Ocean Color Forum
Topic Special Topics / Inherent Optical Properties Workshop / Rrs(0+) -> Rrs(0-) & water coefficients (locked)
By bryanfranz Date 2008-08-05 14:32
Here's a simple question for group consensus:  what is the preferred method of converting from above-water to below-water reflectance.  The competition seems to be between these options:

1) Rrs(0-) = Rrs(0+) / 0.5238

2) Rrs(0-) = Rrs(0+) / [0.52+1.7 Rrs(0+)]

-- bryan
By houghtah Date 2008-08-06 08:43
Following, might we agree upon a standard suite of aw and bbw coefficients for use in our analyses?  I acknowledge the ongoing debate regarding bbw values, but perhaps internal consistency is best, at least to start?  Although, analyses regarding the sensitivity of these "constants" might be prudent for the Workshop.  This is a list of 11-nm average aw and bbw for SeaWiFS, MODIS, and MERIS, as recorded here (aw = Pope and Fry 1997, bbw = 0.0038 * (400. / wl)^4.32).  Thoughts?

`    412      443      490      510      555      670aw  0.00457, 0.00706, 0.01513, 0.03253, 0.05979, 0.44054bbw 0.003344,0.002445,0.001581,0.001330,0.000923,0.000409    412      443      488      531      551      667      678aw  0.00457, 0.00706, 0.01448, 0.04395, 0.05979, 0.43395, 0.46040bbw 0.003344,0.002445,0.001610,0.001118,0.000923,0.000417,0.000389    412.5    442.5    490      510      560      620      665      681.25aw  0.00453, 0.00695, 0.01513, 0.03253, 0.06211, 0.27570, 0.42970, 0.47251bbw 0.003327,0.002457,0.001581,0.001330,0.000888,0.000572,0.000423,0.000381`
By esjboss Date 2008-08-06 10:40
Dear all,

Shouldn't we vary the water IOPs for Salinity and Temperature? bbw is
strongly salinity dependent (30% between zero and 39psu) its value
will not change much between open ocean environment but since we know
it we may as well get used to do it (and thus will not forget to
change it when we do Great Lakes work). Similarly modulating
absorption with temperature is important in the red and NIR (we can
also salinity correct).

Note that salinity and temperature corrections are often applied to
the ac-9 simply so their effect on water is not translated to the
signal relative to water (due to dissolved and particulate materials).
When water is added back to a_pg and c_pg (e.g. when total a and total
c are reported), most often the water added is fresh and at a specific
temperature (e.g. Pope and Fry's) and thus is not representative of
the conditions where the water was measured (e.g. arctic ocean). It is
important that any ac-9 data used here should be those of the
particulate and/or dissolved components only (w/o the added water).

In coastal waters this may not be an issue but in mid-ocean gyre it
will be. In addition this introduces a bias we can easily do without.

Another issue: the value you report below have probably been evaluated
using a 1 or 2nm bandwidth spectrophotometer. The satellite bandwidth
is larger (10-20nm) which means it averages over a certain band range.
We should do our best to perform that average on all optical
properties we use (validation set and water values) to insure that
areas where these properties may be changing non-linearly are not
introducing biases. in-situ ac and bb instrument have typically 10 or
20nm bandwidth (though some c meters have laser sources with less than
1nm bandwidth).

Cheers,
Emmanuel
By bryanfranz Date 2008-08-06 13:24
Emmanuel,

The numbers we actually use for aw and bbw in satellite processing are 11-nm averages derived from 1-nm tabulated values.  The tables can be obtained here:

ftp://samoa.gsfc.nasa.gov/pub/franz/code/data/eval/common/water_spectra.dat.

We use 11-nm averages because, in standard operational processing, we report the satellite retrievals for a 10-nm square band-pass centered on the nominal wavelength (i.e., we correct the nLw retrieval from full band-pass to nominal center band as the last step in the normalization of Lw).  If this correction is not applied, the code uses full-band aw and bbw, as reported here:

http://oceancolor.gsfc.nasa.gov/DOCS/RSR_tables.html

These are the 11-nm averages that we currently use for aw and bbw (as computed internally):

SeaWiFS      aw          bbw
0     412  4.641e-03  3.327e-03
1     443  7.098e-03  2.438e-03
2     490  1.527e-02  1.583e-03
3     510  3.260e-02  1.334e-03
4     555  5.965e-02  9.299e-04
5     670  4.402e-01  4.171e-04
6     765  2.851e+00  2.376e-04
7     865  4.603e+00  1.413e-04

