Sea Surface Salinity (SSS) is an Essential Climate Variable (ECV) that plays a fundamental role in the density-driven global ocean circulation, the water cycle, and climate. The satellite SSS observations from the Soil Moisture and Ocean Salinity (SMOS), Aquarius, and Soil Moisture Active Passive (SMAP) missions have provided an unprecedented opportunity to map SSS over the global ocean since 2010 at 40-150km scale with a revisit period of 2 to 3 days. This observation capability has yielded new findings concerning the SSS variations related to climate variability such as El Niño-Southern Oscillation, Indian Ocean Dipole and Madden-Julian Oscillation, and the linkage of the ocean to different mechanisms of the water cycle such as evaporation, precipitation and continental runoff. This has enhanced the understanding of various upper-ocean processes such as tropical instability waves, Rossby waves, mesoscale eddies and related salt transport, salinity fronts, hurricane haline wake, river plume variability, cross-shelf exchanges. There is also an emerging use of satellite SSS in the study of ocean biogeochemistry processes, e.g. linked to air-sea CO2 fluxes.

Following the success of the initial oceanographic studies on the retrieval and monitoring of this new variable, the European Space Agency (ESA) Climate Change Initiative CCI+SSS project (2018-2021) has aimed at generating improved calibrated global SSS fields over a 10-year period (2010-2020) from all available satellite L-band radiometer measurements (Boutin et al, 2021a, 2021b), currently being extended to 2002 from C-band radiometer measurements at regional scale. It fully exploits the ESA/Earth explorer SMOS mission complemented with SMAP and AQUARIUS satellite missions for the L-Band, and the AMSR-E mission for the C-band. The project gathers teams involved in earth observation remote sensing, in the validation of satellite data and in climate variability study.

Scientific user requirements and product specifications drive the technical developments of the CCI+SSS datasets that are exploited in ocean and climate science research. Requirements of performance, coverage, spatial and temporal resolutions, uncertainty limits, provision, format, data latency, etc... as specified by Global Climate Observing System (GCOS), are iteratively assessed during each cycle of the project, through regular engagement with user groups such as the ESA Climate Modelling User Group (CMUG), and through interaction with European and national projects and international bodies, as well as individual or institutional bilateral contacts.
To build the L-band CCI+SSS product, the merging of the three existing satellite data sets is performed after standardizing the error estimation by comparing SSS acquired at the same time by different sensors, and by extracting systematic uncertainties and random errors on SSS from each satellite mission. They are used to correct individual SSS before merging using a temporal optimum interpolation and to estimate the final uncertainty on level 4 (L4) SSS.
The comparisons of the CCI merged 3-sensor product with in situ ground truth indicate much better performances than the ones obtained with a single sensor data product, with global robust standard deviation of difference with respect to in situ references (ARGO) of 0.15 pss. Large scale interannual variability is successfully represented, and SSS variability in highly variable regions like the Bay of Bengale and in river plumes in the Atlantic Ocean is well marked, confirming the usefulness of the merged product for scientific studies.

This presentation aims at reviewing recent results obtained within the CCI+SSS project, as well as ongoing studies currently addressed in the phase 2 of the CCI+SSS project (2022-2024). In view of the results obtained, we will discuss the needs for future CCI+SSS products as considered with various data users and in view of diverse applications. For example, surface salinity distribution has a crucial impact on marine ecosystems and ocean health, and notably has been shown to be an indicator, and even, an active factor in water eutrophication mechanisms. SSS distribution is much influenced by climate variations, thus CCI+SSS monitoring capability can be a cutting-edge ingredient for the investigation of marine biogeochemistry changes in different regions over the years.

It is now recognised that climate change is accelerating the frequency and intensity of algal blooms, marine heatwaves and hypoxia events with occasionally devastating effects on the marine ecosystems. These changes impact not only the natural wildlife but affect tourism, fisheries and the population’s economies dependent on the sea. Sea surface salinity (SSS) and temperature (SST) are essential variables for monitoring and understanding ocean health. The ability to determine SSS and SST enables key measurements in understanding acidification and de-oxygenation, in response to the changing climate. These are closely related to Ocean Primary Production (OPP), which influences ocean-carbon uptake and is the base of the Ocean food web.

Coastal zones are subject to a mixture of complex natural and anthropogenic influences. River plumes provide influxes of organic material, increasing ecological complexity by the introduction of nutrients, upwelling brings cooler nutrient rich water and reefs and mangroves can protect the shore from erosion. The complexity of processes in these regions requires high resolution differentiation of different temperate and saline zones.

