Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries

The Sea Truth Process

Your comments are invited                            See you on 4SM blog

return to Illustrations


Seatruth at Caicos Bank,   
Bahamas, Landsat TM
The tmnovdataset of the BILKO package  has excellent seatruth dataset from 0 to 25 m of water depth. 
Unfortunately, the 15-25 m depth range  is very poorly represented in this scene. This does not allow to investigate  the limits of the model to full satisfaction 
The results below were achieved  
through the application of 
      • K nir       =4.2 m -1
      • WL blue =472.0 nm
      • WL green=577.5 nm 
      • WL red    =649.5 nm
These results achieve   
a quasi-absolute optical calibration

A good calibration yields  
a fairly good seatruth regression 
  ZZregressor is yet another 4SM feature  
LEFT Lsw green is too low:   this shows as underestimated depths in the 10-20 m range ==> Lsw green   needs to be increased
RIGHT Lsw green has been increased by ~one DN : an offset of 0.73 m is now apparent ==> this is corrected through a Tide height variable

4SM has tools
to help the practitioner 
to conduct
his seatruth exercise
LEFT : a Tide Correction of 0.73 m brings the correlation in a diagonal position:
r 2   =0.81, rms =1.56 m (N   =408)

RIGHT Sea Truth Bottom Reflectance range:
the seatruth regression is limited to 125 pixels with an average bottom reflectance of 0-to-100 over the full range of 0-to-200. 
Any range may be specified; this is handy to 
* ascertain that the calibration of spectral K is correct: this is done using the bright bottoms only where thins are clearer.
* track errors in the determination of path radiance: this is done using the dark bottoms only where slight errors can have dramatic consequences.
Alternately , a SeaTruth Depth range may be specified instead, in order to study the seatruth regression within that range only.

SeaTruth at Negril Shores, Jamaica , Landsat TM  
Through minor adjustments of Lsw, La, WL,  
and using an available Sea Truth dataset, 
the practitioner can progressively improve his calibration.

ZZprofiler is yet another of 4SM features.
It compares computed depths with field depths
along depth sounding lines

Sea Truth
at Prince Edward Island,
Nova Scotia, Canada


  • Sea Truth data collection was conducted one full year after CASI data collection.
  • 4SM processing was conducted as a purely blind test, even before the sea truth data was collected, and attracted the interest of Bedford Institute of Oceanography.
  • This was achieved through the application of K[799]=4.56 m-1, which then leads to a CoefZ=1.33. This optical calibration is shown to have "ball parked" quite nicely:
  • Where depths have not changed to a major extent through severe winter storms in this highly dynamic environment of the St Laurence river estuary, computed depths are in excellent agreement with sea truth depths.
  • This shows that the spectral K values in m-1 adopted in the blind test had reached a "quasi-absolute" calibration, thanks to the precise positioning of CASI's wavelengths.
  • A value of Knir~4.4 m-1 seems to fit most cases.
  • The regression at left
  • is for all ten Depth_Sounding_Lines,
  • where there is clearly a problem
  • over less bright bottoms
  • in the 3.5-6 m depth range!
  • This raises the question
  • of the date of sea truth data acquisition.

Créer un site
Créer un site