Community Research and Development Information Service - CORDIS

FP7

HIGHROC Report Summary

Project ID: 606797
Funded under: FP7-SPACE
Country: Belgium

Final Report Summary - HIGHROC (High spatial and temporal resolution Ocean Colour coastal water products and services)

Executive Summary:
The HIGHROC (“HIGH spatial and temporal Resolution Ocean Colour”) project had the objective to carry out the R&D necessary for the new generation of coastal water products and services from ocean colour space-borne data, giving an order of magnitude improvement in both spatial and temporal resolution and thereby opening up new applications and strengthening existing ones. The HIGHROC project designed, validated and had evaluated by users this new generation of coastal water products and services from satellite missions with optical sensors. In particular products at high temporal resolution (15 minutes from SEVIRI) and high spatial resolution (10m from Sentinel-2 and Landsat-8) are original additions to existing medium resolution products.

The HIGHROC project developed new algorithms dedicated to coastal waters for high spatial/temporal resolution sensors. Those algorithms were validated with in situ measurements achieved during the project and final algorithms were integrated in processor chains in order to process historic and current (near real time) satellite image databases. During one year, HIGHROC satellite products were provided by partners to key users. This one year user service trial allowed to assess the potential impact of HIGHROC products and services to the marine user community. Key users included marine policy makers, marine environmental agencies and the marine/maritime industrial sector.

HIGHROC activities, products and results were also extensively presented to the wider potential user community via dedicated workshops, presentation during targeted meetings, website and brochures. HIGHROC scientific results were presented to the scientific community via publications (11 peer-review publications during the project period), presentations in international conferences and via the organisation by the HIGHROC Science Conference. This conference gathered experts from the international scientific community to discuss the state of the art on coastal ocean colour remote sensing. Finally, HIGHROC activities were also presented to external organisations and space agencies.

A complete version of the project Final Report, including Figures, tables and information on all non peer-reviewed publications with full paper download links can be found at:
http://www.highroc.eu/assets/pdf/highroc_final_report.pdf


Project Context and Objectives:
The HIGHROC (“HIGH spatial and temporal Resolution Ocean Colour”) project had the objective to carry out the R&D necessary for the next generation of coastal water products and services from ocean colour space-borne data, giving an order of magnitude improvement in both spatial and temporal resolution and thereby opening up new applications and strengthening existing ones. This responds to SPA.2013 topic 1.1-06 “stimulating development of downstream services and service evolution”. HIGHROC was expected to design, validate and have evaluated by users the next generation of coastal water products and services from satellite missions with optical sensors. In particular products at high temporal resolution (15 minutes from SEVIRI) and high spatial resolution (10 minutes from Sentinel-2 and Landsat-8) will be the original additions to pre-existing medium resolution products.

In the previous GMES/Copernicus downstream services MARCOAST project, the potential areas and markets of application of coastal water quality products and services have been well-established and prepared. The monitoring requirements arising from the EU Water Framework Directive (WFD) and the Marine Strategy Framework Directive (MSFD) have been clear drivers for end-user requirements and hence service definition. As examples these directives contain mandatory requirements for monitoring of eutrophication, including Chlorophyll a, and turbidity, both parameters which will be improved and provided by HIGHROC.
Pre-existing ocean colour-based products and services, e.g. Copernicus services for marine end-users for WFD reporting, routinely use data from ocean colour remote sensors such as MERIS and MODIS, now followed by OLCI and VIIRS. Despite these existing services successfully providing data to end-users, the MERIS, MODIS, OLCI and VIIRS sensors have critical limitations of spatial and temporal resolution (typically 300m, 1/day) with respect to user requirements.

The HIGHROC objective was to perform innovative R&D to add to the existing service portfolio new coastal water products from Sentinel-2 (S2) and Landsat-8 (L8) at high spatial resolution and SEVIRI at high temporal resolution giving an order of magnitude improvement in respectively spatial (down to 10m) and temporal (down to 15 minutes) resolution. The resulting merger of Sentinel-3/OLCI (plus MODIS and/or VIIRS) with the new HIGHROC Sentinel-2/Landsat-8 and SEVIRI products will dramatically improve information content for nearshore waters, e.g. 10m instead of 300m resolution within first nautical mile of coast for WFD, and improve data availability in periods of scattered or fast-moving clouds, e.g. 30-50 images per day instead of one. The advantages proposed by HIGHROC over the existing products are thus very clear.
In addition to these dramatic improvements for established services, the new HIGHROC products open up entirely new application areas including support for dredging, windfarm construction and operation and aquaculture, that were previously inaccessible because of the limited 300m spatial resolution of MERIS/OLCI.

