Skip to main content

AQUA-USERS: AQUAculture USEr driven operational Remote Sensing information services

Final Report Summary - AQUA-USERS (AQUA-USERS: AQUAculture USEr driven operational Remote Sensing information services)

Executive Summary:
With global population expansion, the demand for high-quality protein is rising dramatically, and fish farming is gaining importance to ensure food security. Aquaculture is one of the fastest growing food production sectors worldwide. Environmental conditions determine the growth and health of the produced species, while the production can also impact the surrounding environment. In particular, harmful algal blooms are a major threat to aquaculture production. Therefore, monitoring is needed on several levels. However, up to now, the available data are often disparate, inconsistent in coverage and of unknown quality. To support the growth of efficient and sustainable aquaculture production, the AQUA-USERS worked on providing the aquaculture industry with user-relevant and timely information based on the most up-to-date satellite data and innovative optical in-situ measurements.

AQUA-USERS was a highly user-driven project with a user board consisting of companies and organisations from 5 countries representing different European aquaculture production systems. Together with the user board, the project partners demonstrated the applicability of the developed methods and tools in several case studies.
The first phase of the project concentrated on user requirements, method development and implementation. Particular focus was on developing new and improved methods for making EO data useful for the aquaculture industry. For the study areas, a ten-year archive of satellite data from the 300m MERIS sensor was processed with regionally adapted and validated algorithms for water quality parameter retrieval. Together with an archive of sea surface temperature data based on a number of different sensors, these data sets were used as a basis for site characterisation and selection.

Based on satellite and in-situ data, a method was established to derive indicators for potential benefits and risks for aquaculture production. Where available, also biogeochemical parameters were included in the analysis. Another focus of the project was improving the methods for detecting harmful algal blooms (HABs) with optical methods. Two approaches have been pursued: training a detection algorithm with spectra of known blooms of certain species on multi-spectral satellite images, and modelling of hyperspectral HAB data based on laboratory experiments on cultures of these species.

To make the project results accessible to our users, two tools were developed: a web portal and mobile application were developed that bring together satellite information on water quality and temperature with in-situ observations as well as relevant weather prediction and met-ocean data.

Two services have been developed in the project and are ready to be deployed: A site selection service and a daily management service.

The first service is the site characterization/site selection. Site characterisation supports the operation of the aquaculture production by increasing the knowledge of the production site. Site selection does on the other hand have a longer perspective by supporting the identification of the optimal production sites. This service can support aquaculture operators in selecting the best locations to establish new production sites and regulators in issuing licences or establishing zones for aquaculture production. In the user survey, 94% of respondents found this service useful or very useful.

The second service is the daily management service that is intended to supply the site manager with relevant and timely information, empowering him or her to make informed management decisions. A large number of data products have been identified together with the users that are available in the AQUA-USERS system for operational management. The data include NRT satellite data, including HAB risk, environmental information from models, buoys and in-situ measurements as well as long-term statistics and production limits. In the final user survey, 94% of respondents found the app very useful or useful to some extent. 92% evaluated the HAB risk maps as useful. In particular for the NRT HAB service, the value adding capability has also been demonstrated in the socio-economic impact analysis.
Project Context and Objectives:
With global population expansion, the demand for high-quality protein is rising dramatically, and fish farming is gaining importance to ensure food security. Aquaculture is the fastest growing food production sector worldwide. Environmental conditions determine the growth and health of the produced species, while the production also has impacts on the surrounding environment. Therefore, monitoring is needed on several levels. However, up to now, the available data is often disparate, inconsistent in coverage and of unknown quality.
To support the growth of efficient and sustainable aquaculture production, AQUA-USERS aimed at providing the aquaculture industry with user-relevant and timely information based on the most up-to-date satellite data and innovative optical in situ measurements. The key purpose was to develop a web portal and mobile application that bring together satellite information on water quality and temperature with in situ observations as well as relevant weather prediction and met-ocean data. Specific focus during the project was put on the development of indicators based on Earth Observation data for aquaculture management including indicators for harmful algae bloom events.
AQUA-USERS was a highly user-driven project with a user board consisting of companies and organisations from 5 countries representing different European aquaculture production systems. Together with the user board, the project partners demonstrated the applicability of the developed methods and tools in two case studies:
• Site characterisation and selection based on historic satellite data
• Daily management using near real-time satellite data and in situ measurements
Project Results:
1) User requirements and interaction
One of the overarching objectives of the AQUA-USERS project was to involve potential users for the products developed by the project by establishing close contact with key users from different geographical locations involved in a diversity of aquaculture activities including fish, shellfish and algal culture. At the outset of the project the key users included two from Norway, two from Scotland, one from Denmark, one from the Netherlands and two from Portugal. As part of promoting interaction between the key users and the project partners, a user board was set up at the first meeting of the project where one of the first outcomes of the board meeting was to produce a questionnaire focused on the characteristics of the user’s production and the environmental and production parameters that were of interest. This information culminated in an initial users’ requirement document that enabled the second questionnaire focusing on the projects two case studies: site selection and daily management using the WISP-3 radiometer.

