tmnov-tutorial-model1-3

Things you can do:
get a feeling of Lw

raw TM1 prominent dark shadow	raw TM2 faint shadow	raw TM3 no shadow??	raw TM4 no shadow
a strong Lwblue normalized Lwblue=10.2	a weak Lwgreen normalized Lwgreen=2.6	a null Lwred Lwred=0.0	hence a null Lwnir Lwnir=0.0

striping/noise in the NIR band
is merily transfered to other bands by the deglinting process.
HowTo Lw and La

For the Blue vs Red pair of bands
the blue dot represents average healthy vegetation

Lwblue and Lwgreen

We posit that
- Lw reflectance of Oceanic waters
  - decreases progressively from high values in UltraBlue through Blue range,
  - then keepsdecreasing through much lower values in Green-Orange range,
  - then is negligible in the Red-NIR range
- Lw can reach up to in excess of 10% reflectance for an UltraBlue band in the Sargasso Sea, or in upwelling of very deep waters
- Lwblue>>Lwgreen
- Reflectance of healthy vegetation is
  - high in Green
  - low in Blue and Red
- La=Lsw-Lw after deglinting
Therefore, for a Blue vs Red pair of bands
- Lwgreen may conveniently be set to 1.5
- Lwblue may be set so that the Soil Line
  - runs right through the scatter of healthy vegetation
  - in a bi-dimensional histogram

Check that yourself: run your script with

-Calibrate/voaI/BdsNpZg/BDh_12_13_34

In 4SM:
What is important is not
that these Lw values be physically exact:
All radiance values are dimensionless!

In 4SM:
What is important is that the Soil Line
be assigned the right ratio LMi/LMj.
4SM is a ""ratio method!

Things you can do:
Adjust Knir

Knir must be estimated using the Calibration diagram

2Knir=4.5 m-1
-KK1_2_0.4960/Knir4.5/CoefK_1.0/mask_3

2Knir=3.5 m-1
-KK1_2_0.4960/Knir3.5/CoefK_1.0/mask_3

What is important is not that the BPL fits nicely
Rather, the BPL model must parallel the linear scatter of BPL pixels

Things you can do:
refine deglinting range

This dates back to 2011
It was using version 4SM.4.08; 4SM has improved a lot since then

Lswnir is now =6.5
-Lsw/60.42/17.41/10.79/06.5
from now on, 4SM shall derive the -Lsw... argument
using Lswnir=6.5 and the glint regressions.
This overrides the values which appear in the commandline

Range of glint correction:
reduce or increase GlintM as you see fit

-deglint/vRbaD/GlintM25.0
- deglinting occurs if Lsnir>=Lswnir=6.5
- deglinting occurs if Lsnir<=GlintM=25

Deglinted=========6.5>=Lsnir<=25.0
- haze and small clouds have been removed
- GlintM=25 is too high
- GlintM=10 is enough

Not deglinted=======Lsnir>25.0
- note that three very shallow areas (and also one small cloud) have not be deglinted

Above is No Deglinting
- -deglint/vRbaD/GlintM00.0
Below is what we want to acthieve locally
- -deglint/vRbaD/GlintM10.0 is enough

-Deglint/vRbaD/GlintM25.0

GlintM=25 was assigned by the AutoCAL process
- all pixels with Lsnir<=GlintM are deglinted
assumes NO bottom detection in NIR band
- therefore, radiances over very shallow areas are damaged
- GlintM must be reduced

-Deglint/vRbaD/GlintM10.0
is more convenient

pixels with Lsnir<=10 are deglinted
- haze is deglinted, but not clouds
radiances over very shallow areas are preserved
allows for modeling very shallow areas using all four BGRN bands

-Deglint/vRbaD/GlintM25.0

-Deglint/vRbaD/GlintM10.0

???GlintM=10 or GlintM=19???

This dates back to 2011
It was using version 4SM.4.08
4SM has improved a lot since then

Profile 1

GlintM=19

<====GlintM=19

Note that the very shallow areas are badly damaged by this "heavy" deglinting
This means that
- GlintM=10 must be applied to the whole image : this does not remove the lump of haze to our satisfaction
- GlintM=19 must be applied on a local basis,
  - knowing that the deglinted radiances in very shallow areas might be badly damaged (less than 1 m deep over bright bottoms).

<====GlintM=19

-deglint/vRbaD/GlintM10.0
10 or 19???

See above that

the difference in depths as computed by Red solution shall be very small in the damaged section
the difference in computed bottom reflectance is badly affected in the damaged section:
- this is something WE DON'T WANT if we are to proceed to bottom typing

-deglint/vRbaD/GlintM19.0
10 or 19???

See above that

the difference in depths as computed by the 3-bands case shall be very small in the damaged section
the difference in computed bottom reflectance is badly affected in the damaged section:
- this is something WE DON'T WANT if we are to proceed to bottom typing

Conclusions:

Lswnir=6.5
GlintM=10.0 over the whole image
- to preclude heavy damages at very shallow locations
GlintM=19.0 on a local basis
- haze, clouds

Things you can do:
remove local haze

This dates back to 2011
It was using version 4SM.4.08
4SM has improved a lot since then

Lswniris now =6.5
-Deglint/vRbaD/GlintM19.0
-E420/530/1250/1530

-deglint/vRbaD/GlintM10.0

Deglinting applied
- inside the range 6.5>=Lsnir<=10.0
See that a lump of haze is not deglinted this would require GlintM=19

-deglint/vRbaD/GlintM19.0

GlintM=19 removes the haze to perfection
In order to protect very shallow areas which cannot be deglinted, this can be obtained
- inside a window: -E420/530/1250/1530
- or under a mask

-E... argument Syntax_E

Things you can do:
refine Lsw

This dates back to 2011
It was using version 4SM.4.08
4SM has improved a lot since then

Can we trust the Glint regressions
for estimating the deep water reflectance Lsw?

-Lsw/60.42/17.41/10.79/06.50 from the glint regression textfile

see in textfile tmnov.mcd :
cm01="24 1_Ldeglinted_485 nm";                    dboc=24 ; r mcd
cm01="25 2_Ldeglinted_578 nm";                    dboc=25 ; r mcd
cm01="26 3_Ldeglinted_677 nm";                    dboc=26 ; r mcd
cm01="27 4_Ldeglinted_840 nm";                    dboc=27 ; r mcd
use OpenEV to display deglinted/smoothed
channels 24, 25, 26 and 27 using a random color pallet

channel_24 TM1

Lsw.............=61.4
orange.........=62
red...............=61
dark purple..=60

channel_25 TM2

Lsw...............=17.4
light green.....=18 has bottom detection
yellow............=17 is optically deep

channel_26 TM3

Lsw..........=10.8
yellow......=11
orange......=10

channel_27 TM4

Lsw...............=6.5
light green.....=7
yellow...........=6

Things you can do:
??which band has deepest bottom detection??
Bottom detection limit deeper in Blue than in Green
Blue TM1 deglinted/smoothed Green TM2 deglinted/smoothed The yellow-green limit is at Lsgreen=17.5: this radiance may be assigned to Lswgreen=17.5 Deeper than ~15 m, sea bottom drops down suddenly to the abyss But Blue_TM1 appears to become extinct approximately 100 m further away than Green_TM2 This consolidates the last optical calibration: ZMblue............~=41 m ZMgreen.........~= 21 m Kblue/KGreen~=0.52 in spite of the quasi absence of bright sandy bottoms deeper than ~15 m