To create valuable coastal ocean colour based products and services, HIGHROC activities included (1) discussions and close interactions with end-users and potential user groups to define the user needs and service requirements and (2) the implementation of a one year service trial with evaluation by the users of the new products and services. These service trials required the definition and validation of specific algorithms as well as the implementation of automated processing chains for each type of satellite data used.

A number of users were selected to evaluate the HIGHROC products and services during this one year User Service Trial period. These users include member state officials responsible for reporting of water quality under the EU Water Framework and Marine Strategy Framework Directives; dredging consultants; government officials responsible for assessing environmental impact of offshore constructions (offshore windmills, ports, etc.) and private consultancies responsible for compiling such assessment reports; and a representative of aquaculture activities.

Theoretical work consisted of developing atmospheric correction and level 2 product algorithms (i.e. atmospheric correction) for the S2/L8 and SEVIRI sensors and Level 3 algorithms (e.g. for suspended particulate matter or chlorophyll-a concentration) for multitemporal and synergistic exploitation of the new products with existing products such as those from OLCI and VIIRS. Image processing chains provided full mission historical and near real time products for local areas including the dedicated test sites. In situ measurements were acquired for dedicated test sites and used to validate the new S2/L8 and SEVIRI products. Exploitation of the products was supported by interaction with user partners and potential user groups with particular focus on the opportunities offered by the new HIGHROC products both for entirely new application areas, e.g. assessing the environmental impact of human activities such as offshore constructions or dredging operations, and for significantly improved spatial and temporal resolution for existing applications, e.g. WFD monitoring and reporting. Post-HIGHROC commercialisation of the products and services is being prepared in the H2020/DCS4COP project, which fully includes the HIGHROC Consortium and one new private company partner.

After success in the User Service Trials the HIGHROC products and services can be upscaled (post-HIGHROC) to similar applications in all EU countries, where similar monitoring requirements apply (WFD and MSFD). The potential for exploitation outside Europe is also high, although different implementations of higher level products (e.g. multitemporal composites) will be required under different national legislations, e.g. Australia, US, Canada, etc. The exploitation of HIGHROC products and services for other non-EU countries without existing monitoring of coastal waters is also technically simple, although end-user requirements may be less clear or less motivated by mandatory legislation.

Project Results:
The main science and technology results are structured by the corresponding RTD work packages on Algorithm Development (WP4), In Situ Measurements (WP5), Image Processing (WP6) and Product Validation (WP7). This structure is adopted in the following subsections. More information, including Figures and Tables can be found in the attached Project-formatted Final Report.

*WP4: Algorithm Development*

Objectives
The objectives of WP4: Algorithm Development were:
• To develop the algorithms required for the generation of coastal water products from a) Sentinel-2 and Landsat-8 (S2plus) and b) Geostationary sensors (GEO)
• To select and define the algorithms to be used within HIGHROC for the generation of coastal water products from the other medium resolution ocean colour sensors: S3/OLCI, MODIS, and VIIRS (S3plus hereafter).
• To develop algorithms for multi-sensors and multi-temporal (daily, monthly, seasonal and annual) products.
• To develop efficient pixel identification algorithms, particularly for cloud and cloud shadow masking.

Results: Algorithms for atmospheric correction (AC)
Atmospheric correction algorithms have been designed and tested for each type of sensors (S2plus, GEO and S3plus). HIGHROC algorithms were designed to be accurate in coastal waters, including turbid waters.

1. S2 and L8
A SWIR based atmospheric correction was developed for Landsat-8 and Sentinel-2 and publicly released in the ACOLITE processor. The use of SWIR bands is preferred in extremely turbid or productive waters where the assumption of a fixed red/NIR water reflectance is no longer valid. The SWIR based aerosol correction was especially useful in the Belgian coastal zone and French river plumes and estuaries. A band filtering technique was developed to reduce noise propagation from the low SNR SWIR bands into the VNIR bands.
The atmospheric correction module ICOR (previously called "OPERA") has also been developed. iCOR can be used to correct both land and water (inland and coastal) scenes and includes the adjacency correction SIMEC (SIMilarity Environment Correction). SIMEC can be used as well as a pre-processor for any type of atmospheric correction including ACOLITE). ICOR consists of 4 modules: land/water masking, land based AOT retrieval, adjacency correction and atmospheric correction. The atmospheric correction parameters are derived from pre-calculated MODTRAN-5 Look-Up-Tables (LUTs) in function of among others sun and view geometry, aerosol optical depth, ozone, water vapour and elevation. The aerosol optical thickness (AOT) is derived above land through a TOA radiance inversion of selected end-members in the scene following the approach described in Guanter et al. (2007). Over water the AOT is retrieved through spatial extension of the derived values of neighbouring land pixels assuming local spatial invariability of the aerosol.
ICOR has been made operational for Landsat-8 and Sentinel-2.

