Skip to main content

Improved monitoring and forecasting of ecological status of European INland waters by combining Future earth ObseRvation data and Models

Final Report Summary - INFORM (Improved monitoring and forecasting of ecological status of European INland waters by combining Future earth ObseRvation data and Models)

Executive Summary:
The 4-years EU FP7-SPACE INFORM project started on 1 January 2014 with the goal to develop novel and improved user-driven products for inland water quality (WQ) monitoring by using innovative remote sensing methods integrated into biogeochemical models which fully exploits the improved spectral, spatial and temporal capabilities of new and upcoming Earth Observation (EO) missions (Sentinel-2, Sentinel-3, EnMAP and PRISMA).
The INFORM project focused on the development of algorithms for new innovative products such as stratification, yellow matter and phytoplankton functional types. Through the collection of an impressive dataset of field measurements, the algorithms have been tested and validated on a diversity of inland waters and rivers. The methodology has been well described and the project resulted in 18 peer-reviewed scientific publications.
The INFORM dataset gathered in the project is unique in the sense that it contains a very large amount of field data, often collected simultaneously with an overpass of optical satellite sensors. This is crucial data for calibration and validation of remote sensing algorithms. The dataset will be made publicly available through the LIMNADES database and will be a basis for future research.
The INFORM project strongly interacted with its End users and was able to create awareness on the different products and their applicability for inland water monitoring. The end users included universities, governmental institutions, public administrations and industry. Next to meetings with the End User Advisory Board, many face-to-face meetings have been organized with national policy makers to create awareness on remote sensing products. Several product flyers have been produced within the INFORM project to clearly communicate about the new developments to the users.
Finally, INFORM provided recommendations for future satellite missions in a dedicated report (D8.9 ‘Report on recommendations for inland water quality monitoring for future satellite missions).
Project Context and Objectives:
The European inland waters (estuaries, rivers and lakes) are used for many purposes such as fishing, recreation, water supply, transport, waste disposal, irrigation, ... The growing number of users and uses, the increasing population, the industrialization, the intensified land use have increased the pressures on Europe’s inland waters requiring a sustainable water management. As a consequence the inland waters are being monitored in an ever increasing manner. A number of legislations aiming to protect and/or to improve the quality of the water have been constructed. Since December 2000 the European Water Framework Directive (WFD, 2000/60/EC) forms the legislative framework for the water management undertaken by the EU Member States aimed at a “good surface water status” for all European water bodies by 2015 both with respect to the ecological status (healthy ecosystems) and the chemical status (low pollution). The EU Habitats Directive, the Shellfish Waters Directive, the EU Drinking Water Directive and the Bathing Water Directive, are established to protect specifically unique and valuable habitats, drinking water resources, and ensure healthy bathing water. Likewise, environmental regulations for e.g. dredging activities (e.g. fairway maintenance) and aquaculture are becoming very rigorous and ask for a detailed Environmental Impact Assessment (EIA) and water quality monitoring to evaluate changes into the system.
Notwithstanding the growing need for environmental monitoring of inland waters, less and less resources are available for in-situ monitoring. Therefore Earth Observation and the integration of Earth Observation data into models could be a very valuable tool to complement these in-situ measurements as it allows for systematic, synoptic observations and forecasting. At the start of the INFORM project, we realized that Earth Observation was underutilized and its value was not compellingly elucidated to the end-users. One of the reasons was the complexity and variability of these inland waters. Inland waters range from extremely turbid to shallow and clear macrophytes dominated waters. They are characterized by a large variability in phytoplankton composition and organic matter. Furthermore, the lack of adequate analysis methods allowing to deal with this complexity, the lack of adequate low-cost EO data at that time for the often small or irregular shaped inland water bodies and the lack of uncertainty estimates on the provided EO products have all contributed to this limited use.
On the other hand positive expectations were given for the very near future through the full utilization of forthcoming sensors with improved spatial, spectral, and temporal resolution. There was clearly the necessity of a continuous effort towards the end-user in order to better understand their needs and to develop algorithms for the future sensors which meet their product expectations with as final goal the use of EO data derived products as ‘legal instrument’ for the implementation of environmental directives. In addition, there was the need to couple these remotely sensed data with aquatic ecological models (also known as model-data fusion or assimilation). Whereas the current status of the water body can be assessed with EO, assimilation of EO products into biogeochemical models allows for analysis of the cause-effect relationships governing a status change, forecast the response to pressures and evaluate different management actions.
New EO data sources
The capability of retrieving water quality parameters such as chlorophyll-a (Chla) with ocean colour satellites has been widely proven and published during the last decade. However, 5 years ago, Earth Observation monitoring efforts were driven by ocean colour satellites such as MODIS and MERIS with rather low spatial resolution, medium spectral resolution and global coverage, and were restricted to open ocean, coastal waters and the largest inland waters. At the start of the project, several new satellites (e.g. Sentinel-2, Sentinel-3, EnMAP/PRISMA) were to be launched providing a wealth of new data at increased spatial, spectral and temporal resolutions. Although not all conceived as being ocean colour missions (e.g. Sentinel-2 has been designed to support GMES land, emergency and security applications), it was clear that the inland water community could benefit considerable from these new data sources:
• A higher spatial resolution extends the monitoring efforts to cover many lakes which are small and have irregular shapes and could not be monitored with the existing missions.
• Higher spectral resolution and new spectral bands could solve problems with atmospheric correction in inland waters and could lead to new products which better serve the user needs and interfacing with biogeochemical models.
New analysis methods and products
The optical and physical complexity of inland waters and the user demand for new and improved products necessitated not only the need for better EO data sources but also required more sophisticated algorithms.
The complexity
The algorithm and product development for existing ocean colour satellites was mainly driven by the ocean and marine community. Applications of these algorithms to rivers and lakes often failed due to their optical and physical complexity which is attributed to the adjacent land. Extension of the spectral coverage of these new EO sensors to both shorter and longer wavelengths allows for an improvement of existing algorithms (e.g. for atmospheric correction, yellow matter retrieval) making them more robust and more generally applicable. Furthermore, through a better knowledge of the spectral signature, the inherent optical properties of inland water bodies and their variability, it was expected that algorithms can be tailor-made to the specific characteristics of inland waters.
The new/improved products
A suite of new products such as phytoplankton composition, light availability for photosynthesis and macrophyte abundances could be retrieved from the new sensors due to an increase in spectral resolution and spectral range. Products could be made available with a higher spatial detail due to an increase in spatial resolution. The increase in information both spatially and spectrally asks for new analysis methods. Innovative analysis techniques developed by the image processing community but which are not yet (fully) taken up by the ocean colour community should be considered for the analysis of the continuous spectrum offered by the upcoming hyperspectral satellites.
Better link with models
While in the ocean community Earth Observation is often used as a standalone tool for water quality information, for most of the inland waters, with the increasing monitoring pressure through European and national legislations, a bulk of in-situ measured and model derived water quality data is available. To optimize the benefits of these complementary sources of information, proper integration of EO and in-situ data in these models is needed. While in the past essential model parameters such as light extinction, stratification of the water body, phytoplankton composition were taken from labour-intensive field measurements at a few sample points and thereby missing its spatial variability, these products are derived in INFORM from new EO satellites providing a full spatial coverage which substantially improve the performance of biogeochemical models.