At the mid-term user board meeting in Venice attended by representatives for Erfjord Stamfisk (Norway), Marine Harvest and Scottish Salmon Producers Organisation (Scotland), Aquapri and Danks Akvkultur (Denmark), Hogeschool Zeeland for Prins & Dingemanse (Netherlands), and Finisterra (Portugal), the partner presented the mid-term status of the project and the users provided information on their sites, especially related to site selection and management. Some very useful discussions then took place between the partners and the users allowing a better understanding of the difficulties and constraints experienced by the users from the different countries. This information provided the basis for the final users’ requirement document, which showed that users were highly interested in met-ocean data and in situ data that could be retrieved from the WISP-3 or derived from satellite products. Additionally, users would like to be able to predict of HAB events and storms.
The final meeting of the user board was at the European Aquaculture Society meeting in Edinburgh in September 2016. This meeting had representatives from Norway, Scotland, Denmark and the Netherlands, with apologies from the Portuguese company, due to work commitments. Although the AQUA_USER products could only be presented towards the end of the project there was a positive feedback, including from Finisterra in Portugal. All the key users contributed to the questionnaire developed for Product and Service Validation and the Socio-Economic Implication and there is an ongoing interaction with them beyond the end of the project.

2) Method development
One of the major activities within AQUA-USERS was developing new and improved methods for making EO data useful for the aquaculture industry. Regional optical algorithms for retrieval of water quality parameters have been tested and validated for the five study regions. The selected and adjusted algorithms have been used for populating the satellite data archive. Based on satellite and in situ data, a method has been established to derive indicators for potential benefits and risks for aquaculture production based on biogeochemical variables. To improve the detection of harmful algal blooms with optical methods, two approaches were pursued: a) training a detection algorithm with multi-spectral satellite images of known blooms of certain species, and b) modelling of hyperspectral HAB data based on laboratory instruments. The improved algorithm has been applied to a large number of satellite images for all study regions. Methods for decision support by multi criteria analysis have been investigated for their applicability within AQUA-USERS.

2.1) Regional optical algorithms
As aquaculture operations are often located in coastal areas with complex water types, finding suitable algorithms still poses a challenge, despite the progress that has been made in the last decades in developing water quality algorithms for complex waters. Therefore, validation studies have been performed for the five study areas of AQUA-USERS to test and validate suitable processors for deriving water quality parameters from MERIS observations. The water types and specific challenges in the different areas vary considerably - from the highly turbid and very dynamic waters of the Wadden Sea to the highly absorbing waters in the Norwegian fjords strongly influenced by adjacency effects from land. For all areas, in situ measurements have been collected from various sources for validation of satellite data. Several candidate processors, selected based on knowledge of the areas and previous experience, were tested and compared per area based on the in situ data.

For the Danish area, the three neural networks FUB (v. 2.2) C2R (v. 1.6.2) and CC were calibrated to reference data, and the results of the calibrated as well as original algorithms were evaluated statistically and by inspection of time series against in situ measurements. The performance based on the statistical measures was very similar for the three calibrated algorithms, therefore C2R (C2R calibrated) was chosen based on the best agreement with the in situ time series.

For the Dutch area, the atmospheric correction component of the three neural networks FUB, C2R and CC has been validated against in situ reflectance measurements available in the MERMAID database (also data from adjacent areas have been considered). The performance of all three algorithms was quite similar, with C2R slightly outperforming the others. Therefore, C2R was chosen for AC in the Dutch area. For WQ parameter retrieval, CC, C2R and the WISP algorithm were validated against in situ measurements of CHL and TSM. For operational processing, CC was found to be the best choice with a scaling factor applied to TSM.

For the Norwegian area, a comparison of L2 products with in situ data showed that C2R fits best for the TSM concentrations, and FUB works best for the CHL-a concentrations.

For the Portuguese areas, a detailed study has been performed comparing MERIS Algal 1 and Algal 2 relative to in situ data, which has been published in Cristina et al. (2014) and Goela et al (2013, 2014), and this data is now used to develop a regional algorithm for Sagres. Validation results, based on matchup analysis, identified a systematic overestimation of standard Algal_1 versus the reference Chla values. The additional comparison of product maps in selected regions of interest confirms this finding, and demonstrates the feasibility and relevance of using regional algorithms for investigating space-born products.

For the UK waters, Chl is generated by the OC5 algorithm (Gohin et al. 2008), Kd(490) by the standard KD2 algorithm (Mueller 2000), CDOM by applying the Quasi-Analytical Algorithm (QAA) (Lee et al. 2006) and non-algal SPM by the IFREMER semi-analytical algorithm (Rivier et al. 2012). The comparison of the results from satellite inversions with modelled data provide confidence on the use of the proposed algorithms for indicators of water quality for aquaculture.