2. GEO (SEVIRI)
The theoretical basis for atmospheric correction of GEO (SEVIRI) data was summarised based largely on the approach previously implemented by [Neukermans et al, 2009, 2012]. Rayleigh correction is based on a look up table generated by 6SV. Absorbing gas corrections are implemented via atmospheric transmittances using auxiliary data for gas concentrations where available. Aerosol correction is based on deriving aerosol type from clear waters and extrapolating in space, assuming uniform aerosol type over turbid waters. A SWIR based atmospheric correction was implemented which avoids the need for assumptions on the water leaving reflectances. This is particularly useful for regions outside the HIGHROC study areas such as the Amazon river plume and the West African coast.

3. S3plus (MODIS, VIIRS and OLCI)
HIGHROC considers “S3plus” to represent S3/OLCI plus one or more of the other medium resolution ocean colour sensors ENVISAT/MERIS, MODIS/TERRA, MODIS/AQUA and/or VIIRS. In general, mature atmospheric correction algorithms for the S3plus sensor family already exist and the corresponding L2R atmospherically-corrected products were used directly by HIGHROC. In cases where the standard products were considered suboptimal, HIGHROC implemented its own algorithms as extensions of existing processors, e.g. namely ODESA for MERIS and/or OLCI and SeaDAS for TERRA, AQUA, VIIRS, MERIS and/or OLCI.
Whereas mature atmospheric correction algorithms existed for the S3plus sensors, mature adjacency corrections were not yet in place. Therefore the SIMEC adjacency correction was tested extensively for MERIS (and OLCI) for several coastal and inland waters in Europe. In addition an evaluation of suitable bands was performed for MODIS and VIIRS.

Results: Algorithms for pixel identification
Different algorithms for pixel identification were evaluated and further developed for Landsat-8 and Sentinel-2. The FMASK algorithm was found to be insufficient for water applications, and the AFAR algorithm was developed for cloud and shadow masking. The IDEPIX toolbox was developed for automatically identifying pixel characteristics, and was implemented in the operational processing screen as it was more generic than the AFAR approach, which relied on the L8/OLI thermal bands not present on Sentinel-2.

Results: Algorithms for coastal water products
Specific algorithms have been developed to retrieve in water characteristics from water reflectance. A default algorithm was defined for each parameter as well as regional algorithms for certain parameters and regions.
Some parameters are not available for each sensor (e.g. chlorophyll-a not available for Landsat8 nor for GEO) because they require specific spectrals band not available. All the algorithms have been developed to be valid in coastal waters which are more complex than open ocean waters because of the presence of suspended sediments and colour dissolved organic matter. A regional band switching turbidity algorithm for the entire Gironde estuary was developed, switching from the green band in the clear offshore waters, the red band in the moderately turbid waters, and the NIR band in the most turbid upstream part of the estuary. The retrieval of chlorophyll concentration from Sentinel-2 was demonstrated for the Belgian coastal zone and the Bourgneuf Bay.

Results: Algorithms for multi-sensors and multi-temporal products
Multisensor (MS), multitemporal (MT) and time-series (TS) products were foreseen in the HIGHROC project. Specific algorithms were designed for each products.

1. MT Products:
MT products are 2 dimensional single-sensor image composites. The creation of MT products is rather straightforward: combine all data from a single product and derive certain statistics (e.g. mean, percentiles, standard deviation). L2 products need to be co-registered to a common grid before aggregation.
Each partner generated MT products for their users using tools provided by HIGHROC partners with a configuration specific to the user needs. One exception is the GEO stream, where both the unfiltered and filtered (5-points running average) datasets were provided as a standard product.

2. MS Products:
The merging of MODIS Aqua and SEVIRI data was demonstrated by (Vanhellemont et al., 2014), and a similar approach was used in HIGHROC. This method uses the assumption that any change observed by the high-temporal/low-spatial resolution GEO sensor is due to vertical resuspension of sediments, and that the spatial features observed by the high-spatial/low-temporal resolution GEO sensor are stable in time.
It was not feasible to merge of S2plus and GEO products due to the large resolution differences and the dynamic temporal (tidal) changes including horizontal advection processes. This is a scientific obstacle, not an implementational difficulty – the merged products just do not correspond to reality because parameters are interpolated over time scales containing too much mixed, i.e. unseparable, spatio-temporal variability.

3. TS Products:
Time-series were evaluated for algorithm performance and anomalies were examined. No standard time-series products were provided by the HIGHROC project, but partners provided support for time-series analysis.

*WP5: In situ measurements*
Protocols and intercomparison studies
The objectives of WP5 were to compare the different instruments, calibrations, and protocols used by the partners for the in-situ measurements in order to identify the critical issues and define standard in situ measurements methods and quality controls to get a coherent and uniform dataset to validate HIGHROC satellite products. A HIGHROC protocol deliverable was produced as well as datasets for target parameters.