A user-driven approach
INFORM was driven by the needs of end-users and the scientific community. They both expressed their needs for better availability and improved products and methods for inland water quality monitoring. Therefore both the end-users and the scientific community were closely involved in the project and suggestions made were implemented by the project team. An end-user committee (EUC) and scientific advisory committee (SAC) formed an integral part of the project.
The main objective of the project was:
To explore and demonstrate how the new capabilities of new sensors, combined with innovative analysis techniques and the coupling with biogeochemical models can be exploited to deliver new and improved products for inland water quality monitoring addressing better the end-user demands.
The project will develop methodologies to improve existing Copernicus services and will provide a basis for new innovative Copernicus service products. This will ensure exploitation of upcoming and future space missions to the greatest possible extent for the monitoring of European inland water quality.
The sub-objectives of the INFORM project were:
To improve atmospheric correction on the basis of the exploitation of SWIR bands (on Sentinel-2, Sentinel-3, and EnMAP/PRISMA) for turbid waters.
2. To improve suspended sediment retrieval algorithms for highly turbid waters based on exploitation of the SWIR bands on Sentinel-3 and EnMAP/PRISMA.
3. To develop, test and validate algorithms for automatic retrieval of the different main phytoplankton functional types and their percentage in the total biomass from hyperspectral spaceborne sensors.
4. To develop a novel algorithm to estimate the absorption coefficient of yellow matter (both in dissolved and particulate form) using hyperspectral data in the UV-visible spectral region and to apply this algorithm to simulated OLCI data on Sentinel-3.
5. To develop innovative methods to discriminate between macrophyte species and to determine their abundance on the basis of multi-temporal hyperspectral spaceborne data such as EnMAP/PRISMA.
6. To develop and validate EO-based lake primary production algorithm exploiting the spatial and temporal dynamics observed in water bodies with Sentinel-2 and Sentinel-3 observations, combined with appropriate models.
7. To develop novel approaches to study the stratification of the water body on the basis of hyperspectral satellite and in-situ data.
8. To develop algorithms to derive light extinction from EO data which is one of the driving parameters in many water quality models.
9. To facilitate the calibration and validation of water quality biogeochemical models for the complex interactions between temperature, nutrients, light availability, phytoplankton and macrophytes.
10. To demonstrate to the end-users for pilot sites these INFORM EO-products and the improved performance of the biogeochemical modelling.
11. To make recommendations on new EO techniques and additional bands for next generation EO sensors.
Project Results:
End user requirements
To collect the end-user requirements, an INFORM End-User Advisory Board (EUAB) was set-up. The INFORM EUAB was composed of 8 members of which each member was proposed by an INFORM partner. Different typologies of end-users (e.g. university, governmental institution, public administration, industry) operating on different aquatic ecosystems were selected in order to collect a broad range of end-user requirements.
In the table below (Table 1) the EUAB member organization, representative, country and INFORM partner linked to the EUAB member is summarized.
Member Number
Linked INFORM partners 1 International Marine and Dredging Consultants (IMDC) Marc Sas, Boudewijn Decrop Belgium VITO
University of Parma, Life Sciences Department
Marco Bartoli
CNR 3 Rijkswaterstaat Ute Menke Netherlands Deltares
Laboratory of Central-Transdanubian Water Directorate
István Kóbor
MTA OK 5 UK Environment Agency Geoff Phillips United Kingdom U STIRLING
Alfred Johny Wüest
7 Marine Research
Department of the
Ministry of
8 Institute for Lake
Research, LUBW
Thomas Wolf Germany EOMAP
Table 1: List of End-User Advisory Board members.
A series of initiatives were carried out to identify the end-users requirements, starting from questionnaires,
followed by two End-User Advisory Board meetings (EUAB01, 20-21 March 2014, CNR, Venice, Italy; EUAB02,
17 March 2016, Brussels, Belgium) . The latter were organized to stimulate the interaction between INFORM
partners and EUAB members. During the EUAB01 meeting, presentations, face-to-face interviews and open
discussions gave the possibility to further explain the requirements expressed in the questionnaire #1 (D3.2).
Figure 1: Participants (INFORM partners and EUAB members) of the EUAB01 (left) and EUAB 02 (right)
The integration of the questionnaire answers, the face-to-face interviews and open discussion resulted in the
definition of a preliminary list of end-user requirements:
SPATIAL INFORMATION: the availability of spatial information about their water body of interest is of utmost
importance for end-users, particularly for upscaling the limited number of in-situ point measurements.
TEMPORAL FREQUENCY: the possibility of having spatial information at different times can help end-users in
monitoring plans both for the regular control of water quality status as well as for the identification of
unusual and unexpected events damaging the environment. Again, with multi-temporal information, the
evolution and dynamics of consequences due to infrastructural projects could be monitored in time.
ACCURACY: products need to be accurate or at least with associated information about the quality of pixel
values (e.g. flags). Furthermore, products have to be retrieved using robust algorithms (with reference to
literature or algorithm theoretical baseline document), which can be adapted to extreme water conditions
too (e.g. high water turbidity).
CONSISTENCY: end-users require consistency between products derived from different sensors; this is important for archiving data and for allowing comparisons of data coming from different sources in different times. A good atmospheric correction (with reference to literature or algorithm theoretical baseline document) is a prerequisite for consistent products.
TAXONOMY: end-users require standardization of the taxonomy used by the project partners and the end-users, in particular for the name of parameters, the measurements units, the legend and the maps color code Moreover, a check, comparison and analysis of the difference between ocean and inland waters remote sensing taxonomy is important. A standardized taxonomy is perceived by the end-users as a prerequisite for a harmonized EU-wide inland water quality monitoring.
ACCESSIBILITY: data has to be easily accessible and downloadable preferably by WMS, no specific knowledge about remote sensing should be needed to work with data but training is requested.
PARAMETERS: since the water environments studied are so diverse, specific parameters will be selected and then investigated with reference to the single study area of interest (e.g. for the Curonian Lagoon Chl-a concentration and cyanobacteria bloom occurrence are useful parameters, while TSM concentration are useful parameters in the Scheldt river).
During the 2nd EUAB meeting, progress of the INFORM project was presented by the INFORM partners with a focus on the INFORM products developed in order to collect some feedback from the EUAB members.
The INFORM End-Users Questionnaire #2, together with the questionnaire #1, was created in order to investigate the main end-users requirements about the typology of products and the delivery of Earth Observation (EO) and biogeochemical model products provided by the INFORM Project (D3.4). The questionnaire #2 was structured in five sections (EO products, Mapping layout, Mapping content, Names and symbols, Data format and delivery) with multiple choice and open questions around specific examples of output products and product description flyers inviting end-users to provide feedback and suggestions for their improvement and to investigate which options better fit end-users needs.
Afterwards, the two questionnaires were distributed to an enlarged end-user group to have different categories of end-users more represented, and to consider the viewpoint of stakeholders not directly involved inside the Project.
In synthesis, the integration of both questionnaire answers resulted in a specific list of end-user requirements:
• EO derived products mostly requested are total suspended matter concentration, turbidity and Chl-a concentration (>80%), followed by the parameters describing the phytoplankton community and its blooming activity.
• Nutrients concentration (P and N) and algal biomass are the two parameters commonly preferred among the biogeochemical modeling derived products.
• The preferred spatial resolutions are 5 m and 30 m, depending on the dimension and the typology of the study area.
• 1 week and 1 month are the principal delivery times request by end-users.
• Standard deviation values is the most useful statistical information, followed by median and range values.
• Regression analysis is the preferred method for uncertainty measures for satellite derived data.
