Red: WZ=5   Yellow: WZ=4   Green: WZ=3

The above results are unfair
to the western part of the image,
where there is hardly
any sign of "pollution"
(is that because the wave breakers
do not hit the beaches?)
 

• By releasing the Lm_red and Lm_yellow thresholds from 10 to 2 while processing the western area, one gets closer to estimating the real potential in terms of error on computed depth.

• The result is:

  • RMSE               =0.96-0.50=0.46 m

  • average ZC-ZR=0.68-0.50=0.18 m

• Increasing Lm_yellow=from 2 to 4 further reduces RMSE  by 0.03 m

  • as it appears, in the depth image below, that the yellow band shows signs of increased turbidity in the 5-8 m depth range

• Remember less glint and less noise in the image should entail further reduction the RMSE figure

-Lm/01.0/01.0/01.0/010.0/010.0/255/255.0/255.0

-Lm/0001.0/001.0/001.0/010.0/010.0/255/255.0/255.0

Red tones signal over-estimated depths
 

• Note the very bad overestimated depths seaward of the fringing reef
  • this is because the 3-bands case cannot assumes greenish bottom signature (this suits a vast majority of cases)
  • whereas corals's signature lacks green radiance

Blue tones signal under-estimated depths

-Lm/0001.0/001.0/001.0/004.0/003.0/255/255.0/255.0
West from row_1 to row_1000

Increasing Lmyellow from 2 to 4
Increasing Lmred from 2 to 3
further improves the results
over the western area

Increasing Lmyellow from 2 to 4
Increasing Lmred from 2 to 3
further reduces RMSE  by 0.12 m
over the 0-10 m depth range

From 0 to 10 m

• 96.2% of all estimated depths  

  • are within +-2 m of DTM

• RMSE=0.84-0.50=0.34 m

• average ZC-ZR = 0.61-0.5=0.11 m

This regression pertains to all pixels
within the 0-10 m depth range along profile_black

Using a smart smoother on a less glinted and less noisy image 

would further reduce RMSE by 0.xx m

Discarding the wild and lousy pixels (see above)

would further reduce RMSE by 0.xx m

Using a full DTM in centimeters  for seatruth

would further reduce RMSE by 0.xx m
 

Profile_black over a backdrop of ZR-ZC
in the western area

Using less bands at any given pixel
increases the RMS error:
this can be because

• illumination is not enough

• the image is noisy

• the bottoms are darker

• the bottoms are deeper

• the waters are less clear

Deeper and/or darker bottoms 
shall contribute more to the RMS error
than shallower and/or brighter bottoms

