Hi, Chiara -

Are you working with Level 1B data?  If so, raw values >= 32767 indicate bad data which should be masked out in your averages.  Depending on your data dumper, that value could appear as -32768.

For specifics, please see the MODIS Level 1B Product User's Guide, found at http://mcst.gsfc.nasa.gov/l1b/.

Of course proper scaling must be applied to the scaled integers stored in the L1B file, using radiance_scales and radiance_offsets as described in that document.  If you see a negative radiance, something has gone wrong in your data extraction.

I would not recommend working with L1A data; raw data numbers are meaningless without the instrument calibrations applied in the L1B step.

Gwyn

Hi Gwin,
actually I am working with Level 2 data, but some of the files I download daily by the subscription (the earlier ones, which are mainly on earth) only show bad data. Maybe I should mask them when I apply the algorithm,as I wrote a procedure which automatically downolads them, applies the algorithm and averages them, so I have no way to directly discard the bad files.
Thank you very much for your quick answer
Chiara

Sorry, I guessed you were working with L1 data, since you mentioned the L2 chlor_a separately.

So you're doing your own derivation of chlor_a from the Rrs values?  Definitely you should mask the data at the pixel level as part of your algorithm, using the l2_flags described on page 9 of this document:

http://oceancolor.gsfc.nasa.gov/DOCS/Ocean_Level-2_Data_Products.pdf

Hi Gwin,
actually I do mask the data at the pixel level.
I apply the algorithm by command mode, and I mask the data as on example 3:

;  initialize the result (new band data chlor_a to be created)
;
result = b1
result(*) = -999.9
;
;  use the IDL function ARRAY_BIT to get all the pixles (iy, array of pixel
;  index) from the l2_flags (r3, band 3 raw data) that have the atmospheric
;  correction algorithm failure (bit 1), land (bit 2), sun glint (bit 4),
;  total radiance greater than knee value (bit 5), or cloud and ice (bit 10)
;  masks set
;  set the result of those pixles to 0.0
; 
iy = ARRAY_BIT(r4, [1,2,4,5,10], jcnt)
if (jcnt GT 0) then result(iy) = 0.0
;
;  the following calculations are based on the MedOC3 algorithm that is
;  proposed by Santoleri
;
;  set the result to 0.0 for pixels in nLw_443 (b1, band 1 geophysical value)
;  or nLw_488 (b2, band 2 geophysical value) or Lw_547 (b3, band 3 geophysical value) that have values less or equal
;  0.0
;
iy = WHERE(b2 LE 0.0 OR b3 LE 0.0 OR b1 LT 0.0, jcnt)
if (jcnt GT 0) then result(iy) = 0.0

Please find attached one of the files which give me troubles.

Thank you

Chiara