Bathymetry and water column correction
LANDSAT 8 OLI 
 of Abu Dhabi, UAE
images courtesy of USGS

This is the Marawaah island study case
see also Shuwayat island 


 
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
 
home
Tide height

Go to depth results october 2013


Tide height

April 20th NIR band
Highest tide in the NIR band: 1.2 m to Chart Datum

July 9th NIR band
Lowest tide in the NIR band: 0.8 m to Chart Datum
May 22nd, June 7th and July 23rd :
NIR shows decreasing tide height
between these two extremes
 



 
  


4SM satellite derived bathymetry (SDB)
 
 



Depth results october 2013  using the full scene
Hastily
Last August, the first 5 scenes had been processed hastily.

For this new exercise,
I worked all 9 scenes: all calibration parameters have been estimated aknew and more rigorously.
Hopefully, all scenes have been processed in a consistent way.

 
No-gap
All results using the PAN solution, which yields a "no-gap" depth result.
That is:  the NIR solution, the Red solution and the Green solution were disabled altogether.
Wavelengths
  • I consistently used the following:
    • all bands at -or close to- mid-wavelength.
    • except the RED band which consistently requires to be set at ~610 nm.
    • except the PAN band which does not require any wavelength  reference.
  • So as to ensure strict match of the BPL model with BPL pixels, which is absolutely required.
Calibration of the PAN band
Proper results depend on proper and very detailed calibration of the PAN solution.
To a degree that I did not suspect.
All this was done manually: I learned a lot, under operational pressure and time constraint.
After all those manual calibrations, I now have an idea for automatically and finely calibrating the PAN band.
But then it needs to be implemented...
Lw
In the course of this UAE exercise,
I think I have identified a more secure and practical way
of estimating the all-important water volume reflectance Lw.
I need a way to validate how I estimate water volume reflectance Lw=Lsw-La
this is all-important for very dark pixels.
Surrounding of Marawaah Island
The quality of the result very much depends on the specific area
  • The surrounding of Abu Dhabi city and the whole Khor Al Bazim quite surely have less clear waters: this is not accounted for.
  • Lousy deglinting can only result in lousy results.
I focused on your stated project: the surrounding of Marawaah Island.
COMPLEX
  • With all those industrial activities, oil riggs and terminals, boats and tankers everywhere, industrial smoke, and heavy load of aerosols, even clouds at times, this scene is daunting.
  • Add to it the electronic noise in the SWIR band, which transfers into other bands upon deglinting.
  • I'd like to know how one is supposed to achieve a formal atmospheric correction of such images.
The PAN solution procedure is still very new.
I keep discovering how to handle it practically.
But I predict it shall become VERY popular, with WV2 as with OLI,
to the extent that its calibration can be refined and hard-coded.
By adding new scenes to this sort of exercise,
and with your fine post-processing,
maybe we can work out a  scene-wise commercial product
  • that eventually shall use pan-sharpened images (if that possibly works).
  • that you would promote for sale.
  • that would become a very usefull depth reference for processing higher resolution images like WV2 for detailed local AOIs.
One key to such project is strict co-registration: your call.
Smoothing?
Smothing or NotSmoothing, that is the question.

I had rather smooth, but it is your call.
Adding the Green solution or not?
It would extend the depth range,
but would exhibit gaps which might be severe locally, depending on the actual bottom signature.
As for enabling the Red solution, I think a well calibrated PAN solution is just as good.
This need to be tested though: your call, but mind the gaps.
BOA reflectance (0-1)
  • I am now equiped in 4SM to output BOA reflectances (range 0-1) for OLI images, thanks to USGS's webpage "Using the USGS Landsat 8 Product"
    • is there anything similar for WV2?
  • Ra: I write out the adopted spectral path radiance Ra into reflectance (0-1): would want to compare with path radiance obtained by formal atmospheric correction processes.
  • Rw: I write out adopted spectral water volume BOA reflectance Rw (0-1): would like to compare with  Hydrolight estimates (but Dekker stays put). 
  • RM: I write out adopted BOA reflectance of the brightest shallow bottom RM: I should normalize the BOA water column corrected reflectancesto a refrence reflectance spectrum of -say- coral sand/mud.
  • see that RM, Rw and Ra are displayed in calibration diagram
 
marawaah20130420F_depth4.pix Purple, Blue and Green against PAN
no smoothing
marawaah20130522F_depth4.pix Purple, Blue and Green against PAN
no smoothing
marawaah20130607F_depth4.pix Purple, Blue and Green against PAN
no smoothing, homogeneous atmosphere, excellent deglinting
THIS POSSIBLY IS THE BEST IMAGE
certainly the best calibration of the PAN solution, with the least compromise or "spying" on the nautical chart
FinalZ = CoefZ*Z + HtideToDatum
marawaah20130623F_depth4.pix Purple, Blue and Green against PAN
no smoothing
marawaah20130709F_depth4.pix Purple, Blue and Green against PAN
no smoothing
marawaah20130725F_depth4.pix Purple, Blue and Green against PAN
Smoothed
CAUTION: swirls of turbid waters in the channel north of Marawaah island
marawaah20130810................. can't get suitable glint regressions: no results
marawaah20130826_depth4.pix Purple, Blue and Green against PAN
No smoothing; I see that this might be a problem: result is bumpy, and there are deep pixel artifacts, but you sure have your ways at cleaning/smoothing!
marawaah20130911_depth4.pix Purple, Blue and Green against PAN
No smoothing
I have a problem of strict co-registration of the 2*2 sum-aggregated PAN channel: this produces deep pixel artifacts such as at pixel 2702/5030 in marawaah20130911