IOP Workshop update, Feb 2010

By @jeremy Date 2010-02-16 14:30

Greetings, all, from a ridiculously snowy Maryland. Hope that the New Year has been treating everyone well. On this Eve of the MODIS-Aqua reprocessing (assuming that tonight's snow doesn't accumulate or freeze), we'd like to reinitiate conversations regarding IOP algorithm implementation at NASA. With the radiometric corrections that accompanied preparation for the SeaWiFS
and Aqua reprocessings behind us, the goal of the OBPG is to refocus on bio-optical algorithms for the next few months. Here, I'll provide a very brief summary on what's been going on here and what we hope to accomplish in the near future, with the hope of sufficiently piquing your interest to provide comments, suggestions, and analyses in the coming months.

We left OOXIX tasked with developing a framework for configuring and testing SAAs, starting with a general spectral optimization model form. To that end, GIOP (Generic IOP) was developed and recently released via SeaDAS. Briefly, GIOP is an infrastructure for building a spectral optimization model on the fly -- the user chooses the inversion/optimization method (e.g., Levenberg-Marquardt, matrix inversion, etc.) and the eigenvectors for aphi (user-defined, Bricaud, Ciotti, etc.), adg (QAA, user-defined S), and bbp (QAA, user-defined eta, Loisel and Stramski eta, etc.) For example, GSM can be run using GIOP by defining the eigenvectors, etc. reported in Maritorena et al. 2002. GIOP will be expanded somewhat soon to include multiple eigenvectors for each parameter (e.g., ad and ag rather than adg). We're currently in the process of analyzing sensitivities of various parameter combinations.

The preliminary configuration for GIOP we decided upon at the conclusion of the workshop (PQN) will be used to generate a long time series of SeaWiFS and Aqua products that will be posted online for your perusal. While results aren't yet available, they will ultimately also be posted on our Level-2 match-up site (http://seabass.gsfc.nasa.gov/seabasscgi/validation_search.cgi)
and Level-2 regional time-series site (http://oceancolor.gsfc.nasa.gov/cgi/regions.cgi). Note that other data products (including IOPs) will be available via these metchanisms as well. Level-3 analyses for PQN will also be generated and distributed. On our to-do list is devising new ways and implementing existing ways to test Level-2 and -3 coverages, uncertainties (Wang et al. 2005, Lee et al. 2010, Moore et al. 2009, GlobColour, etc.), sensitivities, etc. We've also implemented the fuzzy logic methods of Mark D. and Tim M. to identify optical water types, with the goal of merging its capabilities with that of GIOP.

Probably enough news for now. Again, sorry for our absence during reprocessing, we hope this finds everyone well and look forward to hearing about what you've been up to over the year. I imagine you've accomplished or formulated ideas about many of the topics listed above (and hopefully, are willing to share). We'll write again a week or so after GSFC reopens for business.

Best,

J and the gang at the OBPG

By MarkDowell Date 2010-02-16 17:29

Hi Jeremy (and all),
congrats on the start of the reprocessing, hopefully.

A few initial questions and comments.

Can we assume at this point that NOMAD v2 dataset version - nomad_seabass_v2.a_2008200 - is the one we should be using, at least for this phase of the initial parameterisation.

A few requests for NOMAD. If these are out-of-order let me know and I am happy to contribute, but I can appreciate that after all the time and effort you have put into NOMAD in the past you may not want other people dabbling too much.

I think, as previously discussed, it may be useful if you could systematically calculate Sgd (using the wavelength range and fits you normally use).

Also I think we suggested generating "modelled" bb(lambda) using the measured Rrs and Kd and Hubert's method, so as to have a more exhaustive coverage for the bb data (as commented at the workshop this could flagged as generated data).

Finally, for the aph do you have access to the full hyper-spectral data from SeaBASS for the NOMAD stations. It occurred to us that with this, the eta and Sgd we could run a hyperspectral forward model, which may be of use for sensitivity analyses, f/Q parameteristaions, to rapidly pick up wavelengths for future satellites and (shameless self-interested request) eventually to produce a hyperspectral co-variance matrix for the fuzzy membership calculations of the OWT.

Once I have the ok on NOMAD I will start to generate some Kd(lambda0) vs Kd (lambda) relationships for the OWT so as to parameterise the Loisel and Stramski eta calculation and then Tim and I (right Tim) will make a *first attempt* at GIOP model sub-component parameters (i.e. Sgd, eta and aph*) for the OWT. I am hoping to get Tim over to JRC sometime in May.

No rush on this , i just think they would be useful before we start making heavy use of NOMAD for parameterisation.

Regards from India
Mark

PS I made (read:stole from the internets) this for you, BTW it's 35 degC here.

   .--------.
*    . |________| . *
   | __|/\
* .-'======\_\o/.
/___________<>__\
|||||| / (o) (o) \
|||||| |   _ O _   | .
.    |||||| | (_)   (_) |
|||||| \   '---'   / *
\====/   [~~~~~~~~~]
   \\// _/~||~`|~~~~~\_
   _||-'`/ || | \`'-._    *
   * .-` )| ;   || |) ; '.
   / `--.|   || |    | `\
|    \   |||||) |-, \    .
   \    .;    _    ; |_, |
`'''||` ,\    (_)    /, `.__/
||.` '.    .' `.    *
   * ||    ` ' ' `    \
|| ;
. * || | .
|| | *
|| |
.__.-""-.__.-"""|| ;.-"""-.__.-""-.__.
||    /
||'.    .'
|| '-._ _ _ _ _.-'
   `""`

By @jeremy Date 2010-02-17 15:03

Hey Mark,

Good to hear from you. Spend a week in Goa last month and enjoyed a few 30C temps myself. Feels like a million years ago, though, given the latest Maryland weather. Your snowman makes me feel much better, though ;-)

As for your questions:

Simple to add Sdg, eta, and Loisel and Stramski modeled parameters (etas, bb) to NOMAD. Already have the tools in place to add these parameters -- just got distracted recently and ultimately heard a rumor that Hubert had updated his model. Hubert, can you confirm or deny? If you confirm, can I get a preview?

Adding hyperspectral aph is possible, but will take more time. We made a decision long ago (I'll leave it to the group to decide if it was a good or bad one), to limit the wavelength range to permit organization of NOMAD into a single file. Programs like Excel and IDL (structures) cannot handle more than 256 columns (or, they couldn't at the time). IOP optically-weighted near-surface values were calculated on the multispectral (20 wl) data set, so there isn't necessary an simple correspondence to the original aph files in SeaBASS. A solution to this is simply to reprocess and provide hyperspectral aph data in an external file that corresponds in size to the main NOMAD file, which we've done in the past. External files that include raw (unweighted and unsmoothed, but still multispectral) IOPs should be available on the NOMAD Web site. I'll see if we can get this aph thing done, perhaps secondary to the Sdg, eta, LAS modifications. Would something like 2, 3 or 5-nm resolution be sufficient? I think the spectral resolution varies somewhat in the source data set.

Yes, the NOMAD v2 dataset is the most recent version. Plans for v3 evolved over the past year, so nothing to post yet.

Hope things are going well.

J