Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
  
WorldView 2 image Waimanalo Beach, Oahu, Hawaii islands
March 31th 2011, 3289x3241, 2 m ground resolution, 6.5 km * 6.5 km,   courtesy of Ron Abileah
please refer to Mancini, Olsen, Lee and Abileah 2012
home
This data has been deglinted
by Ron Abileah

 


 
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
 
Bottom Typing
Getting started in bottom typing
See final bottom typing results 

Antoine Collin's Live coral colonies index NDR with WV2 images
Using the LIDAR DTM to investigate and to improve 4SM



July 12th
Bottom Typing
waimanalo_deg_3_bottom_type11waimanalo_deg_3_bottom_type12waimanalo_deg_3_bottom_type13waimanalo_deg_3_bottom_type14waimanalo_deg_3_bottom_type15waimanalo_deg_3_bottom_type16waimanalo_deg_3_bottom_type17
  • type 10 is a mixture, should be deleted, shall mostly map as type 11

  • type 11 is clean

  • type 12 is clean

  • type 13 is clean

  • type 14 is clean, should be refined in favour of wZ=5 instead of wZ=4

  • type 15 is clean, should be refined in favour of wZ=5 instead of wZ=4

  • type 16 is clean

  • type 17 needs close attention

 

What's important for bottom typing?
  • NO: precision on computed depths
    • they don't even need to be calibrated in meters
  • NO: precision on water column corrected bottom type signatures
 
  • YES: consistent set of Ki/Kj ratios among all bands
  • YES: consistent water column correction over the whole shallow area
    • provided all water column corrected bottom signatures are twisted the same way, the potential for bottom typing is not affected
    • this means that fully fledged atmospheric correction is a rich man's luxury
    • this means that "spectral libraries" are a rich man's luxury
    • this means: it's OK to work with uncorrected and uncalibrated DNs
      • but best possible deglinting is paramount
 

What should we be wary of for bottom typing ?
 
  • poor deglinting performance
  • system noise:
    • this processing needs maximum smoothing: by a 11 pixels diameter circular window
    • each marine pixel here is a smart average of the 81 pixels which surround it: that's a ~20 m ground resolution
  • heterogeneous waters
    • each water body requires its own set of calibration parameters
 


   
July 25th

Bottom Typing is an art: no ISODATA here!
 


ZC vs ZR regression by 4SM method

ZZRegressor: Statistics of seatruth ZC-ZR
on image waimanalowv2m_deg at Oahu, Hawaii
N=5.589 millions pixels by Morel's 4SM method
HTide=1.00 Smooth=1_5 Using_bands_1_2_3_4_5 cZ=1.00

 1.89% pixels with depth underestimated by more than 5.0 m
4.26% pixels with depth underestimated by more than 3.0 m

6.31% pixels with depth underestimated by more than 2.0 m
15.18% pixels with depth underestimated by more than 1.0 m
31.12% pixels with depth underestimated by more than 0.0 m
23.91% pixels with depth  overestimated by less than 1.0 m
10.80% pixels with depth  overestimated by less than 2.0 m
4.56% pixels with depth  overestimated by less than 3.0 m
1.61% pixels with depth  overestimated by less than 5.0 m
0.37% pixels with depth  overestimated by more than 5.0 m

55.03% of computed depths are within  +-1.0 m of DTM depth
81.00% of computed depths are within  +-2.0 m of DTM depth
91.86% of computed depths are within  +-3.0 m of DTM depth
97.74% of computed depths are within  +-5.0 m of DTM depth
over a total of 100.00% of computed depths












Final water column corrected image   
some locations of AOI targets are shown


July 24th: Final classified image: 15 classes
 

Bottom Typing is an art: no ISODATA here!

 


Bottom types 10 to 14 : possibly sandy areas


Bottom types 15 to 18 : possibly hard substrates with scattered corals
 
 
Bottom types 31 to 35 : probably dense coral constructions

 

Bottom Typing is an art: no ISODATA here!
  • Binning classes into super-classes might be advisable

    • although I think there is room for smarter discrimination, making better use of the behaviour of band 1

  • I had to adapt to WV2's bandwidth. I expect further progress in 4SM for bottom typing of WV2 images

    • WV2 obviously has a nice capabality for discrimination

    • TOO BAD I HAD TO APPLY SUCH A STRONG SMOOTHING on this badly glinted image,

    • which nullifies the claim benefits of very high spatial resolution

STATISTICS OF ALL SHALLOW BOTTOM TYPING: BOA radiance 0-200 for waimanalowv2m_deg WV02 image
 CoefSIG=1.00 nbMISS=0 sizeXY=9248668 -E1/3289/430/3241  on Tue Jul 24 17:53:21 2012
 SIGnbMax=35
            1460484 pixels in Type_0:   15.791% Land area
                  0 pixels in Type_1:    0.000% Unclassified Shallow bottoms
            1697202 pixels in Type_2:   18.351% Optically Deep water
                183 pixels in Type_3:    0.002% No Data area
                  0 pixels in Type_4:    0.000% Wave Breaker area
                  0 pixels in Type_5:    0.000% Cloud/Shadow area
            13667260 pixels in Type>5:  147.775% Classified Shallow bottoms

4559655 pixels  49.301% in type_15 B= 67.5
2076937 pixels  22.457% in type_17 B= 56.3
1775961 pixels  19.202% in type_14 B= 94.8
1630931 pixels  17.634% in type_13 B=101.5
1034929 pixels  11.190% in type_18 B= 50.8
 691391 pixels   7.476% in type_31 B= 46.4
 645851 pixels   6.983% in type_11 B=124.5
 331064 pixels   3.580% in type_33 B= 39.7
 265349 pixels   2.869% in type_10 B=151.7
 215178 pixels   2.327% in type_34 B= 35.6
 176448 pixels   1.908% in type_12 B=114.7
 135054 pixels   1.460% in type_35 B= 33.6
 128512 pixels   1.390% in type_32 B= 37.6

July 25th

this run combines bottom typing with water column correction "on the fly"
 
  • This allows to optimize the computed depth according to the estimated bottom type at each pixel
    • this is akin to a Spectral Library approach 
  • This tends to add undesirable noise to the depth image,
    • which, for now, makes it unsuitable for display
  • But it solves some of the major inconsistencies,
  • Overall, this appears to roughly conserve the precision of computed depth:
    • 81.00% of computed depths are within  +-2.0 m of DTM depth  (was 83% in the July 5th image)
  • this brand new trick seems to be promising,
    • needs to be further tested and augmented
    • would greatly benefit from many more bottom type signatures
bottom typing in 4SM is an art: no ISODATA here!

   
July 5th                                                                          July 25th