents a key parameter to ensure the sustainable exploitation of food chain and face marine environmental threats, considering its strict link with primary production. Satellite retrievals of chlorophyll-a (chl-a) concentration, commonly used as a proxy of phytoplankton, can be easily assimilated into marine forecasting models for detecting harmful algal blooms and investigating the ongoing eutrophication phenomena. In this framework, its efficient exploitation requires a detailed analysis of possible error sources associated with retrieval procedures, as well as the selection of the most suitable chl-a algorithm for a specific study area.
The empirical and semi-analytical chl-a algorithms are the mostly used, mainly because of their easy implementation. The standard chlorophyll-a (chl-a) algorithms based on Moderate Resolution Imaging Spectro-radiometer (MODIS) data aboard the Aqua satellite have been developed for global scale studies, thus showing poor accuracy at basin/regional level.
In this study, we evaluated the most suitable chl-a algorithm for the Basilicata Ionian Coastal Waters (BICW), a sub-set of the Gulf of Taranto (Ionian Sea, South of Italy), by comparing the performance of three empirical and semi-analytical MODIS-Aqua chl-a algorithms, namely Med-OC3, GSM and GIOP. The outputs of one empirical (Med-OC3) and two semi-analytical algorithms, the Garver–Siegel–Maritorena (GSM) and the Generalized Inherent Optical Properties (GIOP) model, have been compared with ground measurements acquired during three different measurement campaigns.
The achieved results proved the poor accuracy (adjusted R2 value of 0.12) of the investigated empirical algorithm and conversely the good performance of the semi-analytical ones (adjusted R2 ranging from 0.74 to 0.79). The co-existence of Coloured Dissolved Organic Matter (CDOM) and Non Algal Particles (NAP) has probably determined large errors in the reflectance ratios used in the OCx form algorithms.
Finally, a local scale assessment of the bio-optical properties, based on the in situ dataset, allowed defining an operational local scale-tuned version of MODIS chl-a algorithm, namely GSM-BICW.
The regression analysis of this algorithm showed the best regression index (adjusted R2=0.86) and both the lowest average absolute percent difference (APD) and root mean square error (RMSE) (11.03% and 0.070 respectively), thus recording a slight improvement in accuracy compared to the GSM-standard algorithm.
The results achieved in this work further confirm the need to regionalize the standard MODIS chl-a algorithms especially in optically-complex or poorly investigated area (i.e. BICW) in order to ensure their adequate exploitation within coastal management systems.

Area D - Board 372

Optical remote sensing (spaceborne, airborne) is a key technology for monitoring environmental systems, and very promising for inland waters. However, the complexity of inland waters and the absence of an ideal spaceborne sensor imposes challenges to existing remote sensing tools. Therefore, in-situ measurements and ground based optical measurements are necessary to validate remotely sensed data. A promising tool for ground truthing of satellite data and monitoring are wireless sensor networks utilizing miniaturized spectrometers. Here we present first results with a low-cost miniature spectrometer (Hamamatsu C12880MA) obtained in the laboratory and in the field, discuss potential and challenges as well as prototype design.

Area D - Board 373

A new autonomous above water radiometer system (WISPstation) was developed by Water Insight based on the experience with their handheld WISP-3 system. With an extended wavelength range (350-1100 nm) and a design tailored to minimise glint (2 viewing directions) the instrument is thought to play a significant role in autonomous water quality monitoring and satellite validation. By design the instrument (WISPstation) is relatively insensitive to calibration drift. In various Copernicus projects (Tapas, EOMORES, CoastObs and Monocle) the WISPstation is being tested in highly different water types and environmental conditions, ranging from case-1 in Mediterranean coastal waters to turbid waters with cyanobacteria proliferation in lakes and lagoons. In view of its initial scientific application, the system is programmed to produce high frequency data with high redundancy to calculate variability in the water system and in the observations. The WISPstation results are stored in the on-line database WISPcloud allowing users to extract data for analysis. We will present spectral results, time series analysis and satellite match-up validation results for various locations. Projects like EOMORES currently investigate the conditions under which water managers could adopt optical monitoring as a viable addition to their regular monitoring programs. For various locations time series of water quality parameters will be shown to demonstrate the capability of operational high frequency monitoring of Chlorophyll-a, Suspended Matter and Phycocyanin.

Area D - Board 374

Sun-induced fluorescence (SIF) from phytoplankton is recognized for its good correlation with chlorophyll-a concentration (chl-a) and has been a widely used parameter in in-situ and laboratory measurements, particularly in ecological studies. SIF therefore provides a good proxy for chl-a estimation and has been applied in ocean colour remote sensing by using absorption regions (i.e. oxygen B-band at 687 nm) where fluorescence emission is high. However, SIF retrieval from optically complex waters is difficult because of the presence of other in-water constituents. The narrow absorption features also necessitate high spectral resolution to lessen inaccuracies in obtaining SIF emission. The upcoming Fluorescence Explorer (FLEX) Mission from the European Space Agency (ESA) provides hyperspectral capability between 500-780nm and can possibly be used to advance SIF retrievals and therefore water quality monitoring of coastal and inland waters.

This study aims to investigate different retrieval techniques for phytoplankton fluorescence in optically complex waters and its potential in distinguishing different algal groups. We use a unique data set obtained between June – August 2018 in the Tyrrhenian Sea (Italy) and Greifensee (Switzerland) comprising concurrent measurements with in situ (Fluorescence Box, FloX), airborne (HyPlant) and spaceborne (Sentinel-3A,-3B) spectrometers. Both FloX and HyPlant are hyperspectral instruments with specifications comparable to the fluorescence module onboard FLEX. We apply several retrieval methods including the fluorescence line height method, a differential absorption spectroscopy approach, and an inversion of a bio-optical model. We quantitatively assess the suitability of the retrieval methods to disentangle reflected from emitted fluorescence light and to compensate for atmospheric disturbances. We discuss possibilities and potential limitations of the individual retrieval strategies.

Results generated from this study will guide the development of novel and robust SIF retrieval approaches for inland and coastal waters. The availability of more efficient techniques in obtaining SIF will be beneficial for water quality monitoring (e.g. harmful algal bloom detection) and will facilitate ecological studies (e.g. identification of phytoplankton functional types, primary production assessments). Finally, this study could further demonstrate the significance of the FLEX data to aquatic research and its many applications.

