Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
Bathymetry and water column correction
at Caicos Bank, Bahamas
4018*4149, 30 m pixel size, UTM zone 18, downloaded from USGS
Using the Panchromatic band for water column correction
to derive water depth and spectral bottom signature:

Landsat 8 OLIP bandset used for this work

Purple_1Blue_2Green_3PAN_4Red_5NIR_6 and SWIR1_7

Please refer to Bora Bora and Sanaa
for use of the PANchromatic band for water column correction

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
scene LC80090452015139LGN00 May 19th 2015
work done august 2016


This is spring in the Bahamas: lots of greeneries.
Sun light is plentyfull
Calibration optimization now operational
Vegetation Line for Lw and La

DTM seatruth   RMSE=0.62 m
see legend
Z4SM vs ZDTM+1.10 m
tide correction+ ZDTM+1.10 m


Data and Deglinting

TOA TCC: raw image
logaritmic enhancement

TOA TCC deglinted image
same logaritmic enhancement


TOA TCC deglinted chequered
same logaritmic enhancement


TOA TCC water column corrected chequered
same logaritmic enhancement

Glint regressions for NIR band at 866 nm
There is no sea surface wave-modulated clutter.
But there is a lot of aerosols, like in the Middle East,
far denser in the South-East part of this scene.

Most often, sea-surface glint and path radiance
have very similar spectral properties,
but not always though!
So what does it take
to "deglint" this scene?
  • Locate an area with lowest NIR signal, most often in a cloud shadow area (could also be some dark lake at sea level):
    • this shall yield the atmospheric path radiance  LaNIR  for the NIR band,
    • which is equal to the deep water reflectance LswNIR   
    • LaNIR=LswNIR
  • Sample areas of variable "glint" signal: here from low to high path radiance
  • Run the pair-wise regressions
    • Knowing LswNIR and the slope and intercept of the regressions, spectral LswWL is computed for all visible bands
  • Run the Deglinter, under the "dark pixel subtraction" assumption

Plot of deglinting along profile_green
for Coastal, Green and NIR bands

Profile_green locator
  • profile_green_A runs from NW towards SE
    • at 136 km, it crosses a very dark cloud shadow
  • profile_green_B runs from SW towards NE
  • increase: see that the path radiance increases steadily from NW_18 towards SE_70
    • this is a huge increase over 150 km
  • "island effect": see that the coastal band at 440 nm exhibits a distinct "island effect" which lowers its deep water radiance
    • see profile_green_B at 120 km, where it extends over ~12 km
  • noise: see that the noise is much weaker over shallow waters

Optical calibration
16U data are scaled to allow for comfortable screen display

Calibration diagram
for bands Blue, Green, Red and NIR
KBLUE/KGREEN=0.54 => Jerlov water type OIB+0.5

Calibration diagram
for bands Blue, PAN, Red and NIR
Reflectance (0-1)
of brightest bottom substrate for this scene
Coastal=1  Blue=2  Green=3  PAN=4  Red=5  NIR=6   SWIR1=7
0.272     0.336   0.405     0.426  0.473  0.613   0.600
2K (m-1) for Jerlov water type OIB+0.5
diffuse attenuation coefficient
Coastal=1  Blue=2  Green=3  PAN=4  Red=5  NIR=6   SWIR1=7
0.114    0.100     0.183                 0.627  2.908              
Reflectance (0-1)
of atmospheric path radiance
0.107    0.082    0.056    0.050   0.037  0.024    0.011
Water volume reflectance (0-1)
0.020      0.016   0.001    0.001  0.000  0.000  0.000

Vegetation Line

Blue vs Red: use the "vegetation line"
to estimate LwBLUE=~13.5


Vegetation Line

Coastal vs Red: use the "vegetation line"
to estimate LwCOASTAL=~24.5


Calibration optimization
is an achievement obtained in july 2016, now hard coded in 4SM
The ratio Kblue/Kgreen
  • Hopefully, assigning a value for the ratio Kblue/Kgreen from the image itself shall be easy enough.
  • This rests on two assumptions: 
    • that the water body is homogeneous.
    • that the brightest bottom type is present over the whole depth range
    • if only as isolated patches/pixels.
Spectral LsM
  • Hopefully, assigning a first order spectral value for LsM
    • retrieved depths shall be quite close to correct
    • let's call that "ball-parking" the calibration.
  • But this is not enough for one who wants to produce a fair water column correction for bottom typing using a time series of images
    • this rests on the assumption that the brightest bottom type sould exhibit a "flat" spectral response
    • meaning: should not be biased toward any particular wavelength.
  • In other words, spectral LsM should be representative of a very bright sand/mud, un-tainted by any discoloration, like clean fine grained coral/quartz sand/mud, or snow
    • achieving this requires a very fine-tuning of LsM
    • achieving this manually can take hours
    • we now use an automatic procedure to do this .

Before optimization of spectral LsM
Average retrieved depth 4.77 m
over a ROI of 5700 pixels
Reflectances are shown (0-1)
  • Depth retrieval may be good
    • and bottom typing for this scene might prove to be useful
  • But, in view of time series of images or scenes, we can't accept such a "sea-saw" water column corrected signature for the brightest bottom type in a coral reef enviromment

After optimization of spectral LsM
Average retrieved depth 4.82 m
over the same ROI of 5700 pixels
Reflectances are shown (0-1)
  • Depth retrieval only changed by a mere 1.1%
  • The optimization is automatic and runs for less than one minute
  • Now we can engage into time series/scenes comparisons of shallow habitats


Calibration before optimization of spectral LsM
appears to be quite good,
but still needs fine-tuning though!

Calibration after optimization of spectral LsM
quite a subtle fine-tuning
has been achieved


Bottom types for SAM classification

Now, a complete suite
of bottom type signatures may be developped,
which hopefully should be suitable
for bottom typing of this time series
of Landsat 8 OLIP scenes
over the Bahamian platform


Now ready for Modeling
Modeling by the PAN solution ;  No smoothing applied
Average bottom brightness
Very bright in spite of ubiquitous seagrass
regrowing in spring
BOA TCC water column corrected
No enhancement
BOA view reveals the details 
SAM mapping
Noise caused by SWIR1 deglinting

see legend for SAM
Retrieved depth in centimeters

see legend for retrieved depth in cm