Typically, thermal and microwave remote sensing is employed to retrieve SST and SSS in aquatic systems. Satellite-derived ocean colour provides the opportunity to process data at a higher temporal and spatial resolution than traditional microwave remote sensing, with ocean colour satellites having pixel resolution on the tens to hundreds metre scale compared to microwave km scale.

This paper explores a new methodology to extract SST and SSS from hyper/ multispectral ocean radiance. Water leaving radiance is linked to the inherent optical properties of the water column, effected by the constituent parts for example from coloured dissolved organic matter, chlorophyll and total suspended matter. These parts contribute to the backscatter, absorption and reflectance coefficients, influencing the spectral signature of the radiance.

This paper considers two sources of spectral information, first hyperspectral water leaving radiance measured at 1nm spectral bands from 350-900nm using ground based radiometers. This was collected on a research cruise in Patagonia at the same time as a thermosalinograph collected in-water temperature and salinity. The second is multispectral satellite images from Sentinel 2, located in the sites of 4 UK Rivers, the Thames, West Gab, Liverpool Bay and Dowsing bay. The satellite images were matched with UK Smart buoys who have temperature and salinity records in 30 minute increments from 2015-2021 matching with the Sentinel 2 satellite timeframe.
The method matched both temporally and spatially, the in-water data of SSS and SST with the hyper and multispectral water leaving radiance, at the satellite or ground level.

This ground based hyperspectral matched data was used to train a linear regression model, with the inputs being the normalised remote sensed reflectance values and output either salinity or temperature, salinity estimations therefore don’t depend on temperature (but there is a correlation). The model was then able to learn the relationship between the spectral input and SSS/SST values and accurately estimates SSS and SST in test regions using just the hyperspectral signature as input.

Further work is also done on using principle component analysis (PCA) to reduce the dimensionality of the inputs whilst maintaining predictive accuracy, 0.99 of the variance was kept just from 5 input variables in the new orthogonal space, with temperature test RMSE 0.63. Feature selection was also examined to understand the spectral bands with the strongest contribution to either the salinity or temperature values, without changing the orthogonal set as PCA does. Feature selection using the Python package sklearn f_regression to select the bands with the largest coefficient values showed the importance of the 400-600nm band region in the prediction of temperature values.

Multispectral Sentinel-2 (S-2) MSI images over the UK smart buoys were used, with the benefit of the 10 m resolution, which enables to monitor smaller aquatic systems. The matched S-2 band information was applied with the same methodology as the hyperspectral ground water leaving radiance, first with a linear regression model, then using feature selection to identify important bands and inputs. The paper then looks at increasing the complexity of the model, and training machine learning algorithms such as neural networks to be able to learn more complex relationships and predict values in different optical water types, with less inputs available as typically found in satellite multispectral images.

This paper demonstrates a link between the ocean colour/ radiance values and the sea surface physical properties temperature and salinity, with a model accurately estimating SST and SSS from just radiance. This relationship between sea surface properties and spectral signatures is then tested in satellite ocean colour images which can be used to monitor these environmental properties with more regular temporal sampling, better spatial coverage and in previously inaccessible locations.
Acknowledgements: Jose Luis Iriarte for in-situ SST/SST data from Patagonia.