MODIS         aw          bbw
0     412  4.641e-03  3.327e-03
1     443  7.098e-03  2.438e-03
2     469  1.050e-02  1.909e-03
3     488  1.458e-02  1.611e-03
4     531  4.394e-02  1.123e-03
5     551  5.697e-02  9.590e-04
6     555  5.965e-02  9.299e-04
7     645  3.267e-01  4.903e-04
8     667  4.339e-01  4.251e-04
9     678  4.610e-01  3.966e-04
10     748  2.848e+00  2.614e-04
11     859  4.421e+00  1.455e-04
12     869  4.748e+00  1.385e-04
13    1240  1.152e+02  2.865e-05
14    1640  6.356e+02  8.564e-06
15    2130  2.513e+03  2.768e-06

MERIS         aw          bbw
0     412  4.641e-03  3.327e-03
1     443  7.098e-03  2.438e-03
2     490  1.527e-02  1.583e-03
3     510  3.260e-02  1.334e-03
4     560  6.206e-02  8.950e-04
5     620  2.756e-01  5.801e-04
6     665  4.282e-01  4.306e-04
7     681  4.716e-01  3.892e-04
8     708  7.827e-01  3.300e-04
9     754  2.865e+00  2.527e-04
10     761  2.866e+00  2.430e-04
11     779  2.700e+00  2.200e-04
12     865  4.603e+00  1.413e-04
13     885  5.564e+00  1.283e-04
14     900  6.415e+00  1.195e-04

-- Bryan
By esjboss Date 2008-08-06 13:39
Thanks Bryan.
I think we need to modulate them with temperature (measured) and salinity (climatology).
Thanks,
Emmanuel
By bryanfranz Date 2008-08-06 14:50
Yes, I think that's a good suggestion. It shouldn't be too difficult to achieve.  For SeaWiFS we'd probably use Reynolds SST (weekly 1-deg or the new daily 0.25-deg product), which we already ingest.  Not sure if there is an equivalent SSS product available.  If you or others can post the dataset(s) to use and the correction method, that would solidify the recommendation.

-- bryan
By jilly Date 2008-08-06 19:25
Hi bryan,

I agree it's worth tackling T and S adjustments now.  The best/newest/only global salinity dataset i know of is the World Ocean Atlas

http://www.nodc.noaa.gov/OC5/WOA05F/woa05f.html

But both T and S on a 1deg grid are not going to be ideal for, say, estuaries with strong, tide-driven changes in T and S over time;  we need to check the magnitude of the T/S corrections at in situ matchups where CTD data are available, and see how suitable the climatologies are.

My 2cents,
jill.
By esjboss Date 2008-08-07 01:23
Bryan,
I agree with Jill on the source for S coefficient.
Even if we may be off on the exact salinity value, we will be better, on average, than not accounting to it.
The source for T, as Bryan suggested, could be within the ocean color output.

The latest numbers on the backscattering coefficient (including S contribution) are in Twardowski et al., 2007 in Biogeoscience. For the S and T contribution to to a_sw I used to use Pegau and Zaneveld, 1997.  Newer values were published by Sullivan et al., 2006, in Applied Optics, using the ac-s.

Cheers,
Emmanuel
By houghtah Date 2008-08-07 09:52
For what it's worth, here's some (sloppy) IDL code to calculate b_sw(wl,temp,sal):

b_sw.pro

Based on Twardowski et al. 2007 and Buiteveld et al. 1994.

I'll use this for some sensitivity analyses.  Will hopefully provide a template for l2gen.

But, someone might want to check my math ...

Will compile similar software for a_sw when back in office next week.
By houghtah Date 2008-08-12 15:52
Here's a crack at IDL software to calculate a_sw(wl,temp,sal):

a_sw.pro

Based on:
(1) aw (22 C, 0 PSU) from Pope and Fry 1997
(2) temp and sal correction factors from Table 2 of Sullivan et al. 2006
(3) the relationship [ aw_corr = aw_meas - temp_corr * (temp - 22.) - sal_corr * sal ]

Returns either (a) 11-nm average of aw centered on input wl, or (b) aw spline fit to 1-nm increments from 400-700 nm.

Again, someone should really check my implementation.

Otherwise, enjoy.
By dem1 Date 2010-03-30 09:45
Bryan,

Are these values still up-to-date or did you have recomputed them with the sst and salinity correction?
And please could you give me the nominal 11-nm square bandpasses values of aw, bbw and also F0 you use for the new MODISA 547 band (old 551 re-defined)?