Area D - Board 375

Hyperspectral optical information can significantly contribute to the monitoring of inland and coastal water ecosystems. The fine spectral resolution and sampling of past (e.g. HICO), current (e.g. DESIS, OMI, TROPOMI) and future hyperspectral sensors (e.g. EnMAP, PACE) improves current water colour remote sensing retrievals, going beyond standard algorithms/retrievals (e.g. band ratio/chlorophyll-a concentration) to the distinction of different phytoplankton functional groups and water optical constituents. An important and challenging step to obtain accurate retrievals over water is the atmospheric correction (AC). In this study, we evaluated the performance of two AC algorithms: Polymer (v.4.8, Steinmetz et al. 2011) and the l2gen (level 2 generator), which is distributed in the SeaWiFS Data Analysis System software package (SeaDAS v.7.5.1, 2018). Polymer is an AC algorithm based on an optimization technique in which atmospheric and oceanic signals are obtained simultaneously. l2gen uses the so-called standard atmospheric correction scheme in which the aerosol signal is estimated in the red and near infrared bands and extrapolated to the visible bands and it has recently been adapted and improved for processing hyperspectral data (Ibrahim et al. 2018). These algorithms were applied to Hyperspectral Imager for the Coastal Ocean (HICO) data. We compared the methods and validated the results against multispectral radiometric measurements at AERosol Robotic NETwork - Ocean Colour stations (AERONET-OC). We also compared the results to data from multispectral sensors as MERIS and MODIS. In addition, we used coupled atmosphere-ocean radiation transfer simulations of top-of-atmosphere and water-leaving-radiance for different water types (Hieronymi et al. 2017) and atmospheric conditions to assess the AC algorithms applied on hyperspectral data. We present and discuss results, potential applications and challenging aspects of hyperspectral missions for the investigation of aquatic ecosystems.

References

Ibrahim et al. 2018. Atmospheric correction for hyperspectral ocean color retrieval with application to the Hyperspectral Imager for the Coastal Ocean (HICO). Remote Sensing of Environment 204 (2018) 60–75.

Hieronymi, M., Müller, D., Doerffer, R. The OLCI Neural Network Swarm (ONNS): A Bio-Geo-Optical Algorithm for Open Ocean and Coastal Waters. Front. Mar. Sci., 11 May 2017.

SeaWiFS Data Analysis System (SeaDAS), 2018. https://seadas.gsfc.nasa.gov/.

Steinmetz et al. 2011. Atmospheric correction on the presence of sun glint: application to MERIS, Optics Express, 19, 9783-9800.

Area D - Board 376

With the recent development in technology and wide coverage of the earth through numerous satellites, the monitoring of the natural resources has been easier than ever before. Remote sensing offers numerous benefits that range from regional scale monitoring to accessibility to historical archival data at any point in time. The frequent field-based studies are costly, time consuming and are impractical in topographically challenged areas. The use of the remote sensing data to monitor inland and coastal water qualities would provide periodical reporting of the water resources using freely available earth observation (EO) data. Through Remote Sensing of Irish Surface Water (INFER) project, we are validating the empirical, semi-empirical and semi-analytical algorithms for inland water bodies in Ireland. We are utilizing Sentinel 2 MultiSpectral Instrument (MSI), which comprises of two European Space Agency (ESA) terrestrial satellites with combined temporal resolution of 5 days. The algorithms for coloured dissolved organic matter (CDOM), chlorophyll-a (Chl-a) and total suspended sediment (TSS) will be validated for inland, coastal and transitional waters using the historical archive of water quality parameters measured in those waters. Depending upon the performance of the semi-empirical and semi-analytical algorithms, additional empirical algorithms will be developed for specific water bodies. The investigation has several challenges such as adjacency problems due to relatively small size of the water bodies and difficulty to find cloud free scenes in the areas where rainfall is more frequent. The end product of this project is the web platform with the access to Sentinel 2 MSI data where users can compute and compare the CDOM, Chl-a and TSS values for different water bodies. This platform will promote the use of EO data for inland water monitoring and would enable sustainable utilization of the water resources. Although, the project focuses on inland and coastal waters of Ireland, the technology is easily replicable in similar places worldwide where freely available remote sensing resources can be utilized to investigate the water resources at a regional scale.

Area D - Board 377

In the last decades monitoring of inland waters has become a major topic in terms of water quality and environmental issues. For this monitoring a lot of in situ measured data is required. In this study we rely on hyperspectral data which has been measured continuously at several inland waters. The objective is to determine chlorophyll a concentration only based on hyperspectral data. Chlorophyll a concentration serves as a proxy for nutrition supply and water quality. During a measurement campaign in the summer of 2018, we collected hyperspectral data and in situ probes of 14 inland waters under real-world conditions in the Karlsruhe region, Germany. The hyperspectral data was measured by a spectrometer mounted on a tripod. In summary, the dataset contains 433 datapoints. Each datapoint consists of hyperspectral data as input data and a chlorophyll a concentration as target value. In a pre-processing step, the hyperspectral data is aggregated respectively to a 4 nm and an 8 nm spectral resolution to investigate the impact of the spectral resolution on the regression performance. The spectral range covers the wavelength between 400 nm and 896 nm. The whole dataset is split into a training and a test subset consisting of 50% of the samples, as a preparation to apply any machine learning model.
Previous studies use different wavelength of spectral data or ratio-based approaches to estimate the chlorophyll a concentration. Maier & Keller (2018) and Keller et al. (2018) have applied supervised machine learning models to estimate several water parameters including the chlorophyll a concentration of the river Elbe successfully. In addition to these two studies, we focus now on a generalization of these machine learning models to estimate the chlorophyll a concentration of 14 different inland water bodies.
The supervised regression of the chlorophyll a concentration of the different water bodies is performed by three different machine learning methods: Support Vector Machine (SVM), Random Forest (RF) and k-Nearest Neighbor (k-NN). No further preprocessing e.g. scaling of the data or dimensionality reduction is applied. The three models are trained on the training subset and are validated with the test subset to avoid bias issues. The regression performance is expressed as the coefficient of determination (R²).
Regarding the regression results for the hyperspectral input data with 4 nm spectral resolution, the SVM model achieves the best result with an R² of about 90%. The RF model performs well with an R² of 82%. The k-NN model achieves only an R² of 76%. The regression results on the dataset with hyperspectral input data of 8 nm spectral resolution resemble the regression performance on the 4 nm dataset. To conclude, the supervised regression models are able to estimate the chlorophyll a concentration of 14 different inland water bodies satisfactorily without any pre-information to which inland water the datapoints belong to. Furthermore, a coarser spectral resolution of the hyperspectral input data has no noticeable impact on the regression results. The latter plays a key role when it comes to applying UAV-measured or satellite spectral data as input data for the machine learning models to estimate the chlorophyll a concentration of a wider area.
In future studies, we will address the upscaling approach. A first step towards this is to aggregate the presented dataset to Sentinel-2 bands with the respective band width. One advantage of the use of satellite data is the continuous repetition rate of the recordings. Due to this automatism, the monitoring of inland waters with satellite data and machine learning models can be an additional way to the expensive in situ measurements. Currently, satellite data is characterized by multi-spectral resolutions, but upcoming satellite mission such as DESIS and EnMAP will be able to provide hyperspectral data.

References:
Maier, P. M., & Keller, S. (2018). Machine learning regression on hyperspectral data to estimate multiple water parameters. arXiv preprint arXiv:1805.01361.

Keller, S., Maier, P. M., Riese, F. M., Norra, S., Holbach, A., Börsig, N., Wilhelms, A. Moladaenke, C. Zaake, A. and Hinz, S. (2018). Hyperspectral Data and Machine Learning for Estimating CDOM, Chlorophyll a, Diatoms, Green Algae and Turbidity. In International journal of environmental research and public health, 15(9), 1881.

