MedEOS is an application development, funded by the European Space Agency as a part of the Mediterranean Regional Initiative. Its main objective is to develop and produce high-resolution, gap-free water quality products based on Earth observation (EO) data. This is achieved through combining the high temporal resolution of Sentinel-3 Ocean and Land Colour Imager (S-3 OLCI) and the high spatial resolution of Sentinel-2 Multispectral Instrument (S-2 MSI) in a process of data fusion.
The project is organized in two distinct phases that make up to 2-years of project lifetime (2021 - 2023). At the end of the first year a full set of demonstration products shall be delivered comprehending 1-year datasets over 5 different test areas around the Mediterranean coasts. In the second year, 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.
Five pilot areas in Egypt, France, Greece, Spain, and Tunisia were selected following specific criteria to maximize the benefit and impact of the project in the scope of the Mediterranean area. They are spread across the Mediterranean exhibiting a variety of biophysical, environmental, and socioeconomic conditions, which are formed under a varying geopolitical and cultural context. This diversity allows the consortium to test the validity of the provided services at different levels:
- scientific and technical - by providing very distinct regions with requirements that cover all services and unique validation conditions;
- technical - service operations and delivery will need to adapt to the different needs and contexts of the end-users, providing extra flexibility and resilience to the overall system;
- user uptake - institutional and societal realities in the different countries will provide an in-depth analysis of the capacity of user uptake at various prevailing conditions.
Three different sets of products are developed in scope of the MedEOS project. The first group consists of five EO directly derived water quality products: Total Suspended Matter (TSM), Turbidity, Chlorophyll-a concentration (Chl-a), Secchi Depth and Colored Dissolved Organic Matter (CDOM). TSM is a key element of water quality in coastal areas. It is a well-known parameter in ocean color applications. A method using a single band, deemed robust for coastal turbid waters, is considered for this project’s application. Turbidity is a water quality feature closely related to TSM. The Turbidity Product is designed according to the ISO 7027 definition: a quantitative measurement of “diffuse radiation” expressed in Formazin Nephelometric Units (FNU). Chl-a concentration is the main pigment in phytoplankton, and a key element in ocean-color applications. Phytoplankton is responsible for primary production through photosynthesis, and an indicator of the natural processes in the water environment. Secchi Depth is used to measure water transparency in the ocean. The Secchi depth is the depth at which the Secchi disk - being lowered in the water column - is no longer detectable by a human observer from the water surface. CDOM is a key element of water quality in coastal areas, and a well-known feature in ocean color applications. Its presence is mostly determined by freshwater outflow, and it can be used as a proxy for dissolved oxygen. All these products are originally generated independently using Sentinel-2 and Sentinel-3 data.
In the next step, data fusion methods are used to generate everyday products with a high spatial resolution, closer to that of Sentinel-2. When needed, this step is preceded by gap-filling to reduce the impact of cloud cover, which can affect the fully operational application of the data fusion methods. The data fusion techniques employed in MedEOS are based on calibrating a model between a pair of fine and coarse resolution images acquired at the same date in the past and applying this model on a coarse resolution image acquired in the present to derive a synthetic fine resolution image of the current situation. In the context of land application, data fusion is usually applied on the ground reflectance level. However, spatiotemporal data fusion of Sentinel-2 and Sentinel-3 data in ocean color applications have to adapt existing approaches tested and validated in land applications to the much more dynamic context of water. Applying data fusion at EO direct products, instead of water reflectance, ensures a better comparability between both satellite sensors and guarantees the use of the best state-of-art EO direct product algorithm for each of them.