Thanks,

Julien
By bryanfranz Date 2010-03-30 14:15
Julien,

The numbers have not changed.  We haven't implemented salinity and temperature effects yet, but that would likely be done as a correction algorithm applied to these values. Here's latest numbers for MODIS with 547.

wave    aw              bbw
412 4.640990e-03 3.327229e-03
443 7.098098e-03 2.437585e-03
469 1.049904e-02 1.909299e-03
488 1.457921e-02 1.610938e-03
531 4.393521e-02 1.122943e-03
547 5.337546e-02 9.892991e-04
555 5.964877e-02 9.298773e-04
645 3.266578e-01 4.902821e-04
667 4.339156e-01 4.251321e-04
678 4.609707e-01 3.965882e-04
748 2.848320e+00 2.613615e-04
859 4.421082e+00 1.454898e-04
869 4.748336e+00 1.385429e-04
1240 1.151705e+02 2.865000e-05
1640 6.356376e+02 8.563636e-06
2130 2.512734e+03 2.768182e-06
By esjboss Date 2010-03-30 17:44
If you are to add the salt effect on bbw (which you should) the latest values are available from X. Zhang papers from 2009 in Optics Express.
All the best,
Emmanuel
By dem1 Date 2010-03-31 04:40
Thanks a lot Bryan,

Do you have also the corresponding F0 for 547?

Best,
Julien
By Loisel Date 2008-08-07 12:57
Hi all,

Here are some elements concerning the conversion of Rrs(0+) to Rrs(0-).
These results are based on hydrolight simulations used in the IOPs
inversion workshop IOCCG, report 5). Only wavelength at 490 nm, and 3 sun
angles (0, 30, 60) are considered here.

A constant values is not appropriated as i) this value change with the sun
angle (see figure called constant-ratio), and ii) a hyperbolic fit
((x/(a+bx), with a and b some constants) is much more appropriated than a
linear fit.

For the hyperbolic fit, and based on the data set used in the IOCCG report,
the constant a and b are different from those given in Lee et al. 1999
(a=0.52, and b=1.7). However, the difference between the Lee et al's fit,
and the one presented here is quite small (not more than 5%, at this
wavelength). However, one may see an effect of sun angle (as expected) on
this conversion (see figure called polyfit).

I believe that we should use the hyperbolic fit instead of a constant
value. Moreover, I think it will be better to include the effect of sun angle
in this hyperbolic form (through a parameterisation or tables). It is quite
easy, and we have the simulations (perhaps two other sun angles will be
useful: Zhong Ping do you still have the inputs of the IOCCG-simulations ?).
We know that it will improve the conversion, so ....

best,
hubert
By zplee Date 2008-08-07 14:56
Hubert and all:

The (0.52 & 1.7) coefficients were developed for nadir view, as an overall average for various Sun angles. As Hubert showes here, apparently not that bad (within 5%). The 0.52 will change a little for view angle away from nadir, but to what extent will depend on angles and VSF, etc. We have an ongoing project looking into that, and would be happy to report our findings later this year or next year.

On the other hand, these coefficients affect more to derived bb or bbp, less to atot. If all measurements are made within a 30 deg view angle and Sun are within 70 deg from zenith, the angle-introduced error to retrieved IOPs will be small (<10%).

As to Hubert's question of IOCCG data set, yes, it is there, but only nadir-viewed Rrs(0+) and Rrs(0-).

zhongping
By sokoletsky Date 2011-11-23 16:02 Edited 2011-11-23 16:59
Hi Hubert,

Thank you for interesting information! I am also interested by this issue for my ocean color modeling. I did my own attempts to find Rrs(0+)/Rrs(0-) ratio. Actually, a form of this ratio was established by Saunderson in 1942 (so called "Saunderson's correction"), and it is just in the form (hyperbolic) that use you or Lee at al. I also used this form, attracting some theoretical (Fresnel & Snell's laws; Quan and Fry, 1995; Morel and Gentili, 1996; Molenaar et al., 1999), observational (Aas and HÃ˜jerslev, 1999; Toole et al., 2000) and radiative transfer modeling (Mobley, 1999; Aas, 2010; Kokhanovsky & my own, unpubl.) stuff. I have results now for any wavelength and solar zenith and viewing angles, but for the better comparing with your results, I may give "a" & "b" for your lambda (490 nm),  3 sun angles (0, 30, 60) and nadir viewing:

0 deg: a = 0.515, b = 1.642
30 deg: a = 0.515, b = 1.744
60 deg: a = 0.515, b = 2.057

Thus, my results are extremely close to Zhongping's results, and, of course, close to yours.

Interestingly to analyze an impact of temperature and salinity. My model allows to do this. As an example, I can show coefficients for the normal temperature (25 deg), extremal high salinity (280 per mil) and previous wavelength and angles:

0 deg: a = 0.479, b = 1.768
30 deg: a = 0.479, b = 1.877
60 deg: a = 0.479, b = 2.214

Best wishes,

Leonid
Topic Special Topics / Inherent Optical Properties Workshop / Rrs(0+) -> Rrs(0-) & water coefficients (locked)