2.2) HAB detection
The information about harmful algal blooms is highly important and valuable for management in the aquaculture industry. In high concentrations, harmful algae may cause respiratory irritation for humans, shellfish neurotoxic poisoning, fish mortality and can lead to significant economic losses by damaging fish farms (Stumpf et al., 2003; Miller et al., 2006). Consequently, significant efforts were put in AQUA-USERS towards timely detection of HAB events and providing this information along with other parameters of water quality for the users of aquaculture sites to assists them in taking management decisions.
The developed method for HAB detection uses spectral measurements of water leaving radiance to estimate harmfulness of algal blooms (Kurekin et al., 2014). It employs a fully automatic data-driven approach to identify key characteristics of water leaving radiances and derived quantities, and to classify pixels into “harmful”, “non-harmful” and “no bloom” categories. The application of HAB classifier consists of training and classification stages. It includes collection of training data, automatic selection of ocean colour features for different HAB species using stepwise discriminant analysis (SDA) and calculation of discrimination function parameters in the feature hyperspace.

Two alternative approaches have been developed for training the HAB classifier to discriminate different HAB species identified in the key areas. The first approach is based on using multispectral measurements of HAB events provided by satellite ocean colour sensors. The image scenes with HABs were identified using case studies and historical records of HAB events. The HAB classifier was developed for the following species: K. mikimotoi, Pseudo-nitzschia spp., Noctiluca scintillans, L. polyedrum, Phaeocystis globose, Pseudochattonella sp., Pseudo-nitzschia, Dactyliosolen fragilissimus. The classifier performance was estimated on training data by using cross-validation method with the total probability of correct classification Pt>87%.

The second approach for training the HAB classifier was based on growing cultures of HAB species under laboratory conditions, measuring their inherent optical properties (IOPs), and then modelling the water reflectance for the specific HAB species. Cultures of five harmful algae species were grown at the University of Lisbon: Lingulodinium polyedrum, Gymnodinium catenatum, Karenia mikimotoi, Pseudo-nitzschia australis and Phaeocystis sp. The IOPs of these species were measured with two Wetlabs instruments, an ACS and an ECO-BB3, placed in-line. The measurements of water IOPs for several HAB species obtained during the laboratory experiment were applied to extend the training dataset and improve the results of the HAB classification. In the experiment the HAB classifier was trained to discriminate the K. mikimotoi using modelled water reflectance and demonstrated better accuracy than the classifier trained on MERIS data. Overall, the total probability of correct classification Pt was improved by 10.7% and the false alarm rate of K. mikimotoi bloom classification has reduced from 1.8% to 0.05%.

The developed classification algorithm was applied to generate the HAB risk maps of user sites in SW Portugal, UK, the Netherlands, W Norway and Denmark. Individual maps were generated for the key phytoplankton species that affect aquaculture in these regions: Lingulodinium polyedrum, Phaeocystis globosa, Karenia mikimotoi, Dactylosolen fragilissimus and Pseudochattonella farcimen. The HAB risk maps were produced by processing the AQUA-USERS satellite data archive for the period from 2002 to 2012. Validation was carried out by comparison of HAB risk maps with historical records of HAB events, in situ measurements of chlorophyll concentration and cell counts of harmful species.

Visual analysis has been applied to overcome the difficulties of comparing satellite and in situ data when the number of matchups was relatively small due to limitations of the satellite resolution, cloud cover, and optical complexity in the near-coast pixels. To achieve this the HAB risk maps (7-day composites) coincident with HAB events were visually compared with the location of the sampling sites producing the highest counts of HAB species.

2.3) Aquaculture indicators
Aquaculture monitoring is imperative not only because environmental conditions are crucial to ensure the growth and health of the produced species, but also because the production often releases large amounts of nutrients affecting the surrounding environment. Monitoring is therefore essential on both levels; however, difficulties arise from lack of quality data, and inconsistent coverage. For that purpose, satellite remote sensing (RS) can be an effective tool as it can provide the necessary time and spatial resolution at reasonable costs.

Measuring the oceans by satellite remote sensing has proven to be a unique resource to achieve local to global scale observations of a vast range of oceanic parameters (Robinson, 2007). Ocean Colour provides estimates of variables such as chlorophyll-a concentration (Chl-a), which is a proxy for phytoplankton biomass and ocean primary productivity (e.g. Behrenfeld and Falkowski, 1997), and thus, it can be used as an ecological indicator for estuaries and coast sites (Platt et al., 2010).
Aquaculture management involves different actions (e.g. lowering cages, harvest, feed) that change according to type the of aquaculture (e.g. mussel, fish, seaweed), but, when considering monitoring purposes, there is a set of parameters that is common to the needs of the majority of users. All users in the AQUA-USERs user-board expressed their interest in having information on weather and sea state conditions, as well as water temperature and Chl-a concentration data. The latter is referred to as a relevant parameter to monitor as it can be used as a biomass proxy (i.e. food availability for mussels) or as an indicator of water quality. For weather and sea state conditions, instant information may be sufficient for real-time management; however, temperature and Chl-a data have also to indicate gradients and fluctuations in the environmental conditions for effective evaluation and decision making. Therefore the evaluation of the natural variability of these parameters was prioritised. Ten years of satellite data (2002-2012) were processed for all the users’ sites and regional-specific time-series were determined for Chl-a, Sea Surface Temperature (SST) and total suspended matter (TSM).