The resultant daily high resolution EO direct products are then used to generate the second set of MedEOS products - EO indirectly derived water quality products. This group contains the following products: Faecal Bacterial Contamination Indicators, Eutrophication Indicators, Harmful Algal Blooms, and Global Environmental Anomaly Detection. Faecal bacterial contamination indicators depend on continental inputs (e.g., river discharges, sewer system outflows) and on bacteria resilience capacity in the coastal ocean. Three parameters are used to monitor the risk of coastal waters contamination: Faecal Bacteria Decay Rate (T90); Faecal Bacteria Vulnerability Index; and Local Bacterial Concentrations. Eutrophication indicators are highly conditioned by nutrient abundance, therefore phosphorus and nitrogen concentrations are extracted from Copernicus biogeochemical model for the Mediterranean Sea. Derived concentration using multiple regression models based on image spectral bands and other EO direct products (Secchi Depth and Chl-a) and/or provided by in-situ measurements if available will be considered. These concentration estimates are to be used together with Chl-a to compute the Eutrophication Index. Chlorophyll-a concentration derived from EO datasets and/ or the Eutrophication index are used to narrow down areas where Harmful Algal Blooms (HABs) could be detected. Spectral indices designed for specific HABs communities are further employed to derive either qualitative or, when possible, quantitative information about the intensity of the blooms. Based on the multi-feature datasets obtained from both EO direct and indirect products, a global model will be developed to automatically detect anomalies in coastal water quality, which shall indicate probable anthropogenic pollution.
The third MedEOS set of products is dedicated to river plume monitoring. Rivers are one of the main conveyors of land-based pollutants into nearshore and coastal water. To support the monitoring and assessment of river plume impact on coastal water quality, a river plume monitoring dataset will provide a systematic detection of plumes related to major rivers discharging freshwater into the Mediterranean basin. The methodology is also applicable to large plumes that can be generated by urban sewer system outflows (e.g., major marine outfalls), which can largely affect coastal water quality. The detection includes the plume spatial delineation, out of which a set of characteristic features are evaluated, e.g., plume extension, orientation, extent and concentration threshold. The MedEOS algorithm produces this information by using multiple inputs, which should deliver a signature related to the river or sewer system plume. The main input EO products considered for now are Sea Surface Temperature, turbidity, Total Suspended Matter, Chl-a and Sea Surface roughness. Synthetic Aperture Radar (SAR) imagery can be also used to provide plume extension, because e.g., runoff plumes are associated with specific surface slick characteristics. Moreover, the algorithm is also applicable with numerical model outputs; thus, offering perspectives to complement the daily tracking obtained from EO products by higher frequency (1 h to 1 day) tracking from ocean models. User requirements related to the river plume monitoring service are populated following dedicated activities of the project and as a result of the dissemination and communication campaign of MedEOS.
All the services described above will be implemented in services4EO, Deimos EO Exploitation Platform solution. This platform will hold all service development, integration, deployment, delivery and operation activities. Its design and deployment will be driven by the need to come up with services that are easily tailored to the real operational conditions, accepted by the users, and become a constituent element of the users’ business as-usual working scheme.
For more information please visit the MedEOS website: https://medeos.deimos.pt/. This project has received funding from the European Space Agency Contract No. 4000134062/21/I-EF.
Sea surface salinity (SSS) represents one of the Essential Climate Variables (ECVs) defined by the Global Climate Observing System (GCOS). Ocean circulation, climate variability and water cycle are deeply impacted by salinity variations. Moreover, salinity is strongly affected by freshwater input from rivers and land run-off as well as atmospheric forcings, as precipitation and evaporation. Defined as a concentration basin, the Mediterranean Sea represents a hot spot for the characterization of salinity variability, requiring dedicated efforts to monitor and understand ongoing changes and their potential impacts at regional and global scales.
Nowadays, an increased number of moorings and floating buoys provide accurate SSS measurements in the Mediterranean Sea. However, in situ data have poor coverage in time and space, hindering the monitoring of SSS patterns and their space-time variations and trends. Conversely, satellite remote sensing provides SSS surface data at high spatial and temporal resolution, complementing the sparseness of in situ dataset.
Here, we describe an improved configuration of the multidimensional optimal interpolation algorithm originally proposed by Buongiorno Nardelli et al. (2012; 2016) and Droghei et al., (2016; 2018), specifically designed to provide a new daily SSS dataset at 1/16° grid resolution covering the entire Mediterranean Sea (hereafter Med L4 SSS). Two main improvements have been introduced in this regional algorithm: the inclusion of remotely-sensed salinity observations data from multiple satellite missions [i.e. NASA’s Soil Moisture Active Passive (SMAP) and ESA’s Soil Moisture and Ocean Salinity (SMOS) satellites] and a new background (first guess) field built by blending a Mediterranean in situ monthly climatology close to main rivers’ outflow, with the upsized and temporally interpolated weekly global SSS product distributed by the Copernicus Marine Environment Monitoring Service (CMEMS).
