1. Tailored Algorithms for Ocean Colour applications at regional scales: geographic and optical approaches Mark Dowell mark.dowell@jrc.it & Stewart Bernard sbernard@csir.co.za

2. Rationale There is necessity to describe a considerable amount of variability in Inherent Optical Property (IOP) subcomponent models. This is particularly true, if inversion algorithms are to be applicable at global scale yet remain quantitatively accurate in coastal & shelf seas. This is unlikely to be achieved in the foreseeable future, with a single representation of IOP subcomponents. –BEAM – Case2R The proposed approach is an algorithm framework more than a specific algorithm.

3. Defintitons Geographic Class-based

4. Regional vs. Class-Based make table! Regional Advantages –Explicitly link to locally measured in-situ data –May be “simpler” –Accounts for physiological differences Disadvantages –Explicitly link to locally measured in-situ data – not generalized –May result in regional discontinuities Class-based Advantages –Generic, “global”, can be generalized –Can be used as a tool to identify “black holes” –Seamless transitions –Continuous improvements through additional on in-situ data Disadvantages –More complicated to implement –Computational more expensive – not much!

5. Stewart’s regional algorithm Background

6. Figure 7. Satellite-derived images of 1-km resolution from the Medium Resolution Imaging Spectrometer (MERIS) sensor, detailing the development of the Gonyaulax polygramma bloom in the False Bay area in February–March 2007 and its transport along the shelf edge of the West Coast. The simultaneous development of a bloom dominated by the toxic Alexandrium catenella in the St. Helena Bay region is also evident. Chlorophyll a (Chl) products were calculated using merged data from the standard MERIS Algal 1 algorithm for Chl 25 mg m-3. No flags were applied to the data so as to allow bloom transport to be observed in images where absolute Chl is relatively less important (except where high sun glint caused the failure of the Algal 1 product). [Pitcher et al 2008] The Southern Benguela A highly productive upwelling system, frequently suffering from harmful algal blooms, with considerable atmospheric variability. Both empirical and analytical algorithms have been developed for these high biomass waters, with buoy based validation of water leaving radiances. The next few years should see more buoy – and AERONET based validation of ocean and atmospheric algorithms

7. MERIS Sample Products Mozambique Channel 27 th October 2008 Algal 1 Chlorophyll a Case 1 algorithm Algal 2 Chlorophyll a Case 2 algorithm Yellow Substance Case 2 algorithm Total Suspended Matter Case 2 algorithm South African East Coast: Natal and Delagoa Bights, Sofala Banks Dynamic coastal systems subject to extensive riverine influence and the oceanic influence of the Agulhas current; these are extremely good subjects regional algorithms dealing with high Case 2 variability. The next few years should see extensive radiometric and geophysical validation efforts, focusing initially on the Natal Bight.

8. Stewart’s regional algorithm Example Products

9. What to parameterize? Variance and Co-variance of Optically Active Constituents Parameterising IOP subcomponent models (or fit coefficients – for empirical algorithms) Different OWT different inversions method Avenue to spatial uncertainty estimates Regional value-added products

10. 0102030 Forest Wetland Water Reflectance Band 1 Reflectance Band 2 Mean class vector Unknown measurement vector Traditional minimum-distance criteria Hard 0102030 Forest Wetland Water Reflectance Band 1 Reflectance Band 2 Fuzzy graded membership Water = 0.05 Wetland = 0.65 Forest = 0.30 Fuzzy The approach undertaken adopts fuzzy logic to define and identify, in radiance space, distinct bio-optical provinces that implicitly reduce the variance in the IOP subcomponent models.

11. In-situ Database Rrs( ) c Cluster analysis c Sgd, aph*,……. c Station data sorted by class c Class based relationships 8 classes Class Mi,  i Satellite Measurements c Individual class derived products Merged Product c c c Calculate membership Rrs( )

12. Advantages of fuzzy logic defined provinces They allow for dynamics both seasonal and inter-annual in the optical properties of a given region. They address the issue of transitions at the boundaries of provinces (through the fuzzy membership function of each class) thus resulting finally in the seamless reconstruction of a single geophysical product.

13. 8 objectively identified classes in radiance space

14. Class 1Class 2Class 3Class 4Class 5Class 6Class 7Class 8 May 2004 MERIS Global Composite

15. Benguela example I’m using the 2006 MODIS image ( A2006082123500.L2_LAC) it would be good to have your regional algorithm for the same product and a RGB image

16. Benguela example Mark’s image of classified optical water types+ membership maps

17. MERIS October 8 th 2008

19. Relation to current understanding turbid water flag After Morel and Bélanger 2006

20. Relation to current understanding turbid water flag Class 1Class 2Class 3Class 4Class 5Class 6Class 7Class 8

21. Class persistence 36 month Time-series Class 5Class 6 Class 7 & 8 01

22. Class Persistence distribution of classes dominant for more than 70% of observations Class 1Class 2Class 3Class 4Class 5Class 6Class 7Class 8

23. Spatial Uncertainty Analysis derived from class based assessment of OC3M with NOMAD2

24. Why Not !?

25. Class –based GSMClass –based QAA S gd varies based on class [0.0175,0.0164,0.013 9,0.0147,0.0153,0.01 28,0.0138,0.0121] a ph *() varies dependent on class  (i.e. slope of bbp) using Carder’s relationship S gd variable based on class a t (443) versus r rs (443)/r rs (555) class based a t (555) versus a t (443) class based a ph (443) versus Chl class based a ph *(443) One could imagine applying a tuning algorithm (e.g. simulated annealing) to each class to determine optimimal class based model coefficients.

27. Amoeba - NLO Spectral Unmixing

28. Conclusions

29. Future initiatives MERIS specific class-based model parameterization Investigate feasibility of implementing BEAM module for class distribution –Eventual link to Regional C2R processor Consider the possibility of a proposal to G-POD to systematically process class distribution at basin and global scales. GSFC to implement code for mapping of Optical Water Type distributions in SEADAS

30. Proposed ROI WG Initial proposal for a Regional bio-Optical algorithm Initiative (ROI) presented at 2008 IOCCG meeting Establish central portal for information on regional OC algorithms, main focus: –Capacity building –Maps of water types –Geographically located bibliography –Protocols –Round robin Expected to start some time in 2009 If interested contact Mark Dowell (mark.dowell@jrc.it) or Stewart Bernard (sbernard@csir.co.za) to be added to mailing list

31. ChloroGIN ChloroGIN is the Chlorophyll Globally Integrated Network Conceived during a IOC/GOOS/POGO/GEO sponsored workshop in Plymouth, Sept 2006 Addresses GEO task EC-06-07 – “Build upon existing initiatives e.g. ANTARES in South America … to develop a global network of organization-networks for ecosystems, and coordinate activities to strengthen observing capacity in developing countries. “ ChloroGIN aims to promote in situ measurement of chlorophyll in combination with satellite derived estimates and associated products.