• For the accuracy of satellite derived data in comparison to in-situ data difference in % from in-situ values are the preferred choice.
As a conclusion, we can state that the products developed within the project and presented in the format as requested by the EUAB members will be relevant for the monitoring of inland water quality and for the comprehension of ecological processes.
Results uptake by end-users (EUAB03)
The objective of the two-days workshop “Earth Observation for Inland and Transitional Water Quality and Aquatic Vegetation Monitoring” held in Copenhagen in 2017, was to present the FP7 SPACE INFORM final results and products to the EUAB members and other stakeholders from industry, ESA, EEA and the EC and to collect feedback and requirements for future products.
In the Day 1, after welcome presentations and a brief overview of other European initiative on Earth Observation data and products, VITO and CNR presented an overview on the INFORM project and the geoportal, and the history of end-users interaction activities. Then each INFORM partners made a presentation explaining product typology, study site, methodology, algorithms, validation, and showing the final products. An interactive corner sessions was organized followed by an open discussion with end-users. The interactive session was displayed in five interactive corners: one corner was dedicated to water quality model, and the others to posters showing EO products for four categories of aquatic environments (River, Lake, Wetland, and Lagoon).
In the Day 2, recommendations for future satellite missions were presented by RBINS. VITO and RBINS presented an introduction to iCOR and ACOLITE software for atmospheric correction for Landsat-8 and Sentinel-2. Two practical hands-on session followed.
In conclusion, all end-users gave positive feedback about the water quality products developed in the INFORM project. Some valuable suggestions to improve future products were also collected. During the project, the users pointed out that the major benefit of using EO data is given by the possibility of having spatial, multi-temporal, and updated information derived from satellite images with respect to the traditional in-situ monitoring techniques based on point measurements. The main general requirement, during the workshop, was about the accessibility at different levels of complexity of the INFORM EO/model products and their algorithms after the end of the project. In addition, emerged that now end-users are willing to pay for ready to use standardized non-complicated products. This could be a first step to integrate remote sensing data in a routinely based monitoring. There is also a need for a cohesive data set across Europe for the purposes of the Water Frame Directive (WFD) reporting.
All details about comments, feedback, and recommendations for future products/projects from the interactive session and open discussions were described in the deliverable 8.12.
Datasets gathered in the project
Curonian Lagoon (+ airborne COOLAPEX partly funded by EUFAR), Lake Balaton, Kis Balaton, Lake Marken, Venice Lagoon, Lake Garda, Po river, Loch Leven, Loch Lomond, Danube Delta & Back Sea
Collected in situ data includes:
• Land reference targets: the reflectance of several land reference targets was measured for validation of the atmospheric correction
• Apparent optical properties including water reflectance using different instruments (TRIOS RAMSES, WISP, Satlantic HyperSAS, ASD FieldSpec PRO FR), light penetration and turbidity
• Aerosol Optical Thickness, visibility and water vapour concentration from sunphotometer readings
• Bio-geochemical data including Chlorophyll-a, Phycocyanin, Total Suspended matter (TSM), Coloured dissolved organic matter (CDOM), Particulate and dissolved organic carbon, Phytoplankton biomass, Phytoplankton primary production
• Inherent Optical Properties including Spectral absorption and scattering, Particulate backscattering, Particulate absorption
• Macrophytes in situ data including canopy and leaf reflectance, leaf area index, fractional cover and above water biomass
Partners agreed to make the field data available in the LIMNADES database after publication ( ). LIMNADES is a centralised database for in situ bio-optical measurements and satellite match-up data from lakes and other inland waters worldwide. The database is held in trust and maintained by the UK GloboLakes project ( LIMNADES currently contains observations from over 1500 lakes worldwide including 4500 hyperspectral water-leaving reflectance from 250 lakes around the world.
The satellite products (incl. Sentinel-2, Landsat-8, Sentinel-3 and Airborne APEX data) were made available to the user community through the INFORM geoportal:
Algorithms designed
The INFORM project partners developed a suite of algorithms for the atmospheric correction and level 2 products retrieval from airborne hyperspectral data, Landsat-8 and Sentinel-2. The algorithms go well beyond the state-of-the art, are described in peer-reviewed scientific publications and several are made available to the wider public as standalone tools or plugins into existing toolboxes.
• Atmospheric correction
The iCOR atmospheric correction software was developed for Sentinel-2 and Landsat-8. It is available as a plugin the SNAP toolbox for Sentinel-2 and Landsat-8. Details on the updated atmospheric correction procedures can be found in De Keukelaere et al. (Submitted).
In the project, an inventory was made of existing field and airborne data and data gaps were identified. Based on this inventory, new datasets were acquired for development and validation of the algorithms. The datasets gathered within the project are highly valuable as they were collected with great care and represent matchups with airborne and satellite data.
Development campaigns (in-situ, spaceborne) were organized in the first years of the project at:
Lake Balaton (+ airborne), Kis Balaton (+ airborne), Mantua Lakes (+ airborne HYPPOS funded by EUFAR), UK lakes (+ airborne funded by NERC), Lake Geneva+Lake Biel (+ airborne HILBILLY funded by EUFAR)
Testing campaigns (in-situ and spaceborne) were organized in 2016 and 2017 at:
The ACOLITE atmospheric correction was developed and is a stand-alone version of the IDL processing for Landsat-8 data and Sentinel-2 data. Details on the updated atmospheric correction procedures can be found in: Vanhellemont & Ruddick (2016) and Vanhellemont & Ruddick (2015).
Extensive validation has been performed of the reflectance products in European lakes by comparing water leaving reflectance from L8-OLI and S2-MSI imagery to in-situ observations gathered in Lake Balaton, Mantua Lakes, Lake Garda, Lake Geneva, Curonian Lagoon and Lake Marken by project partners (CNR, VITO, U Stirling, MTA OK, and Klaipeda University). Next an intercomparison was performed between both ACOLITE and iCOR based on all available in-situ data. Details can be found in D5.15.
The atmospheric correction processors are highly mature for S2 and L8. They are currently integrated in operational processing chains. They can be applied to inland waters, coastal waters and land.
• Attenuation and euphotic depth
The quantification of available light in the water column is crucial to evaluate water quality in lakes as it is one of the major factors determining primary production. The light environment in water is generally described in terms of vertical attenuation coefficient (Kd). The validation of the adapted multiband quasi-analytical algorithm (QAA) was reported in Lee et al. (2002). The QAA, originally developed for MERIS/MODIS, is designed to derive absorption (a) and backscattering (bb) coefficients by inverting the spectral remote sensing reflectance (Rrs(λ)). The absorption and backscatter coefficients are subsequently used to estimate Kd (Lee et al. 2005a; Lee et al., 2005b, Lee et al., 2007). In INFORM this algorithm was adapted to be used for new and upcoming sensors, such as Landsat-8 OLI (L8-OLI) and Sentinel-2 MSI (S2-MSI) for inland water cases
This process consists of a spectral shift of the broad band set of L8-OLI and S2-MSI data to narrow band (i.e. 10 nm) and a recalibration of the different steps of the QAA based on the IOP-Rrs relationships for inland waters. The adapted QAA has been integrated into ACOLITE which is a binary distribution of the Landsat-8 OLI processing. It allows simple and fast processing of L8 images for marine and inland water applications and the following output products have been added to the portfolio: Kd443, Kd490, Kd560, Kd665, KdPAR, a443, a490, a560, a665, bbp443, bbp490, bbp560, bbp665.