Area D - Board 378

Low levels of pelagic productivity characterize a substantial number of inland waters. As such, they are important sources for water supply and indicators of relatively unimpaired ecological conditions. Despite their importance, only a very small fraction are systematically monitored with most of the monitoring effort focusing on eutrophic systems. In recent years, remote sensing measures of phytoplankton chlorophyll-a (chl-a) evolved to complement in-situ measurement efforts offering spatial coverage and temporal revisit advantages. Nonetheless, the use of optical remote sensing in inland waters still faces several challenges, one of them being the accuracy and reliability of the constituent retrieval algorithms in inland waters with low phytoplankton biomass. From an optical point of view, phytoplankton are efficient light absorbers, but generally only make modest contributions to the total backscattering coefficient. As a result, remote sensing algorithms focused on the retrieval of chl-a concentrations generally rely on the absorption signal. However, in low biomass waters often non-phytoplankton materials, namely coloured dissolved organic matter (CDOM) and non-algal particles (NAP), compete strongly with phytoplankton to the total light absorption coefficient. The inherent optical ambiguity leads to errors in the retrieval and is often unaccounted for in present chl-a retrieval schemes warranting the exploration of new algorithmic approaches. There is an urgent need to develop, advance and validate algorithms for chl-a retrieval purposes to refine our quantitative understanding of the water quality status in these environments. In this study we present new and revised algorithms to improve the accuracy of chl-a retrievals (within the context of CDOM and suspended particles) in waters with low phytoplankton biomass. The presented algorithms are based on a large set of field data collected to adequately represent these water types. The datasets are enhanced by hyperspectral radiative transfer simulations to determine optimal configurations of said algorithms in the context of current and future satellites to ultimately assess which (future) satellite configurations offer essential ‘breakthrough’ capabilities to monitor these water bodies.

Area D - Board 379

Water quality is one of the major issues that societies are already facing with impacts on human health, food production, ecosystem functions, biodiversity and economic growth. Climate change and the increase in population and in water demand will get the situation even worse in the coming years. In that context, there is a strong need of a global monitoring on inland water quality and satellite data can help to answer this request.
We develop a specific inland water color processing chain for the Sentinel-2 images which resolution makes possible to assess very small objects of a few tenths of meters, such as rivers and minor lakes. The processing chain handles Sentinel-2 Level-2 products and a specific atmospheric correction for inland water integrating sunglint correction has been validated and first tests are conclusive to fully process Level-1C images. The final goal is to produce water quality products operationally (time series and images in near real time) such as water transparency, suspended particulate matter (SPM) concentration, chlorophyll-a content and colored dissolved organic matter (CDOM). A large database of water quality and field radiometric measurements (i.e. > 1000 points) collected in river and lakes is used to evaluate the performances of the processing chain. The first water color products from Sentinel-2 images are distributed through THEIA the French land datacenter https://theia.cnes.fr/ for several sites located in France and across the tropics.
Coarse spatial-resolution sensors such as MODIS, VIIRS (Visible Infrared Imaging Radiometer Suite) and Sentinel-3 will be progressively combined with high resolution sensors (Sentinel 2, Landsat) to develop a multi-mission water color product in order to improve the monitoring frequency.
We will explain the method, calibration of the model and its validation over various sites worldwide and present the freely distributed water color products.
Another major challenge is the assessment of river sediment discharge essential for hydropower impoundment management and quantification of the geochemical fluxes from continents to oceans. It is possible to retrieve river sediment discharge by coupling river discharge estimated by hydrological model and water height from altimetry with SPM load assessed from optical images. We will present results over the Amazon basin validated with conventional measurements. This work shall prepare the arrival of the future SWOT data (Surface Water and Ocean Topography). The SWOT mission which launch is foreseen in September 2021 will provide global measurements of water surface elevation at unparalleled spatial resolutions making possible to retrieve suspended sediment load over all river wider than 100m (with a target at 50m).

Area D - Board 380

Monitoring river solid transport plays a key role in the identification of changes in sediment supply, especially under climate change scenarios. Sediments are an essential component of the ecosystem, they provide ecosystem services, namely the benefits that people obtain from ecosystems, like sediment provision, water quality regulation, habitat provision raw abiotic material and carbon sequestration. Measurement of changes in the volume of sediments flowing in rivers and entering water reservoirs is indispensable to understand the impact on water reservoirs function and the ecosystem services provided.
Inland water monitoring using Remote Sensing may complement in-situ measurements supplying continuous spatially explicit representation of parameters related to water quality and solid transport. Total Suspended Matter (TSM) concentration is a proxy for solid transport, which is result of the sediment export, and can be estimated from optical satellite sensors and used to analyze spatio-temporal dynamics of sediments. Sentinel-2 and Landsat 8 satellites, equipped with MSI and OLI optical sensors whose spectral bands allow to perform a more accurate atmospheric correction, enabled the development of methodologies for monitoring river water color from space thanks to an unprecedent combination of high spatial resolution and short revisit time.
This research study aims to analyze inland water constituents, estimated for main rivers basins in northern Italy from optical remote sensing data, in order identify the spatio-temporal dynamics of suspended sediments. A processing chain, developed in order to systematically perform atmospheric correction over water and retrieve water constituents from Sentinel-2 MSI and Landsat-8 OLI data, was used to process satellite images with both the C2RCC and the ACOLITE algorithms. Later multi-temporal statistics of the retrieved water constituents was done in order to compute the spatial and temporal variability. Finally, a dataset of simultaneous acquisitions of Sentinel-2 MSI and Landsat-8 OLI images was used to compare the performances of C2RCC and ACOLITE algorithms and evaluate the suitability of the synergistic use of optical data acquired by multiple sensors for the monitoring of river color from space.
Results demonstrate that further development of the algorithms is needed in order to foresee an operational monitoring of water quality for riverine systems. In example, sun-glint significantly affects Sentinel-2 MSI and requires a proper correction algorithm in order to get more accurate estimates of water constituents concentrations. This research study demonstrates the potential synergic use of recently available remote sensing data at high resolution and dense temporal series for the analysis of sediment dynamics in rivers.

Area D - Board 381

Lakes and rivers in cities are an important natural environment which are also heavily used by human activities. They are of great economic (e.g. harbours), ecologic (e.g. urban climate, habitat) and touristic (sailing, swimming, recreation) importance. Some cities have large water areas, and these even determine the character of the city (like Hamburg or Stockholm). Thus, monitoring of urban water bodies is relevant for many different aspects and they also fall under the European Directives such as the Water Framework Directive or the Bathing Water Directive. It is in the responsibility of authorities to warn the people if any harm is caused by the water, e.g. by harmful cyanobacteria blooms.
The interest in using Earth Observation for inland water quality monitoring has been growing during the recent years and many activities are ongoing in algorithm development for atmospheric correction and in-water retrieval for inland waters and services are set-up. The challenges for water quality retrieval for inland waters from EO are the large variety of different water types and concentration ranges, as well as the strong influence or adjacency effect due to the surrounding land. In cities, remote sensing is even more difficult due to the spatial dimensions, more diverse influences of the surrounding and dealing with shadow from high buildings, strong adjacency effect by bright surfaces or absorbing aerosols in the atmosphere. Water bodies are often relatively small and used by numerous boats (recreation, sports) and larger ships (harbours, rivers). We are investigating the impact of these influences and develop for solutions where possible, e.g for advanced pixel classification for shadows and ship detection, improved atmospheric correction for absorbing aerosols or detection of different algal categories. In-situ measurements from the monitoring stations nets as well as dedicated measurements are planned for algorithm development and validation. We will showcase this for the water bodies of Hamburg and the goal is to introduce the remote sensing techniques into the measurement concept of the city.