The present work is developed within the MedEOS project, an application development project funded by the European Space Agency (ESA) as part of the Mediterranean Regional Initiative. Its main objective is to develop and produce daily high-resolution, gap-free water quality products based on Earth observation (EO) data for the whole Mediterranean coastline. This is achieved through combining the high temporal resolution of Sentinel-3 Ocean and Land Colour Imager (S3 OLCI) and the high spatial resolution of Sentinel-2 Multispectral Instrument (S2 MSI) in a process of data fusion.
Within MedEOS, five EO directly derived water quality products are to be developed: Total Suspended Matter, Turbidity, Chlorophyll-a Concentration, Secchi Depth and Colored Dissolved Organic Matter. In addition, EO indirectly derived water quality products shall also be produced, largely relying on combination of satellite and in situ data with numerical model results: Faecal Bacterial Contamination Indicators, Eutrophication Indicators, Harmful Algal Blooms, and Global Environmental Anomaly Detection. Finally, a river plume monitoring algorithm will provide a systematic detection of plumes related to major rivers discharging freshwater into the Mediterranean basin.
Given state-of-the-art approaches for spatiotemporal data fusion available in the litterature (see review in [1]), the proposed strategy for data fusion on EO derived water quality products is a combination of reconstruction-based and learning-based approaches.
The reconstruction-based STARFM (Spatial and Temporal Adaptive Reflectance Fusion Model) [2] is widely used in the litterature for spatiotemporal data fusion for land applications. Reconstruction-based approaches such as the STARFM relies on the hypothesis of reflectance similarity between both high and coarse-resolution sensors. If this hypothesis is often verified for land applications, it is less true on water. Lower values of water surface reflectances in comparison to land reflectances makes the reflectance similarity much more delicate to reach between two optical sensors. The higher sensitivity to atmospheric correction, higher impact of BRDF effects, heterogeneous sunglint effects and differences in spectral response for the same spectral band are all sources of discrepancies between water reflectance levels estimated from each sensor. It was therefore decided not to apply STARFM at the reflectance level but rather at the water quality products level to ensure a better consistency between products derived from both optical sensors.
Besides, STARFM is also strongly based on the assumption that objects do not drastically change between the date at which a pair of S2/S3 images exists (T0) and the date for which only a S3 image is available (T1). As it is most of the time not the case on highly dynamic ocean waters, an alternative approach is needed to enable the use of algorithms such as STARFM on water applications.
Our proposed strategy is to trick the STARFM algorithm by feeding it with an artificial pair of S2 and S3 data. In order to do so, a reference database made of small image patches is built from a S2/S3 water quality product database derived from a large time series of data. The S2/S3 pair is then created artificially through a coarse and fine resolution matching analysis between the reference database and the S3 water quality product observed on the targeted day. The matching analysis is first carried at the coarse resolution and further refined at the fine resolution. Each matching patch of coarse S3 pixels and its corresponding group of fine S2 pixels are further used to create an artificial S3 T0 product at a coarse resolution and an artificial S2 T0 product at fine resolution. Both the pair of S2/S3 artificial data at T0 and the S3 gap-filled product at T1 are given as input to the STARFM algorithm to produce the simulated product at S2 fine resolution.
The gap-filling part of the process is performed using the DINEOF algorithm [3]. The database of water quality products derived from Sentinel-3 images are combined into multiple overlapping time series of approximately 2 months. Masks are extracted from each daily data to differentiate clouds from land pixels and perform the gap-filling only on desired pixels. The daily S3 products are also subsetted to fit the extent of a targeted S2 tile.
A first run of DINEOF is performed on the target time series. Using an approach proposed in [4], outlier pixels, mainly located near the edges of clouds, are detected and further removed from the dataset before applying the second round of DINEOF. The target day gap-free product is further extracted from the gap-free time series and used as input to the data fusion part of the processing chain.
An example of results of data fusion is given in Figure 1.
Next steps in the MedEOS project will be dedicated to the integration of all the different algorithms, developed by different service providers, in a single EO Exploitation Platform solution. This solution allows for the deployment of different components of the services supply chain in separated ICT cloud providers, making all service outputs available in a unique archive and catalogue module and visible in tailor made web-based geoportal. MedEOS implementation, therefore, follows the latest guidelines from the data science community, setting up the service processors where the input data is readily available, thus avoiding unnecessary download and transfer of data and increasing the performance of the services.
Production of water quality products and application of the proposed data fusion approach shall be performed initially for one complete year (2020) within five Pilot Areas in Egypt, Greece, France, Spain and Tunisia. Validation of these delivered products will be performed by comparing results with CMEMS products OCEANCOLOUR_MED_BGC_HR_L3_NRT_009_205 [5] and with in-situ data collected by engaged users from each pilot. In a second phase, those products shall be derived and validated over the entire Mediterranean Sea coasts for a 3,5 year period, from March 2019 to September 2022.
For more information please visit the MedEOS website: https://medeos.deimos.pt/

[1] Belgiu, M., & Stein, A. (2019). Spatiotemporal image fusion in remote sensing. Remote sensing, 11(7), 818.
[2] Gao, F., Masek, J., Schwaller, M., & Hall, F. (2006). On the blending of the Landsat and MODIS surface reflectance: Predicting daily Landsat surface reflectance. IEEE Transactions on Geoscience and Remote sensing, 44(8), 2207-2218.
[3] Beckers, J. M., & Rixen, M. (2003). EOF calculations and data filling from incomplete oceanographic datasets. Journal of Atmospheric and oceanic technology, 20(12), 1839-1856.
[4] Alvera-Azcárate, A., Sirjacobs, D., Barth, A., & Beckers, J. M. (2012). Outlier detection in satellite data using spatial coherence. Remote Sensing of Environment, 119, 84-91.
[5] https://resources.marine.copernicus.eu/product-detail/OCEANCOLOUR_MED_BGC_HR_L3_NRT_009_205/