RMS error increases exponentially 
with bottom depth
with bottom darkness

for a "best case", see AAHIS at Kanehoe Bay

nice -20  4SM  -Process  -Origin/Olsen_NPS               
-DB/waimanalowv2m_deg/47_3_16_0/8_51/3289_3241/0.001_-0.001/1_1   
@DB/waimanalowv2m/47_3_8_0/8_51/3289_3241/0.001_-0.001/1_1       
-Mis/Hawaii/Oahu/WV02/Multi/TIF/UTM_4_008/0.001_-0.001/31_MAR_2011   
@-WL0427.0/477.5/546.0/607.5/659.0/724.0/831.0/908.0           
-cWL/0.500/0.500/0.550/0.500/0.500/0.500/0.500/0.500           
-M/@000001/@0002/00003/00004/00005/@0006/@0007/00008           
-Lm/0001.0/001.0/001.0/004.0/002.0/010.0/255.0/255.0           
@Lm/0001.0/001.0/010.0/010.0/010.0/245.0/255.0/255.0           
@Lm/0001.0/001.0/001.0/001.0/001.0/010.0/255.0/255.0           
@Lm/0001.0/001.0/001.0/001.0/001.5/001.0/255.0/255.0Land-E1/600/1900/2700        
@Lm/0001.0/001.0/001.0/001.0/001.0/001.0/255.0/255.0Z21        
-Lsw/045.7/030.1/009.9/009.5/009.4/000.0/000.1/000.1_new       
@dLsw/0000/00000/00000/00000/003.5/00000/00000/00000           
-Lw/0041.7/017.8/002.7/000.9/000.0/000.0/000.0/000.0_new       
-LsM/220.0/195.0/225.0/215.0/200.0/200.0/200.0/200.0_new       
@LsM/220.0/195.0/225.0/240.0/230.0/200.0/200.0/200.0_ZEDland       
@Veg/027.0/024.9/041.0/030.8/022.8/113.8/194.8/202.3original       
-Veg/027.0/024.9/027.0/039.8/022.8/113.8/194.8/202.3           
-SCL/00038/00042/00078/00101/00058/00106/00078/00114_new       
-LS/005000/05000/05000/05000/05000/05000/05000/05000           
-LsT/000.1/000.1/000.1/000.1/000.1/000.1/000.1/000.1           
-KK2_3_0.3500/Knir8.0/Bm400_BG510_GR600_RN700/mask_3           
@Z/LAT1.00/N_1/land_1/z21 /RLN_46.0/RLN_10.0/RLN_7.5   /cZ1.12  /mask_3       
-Z/LAT1.00/N_1/land_1/z21 /cZ0.943 /RLN_46.0/RLN_10.0/RLN_7.5   /cZ1.12  /mask_3       
-B/tclNe/cLM1 .2                                       
-deglint/VRbaD/DegTol1.00/F1/L8/N1/FN/mDEGLINT1/GlintM255                   
@extract/v/rawBDH/NIRband8/NIRmax5.0/FullBDH/mSOIL21/mBPL3/ext_1500_1700_1_1000/Mask_3   
@extract/v/rawBDH/NIRband8/NIRmax5.0/fullBDH/mSOIL21/mBPL1/Ext_2260_2360_400_500-estuary   
-extract/v/rawBDH/NIRband8/NIRmax5.0/fullBDH/mSOIL21/mBPL2                    
@Calibrate/v/BdSNpZg/BC_8_7_6_5/BDh_12_13_14                           
-calibrate/v/BdSNpZg/BC_7_6_5_1/BDh_12_23_14                           
-calibrate/v/BdSNpZG/BC_5_4_3_2/BDh_12_23_14                           
@Model /mask_3 @-E1/3289/430/3241  @-E1900/2200/800/1000 -E1/1000/1000/3241         @-LL1           
-Smooth/5/D/smart+    @-DD2323/1865-DD2360/1854-DD1400/2620-DD2411/1790 @-DD220/2045-DD2066/515    
@ProfileAB/Z/profile_black/chAB_1_2_3_4_5                       
-RegressZZ/v/p/DTM_1/0.001_0_0/1.000_0.0/ZM/3_5/0_201/0_10_50/*  @-E1/500/1300/3241 @-E800/1200/800/1600   
@Classify/Typer40_0.2/Classifier_1_0_Scef_LastChType3/mode_5_11_16               
@Classify/typer_0.2/Classifier_1_0_scef_LastChType3/mode_5_11_16               
@ProfileAB/L/profile_black/chAB_4_5 1_2_3_4_5                           

About the pollution:
in the end, we have no clue
aout the cause of this "pollution".

• pollution by the  Waimanalo Stream would be expected to result in a distinct increase of absorption coefficient in both blue bands due to a strong content in yellow substances:

  • this does not appear to be the case, although we could estimate that 0-5 deep waters just at the stream's outlet are of a Coastal 1 water type of Jerlov: these are still clear "green" waters with a hint of dissolved organic matter.

  • it all looks like remediation efforts since 2001 have been quite successful

• pollution by suspended particles: it would not take a lot of suspended particle load to prevent genuine bottom detection by the green band deeper than ~8 m in the gully

  • we tend to favor this hypothesis

  • but how do we explain that this "pollution" spreads so wide and large at the base of the water column, and even fills isolated sandy ponds where their depth is badly underestimated?

• linearity: we found no evidence which could lead to blame the lineariry of the sensor's response curve