For each time-series the 10th, 50th, and, 90th percentiles, the average and the standard deviation, were computed for each 7 days of the year (note that the 50th percentile is also known as the median).
The climatological data with respective statistics data have been implemented in the AQUA-USERS app to provide alert conditions to site managers. Users will be able to compare near-real time measurements with historical climatological data. The app will also save the new measurements taken by the user to allow the analyses of trends.

As mentioned, information on marine environmental parameters like Chl-a is essential to monitor water quality, not only for aquaculture managing, but also to comply with environmental directives and to evaluate aquaculture impact in the surrounding media. The Chl-a 90 percentile (P90) has been recommended as an indicator of eutrophication in European coastal waters and different studies have highlighted the effectiveness of using ocean-colour remote sensing data with that purpose (Gohin et al. 2008, Novoa et al., 2012). The monitoring of areas relative to their natural variability should alert to abnormal conditions, eventually undesirable noxious blooms.
Two case-studies have been evaluated to exemplify the applicability of the thresholds derived based on climatology in the detection of HABs. Two historical bloom occurrences were analysed, to check if the corresponding Chl-a records would be identified as an anomaly. A bloom of the toxic dinoflagellate Lingulodinium polyedrum that occurred in the coast of Algarve, on the 17th August 2004, which attained 1x106 cells/L. Although no Chl-a measurements were taken in situ, according to our pigment results obtained in cultures, this cell concentration would correspond to a Chl-a range of 12-20 mg m-3, which should be added to an unknown value for the rest of phytoplankton community (Brotas et al. 2014). The second is a Karenia mikimotoi bloom occurring off Cornwall in June 2010, with a cell abundance of 1x106 cells/L (Kurekin et al 2014). In both cases, pronounced Chl-a anomalies were found.

2.4) MCA / Decision support
A decision support for daily management operations had been envisaged for daily management operations at the aquaculture sites to the users. The idea was to base suggested decisions on previous recorded decisions by the user under similar environmental conditions, and their evaluation of how successful the management decisions were. However, it became clear that during the lifetime of the project, it would not be possible to collect enough data for to derive robust decisions. It was still considered useful to ask the users to record management decisions and their subsequent evaluation. However, the step to base a management recommendation on this data has not been taken.

3) Technical development
AQUA-USERS is based on aggregation and analysis of data gleaned from a wide variety of sources, including Earth observation (EO) and model data. This section summarises the EO data processed and tools implemented to achieve the aim of assisting aquaculture farmers with day-to-day and longer term decision-making related to water quality and risks. These tools are the web portal and the mobile app, which are to be made commercially available to the aquaculture industry.

3.1) Satellite data archive
MERIS was the most relevant sensor for aquaculture applications such as estimation of water quality, chlorophyll-a concentration, and HAB discrimination, due to its high-quality ocean colour channels and 300m resolution allowing observations much closer to the coast, estuaries and fjords where farms are located, compared to other polar-orbiting sensors such as Aqua-MODIS or Suomi-VIIRS with 1000m and 750m resolutions, respectively. It was originally proposed that ocean colour data from the follow-on Sentinel-3 OLCI sensor would enable comparable data to be provided in near-real time during the final year of the project, but unfortunately several launch delays meant that we had to revert to the coarser MODIS and VIIRS data for this purpose.
The satellite processing was shared across three of the AQUA-USERS partners: PML, GRAS and WI. Full details of satellite data processed are in deliverable D4.1 “Implementation of satellite archive”.
PML processed the entire MERIS 300m ocean colour archive for the period from 2002 to 2012, covering two geographical regions in the UK (Cornwall and Scotland) and two regions in Portugal (Algarve and Iberia). Products comprised remote sensing reflectance (Rrs), absorption, backscatter, chlorophyll-a estimated using the OC4v6 (NASA) and OC5 (IFREMER) algorithms, suspended particulate matter (SPM), and coloured dissolved organic matter (CDOM). Sea-surface temperature (SST) was generated using AVHRR data.
GRAS processed the entire MERIS 300m archive for the Inner Danish Waters including eastern North Sea and the western Baltic Sea. A similar set of products were generated using the Case2Regional, FuB/WeW and CoastColour packages.
WI processed the entire MERIS archive for two geographical regions: the Dutch coastal waters and the south west coast of Norway. For the Dutch coastal area, products were generated using the C2R (case 2 regional) neural network approach, WISP, and Lee 2005 algorithms. For Norway, parameters were calculated using the FUB/WeW neural network. For both areas, SST data were acquired from the GHRSST Level 4 MUR Global Foundation Sea Surface Temperature Analysis dataset.

