We've been working on an over-ocean aerosol retrieval algorithm from SeaWiFS and I am comparing my results to those from the OBPG algorithm (AOD at 865 nm). I'm also exploring the differences in the context of AERONET direct-Sun observations at 49 different island/coastal sites, and had a couple of quick questions. Part of the difference will be because of the aerosol models we are using in the retrieval (we have a bimodal lognormal distribution and vary the fine/coarse split, and it also changes to be dust-like or smoke-like at high AODs), I imagine surface reflectance assumptions will also play a part in many cases.

Last year's paper (Ahmad et al, AO 49, 2010) presented new aerosol models for use in the SeaWiFS ocean colour processing. Over here we have been testing our retrieval using SeaDAS 5.2.0 for the L1a->L1b conversion, and using l2gen from that version to create the OBPG level 2 AOD data. We will do our final processing run (and regenerate the OBPG data) using the current SeaDAS (6.2.0?) so I expect the final results will change in part at least due to calibration updates. My question is, does SeaDAS 5.2.0 use the Ahmad et al (2010) aerosol models or the older Gordon & Wang (1994) set? As mentioned I will run with the newest SeaDAS for our final processing, but I realise now that by using SeaDAS 5.2.0 it is possible that we will have been using the outdated aerosol models, so I am not sure how useful the statistics I have generated will be now.

If there is a convenient time for you guys, I would be happy to pop over (I am at GSFC) and discuss the comparison in person some time?

SeaDAS 6.0 was the first to include the new aerosol models (and many other updates related to reprocessing). See:

  http://oceancolor.gsfc.nasa.gov/REPROCESSING/R2009/

-- bryan

Hi Bryan,

Thanks for the confirmation. We are going to rerun l2gen using the current SeaDAS before I take this further, then (no point comparing outdated results!). We will also need to see how the new calibration affects our retrievals, although those will take somewhat longer to reprocess.

Andy

If you are thinking of using SeaWiFS L1B from SeaDAS 5.2 with l2gen processing from SeaDAS 6.x, I would advise against it.  The prelaunch calibration, as applied in L1B, was changed for the reprocessing.  The vicarious calibration, as applied in l2gen, was previously compensating for error in the prelaunch calibration.  The l2gen in SeaDAS 6.x does not have this correction, so the 5.2 L1B is radiometrically incompatible with 6.x l2gen.  There is really no reason to produce L1B, since l2gen takes L1A SeaWiFS as input. 

-- bryan

For the OBPG processing, we're directly using SeaDAS to go from L1a to L2 (so there should be no inconsistency with the calibration). Our own processing uses SeaDAS for L1a->L1b but we're generating these on-the-fly, so again when we switch to the newest SeaDAS it shouldn't generate any inconsistency in the calibration. In both cases the L1b are generated only temporarily and deleted. So we will not be using SeaDAS 5.X and SeaDAS 6.X products together.

Just to confirm, are you saying that the SeaDAS 6 applies all calibration in the L1a->L1b stage (i.e. l2gen in SeaDAS 6 does not apply any calibration corrections which are not applied by l1bgen)?

Ah, ok--I will have to check and make sure that we are applying it.

When we've got both the OBPG data and our algorithm processed with the newest SeaDAS, are you the best person to call/email to discuss the comparison?

Andy

I would suggest you discuss with Zia Ahmad for starters.  He's been doing other AOD comparisons.

-- bryan

Thanks, will do.

One more question: going through a previous post about calibration from a few months ago, http://oceancolor.gsfc.nasa.gov/forum/oceancolor/topic_show.pl?pid=16351 , suggested that the vicarious calibration *is* applied by l1bgen. Could I just confirm quickly, which of these is correct, if we want to process SeaWiFS data with our own algorithm:

1. For the most current SeaWiFS calibration, apply l1bgen from SeaDAS 6.2.0 to l1a files.

2. For the most current SeaWiFS calibration, apply l1bgen from SeaDAS 6.2.0 to l1a files, and then apply additional calibration which is contained in l2gen to the result.

Andy,

l1bgen applies the default vicarious gains to the level-1b data unless it is told to do otherwise through the command line parameter 'gain=[xxx]' where xxx is the vicarious gain applied to each band.  If you then use the level-1b data as input to l2gen, l2gen also applies the default vicarious gains to the level-1b data unless told to do otherwise.  You have to be careful in using l1bgen and l2gen so that you do not wind up applying the vicarious calibration twice.  When I use l1bgen for the SeaWiFS lunar data analysis I use 'gain=[1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0]' on the command line to turn the vicarious calibration off.

Gene Eplee

Thanks, Gene, Bryan, I think I have it now.

So, when we process with the OBPG algorithms, we will use l2gen on the l1a data (SeaDAS 6.2).

When we process with our own algorithm (which takes l1b as input), we will use l1bgen on the l1a data (SeaDAS 6.2), applying the vicarious gains with that, but then no additional calibration.

Andy