Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
WorldView 2 image at Waimanalo Beach, Oahu, Hawaii islands
3289x3241, 2 m ground resolution, courtesy of Ron Abileah

Pollution    and    Error

see also http://www.watercolumncorrection.com/waimanalo-errors.php

  • Screen displays of Red and Yellow bands exhibit weird features, which suggest some kind of "pollution"
  • Whatever the cause of this alledged "pollution", whether Waimanalo Stream or breaking waves along the shoreline or ...,
    • it is observed to affect the Red, Yellow and Green bands in the central part of the image
    • it would be caused by a build-up of suspended load in a deep-seated layer of slightly turbid waters which simply prevents bottom detection and cause badly underestimated bottom depths, 
      • as we failed to substantiate any significant build-up of dissolved organic matter (yellow substances) in these extremely clear "blue waters"
  • Use of the DTM and detailed insight into these pollution aspects leads us to conclude that the effective wavelength for the Green band would be 549.5 nm, rather than 546 nm
  • Allowing for this last bit of tuning,
    • the overall precision on computed depth improves substantially
    • an initial suspition of "non-linearity" of the response curve of the WV2 sensor seems to vanish away
  • It's hard to see how sophisticated fully fledged, time consuming, and costly  atmospheric corrections would help

Evidence of pollution
Under-estimation of depth in modeling results

Operational wavelengths: NoNeed for field data for WV2 images
Final results
Observed error

Reducing the error: work on the western area
Command Line

Evidence of pollution
 see report by Hawaii State Department of Health 2001

depths converted into centimeters

  • Waimanalo stream was heavily polluted in 2001

  • Distinct features in the water column correction of the WV2 image over Waimanalo Bay (March 31 2011)  might be caused by pollution of the waters by the outflow of Waimanalo stream

  • the whole purplish area would be affected in the water column corrected TCC view

From left to right, images below show deglinted bands 6 to

  • NIR band 6 hardly exhibits any water penetration
  • Red band 5 at two small areas indicates that the Red band reaches complete extinction of bottom reflected light at a DTM depth of 6-7 m
    • whereas it exhibits distinct and heterogeneous water leaving reflectance over a domain which extends from 0 to ~11 m in tide corrected DTM depths: this is contributed by water volume reflected light rather than by bottom reflected: in other words, the bottom is mostly not detected through this "mist"
    • deep water radiance is ~9.5 , but it is observed to reach 13-14 over the polluted area
  • Yellow band 4 maximum depth of penetration is ~12 m (i.e.  ~5 m deeper than Red band in this water body)
    • this does not show here, as Red and Yellow bands mostly look the same in these displays
    • this means that the Red and Yellow bands are heavily affected by the water pollution (possibly suspended particle load)

  • Bands 3, 2 and 1 clearly depict ~2-3 m high coral patches which are not seen in Yellow band 4
    • this means that bands 3, 2 and 1 are relatively less affected by the water pollution
    • although we might expect diffuse attenuation coefficient in  the Purple band to be distinctly stronger due to higher content in dissolved organic matter (yellow substances): optical calibration shows that this is not the case
  • The area affected by pollution would warrant special calibration and modeling under a specific mask
  • I just wonder if the 'semi-analytical methods' would adequately account for these subtleties


No sign of "pollution" in the NIR1 band

Location of profile_green

Profile_green_C and profile_green_B run over optically deep waters in the RED and NIR bands

NIR  shows no sign of significant "pollution"
RED shows consistently several DNs higher than deep water radiance

Add tide height=100 cm to all DTM readings

Read further comments

No sign of "pollution" in the NIR1 band

Profile_green for bands 4, 5 and 6
section C is north                 section A is south

profile_green for band 7

Pollution is for real
Well maybe not real "pollution" as we know it.

Still there is something that causes the increase of deep water radiances in the Red and Yellow bands,
but does not significantly affect the water optical properties of these Oceanic Type 1 blue waters.

Waters that pour out of the -small- Waimanalo Stream are observed to be of the Coastal 1 water type:
clear waters with a buildup of dissolved organic matter.

There is a large army base, and also a marine park in the surrounding.
There is also aquafarming at Oceanic Institute at Makapuu point: they pump up deep cold waters?

My guess is that, for some reason,
a plume of suspended particles hovers over the 'polluted' area.


Lm=1 ==> extent of

  • red: modeling with bands 1, 2, 3, 4 and 5, where the BOA signal is >=Lsw+1 in the red band

  • yellow: modeling with bands 1, 2, 3 and 4, where the BOA signal is >=Lsw+1 in the yellow band


my interpretation is that

  • most depths computed can only be underestimated to a point that is unacceptable
  • unless very severe Lm thresholds are used so that depths are only computed where the bottom reflected signal -or bottom contrast- is strong enough that the potential error remains small and acceptable

SHOALS seatruth depths
add tide height=100 cm to DTM readings


  • only the westernmost part of this image is not affected by the pollution,
    • where both limits run over acceptable SHOALS depths
  • most of the image is very affected:
    • where both limits run over SHOALS depths which by large exceed the maximum bottom detection depth: i.e. ~6 m for bright bottoms in the Red band, and ~11 m for bright bottoms in the Yellow band
  • remember that SHOALS depths are ~ 1 m shallower than uncorrected 4SM depths (tide height=~ 1 m)

Bad results
Thresholds -Lm/1/1/1/1/1/255/255/255
 Water column corrected color composite:
a dramatic display of polluted areas: 
they are rendered in blueish and purplish tones


Final results
Thresholds -Lm/1/1/1/10/10/255/255/255
 Water column corrected color composite:
polluted areas are still there
rendered is pinkish tones

Bad results
Thresholds -Lm/1/1/1/1/1/255/255/255
Normalized water column corrected color composite:
  • a dramatic display of polluted areas:
    • they are rendered in deep purple tones
    • there is no way for us to say how far away they extend seaward (likely a bit further away than shown here)
  • note the presence of bright green tones:
    • they are obtained by fully enabling the Red band
    • I think they are realistic


Final results
Thresholds -Lm/1/1/1/10/10/255/255/255

Water column corrected color composite

  • a less dramatic display of polluted areas: they are rendered in dull tones
  • note the absence of bright green tones:
    • by disabling the Red band for all but the brightest pixels
    • modeling with bands 1, 2, 3 and 4 cannot render the existence of those bright and green features: they shall be missing in the bottom typing exercise
Did you say "a two substrates" scene?

Bad results
Thresholds -Lm/1/1/1/1/1/255/255/255

Blue tones: underestimated depths "invade" the scene


Final result

thresholds are raised in order to alleviate artifacts
(cannot block out  artifacts though!)



Créer un site
Créer un site