Area D - Board 382

Water quality monitoring is a topic that has garnered more and more importance in the past few years, given the increase eutrophication faced by many inland waters, as well as the increased occurrences of harmful algae blooms. Next to the development of reliable atmospheric correction and water constituents algorithms, a correct pixel identification is paramount, and most especially the correct detection (and removal) of clouds, as well as cloud and mountain shadows. There is still a lot of work to be done in this domain, especially considering the expanding use of high-resolution (HR) sensors such as Copernicus Sentinel-2 for inland water quality monitoring. The error in chlorophyll retrieval for unmasked semitransparent clouds or in cloud shadow conditions can reach 50%. Cloud shadows are most relevant for HR sensors, while for medium resolution sensors such as Sentinel-3 OLCI, the percentage of pixels impacted by a cloud or a mountain shadow is smaller due to geometrical conditions, but still necessary to be identified and either masked or corrected for.

Besides clouds, other conditions such as floating vegetation (sub-pixel) or bottom reflection lead to errors in the atmospheric correction and water retrieval if not detected and treated properly. Based on experience, operators for pixel identification in general, and cloud detection in particular, should combine several physically based as well as statistical and geometrical methods to be optimal for a given sensor and to offer users a more defined classification. For example, spectral tests (decision tree based on band(s) thresholds) can be used with neural networks and both methods can be individually parameterized and applied to medium and high-resolution sensors. Shadow detection methods require top height and base information which often must be assumed to assess a potential shadow area. This potential shadow area can then be merged with a geometric test to precise the shadow area. The IdePix operator in the SNAP toolbox is a multi-sensor pixel identification tool where we collect the result of many years of experience of pixel identification for water (and land) applications.

This presentation will focus on the Sentinel-2 cloud and cloud/mountain shadow detection algorithms implemented in the SNAP IdePix Operator, as a showcase of the currently available techniques. We will detail the methods and show the impact on the retrieved surface reflectances as well as chlorophyll, TSM and CDOM products in different conditions. Single image examples as well as the impact on time series and temporal composites will be shown. The IdePix operator can be easily implemented in an operational processing environment, since the methods used are mono-temporal, thus only requiring the currently processed product, and have been optimized for fast processing.

Area D - Board 383

Regular observations of water quality in lakes provide essential information for resource management. To the aim, integration of satellite remote sensing with in situ monitoring from optical sensors might provide multi scale and multi-frequency data. The satellites Sentinel-3 (S3) with OLCI and Sentinel-2 (S2) with MSI instruments, both of them having a twin configuration, offer a unique opportunity to monitor inland waters phytoplankton phenology with high frequency and fine spatial scale. Then, as the dynamics of phytoplankton growth might have significant variations during the day, satellite observation might be further integrated with hyperspectral observations gathered from continuous measurement sensors, which also provide valuable data for validating satellite products.
This study presents the application of such integrated approach, as part of the European H2020 EOMORES project, developed in Lake Trasimeno, a turbid shallow lake located in central Italy. The aims are: i) the evaluation of suitable processing chains for chlorophyll-a concentrations (chl-a) from S3 and S2; ii) the application of validated satellite-derived chl-a products for lake monitoring, and iii) a time series analyses that from S2 and S3 goes back to MERIS lifetime. More in detail, S3 and S2 are first converted into remote sensing reflectance with Polymer and a 6SV-based codes. Then, chl-a concentrations are obtained applying a bio-optical model inversion implemented in BOMBER tool. The satellite-products are evaluated based on field data and then analysed to capture phytoplankton phenology. The consistency between OLCI- and MSI-derived products was analysed from synchronous pairs of S2 and S3 imagery. Satellite data analysis from 2015 to 2018 was also integrated with in situ observations acquired from a WISPStation (an operational station based on a WISP-3 spectroradiometer) that, since April 2018, is autonomously gathering and serving data to evaluate inter- and intra- day changing of remote sensing reflectance. Finally, previously validated MERIS-derived chl-a maps, were exploited to enlarge the time series back to 2003.
The results show that despite the atmospheric correction overestimated Rrs spectra with respect to field data, good accuracy was obtained in terms of the shape of spectra, with χ2 ranges of 0.03-0.2 for MSI, and of 0.008-0.07 for OLCI. Validation of chl-a products showed a mean average error of 1.16 and 2.5 mg/m3 for MSI and OLCI, respectively. The consistency analysis conducted across the products obtained by the pairs of MSI and OLCI showed a good agreement between OLCI and MSI (r2 0.87). A slight overestimation was detected in some cases from MSI respect to OLCI products.
Seasonal trend shows a highest peak in late summer-autumn, when usually cyanobacteria blooms occur in the lake. Intensity of blooms varied during years, with lower values in 2003, 2006, and 2007. Phenology showed strong differences in both starting date of phytoplankton onset and on maximum peak date. In particular the bloom reached its maximum on the earliest date in 2005, and on the latest in 2006. Duration of the bloom ranged between 3.36 days in 2003 and 29.26 in 2017. A seasonal Kendall’s test with a median annual slope of 0.173 mg m-3 per year was also observed as increasing trend. Seasonal variability was also observed by data acquired by the WISPStation in 2018. These hyperspectral data finally showed that the typical peaks/dips due to phytoplankton at 670-700 nm wavelengths of remote sensing reflectance, appeared behind middle of July. When ingested into the bio-optical modelling inversion, the WISPStation data allowed to estimate that chl-a can significantly change during the day (e.g. up to 200% as on 6 August 2018 when chl-a ranged from 5 to 37 mg/m3).