The multi-sensor regional SSS dataset has been validated against independent in situ SSS observations collected in the Mediterranean Sea between 2010-2018 and also compared with global weekly CMEMS product and Barcelona Expert Centre (BEC) regional product. The statistics of validation highlighted an improved performance of the new Med L4 SSS. The results also demonstrated that the use of a background blending in situ monthly climatology and CMEMS SSS L4 weekly product determines a reduction of the SSS errors along the coast. A power spectral density analysis highlighted that, among all the datasets, the Med L4 SSS field achieves the highest effective spatial resolution.
Mediterranean hurricanes (Medicanes) are synoptic-scale cyclones typical of the Mediterranean area which share some dynamical features with the well-known tropical cyclones during their mature stage, even if they are smaller in size: the presence of a quasi-cloud-free calm eye, spiral-like cloud bands elongated from the center, strong winds close to the vortex center and a warm core. The most intense Mediterranean Hurricane (Medicane) on record, named Ianos, swept across the Ionian Sea between 14 and 18 September 2020, affecting Southern Italy and especially Greece and its Ionian islands, where torrential rainfall and severe wind gusts caused widespread disruption, landslides, and casualties. In this study, satellite measurements from the NASA/JAXA Global Precipitation Measurement Core Observatory (GPM-CO) active and passive microwave (MW) sensors are used to analyse the precipitation structure of Ianos. Two GPM-CO overpasses, available during Ianos development and tropical-like cyclone (mature) phase, are analysed in detail. GPM Microwave Imager (GMI) measurements are used to carry out a comparative analysis of the medicane precipitation structure and microphysics processes between the two phases. The GPM-CO Dual-frequency Precipitation Radar (DPR) overpass, available for the first time during a medicane mature phase, provided key measurements and products to analyse the 3D precipitation structure in the eyewall and in the rainbands, offering further evidence of the main precipitation microphysics processes inferred from the passive MW measurement analysis. Substantial difference in the rainband precipitation structure and microphysics processes is observed between the development and the mature phase. The inferred deep convection features (updraft strength, graupel growth, presence of supercooled droplets) are related to the observed change in lighting activity between the two phases. During Ianos mature phase, the GPM-CO measurements provide evidence of weaker updraft and limited graupel size which, combined with the strong horizontal wind, inhibits cloud electrification processes, while shallow convection/warm rain processes are observed in the inner region of the eyewall. The study demonstrates the value of the GPM-CO not only to characterise medicane Ianos precipitation structure and microphysics processes with unprecedented detail, but also to provide evidence of its similarities with tropical cyclones.
The Mediterranean region has been referenced as one of the most responsive regions to climate change. The last report from the International Panel on Climate Change (IPPC), and the First Mediterranean Assessment Report of the Mediterranean Experts on Climate and environmental Change (MedECC) highlight the Mediterranean as one of the most vulnerable regions in the world to the impacts of global warming. Climate change impacts, already effective in the region are bringing with them a number of challenges in terms of agriculture sustainability and food security. Climate events are having an increasingly extreme connotation, extreme agro-meteorological events such as droughts, heatwaves or rainstorms are expected to occur with higher frequency in the coming years, making agricultural systems more fragile and increasing the inter annual variabilities of crop production and grain quality.
In this context, enhanced cooperation and information sharing in the field of agriculture became an asset among Mediterranean countries to prevent food insecurity and social instability, but also to deal with commodities market distortions, as for example unjustified price volatilities.
Earth observations (EO) data are key inputs into all the analyses and decision-making processes that are critical to guarantee food security and cropping systems resilience to recurrent agro-meteorological stress. Moreover, the combination and integration of EO derived indicators with other information layers such as agro-meteorological indicators or specific agronomical expertise provide an increasing potential to monitor the seasonal response of agricultural systems to climatic stressors.
