Coastal Zones 3

The first satellite of the Sentinel-2 constellation, Sentinel-2A (S2A) was successfully launched on the 23rd of June 2015. It has on board the MultiSpectral Imager (MSI), an optical imager with 13 spectral bands spanning from the blue to the shortwave infrared (SWIR) with 10, 20, or 60 m ground resolution. As previously demonstrated with Landsat-8 (L8) imagery at 30 m resolution, many human activities in coastal waters such as offshore construction and dredging, and their impacts, are spatially resolved at these resolutions. Thanks to the free dissemination and increasing quality (in terms of signal-to-noise ratio, SNR) of these data, we are entering a new era of high resolution water colour remote sensing.

Here we present the application of MSI imagery for coastal waters, and the processing and atmospheric correction developed in the EC-FP7 HIGHROC project. The MSI has a pair of 20 m SWIR bands at 1.6 and 2.2 µm, allowing for a robust image-based atmospheric correction, even over extremely turbid waters. One of the main advantages of S2 over L8 is the inclusion of a red-edge band at around 705 nm, allowing for the determination of chlorophyll a concentration in turbid and productive waters, where open ocean blue-green ratio algorithms fail. MSI has red (665 nm) and NIR (842 nm) bands at 10 m spatial resolution, allowing for the retrieval of turbidity or suspended sediment concentration, even in narrow inlets and ports. This makes it an invaluable dataset for validating sediment transport models that are needed for optimization of dredging operations and coastal defense around turbid water ports.

The Operational Land Imager (OLI) on board of L8 has the advantage of higher SNR in its 30 m bands, with potential for sharpening using the 15 m panchromatic band. L8 also has on board the two-band Thermal Infrared Sensor (TIRS) that allows for lake and sea surface temperature retrieval, as well as the estimation of cloud height. Due to the lower revisit time of the S2 (10 days with one unit, 5 days with two) and L8 missions (16 days), and the advantages of both, they have to be seen as a virtual constellation for observing the rapidly varying coastal systems.

The quality of coastal and inland waters is increasingly threatened by population growth, urbanization, and climate change. In particular, the interfacial nature of the urban coastal zone, bridging aquatic, terrestrial, atmospheric and anthropogenic domains, subjects itself to significant impact from dynamic and complex processes. Timely, accurate, consistent, and scientificly based assessments, monitoring, and forecasting of coastal water quality are crucial across global, regional, and local scales. European Sentinel and other multi-sensor satellite data represent invaluable contributions in this regard. For example, satellite-based observations from ocean colour radiometry (MERIS, VIIRS, OLCI et al.), synthetic aperture radar (SAR: Sentinel-1, ENVISAT, RadarSat, ERS et al.), and high-resolution multi-spectral imagers (Sentinel-2, Landsat-8), coupled with in situ data and data-assimilation approaches, provide an essential synoptic perspective at increasingly relevant spatio-temporal scales. Here we describe experimental and operational data products generated from these sensors/platforms (chlorophyll-a, total suspended matter, optical properties, oil/biogenic slicks, platform/ship detections et al.), and demonstrate utilization of these products for diverse urban water quality applications and services. Specific phenomena and features addressed in this context include river discharge, sediment and runoff plumes; algal blooms (harmful and nuisance); oil spills; and, platform/ship detection and discharge events. The pathway from multi-sensor data to products, and then to integrated information in support of end-user needs (e.g., water quality management, policy, and remediation efforts) is also discussed. This work is an advancement toward the objectives of the International Water Quality Summit held in 2015 by the Group for Earth Observations (GEO), particularly toward development of urban water quality monitoring and forecasting services in developed and developing nations.

Improved data access along with advances in sensor and computing technology have increased quantitative use of satellite remote sensing data for the synoptic assessment of the extent of coastal resources. However, satellite monitoring of coastal regions is often hampered by the vast range in the amount and type of water column constituents. Additionally, there are many environmental variables that impact on the quality of the satellite imagery, such as atmospheric and water-surface conditions, and variations in terrestrial catchment inputs. Without sophisticated and objective methodologies, the imagery can be difficult to use for reliable trend assessments or factor analysis.

A physics-based inversion model was developed for hyperspectral data to retrieve bathymetry, substrate composition and water quality information [(concentrations of chlorophyll-a (CHL), non-algal particulates (NAP) and colored dissolved organic matter (CDOM)]. Based on optical modeling, this approach requires an understanding of the interactions between light and the atmosphere, the water surface, water column constituents and the substratum, if optically shallow.  Retrieving information from satellite data on water quality and benthic substrata from (often multispectral) satellite data through the inversion of constituents and substratum is constrained by the spectral and spatial characteristics of the satellite imagery.