Area D - Board 384

Investigation of the dynamics of submerged aquatic vegetation (SAV) in lakes is a challenging topic due to the complex processes involved in driving SAV responses to internal and external determinants. SAV acts as ecosystem engineering, able to regulate deeply the colonized habitats; however, they it is also extremely sensitive to human perturbations, especially to eutrophication. In view of the significant role played by SAV in freshwater ecosystems, understanding and quantifying the environmental factors that influence the its distribution patterns is indispensable for integrated management practices of the ecosystems.
Our main objectives were to map the distribution of SAV in the southern part of the Lake Iseo (a deep subalpine lake in Northern Italy) in the period 2015-2017, and to evaluate its spatiotemporal variations and the possible factors of change. Specifically, we focused our attention to the bathymetric layer between 0 and 10 m of depth dominated by Vallisneria spiralis, Najas marina and Ceratophyllum demersum.
To the aim 17 Sentinel-2A/B images, at 10m spatial resolution, were atmospherically corrected through the radiative transfer code 6SV to produce the Remote Sensing Reflectance (RRS) that were then used as input of the code BOMBER (Bio-Optical Model-Based tool for Estimating water quality and bottom properties from Remote sensing images) for estimating the substrate type and coverage. The parameterisation of the bio-optical model implemented in BOMBER was based on a comprehensive site-specific dataset of concentrations of optical active parameters, specific inherent optical properties and bottom albedo spectra (i.e. three spectra describing: SAV with high albedo, SAV with low albedo and sand). The results of the image-processing chain, both in terms of atmospheric correction and bio-optical model inversion, were compared to in situ measurements of RRS and substrate coverage classes observed during field campaigns performed synchronous to satellite overpasses (or with a maximum of 1-day time difference). Root Mean Square Error (RMSE) for atmospheric correction products showed good agreement between in situ and satellite RRS (the RMSE values in the visible region were between 0.002 and 0.003 sr-1, with lower values in the near-infrared band). Analogously, the confusion matrix, depicting the agreement of satellite-derived substrate coverage classes with respect to in situ surveys, show a good accuracy too.
The results show that the southern part of Lake Iseo is characterized by significant coverage of SAV: between 0 and 5 meters deep there are the major coverings, which decrease with increasing depth, moreover the spatial analysis has shown that close to the coast there are mainly sparse macrophytes, while dense macrophytes are usually located between 10 and 50 meters from the coast line.
The temporal analysis showed a decrease in the area of SAV between 2015 and 2017, equal to 19% of the rate of coverage of the bottom. Decrease that is also confirmed by the presence of floating macrophyte stands on the surface of water.
In order to understand the potential reasons of this changes, we have produced water quality maps (Chlorophyll-a, and Total Suspended Matter) from Sentinel-3 data. Sentinel-3 images were atmospherically corrected with polymer and to the obtained reflectance products was applied the bio-optical BOMBER in the deep-water mode. The analysis showed that prevalent causes of the decrease in SAV distribution could be water turbidity and the effect of the wind.
All the results have been shared with the regional environmental protection agency (ARPA Lombardia) to supplement the management plan of the Lake Iseo. The activities are included in the ISEO project funded by the CARIPLO Foundation and the H2020 EOMORES project.

Area D - Board 385

In the frames of Copernicus mission, it is possible to retrieve information from optically complex inland waters from specifically designed satellite Sentinel-3A (S3A), which was accompanied with Sentinel-3B in 2018. Those satellites allow daily overview of sea areas and inland water bodies with sufficient size, but since S3A/OLCI pixel size is 300 m, it is not suitable for monitoring of very small lakes. Our test site is L. Ülemiste, which surface area (943.6 ha) is large enough for S3 usage. L. Ülemiste is a shallow eutrophic water body (max depth 4.2 m), with cyanobacterial blooms present in summer. Due to problems with excess phytoplankton, various biomanipulation activities took place on the lake from 2002 until 2009, including selective fish removal (mainly bream) during winter time together with introduction of predatory fish larvae. This resulted in increase in transparency and reduction of Chl a during 2007-2009, but expected long-term domination of large zooplankton did not occur. Phytoplankton community shifted from filamentous cyanobacteria (Limnothrix redekei as dominant species) to dominance of colonial cyanobacteria including various Microcystis species, which may lead to potential problems with cyanobacterial toxins.
Lake Ülemiste is the main water supply for Estonian capital Tallinn, giving drinking water to 90% of inhabitants. This is the main reason, why this waterbody is under frequent monitoring, allowing to get information about in situ measured Chl a much more often than from other lakes. Regular monitoring activities in Estonian small lakes take place at most 4 times per vegetation period, whereas in L. Ülemiste the frequency is once a week. This gives the possibility to compare in situ measured Chl a with the same day Chl a values retrieved from the satellite and expand results to other shallow eutrophic lakes with less available in situ measurements. Sentinel-3A/OLCI L1 and L2 FR images were downloaded from the Copernicus Online Data Access (CODA), SNAP 5.0 was used for image processing. Different algorithms using red and NIR wavelengths (e.g. 2 band algorithm (2BDA), 3 band algorithm (3BDA), 4 band algorithm, NDCI, MCI and FLH) were calculated from L1 data (top of the atmosphere radiances) to get the most suitable algorithm for Chl a retrieval. 3BDA gave most accurate result (R²=0.76). 4 band algorithm was originally developed for very turbid lakes (e.g. Taihu) and did not suit for L. Ülemiste. 2BDA and NDCI gave similar results (R² around 0.65). Standard L2 calculation scheme - Chl a retrieval via neural network was also studied. Developed algorithms can be used for characterizing seasonal dynamics and spatial overview of Chl a in shallow eutrophic lakes.

Area D - Board 386

River-lake systems are increasingly suffering from multiple stressors such as eutrophication and disturbance by more frequent extreme weather events induced by climate change. Eutrophication in combination with disturbance events increases the likelihood of harmful algal blooms, which might lead to elevated greenhouse gas emissions from lakes. Furthermore, heavy rain events often related with nutrient run-off from the catchment area influence nutrient levels and bloom development of downstream lakes along river-connected lake chains. However, phytoplankton dynamics and algal bloom development along river-connected lake chains are poorly understood. This is of special concern, since, e.g., the water quality of river-connected German lowland lakes is in poor or bad ecological status according to the European Water Frame Directive.
The interdisciplinary CONNECT project, therefore, aims for comprehensive insights in river-lake chain ecology and related processes, i.e. lake coherence along river-connected lake systems. We will use high frequency in-situ measurements in combination with remote sensing data to enable lake monitoring at high spatial and temporal scales. We hypothesize that lake connectivity is of major importance for the spreading of eutrophication events and thus algal blooms increasing the coherence between lake ecosystems along a lake-river chain.
To test our hypotheses of lake-to-lake connectivity we are conducting a field campaign on the Havel river-lake chain in the NE German lowlands and a large-scale enclosure experiment in the LakeLab (well instrumented with automatic profilers) at Lake Stechlin to manipulate lake connectivity. Here, we present how we combine high frequency in-situ measurements, high throughput laboratory approaches by FlowCam, HPLC algal pigment analyses and ground based as well as airborne and satellite remote sensing data to study lake connectivity. Our study i) builds a collaborative research network between remote sensing experts and freshwater scientists, ii) analyzes lake-to-lake connectivity of deep and shallow lakes, iii) will further improve ground truthing of remote sensing data in these aquatic systems, and iv) provides advances in monitoring and management of freshwater ecosystems with special emphasis on river-lake chains.