3.2) Web portal
During AQUA-USERS, a web portal was developed: a simple-to-use but powerful web application that allows users, both scientific and non-scientific, to display and compare selected data sets in the context of a background map and generate plots showing the variation of the selected datasets over time. The portal was developed to allow the selection, visualization and analysis of any geospatial data, although we focussed on EO-based indicators of ocean colour and temperature, in particular for the 2002-2012 archive of 300m resolution data from the Envisat MERIS ocean colour sensor. You can access the portal at
Key features include: easy browsing through a long time-sequence of products using a time slider; fast panning and zooming through large EO maps; time-series chart generation for an arbitrary-shaped region of interest; interactive changes to colour palette to highlight features of interest, and transparency for combining several products.
Although the portal was implemented in year 1, further functions were incorporated during the remainder of the project, along with many EO data layers, including some near-real time data on sea-surface temperature, chlorophyll-a, and HAB risk maps for two of the key species: Karenia mikimotoi and Pseudo-nitzschia. The time-series chart functions proved particularly useful, even within the consortium, for analysing and validating the regional algorithms and HAB risk at specific locations.
Data are provided to the portal through community-standard interfaces and metadata formats: Open Geospatial Consortium standard interface Web Map Service (WMS), and Web Coverage Service (WCS). This allowed large EO datasets to be distributed across PML, GRAS and WI institutes, and served to the Web Portal using Thematic Real-time Environmental Distributed Data Services (THREDDS).

3.3) Mobile app
The mobile app is the front end of the AQUA-USERS system that allows the users to interact with the information stored in the AQUA-USERS database. When a user logs into the system, the first step is to retrieve up-to-date data for his/her production site(s). The app sends a request to the database to retrieve all relevant information stored there. An overview of all the data sources connected to the database is given in D6.2. All recent in situ measured data are presented to the user, and, where available long-term statistics and an indication of production requirements are given to help the users make sense of the data. The user has also the option to enter additional measurement data via the app.

In the next screen, the user is asked to record the management decision taken. The decision together with all the data that was available to the user at the time of decision making is stored in the database. The final screen allows the user to view and review all previous decisions. This evaluation is also stored in the database and will allow a more detailed analysis of the decision process and the considerations that are important to the users.

4) Data collection
Data collection has been a vital part of AQUA-USERS, therefore common sampling protocols and methods for quality control were established quite early in the project. A data policy was also established. This policy is a guide to help generate, maintain and safeguard high-quality data, and to share and gain access to data within the project according to levels of sensitivity agreed upon by project consortium.

4.1) Satellite data
Automated scripts were implemented to harvest and average recent EO data within a small neighbourhood (10 x 10 km) of each aquaculture site. The Web Coverage Service provided a standard interface to allow any of the massive EO datasets to be queried remotely from WI for data values for specified dates and coordinates, which were then inserted into the AQUA-USERS Database for display on the mobile app.

4.2) In situ data
An important aspect of the AQUA-USERS project was the collection of in situ data and integration with the other data sources used. During the whole project, all core users together with their service providers have been provided with a WISP-3 spectroradiometer to measure optical parameters (Chl-a, TSM, CDOM and Kd) and the spectral reflectance and inherent optical properties. More than 1200 WISP-3 measurements have been collected at dozens of sites in five countries. In addition to this large spectral in situ dataset, several field campaigns have been performed by the partners and users as well as routine sampling by the users at their production sites. These campaigns included a variety of environmental parameters, depending on the specific environment and research question.

In particular, in Norway and Portugal, intensive measurement programmes were carried out to understand the complex relationships between environmental conditions and production variables such as fish health or mortality and mussel growth.

4.3) Data from external sources
In addition to data from satellites and in situ data, data were also collected automatically from multiple external sources.
The data from external data sources are primarily physical variables of relevance to the aquaculture operation. The list of retrieved variables includes but is not limited to meteorological variables, e.g. wind speed and air temperature; ocean variables, e.g. current speed, water temperature, salinity and wave height.
The primary source of ocean model data is Copernicus Marine (formerly MyOcean), which provides a range of models, both global and regional, covering all areas of interest. The spatial resolution of the Copernicus data is by necessity relatively coarse, and using the data directly without downscaling models is not optimal.
Data from automated buoys available via the EMODnet portal were also used when a buoy was located close enough to be relevant for a particular aquaculture site.

5) Case studies
To bring the findings and technical developments produced by AQUA-USERS into real life, case studies were conducted. The first case study applied the analytical framework set up for site characterisation and site selection. The cases comprised different types of aquaculture in different European regions. The second case study was planned to test and improve the first version of the AQUA-USERS app, which was primarily to be based on optical data from the WISP-3 instrument and earth observation data of chlorophyll. After populating the AQUA-USERS database with more data and expanding the functionalities, the third case study should validate the final app service as well as the web portal. Due to changes in the scheduling of precursor activities, it was however decided to merge case study two and three into one.

