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 The information existing below will be retained for historical reference.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.
All,
I have posted an analysis comparing IOP inversion at Level-2 (composited to global monthly) with IOP inversion on Level-3 monthly mean Rrs.
http://oceancolor.gsfc.nasa.gov/MEETINGS/OOXIX/IOP/l2_vs_l3.html
In summary: small effect on average.
-- bryan
I have posted an analysis comparing IOP inversion at Level-2 (composited to global monthly) with IOP inversion on Level-3 monthly mean Rrs.
http://oceancolor.gsfc.nasa.gov/MEETINGS/OOXIX/IOP/l2_vs_l3.html
In summary: small effect on average.
-- bryan
Bryan:
Very nice results showing! As for the QAA difference between the two routes, one factor needs to be kept in mind is that the present version is designed for nadir-looking Rrs (ie works better with normalized Rrs). I will share with the group at the workshop of the next version that will take account of the angular variations.
Cheers!
zhongping
Very nice results showing! As for the QAA difference between the two routes, one factor needs to be kept in mind is that the present version is designed for nadir-looking Rrs (ie works better with normalized Rrs). I will share with the group at the workshop of the next version that will take account of the angular variations.
Cheers!
zhongping
Ping,
I would expect that such a change would make the differences between L2 inversion and L3 inversion larger. Do you agree?
-- bryan
I would expect that such a change would make the differences between L2 inversion and L3 inversion larger. Do you agree?
-- bryan
Bryan:
If I understand right, in L2 inversion, it is Rrs(angle) used; in L3 inversion, it is Rrs(0) used. This "0" means nadir-looking with Sun at zenith. And, it is the case-1 f/Q used to convert Rrs(angle) to Rrs(0) at present.
Now, if we use another scheme to convert Rrs(angle) to Rrs(0); or, with angular-specific parameters for IOP inversions in L2, the difference should be reduced if everything is consistent (especially the two kinds of Rrs(angle)->Rrs(0) conversions). However, if the case-1 f/Q is not designed for waters such as eutro, then hard to say.
I guess I am trying to say is that the 'gap' between L2 and L3 inversions are understandable, as most inversion models use same set of parameters for Rrs at all angles. On the other hand, for eutro water, the case-1 f/Q will have larger uncertainties.
Cheers,
zhongping
If I understand right, in L2 inversion, it is Rrs(angle) used; in L3 inversion, it is Rrs(0) used. This "0" means nadir-looking with Sun at zenith. And, it is the case-1 f/Q used to convert Rrs(angle) to Rrs(0) at present.
Now, if we use another scheme to convert Rrs(angle) to Rrs(0); or, with angular-specific parameters for IOP inversions in L2, the difference should be reduced if everything is consistent (especially the two kinds of Rrs(angle)->Rrs(0) conversions). However, if the case-1 f/Q is not designed for waters such as eutro, then hard to say.
I guess I am trying to say is that the 'gap' between L2 and L3 inversions are understandable, as most inversion models use same set of parameters for Rrs at all angles. On the other hand, for eutro water, the case-1 f/Q will have larger uncertainties.
Cheers,
zhongping
Ping:
In both inversion approaches it is Rrs(0) that is used for input to QAA and GSM. The method is that of Morel, and it is applied at Level-2 with exact geometry. The "exact" Rrs at Level-2 is then either input to the IOP model and then binned, or binned and then input to the IOP model.
For PML we used Rrs(angle) for both L2 or L3 inversion, since the PML LUTs are expecting Rrs(angle). This is a problem for the Level-3 inversion simply because the geometry (especially sensor zenith and relative azimuth) is ill-defined for binned Rrs. I am using mean solar zenith, mean sensor zenith, and mean relative azimuth in that case. In a monthly Rrs mean, the observation going into the bin come from both sides of the swath and everywhere in between. I suspect that mean geometry is not particularly representative.
I agree that our present approach to f/Q correction via Morel will be in greater error in eutrophic waters. I hope that you (and/or Jill, Taka, Tim) can demonstrate that this issue, or IOP retrieval in general, is improved by incorporating the f/Q correction into the IOP inversion.
-- bryan
In both inversion approaches it is Rrs(0) that is used for input to QAA and GSM. The method is that of Morel, and it is applied at Level-2 with exact geometry. The "exact" Rrs at Level-2 is then either input to the IOP model and then binned, or binned and then input to the IOP model.
For PML we used Rrs(angle) for both L2 or L3 inversion, since the PML LUTs are expecting Rrs(angle). This is a problem for the Level-3 inversion simply because the geometry (especially sensor zenith and relative azimuth) is ill-defined for binned Rrs. I am using mean solar zenith, mean sensor zenith, and mean relative azimuth in that case. In a monthly Rrs mean, the observation going into the bin come from both sides of the swath and everywhere in between. I suspect that mean geometry is not particularly representative.
I agree that our present approach to f/Q correction via Morel will be in greater error in eutrophic waters. I hope that you (and/or Jill, Taka, Tim) can demonstrate that this issue, or IOP retrieval in general, is improved by incorporating the f/Q correction into the IOP inversion.
-- bryan
Bryan:
Thanks for your further explanation, and that means my earlier "understand" is not right.
Cheers!
zhongping
Thanks for your further explanation, and that means my earlier "understand" is not right.
Cheers!
zhongping