Area D - Board 387

Within the project CONNECT we are establishing a collaborative network between experts in remote sensing (RS) and freshwater ecology to study connectivity and coherence of lake ecosystems in a regional context at unprecedented temporal and spatial resolution. The overall aim is to understand the yet unexplained variation in phytoplankton dynamics among river-connected German lowland lakes, many of which are presently classified as in poor to bad ecological status. These lakes often face a high risk of eutrophication, mass development of harmful algal blooms, and high production of greenhouse gases. We suggest if measured on adequate temporal and spatial scales much of the among-lake variation in phytoplankton dynamics to be explained by the strength of hydrological lake-to-lake and lake-to-catchment connectivity as modulated by lake depth and mixing regime. This may have profound implications for the maximum intensity, spatial range and regional-scale magnitude of eutrophication impacts. We use (i) a large-scale experimental manipulation of lake connectivity, and (ii) observational field campaigns contrasting deep and shallow river-connected lakes. We will tackle this research frontier by an innovative combination of automatic high-frequency in situ measurements with state of the art RS technology combining ground based, airborne and satellite remote sensing data. Climate change is expected to alter the hydrology, and thus the connectivity of lake-river systems. However, it is also predicted to increase extreme weather events leading to an increased input of nutrients as well as colored dissolved organic matter (cDOM). By providing data of high spatio-temporal coverage, CONNECT will provide basic high quality data to better understand mechanisms of eutrophication at the local and regional scale. Our data, thus, provide a valuable basis to improve current management of such river-connected lake ecosystems under future climate scenarios. To reach this ambitious goal, the project will (i) build a cross- disciplinary collaborative network of excellence, (ii) develop a mechanistic understanding of lake ecosystem functioning at local and regional scale, (iii) improve future environmental monitoring and interpretation of available data from inland waters, and (iv) support more effective integrated management of river-connected lakes to mitigate eutrophication impacts.

Area D - Board 388

AquaWatch, the water quality community of practice, achieved Initiative status with the Group on Earth Observations (GEO) in 2017. Focused on the synthesis of remotely sensed and in situ data for improved global data products that are open access, AquaWatch offers subject matter experts a unique forum to tackle community-wide issues and to offer expert-informed guidance and advocacy to governments and programs with a global reach. During the recent workshop in Stirling, UK, jointly held with GLOBOLAKES, participants compiled water quality knowledge gaps, challenges and opportunities, and AquaWatch identified several tasks volunteers from the water quality community will advance in 2019. Some of the activities include support of validation efforts, algorithm documentation, capacity building, education and developing pilot projects. We are seeking volunteer subject matter experts to participate in these efforts. If you are interested, please reach out to us: www.geoaquawatch.org ; @geoaquawatch; info@geoaquawatch.org

Area D - Board 389

Globallakewatch.org: global online near real-time case 2 water quality monitoring for the masses

In the recent decades, algorithms for satellite-based monitoring of chlorophyll, suspended sediment and coloured dissolved organic matter (CDOM) have been developed and evaluated across many lakes and costal areas. While accuracies are only rarely perfect and transferability between sites is limited, such applications have already proved highly useful for understanding lake processes. Satellite imagery will never provide data with accuracy comparable to field sampling, but the spatially explicit view offered by imagery can be extremely informative. The heterogeneity of lake conditions revealed by such images and the currents, eddies and plumes identified can provide essential background data for ecological studies. Additionally, satellite imagery can act as an early warning system or complement local monitoring activites. However, most satellite water quality information was so far only accessible to the remote sensing scientists who produced them and their immediate collaborators. Meanwhile, there is increasing interest from the public in water quality issues. Additionally, in many remote locations, satellite data is the only source of up-to-date information, but the local population has no access to these datasets. Opening satellite water quality in an easy-to-use global data platform is expected to contribute to education and raise awareness to the importance of water quality.
We use Google Earth Engine to access and process archive and current satellite data. Google Earth engine is a development, analysis and visualization framework that provides access to freely available satellite datasets (MODIS, Landsat and Sentinel) and provides processing capacity at petabyte scale. Satellite imagery is processed online and offered for analysis typically a few hours after collection. Global Lake Watch is a free interactive platform for querying and visualizing satellite data for water quality applications globally based on Google Earth Engine. We offer a set of robust algorithms for chlorophyll and suspended sediment concentration estimation in addition to true colour visualizations. Based on comparative studies from the literature (mainly based on MERIS), the Fluorescence Line Height index is recommended for most Case 2 waters, but Maximum Chlorophyll Index and Reflectance Line Height is also offered. Suspended sediment is estimated from band intensities close to 700 nm. The integration of Landsat 7 and 8 and Sentinel 2 and 3 means that if cloud cover allows, imagery is collected daily for most areas. Obviously chlorophyll concentration estimates are strongly influenced by water transparency and the maps produced are only qualitative. Nevertheless, we expect that Global Lake Watch can be a starting point for viewing lake conditions for a wide range of users including scientists, decisionmakers, fishermen and the interested public.

Area D - Board 390

Surface waters play a pivotal role in the Sahelian region as they provide a critical water resource for livestock and people, they have an important impact on the ecosystems in terms of biodiversity and emission of greenhouse gases and they influence the spread of water-borne diseases. In particular Suspended Particulate Matter (SPM) play a role on the development and the survival of bacteria responsible for diarrheic diseases, the major cause of mortality for children under 5 years of age in this area.
Monitoring turbidity and SPM in water bodies in this region is therefore a key issue. However, their spatio-temporal dynamics is poorly known given, on the one hand, the scarce in-situ data available and, on the other hand, the high spatial and temporal resolution necessary to follow Sahelian water bodies from space, that only recent satellites can provide. In addition, optical remote sensing in this region can be quite challenging given the high aerosol loading (mineral dust and biomass burning) that makes atmospheric corrections more difficult as well as the high water vapor content and cloud cover during the rainy season. Moreover, the extremely high values of turbidity found in West African surface waters may invalid inversion algorithms classically employed for other regions.
This presentation addresses the capability of Sentinel2 level2 products produced by the Theia pole to monitor SPM in Sahelian waters.
In-situ data, collected over different sites (the Bagré reservoir in Burkina Faso, the Agoufou lake in Mali, the Niger river and some ponds in Niger and Senegal), are used to evaluate different indexes to derive SPM from the reflectance in the visible and infrared bands of Sentinel2 Theia products. We show that the NIR reflectance is well suited to map SPM up to high values and that the same retrieval algorithm can be applied to estimate SPM for the different kinds of water bodies (ponds, lakes and rivers) investigated. In addition, the atmospheric correction scheme employed to retrieve surface reflectance for Sentinel2 products provided by the Theia pole, is found to be accurate enough for monitoring SPM in this region, even if there is still potential for improvement.
Overall Sentinel2 proved very useful to follow the SPM spatio temporal variability in different water bodies across the Sahelian area. For example, it allowed to quantify SPM in the Niger river during the “red flood” hydrological phenomenon, to highlight SPM propagation from the upstream to the downstream in the Bagré and Guiers reservoirs, the latter providing water to Dakar, and to better understand SPM drivers and their complex interactions with meteorological and hydrological variables (precipitation, wind, water levels and lake size) in ponds and small lakes in northern Mali.