Intercalibration exercises
To define protocols and intercompare instruments, several dedicated cruises and experiments were organized within the HIGHROC project.
(1) Some partners were represented at a field campaign held on board the RV Belgica in April 2014. The purpose was to intercompare the measurement protocols and to acquire in situ measurements for marine remote sensing reflectance and the corresponding laboratory measurements of optically active compounds. One lesson-learned was that doing proper radiometric intercomparison for a moving ship in waters with string tides is challenging and can easily cause errors.
(2) An intercomparison field campaign was organized by JRC at Venice Tower AquaAlta in the period June 23-27, 2014. The aim was to intercalibrate Trios Systems and different optical instruments and profilers. Among HIGHROC partners, both RBINS and NIVA were represented with their Trios RAMSES systems.
(3) Based on the experience from the Belgica cruise the next intercomparison workshop was performed at NIVA Research Station in Oslo in March 2015. This workshop included also sensors for Inherent Optical Properties. During this workshop, radiometry calibration and intercomparison were performed. Turbidity sensors were compared for different types of waters (river water, high Chl-a water, and sea water added kaolin or formazin), in parallel with filtrations for suspended particulate matter (SPM). SPM filtration protocols as well as Chl-a analysis were also compared for the cruise on the RV Belgica.
The exercise showed that measurements made of turbidity using a hand-held turbidimeter was easy to perform and the results were reliable and comparable between the partners. Significant differences between turbidity sensors were identified under the different water types when compared to the ISO turbidity standard using the handheld turbidimeters. Overall, the workshops have led to a good practices guide, describing the measurements methodology and the recommended quality controls.
(4) An intercomparison exercice was organized by University of Stockholm as part of the Nordic Network collaboration. It took place at field station in Askø 10-14 May 2016. The aim was to intercalibrate radiometers in case 2 water with high CDOM absorption. Among HIGHROC partners NIVA and RBINS were taking part in the Trios RAMSES Systems measurements onboard R/V Oceania. All Trios systems had been calibrated at Tartu Observatory shortly prior to the experiment. Additional reference measurements and sensor controls were carried out at the laboratory facility in Askø.

In situ measurements for HIGHROC datasets
(1) UPMC partner has acquired data in the Rhone river and Gironde estuary (one cruise with RBINS participation). New buoys and upgrades of sensor were made and a large dataset of validation data has been produced and used in the other validation workpackages.
(2) RBINS has maintained two AERONET-OC stations in Belgian waters (Zeebrugge MOW-1 and Thornton/Cpower) and University of HULL has installation an AERONET-OC station at the Blyth wind site. These data are acquired autonomously by a CIMEL-SeaPrism system and transmitted on a daily basis to the centre managed by NASA for near real time processing. An upgrade was made of one AERONET-OC stations and the 3 stations have been successfully operated and produced important validation data. AERONET-OC systems provided a very good dataset of water leaving radiance to validate atmospheric correction algorithm.
(3) CEFAS has maintained three North Sea SmartBuoy moorings system throughout 2014-2016 equipped with turbidity sensor and PAR sensor for calculating Kd. CEFAS surface buoys are smartbuoys deployed for 3 months at a time and serviced in February, May, August and November. The turbidity sensors are used with wipers to reduce interference from biofouling. Measurements are performed by mean burst of 5 or 10 minutes every 15 or 30 minutes.
(4) NIVA has operated Ferrybox-lines along the Norwegian coast throughout 2014-2017 producing validation data of both water reflectance and Chl-a, Turbidity data and CDOM. NIVA’s Ferrybox system is mounted on the Coastal Express sailing along the Norwegian cost between Bergen and Kirkenes and a Ferry between Oslo and Kiel in Germany. A water sampler was activated in selected areas to provide additional data for matchups at fixed stations. The vessel is also equipped with hyperspectral Trios RAMSES Systems for determining water reflectance.
CEFAS has used FerryBox onboard the research vessels in 2014-2016 in addition to sampling on clear sky occasions during research cruises in the North Sea. Several in situ campaigns has also been performed with RV Belgica in the North Sea in 2014-2015, when also other WP5 partners participated, and in 2017.

