A project initiated and funded through the SEOM Programme of ESA – the European Space Agency.
Ocean Colour remote sensing has significantly improved over the last 15 years, but the emphasis was still mainly on chlorophyll-a retrieval in Case 1 waters. In recent years progress has also been made to derive water quality parameters in Case 2 waters. This is much more difficult due to the optically complex composition of Case 2 waters, the large variability of the specific optical properties and the large value range of concentrations of water constituents, such as chlorophyll-a, total suspended matter or CDOM absorption. This high variability in spectral shape (SIOPs) and value ranges (concentrations) lead to a multitude of possible spectra from Case 2 waters, which cause the big problems in their inversion. This has been recently addressed by various projects (e.g. CoastColour, SeaSWIR, Diversity II, ESA MERIS Lakes project and others), without however being fully satisfactory, and while still remaining within a rather narrow range of concentrations and variability.
It is the objective of this project to study and advance the state of the art of retrieval of water reflectances, IOPs and concentrations from ocean colour data of extreme absorbing and extreme scattering Case 2 waters, demonstrating this with MERIS historical data and preparing algorithms and processors for the near future Sentinel 3 OLCI and SLSTR sensors. The work shall be done in constant dialogue with the user community and the products and software processors shall be made free and publicly available for rapid uptake by the community.
Challenges of extreme waters and C2X project approach
There are a number of challenges that need to be addressed by the C2X team:
The extreme case 2 waters with very high TSM loads (>100 mg/l) are termed here C2SX (Case 2 Scattering Extreme) waters. There is a significant user interest in such waters for management of coastal and estuarine sediment transport and associated dredging/dumping operations, for ecosystem modellers considering primary production and associated eutrophication and/or carbon fixation and for quantifying supply of organic carbon from the world’s major rivers (Amazon, Yangtze, La Plata, etc) to the coastal oceans.
The extreme case 2 waters with high CDOM absorption coefficient, aCDOM(440nm)>1/m are termed here C2AX. In Europe, the Baltic Sea is a specific and large example case of C2AX waters with intense user interest, but also the Black Sea and boreal inland water are characterised by very high CDOM absorption. The user interest in C2AX is very similar to that of any coastal waters and include: water quality monitoring, harmful algae bloom detection (cyanobacteria in the Baltic) and the needs of ecosystem modellers for primary production and carbon cycle studies. Under low CDOM conditions, the reflectance has a distinct spectral shape in the visible range. Addition of CDOM lowers the broadband magnitude to a (very) low level, from which it is very difficult to derive further information on water constituents, e.g. Chlorophyll a, out of the signal, quite apart from issues with the atmospheric correction in these cases.
Atmospheric correction of ocean colour data is probably the most critical problem for exploitation of ocean colour data because most of the light reaching a satellite sensor comes from photons scattering in the atmosphere that contain no information on water constituents. The most critical component, particularly for turbid C2X waters, is generally the aerosol correction because it cannot be calculated a priori from geometry and/or meteorological data, but must be determined by inversion of imagery. In extremely turbid C2SX waters, the marine reflectance may be non-zero for the shorter SWIR wavelengths and the NIR marine reflectance is no longer a linear function of TSM. Two general classes of aerosol correction can be distinguished: Firstly, the NIR/SWIR extrapolative approaches and secondly, the multispectral coupled ocean-atmosphere model full spectrum inversion.
A main challenge for retrieval of water properties lies in the chlorophyll retrieval under very high absorbing conditions. Any method to estimate chlorophyll a from marine reflectance assumes that the absorption of phytoplankton has an impact on reflectance that is both measurable and distinguishable from the absorption of non-algae particles and CDOM. The problem in C2X waters is that absorption by non-algae particles (C2SX) or CDOM (C2AX) in the blue-green spectral range, normally used for chlorophyll a retrieval, is much greater than absorption by algal particles. The effect of the latter cannot be distinguished. This “masking effect” gives an effective detection limit for chlorophyll a retrieval in C2X waters.
Our approach to address these challenges builds on the following corner points:
- Comprehensive review of state of the art including experimental research. Combining our own expertise with that of the science support team and a literature study we will identify and classify potential algorithms for atmospheric correction and in-water retrieval. We are open to any potential solution and will assess them by potential performance, complexity etc. This will be supported by testing on breadboard level. The use of the OLCI 1020nm band should improve atmospheric correction in C2SX waters, and we will perform an analysis of the expected performance of this band (signal:noise, out of band response, asymmetric spectral response function, etc.). Similarly the OLCI 400nm band should improve the atmospheric correction in C2AX waters, and the use of the SLSTR SWIR bands should improve the atmospheric correction in C2SX waters, drawing on recent experience from the similar Landsat-8 SWIR bands. Finally, the Requirements Baseline Document and the ATBD v1 will be iterated with the science support team before delivery to ESA.
- Algorithm intercomparison and improvement. We will use processors for atmospheric correction (polymer, neural network) and in-water retrieval (ratio algorithms, coastcolour) which are available within the team, extend them to include the new S3 bands (OLCI + SLSTR), and implement as S3 Toolbox prototypes new algorithms following the scientific analysis (SWIR AC, multi-band in-water algorithms). We always intercompare our results with the standard processing from ODESA. From the start we will implement algorithms as processors for both MERIS and for OLCI&SLSTR. We will study their performance on a set of test sites covering extreme turbid and extreme absorbing water bodies (coastal, estuarine and inland waters) with MERIS band set. Test sites include amongst other the Belgian coastal waters (new AERONET-OC site, operated by RBINS), Schelde, Gironde, La Plata, Yellow and Jangste river estuaries for C2SX waters, and the Baltic Sea, Finish Lakes and the Lena Delta (Arctic) as C2AX waters. We will use Hydrolight and SMART-G (Monte Carlo) radiative transfer models to generate simulated test data for sensitivity studies and for systematic test of the MERIS and OLCI&SLSTR band sets processors. The SMART-G code allows to simulate adjacency effect, and of course to include absorbing aerosols in the atmospheric simulations. This will also give us indication of algorithm performance, and accuracy depending on combination of IOPs. We will complement this by comparison with in-situ, where available, and by applying a spectral decoupling analysis. We will also address the problem of cloud screening in C2SX and C2AX waters as an important aspect of ocean colour retrieval in extreme waters. Finally we will address in the ATBD a per pixel uncertainty, knowing that this is still a matter of research. We will also discuss vicarious calibration and study the performance of vicarious calibration coefficients over the extreme water bodies.
- As described above we will address both, MERIS as well as OLCI&SLSTR in parallel from the start of the study. We will continuously collect findings relevant for the Sentinel 3 sensors in order to draw conclusions, identify gaps in understanding, and formulate recommendation for additional studies. For example, we will use the results of the analysis of vicarious calibration on MERIS w.r.t. to the extreme waters, to give recommendations for vicarious calibration of Sentinel 3. It is expected that this task will be carried out when first Sentinel 3 data are available so that the work is complemented with results from running the processors on real data. Results will be presented to and discussed with the ocean colour community during a SEOM Ocean Colour Case 2 Extreme Waters Workshop in autumn 2016. This will be held in the tradition of the successful CoastColour User Consultation Meetings.