The forum is locked.
The Ocean Color Forum has transitioned over to the Earthdata Forum (https://forum.earthdata.nasa.gov/). The information existing below will be retained for historical reference. Please sign into the Earthdata Forum for active user support.
Topic Special Topics / Inherent Optical Properties Workshop / some points we have to discuss (locked)
Hello all,
Just a small contribution to some of the points we have to discuss well before we gather in October.
Concerning the products we have to consider (a, adg, aph, bb), i strongly believe that these IOPs have to be assessed at different wavelengths, that is in the part of the visible spectrum where their biogeochemical interpretation is "much easier". For example, adg at 412, aph at 443, and bbp in the green (at least not at 412 and 443 where absorption effect is too "strong").
Concerning the following question: Should Rrs be "screened"/evaluated prior to IOP calculation ? I will say YES, or at least a flag should be given. This question raises another one which is related to the propagation of errors throughout the bio-optical algorithm due to imperfect atmospheric correction. The sensitivity of the selected/new-developed inverse method to this specific point is crucial in my opinion.
Concerning the validity of the in situ data set, we know that the inversion of a and bb from Rrs and Kd is straight forward when both Rrs and Kd are available (see for exemple the review of gordon in 2002 in oceanologia). The consistency of the IOPs data set with the radiometric data set could then be first tested this way.
best,
hubert
Just a small contribution to some of the points we have to discuss well before we gather in October.
Concerning the products we have to consider (a, adg, aph, bb), i strongly believe that these IOPs have to be assessed at different wavelengths, that is in the part of the visible spectrum where their biogeochemical interpretation is "much easier". For example, adg at 412, aph at 443, and bbp in the green (at least not at 412 and 443 where absorption effect is too "strong").
Concerning the following question: Should Rrs be "screened"/evaluated prior to IOP calculation ? I will say YES, or at least a flag should be given. This question raises another one which is related to the propagation of errors throughout the bio-optical algorithm due to imperfect atmospheric correction. The sensitivity of the selected/new-developed inverse method to this specific point is crucial in my opinion.
Concerning the validity of the in situ data set, we know that the inversion of a and bb from Rrs and Kd is straight forward when both Rrs and Kd are available (see for exemple the review of gordon in 2002 in oceanologia). The consistency of the IOPs data set with the radiometric data set could then be first tested this way.
best,
hubert
All,
Regarding the screening of Rrs prior to IOP retrieval: the primary issue is that Rrs may go negative in some bands. This obviously indicates an atmospheric correction and/or calibration problem. The question is, do we fail IOP retrieval if just one band goes negative (e.g., 412 or 670), even though we may not require that particular band for the inversion. A negative reflectance in any band should raise suspicion for all bands, but slightly negative Rrs at 670 in very clear water, or at 412 in high CDOM water, is not necessarily a non-physical result (i.e., weak signal + random instrument noise = negative Rrs).
For some inversion models, we could consider dynamically changing the band set used for inversion, if some bands go negative. Alternatively, we can allow negative Rrs in the inversion if it is only slightly negative (we'd need to decide on the thresholds).
We can flag suspect cases as either failure conditions or warning conditions. We then need to decide if we should mask those cases in Level-3.
-- Bryan
Regarding the screening of Rrs prior to IOP retrieval: the primary issue is that Rrs may go negative in some bands. This obviously indicates an atmospheric correction and/or calibration problem. The question is, do we fail IOP retrieval if just one band goes negative (e.g., 412 or 670), even though we may not require that particular band for the inversion. A negative reflectance in any band should raise suspicion for all bands, but slightly negative Rrs at 670 in very clear water, or at 412 in high CDOM water, is not necessarily a non-physical result (i.e., weak signal + random instrument noise = negative Rrs).
For some inversion models, we could consider dynamically changing the band set used for inversion, if some bands go negative. Alternatively, we can allow negative Rrs in the inversion if it is only slightly negative (we'd need to decide on the thresholds).
We can flag suspect cases as either failure conditions or warning conditions. We then need to decide if we should mask those cases in Level-3.
-- Bryan
Tangentially related to "changing the band set used for inversion" ...
Been reviewing the performance of SeaWiFS and MODIS-Aqua in Chesapeake Bay. Some focus was placed on Larry Harding's IOP inversion model (Magnuson et al. 2004, ECSS 61, 403-425; builds upon GSM inversion approach). Looked at wavelength sensitivity of this model, at least as it relates to SeaWiFS chl match-ups in the Bay. Only retrievals meeting 0.001 < chl < 100 mg mg-3 are considered valid.
Here are chl match-ups when Rrs(412) - Rrs(670) are considered (6 wavelengths):
(upper left image)
Here are a series of chl match-ups when Rrs at specific wavelengths are omitted (three 5 wavelength runs):
(upper right image)
Scatter noticeably increases. Note that the "no 670" run returns only half the sample size; specifically, ~75 match-ups were lost to chl falling below 0.001 mg m-3. If you compare adg(443) and bbp(443) returned from the full (6 wavelength) run to those from this "no 670" run (5 wavelengths), you can see how the loss of constraint in the red part of the spectrum can lead to erroneously low chl and high adg and bbp. The grey circles are values retrieved when 0.001 < chl < 100 mg m-3. The black circles are values retrieved when chl < 0.001 mg m-3.
(lower image)
I'll spare you our full discussion and interpretation (a document is available) -- but, figured the group might have some passing interest in this analysis, even if the results aren't surprising or 100% relevant to the "global scale" interests of this Workshop.
Been reviewing the performance of SeaWiFS and MODIS-Aqua in Chesapeake Bay. Some focus was placed on Larry Harding's IOP inversion model (Magnuson et al. 2004, ECSS 61, 403-425; builds upon GSM inversion approach). Looked at wavelength sensitivity of this model, at least as it relates to SeaWiFS chl match-ups in the Bay. Only retrievals meeting 0.001 < chl < 100 mg mg-3 are considered valid.
Here are chl match-ups when Rrs(412) - Rrs(670) are considered (6 wavelengths):
(upper left image)
Here are a series of chl match-ups when Rrs at specific wavelengths are omitted (three 5 wavelength runs):
(upper right image)
Scatter noticeably increases. Note that the "no 670" run returns only half the sample size; specifically, ~75 match-ups were lost to chl falling below 0.001 mg m-3. If you compare adg(443) and bbp(443) returned from the full (6 wavelength) run to those from this "no 670" run (5 wavelengths), you can see how the loss of constraint in the red part of the spectrum can lead to erroneously low chl and high adg and bbp. The grey circles are values retrieved when 0.001 < chl < 100 mg m-3. The black circles are values retrieved when chl < 0.001 mg m-3.
(lower image)
I'll spare you our full discussion and interpretation (a document is available) -- but, figured the group might have some passing interest in this analysis, even if the results aren't surprising or 100% relevant to the "global scale" interests of this Workshop.
Topic Special Topics / Inherent Optical Properties Workshop / some points we have to discuss (locked)