Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
 
A 2138*1862 IKONOS 4 bands image of Lee Stocking island, Bahamas
Work updated in 2005

4SM licensekey is 346 Euros for this image
Note that this cost could be lowered by masking out deep water areas

Wavelengths at mid-response curve are 480, 551, 665 and 805 nm;
but WL[2]=565 nm and WL[3]=640 nm here for sake of consistency of optical calibration.

A very nice image indeed
Please refer to "Optical remote sensing of benthic habitats and bathymetry in coastal environments
at Lee Stocking Island, Bahamas: a comparative study
"
by Louchard Eric, Reid Pamela, Stephens Carol, Davis Curtiss, Leathers Robert and Downes Valerie,
Limnology and Oceanography, 2003, vol. 48, pp511-521
see also Landsat TM image
see also WV2 image
see also HICO 
see also Landsat 8 OLIP

home


 
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
 
So
I keep digging
until suitable data
become available
 
 

 

Some sea truth on IKONOS image of Lee Stocking Island, Bahamas
by Serge Andrefouet, U. of South Florida (2003), 
now at IRD in Noumea, New Caledonia

The sea truth below was obtained from S. Andrefouet on Feb 27th 2003
i. e. several months after "totally blind processing in sept 2002" by Y. Morel 
of a "deglinted" 3-bands BGR image
Wavelengths used were 495, 550 and 675 nm

 
 
Only for oceanic side GCPs
(similar estimation in progress for channels and lagoon sides)
Depth in meters , range 0-25 m
X=bathymetry model
Y=field data

Y=1.0752X + 0.9408 (R 2 =0.8845)
"Trend: underestimation of the depth"

 
 
As for the bottom types, there are some variations,
but the trend is the same, which is good
." 

more seatruth

 




Presentation below for work done in 2005
please refer to deglinting at Lee Stocking Island by Hochberg et al 2003

 
Raw image TCC
 histeq enhancement
 
 
Zoom : Raw image TCC
histeq enhancement
 
 
Deglinted image TCC
Same histeq enhancement as above.
  •  A fairly strong level of baseline skyglint has been removed,
  • as well as the swell modulated glint

Deglinted image, red band_3
  • Further histeq enhancement reveals some problems with radiometric quality,  
    • particularly in the Red band[3]
  • which will warrant a fairly high Lm[3] threshold value
Because this image is already geocoded, 
there is no  simple way to correct
for this very bad feature.

 
 
Before Automatic Deglinting
linear stretch enhancement
"good glint makes for good deglinting prospect!
 
After Automatic Deglinting
enhancement unchanged
"good glint makes for good deglinting performances!
 
After Automatic Deglinting
Blue band, histeq enhancement
"good glint makes for good deglinting performances!
  
After Automatic Deglinting
Green band, histeq enhancement
"good glint makes for good deglinting performances!

Deglinted image TCC
  • Deep water areas are not smart-smoothed
  • Further histeq enhancement reveals some problems with radiometric quality,
    • particularly in the Red band[3]
    • which will warrant a fairly high Lm[3] threshold value
  • Some scattered clouds resisted deglinting
  • Scattered cloud shadows remain anyway
 
Zoom : Deglinted image TCC
  • Deep water areas are not smart-smoothed
  • Further histeq enhancement reveals some problems with radiometric quality,
    • particularly in the Red band[3]
  • which will warrant a fairly high Lm[3] threshold value
 
Deglinted blue band
Note the prominent dark bed in the blue band
along the outer slope...
 
 
 
Deglinted green band
....which has no equivalent in the green band:
this is a likely candidate for a dense seagrass bed
 
 
 
image_Z 
the likely "dense seagrass bed" along the outer slope
apears to be distinctly shallower  
image_Z b&w
 
 
 
image B
 
 

 image LBS 
  • this is a "low-tide", or "water column corrected", RGB true color composite of the shallow bottom reflectance over the scene (linked to a false color composite)
  • in this view, each channel is scaled linearly over the range 0-250 for both dryland and shallow areas:
    •     0  for the BOA reflectance of a black bod
    • 250 for the BOA reflectance of the brightest shallow bottom type that exists over the scene
  • the radiometric content of the bottom reflectance signature is not readily "visible" in this view 

 image WZ
red:     areas modeled using bands 1, 2 and 3  
green: areas modeled using bands 1 and 2 only  

As the NIR band was used for deglinting, 
it was not available for modeling of very shallow areas 

 

image WR
The color displays the diameter of the circular smart-smoothing window
  • from Green (3 pixels for very shallow areas), Blue (5), Yellow (7) , Orange (9), 
  • through to Pink (11 pixels for deepest areas)  

Optical calibration
Because of the heavy glint, and quite distinctly here because clean bright  sandy detritic bottoms would appear to be scarce in this environment (biofilms, seagrass, algae and oolit formation all involve photosynthetic acitvity at the sea bottom surface), the calibration of optical properties is requires some experience.  

Optical calibration
Add to it that we do not have any seatruthed experience of the proper way to set operational wavelengths for the 4 wavebands: this means that a significant CoefZ factor might be needed in order to adjust computed depths to seatruth reality. 
 
 
 Optical calibration
Based on in-situ data, Louchard et al. state that "the  IOPs  were assumed to be uniform over the study site".
  • But we found clear signs of significant variations of the operational attenuation coefficients at various locations in the image:
  • In particular, less clear waters are associated with the crest of the oolitic sand wave.
  • This does not help to secure the Brightesst Pixels Line. 
 
 
 
The normalization
of spectral bottom reflectances
This is to illustrate the process of normalization
of the computed bottom reflectance signature in 4SM 

image LBS-both
  • upper: image LBS true color composite, a "low tide view"
  • lower : image LBS normalized: the radiometric contrast  of the computed bottom reflectance signature, irrespective of its average value,  increases 
    • from quasi-null in grey shades
    • to very strong green shades for a greenish bottom  like a seagrass meadow, 
    • or to blue-brown shades for a blueish bottom like some coral beds or red-brown algae (none in this scene apparently?)  

 image LBS normalized
 the radiometric contrast
of the bottom reflectance signature
is dramatically enhanced
by this final optional process