• disable the -Extract... and -Calibrate... argument
  • -extract...
  • -calibrate...
• check that -Model... argument is enabled
• ensure no mask is used   -Model/mask_4

• The two -CP... arguments can be taken out or disabled,
  • as this refinement only applies to very limited environment,
  • and is certainly not justified over other areas.
    • @CP/196.96/157.85/058.64_0.25m @@
    • @CP/185.04/130.29/029.71_0.50m @@ 

• run the script tarawa-subset.sh
  • if -Model/Mask_4 is left enabled, only pixels under Mask_4 are processed
    • while other pixels are simply duplicated.
  • to prevent this, please ensure Mask_4 is disabled into mask_4 with a lowercase m,
    • so that 4SM processes the whole image.

• Smoothing: -Smooth_3d/smart+
  • -smooth...,   smoothing disabled
  • -Smooth...,  smoothing enabled(for marine pixels only)
    • WinRadMax=5 is applied: the circular smoothing window has a radius of 5 pixels
      • that is 81 pixels in the smoothing kernel
    • d no smoothing of optically deep waters
    • D smoothing of pixels masked at mSE=240 (optically deep waters)
  • /smart...  plain smoothing applies
  • /Smart... smart smoothing applies
  • + SmoothPlus is enabled
  • - SmoothPlus is disabled
• Smart-Smoothing is a complex feature in 4SM. 
  • Enough of that for the moment!

The two -CP... arguments can be taken out or disabled,
as this refinement only applies to very limited environment,
and is certainly not justified over other areas in this image. 

• Minimum threshold radiance values are applied in order to control artifacts:
  • -Lm0001.0/0004.0/0005.0.
• You need to experiment with those thresholds, see what happens, until satisfied, as -deglint... does not solve all problems.
  • Set Lm[2] to 1.0 instead of 4.0, see that
    • we seem to have foul waters at some locations inside the lagoon
    • the electronic noise suddenly invades the output.
  • Set Lm[3] to 255 to disable the use of the NIR band altogether,
    • see that this does not affect computed depths over bright sandy bottoms.
    • but over very shallow seaweeds or algae, which is not the case here,
      • the use of XS3 can make a world of a difference.
  • Set Lm[2] to 255 to see the extent of RED coverage.
  • set Lm[2] to .../004.p/... to enable progressive Lm

• NIR solution must be enabled: - M0001/00002/00003
  • it is controlled through the -Lm[NIR]
• ONE-band-case enabled: note in the - M00001/00002/00003 argument that all three bands are enabled:
  • in deeper areas where the Red band is extinct
  • depths are computed using XS1 alone (the one-band solution), assuming constant bottom brightness Lbref[1]
  • this is obtained by enabling the first band in the -M... argument and choosing an appropriate value for Lbref[1] in -B... argument
• TWO-band-case only
  • to disable the one-band case
  • change -M00001/00002/00003 into - M@00001/00002/00003 .
  • ==> deeper areas are mapped to optically deep waters instead.

• Lbref[1] : is used to compute Z in the ONE-band-case :
  • -B/tclne_3.0/200_100/1/1.00
  • LBref[1]=200 is used for the One-band case for pixels that are assumed to be bright bottoms
  • change its value to other values like 170 or 220, see what happens:
    • this affects retrieved depth
    • Z=(log(LBref-Lw)-log(Ls-Lsw))/K
  • choose your prefered value for LBref[1].
• Lbref[2] : is used to compute Z in the ONE-band-case
  • -B/tclne_3.0/150_040/1/1.00
  • this used for the one-band case for pixels masked at 236:
  • typically in areas that are assumed to be dark bottoms, like coral reefs, outer reef slope...
• CoefLM : 
  • note that the last parameter is set to CoefLM=1.00 
    • -B/tclne_2.0/150_040/1/cLM1.00
  • All water column corrected reflectances are multiplied by this CoefLM value.
  • Change its value to any other value, like 1.2 or 0.87, see what happens
    • see that computed bottom reflectances are affected accordingly
    • while computed depths remain unaffected.
  • Once/if a new -LsM... argument has been settled and enabled
    • make sure you return CoefLM to 1.0
      • -B/tclne_2.0/150_040/1/cLM1.00 

• This argument is comprised of many variables
  • which are all provided in the 4sm.def defaults textfile
  • try "gedit 4sm.def"
• -Z/MSL0.0/...   provides a tide height and datum
  • it is only used in the following event:
    • seatruth:      ?tide height is added to tide-corrected seatruth depth?
    • deliverable: tide height is subtracted when writing final deliverable SDB
      • ?in order to write a tide-corrected deliverable product

• the 4sm.def defaul text file is written upon the AutoCAL process
• its content may be manually edited as desired
• it is read at the start of reading the command line
  • any of its content may be overriden by the command line

• It is not possible to account for the many lacks of homogeneity in this scene
  • the lagoon water body is open to the ocean westwards
  • the lagoon also receives oceanic waters through the hoas
  • the reef flat and the reef slope should be modeled with specific calibration for very clear waters
  • over 50,000 persons live in the southern part of the atoll
    • this must have an impact on the attenuation properties of the waters at a very local scale
    • try to extract calibration data for some sub-areas along the southern part of the lagoon, see for yourself
• I have seen the South-West part of the lagoon affected by "milky" waters
• etc
• Deeper than 3-5 m, the green band is the only source of information
  • the only way to compute a depth is to assume a constant bottom type reflectance:
  • LBref[1]=150 in a range of 0-200 is assumed for of supposedlybright bottoms
  • LBref[2]= 40 in a range of 0-200 is assumed for areas of supposedly dark bottoms which are (or should be) masked at 236, like the outer reef flat and slope
• Still: a very nice opportunity for training
  • existing nautical maps show a maximum depth of ~16 m just north of the main pass at lowest tide
  • this shows that our combination of LBref[1]=150 and Kgreen=0.269 seems to "ballpark" the problem quite close!
• A 4-bands multispectral image certainly should be better
  • although the attenuation coefficients for Kblue~=0.28 and Kgreen~=0.27 are probably very close to each other for a OIII water type of Jerlov
  • this would yield a ratio Kblue/Kgreen very close to 1
  • if this would be the case, then the one-band case would be the only way to produce an estimate of water depth deeper than 3-5 m
  • it is our experience that the ratio Ki/Kj must be <0.8 for water column correction to be achievable using a multispectral image: Ki<<Kj