5.1)Case study “Site characterization/site selection”
Being able to characterise the conditions important to aquaculture production is essential for future expansion of the industry. Knowledge on specific sites and on regional variability are necessary to select the most optimal production sites where conditions support maximum productivity combined with acceptable environmental impacts. Data, which describes these conditions spatially and temporally are therefore of high value. To demonstrate this an analytical framework was developed and tested together with the end users.

The aim of some cases was to improve the knowledge of an existing production area (site characterisation) while for the other cases it was to support selection of new production sites (site selection). Site characterisation supports the daily operation of the aquaculture production by increasing the knowledge of the production. Site selection does on the other hand have a longer perspective by focusing on the optimal locations for future production sites. Site selection does also support administrative management in regulation and licensing.

The case studies were defined in close cooperation with the end users. In all, five cases were studied using available spatial and temporal data. The data comprised processed remote sensing observations, historical in situ measurements and results of deterministic modelling. The cases covered finfish (salmon, rainbow trout) in Scotland, Norway and Denmark as well as mussel production in Portugal and the Netherlands.

Based on the objectives of the case, determinant factors for the site characterisation/selection were identified and (historical) data collected. In most cases, the primary data source was Earth Observation data, but in situ data measured by the project and external parties as well as deterministic modelling data were also important as data sources. For each case, the analytical framework was adapted to the case and implemented and analysed in a GIS environment. The outcome of the GIS analyses was maps describing production sites / identifying sites suitable for aquaculture production of the given species and in the given geographical area.
The conducted case studies are described in Deliverable 6-1. After submission of the deliverable, further work was done on the case for selection of suitable sites for trout production in Danish marine waters to refine the data processing and the selection of indicators. At the end of the project, this led to a contract between DHI and the Danish Ministry of Environment and Food Production building on the work done in AQUA-USERS. This project expands the concept further and involves more analytical approaches.
In summary, the AQUA-USERS case studies confirmed the applicability of defined framework and the concept of integrating spatial and temporal data from different sources.

5.2) Case study “Daily management”
After introduction of the AQUA-USERS web portal and app, users and stakeholder were interviewed via the AQUA-USERS web questionnaire. The user survey showed that one added value for the AQUA-USERS app would be to bring the data provided to the next level comprising interpretations of the data with specific focus on key information required in daily management. Such derived data provision was not part of the scope of the AQUA-USERS project, however for demonstration purposes, examples of further analysis were provided.

To support aquaculture daily management is a challenge that requires data from different sources. The integration of various external sources was achieved in this project and users can access data both historical and near-real time, from models, buoys, satellite or in situ measurements available for their site.
During Case study 2, the assessment and validation of data and products provided to users through the app and web portal were assessed and optimized. Overall, satellite data proved to be very useful in the characterization of the sites natural variability and establishment of thresholds. However, the location of the sites, mainly in coastal waters or in enclosed bays or fjords present a challenge to the use of satellite data. This was particularly verified in the Norwegian case where the assessment of Chl-a natural variability was not possible to determine due to lack of good satellite data for the region. Regarding the HAB products, detection methods rely in satellite data available and training of algorithms with past events. Although validation results were encouraging for the detection of K. mikimotoi in UK or L. polyedrum in Portugal, other regions revealed the need of collecting more data with wider geographical and seasonal distribution to allow further adjustments.
The amount of data gathered and processed taking in consideration the users requirements and made available through the AQUA-USERS app and web portal are important tools for the daily management of aquaculture sites. However, to develop a daily management service that also includes a warning of upcoming problems it is important to identify the relationship between the water quality parameters monitored and the response of the fish and mussels to these parameters. First attempts to establish these links have been made in this case study.

In Portugal, where mussel growth was linked to Chl-a concentration (r=0.66 n=6), which is an indicator of food for filter feeders and first trials were conducted to simulate the growth of an individual mussels with the in situ environmental drivers and additionally experiments are planned to model growth at farm scale. However, more samples are needed to establish the relationship between growth and environmental drivers as well as to improve the parameterisation of the models.

For seaweed farming, the occurrence of phytoplankton blooms impacts the growth of the macroalgae. One re-occurring bloom influencing production is the spring bloom and early warning of the onset of these blooms can benefit farming of seaweeds as ideally harvest time for the seaweed is just before the bloom. To investigate if temperature can be used as a precursor warning of a coming bloom, for a seaweed culture farm in Norway a possible relationship between time series of sea surface temperature (SST) and the phytoplankton biomass (expressed by chlorophyll-a) was investigated that would confirm the usability of satellite SST observations to forecast the onset of the spring bloom. The case study showed that for the area and the 5-year period investigated, a relationship exists between the SST and the onset of the algal blooms. These first results are promising but further investigations are necessary to validate the predicted onset with actual occurrence of species such as bryozoan. Only if the results can be further validated, can a service be developed to forecast optimal harvest time based on SST data.

