Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
 
  
WorldView 2 image Marmion Marine Park, Western Australia
June 2nd 2012, 2875*4096, 2 m ground resolution, 5.7 km * 8.2 km,   courtesy of Simon Allen, CSIRO
       "(c) DigitalGlobe, Inc. All Rights Reserved"
"Includes copyrighted material of DigitalGlobe, Inc., All Rights Reserved"
Watch DigitalGlobe's bathymtery webinar 2013

back to study cases

Marmion in Simon's image below
Marmion in Perth twin stripes mosaic

This dataset, courtesy of Simon Allen, is comprised of
  • tiles R1C1 and R2C1 12JUN02025339-M2AS_R1C1-052749202010_01_P002 Multi
  • tiles R1C1 and R2C1 12JUN02025339-P2AS_R1C1-052749202010_01_P002 Panchro 0.5 m resolution


 
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
 
 






bands 1, 2, 3, Panchro, 4, 5, 6, 7 and 8        
become       
bands 1, 2, 3, 4, 5, 6, 7, 8 and 9
Panchromatic band was not used here



Comments
The data
Optical calibration

Facts
Sewage outfall

System noise and pollution
Results 
CommandLine
 

Nov 23rd 2012
Comments

Simon, I beg for indulgence,

as I lack the deep water coverage
which is required for estimating deep water radiances.

When I get some seatruth data, then I can use it to ascertain blue-green deep water radiances.

The deeper/darker the bottom, the more hazardous the results.
The least bands usable, 
the more hazardous the results.

No tide correction applied.
Smart-smoothing applied.

FinalZ = CoefZ * ComputedZ - HTide
CoefZ and Htide to be derived from seatruth regression

  • I used only the MULTI bands, so far
  • I used all the experience gained since november 2011 when I pocessed a Geraldton W2 image: this involves
    •  estimating the spectral deep water radiances Lsw for the purple, blue and green bands
    • estimating the spectral water volume reflectances Lw for the purple, blue, green and yellow bands
    • estimating operational wavelenghts for the yellow and red bands
  • This was made possible thanks to the absence of adverse surface glint 
    • and lso needed masking all  those countless alien objects that populate the image: boats and wakes, all sorts of objects at the sea surface in this Marine Park
  • The nice radiometic quality of the image allowed me to operate the smart smoothing scheme
    • It shall be observed that the smart smoothing does a good job of smoothing homogeneous features while respecting their sharp limits

Existing depth data

  • As a general rule, keeping secret the existing depth data does not serve any purpose
  • By making it available along with the image data, the end user can save the practioner a lot of time and anguish, particularly when the image is a hard case
 

No optically deep waters

  • The lack of optically deep waters is a real pain for operating the simplified RTE

    • as it relies on the estimation of the so-called "bottom contrast" LsB-Lsw, or LB-Lw, where the deep water radiance takes a dominant role over darker or deeper bottoms

    • TOA : Ls=Lsw + (LsB-Lsw)/exp(K*Z)           or BOA : L=Lw + (LB-Lw)/exp(K*Z)

  • I TUNED THE CALIBRATION SO AS TO OBTAIN COHERENT RESULTS

    • Deep water radiances Lsw in the blue-green range are pure fantasy

  • So it would be required that the dataset extends farther seaward so as to reach areas where the water is optically deep in the blue-green range, where to ascertain the value of Lsw for these bands (and therefore also the value of Lw)

  • It would also be desirable that the dataset extends a bit further inland, so that I can get a more reliable Soil Line

 

Sewage outfall

Failed

  • Panchro band : up to now, I failed to obtain consistent results

Dark vs Bright

  • The big question is: very dark features, probably rock constructs, would be expected to raise a bit higher than surrounding sedimentary areas.

    • I feel uncomfortable that those dark  features are mapped ~2 m deeper than surrounding bright features 

Aliens

  • Thorough masking of those aliens yields a more secure and much less clear water type

    • This takes an awful lot of time

    • Sorry for that: very high ground resolution in a popular marine park requires a lot of learning

  • All manners of radiometric aliens yield fancy depth results: beware of that when running a seatruth regression!!

    • this includes fancy pixels all around the 4x4 sum-aggregated Panchro image

    • I masked them out of the calibration process, but they do appear in the modeling results, lest I would punch holes to weed them out, like they were clouds

    • you may use the value of 1 in this mask

Altogether
I can get NOOO sa-tis-fac-tion!

  • A frustrating experience, like at Geraldton.

  • After days of battling, I give up: the results are hazardous, to say the least; although the southern part of the image might be less of a problem because the pollution does not seem to affect it.

  • The data is not at its best:

    • system noise

    • pollution: water quality is NOT homogeneous

    • lack of optically deep water coverage

    • lack of land coverage

  • The site is a difficult one:

    • most of the shallow area is deeper than 6 m: the Yellow band is of little help over those very dark bottoms, even more as it needs a high threshold because it is affected by pollution.

    • most bottom substrates are very dark, even the "bright" one, apart from a few small areas.

    • the topography is quite hard to guess: dark vs bright. Seems we have a dark rocky platform upon which stand hydraulic sandwaves a few meters higher.

  • Essential calibration parameters cannot be estimated to satisfaction:

    • deep water radiances Lsw for the Purple, Blue and Green bands: I need optically deep water coverage.

    • atmospheric path radiance La: La=Lsw-Lw : I need more land coverage.


 










Nov 23rd 2012

The data


mosaic FCC bands 7, 3 and 2


mosaic Panchro

 



Purple band 1 at 427 nm

 



Blue band 2 at 477 nm



Green band 3 at 546 nm


Yellow band 5 at 596 nm

  • In the clearest waters, Yellow does not show bottom detection in excess of ~12 m of depth over very bright bottoms.

  • So we can infer that "bright" bottoms here are quite deep in most of the image: this is a precious indication, and I used it when picking calibration details.


Red band 6 at 641 nm
this band is rotten
things do happen!


NIR1 band 8 at 831 nm
No glint
No adjacency effect






Nov 23rd 2012
 Optical calibration 


Calibration diagram for bands 2, 3, 5 and 6
Waters are quite clear: ~OIB water type of Jerlov


Calibration diagram for bands 1, 3, 5 and 8