A time series of landsat 8 images
at San Lorenzo Channel, Baja California

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
Collaboration with 
Fabio Favoretto, Ph.D Student, Coralline Algae Ecology
Grupo Interdisciplinario de Ciencia Ambiental, Universidad Autonoma de Baja California Sur
Carretera al sur km 5.5  | La Paz
Using pan sharpened images in this study
Pan sharpening by Fabio Favoretto, using GRASS Brovey method

work done october 2016

LANDSAT_SCENE_ID = "LC80340432016029LGN00"
DATE_ACQUIRED = 2016-01-29
    SCENE_CENTER_TIME = "17:47:55.3960143Z"
January 29th 2016

No smoothing applied

January 29th 2016    vs   Sonar Depths
RMSE=1.38 m

Sonar Two Days seatruth depths raster
not corrected for tide

15 m GSD
see depth legend

Z4SM    vs   ZDTM


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

A tide height of 0.8 m is added to ZDTM
This is compatible with tide tables for La Paz


Depth residuals ZDTM+0.80m   -   Z4SM

see legend

15 m pansharpened resolution
Pansharpening by Fabio using R  October 15th 2016

BROVEY method : Formula_OTB_pansharpening
Where i and j are pixels indices, PAN is the panchromatic image, XS is the multi-spectral image and PANsmooth is the panchromatic image smoothed with a kernel to fit the multi-spectral image scale.
Brovey transform
'IHS and BT suffer from individual color distortion on saturation compression and saturation stretching, respectively'
"BT is very efficient and highly accurate for merging SPOT-5 images"

CO-REGISTRATION: I observe a slight misregistration of pansharpenned  MULTI bands versus PAN band
Like one pixel rowwise and one pixel linewise
Seems this is only over land
Boats at sealevel are OK
USGS.............................data are coded U16
GRASS pansharpenned data are coded 
4SM reads LC80340432016029LGN00_MTL.txt 
This is required in order to convert DNs into reflectances (0-1)
offset=-REFLECTANCE_ADD_BAND[c]/REFLECTANCE_MULT_BAND[c] is subtracted upon importing the USGS data
seems pansharpened data have been added some big offset in the process
seems pansharpened DN radiances have been expanded over a wider S16 DN range
I suppose the LC80340432016029LGN00_MTL.txt received from USGS does not apply any more to GRAS pansharpened data:
but my stuborn code use the MTL metadata no matter what
we need to sort that out for further time series applications
as 4SM is only concerned with ratios among numbers for optical calibration of the attenuation of sun light through water
but still: be cautious when comparing SAM results which operate calibrated reflectances!!!
4 HOURS: it took me 4 hours to import pan-sharpenned data into a 4SM working database, calibrate the model (including wondering about the above!) until finished the first processing
now some fine tuning is in order
finally.........SUCCESS........stratified waters
  • All too often, things do not settle until I acknowledge that waters along the coastline are frequently less clear than farther away
  • This, after all, is what we can commonly experience while scuba diving not too far from a major city?
  • Last time I had to fight hard was at Gulf of Laganas, Greece 
  • It translates into a model for stratified water layers, becoming clearer as depth increases
    • ratio Ki/Kj decreases as depth increases
    • 4SM has a gimmick to account for it

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

Calibration diagram for the whole scene
for bands Blue, Green, Red and NIR
stratified waters
0-5 m: KBLUE/KGREEN=0.93 Jerlov water type Coastal

then: KBLUE/KGREEN=0.79 Jerlov water type OII+0.5
quite a subtle feature!


Calibration diagram for the whole scene
for bands Blue, PAN, Red, NIR


Ready for modeling
No smoothing applied

BOA TCC: water column corrected

Retrieved depth

see legend

Bottom typing

No smoothing applied

SAM classified image
WHITISH: spectrally neutral signature
GREENISH: greenish signature 

see legend for SAM

Average bottom brightness

see legend for image B

Bottom type signatures
  • signatures obtained in shallow areas using the RED solution, mostly at the upperleft beach
    • type10 is taken over brightest at the beach around UTM565.470/2702.416
    • type26 is over very dark reefs around UTM569.573/2967.200
  • Fabio, I want/need to develop a better understanding of, and control on this SAM mapping, as it is very new in 4SM and involves quite some assumptions
  • When scuba diving, please test this result