Finally, for a fish farm in Norway, an initial analyses of the link between water quality and fish welfare were undertaken. Within the Norwegian salmon industry, relatively high mortality rates in the grow-out phase in the sea, and incidents of high unexplained mortalities during salmon lice treatments have raised the awareness about water quality, and how it impacts fish welfare either as a single factor or in combination with e.g. pathogens. Considering only the data collected during the AQUA-USERS period, there was no apparent relationship between mortality and algal biomass, TSM or Secchi depth. However, analysing AQUA-USERS collected data together with the historic data from the site revealed some clear correlations between environmental variables and mortality. The most robust result was that increased temperature and increased total abundance of phytoplankton led to increasing mortality. Among phytoplankton taxa, a high abundance of Prymnnesium parvum was associated with increased mortality. For the same total amount of phytoplankton, mortality decreased with increasing abundance of diatoms and flagellates. An easy measurable water quality parameter such as Secchi depth was not related to mortality. More research is required to reveal complex interactions between fish welfare and water quality that can form the basis for robust indicators to build into monitoring services and products, in order to reduce losses.

In summary, there is potential to develop a daily management service that also includes a warning of upcoming problems, but interactions between the different environmental factors and the specific aquaculture are very complex and significantly more research is needed to develop such a service.

5.3) Product and service validation
The work within AQUA-USERS has been carried out together with the user board, where the user board represents the interests and points of view of both end-user organizations and customers. The user board has throughout the project provided an independent assessment of users' needs and priorities and validates the service development. Furthermore the requirements defined by the user board have guided the method and tool development within AQUA-USERS in an iterative process to ensure that the services meet the users' needs. The validation of products and services and their performance and utility was the last step in this iterative process, and provided crucial information necessary for the further development of a commercial relationship with the involved users as well as potential new customers inside and outside Europe.

In quality management systems, such as the ISO 9000, “validation” is together with “verification” a very important component. In principle, “verification” and “validation” are different but, in practice, the usage of these terms varies and sometimes they are even interchanged. In everyday language, verification is assessing whether “You built (designed) the product/service right” while validation is assessing whether “You built (designed) the right product/service”. The validation process uses objective evidence to confirm that the requirements which define whether an intended use or application of a product or service have been met. This is most often an external process and involves acceptance of fitness for purpose from end users and other product stakeholders.

The requirements defining the intended use of the AQUA-USERS products have been identified through an ongoing dialogue with the user board throughout the project. Requirements have thereafter been discussed and the technological progress towards fulfilling the requirements has been on the agenda of user board meetings.
A full verification and validation procedure to the quality of an ISO management system was not possible because of technical delays in producing the AQUA-USERS products. Nonetheless, despite the technical difficulties and the delay in the launch of Sentinel 3, the AQUA-USERS app, the Web portal, the HAB risk maps, the site characterization and selection service products were all available towards the end of the project and through the use of demonstrations, interviews and online questionnaires, we were able to carry out an initial verification and validation of the products with the user board members and other potential users of the products.

The questionnaire revealed that all products were considered useful both among the user board members and the wider group of potential users. Between 86 and 95% of the respondents replied “very much so” or “to some extent”, “strongly agree” or “agree” when they were asked to rate the usefulness of the products/services. A high proportion (91-95 %) answered that the products gave and added value for their purposes, but the willingness to pay for the products/services was somewhat lower (47-71 % of those who answered the question confirmed a willingness to pay).

The questionnaire also had some general questions regarding monitoring of water quality and HABs. The replies revealed that a significant proportion (around 60%) of the respondents (both user board members and other potential users) had either experienced HABs or knew about HABs affecting their area, and more than 80% replied that they would use a HAB prediction/warning service, where such a service was available.

Potential Impact:
1) Socio-economic impact and the wider societal implications
AQUA-USERS has demonstrated value adding capability and has developed geo-information service provision to coastal aquaculture. A straightforward socioeconomic analysis was performed through a targeted survey of several potential user groups, including stakeholders from the aquaculture, tourism and health sectors. Additional socio-economic spin-off impact was assessed through exploration of the value chain for the Dutch and Portuguese mussel sector.

1.1) Impact of prediction of HABs on the Dutch mussel sector
Better prediction of HABs will give time for the mussel production companies, wholesaler or distributor to take mitigating measures. Mitigating measures can only be taken before the mussels are harvested. The impact of HABs on the mussel yield and the accompanying damage to the mussel production companies, fish wholesalers, distributors and consumers will be reduced.

Since a prediction will always include some uncertainty, there is a chance that the prediction is wrong. The costs of unnecessary mitigating measures will cause additional damage. However, some parties already use a prediction system, e.g. a national HAB monitoring system, one used a satellite service, own observations or other newspapers. Also, the predictions of these systems will be uncertain. We assume that the AQUA-USERS satellite prediction will reduce the uncertainty level. Therefore, we assume that less unnecessary mitigating measures are taken.