CIHEAM is an intergovernmental organization devoted to the sustainable development of agriculture and food and nutrition security around the Mediterranean Basin. It is composed of 13 Member States (Albania, Algeria, Egypt, France, Greece, Italy, Lebanon, Malta, Morocco, Portugal, Spain, Tunisia and Turkey). In September 2012, the Agriculture Ministers of the 13 CIHEAM members, recommended that CIHEAM countries should "contribute, in close collaboration with G20 follow-up group, to the development of an information system on Mediterranean markets linked to AMIS (Agricultural Market Information System) and to share information so as to fight prices volatility within agricultural markets".This led to the creation of MED-Amin, MEDiterranean Agricultural Market Information Network (http://www.med-amin.org/en/home/), which was officially endorsed by agriculture ministers of the 13 countries in February 2014. Hosted at CIHEAM Montpellier, the MED-Amin network gathers representatives from Agricultural Ministries, statistical services and Cereals Offices from CIHEAM member states, with the aim to cooperate and share information among the national information systems on cereal markets.
Since 2017, in collaboration with the Joint Research Center of the European Commission, the MED-Amin network has developed a pilot action for monitoring crop conditions of cereals (common wheat, durum wheat and barley) taking into consideration the exploitation of remote-sensing indicators, agro-meteorological variables and quality statistical data, to provide a robust indication of the shocks (positive or negative) to be expected on future harvests.
The present contribution introduces the qualitative crop forecasting activities carried out in the frame of MED-Amin across the 2021-2022 winter crop campaign. Emphasis will be put on (i) the synergy in the use feedback from national contact points and earth observation information to derive seasonal hot spots of winter crops production and on (ii) the participative approach the partners are adopting which is made up of seasonal information exchanges between the MED-Amin Secretariat and its focal points.
The MED-Amin forecasting exercise is made of four main stages:
- A statistical data collection (MED-Amin baseline) in February, gathering area, yield and production records crop-wise and region-wise for all the thirteen partners.
- A first pre-screening analysis in March of EO derived biophysical indicators, highlighting the main negative and positive biomass accumulation hot spots on wheat and barley crops. This information is provided to the focal points for a first feedback from the field.
- A second round of pre-screening analysis takes place in May, where the previous remote sensing analyses are updated and shared again to be cross-validated.
- Results are consolidated in June based on crop conditions reports from the focal points to capture the final crop evolutions before harvest.
Three bulletins are jointly developed and disseminated through the network and the media along the crop season.
Results are discussed in view of improving information on cereal markets (production, utilization, stocks, prices, trade) within the Mediterranean region and for tending to real-time transmissions of early warnings in a context of fragility due to climate change stressors (e.g. lack of hydro-meteorological events) and global prices volatility. It illustrates the necessity to mobilize human expertise in the field for ground observations. Obtaining data on precise development stage calendars and/or calibrating indicators and models require cooperative actions between ground experts and technical services specialized in the analysis of satellite and meteorological data and using reliable and easy-to-use indicators.
The project
Soil sealing – also called imperviousness – is defined as a change in the nature of the soil leading to its impermeability. Soil sealing has several impacts on the environment, especially in urban areas and local climate, influencing heat exchange and soil permeability; soil sealing monitoring is crucial for the Mediterranean coastal areas, where soil degradation combined with drought and fires contributes to desertification.
Some artificial features like buildings, paved roads, paved parking lots, and other artifacts can be considered to have a long duration. In general, these land cover types are referred to as permanent soil sealing because the probability of coming back to natural use is low. Other land cover features included in the definition of soil sealing can be considered reversible. For them, the probability of coming back to natural use is higher. The land cover classes that are included in the reversible soil sealing have been defined with the users of the project, and include solar panels, construction site in early stage, mines and quarries, long-term plastic-covered soil in agricultural areas (e.g., non-paved greenhouses).
The project Mediterranean Soil Sealing, promoted by the European Space Agency (ESA) in the frame of the EO Science for Society – Mediterranean Regional Initiative, aims to provide specific products related to soil sealing, its degree and reversible soil sealing over the Mediterranean coastal areas by exploiting EO data with an innovative methodology capable to optimise and scale-up their use with other non-EO data. Such products have to be designed to allow – concerning current practices and existing services – a better characterisation, quantification and monitoring within time of soil sealing over the Mediterranean basin, supporting users and stakeholders involved in monitoring and preventing land degradation. The project started in March 2021, will produce the first results in March 2022 and the final products in March 2023.
