Bathymetry and water column corection
World View 2 at Saint Croix, US Virgin Islands
8967*8805, MULTI, UTM zone 8

LIDAR seatruth of 4SM computed depths at Buck Island Reef, US Virgin Islands
Investigation of Wavelengths and Tide Height, etc


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
Turbid plumes

How do Semi-analytical methods
handle these?

4SM can't
I gather DG can't


Relative difference in percents
  • Blue depths are underestimated
  • Red depths are overestimated

ZZRegressor: Statistics of seatruth ZC-ZR
on image buckislandreefwv at BuckIslandReef, USVI
N=8.647 millions pixels
HTide=0.40 Smooth=1_5 Using_bands_1_2_3_4_5
 2.19% pixels with depth underestimated by more than -5.0 m
 0.44% pixels with depth           ""                -4.0 m
 0.65% pixels with depth           ""                -3.0 m
 1.44% pixels with depth           ""                -2.0 m
15.37% pixels with depth           ""                -1.0 m
38.90% pixels with depth           ""                -0.0 m
33.27% pixels with depth  overestimated by more than  0.0 m
 6.27% pixels with depth            ""                1.0 m
 0.95% pixels with depth            ""                2.0 m
 0.31% pixels with depth            ""                3.0 m
 0.13% pixels with depth            ""                4.0 m
 0.07% pixels with depth            ""                5.0 m
72.18% of computed depths are within  +-1.0 m of DTM depth
93.81% of computed depths are within  +-2.0 m of DTM depth
96.20% of computed depths are within  +-3.0 m of DTM depth
97.17% of computed depths are within  +-5.0 m of DTM depth
over a total of 100.00% of computed depths
Coral heads
  • coral heads are deglinted and smart-smoothed correctly
  • coral heads are so dark that they are modeled by the 3 bands case: wZ=3
    • their computed depth is commonly overestimated by several meters
    • they stand out as dark red in the above difference image
Strong whitecaps
  • many whitecaps too strong deglinting to be meaningful
  • as they are not masked,
    • they generate fancy shallow computed depths which are totally unrelated to seatruth depth


Very dark coral heads and mounds:
depth is badly overestimated

Location of Profile Blue

Profile Blue

  • Yellow: on section B at 0.330 km, one coral head is modeled by wZ=4 (yellow along the abcissa):
    • its computed depth is much less overestimated


dated May 20th 2013,  30 MB
from GoogleDrive

Computed depth corrected to LIDAR datum: a tide height of 0.4 m has been subtracted
Comparison with DTM reveals that my 4SM depth image is slightly offset in both X and in Y

Command Line
cp buckislandreefwv.049_BUIS_Bathy_3m buckislandreefwv.049

nice -20  4SM.4.13  -Process  -Origin/DigitalGlobe  
@ProfileAB/ZrvwB/profile_purple/DTM/chAB_1_2_3_4   /3_5/0_201/0_50_50/*  

Tide height
"NOAA web site indicates a tide offset of about 10cm"
I read ~20 cm on Jan 14th 2012 at Christiansted Harbor
Now, I'm a bit confused
In the course of all these above tweaking,
I seemed to witness the need to add a tide height of 0.6 m to Lidar DTM depths
which now appears to be un-justified, from looking at NOAA's website.
  • So, I should review all this tweaking, which included tweaking of the Soil Line, see if I can do without subtracting any tide height
  • Then, applying a tide correction is only justified in two events
    • seatruthing: upon reading the seatruth depths, in order to preserve the radiometric consistency while seatruthing
    • deliverable: the end user must be delivered WV2 depths to specific hydrographic datum

Profile Purple 
BLACK profile are 4SM depths

RED profile are DTM depths: 
a tide height of 0.4 m has been added to DTM depths

Location of Profile_Purple
backdrop is the Lidar DTM
sections A, B and C from west to east

  • This sort of plot is valuable to appreciate Htide
  • water column corrected reflectances are a great contribution for the practioner
    • to develop a fair understanding of what is going on, locate weird features, and keep things under control
    • to start locating suitable training sites for bottom type classification

Water column corrected view
4SM computes both bottom depth and water column corrected spectral reflectance "on the fly"

This color composite is "normalized"
so that subtle variations show up 
  • weird results stand out dramatically, and call for some explanation or further tweaking
  • potential bottom types may be sampled through training sites for bottom type classification

Like at low tide:
The natural color composite is less dramatic!


June 10th 2013

Much improved results

This much improved result
benefitted greatly of the LIDAR ground truth.
At least it shows what is achievable
with 4SM under ideal conditions.
  • The most delicate parameter is spectral deep water reflectance Lsw, a choice which is all the more uncertain in view of the WV2 system noise.
  • Then, glint parameters must be taken over whitecaps (rather than over clouds, which did not yield consistent results here).
  • This was obtained through hard-coding in 4SM some observed behaviour of bottom types, under LIDAR control
    • it remains to be seen whether or not this is applicable too other sites: the odds are: maybe not!
  • I still need to work out how to flag whitecap and cloud pixels, as they should be excluded from the seatruth exercise!

Ground truth depth - Retrieved depth, in centimeters
  • Note the large blue area in the easten part: depths there are overestimated by ~2 m
    • this translates into a sharp red peak at ~18 m in the red histogram below (same as in Compendium)

No pixel excluded
No tide correction
over the 0-30 m ground truth depth  range
No pixel excluded
No tide correction


Profile Red: see its location