In a validation study, a total of 5000 different remote sensing reflection spectra from L8-OLI were simulated using HydroLight 5.2. This data set provides reference material for each of the processing steps enabling a direct validation of the adapted QAA to data from new sensors. The scatter plot in figure 2 shows the comparison of the Kd(490nm) product calculated using the adapted QAA on L8-OLI data to the reference values from HydroLight. These results demonstrate that the adapted QAA used with L8-OLI data is capable of providing high quality Kd products with a high spatial resolution which is highly important for inland water applications.
Figure 2: INFORM product flyer attenuation and euphotic depth
• TSM and turbidity
Total Suspended Matter (TSM), which is the amount of organic and inorganic particles suspended in the water, is an important feature for the monitoring of water quality. High TSM concentrations at the surface will lead to a reduction of light in the underlying water layers which affects the aquatic vegetation. But also the transport of TSM, including toxins is important to monitor. Remote sensing can play a role in TSM mapping since TSM is, next to chlorophyll and CDOM, one of the optically active constituents. Therefore satellite or airborne data can provide useful information about the spatial patterns or historical characteristics of TSM.
Here a generic multi-wavelength switching algorithm is proposed based on the Nechad et al (2010) single band semi-analytical TSM algorithm. It has the advantages over the previously established algorithms that it can be applied to any sensor with bands in the Red and NIR and that it is also suitable for extremely high TSM concentrations when the sensor has a spectral band in the SWIR I region.
RBINS also implemented multiple algorithms for TSM and turbidity into ACOLITE for L8 and S2 data (Fig. 3).
Figure 3: INFORM product flyer TSM
Since most of the inland lakes investigated/studied showed low concentrations of TSM, only in the highest
turbid waters like the Scheldt River, a switch was made to the higher wavelengths for TSM retrieval.
The validation process showed promising results for Sentinel-2 derived TSM products for Lake Marken (R² =
0.95) and Mantua Lakes (R² = 0.78). A cross-comparison between a Landsat-8 and a Sentinel-2 images of
Lake Marken acquired on the same date (01/05/2016) yielded a slope of 0.96 showing the robustness
between the two sensors. For Lake Balaton, known for it’s small and very light sediment particles, a
recalibration of the coefficients was needed to derive TSM maps.
• Yellow Matter
Yellow matter, and in particular the fraction of coloured dissolved organic matter (CDOM), is a proxy for
dissolved organic carbon (DOC) and the availability of light for photosynthesis by phytoplankton and
submerged macrophytes. CDOM is difficult to quantify from remote platforms due to often poor signal
quality at the blue end of the spectrum, where CDOM absorbs light most prominently. Overlapping
absorption signals, particularly in very turbid or productive waters, may also partially mask the CDOM
absorption. We investigated whether an analytical (globally valid) algorithm can quantify CDOM absorption
in different water types.
In addition to in-situ validation described in D5.13 the new inversion method was applied to an APEX image
of 19 July 2014 obtained over Lake Balaton. Only the imagery over the western part of the lake, at the mouth
of the Zara River, is used. Results are summarized in the flyer (Fig. 4).
The proposed inversion scheme shows significant promise in the sense that a wide range of water types can be handled. However, improved results should be expected once a more sophisticated bio-optical model is linked to the inversion.
The inversion scheme makes use of the hyperspectral nature of the reflectance spectrum and can as such be used to determine whether closely located bands provide similar solutions for the magnitude of the backscattering coefficient. This allows automated filtering of poor spectra, for example affected by sun glint, low signal levels, or shallow mixed layers. These features should be further explored and exploited.
The work carried out indicates that there is a defined limitation to the use of the analytical method for inversion of reflectance to absorbing components at short wavelengths. This limit is not primarily imposed by extrapolation errors from red and near infra-red bands, but by overlapping absorption signatures of particulate and dissolved components. Larger data sets should be explored to determine the exact limit up to which analytical inversion performs adequately. Beyond this limit, only empirical algorithms that are locally tuned would be of use, with the advantage of better error performance but without transferability of algorithms to other regions.
PML finished the validation of the analytical inversion to ultimately derive yellow matter absorption from hyperspectral reflectance, focusing on minimizing the propagation of errors from the inversion of reflectance into bulk inherent optical properties. In waters where a strong phytoplankton component overlaps with CDOM absorption, errors in the CDOM absorption component retrieval remain problematic and should be at priori flagged as uncertain.
Two case studies using hyperspectral APEX imagery were carried out, including validation against in-situ data, on Lake Balaton and the Curonian Lagoon. In addition, a large number of pre-existing algorithms tuned against the LIMNADES dataset (part of the GloboLakes project) were compared for CDOM-rich lakes, further demonstrating the need for full spectral inversion approaches to derive CDOM absorption from lakes.
Figure 4: INFORM product yellow matter
• Phytoplankton functional types
As the foundation of the aquatic food chain, phytoplankton is an integral part of the ecosystem, affecting
trophic dynamics, nutrient cycling, habitat condition, and fisheries resources. Phytoplankton are responsible
for 45% of the total primary production of plants on Earth and uptake of the greenhouse gas carbon dioxide,
at the same time they produce almost 70 % of world’s atmospheric oxygen and they contribute to the
biological pump.
Several studies demonstrated the capability of remote sensing to map phytoplankton in inland waters
(Odermatt et al., 2012). Various approaches have been applied successfully and good results have been
achieved both using site-specific semi-empirical algorithms (e.g. Zimba and Gitelson, 2006) and bio-optical
modeling (e.g. Mishra et al., 2014). Bloom events have been identified in different ways: by mapping the
concentration of chlorophyll-a (chl-a), which is the main photosynthetic pigment of all the phytoplankton
(e.g. Moses et al., 2012), or the concentration of characteristic accessory pigments for cyanobacteria, such as
phycocyanin (PC) and phycoerythrin (PE) (e.g. Duan et al., 2012) or by using qualitative approaches
exclusively based on the spectral characteristics found in water reflectance spectra (e.g. Bresciani et al., 2011).
Within INFORM we used a combined approach to estimate Chl-a and accessories algal pigments such as PC and PE: i) the semi-empiric approach might be used for a preliminary detection of the presence of pigments common to pre-defined algal species; and ii) spectral inversion of bio-optical modelling where the absorption and backscattering coefficients of different phytoplankton species are included (e.g. Simis et al. 2007; Li et al. 2013).
Then, for moving towards a Phytoplankton Functional Types (PFT) mapping, phytoplankton has to be considered in terms of biogeochemical function that is not only depending on pigment composition but also on Phytoplankton Size Class (PSC) and on supplementary variables (e.g. habitat conditions) which cannot be obtained from remote sensing techniques. Therefore, for producing a PFT map from satellite data we need to combine the information achievable from space to ancillary information.
Figure 5: INFORM product flyer phytoplankton products
APEX and satellite (S2, S3 and L8) images and in-situ data were combined to develop semi-empirical algorithms. We implemented a bio-optical model (with spectral inversion techniques) to distinguish the phytoplankton pigments and we implemented a robust adaptive semi-empirical algorithms (two-bands adaptive) to quantify Chl-a and pigment concentrations (PC and PE) in cyanobacteria.