The targeted products are high-resolution maps of the degree of soil sealing and the reversible soil sealing over the Mediterranean coastal areas (within 20km from the coast) for the 2015-2020 time period, at yearly temporal resolution with a targeted spatial resolution of 10m.
The team
The project team is led by Planetek Italia, and composed by ISPRA and CLS.
Planetek Italia is in charge of the development of the infrastructure, the engineering of the algorithms and the communication activities. CLS is in charge of the soil sealing mask and of the experimental reversible soil sealing processing algorithms, ISPRA of the soil sealing degree processing algorithms. The interaction with the users is led by ISPRA, institutionally involved in the land degradation theme into international and regional organisations and the national body responsible for the theme in Italy.
Methodology
Introduction
The general workflow for the production of Ulysses products is shown in Figure 1.
The processing chain is split into four parts: Pre-processing; Soil Sealed Masks Production; Permanent/reversible Soil Sealing Production; Computation of the Soil Sealing Degree.
In the pre-processing, L2A and L3 images are derived from L1C Sentinel 2 data, while, from the Sentinel 1 acquisitions, the backscatter images and the coherence images are derived. The core of the processing chain is the second step in which AI algorithms are applied to derive the soil sealing mask: a binary image in which artificially covered pixels are identified. A refinement step is required to improve the quality of the mask. In parallel to the production of the sealing mask, the identification of Reversible Soil Sealing is performed. In the final step, the soil sealing degree is computed.
The soil sealing mask and reversible sealing processing workflows
The project uses as optical source Sentinel-2 data Level 1C. All the L1C images are corrected to level 2A corresponding to Top Of Canopy calibration using MAJA processor. A final pre-processing consists of producing the level 3A cloud-free composite for each month from the previous level 2A. At this point from level3A composite images a several numbers of indices are computed: NDVI, NDTI & NDBI. From level2A are derived the PANTEX band for each cloudless acquisition date. PANTEX is a texture descriptor very powerful specially to describe urban areas. The processing of Sentinel-1 SLC data consist of producing coherence data as well as calibrated and orthorectified sigma naught product from which will be derived monthly mean maps. Then texture elements, e.g Mean and Variance, will be derived from each mean map.
Pre-processed data are supplied to a machine learning tool called “Broceliande” along with the set of in-situ data for extracting general soil sealing mask or to discriminate permanent from reversible impervious object. “Broceliande” is mainly based on 2 pieces of software, Triskele used to produce hierarchical representations of images, and Shark library used for Machine Learning. From the set of selected bands, we will build multiscale representations through the model of morphological trees from which we derive multiscale features called attribute profiles (“AP/DAP generator” stage). Such trees can also be seen as a stack of nested segmentations and thus as a generalization of the concept of mono-scale segmentation layer in GEOBIA tools. For each hierarchical representation, we will measure specific attributes (e.g., area, weight, compactness…) for all objects appearing at different scales for a given pixel. These features are thus assigned to each pixel. The next step is the use of the shark random forest classifier. Due to its relatively simple parameterization, computation efficiency, and high accuracy, we will use as a basis the Random Forest (RF) classification algorithm based on its proven experience.
The soil sealing degree
The objective of the developed methodology is to compute, for each pixel of the soil sealing mask, the degree of soil sealing as the fraction of pixel area covered by artificial surfaces. The estimation of soil sealing degree (in the range 1-100%) at subpixel level (10m spatial resolution) is challenging because of the mixed pixel and spectral similarity between natural soil and artificial surfaces. Therefore, the methodology was designed to be suitable for the Mediterranean coastal areas through automatic processing of Sentinel data.
The methodology is based on the NDVI calculation using Sentinel-2 L2A time series and exploits the correlation between the derived quasi-maximum NDVI and the soil sealing degree. The aim of using a long time series is to remove fluctuations due to the seasonality of vegetation that can partially cover the sealed areas. After several tests, Sentinel-1 GRD data were excluded from the processing because of the lower spatial resolution (compared to Sentinel-2) that makes the computation of soil sealing degree at subpixel level difficult and less reliable.