Description of HIGHROC datasets
All partners collected in situ data up to summer 2017 and ended with the final dataset. All the data were formatted within the standard defined by the consortium and include the following contributions (formatted as Provider Platform Parameters Site):
CEFAS Surface Buoy CHL, Kd, SPM, TUR Dowsing
CEFAS Surface Buoy CHL, Kd, SPM, TUR West Gabbard
CEFAS Surface Buoys CHL, Kd, SPM, TUR West Gabbard 2
CEFAS Surface Buoys CHL, Kd, SPM, TUR Liverpool Bay
CEFAS Surface Buoys CHL, Kd, SPM, TUR Warp
CEFAS Stations CHL, Kd, SPM North Sea and Celtic Seas
LOV Stations Ed, Ld, Lu, TUR, SPM Gironde
LOV Stations Ed, Ld, Lu, TUR, SPM Rhone Plume
NIVA Transect/Ferrybox Ed, Ld, Lu, CHL, CDOM Color Fantasy, Port and Starboard radiometry
NIVA Stations/Ferrybox Ed, Lu, Ld, CDOM, SPM Trollfjord, Herdla, Bow and Stern radiometry
RBINS Stations CHL, RHOW, SECCHI, SPMGFFA, TUR, dRHOW, dSPMGFFA, dTUR Mediterranean
RBINS Stations CHL, RHOW, SECCHI, SPMGFFA, TUR, dRHOW, dSPMGFFA, dTUR Scheldt
RBINS Stations CHL, RHOW, SECCHI, SPMGFFC, SPMGFFA, SPMGFF, TUR, dRHOW, dSPMGFFA, dTUR Southern North Sea
RBINS AERONET-OC platform Lwn Belgian coastal waters: MOW-1
RBINS AERONET-OC platform Lwn Belgian coastal waters: C-POWER
U HULL AERONET-OC platform Lwn North Sea: Blyth

*WP6 Image Processing*
Objective:
The main objective of WP6: image processing was to develop prototype image processing chains for the new S2plus and GEO HIGHROC products and derived multi-sensor and multi-temporal products, integrated with the existing S3plus products.

Results:
Highroc produced three main prototype image processing chains: S2plus, GEO and S3plus and a derived multimission processor. The S2plus processor, which includes Sentinel-2 and Landsat-8 runs at VITO, the GEO processor runs at RBINS and the S3plus processor which includes MODIS, VIIRS and Sentinel-3 runs at BC. For all processors a dedicated workflow was developed including pre-processing, atmospheric correction and the derivation of L2W products. The L2W products feed into the Multimission processor producing multi-temporal and multi-mission synergy products. All processors share a common image and metadata format.
All products produced by the processors were made available to the HIGHROC partners, which added further value and distributed them to their end users as specified in the Service Level Agreements. The processors were operationally in the period of the User Service Trials from October 2016 to October 2017.

S2 plus prototype processor
The S2plus processor ran operationally in Near Real Time (NRT) during the period of the user service trials and produced data for more than 15 sites (72 unique Landsat-8 tiles and 67 unique Sentinel-2 tiles).
Landsat L1 data is downloaded from USGS. The data is then processed in several steps including pre-processing, atmospheric Correction (A/C) and Level 2 Water (L2W) processing. Post processing includes building of the NetCDF file, archiving and distribution via FTP. In total more than 1050 tiles have been processed resulting in a total volume of 1 TB.
Sentinel-2 Level 1C data is downloaded from the ESA hub and further processing is done similarly to the L-8 processing. The L2W processor delivers 2 additional outputs compared to the L-8 processor.
In total more than 8000 tiles have been processed resulting in a total volume of 16 TB.

GEO prototype processor
The operational version of the GEO processor ran in near real time (NRT) during the whole User Service Trial period (Oct. 2016 to Dec. 2017). The NTR GEO processor is composed of the 3 modules to (1) process raw files, extract and save the european waters region, (2) apply the atmospheric correction in 4 local regions (Southern North Sea, Bay of Biscay, Golf of Lion and Skageratt region) and (3) produce image file of the SPM distribution for each product. At the end of each day, daily netcdf are produced. They also contain time-filtered (5 data points) time-series. After that, all GEO products created during the daytime are copied to a ftp server shared with all partners. The GEO processor is coded in IDL (atmospheric correction processor), bash and python. It is run every 15 minutes. During the User Service Trial, 88640 tiles have been processed and HIGHROC GEO products produced during the period Oct. 2016 - Dec. 2017 represent 49 GB of data.

S3plus prototype processor
At BC, the near real time (NRT) processing for the S3plus chains consisted of several steps: setting up a download order for the predefined regions (15), running the preprocessor (Idepix, subsetting) and respective L2 processor(s) with region specific parameterisation(s), applying the HIGHROC water algorithms, and finally uploading the dedicated ftp.
Processing chains for MODIS Aqua, and VIIRS, were finalised and ran by the start of the Service Trials. MODIS Terra processing was implemented during the course of the Service Trials upon request from a few partners. As for OLCI, since the launch was delayed, the availability of standard L2 products was also delayed for the project (July 2017, too late for the Service Trials). It was then decided to implement a processing chain based on the C2RCC processor, which is equivalent to the one used in the ground segment for the alternative atmospheric correction products (chl_nn and tsm_nn). A similar process as with MODIS and VIIRS was as such implemented for NRT OLCI delivery on the ftp

Two kinds of multi-temporal products were offered to HIGHROC users, a temporally aggregated (L3, daily, monthly, seasonally, yearly aggregated) product derived from each sensor of the S3+ stream to observe and analyse trends, or a temporally aggregated GEO product to reduce the inherent noise of the individual 15 minutes products.

The multi-sensor product integrated in the multimission processor is a S3plus/GEO synergy product. The algorithm for the multisensor synergy product was implemented in python for on-demand processing.

*WP7 Product Validation*

Introduction
The quality of the HIGHROC satellite products was assessed based on the field datasets acquired and/or compiled as part of the HIGHROC project, and the corresponding satellite-derived measurements generated from the prototype processors. The different field test sites covered various types of European coastal waters, from clear to highly turbid environments. For each test site, taking into account the regional specifications defined in terms of satellite data processing and match-up protocol, the satellite-derived water reflectance and biogeochemical products were compared to the corresponding quality-checked products measured in the field.
Based on the scatterplots produced, the differences between the satellite-derived and field measured values were used to express statistically the confidence (uncertainty) level in the satellite products taking into account the uncertainty associated to the field data. The results obtained were analysed to give recommendations on where algorithm improvements are required or validation techniques need adjustment.
Satellite data and products included:
• high spatial resolution data (called S2plus) provided by the S2-MSI and L8-OLI sensors
• medium spatial satellite data (called S3plus) provided by the MODIS and VIIRS sensors
• high temporal resolution data (called GEO) provided by the MSG-SEVIRI sensor.
The different satellite data and products considered are illustrated in Figure 1.3.4.A. S2plus satellite data have a spatial resolution of about 30 m and a temporal revisit of about 5 days. S3plus satellite data have a temporal resolution of about 300 m and a temporal revisit of 1 day. GEO satellite data have a much coarser spatial resolution (about 4 km) but a high temporal revisit (15 min).

The first step in the validation exercise was the validation of the atmospheric corrections applied to the satellite data to retrieve the water reflectance, here the remote sensing reflectance (Rrs) or normalized water-leaving radiance (Lwn). The second step in the validation exercise was the validation of the derived water biogeochemical products, typically the water turbidity, concentrations of suspended particulate matter (SPM) and chlorophyll-a (Chla).

Field datasets
Only a part of match-ups between field and satellite data were obtained using field measurements carried out during regular field (shipborne) campaigns. Most of the match-ups were obtained using field data recorded by autonomous sensors onboard moving such as ferries or fixed platforms such as Smartbuoys , Aeronet-OC stations and the instrumented Mesurho platform .

Results
The first part of the exercise provided satisfactory results in terms of validation of the atmospheric corrections applied to satellite data. As could be expected, applying appropriate atmospheric corrections (here the MUMM algorithm) to S3plus data recorded over moderately turbid waters resulted in quite limited differences (<25%) between field and satellite data, at least in the green and red spectral regions (not shown). The atmospheric corrections applied to S2plus satellite data recorded over moderately to highly turbid waters could also be validated, which represents a step forward in the remote sensing of coastal and estuarine waters. GEO satellite products were also validated based on direct comparisons with simultaneous S3plus satellite products.

The next step was to analyse match-ups between field data and satellite biogeochemical products. Both the water turbidity and SPM concentration are functions, as a first approximation, of the water reflectance in the red spectral region (or the near-infrared in the case of highly turbid waters). It was not surprising to obtain good results in terms of turbidity, SPM concentration and diffuse attenuation coefficient (Kd) remote sensing retrievals. Over a 13 year period, numerous match-ups were obtained between S3plus and Smartbuoy data, resulting in a robust validation of these key products. The good agreement between field and satellite observations support the future monitoring of seasonal to multi-year dynamics of SPM concentration in coastal waters.

It is a priori more difficult to remotely sense Chla concentrations in turbid coastal waters where non-algal particles and dissolved substances contribute significantly to both light absorption and backscattering. The contribution of Chla to the water reflectance is more difficult to detect and the blue-to-green Chla algorithms usually fail. This failure has been often observed based on HIGHROC datasets. Another issue to deal with when using in situ Chla fluorescence measurements to validate satellite products is the quenching effect on phytoplankton cells around mid-day. Despite these issues, a quite satisfactory agreement was observed between field-measured and satellite-derived observations of the seasonal dynamics of Chla. Best results were usually obtained considering weekly-averaged field and satellite observations in test sites with low turbidity.

Conclusions and perspectives
This multi-sensor and multi-site match-up exercise conducted for S2plus, S3plus and GEO satellite products provided many satisfactory results and also highlights required improvements in the development and validation of future algorithms.
Results obtained in terms of atmospheric corrections of satellite data demonstrate that algorithms specifically designed for turbid coastal waters are now quite mature and operational in the visible and NIR parts of the spectrum. SPM concentration, turbidity and Kd algorithms also proved to operational but are still regional or at least require prior knowledge of the turbidity water range.
At the end of this validation exercise, most of the OLI, MODIS and VIIRS satellite products can be considered as ‘validated’ thus operational for delivery to users. More quality match-ups and maybe even algorithm development are required to assess the uncertainties associated to MSI and especially OLCI satellite products in the near future.
In all sites, improvements are required to minimize the uncertainty on the satellite-derived Chla concentration. The validation of future Chla products in turbid coastal waters require: (i) an efficient correction or filtering for Chla-fluorescence measurements affected by quenching effects; (ii) more complex Chla algorithms able to first remove the optical contributions of NAP and CDOM from reflectance spectra to accurately retrieve the Chla spectral signature. Neural network and/or multi-step regional algorithms could represent solutions.

Potential Impact:
*Potential impact on marine user community*

HIGHROC project and activities had an important impact to the marine user community by providing quality control, high resolution, maps of seawater parameters such as suspended particulate matter (SPM), Chl-a, etc. During the one year period of the User Service Trials, the full suite of HIGHROC products was supplied to key users in order to assess the utility of HIGHROC products and service. Feedback was collected from those users.
The key-users involved in the HIGHROC user trial are listed below.
• Environment Agency (EA) in the UK is a non-departmental public body, sponsored by United Kingdom government's Department for Environment, Food and Rural Affairs (DEFRA), with responsibilities relating to the protection and enhancement of the environment in England.
• SUMO is a research group of the RBINS (OD Nature) institute. Its aim is to better understand sediment transport and dynamics in Belgian waters.
• SHOM (Service hydrographique et océanographique de la marine) is in Brest, France. It is the national hydrographic service. Its objectives are to support the Navy on environmental information and to support public policies.
• The Scientific Service Management Unit of the Mathematical Model of the North Sea (MUMM, RBINS-OD Nature) has the objective to improve knowledge of the North Sea and provide scientific marine services. It is in charge of reporting about coastal water quality for the WFD and MSFD
• International Marine and Dredging Consultants (IMDC) is a Belgian engineering and consultancy company specialised in a vast range of water related projects. The main topics of their studies are: dredging, offshore energy, flood Risk, waterways, coasts & estuaries and ports & offshore structures. The main clients are governmental agencies, dredging companies, marine industries and LNG companies.
• BSH, the German Federal Hydrographic Agency, is obliged to assess the state of the North Sea and Baltic Sea at every time, and to give reports about the state for defined observation periods.
• Brockmann Geomatics (BG), Sweden, is a service provider itself and serves several Swedish Authorities responsible for water management and water quality, such as SwAM (Swedish Agency for Marine and Water Management), the Swedish Water Authorities and the regional County Administrative Boards.
• The Norwegian Environment Agency (NEA) is a government agency under the Ministry of Climate and Environment working for a clean and diverse environment. Their primary tasks are to reduce greenhouse gas emissions, manage Norwegian nature, and prevent pollution with principal functions including collating and communicating environmental information, exercising regulatory authority, supervising and guiding regional and local government level, giving professional and technical advice, and participating in international environmental activities.
• i-Sea, in France, is a private company created in 2004 from the GEO-Transfert Institute, a technology transfer cell of the Université de Bordeaux 1 and ADERA. The RIVERCOLOR project is considered as an (intermediate) user of the HIGHROC project.
• Pierre Gernez (University of Nantes, France) oversees the remote sensing of environmental parameters in the Bourgneuf Bay for the GIGASSAT project.
User expectations and needs were very different in the HIGHROC service trial. Some of the users were only interested in a final product (e.g. averages over time of SPM, chlorophyll concentrations, or Kd), while some wanted to use intermediate products issued from the HIGHROC processing chain. Nevertheless, the general feedback of the service trial was very positive (80 % satisfaction on average) despite the diversity of the products requested and the expertise of the provider. In general, all the users acknowledge the scientific quality of the consortium, the communication and the reactivity to problems in the partnership. The service trial presented the highest score for communication (83.3%) followed by the benefits (82%) and the performance (77%). Users recognised the benefits of remote sensing information to complete scarce in situ observations. They are happy to continue to discuss with the consortium and they recognized its scientific expertise.

*Potential impact and dissemination of results in scientific community*

HIGHROC project had an important impact in the scientific community with the publication of 11 papers in peer-review journal. HIGHROC results have also been presented in numerous remote sensing oriented scientific meetings, including strong participation in major events such as IOCS conferences in 2015 and 2017, the Ocean Optics conference in 2016, the Sentinel 3 for Science symposium in 2015, etc. The HIGHROC consortium was also very active in the international working groups such as Sentinel-2 Validation Team (S2VT) and Sentinel 3 Validation Team (S3VT) with presentations from different partners during S2VT and S3VT meetings.