The products include Chl-a concentration maps obtained from APEX imagery of Mantua Lakes (2011 and 2014) that show a very good agreement (r=0.95) with in-situ data. Depending on phytoplankton composition at the time of APEX acquisitions a map depicting the spatial variation of PC pigments and a map showing PFTs have been also produced in 2011 and 2014, respectively (Fig. 5). From the APEX data we produced maps of Chl-a and PC concentrations for Curonian Lagoon.
Algorithms and bio-optical models were implemented to be applied to S2 and S3 images. The algorithms dedicated to different satellite images for the different case studies have provided robust results. Through the semi-empirical algorithms dedicated to eutrophic environments, Chl-a and PFT maps were produced for Mantua Lakes from S2 images for the 2015-2017 data-set and for the Curonian lagoon from S3 images. In addition, Chl-a maps were also produced from S3 images for Lake Garda through the bio-optical modeling, parameterized specifically for oligo-meso trophic environments.
We implemented an algorithm dedicated for retrieval scum phenomenon. Based on particularity at this phenomenon we have used Curonian lagoon as test site. The maps wereproduced from S2 and L8.
Remote-sensing data to determine phytoplankton size class (PSC) in lakes was investigated by testing existing algorithms developed for marine environments (e.g. Brewin et al. 2010). The PSC model was tested in Lake Balaton (Hungary) using abundance data collected from August 2007 to December 2013.
We assessed the sensitivity of algorithm performance in relation to variability in the presence of additional optically active constituents and different phytoplankton photosynthetic pigments. The sensitivity analysis of the bio-optical model and semi-empirical algorithms was performed with the algorithm based on the variance method.
The analysis recognized three spectral ranges with specific level of interactions between the inputs. The first part of the spectrum up to 500 nm had average level of 10% of interaction; the second up to 600 nm showed values of 5% with a peak around 580 nm; the third showed an increasing interaction level until 15% near 715 nm.
• Stratification
The vertical gradients of water constituents and stratification in the water column due to density gradients and hydrodynamic processes are well known natural phenomena. They can be quite pronounced close to river inflows, shallow water areas with resuspension events or for phytoplankton distributions.
However, remote sensing algorithms deriving in-water properties from satellite data usually simplify the water body as vertical homogenous one-layer system. Some scientific publications describe the impact of the non-uniform vertical water (Gordon and Clark 1980, Zaneveld et al., 2005) and solutions to derive also information about the vertical structure in studies (Frette et al., 2001), but were not yet applied on remote sensing data. The parameter frequently derived from satellite products and indicating the depth integral
where 90% fraction of light is scattered back and contributing to the remote sensing signal is called z90. This parameter is strongly dependent on the total absorption in water and therefore wavelength dependent. Vice versa, the wavelength dependency of light extinction can be used to derive in-water properties from various depth intervals, when the retrieval algorithms can be adapted to different sensor channel dependencies using a sensor independent inversion approach (Heege et al., 2014).
In an attempt to retrieve the vertical profile of the water body, a 2 layered model were used to establish a parameterization of the water body in dependence of these two layers. The algorithm determines a best-fit two layered water body with different levels of TSM in the two layers. Thus it has three resulting values: 1) the upper layer TSM concentration, 2) the interface depth and 3) the lower layer concentration.
Figure 6: INFORM product flyer stratification
For testing the newly developed stratification algorithm, several satellite images of Sentinel-2A, Landsat 8 and MERIS of Lake Constance have been processed (Fig. 6). The dates selected according to existing validation data and suitable image quality. The in-situ data were gathered through routine measurements of the INFORM end-user Lake research Institute of the State Institute for the Environment, Measurements and Nature Conservation of Baden-Wuerttemberg. In general, the resulting stratification images show a high degree of smoothness and the regional distribution coincides between Landsat and MERIS. Also, the trend in stratification calculated by the simulation coincides with the trend in the in-situ data. When using Sentinel 2 imagery, the interface depth was relatively stable but increase factors tend to reach the fit limits. Still, it was possible to obtain information about whether the concentrations decrease or increase and some indication
of how strong the concentration changes are (tendency is displayed correctly). A detailed description of the results and resulting images can be found in deliverable D5.9 and D5.13.
• Macrophytes
Aquatic plants, or macrophytes, are key components of inland freshwater ecosystems (Jeppesen et al., 1997), playing a significant role in the global carbon and nutrient cycles (Jordan et al., 2011), as well as in the provision of suitable niches for nursery and feeding activities for several aquatic faunal species and threatened taxa (Schriver et al., 2005). Littoral freshwater environments have experienced a dramatic reduction of extent and a decline in quality and functionality in the last decades, all around the globe (e.g. Bresciani et al., 2012), mainly due to anthropic impacts (Jeppesen et al., 2010). Even if preliminary evidences suggest an ambiguous role of climate warming on macrophytes, an increase in their growth rates and spatial distribution is expected (Carmichael et al., 2014; Jacobs and Harrison, 2014). Alongside this, a reinforcement of water eutrophication symptoms is foreseen (Kosten et al., 2011), especially in shallow lakes and wetland habitats located in temperate to high latitude regions (Finlayson et al., 2013).
New generation of operational satellite platforms, such as Landsat 8 and Sentinel-2 (and future hyperspectral missions EnMap and PRISMA, among others) allow the provision of EO data with adequate for punctual monitoring of macrophyte dynamics in both space and time at unprecedented resolutions (spectral, spatial and temporal).
As part of the EU Water Framework Directive (WFD), biomonitoring indices based on macrophytes has been implemented for assessing freshwater ecosystem status (Jagtap et al., 2003). Within this framework, EO data are effective in providing consistent information on key macrophyte parameters at local to regional scales, such as macrophyte presence, coverage and functional type, and synthetic biophysical parameters (leaf area index – LAI, fractional cover, leaf biomass).
Figure 7: INFORM product flyer macrophytes
Within INFORM we implemented an approach for mapping macrophyte community types based on multi-temporal aquatic vegetation indices features derived from Landsat data (ETM+ and OLI) and tested over a set of heterogeneous case studies, merging archive data with new in situ data collected during the project. Resulting macrophyte functional type products allow the delineation of four macrophyte classes (helophyte, emergent rhizophyte, floating macrophyte and submerged-floating association) at 30 m spatial resolution (Villa et al., 2014; Villa et al., 2015).
Alongside community type mapping, algorithms based on semi-empirical regression models fed with optimal spectral indices derived from EO data (Villa et al., 2017) were developed for estimation of macrophyte biophysical parameters - namely fractional cover (fC), leaf area index (LAI) and above-water biomass - using narrowband (APEX) and broadband (Sentinel-2 MSI) data.
Macrophyte community type maps were produced for two INFORM test sites: Kis-Balaton and Mantua lakes system, in 2014 and 2016. Macrophyte biophysical parameters products were derived over Kis-Balaton and Mantua study areas, using hyperspectral APEX data acquired during 2014 INFORM development campaigns, as well as using Sentinel-2 multi-temporal data covering 2016 season (January-October).
Validation of both community type and biophysical parameters products was carried out by comparing 2016 results to reference data acquired in situ. Community type mapping showed good performance, with 87-88% overall accuracy, and low per-class error (omission < 10%), with the exception of the association of submerged and floating species. Biophysical parameters using both narrowband and broadband data estimation was quite accurate for fC and LAI, i.e. with relative error < 18%, and acceptable for biomass, i.e. with relative error in the range 34-42%.
EO-based products of macrophyte parameters can be effectively employed for assessing ecological status of inland waters, and informing management actions by local authorities, such as:
• the identification of different macrophyte community types, possibly extending to invasive species, can be used in natural protected area for species census, quantification of areal coverage, and evaluation of conservation strategies;
• temporal information about the macrophyte assemblages coverage and density (e.g. fractional cover) can support active control of invasive species and prevention of risks connected to infilling and hypoxia in water column;
• quantitative mapping of macrophyte biomass can support management decisions dealing with vegetation spread control and cutting, including estimation of biomass removal costs;
Furthermore, these products can be used as synoptic, quantitative information source for the study of freshwater biogeochemical processes and ecosystem status, and in particular (some prominent examples):
• for relating functional traits and biophysical parameters of different macrophyte groups to water and sediments physico-chemical parameters;
• for studying the coexistence and balance of different macrophyte groups and phytoplankton in shallow aquatic environments;
• for analysing the spatial and temporal evolution of different macrophyte groups in inland waters along a gradient of climatic and trophic conditions;
• for linking macrophyte parameters and macrophyte metabolism (e.g. N uptake and removal).
• Phytoplankton primary production
Primary production by phytoplankton is a fundamental process underlying lake metabolism. The uptake, transformation and respiration of carbon (C) by phytoplankton in lakes contributes significantly to carbon transfer across the air-water interface and is thus an integral processes in regional and global carbon cycling.
There are several published models for the estimation of phytoplankton production from satellite-derived data, but these have been almost exclusively developed and validated using data from marine waters. In the INFORM project, we have tested and, where appropriate, adapted a number of satellite-based primary production models using data from a number of case study lakes.
Figure 8: INFORM product flyer primary production
U STIRLING completed the evaluation of satellite-based primary production models using in situ
measurements from Lake Balaton and Loch Leven from MERIS satellite data. Results show promise with a
modified version of the VGPM algorithm providing good estimates of primary production in Lake Balaton and
Loch Leven. Further work has focused on improving the accuracy of the input data (especially Kd), revising
the parameterization of phytophysiological parameters (i.e. Pbopt) and extending the model comparison to
include a wavelength-resolved model for turbid waters. In addition, U STIRLING has designed and built a
photosynthetron for simulated in-situ 14C incubations and is using this to collect new data for model
parameterization and validation
Validation work based on Lake Balaton and Loch Leven identified sources of model uncertainty. This is now
informing the improved selection of algorithms used to derive the input products to reduce uncertainties on
primary production estimates.
• Sun-induced chlorophyll fluorescence
The presence of Chl-a increases the magnitude of radiance measured over the fluorescence spectral band.
Therefore, the relative contribution of chlorophyll can be estimated by quantifying this increase in upwelling
radiance. This is achieved by first defining a baseline below the fluorescence radiance peak and using the
coordinates of the line to estimate a radiance value that would be expected in chlorophyll-free water. The
chlorophyll fluorescence contribution is then simply the measured radiance at the fluorescence peak minus
the baseline estimate at the peak wavelength. The maximum peak height (MPH) algorithm (Matthews et al.,
2012) is designed with a conditional peak position selector which searches for the maximum radiance over
three bands. MPH is an operational ocean colour algorithm designed to estimate Chl-a concentration from
SICF measured in optically complex, inland and near-coastal waters.
Global in-situ observations of reflectance and Chl-a were used to test the existing MPH method as well as
validate an improved MPH algorithm, as described in D5.13. These methods were also applied to MERIS
images taken from Lake Geneva and Lake Balaton. Results are summarized in the flyer below (Fig. 9).
Figure 7: INFORM product flyer sun-induced chlorophyll products
We developed and improved algorithms for chlorophyll estimation in low biomass waters primarily using fluorescence-based approaches. We assessed chlorophyll-fluorescence signal across different optical water types using in-situ data and radiative transfer modeling. In addition, we completed an initial validation of a combined fluorescence-scattering peak algorithm (MPH) and developed an adaptive parameterization for different optical water types. This was undertaken using in-situ data and is was extended to satellite data. This approach results in a substantial improvement in algorithm performance, but the algorithm is still subject to high uncertainties in some lakes or optical water types. This work is being extended using a large (>10,000 spectra) simulated dataset. We also started to explore other approaches for chlorophyll retrieval in low biomass waters not based on chlorophyll fluorescence.
The accuracy and uncertainties associated with the SICF algorithms was assessed against in-situ data for a large number of lakes; this is now being extended using a large simulated dataset (D5.13). We identified optical-water types where SIFC contributes significantly to the water-leaving radiative signal as a means of containing their application to suitable water types.
EO-model integration
Deltares worked with RBINS on the implementation of the spectral Kd (light attenuation) model in the Delft3D WAQ (biogeochemical) model for Lake Balaton (Hungary) and Markermeer (The Netherlands). The Kd model is one of the elements of the biogeochemical model that aims to describe the resuspension, settling and dispersion of non-algal suspended matter and phytoplankton as well as their mutual interactions. The report describing this Kd model and the first version of the model for Lake Balaton were delivered as part of D6.1. The spectral Kd model showed good correlation with the measured Kd in Lake Balaton over many years. There were some differences between the relative contributions of the different constituents to the overall Kd. The Balaton model simulations for Chl-a were good when compared to the average in-situ measurements for the years 2004-2013 but higher than 2014. This is due to the fact that in-situ measurements on nitrate were lower than simulated by the model.
Deltares finalized D6.2 whish was an application of the Delft3D biogeochemical model for Lake Markermeer. The model was set up for the year 2006 for which EO data from the FP7 project FRESHMON were available to compare the model outcome on TSM and Chl-a. The model simulates TSM and Chl-a concentrations. Comparison with the EO data shows that TSM is well simulated but Chl-a gives more problems. The Chl-a simulations in spring/summer are good because of the low wind speed conditions. In winter (high wind speed) the model underestimates the Chl-a concentrations. The reason for this is that algae flocculate with suspended solids during summer, sink to the bottom where they accumulate. In winter, this material then resuspends but it remains in the lake because this system is physically isolated from the adjacent lake IJsselmeer and the Wadden Sea.
VITO developed an algorithm to validate model performance with EO data. The algorithm was a set of Matlab codes to remove the satellite images and plot the image into the grid of the model. Consequently, for each model grid cell the level of agreement between model and EO is calculated using the Target Value as the statistical method. The results were published in D6.3.
Deltares also finished Task 6.3 (resulting in delivery of D6.4). Here, both case studies underwent through optimization steps. For Balaton, the correlation between observed and modelled Kd(PAR) slightly improved
in comparison to Task 6.1. The updated model systematically overestimates the Kd(PAR). The correlation between computed Secchi depth and in-situ measurements improved in the updated model. The comparison between Kd(PAR) from UITZICHT and Landsat-8 recordings did not show an improvement in comparison with deliverable D6.1 but Kd(PAR) is well calculated by UITZICHT for the eastern basins of the lake.
For the case Markermeer, Deltares switched to another primary production model, DYNAMO, to see if the model could better match the observed Chl-a patterns in the winter of 2006. DYNAMO did show higher Chl-a values for the winter period and therefore matches the observed winter Chl-a values in 2006 better than the BLOOM module. The peak Chl-a values in the spring period however, were much higher than observed for an unknown reason. BLOOM was able to reproduce the spring situation well, but DYNAMO not. Lake Marken is perhaps the most heavily modified water body in Europe as it went from an estuary to a freshwater lake and is surrounded by dikes that do not allow for a natural transport of suspended solids to the North Sea. These suspended solids cause several complex interactions with algae, nutrients and bacteria in the pelagic zone that are still unknown and hence difficult to model.