Area D - Board 391

Submerged aquatic vegetation (SAV) plays an important role in freshwater lake ecosystems. Due to its sensitivity to environmental changes, several SAV species serve as bioindicators for the trophic state of freshwater lakes. Variations in water temperature, light availability and nutrient concentration affect SAV growth and species composition. To observe the trophic state of freshwater lakes as recommended by the European Water Framework Directive (WFD), SAV needs to be monitored regularly. This study analyses the development of macrophyte patches at Lake Starnberg, Germany, by exploring four Sentinel-2A data sets recorded within the main growing season in August and September 2015. Two different methods of littoral bottom coverage assessment are compared, i.e. a semi-empirical method using depth-invariant indices and a physically based, bio-optical method using WASI-2D (Water Colour Simulator). The lake bottom spectra required from both methods were generated by using the reflectance models developed on base of in situ multiyear daytime and across the vegetation period measured species specific spectral libraries. Both methods identified vegetated and non-vegetated patches in shallow water areas and allowed tracing the seasonal change of SAV pattern. Furthermore, tall- and meadow-growing SAV growth classes could be differentiated. Both methods revealed similar results when focusing on the identification of sediment and SAV patches (R² from 0.5603 to 0.8137), but not for a differentiation on SAV class growth level (R² from 0.002654 to 0.4194). For relating the observed seasonal changes in pattern of SAV to effects of climate change a multiyear monitoring seems necessary. Referring to the outcomes of an analysis of multiyear in situ data series collected over the daytime and the SAV growing season by a submersible field spectroradiometer based system conducted by the authors (Fritz et al., 2017a), even the identification of invasive species seems possible by multiseasonal observations, because invasive species seems more resistant to changes of the environment and their canopy collapse later in the vegetation period.

Area D - Board 392

Currently the status of small lakes in Estonia has been evaluated with a 6-year interval according to the establishment of the River Basin Management plans. With the limited available resources the prioritization of water bodies to be included in the field-campaign needs to be done. The main goal of the current study was the development of an analysis that could contribute to such prioritization for the annual in-situ campaign relying on operational information from different databases. The analysis was performed for the lakes in the region of the Koiva River basin, Estonia. We analyzed datasets that provide operational/routine information about man-induced pressures such as clear-cuts, agricultural fields and nutrients input from wastewater outlets in the catchment area of the water body for the years 2013-2016. The percentage of clear-cut was calculated from the clear-cut maps compiled on the basis of remote sensing imagery (Sentinel 2 and Landsat 8). Changes in agricultural land use were estimated from datasets obtained from Estonian Agricultural Registers and Information Board. To identify the possible shift in water quality compared to in-situ data from the earlier status evaluation, the transparency and chlorophyll- a concentration were determined from Sentinel-2 imagery (2017-May 2018) by implementing the standard C2RCC (Case-2 Regional/Coast Colour) processor. The analysis revealed not only the water bodies with increased man-induced pressures in the catchment area and lakes with the shift in water quality parameters but highlighted also the changes expressed in ecological status evaluation in 2017. As the analysis revealed the possible shifts in lake status class adequately, it could be used for planning field campaigns in the future. For instance, priority should be given to lakes in which remote sensing data indicates significant changes in water quality or where the nutrient pressure of a water body has changed due to (1) changes in land use (increased clear-cuts or agricultural land) or (2) increased flow rates from wastewater outlets.

Area D - Board 393

The importance of inland water bodies has never been greater. Inland waters, including lakes, have become one of the key points of the carbon cycle and therefore the climate change. Lakes are also economically important as a source of drinking water, and also providing fishing and recreation opportunities. In this study, we compared two sensors on different Copernicus satellites: MSI on Sentinel-2 and OLCI on Sentinel-3. We used in-situ data from 51 sampling points across 4 different lakes collected during 2018 to validate different processors and methods. Level-2 optical water quality products, such as chlorophyll a concentration (Chl-a), water transparency (Secchi depth), total suspended matter (TSM), and coloured dissolved organic matter (CDOM) were compared against in-situ data. Three out of four of the studied lakes are situated in Latvia: (1) a clear oligotrophic Lake Razna, where Secchi depth was measured 3.2-6.6 m and the average Chl-a was 5 mg/m³ during 2018; (2) a shallow, eutrophic and macrophyte-dominated L. Lubans, where Secchi depth varied 0.4-1.2 m and the average Chl-a was 16 mg/m³; (3) and a shallow, eutrophic and CDOM-dominated L. Burtnieks in northern Latvia, where Secchi values are similar to L. Lubans, 0.4-1.2 m, the average Chl-a was 27 mg/m³. The fourth of study lake is large and shallow, very turbid and eutrophic L. Võrtsjärv located in Estonia. The measured Secchi depth was 0.5-0.7 m and the average Chl-a 28 mg/m³ in 2018.
Our main objective of this study was to find the best algorithms for estimating Level-2 optical water quality parameters from both Sentinel-2 and -3 satellites using in-situ match-ups. We used the standard Level-2 products, some available processors (C2RCC/C2X and MCI), band ratio algorithms based on the reflectance spectra, and reflectance based optical water type (OWT) guided approach. In our interest was also to compare the OWT results from both satellites. The dynamics of OWTs of the two satellites were generally agreeing. The Turbid and Very Turbid water type classes had the strongest disagreement, that can be caused by the band resolutions, the shape of the reflectance spectra or the saturation of the signal. This shows that the reflectance spectra are an important factor if using the OWT guided approach to estimate the Level-2 products. To estimate how useful OWT guided approach is, we compared the spatial variability of the optical water quality parameters derived with and without the OWT guiding. The best methods to obtain the Level-2 water quality products were different for Sentinel-2 and -3. Most cases the band ratio algorithms gave the best results for both satellites. Deriving the water quality parameters from Sentinels is a first step in estimating the ecological status of the lakes. The symbiosis of the two satellites gives us the chance to do it in most of the lakes, big or small.