The Black Sea is a small enclosed basin where coastal regions have a large influence and mesoscale signals dominate the dynamics (the Rossby radius of deformation is about 20km). Large riverine inputs, mainly on the northwestern shelf, induce well-marked horizontal gradients in the distribution of the Black Sea salinity and optical characteristics: coastal and shelf waters have very low salinity and contain large amounts of optically active materials (e.g. coloured dissolved organic matter) and its oligotrophic deep sea has a salinity around 18.2. The presence of these contrasting water characteristics in a relatively small enclosed environment, combined with land contamination and the specificities of its atmospheric composition(e.g. high cloud coverage, aerosols) make the Black Sea a challenging area for the development of high quality satellite products.

We present results from the ESA EO4SIBS (Earth Observation for Science and Innovation in the Black Sea) project dedicated to the development, and subsequent scientific analysis, of new algorithms for the development of satellite products in the Black Sea. In particular, ocean colour products (chlorophyll-a, total suspended matter concentrations, turbidity) are produced from Sentinel 3 (S3) OLCI data combining different algorithms and then, a classification of water masses is proposed. A revised gridded altimetry product based on 5-Hz along track data (combining Cryosat and S3 SAR over 2011-2019) is produced and validated in the coastal zone with tide gauge data. L2, L3 and L4 Sea Surface Salinity is derived over 2011-2020 from the L-Band measured by SMOS and compared with in-situ surface salinity data from field sampling and Argo. In the presentation, we will describe the technical development that are needed to obtain high quality products in the Black Sea. An experimental CDOM product is also proposed.

All these products are integrated to further understand the Black Sea physical and biogeochemical functioning. For instance, the Black Sea mesoscale dynamics are inferred from the 5-Hz altimetry product using an eddy detection and tracking algorithm. The benefit of assimilating ocean colour data in the Black Sea operational model (also known as CMEMS BS-MFC BIO) for the prediction of the Black Sea ecosystem is illustrated as for instance the simulation of particularly intense blooms in June 2017.

EO4SIBS Gridded products are archived as CF-compliant NetCDF files and disseminated through ncWMS protocol.
We conclude with a set of recommendations for observational requirements needed to increase the quality of satellite products in the Black Sea and to be able to use the full potential of current and new information provided by satellites.

Dissolved oxygen (DO) is one of the most critical parameters for aquaculture, as it is vital for all organisms living in water. The survival, growth and food intake of fish is directly affected by changes in DO concentration. Therefore, the systematic and continuous monitoring of DO is of crucial importance for proper production management.
Remote sensing is widely used for the detection of optically active parameters such as chlorophyll-a, total suspended matter, and more. DO does not change the optical properties of water, so it is impossible to estimate its concentration directly from the reflection values of satellites. Literature has shown that the concentration of DO can be estimated indirectly as a result of its correlation with other parameters (e.g. temperature). Nevertheless, only few studies have addressed the issue using remote sensing data [1]–[4]. These studies have limitations, either because they are applied in a small area, thus, giving only local information which is difficult to transfer in a different region, or because the available in-situ data are limited. Our work aims to integrate satellite data, along with in-situ observations using machine learning techniques in order to bring forth innovative approaches on how DO can be estimated and monitored on large scale across the Aegean Sea.
Most recent satellite sensors offer a wide range of data products useful for estimating water quality parameters at large scale temporally, however, they are not suitable to be adopted as a standalone solution in real case scenarios. Copernicus Marine Environment Monitoring Service (CMEMS) provides regular and systematic biogeochemical and physical information on the marine environment. In the framework of our study we exploit Level-4, daily, gap-free satellite observations for chlorophyll-a concentration and sea surface temperature from multi-platform observations. We use CMEMS data to describe environmental conditions and study their correlation to the concentration of DO. Knowing this relation, in-situ observations of DO are used to train a machine learning regression model.
For the present study, we developed an approach for DO estimation based on a Support Vector Regression machine learning model, using CMEMS data and in-situ observations. A similar approach was presented by Guo et al. [2], however the study was focused on monthly and annual data, which does not correspond to aquaculture needs. Our in-situ dataset includes daily measures of DO from 3 different aquaculture sites (Agrilia Lesvos, Souda Crete, Astakos Ionian Sea), collected from May 2021 to October 2021. 80% of the dataset was used as training sample and the remaining 20% was used as test sample. For every set of coordinates an array was created including the values of chlorophyll-a and sea surface temperature, as well as the values of one, two and three days prior to the sampling.
Our preliminary results are promising and indicate a strong correlation between sea surface temperature and dissolved oxygen. The predicted and the measured values differ less than 3% in most cases. The lack of extreme incidents during the study period is more challenging for the prediction, since the algorithm is only trained for a limited range of values. When the values exceed the normal range the differences between the predicted and measured values are higher. Expanding the dataset to include more cases is essential for proper training of the algorithm and, consequently, for improving our results.

