Tarawa subset : First Modeling
based on the calibration obtained  using pixels masked at mSE=3 only
Things we have learned so far
From Profiling
  • -dLsw000.0/000.0/0-1.0
  • Lswgreen=25.9   
  • Lswred=13.9   
  • LswNIR=5.5, to be confirmed by Ghost pixels
From calibration under mask_4
  • -CP/159.75/094.53/013.56_0.75m
  • -KK0.29/Knir3.5...    water type OII inside the lagoon
  • -LsM/180.7/169.2/156.3
  • LswNIR=5.5 is confirmed by Ghost pixels
  • -Lw/0001.5/000.0/000.0

Run the Modeler for Line_5
Activity : use LINE_5
Run the Modeler
First Modeling 
  • based on the calibration obtained under Mask_4 only:
  • please, append a copy of LINE_5 to your script
  • then do as directed, step by step
After the optical calibration, we are satisfied that
  • the deglinting scenario seems to be  OK
  • we understand that LswNIR must be set at ~=5.5
    • so that approximately three DNs of sky glint shall be removed from the NIR band
    • please note that , unlike sun glint, sky glint is not modulated by waves
  • a good fit has been achieved between the BPL pixels and the ratio KI/KJ=0.29
    • KNIR~=3.5 m-1 is in good agreement with Kred~=0.634 m-1 and Kgreen~=0.184 m-1 
  • the Z2 vs Z3 plot is quite good:
    • the NIR_solution=Z3 and the Red_solution=Z2 are in good agreement over bright substrates
  • therefore, the calibration of the simplified shallow water optical model
    • under mask=4 is now complete
we may now proceed to run the modeling tool 
About 4SM license
  • By default, 4SM writes chequered results upon deglinting or modeling, unless
    • the image is verified to be licensed
    • or it is a tutorial
  • Notice   -Process/nosu in Line_5 command line:
    • this is to preclude any priviledge
  • Enable the -LK argument, and run the script
    • this wrtites and displays a  tarawa-subset.lic text file
    • please, review text file  tarawa-subset.lic:
      • "4SM.8.07_Required_LicenseKey for file tarawa-subset.pix is
    • this license key is already included in 4SM code for free tutorials
  • Remove or disable the @-LK argument
See here for details
 Color tables  in OpenEV
dir ./4sm.calls/pct
  •  pctmSE.leg .........is the pseudocolor table to display mask mSE...............in channel bands+1=4
  • pctWZ.leg.............is the pseudocolor table to display image_WZ..............in channel bands+2=5
  • pctWR.leg.............is the pseudocolor table to display image_WR..............in channel bands+3=6
  • pctZ250.leg..........is the pseudocolor table to display image_Z250....... ....in channel_7
  • pctB.leg................is the pseudocolor table to display image_B..................in channel_9
  • pctZG.leg..............is the pseudocolor table to display image_ZG in  cm... ..in channel_28
mask_4 is enabled
mask_4 is red
  • ONLY pixels masked at 4 are processed
    • other pixels are just left unchanged
    • this is because the mask_4 option is enabled
Bottom brightness B
  • The average water column corrected bottom brightness
    • is written to image_B in channel_9 
    • coded on a scale 0-200  
  • It is shown below left; see legend for image_B
    • this display shows very dark hues where bottom is deepest
    • this is unlikely and unacceptable in this environment:
    • it can be countered by any of the following  tricks:
      • increase Lm[red] or Lsw[red]
      • increase Lsw for the red band
      • increase the threshold Bmin in -B/... argument
Lm[2] can be increased: try that now
  • -Lm/0001.0/002.0/255.0 
  • -Lm/0001.0/004.0/255.0  ...
  • increasing this threshold Lm[red] suppresses less acceptable results, and has no efefct on retrieved depth
  • this is what we want

image B in channel_9, low threshold Lm

 image B in channel_9, high threshold Lm
Lsw and dLsw
Lsw[2] can be increased: Lsw=Lsw+dLsw  
try that now  
  • -dLsw001.0/000.2/000.0
  • -dLsw001.0/000.6/000.0
  • -dLsw001.0/001.6/000.0
  • -dLsw001.0/002.6/000.0
  • -dLsw001.0/003.6/000.0  ...
  • increasing Lsw[red] increases both retrieved depth and bottom brightness accordingly
    • and vice-versa
  • do we want that?
    • rather a dubious trick!
  • BminN is used to exclude image brightness less than Bmin=N to appear in the results:
  • For example, if a pixel is modeled by the Red solution and yields B<Bmin, then
    • it is processed by the next solution (Green solution) if possible
    • or it is mapped as deep water
  • By doing this, the practitioner assumes that a brightness less than 10 does not exist in this scene, and therefore may not be accepted