The number of HABs in the Dutch coastal waters have been somewhat limited, though an estimated economic damage in terms of 35 % reduction of turnover is substantial, and major damage was reported for the Phaeocystis bloom in 2001. The analysis shows that the potential damages of HABs for the mussel production industry, fish wholesalers and distributors are limited due to impact of a lower supply on the price. However, there may be substantial indirect damages to consumers, distributors and the processing industry.

Currently, the value of earth observation information to predict HABs will be limited due to the low probability of HABs in the Dutch waters. In the future, the probability of HABs may increase, which will increase the value of better HAB prediction. The largest benefit will be the reduction of the indirect damages. The timing of the prediction is crucial for the reduction of the impact. In summary, an on-time prediction of HABs in a situation with more HABs will be most beneficial.

1.2) Impact of prediction of HABs on the Portuguese mussel sector
Prediction of HABs in the Portuguese region would allow the mussel farmers to harvest before bioaccumulation of toxins by the mussels. However, in this area, phytoplankton species produce toxins at low cell/L concentrations constraining the role of satellite remote sensing. The most common blooms such as Pseudo-nitzschia and Dinophysis, do not discolour the water and in the case of Dinophysis, the blooms are usually sub-superface which limits the capability of detection by satellites. Better optical characterization of the toxic species and the future hyperspectral satellite missions may be beneficial. For the detection of intense blooms of Lingulodinium polyedrum, the AQUA-USERS HAB detection tool may be promising, but currently, the value of earth observation information to predict HABs in the Portuguese waters is limited. Timing of the prediction would be crucial for the reduction of the impact.

2) Main dissemination activities and exploitation of results
2.1) Dissemination
AQUA-USERS principle sources of outreach to the wider community are through its web site at which was kept up to date throughout the project, a twitter account with an increasing number of followers (more than 200 at the end of the project), two newsletters, flyers and a policy briefs.

In addition, over the duration of the project, there has been continuous contact with the key users through questionnaires, meetings, and demonstration of products, to involve them in the development and the validation of the products, using as far as possible, data from their farms. At the same time, there has been a continuous effort to disseminate the project through a variety of oral and poster presentations at events aimed at both national and international audiences in both science and industry. These comprise 26 events with 42 presentations representing all the partners in AQUA-USERS. The project has so far yielded a total of 10 scientific publications, with several more being currently prepared.

Apart from these events, the project has been disseminated through trade journals and other fora for non-specialists, numbering 8 so far (more in preparation). Finally, the project has participated in 6 workshops with stakeholders including one in Norway, two in Denmark, one in Portugal, one in Scotland and one in Chile. The one in Scotland was particularly important, as project partners were able to present AQUA-USERS to a workshop and a trade booth at the EAS conference at Edinburgh in September 2016. This effort enabled the partners to obtain many responses from stakeholders to the questionnaire on the socio-economic impact and the wider societal implications of the project presented in section 1 above.

2.2) Exploitation
Right from the start of the project, the consortium focussed efforts on exploitation and the business plan, with a draft and updated plan (D7.4) and a separate final plan at the end of the project (D7.9). Activities included the definition of the final AQUA-USERS services and their set up, a market analysis, analysis of competitors, SWOT analysis, cost estimation of the services, the IPR model, a business model as well as a discussion of the future plans with the services of the individual partners. Two services were identified: a site selection service and a daily management service. These services are composed of data products and tools developed within AQUA-USERS (such as the app and the portal, the WISP-3 and the HAB risk maps).

While there is a larger number of stake holders, the key target customers are aquaculture producers. The market analysis focussed on the HABs as there is a clear business case for early warning of HAB events, based on the documented losses that aquaculture producers suffer from HABs. The evaluation of the services in the user questionnaires was very encouraging. The analysis of 51 responses to questionnaires has shown possibilities for offering a service level agreement to include:
- assembled information for daily-management in the app;
- access to near real time HAB maps for specific locations in the web mapping portal;
- a strategic site characterisation /selection analysis.
65% of respondents found the app very useful and 29% to some extent. 62% of the respondents confirmed that they would be willing to pay for a subscription to the app. 49 % strongly agreed and 43% agreed that NRT HAB risk maps would be useful for them. As for historic HAB risk maps, 61 % of the respondents found these very useful and 39% useful to some extent. 60 % of respondents indicated that a site selection/site characterization tool that includes Earth observation (satellite) data would have very much added value for them, and 34% indicated some added value. 42 % were willing to pay for site selection analysis.
For future collaboration of the consortium, a flexible model is followed which allows partners to group into service lines when customers request specific services. The composition of the service line for a particular service depends on the kind of service and area. All partners have stated which services they are interested in as well as indicated if any of their (shared) foreground comes with specific conditions for use. So far, there is no commercial AQUA-USERS service set up, but a number of potential users have indicated interest. Also, at least two partners already had some commercial spin off from AQUA-USERS.

List of Websites: