Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
WorldView 2 image Cockburn Sound, Western Australia
June 2nd 2012, 11210*14500, 2 m ground resolution, 22.42 km * 29 km
"(c) DigitalGlobe, Inc. All Rights Reserved"
"Includes copyrighted material of DigitalGlobe, Inc., All Rights Reserved"
image courtesy of Gregory Miecznik, DigitalGlobe, under a Data Exchange agreement

the "PAN solution"
uses the PANCHRO band,
4x4 sum-aggregated and coregistered
bands 1, 2, 3, Panchro, 4, 5, 6, 7 and 8       
bands 1, 2, 3, 4, 5, 6, 7, 8 and 9
This new dataset from  DigitalGlobe is comprised of

MULTI is comprised of
12JUN02025337-M3XS_R01C1-052870327010_01_P001, tiles RO1C1 to R11C1

12JUN02025403-M3XS_R01C1-052870327010_01_P001, tiles R01C1 to R11C1

PANCHROMATIC is comprised of
12JUN02025337-P3XS_R01C1-052870327010_01_P001, tiles R01C1 to R11C1
12JUN02025403-P3XS_R01C1-052870327010_01_P001, tiles R01C1 to R11C1
work done in november 2013

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

From "The Geomorphology and Sediments of Cockburn Sound",

This image really is shit
(low light in june, bad S/N ratio, electronic noise, very dark bottoms),
but these results still might useful(?).
Anyway, there is much to learn by working on it.

Please refer to

The data
Optical calibration
Command Line
work done in november 2013

The data
I had to mosaic six tiles, not a simple task using just 4SM and OpenEV.

TCC, linear enhancement
Looks good!
Very dark, though, in spite of clear sky.
This is June, early winter in Perth.
End spring might be better for sun illumination.

TCC, logarithmic enhancement
Not any more!

Blue band

Purple band

Green band

PAN band
Note the ominous dark area north of Success Bank:
this smacks of negative bottom contrast
"Cockburn Sound has developed within an elongate depression that forms a margin between the Swan
Coastal Plain and the Rottnest Shelf, known as the Cockburn-Warnbro Depression (Fig. 2)"
All we can do in 4SM is try to save the day:
  • produce some results,
    • the deeper the bottom and/or the darker the bottom, the worst the accuracy and precision
  • and inform the end user that
    • the image is not worthy of shallow water work in 4SM.
The city of Perth spews
urban, riverine and industrial pollution
into the atmosphere and waters.

This very much looks like
a desperate case for 4SM shallow water work,
which uses a simplified radiative transfer equation.


PAN, histeq enhancement
  • the panchro band has its own vertical striping
  • upon deglinting by the NIR1 band, the amount of system noise increases dramatically
  • upon modeling by the "PAN solution", the result looks awfull

NIR1 830 nm, histeq enhancement
  • it has its own vertical striping
  • deglinting transfers this noise into the the Panchro band
  • upon modeling by the "PAN solution", the result looks awfull

PAN, smart-smoothed, histeq enhancement
  • the northern part is not deglinted, just smart-smoothed
    • upon modeling by the "deglinted_PAN solution", the result looks more acceptable
  • the southern part is deglinted_smart-smoothed
    • upon modeling by the "deglinted_PAN solution", the result looks awfull
Trade off, hands on
  • No need to deglint: lucky this image does not really need deglinting
    • apart from a very few small clouds
    • so that we can disable the deglinting
  • Need for the "PAN solution":  lucky we have the PAN band
    • because most bottoms are too deep/dark for the "Yellow solution" to be of any use
    • because the "Green solution" is quite happy to be off by +/-  several meters of retrieved depths over a dark bottom depending on its spectral bottom signature
  • Need for some field data: lucky we have the GA bathymetric map to refer to
    • so that we can fine-une the deep water radiance in the PAN band so as to achieve an acceptable fit on the retrieved depths with the most proeminent bathymetric features over dark/deep bottoms
  • Foolish: this WV2 take over Marmion Marine Park and Cockburn Sound  is a very hard case
    • why not use existing seatruth data for fine-tuning the optical calibration parameters?
    • this does not make sense
    • it shall take a long time before the radiometric quality of the imagery improves so that "no need for field data" becomes routine!!
    • until then:
      • either 4SM uses images of exceptional radiometric quality with high S/N ratio.
      • or access to existing field data allows the 4SM practioner to save the day.
My personal views
For shallow water work,
the specification of the future WV sensors
should concentrate
  • less on reducing the pixel footprint
  • more on reducing the system noise at the sensor
  • more on reducing the system noise at pre-processing stage: enrol a Fourier transform geek
  • more on suitable imaging conditions: near nadir viewing, maximize illumination
  • more on "deglintability" : no more of those time gaps
  • more on improving the PAN channel further
  • things keep improving
  • I shall need to go back to Marmion Marine Park, see how I can improve the results
  • that's if Simon resumes cooperation

Optical calibration

Calibration diagram
for Blue, Green, Yellow and Red bands
  • waters are distinctly less clear here
    • Kblue/Kgreen=0.75 at Perth
  • than in the same image at Marmion Marine Park
    • Kblue/Kgreen=0.68 at Marmion

Calibration diagram
for Blue, PAN, Yellow and Red bands
  • the BPL model for the pair Blue/Pan is extracted automatically from the image by 4SM.
  • it would appear that this yields acceptable results.