Syntax for the -Lsw and -dLsw arguments

back to Help

1 - NO NEED for field data, nor for atmospheric correction
2 - this is demonstrated in this website, using a variety of hyper/multi spectral data
Requirements are
1 - homogeneous water body and atmosphere
2 - some coverage of optically deep water
3 - some coverage of dry land
Problems are
1 - the precision on estimated depth is found wanting, because the noise-equivalent change in radiance  of accessible data is too high for shallow water column correction work 
2 - radiance data should be preprocessed by the provider at level 1 in order to improve S/N ratio
3 - exponential decay: the deeper/darker the bottom, the poorer the performances
I keep digging
until suitable data
become available
  • May be modified by a -dLsw/00.00/00.50/00/00 argument
  • Deep Water Radiance, may be expressed as a real value
  • If any Lsw value is greater than 255, all Lsw values are assumed to be expressed as 16U values: they shall be scaled accordingly

How to choose dLsw
  • Lsw is derived by use of the deglint regressions
    • the glint regressions must be estimated using a glint.shp shapefile
      • this shapefile must sample a wide range of glint intensity, from least glint to high glint
    • Lsw[c]=SGbias[c][cnir]+Lsw[cnir]*SGslope[c][cnir]
  • So depends very much on the adopted deep water radiance Lsw[cnir]
    • which is calculated as the average of the lowest radiances inside a cloud shadow ROI.
  • So Lsw.shp and dLsw.shp shapefiles
    • may be identical over an homogeneous scene: dLsw may be negligeable
    • may be separate if the glint regressions are estimated well away from the area of interest: dLsw may be quite significative , ib order to be representative of the area of interest
  • In case there are dark cloud shadows:
    • Lsw[cnir] is fairly easy to estimate as an average for a ROI located under darkest cloud shadows:
    • this is where both Lsw.shp and dLsw.shp are to be located,
    • and dLsw shall turn out to be ~negligeable.
  • In case there are no dark cloud shadows,
    • one must choose a ROI for Lsw.shp, to be located where there is least glint.
    • Lsw[cnir] is calculated as the average of the lowest radiances inside that least glint ROI.
  • If the above does not suit a particular area of interest
    • as was the case at San Lorenzo Channel,
    • glint regressions and spectral Lsw are estimated over open waters,
    • and a dLsw.shp shapefile may be located so as to be representative of the particular area of interest, like the shoals  of San Lorenzo Channel.


How to Choose Lsw

  • Spectral Lsw is to be set at a value which is compatible with average readings and the level of noise over optically deep areas of the image
  • The -Lsw... argument may be complemented by a -dLsw.. argument
  • In practice, and because of the system noise, it is customary to look at the histogram of radiances over such area, and choose a value which is distinctly higher than the average
  • Note that -dLsw and -dLa  arguments are available:
    • Lsw=Lsw+dLsw+dLa
    • La=La+dLa+dLsw
    • La= MIN(La,Lsw)
Choosing Lsw for Landsat TM image tmnov.pix
three sources of evidence
your duty is to make up your mind and come to an operational trade-off
That's assuming
  • homogeneous illumination conditions
  • homogeneous optical water properties
  • homogeneous level of noise...
I have come to the conclusion that 
  • Lsw may contain some glint as long as it is not modulated by the swell
  • this shall not entail any adverse artifacts upon modeling
Lsw from graphic evidence upon calibration for tmnov tutorial
  • Quantization: 
    • an integer value of 14 represents the range of real values from 13.5 to 14.5
    • therefore, for a reading of 14 in the image, we may specify up to Lsw=14.5, subject to other pieces of evidence
  • Bending: one such piece of evidence is that, under the BPL assumption,  the linearized BPL pixels should display as a straight line: 
    • bending up      signals a Lsw that is too high:
      • see below the green series using Lsw[green]=15.5
    • bending down signals a Lsw that is too  low:
      • see below the     red  series using Lsw[   red]=13.5
    • using Lsw[green]=14.5 yields the best rectilinear fit of the six deepest BPL pixel
  • Seatruth on image tmnov.pix appears to confirm this approach
Lsw from investigating a slightly smoothed image
  • Search your image thoroughly for deep water radiance: 
    • below is a display of the image, slightly smoothed by a 3*3 circular window
      • -Prepare/scale/delimb/fill/Smooth/linearize
    • it illustrates the difficulty of choosing a deep water radiance
      • no wonder some practitioners prefer to evade the problem altogether!

Lsw from the glint regression

If a -deglint... parameter is specified in the commandline, look at the glint textfile
-Lsw/53.6/14.0/09.6/04.5 - - - - - - - -
-Lsw/53.7/14.1/09.7/04.6 - - - - - - - -
-Lsw/53.9/14.2/09.8/04.7 - - - - - - - -
-Lsw/54.1/14.3/09.9/04.8 - - - - - - - -
-Lsw/54.2/14.4/10.0/04.9 - - - - - - - -
-Lsw/54.4/14.5/10.1/05.0 - - - - - - - -
-Lsw/54.6/14.6/10.3/05.1 - - - - - - - -
-Lsw/54.7/14.6/10.4/05.2 - - - - - - - -
-Lsw/54.9/14.7/10.5/05.3 - - - - - - - -
-Lsw/55.1/14.8/10.6/05.4 - - - - - - - -
-Lsw/55.2/14.9/10.7/05.5 - - - - - - - -
-Lsw/55.4/15.0/10.8/05.6 - - - - - - - -
-Lsw/55.6/15.1/10.9/05.7 - - - - - - - -
if/when you've decided on a suitable value for LswNIR
  • by specifying -Lsw/..../05.0 in the command line, 
    • then 4SM shall refer to the glint regression
    • to specify all other Lsw values: -Lsw/54.4/14.5/10.1/05.0 
    • and use them to override whatever values specified in the commandline
  • optional -dLsw and -dLa arguments still apply