The method computes soil sealing degree performing the NDVI calibration assuming a linear relation with soil sealing degree, similarly to the method described by Gangkofner et al. (2010); the main advantage of this approach is that it does not require any training input, but only the definition of a minimum and maximum value of NDVI related respectively to 100% and 1% soil sealing degree. Moreover, a method for the automatic estimation of the calibration parameters from the quasi-maximum NDVI raster was developed in order to adapt the correlation to the different vegetation types and phenology characteristics of the Mediterranean coastal areas.
The innovation
Project’s innovations reside in multiples aspects. First, products are generated over all the Mediterranean basin for a period of 5 years. Considering the update frequency, the extent of the production area and the thousands of images involved this is a major innovation. Second, innovation resides in the technique applied for the soil sealing extraction by using “Broceliande” tool where is nested the hierarchical tree approach in previous of the classic classification operation. This innovation was the subject of two published papers :Merciol & al, GEOBIA at the Terapixel Scale: Toward Efficient Mapping of Small Woody Features from Heterogeneous VHR Scenes, 2019 and Merciol & al, Broceliande: a comparative study of attribute profiles and features profiles from different attributes, 2020. Third, the project aims at producing in an efficient and automatic fashion the soil sealing degree. The innovation of the developed methodology is related to the annual update of the map, fostering the consistency of the NDVI time series at pixel level. Moreover, the processing related to the calculation of the NDVI quasi-maximum and the definition of the calibration parameters is a major result of the development phase.
Olive trees are a drought-resistant species traditionally grown in the Mediterranean Basin, where 95 percent of global olive cultivation is located. In the last three decades, super high density (SHD) planting systems have been taking over the olive sector worldwide, since they allow mechanization of the pruning and harvest processes, and thus reducing costs. Furthermore, when irrigation practices are adopted, these so-called hedge-row plantations can reach highest yields compared to other planting systems. However hereby, the limited water resources of many drought-prone areas dedicated to olive cultivation will be faster exploited and depleted, presenting not only an environmental risk but also affecting the local population dependent on such water resources. At the same time, a surge in olive oil production may lead to higher amounts of residues, which if wrongly disposed of, can put additional pressure on the limited water resources.
In the last ten years, Moroccan olive production has more than doubled, placing the country among the five main producers worldwide. The leading producing region in the country is the Fès and Meknès Region. Here, the Saïss plain has been subject to several sociological and environmental studies, due to an ever-increasing risk of groundwater depletion. Against this background, our study aimed to assess with the help of Remote Sensing data and methods whether an intensification process has occurred in the Saïss area, as well as to evaluate the extent and dimensions of the potential impact on water resources .
In our study we worked with high resolution (HR) Rapid Eye and Planet Scope imagery of 5 and 3 metres resolution, respectively, acquired via PlanetLabs’ Education and Research Program. Due to quota limitations, only two seasonally representative HR images per year were used to extract SHD olive plantations in 2010 and 2020. For this, an unsupervised approach was developed, consisting in applying an adapted form of hierarchical clustering using a two clusters k-Means algorithm. This allowed discriminating SHD olive plantations without requiring any labelled data. Furthermore, incorporating cloud-based geospatial computing in Google Earth Engine allowed a very low computational cost using HR data to an extent of 140.000 ha.
For accuracy assessment, a supervised land use land cover (LULC) classification approach with 2020 Sentinel-1 and -2 imagery was used adopting methods based on recent findings from the literature on tree crop and land use mapping in semi-arid regions. Besides evaluating the unsupervised approach, this second step reached an overall accuracy of 89.9% and identified other land use classes with high crop water requirements in the study area and allowed addressing the main land use conversions from SHD olive plantations between 2010 and 2020.
The main findings in this study highlighted the importance of using multispectral information over vegetation indices and the good performance of high-resolution imagery for olive mapping despite of lower temporal resolution. The results of the analysis also confirmed that despite of a considerable number of new SHD plantations in the study area, there was a comparable number of abandoned orchards and land use conversions within the last ten years. Thus, the intensification process in the study area had been counterbalanced by the degradation of old plantations, and the land use change to annual crops and urbanization.