Multi-temporal Landsat data acquired from the Australian Geoscience Data Cube (AGDC) when combined with the optical modeling approach, enabled a consistent analysis of coastal water quality through the time-series. From this data, water quality variables such as total suspended matter and the light environment of the water column (vertical attenuation of light) were retrieved.

Westernport experienced extensive loss of seagrass coverage between 1970-2000 and it is well-known that seagrass can be affected by coastal water quality due to their sensitivity to light availability and sedimentation. The extent to which seagrass loss or the water quality of the bay is impacted from the catchment or riverine sediment loads is unclear, as other physical processes, such as tides or wind-driven benthic sediment re-suspension may be completely or partially responsible. The temporal assessment of the AGDC derived water quality information was compared with sediment load monitoring in the rivers draining into Westernport. These analyses are used in support of the Westernport hydrodynamic modelling and seagrass modelling activities.

The Sentinel-2 MSI sensor was primarily designed for land applications. However, our results show that the Sentinel-2 data is very useful in aquatic applications, at least in the cases where the water leaving signal is relatively strong. Examples of such applications are mapping of shallow water benthic habitat and monitoring of algal blooms.

We evaluated the performance of Sentinel-2 10 m imagery in mapping benthic habitat by comparing Sentinel-2 data from August 2015 with HySpex airborne spectrometer data from the same area in Western Estonian Archipelago collected a year earlier. The preliminary results show that water reflectance spectra of Sentinel-2 match with the HySpex results. Multispectral data of Sentinel-2 does not allow to map as many benthic habitat types as it can be done with hyperspectral data. However, it should be possible to map broad benthic algae groups (red, green and brown algae), especially when in situ data is available from the study site and supervised classification methods can be used. Mapping the extent of benthic habitat, that is also an important parameter in benthic ecology, is straightforward in shallow waters although defining the border between optically deep water and dark vegetation becomes difficult in 5-6 m deep water (in the Baltic Sea conditions). On the other hand narrow underwater sand dunes with just 10-20 cm depth variance are clearly seen in Sentinel-2 imagery in shallow sandy areas. The red edge bands of Sentinel-2 are potentially very useful in benthic habitat mapping. However, in some cases (highly variable bottom) using the 20 m spatial resolution data may cause problems due to mixed pixels.

10 m spatial resolution of Sentinel-2 allows to study cyanobacterial blooms with great detail. In many cases the Sentinel-2 reflectance resembles terrestrial vegetation meaning that the floating scum stripes are wide enough for the 10 m spatial resolution. In some areas the scum stripe spectra are similar to dense underwater bloom meaning that even the 10 m spatial resolution is too coarse compared to the actual width of the scum stripes. The red edge bands are very useful in bloom remote sensing as the height of the reflectance peak near 710 nm is one of the best indicators of phytoplankton biomass in turbid waters and is often utilised in chlorophyll retrieval algorithms. Therefore, using of the 20 m resolution data is preferable in quantitative mapping of phytoplankton biomass unless the spectral heterogeneity of the bloom is extremely high (e.g. narrow stripes of floating scum).

The Case 2 Regional Processor for MERIS, freely available in ESA’s BEAM Toolbox (Doerffer and Schiller 2007 & 2008), has been proven to address well many kind of optically complex waters (Palmer et al 2015, Attila et al 2013, Vilas et at 2011, Odermatt 2010). In the framework of ESA DUE CoastColour project it has been further developed to cope with very high scattering coastal waters (Doerffer & Brockmann 2014, Sá et al 2015). In this so called “C2RCC” evolution of the algorithm the bio-optical model has been revised and comprises 5 optical components and thus includes a flexibility model for the specific optical properties. The atmospheric model has been changed and it based on a parametrised version of the SOS code (Lenoble et al, 2007). A special coastal aerosol model derived from coastal Aeronet stations (Aznay & Santer, 2009) has been introduced.

Further evolution of the C2RCC processor concerns the internal quality checking through an auto-associate neural network using the top-of-atmosphere spectra, as well as dedicated error networks which have been derived from comparison with a large reference dataset based on the NOMAD database. As a result the processor provides both, per pixel uncertainties and quality flags.

This version of the algorithm is also used in the Sentinel 3 OLCI ground segment processor and will be applied for the MERIS 4^th Reprocessing, planned for early 2016. Recently it has demonstrated its good performance in the Ocean Colour CCI Round Robin.

