Bathymetry and water column correction
at SanLorenzoChannel, Baja california
Image courtesy of the U.S. Geological Survey
7671*7841 30 m pixel size, UTM zone 12
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

scene LC80340432014007LGN00, January 07th 2014
DATE_ACQUIRED = 2014-01-07
    SCENE_CENTER_TIME = 17:49:09.8591999Z

Work done october 2016

January 07th 2014
15 m resampled resolution
Resampling by 4SM

Ground truth

January 7th 2014    vs   Sonar Depths
RMSE=1.76 m


Sonar Two Days seatruth depths raster
not corrected for tide

15 m GSD
see depth legend

January 7th 2014    vs   DTM


DTM from U. of La Paz, Mexico
resampled to 15 m GSD
see depth legend

ZDTM+0.5m - Z4SM
DTM - January 7th 2014
see legend 

GREEN solution or PAN solution?
15m resampling or not?

15 m PAN-sharpening or not?
Smoothing or not?

One more trick: apply ZDTM
15m GSD PAN sharpened by 4SM
Dec 7th 2016:
this was a gross mis-understanding

of the Brovey magic formula!
but nonetheless: it worked fine!

GRASS tutorial: with a Brovey pan sharpening,
each of the 3 lower resolution bands and panchromatic band
are combined using the following algorithm
to calculate 3 new bands at the higher resolution (example for band 1):

new band1 = --------------------------------- * panband
  band1 + band2 + band3
One oldband1 30 m GSD pixel is split into four 15 m GSD newband1 pixels:
Final touch requires to return pansharpened bands to original scaling of Landsat 8 data.
I achieve that in 4SM, as follows:

newband1=newband1*oldband1/average(four newband1)

This means that I can calibrate 
any pansharpened  radiance into reflectance (0-1)
New SAM in 4SM
  • previous version only mapped GREENish  spectral angle bottom signatures
    • in the range type10 to type30: shades of green
    • "PURPLEish" bottom signatures did not pass the test
      • they were mapped a type10, which is brightest bottom type
  • from now on, purple signatures are mapped
    • in the range type30 to type 50: shades of purple
Green Coastal_1  water
This image offers an opportunity for calibration of a Coastal water type
Seems that this water has a disctintly green colour

Now I want to
  • process all SanLorenzoChannel images at 15 m GSD
    • using 4SM pansharpened data
  • redo all SAM mapping for
    • SanLorenzoChannel
    • Caicos Bank time series
    • LeeStockingIsland time series

Data and Deglinting

TOA TCC: raw image
logarithmic enhancement

BOA TCC deglinted image
logarithmic enhancement
30 to 15 m resampling of MULTI bands
no pan-sharpening

  • Image shows SWIR1 band over open waters
    • it has been resampled from 30 m GSDto 15 m GSD by splitting one pixel into 4 identical pixels
    • no pan-sharpening has been applied
  • Red vector shows which 15m GSD panchro pixels have high glint signal
  • So good co-registration has been achieved
    • this required offset by one PAN pixel in both row and line
  • Maybe deglinting a pan-sharpened image is a bad idea, as it is bound to produce a very noisy deglinted image.
  • Or should the sharpening be applied on deglinted images?
Two distinct water bodies
Bay of LaPaz..........vs.......LaPaz Basin
  • WEST: BayOfLaPaz:
    • hardly any seasurface glint
    • deep water radiances are higher in the blue-green range
  • CENTER: crosses the SanLorenzoChannel
  • EAST: LaPazBasin:
    • a lot of  seasurface glint
    • deep water radiances are lower in the blue-green range
Sea surface glint

glint is fairly strong
over the LaPaz Basin

Sea surface glint

  • deglinting by SWIR1 is fairly good, as shown by R2 values
    • deglinting by NIR exhibit lower R2 values
  • deglinting of the pansharpened image reveals that
    • co-registration is good
    • 4SM Brovey pansharpening did not damage the radiometric quality of the original data
    • deglinting shall be quite noisy (is this because of pansharpening?
Glint regressions using SWIR1
No pan-sharpening applied

R2 values are fairly high
15 m resampling does not increase noise
Glint regressions using SWIR1
Pan-sharpening applied

R2 values are low
15 m pan-sharpening increases noise


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

Calibration diagram
for the whole image
for bands Blue, Green, Red and NIR


Calibration diagram
for the whole image
for bands Blue, PAN, Red, NIR
Coastal waters 0-5 m
Xblue vs Xred
  • KBBLUE/KRED=0.33
    • KBLUE  =0.245 m-1
    • KGREEN=0.268 m-1
    • KRED    =0.744 m-1
  • This is Coastal 1 water type of Jerlov
  • Blue BPL pixels are located at the beach
  • SOIL LINE  is represented by small white dots
    • it runs from the LsM point all the way to a black body point
    • its shape offers a way for us to estimate the water volume reflectance, and therefore also the atmospheric path radiance
    • this assumes Lsw=La+Lw
    • it also assumes that LwRED~=0
Deeper waters  are clearer
    • KBLUE=0.163 m-1
    • KGREEN=0.221 m-1
  • This is Oceanic OII+0.4 water type of Jerlov
This stratifification of water clarity
  • is not big deal
  • is observed in all three images  I worked on so far
  • was the cause of "saturation" in my preliminary results
  • may be confirmed by visual observation, maybe by underwater photos,  maybe also using Secchi disk (horizontal sighting??)
Calibration diagrams are just as good
for a 15m GSD pansharpened image
as it is for original 30 m GSD data

  • 0-5m: KBLUE/KGREEN=0.95
    • Jerlov water type Coastal_1
  • 5-30m: clearing progressively to OII+0.3

Greensish Coastal_1 water
  • Thanks to abundant coverage of low_altitude mountainous desertic dryland with many very dark shaded areas, the following can be estimated with reasonable confidence:
    • atmospheric path radiances
    • water volume reflectances
  • The result hints at
    • a fairly low water volume reflectance in the COASTAL band,
    • a fairly high water volume reflectance in the GREEN and PAN bands
  • This is a remarquable result,
    • it is consistent with Coastal water type in the 0-5 m depth range
    • this bodes well as we wish to ascertain water column correction in view of SAM botttom typing
  • Failing to detect  and account for that entails frustation of the practioner

"Stratified waters"

The central channel exhibits clearer waters

"Stratified waters"

Profile Blue location
backdrop is a BOA TCC WCC
 The RED and NIR bands exhibit higher deep water radiance where Coastal water type rules:
this is suspended particles load
  • North: from water line to WP2
  • South: from WP3 to water line 


Créer un site
Créer un site