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 LC80340432015010LGN00, January 10th 2016
Work done march 2017

GSD 15 m PANsharpened by 4SM
a strong case of Coastal 1 water type down to ~8 m, followed by OII+0.4 at greater depths

The good and the bad:
useful in a time series study case
but very risky as a one-off processing

A good study case for analytical methods, see if they can cope

Ground truth
uses CombinedZ as seatruth DTM

only under all of San Lorenzo Channel and all of Isla Espiritu Santo

January 10th 2015    vs   CombinedZ
  • Lsw: retrieved depths in the 10-30 m range along the diagonal  very much depend on the right choice of a location for estimating deep water radiance Lsw: a bold move in this complex swirling blooming situation: looks like I could reduce Lsw for the Panchro just a wee bit
  • Cut-off: then there is that strong forking in the 10-20 m range, where retrieved depths level at 10-12m: see how retrieved depths all around Isla Espiritu Santo are underestimated (red tones)
  • Noise: very noisy: might be caused by lax co-registration of panchro with Multi bands

CombinedZ - January 10th 2015
see legend
Blue tones: depth is over-estimated
Red tones: depth is underestimated

This explains that
  • 1-3 m: excellent, assuming tide height 0.35 m
  • 3-8 m: fairly good
  • 8-13 m: slightly under-estimated
  • 13-20 m: badly over-estimated
  • 20-25 m: very good

only under all of Channel
and all of Isla Espiritu Santo

Data and Deglinting
a very complex atmospheric and hydrologic situation

TOA TCC: raw image
logarithmic enhancement
Note the red discoloured features: these are discoloured waters which do not show up in the NIR or SWIR1 bands (meaning that nothing actually is floating at sea surface)

BOA TCC deglinted image
logarithmic enhancement

"deglinting" of lumps of haze
and of slight adjacency effect is quite good
Glint regressions using NIR
there is no hint of sea-surface glint

RMSE values are very high
deep water radiance Lsw is sampled here:
a bold move!

Optical calibration
only under San Lorenzo Channel and and Isla Espiritu Santo
surprinsingly clear-cut, in spite of hydrological complexities

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
What we see is:
  • 0-8m: KBLUE/KGREEN~=0.94
    • Jerlov water type Coastal_1,
    • nice linear BPL line: denotes homogeneous brightest bottom in homogeneous waters
  • 8-13m: clearing  to water type  OII+0.4
  • 13-30m: KBLUE/KGREEN~=0.74
    • homogeneous OII+0.4
Distinctly stratified waters

Water appears to be distinctly clearer
deeper than 8 m:
a clear view of what I call "stratified waters"
Under the BPL assumption..............==>
Clearest: the above describes the clearest water over the whole shallow depth range, whatever the location: these shall exhibit correct depths: see areas in light-red_to_light-blue tones in the ZDTM- Z4SM image.
Less clear: but the above masks areas where the brightest pixels are in less clear: these shall exhibit under-estimated retrieved depths: see areas in dark red tones in the ZDTM- Z4SM image
==>.....So now we offer some understanding
of the hydrological situation on that day
  1. Coastal waters rule everywhere, from West to East, from North to South, from sea-surface down probably to in excess of 20 m deep
  2. Oceanic OII waters well up from below at restricted locations in the channel, probably powered by tide oscillations
    1. but they certainly do not reach shallower than ~8-13 m deep on that day

under GREEN window

under BLUE window
coastal waters all the way down


Three windows for calibration

for each window,
extraction of calibration data
is restricted to that window,
in order to probe local conditions


under RED window
As opposed to
Other scenes exhibit intermediate situations

The bay area, of course,
should be described on its own!

 from West to East,
from North to South

You wrote 25/02/17 :
"it is very important to support the theory
that the October 2014 image is the result of the Odile hurricane,
that provoked huge movements of sediments,that disrupted the northern reef...etc,etc"

Hurricane ODILE was on September 10-19th 2014

Z4SM with location of profile Black

Profile Black october 22nd 2014.............==>
4SM computed depths
mostly conform with CombinedZ
the theory that the October 2014 image is the result of the Odile hurricane"

==>................Profile Black January 10th 2015 4SM computed depths
exhibits two distinct differences
with CombinedZ
January 10th 2015
  • Overestimated:
    • The southern half of sections C (5 to 7.2 km) and D (5-8.5 km) exhibit depths over-estimated by 2-4 m
    • They have been processed assuming OII water type
  • Underestimated
    • section A (4.8 to 5.4 km) exhibit underestimated depths by ~7 m
January 10th 2015

  • A: either the hurricane has moved large amounts of sediment around
  • B: or this is a trick of
    • B1: either complex water optical properties
    • B2: or bottom substrate spectral signature
Solution A
is impossible because both images are post-ODILE
  • the CombinedZ DTM represents a "stable" situation
    • which has been modified by hurricane ODILE in 2014 at two major locations:
    • eroded sediment over sections C and D as observed
    • deposited sediment over section A as observed (progradation)
  • but we do not observe such modifications on October 22nd 2014
Solution B
  • B1:
    • turbid waters flowing at depth at section A (4.8 to 5.4 km) would cause depth underestimation by 4SM
  • B2:
    • strong modifications of bottom substrate spectral signatures can bring about such observed variations in the computed depth results
I am shall download
more post-Odile images:
It'll be interesting

  • to compute a second CombinedDepth DTM, using a new series of scenes, see if it matches the first CombinedDepth.
  • to streamline that process, as it obviously is
    • an interesting workaround lack of homogeneity of atmospheric and water optical properties, assuming such inhomogeneities are only transient 
    • and an likely/acceptable alternative to analytical methods(?).
Such proeminent discrepancies in retrieved depth would be eliminated
 by the combining depths process

Ready for modeling

Retrieved depth
see legend

Bottom brightness

BOA CC WCC normalized
BGR color composite:
  • Blue  =Coastal band
  • Green=Green  band
  • Red    =Blue   band

very pale indeed!!!



Créer un site
Créer un site