The HIGHROC Science conference, open to all experts from the scientific community, was organised in Brussels from 7 to 9 November 2017. It allowed presentation of the HIGHROC results to the scientific community but more importantly opened up the discussion with the international community about new challenges for coastal ocean color remote sensing. The scientific committee was composed half of HIGHROC partners and half of external experts in coastal ocean colour.
The scientific program was divided into 4 main sessions:
• In situ measurements and validation
• Coastal water applications
• Satellite data processing
• New satellite sensors and algorithms
Regarding participation, over the 68 confirmed registrations (80 participants were expected at maximum), 62 participants attended the conference. Participants came from 17 different countries (14 Europeans countries + South Korea, Canada and Argentina). 47 scientific abstracts were submitted to the conference (25 oral presentations and 22 poster presentations). Most of the scientific contributions were of very high quality, stimulating a state-of-art discussion during the last meeting session. The conference was also the opportunity to train and advertise to the scientific community the software developed during the HIGHROC project such as ACOLITE and iCOR which are both dedicated to atmospheric correction of S2 and L8 images over inland and coastal waters.

*Dissemination of HIGHROC activities to users, public and relevant organisations*

A website has been created and updated throughout the project to describe HIGHROC activities (www.highroc.eu). It describes the project and its general objective, shows some user stories, presents the HIGHROC products and lists all the HIGHROC publications.
Two brochures have been produced to describe the HIGHROC project, one at the beginning emphasis the objective and one at mid-project with some results (http://www.highroc.eu/assets/pdf/highroc_brochure.pdf , http://www.highroc.eu/assets/pdf/highroc_brochure_2.pdf ). These brochures has been distributed to potential users and are available online.
HIGHROC partners also attended conferences to meet and present HIGHROC products to external communities which might be interested by products as potential users (e.g. dredging community during the CEDA dredging days 5-6 Nov 2015; FerryBox community during a Workshop on MS Colro Fantasy on October 2017).
Two user workshops were organised during the project. The first HIGHROC User Workshop “Earth Observation methods for the coastal zone: progress and opportunities” was held in London on 8-9 June 2015. The aim of the workshop was to bring together as wide a range as possible of potential end-users for the project, including marine policy makers, marine environmental agencies and the marine/maritime industrial sector.
The workshop was organised in four sessions to answer:
(1) What is the state-of-play for the measurement of in situ data for remote sensing validation purposes?
(2) What are the current applications of Earth Observation (EO) data to solve marine policy questions?
(3) What will HIGHROC bring to the field?
(4) How will we make the most of the new Sentinel data?
With more than 30 participants from five different countries representing the core user base of HIGHROC, and 18 presentations over the course of 1 ½ days, the workshop was successful in engaging HIGHROC with a strong user base in the UK, as well as showcasing (potential) HIGHROC user applications in e.g. France and Belgium.
The second HIGHROC User Workshop was held in Hamburg on 9-10 October 2017. The aim of the workshop was to give the floor to the HIGHROC core users after the one-year Service Trials period to report about the usage and integration of EO and HIGHROC products into their daily work. The workshop was successful in gathering over the course of 1 ½ days 22 participants from six different countries, to hear presentations and discuss coastal applications and how HIGHROC products were used. The discussions focussed on the usability of the HIGHROC products in pre-existing working schemes for each user and the combination of different data sources, among others HIGHROC products and other EO based results.
HIGHROC was also presented to external organisations such as the CMEMS community with several presentations during the project period, (e.g. Copernicus workshop for coastal zone monitoring and management, Brussels, 29/06/2017, Copernicus Marine week, Brussels, 28/09/2017), and the space agencies (e.g. ESA, EUMETSAT and active participation in S3VT and S3VT groups).

*Exploitation of the results*

HIGHROC atmospheric correction algorithms developed by partners RBINS and VITO were integrated into publicly available software, ACOLITE and iCOR respectively, to be used by a larger community. In addition, specific training sessions were organized during the HIGHROC Science Conference.

HIGHROC processing chains will be use in the DSC4COP project whose aim is to propose integrated oceanographic data layers to Intermediate Business Users in a pre-commercial licensing framework. Each HIGHROC product will consist of a specific data layer. Hence, HIGHROC results will be used in an operational mode in the future.

List of Websites:
www.highroc.eu
Contact via form at http://www.highroc.eu/contact

Reported by

INSTITUT ROYAL DES SCIENCES NATURELLES DE BELGIQUE
Belgium

Subjects

Other Technology
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top