[1] E. Batur and D. Maktav, “Assessment of Surface Water Quality by Using Satellite Images Fusion Based on PCA Method in the Lake Gala, Turkey,” IEEE Trans. Geosci. Remote Sens., vol. 57, no. 5, pp. 2983–2989, May 2019.
[2] H. Guo et al., “A generalized machine learning approach for dissolved oxygen estimation at multiple spatiotemporal scales using remote sensing,” Environ. Pollut., vol. 288, p. 117734, Nov. 2021.
[3] N. Karakaya and F. Evrendilek, “Monitoring and validating spatio-temporal dynamics of biogeochemical properties in Mersin Bay (Turkey) using Landsat ETM+,” Environ. Monit. Assess., vol. 181, no. 1–4, pp. 457–464, Oct. 2011.
[4] Y. H. Kim et al., “Application of satellite remote sensing in monitoring dissolved oxygen variabilities: A case study for coastal waters in Korea,” Environ. Int., vol. 134, no. October 2019, p. 105301, 2020.

Ocean fronts are moving ephemeral biological hotspots forming at the interface of cooler and warmer waters. In the open ocean, this is where marine organisms, ranging from plankton to mesopelagic fish up to megafauna, gather as well as where most fishing activities concentrate, generating bycatches and conflicts such as depredation. Fronts are critical ecosystems and understanding their spatio-temporal variability is essential not only for conservation goals but also to ensure sustainable fisheries. Frontal activity is thus an integrated indicator of the health status of oligotrophic oceanic systems.
The pelagic realm of the Mozambique Channel (MC) is an ideal laboratory to study ocean front variability due to its energetic turbulent flow and its high biodiversity, currently under consideration of protection (marine protected areas).
We use satellite Sea Surface Temperature (SST) from GHRSST Level 4 Multi-scale Ultra-high Resolution (MUR) 0.25deg Global Foundation Sea Surface Temperature Analysis v4.2 and Sea Surface Height (SSH) from Global Total Surface and 15m Current (COPERNICUS-GLOBCURRENT) from Altimetric Geostrophic Current and Modeled Ekman Current Reprocessing. The surface perspective provided by these remotely-sensed data is complemented by analyses at greater depths using high-resolution nested ROMS model output in the MC. Based on horizontal thermal gradients computed with the Belkin and O’Reilly Algorithm and dynamical frontal structures derived from the Finite Size Lyapunov Exponents (FSLE), we define strong and moderate fronts based on their intensity. Moderate thermal fronts are uniformly spread across the MC, whereas stronger thermal fronts are more prevalent in the southern MC and along continental shelves, shedding light on frontogenesis processes. The total front count in the MC does not exhibit much seasonal change, but moderate fronts are more frequent in austral summer and autumn while stronger fronts occur during austral winter and spring. Using Empirical Orthogonal Functions and searching for relationships between fronts statistics and climatic indices (e.g. Indian Ocean Dipole, El Niño Southern Oscillation, etc.), our results suggest that a substantial part of MC frontal activity is controlled by basin-scale teleconnections while local dynamics do also play a substantive role.
Additionally, our results suggest that current satellite SST products do not reveal as much frontal activity as the one estimated by dynamical structures derived from satellite SSH. This difference apparently stems from the artificial lack of small-scale structures in oceanic temperature fields due to both blending procedure and cloud-cover-induced corrections. Improved SST products (e.g. using smaller/shorter length/time scales for the merging procedure and improving the flagging criteria) would allow a finer comparison between tracer-based and dynamically-based front detection methods.
Ongoing analyses combining our multidecadal front database with a compilation of historical biological data in the region will bring critical insight to understand how marine life interact with ocean fronts in a context of global change and how frontal variability can be used as an ocean health indicator worldwide.