Beside MERIS and OLCI, the C2RCC processor is also able to process the US ocean colour sensors MODIS, SeaWiFS and VIIRS. Validation for these processor is currently undertaken and will be presented at the conference.

A further development of the C2RCC algorithm is the application to spatial high resolution sensors by adapting it to the spectral characteristics of Landsat 8 and Sentinel 2. In the framework of the EU FP7 GLaSS project good validation results were obtained when applied to Landsat 8 data over inland waters. At the time of this abstract the forward calculations for Sentinel 2 have been completed, and the adaptation to Sentinel 2 will be completed in Q1 2016. First validation results with Sentinel 2 will be available for the Living Planet Symposium.

By construction the C2RCC processor works over those optical water types and atmospheres, which are covered by the forward modelling. However, the most challenging waters are still the extreme scattering (i.e. TSM > 100mg/l) and extreme absorbing (adg > 1/m) waters. These waters are in the focus of the ESA SEOM Case2 Extreme Water project, and the investigations include the atmospheric correction as well as the in-water retrieval. The C2RCC processor will be tested together with other approaches. Results will be presented as part of this presentation.

The C2RCC processor will be distributed free and open as plug-in of ESA’s Sentinel Application Platform SNAP Toolbox and will thus be available to a large user community.

References:

Attila Jenni, Sampsa Koponen, Kari Kallio, Antti Lindfors, Seppo Kaitala, Pasi Ylöstalo (2013): MERIS Case II water processor comparison on coastal sites of the northern Baltic Sea, Remote Sensing of Environment, Volume 128, 21 January 2013, Pages 138-149, ISSN 0034-4257, http://dx.doi.org/10.1016/j.rse.2012.07.009.

Doerffer R, & Carsten Brockmann (2014): DUE CoastColour Consensus Case 2 Regional Algorithm Protocols. http://coastcolour.org/documents/DEL-26_Consensus_Case_2_Regional_Algorithm_Protocols_v1.0.pdf

Doerffer, R., & Schiller, H. (2007). The MERIS case 2 water algorithm. International Journal of Remote Sensing,28(3), 517−535.

Doerffer R. and H. Schiller (2008): MERIS Regional Coastal and Lake Case 2 Water Project - Atmospheric Correction ATBD, GKSS Research Center 21502 Geesthacht Version 1.0 18. May 2008

Lenoble J., M. Herman, J.L. Deuze, B. Lafrance, R. Santer, D. Tanre ,. (2007). A successive order of scattering code for solving the vector equation of transfer in the earths atmosphere with aerosols Journal of Quantitative Spectroscopy & Radiative Transfer 107 (2007) 479507

Odermatt Daniel, Claudia Giardino, Thomas Heege (2010): Chlorophyll retrieval with MERIS Case-2-Regional in perialpine lakes, Remote Sensing of Environment, Volume 114, Issue 3, 15 March 2010, Pages 607-617, ISSN 0034-4257, http://dx.doi.org/10.1016/j.rse.2009.10.016.

Palmer Stephanie, Peter D. Hunter, Thomas Lankester, Steven Hubbard, Evangelos Spyrakos, Andrew N. Tyler, Mátyás Présing, Hajnalka Horváth, Alistair Lamb, Heiko Balzter, Viktor R. Tóth (2015): Validation of Envisat MERIS algorithms for chlorophyll retrieval in a large, turbid and optically-complex shallow lake, Remote Sensing of Environment, Volume 157, February 2015, Pages 158-169, ISSN 0034-4257, http://dx.doi.org/10.1016/j.rse.2014.07.024.

Sá C., D. D'Alimonte, A.C. Brito, T. Kajiyama, C.R. Mendes, J. Vitorino, P.B. Oliveira, J.C.B. da Silva, V. Brotas (2015): Validation of standard and alternative satellite ocean-color chlorophyll products off Western Iberia, Remote Sensing of Environment, Volume 168, October 2015, Pages 403-419, ISSN 0034-4257, http://dx.doi.org/10.1016/j.rse.2015.07.018.

Vilas Luis Gonzáles, Evangelos Spyrakos, Jesus M. Torres Palenzuela (2011): Neural network estimation of chlorophyll a from MERIS full resolution data for the coastal waters of Galician rias (NW Spain), Remote Sensing of Environment, Volume 115, Issue 2, 15 February 2011, Pages 524-535, ISSN 0034-4257, http://dx.doi.org/10.1016/j.rse.2010.09.021.

Zagolski R., Santer R., Aznay O., (2007). A new climatology for atmospheric correction based on the aerosol inherent optical properties. Journal of Geophysical Research, Vol. 112, No. D14, D14208