Bresciani, M., Giardino, C., Bartoli, M., Tavernini, S., Bolpagni, R. & Nizzoli, D., (2011). Recognizing harmful algal bloom based on remote sensing reflectance band ratio. Journal of Applied Remote Sensing, 5(1), p.053556.
Bresciani, M., Bolpagni, R., Braga, F., Oggioni, A. & Giardino, C., (2012). Retrospective assessment of macrophytic communities in southern Lake Garda (Italy) from in situ and MIVIS (Multispectral Infrared and Visible Imaging Spectrometer) data. Journal of Limnology, 71(1), p.19.
Brewin, R. J. W., Sathyendranath, S., Hirata, T., Lavender, S. J., Barciela, R. and Hardman-Mountford, N. J. (2010). A three-component model of phytoplankton size class for the
Atlantic Ocean. Ecological Modelling, 221, 1472–1483, doi: 10.1016/j.ecolmodel.2010.02.014.
Carmichael M.J. Bernhardt S.E. Brauer S.L. & Smith W.K. (2014). The role of vegetation in methane flux to the atmosphere: Should vegetation be included as a distinct category in the global methane budget? Biogeochemistry, 119(1-3): 1-24.
De Keukelaere, L. , Sterckx, S. , Adriaensen, S., Knaeps, E., Reusen, I., Giardino, C., Bresciani, M., Hunter, P., Van der Zande, D., Vaiciute, D. iCOR Image Correction for Atmospheric Effects, Results for Landsat-8 OLI and Sentinel-2 MSI . special issue Sentinel-2 data calibration and validation of the European Journal of Remote Sensing. Submitted.
Duan, H., Ma, R., & Hu, C. (2012). Evaluation of remote sensing algorithms for cyanobacterial pigment retrievals during spring bloom formation in several lakes of East China. Remote Sensing of Environment, 126, 126–135.
Heege, Thomas, Kiselev, Viacheslav, Wettle, Magnus, & Hung, Nguyen Nghia. 2014. Operational multi-sensor monitoring of turbidity for the entire Mekong Delta. International Journal of Remote Sensing, 35(8), 2910–2926.
Jacobs A.E. Harrison J.A. (2014). Effects of floating vegetation on denitrification, nitrogen retention,
and greenhouse gas production in wetland microcosms. Biogeochemistry 119:51-66.
Jagtap, T. G., Komarpant, D. S., & Rodrigues, R. S. (2003). Status of a seagrass ecosystem: an ecologically sensitive wetland habitat from India. Wetlands, 23(1), 161-170.
Jeppensen E., Jensen J.P. Søndergaard M., Lauridsen T., Pedersen L.J. Jensen L., (1997). Top-down control
in freshwater lakes: the role of nutrient state, submerged macrophytes and water depth.
Hydrobiologia 342/343:151-164.
Jeppesen E., Kronvang B., Olesen J.E. Audet J., Søndergaard M., et al. (2010). Climate change effects on nitrogen loading from cultivated catchments in Europe: implications for nitrogen retention, ecological state of lakes and adaptation. Hydrobiologia 663: 1-21.
Jordan S.J. Stoffer J., Nestlerode J.A. (2011). Wetlands as sinks for reactive nitrogen at continental and
global scales: A meta-analysis. Ecosystems 14:144-155.
Kosten S., Jeppesen E., Huszar V.L.M. Mazzeo N., Van Nes E.H. Peeters E.T.H.M. Scheffer M., (2011).
Ambiguous climate impacts on competition between submerged macrophytes and phytoplankton
in shallow lakes. Freshwater Biol. 56:1540-1553.
Lee, Z., A. Weidemann, et al. (2007). Euphotic zone depth: its derivation and implication to ocean-color remote sensing. Journal of Geophysical Research 112 C03009)
Lee Z., Kendall L., Arnone R. (2002) Deriving inherent optical properties from water color: a multiband quasi-analytical algorithm for optically deep waters. Applied Optics, 41(27), 5755-5772
Lee, Z., K. Du, et al. (2005). "A model for the diffuse attenuation coefficient of downwelling irradiance." Journal of Geophysical Research 110(C02016).
Li, L., Li, L., Song, K., Li, Y., Tedesco, L. P., Shi, K., & Li, Z. (2013). An inversion model for deriving inherent optical properties of inland waters: Establishment, validation and application. Remote Sens. Environ., 135: 150-166.
Matthews, M. W., Bernard, S., & Robertson, L. (2012). An algorithm for detecting trophic status (chlorophyll-a), cyanobacterial dominance, surface scums and floating vegetation in inland and coastal waters. Remote Sensing of Environment, 124, 637–652.
Mishra, S., Mishra, D.R. Lee, Z., (2014). Bio-optical inversion in highly turbid and cyanobacteria dominated waters. IEEE Trans. Geosci. Remote Sens. 52: 375–388.
Moses, W. J., Gitelson, A. A., Perk, R. L., Gurlin, D., Rundquist, D. C., Leavitt, B. C., et al. (2012). Estimation of chlorophyll-a concentration in turbid productive waters using airborne hyperspectral data. Water Research, 46, 993–1004.
Nechad, B., Ruddick K.G. Park, Y. (2010). Calibration and validation of a generic multisensor algorithm for mapping of Total Suspended Matter in turbid waters. Remote Sensing of Environment, 114 (4): 854-866.
Odermatt, D., Gitelson, A. A., & Brando, V. E. (2012). Review of constituent retrieval in optically deep and complex waters fromsatellite imagery. Remote Sensing of Environment, 118, 116–126.
Simis, S. G. H., Ruiz-Verdu, A., Domingues-Gomez, J. A., Pena-Martinez, R., Peters, S. W. M., & Gons, H. J. (2007). Influence of phytoplankton pigment composition on remote sensing of cyanobacterial biomass. Remote Sensing of Environment, 106, 414–427.
Schriver P., Bogestrand J., Jeppensen E., Sondergaard M., 2005. Impact of submerge macrophytes on fish-zooplankton interactions: large-scale enclose experiments in a shallow eutrophic lake. Freshwater Biol. 33:255-270.
Vanhellemont Q. & Ruddick K. ACOLITE For Sentinel-2: Aquatic Applications of MSI imagery (2016) Submitted for the proceedings of the 2016 ESA Living Planet Symposium held in Prague, Czech Republic, 9-13 May 2016, ESA Special Publication SP-740
Vanhellemont Q. & Ruddick K. (2015). Advantages of high quality SWIR bands for ocean colour processing: Examples from Landsat-8 Remote Sensing of Environment. Volume 161, pp. 89–106. Open Access: Villa, P., Mousivand, A. & Bresciani, M., (2014). Aquatic vegetation indices assessment through radiative transfer modeling and linear mixture simulation. International Journal of Applied Earth Observation and Geoinformation, 30, pp.113-127.
Villa, P., Pinardi, M., Tóth, V.R. Hunter, P.D. Bolpagni, R. & Bresciani, M., (2017). Remote sensing of macrophyte morphological traits: Implications for the management of shallow lakes. Journal of Limnology, 76(s1).
Villa, P., Bresciani, M., Bolpagni, R., Pinardi, M., & Giardino, C. (2015). A rule-based approach for mapping macrophyte communities using multi-temporal aquatic vegetation indices. Remote Sensing of Environment, 171: 218-233.