Area D - Board 394

In this study, we are aiming to test suitability of S2 MSI data for mapping water quality over four morphologically and optically different water bodies, as discovered from in situ measured remote sensing reflectance, in Lithuania: Plateliai and Dusia lakes – categorized as oligo-mesotrophic water bodies; Kaunas reservoir and the Curonian Lagoon – categorized as eutrophic water bodies with massive growth of harmful cyanobacteria blooms. Lithuanian water bodies (lagoon, lakes, reservoirs) are extremely important recreational sites, however strongly impacted by eutrophication with appearance, growth and proliferation of cyanobacteria. The ongoing eutrophication with all the consequences leads to the high demand for comprehensive operational monitoring of water quality both at local and global scales. However, majority of Lithuanian surface waterbodies are relatively small and therefore Earth Observation-based monitoring of algal blooms with previous satellite data (i.e. MERIS/Envisat, MODIS Aqua/Terra) was always limited. Sentinel-2 Multispectral Imager’s (S2 MSI) opened a great new potential of small and large water bodies remote sensing with a comprehensive spatial and spectral resolution.
For this study, in situ reflectance (Rrs) measurements concurrent with S2 overpasses were collected with WISP-3 instrument (Water Insight) over the Curonian Lagoon and for a first time over Kaunas reservoir, Plateliai and Dusia lakes during summer 2016-2018. Simultaneously, water quality parameters were measured at each water body. The performance of three atmospheric correction algorithms (Sen2Cor, C2RCC and 6SV code) was tested. For chlorophyll-a (Chl-a) retrieval semi-empirical band-ratio algorithms and Chl-a product of C2RCC processor were tested against in situ measurements. Continues Rrs measurements were taken by a WispStation (fixed position spectrometer system, Water Insight) installed in the Curonian Lagoon.
The comparison between Rrs measured in situ and derived from S2 MSI reviled that EO-based Rrs after application of 6SV code showed a slight overestimation, however the shapes of spectra were in a good agreement for all tested study sites. In the case of eutrophic water bodies the performance of Sen2Cor showed also a good result. The results show that S2 MSI data has great potential for inland water remote sensing in Lithuania, as we were able to map Chl-a concentration in lakes, Kaunas reservoir and the Curonian Lagoon by means of commonly used band-ratio type algorithms. Significantly strong correlation was determined between band-ratio at 704 and 665 nm S2 MSI bands and in situ measured Chl-a in the eutrophic water bodies: Kaunas reservoir (R2 = 0.80) and the Curonian Lagoon (R2 = 0.83), while in the oligo-mesotrophic lakes significant correlation was determined between 490 and 560 nm band-ratio and in situ measured Chl-a (R2 = 0.58). Chl-a retrieved after application of C2RCC underestimated in situ measured values and therefore the further investigation and tuning of parameterization parameters to local conditions and water type is needed. Overall, a great success of the band-ratio algorithm that uses red and NIR band to be suitable for Chl-a mapping over the productive water bodies was demonstrated globally, therefore could be used operationally for eutrophic status determination and as warning indicator for cyanobacteria presence using data from various platforms: satellite data and handheld or fixed position hyperspectral spectroradiometers (like WispStation), that are providing the continues Rrs measurements. Examples from Lithuanian waters will be demonstrated and discussed.

Area D - Board 395

Finnish Environment Institute (SYKE) has over twenty years of experience in utilizing satellite observations for monitoring surface algal blooms in the Baltic Sea. Right from the beginning, the monitoring has been developed as an integrated part of SYKE’s national weekly algae bloom press releases. Cyanobacteria, a blue-green algae, forms toxic surface floating blooms during the summer holiday periods in July-August, therefore are of high interest to public and press in all Baltic countries. The algae bloom service of SYKE has mainly utilized medium resolution instruments, such as MODIS, Envisat MERIS and Sentinel-3 OLCI for generating the surface algal bloom map products. In 2018 high resolution data from Sentinel-2 series (S2A and S2B) MSI and USGS Landsat-8 (LC8) OLI were included in the algal bloom monitoring. This quickly increased the public interest on the service.

In the Northern Baltic Sea, the cyanobacteria bloom season typically starts as the waters warm up by the end of June and continue to September. In 2018, the blooms were especially intensive in certain parts of the Baltic Sea, namely in the Gulf of Finland and in the Northern Baltic Proper. During a two-week period in July, extensive cyanobacterial blooms were packed tightly in the coastal waters of Finland, and was of serious and intensive concern to the public, authorities and press. As an organization responsible to this national concern, SYKE formed a special group to provide information on the algae and its toxins. During these intensive weeks of mid-summer holiday period, the EO services of SYKE provided daily information on the extent of surface blooms. In addition to the bloom observations, EO data was used as input data for a drifting model predicting the movements of the surface blooms (run by the Finnish Meteorological Institute). Also citizen observation data of cyanobacterial blooms were collected and provided together with EO data to give larger and more specific information on algal blooms.

SYKEs algae bloom services are multi-instrument approach: S2A & S2B MSI, LC8 OLI and S3 OLCI are utilized whenever a non-cloudy overpass is available. The products have four classes: No, Potential, Likely and Evident algal bloom, determined for each pixel and then generalized over coarser area. Algal bloom maps based on OLCI data are produced and published daily via Web service PINTA (www.syke.fi/pinta/en). PINTA also includes annual composites of algae occurrences provided after the algal bloom season. The high resolution products from S2A & S2B MSI and LC8 OLI instruments are available in TARKKA, a service for various types of high resolution EO products and complementary GIS (www.syke.fi/tarkka/en). During summer 2018 a total datasets of true color images (RGB) of S2A & S2B MSI, LC8 OLI as well as S3 OLCI were available in TARKKA and PINTA user interfaces. They were used largely in SYKEs press releases and in various newspapers.

In the poster we will present how the cyanobacterial blooms advanced in SYKE’s products and in RGB images. Satellite products will be compared with the citizen observations.

Area D- Board 396

In addition to 450 000 lakes, Norway has the world’s 2nd longest coastline. The Water Framework Directive are one of EU's most ambitious and comprehensive directives for the environment. Its main purpose is to secure protection and sustainable use of the water environment in fresh, ground and coastal water. Norway has committed to this through the Water Regulation in the Norwegian law (“Vannforskriften”). Through this commitments, marine and fresh water bodies are monitored each year and each month to be able to set the ecological status. Both in the marine water and in fresh water, one of the biological water quality elements is the phytoplankton. In marine waters biomass is measured through the biomass proxy Chlorophyll a. For freshwater there are also other parameters included such as the total algae volume, species composition and cyanobacteria maximum volume. With the very long and complex coastline of Norway, the many lakes and the phenology of the phytoplankton with a production period from February/March to September/October (with spring bloom in February/March) depending on the region, the monitoring program is comprehensive both in number of water bodies to monitor but also in the sampling frequency. NIVA together with the Norwegian Space Centre (NSC) and the Norwegian Environment Agency (NEA) are investigating the possibilities of using and including remote sensing (the Sentinels, Landsat-8 and other sensors) in the work of calculating the ecological status. Various algorithms are developed for case 1 (open ocean) and case 2 (coastal waters) waters for the various satellites available with different regional focus. Some algorithms suit better than others for the different water regions (marine and fresh water) and one processing procedure can not be done covering all the water types. With focus on Norwegian coastal and inland waters, various algorithms are being tested and IOP and in-water products are developed for its potential use in future monitoring programs. In addition to this, remote sensing reflectance (Rrs), water inherent optical properties (IOP) and satellite end user products derived from the Sentinel-3 OLCI sensor using a comprehensive method for atmospheric correction and IOP retrieval has been developed based on data from Norwegian fjords and coastal water areas. In this work special emphasis has been placed on accurate retrieval of Rrs values and water IOPs over open ocean as well as over cDOM rich and turbid coastal and/or inland water. For validation match-up data obtained from the Norwegian national FerryBox network (NorSOOP) and the national monitoring programs of the NEA are used. Along the Norwegian coast, from Bergen to Kirkenes, through the Barents Sea, and also the North Sea and Skagerrak, the NorSOOP ferries network is covering a large diversity of water types. The FerryBox systems measure core parameters such as salinity, temperature, oxygen, turbidity, cDOM and chlorophyll fluorescence, wind speed and direction, and hyperspectral above water reflectance. The NorSOOP Ferrybox network provides crucial insight into the dynamics of environmental conditions and a large and diverse dataset for remote sensing algorithms validation.

Area D - Board 452