updated November 13th 2016
slcOLI_20160129_15mresults.zip 62 MB
updated November 13th 2016
Channel Descriptor    file="slcOLI_20160129_15mresults.pix";
 1 U8 SAM bottom typing-----------------------------
 2 U8 image B average bottom brightness (0-200)-----
 3 U8 image Z in decimeters-------------------------
 4 U8 WCC reflectance (0-250) at 440.0 nm-----------
 5 U8 WCC reflectance (0-250) at 480.0 nm-----------
 6 U8 WCC reflectance (0-250) at 560.0 nm-----------
 7 U8 WCC reflectance (0-250) at 590.0 nm-----------
 8 U8 WCC reflectance (0-250) at 655.0 nm-----------
 9 U8 WCCnormalized (0-250) at 440.0 nm>B-----------
10 U8 WCCnormalized (0-250) at 560.0 nm>G-----------
11 U8 WCCnormalized (0-250) at 480.0 nm>R-----------
12 S16 Z4SM in centimeters--------------------------
13 S16 ZDTM in centimeters--------------------------
14 S16 ZDTM-Z4SM in centimeters---------------------

Congratulation Brovey method
works very well for 4SM shallow water work

Comments following your email  dated Mon, 17 Oct 2016 17:28
- BOA-TCC: what i see here is a representation of the “Below of Atmosphere” radiance, water column corrected images, from which you can retrieve depth, right? so everything i see there is topology of the bottom? red points along the coastline and are outliers? 
What you see is a "Low-tide view" of this scene, meaning it is a water column corrected view, like if at no tide, and like if on the moon where there is no atmosphere
- SAM Classified image: a map of the BOA water column corrected spectral bottom reflectance:  should i interpret the different green range of values as different ground types? 
  • Water column corrected spectral bottom reflectance is not always spectrally neutral.
  • Type10: in 4SM, I assume that the brightest shallow bottoms are devoid of any chlorophyl, so are spectrally neutral, like clean/fine-grained coral sand or terrigeneous sand. So In enforce that assumption in the "optimized" optical calibration
  • Type11 to ...: then 4SM starts to render a certain ammount of greeneness of the spectral water column corrected bottom signature
    • starting from  endo fauna/flora (cyanobacteria, foraminifera, mucus, etc),
    • through sparse vegetation/corals,
    • all the way to dense and healthy vegetetion cover (seagrass meadows, encrusted hardgrounds, ...)
  • SAM uses BOA water column corrected bottom spectral signatures in units of calibrated reflectance (0-1), like if at low tide on the moon
    • by the book, the spectral angle should be converted into radians
    • which I don't do, hence the scaling see legend for SAM
  • In 4SM, SAM is a direct translation of how proeminent the Green features relative to the average (Coastal+Blue) in a water column corrected BOA spectral bottom signature  
    • SA=(Rwcccoastal+Rwccblue)/(2*Rwccgreen)
    • Rwcc are scaled  to reflectance (0-1),
      • this is what is written as BOA WCC signatures in channels 4 to 8 of the results you downloaded
    • see caicosoli-20130513
    • see caicosoli-20151111
-the color is chosen because it is within the green wavelengths? so this means that i can interpret more green intensity to the presence of Chlorophyll? where the black area is the area “””greener”””??
I leave it to you to interpret what all that gimmick represents, and how you can benefit from it.
One thing that surprises me is that "very green" (like SAM type 25) in fact describes a fairly dark signature, i.e. all reflectances are very low, just the Green is a bit higher.
How much of chlorophyll activity is involved, I don't know.
How about the seasonal cycle: we all know about big heaps of dead seagrass at mediterranean beaches during winter: this is all dead and black (any chlorophyll left??)
SO: very dark cannot be equated to very green. There is more to it!
You are the one to test it all on site
I see that you are developing sound and learned understanding
I am elated
-average bottom brightness: i think this is self-explenatory (well hopefully i got this right)... so excluding pixels that tend to be higher than 200 (violet), this is another way that could help us identify bottom types. 
For Landsat 8 or WV2, average bottom brightness is (WCCCoastal+WCCBlue)/2  on a scale of 0-200 see legend for image B
Echosounding for RMSE on depth
As you probably have no access to MBES/LIDAR mapping of your site,
nothing beats a few DGPS  navigated depth sounding profiles
for one who wants to derive RMSE on retrieved depth.

Créer un site
Créer un site