Zaneveld J. R., Barnard A., and Boss E., (2005). Theoretical derivation of the depth average of remotely sensed optical parameters. Opt. Express 13, 9052-9061
Zimba, P. V., & Gitelson, A. (2006). Remote estimation of chlorophyll concentration in hypereutrophic aquatic systems: Model tuning and accuracy optimization. Aquaculture, 256, 272–286
Potential Impact:
Impact of the project
• From algorithm development towards operational services
The INFORM project focused on the development of algorithms for new innovative products taking advantage of the new increased spectral and spatial resolution of new satellite missions such as Sentinel-2. Through the collection of an impressive dataset of field measurements, the algorithms have been tested and validated on a diversity of inland waters and rivers. The methodology has been well described and the project resulted in 18 peer-reviewed scientific publications. In addition, all the ATBD documentation for the key bio-physical parameters is open free for consultation. Through these scientific publications, the work performed by INFORM can be carried further by other researchers and can be integrated in operational
services. As such, several partners which already provide operational services (e.g. EOMAP and VITO) already extended their product portfolio based on the outcomes of the INFORM project.
The newly developed algorithm will also feed into the Copernicus services, mainly the Global Land Service Lot 2 where both PML and VITO are involved as service providers. Today a number of products are in development such as reflectance, turbidity and trophic state. INFORM developments allow to extend this product portfolio with new advanced inland water products.
Finally, several algorithms are also implemented in the workflow of the Belgian collaborative ground segment, TERRASCOPE. TERRASCOPE provides easy access the Sentinel data for Belgian users and provides an exploitation platform for users willing to run their own algorithms close to the data.
• Impact for the scientific community:
o The dataset - a basis for future research
The INFORM dataset is unique in the sense that it contains a very large amount of field data, often collected simultaneously with an overpass of optical satellite sensors. This is crucial data for calibration and validation of remote sensing algorithms. The dataset will be made publicly available through the LIMNADES database and will be a basis for future research. In synergy with INFORM three EUFAR TA projects with APEX campaigns have been also accomplished during the INFORM lifetime for three test sites: Balaton, Mantua and Curonia; such unique data set provide relevant hyperspectral data for algorithm testing and validation.
o Teaching and training of researchers
Two software tools for the atmospheric correction of Landsat-8 and Sentinel-2 were made publicly available: iCOR as a plugin in the SNAP toolbox ( and ACOLITE as a standalone software tool ( Together they have already more than 400 registered users. Users include scientists, policy makers, and business people. Scientists are using both tools intensively resulting in a number of presentations in scientific conferences and publications.
Training sessions on iCOR and ACOLITE were already organized at the INFORM User Uptake workshop in Copenhagen (17 November 2017) and a new training session is foreseen at the 2018 International Geoscience and Remote Sensing Symposium (IGARSS) on 23-27 July in Valencia (Spain).
iCOR is currently being developed for Sentinel-3 and a SNAP plugin for iCOR –S3 is foreseen to be released in October 2018 (ESA funded iCOR4S3 project).
Making available these software tools to the wider scientific community has a great impact. As satellite data has become freely available from the space agencies -which gave an enormous boost in the use of the data- scientists have now also the possibility to further process the images and derive end products. This makes that single researchers do not have to start from scratch and can built further on these tools, focusing on more value adding.
A summer school titled “SENTINEL FOR WATER RESOURCES” was organized for PhD students, young post-doctoral scientists, technician, expert in the sector of remote sensing. The topics were the Copernicus, ESA Sentinel 1, Sentinel 2 and Sentinel 3 missions with focus on Water Resources, water quality measurements and analysis, and operational applications.
• Impact for the end user: Integration of remote sensing in current working practices of users:
The INFORM project strongly interacted with its End users and was able to create awareness on the different products and their applicability for inland water monitoring. The end users included universities,
governmental institutions, public administrations and industry. Next to meetings with the End User Advisory Board, many face-to-face meetings have been organized with national policy makers to create awareness on remote sensing products. Several product flyers have been produced within the INFORM project to clearly communicate about the new developments to the users. These flyers contain information on the product and the methodology and are made available on the INFORM website.
For governmental institutions and public administrations, INFORM particularly contributed to integrate remote sensing as an innovative technology in the current reporting obligations. As such, both Rijkswaterstaat (the executive agency of the Ministry of infrastructure and the Environment in the Netherlands) and the Environmental Protection Agency of Ireland already issued a call for tenders on the use of remote sensing for water quality monitoring and reporting. Remote sensing has important advantages as being transboundary and objective and lower cost than traditional field measurements.
In addition the products obtained in the project were used in the management of cutting the macrophytes and water level by Mincio River National Park, by Mantua Province and by Po River Authority in Italy.
For the industrial users, INFORM contributed to include remote sensing derived products for optimization of dredging locations, the better spatial planning of new harbor locations and the calibration and validation of sediment transport models. Today, several companies in the dredging sector are using remote sensing data for this purpose.
• Impact for the intermediate business user
An intermediate business user is a company, consultancy already providing services based on remote sensing data. Several of these intermediate users lack technical expertise and processing capabilities. INFORM has significantly reduced the technical barrier for these intermediate end user to use high resolution satellite data for aquatic applications. INFORM has clearly shown the benefits of the new satellite missions for inland water quality monitoring.
• Teaching and training of students
CNR-IREA organized a training course on optical remote sensing and on inland water ecology to six classes of middle schools (about 120 students). The first step was the training workshop to the professors of the classes involved in the project. Then followed three teaching meetings addressed to students to explain the basic concepts of lake ecology and of remote sensing applied to water. Field campaigns were performed with students to complete the training course with the use of radiometric instruments. A final meeting was held in each schools to match in situ data and satellite products (chlorophyll-a and macrophyte maps) obtained by S2 images.
Moreover, INFORM project aims and products were presented to university students of the master's degree on Environmental Sciences of the University of Parma, of the Msc. and PhD students on optical remote sensing at Klaipėda University and Hydrological PhD students in Hungary.
• Recommendations for the Space Agencies
INFORM provided recommendations for future satellite missions in a dedicated report (D8.9 ‘Report on recommendations for inland water quality monitoring for future satellite missions), and partners also contributed to the CEOS Feasibility Study for an Aquatic Ecosystem Earth Observing System report edited by
Arnold Dekker and Nicole Pinnel ( The study focuses on benefits and technological difficulties of designing an Earth observing satellite mission focused on the biochemistry of inland, estuarine, deltaic and near coastal waters as well as mapping macrophytes, macro-algae, sea grass and coral reefs. The study clearly defines the needed spatial resolution and the spectral bands needed for retrieving aquatic ecosystem variables and removing atmospheric and air-water interface effects. To support the definition of global hyperspectral satellite mission, INFORM results have been also presented at the international at the workshop ‘Exploring the Earth's Ecosystems on a Global Scale: Requirements, Capabilities and Directions in Spaceborne Imaging Spectroscopy (held at the International Space Science Institute: 21- 25 November 2016)’.
Table A1 contains all scientific (peer reviewed) publications, and Table A2 contains all dissemination activities relating to the foreground of the project. All papers fully or partially funded by INFORM can also be accessed (after the end of the journal embargo period) from the FP7 INFORM Project Community in the ZENODO repository (
List of Websites:
Project website:
contact details:
Ils REUSEN, Dr., Vlaamse Instelling voor Technologisch Onderzoek N.V. (VITO)
Tel: +32 14 32 68 62
Fax: +32 14 32 27 95