Appreciate if someone can help in the questions 

1)  By seadas msl12,  i processed L1a to get L2 images ot get more variables
including chl_gsm01, acdm_gsm01 and bbp_gsm01 that are the products
by optimization of the Garver-Siegel-Maritorena-2001 model.  I realized that most
of the values in chl_gsm01 are negative and the other two  are not in the region
( coastal waters near the Cap Cope) region).  Does that suggest the algorithm may not work properly in this region?

2)  I selected brdf_opt = "Morel f/Q & gothc-R" for msl12 processing. The Morel et al's BRDF
needs chlorophyll  as input. I assume that msl12  is using OC4 chl, right? and the retrieved
water-leaving  readiance Lw_xxx has been corrected to nadir-viewing Lw, right?

Thanks

Yes, the implementation of Morel f/Q uses the sensor default chlorophyll algorithm for indexing into the tables.  For seawifs this is OC4.

I would also add that the use of alternate BRDF options should be accompanied appropriate by recalibration.  The current default vicarious calibration is:

  gain=[1.0186,1.0018,0.9685,0.9876,0.9978,0.9625,0.941,1]

with full Morel f/Q + Gothic-R, it should be:

  gain=[1.0176,1.0008,0.9676,0.987,0.9973,0.9625,0.941,1.00]

this might help.  Just plug the new gains in the seadas GUI and set vcal_opt=1.

Brian and Sean

Thanks for your replies. Still need make sure:

I am processing in a batch-mode to call the msl12 procedure:
     msl12,
     .......
     brdf =3,     ; for full Morel f/Q + Gothic-R
     vcal_opt=1, [1.0176,1.0008,0.9676,0.987,0.9973,0.9625,0.941,1.00]
     .....

    Is that  setup right?

Brian mentioned that  "the use of alternate BRDF options should be accompanied appropriate by recalibration".  I do not really get it.
That means both options need selecting if BRDF correction is considered. Any document  can be pointed to me?

Thanks again

Hui

Yes, for f/Q + Gothic-R:
brdf=3,vcal_opt=1,gain=[1.0176,1.0008,0.9676,0.987,0.9973,0.9625,0.941,1.00]

This is not documented.  I can only tell you that these are the proper gains because we happen to be testing this option.  We can not reccomend it at this time.  As currently implemented, it seems that the flat-ocean Fresnel correction results in the lowest residual scan-dependence for SeaWiFS retrieved nLw.

-- bryan