Below are some explanations
-M@0001/0002/0003    controls the solutions
  • -M@0001/0002/0003...............prevents the One-band solution
  • -M00001/0002/0003...............enables the One-band solution
  • -M@0001/0002/0003..............enables the Red solution
  • -M@0001/0002/0003..............enables the NIR solution
 Lm   -Lm/0001.0/001.0/255.0
threshold value for bottom contrast Ls-Lsw 
  • Lm thresholds are scaled on a range of normalized radiance  0-200
  • if no -Lm... argument is provided, defaults are -Lm/001/001/255
  • for the tarawa-subset tutorial
    • Lmnir=255: the NIR solution is prevented here
    • Lmred=1: see for yourself that Lmred=1 gives a chance to noisy/foul low radiances in the red band
      • to produce conspicuous dark artifacts in bottom reflectance
    • find your own prefered -Lm... values :
      • we use Lmred=4
Syntax dLsw
-LL... argument

Line_message on screen upon modeling
  • -LL10 causes a message on your session screen for one in ten lines processed
    • this is the default
  • -LL100 causes a message on your session screen for one in a hundred lines processed
  • -LL1 causes a message on your session screen for each and every line processed
  • etc
mask_SE   in -Model... argument
  • -Model/mask_4
    • mask_4 is disabled: therefore all shallow pixels shall be processed
  • -Model/Mask_4
    • Mask_4 is enabled: therefore only shallow pixels which exhibit mSE=4 shall be processed
    • other pixels are left unchanged
 Smooth or NoSmooth
  • try your own -Smooth... argument
  • review radius of circular smoothing kernel in image WR (in channel_6
 is done in 0.033 minutes on my machine
is done in 0.050 minutes on my machine

One-band case   -M@00001/00002/00003
optically deep? 
  • see that no modeling is performed if Lsred<Lmred: such pixels are treated as optically deep
    • -Lm/0001.0/001.0/255.0
  • then most of the lagoon waters are considered optically deep, whereas there is valuable information in the green band
    • we can decrease Lmred to 0.5, or even to 0.2
    • beware that this opens the door to the fox: system noise shall put out a nice show
    • NOT a good solution!
image_B in channel_9: average bottom brightness 
we want to assume that the lagoon bottom is made of bright/clean coral sand/mud
Please change -B/tclNe/Bmin0/cLM1.0     
into      -B/tclNe5.00/LBref200_100/Bmin0/cLM1
  • if this is right, then we may apply the one-band case using the green band only
  • but a "brightness" must be assigned to those deeper lagoon bottoms
    • LBref[1]=200 on a scale from 0 to 200 in normalized BOA radiance
    • we shall use a value of LBref[1]=170 
to enable the "one-band case",
  • you must change from -M@000001/00002/00003       
    • to        -M000001/00002/00003
one-band case depths are computed as
Z=(log(LBref[1]-Lwgreen)  -  log(Lgreen-Lwgreen))   /  Kgreen
try your own prefered value for LBref[1]

  • First run using
  • Then run using
  • Then run using
  • Then run using
  • See how -Lm001.00/04.00/005.0 helped a lot to prevent lousy/unexpected dark bottoms
    • try -Lm001.00/02.50/005.0, might be sufficient to counter some turbidity that seems to affect the water at depth in the red band
  • Progressive Lm
    • see that -Lm001.00/04.0p/005.0 does a better job
    • try other values; make up your own mind: it's all there for you to decide by yourself
  • -Model/Mask_3
    • causes modeling/updating only of shallow pixels which read 3 in mask mSE
    • for other pixels, 4SM does not change anything
    • this allows the practionner to apply specific modeling conditions to areas that deserve them
  • -M0000001/00002/00003
    • see that the one-band case has been applied
  • finally, disable the mask option: -Model/mask_3
 Also note that some very shallow areas under Mask=3
are saturated to B=201 and colored in deep blue.
Why is that? Something weird happens there.
 We need to know!
The above modeling applies to the calibration
obtained inside the lagoon under mask=4.
We need to know whether these conditions apply to the whole image subset.
We are in for a big surprise!

Modeling over the reef rim or the outer reef slope 
would require specific parameters,
as the water there is cristal clear! 

Tarawa atoll is the home to a population of 56,000 persons,
so the lagoon waters are under high anthropogenic pressure.


Next to Calibration under mask_3

Créer un site
Créer un site