WorldView 2 image Bay of Laganas, Zakynthos Island, Greece
8114*6968 image courtesy of Emmanouil Oikonomou, Technological Institute of Athens, Feb 11th 2016
work done in february 2016

the "PAN solution" uses the PANCHRO band,
4x4 sum-aggregated and coregistered
bands 1, 2, 3, Panchro, 4, 5, 6, 7 and 8 become  1, 2, 3, 4, 5, 6, 7, 8 and 9
This is a case of extremely dark bottoms  with near-null reflectance
like along the coast of Western Australia


The data
Four tiles had to be mosaicked.
Data includes a 4*4 sum-aggregated, 2 m resolution,
co-registered Panchromatic band

TCC raw

PAN raw
  • 4*4 sum-aggregated, 2 m resolution, using OpenEV
  • PAN exhibits much deeper penetration than Yellow

Glint Regressions

TCC image
Region of interest where to sample "glint"
is very difficult to identify.

There is hardly any sea-surface glint effect.

NIR2 image

But a fairly strong adjacency effect MUST be corrected for
in view of optical modeling.


Glint regressions
based on NIR2 band 9

  • Note that adjacency effect can only be sampled where bottom is assumed to be optically deep
  • Therefore, glint regressions for this scene are not very robust

Glint regressions
based on NIR1 band 8


removal of the adjacency effect

  RED deglinted
  •  For comparison, a vertical strip has been left un-deglinted in this image
  • Note distinct turbidity off Argassi (far north)
    • it extends far away offshore


TCC deglinted
Please note negative bottom contrast
in bands 1, 2, 3 and PAN
over very dark bottoms

BLACK: horizontal stripes of NoData
occur in the NIR1 and PAN wavebands

Pan deglinted
Please note negative bottom contrast
in bands 1, 2, 3 and also PAN
over very dark bottoms

  GREEN deglinted
Please note negative bottom contrast
in bands 1, 2, 3 and PAN
over very dark bottoms

Deglinting along Profile GREEN
  • Note the negative bottom contrast from 0 to ~5 km in both blue bands, and also in green band

Raw Blue_2_band
with profile locator

  • Note a slight offset of low radiances in the Blue_band_2 over a westernmost vertical strip
Deep water radiances
see Profile RED

Negative bottom contrast
This is where Ls<Lsw: the TOA radiance is less than the Deep water radiance, over the whole depth range.
  • For a blue band,  this means a very low reflectance:
    • this is customary for seaweed/grass beds, and for healthy vegetation on land.
  • For a green band, this means a near-null reflectance: this is sometimes the case for coconut grooves in atolls;
    • it is very rare underwater.
  • At this extremely low level of reflectance, optical modeling would require top notch sensors designed for coastal work.
    • this is not the case of current EO spaceborne sensors like OLI or WV2.

Homogeneous water body?
  • Unlike sophisticated analytical methods, 4SM normaly assumes a homogeneous water body over the whole scene and over the whole depth range.
    • where this condition is not met, results can be very badly affected.
    • but we have designed a way in 4SM to overcome a situation where waters get progressively clearer away from the waterline.
  • For this scene, we seem to observe slightly turbid waters in the RED range of the solar spectrum in the 0-5 m depth range over brightest bottoms -see red arrows.
    • in the optical calibration, we shall attempt to account for a progressive clearing of waters deeper than ~5 m deep in this image,
    • from a consistent OIII marine water type of Jerlov in the 0-5 m range
    • to OIB at ~30 m.

Using metadata:
reference to reflectance of coral reef sands
Landsat 8
  • Landsat 8 OLI data include a .MTL.txt metadata text file with which to convert TOA radiance into units of TOA reflectance:
  • We have processed many Landsat 8 images of coral reef environments, and so derived an average BOA reflectance signature spectrum for coral reef sands, calibrated in the 0-1 range.
  • This signature is then used to scale water column corrected of any current scene relative to coral reef sand signature.
  • Using the .IMD metadata text file of a WV2 image, we do the same: 
    • absCalFactor, effectiveBandwidth, etc.
  • It shall be observed that the reflectance of the brightest shallow substrates in this Laganas scene is only ~0.25 that of coral reef sands,
    • although the whole of Zakynthos island is mainly made of limestone.

This is a marked recent improvement in 4SM,
as it allows for in a tighter control of the Soils Line.


Using the PANchromatic band:
The "PAN solution"
  • Deeper than ~10 m over bright bottoms in a WV2 scene, bottom reflection only occurs in the Green and Blue bands even over bright bottoms: that's three bands.
    • in 4SM, to retrieve bottom depth, Green is weighed against the average of two blue bands.
    • this results in underestimation of depth over green submerged vetetation:
      • the darker the bottom, the largest the underestimation.
      • this is sometimes very nasty at shallow depths.
  • By adding a PAN channel, we now have four bands.
    • in 4SM, to retrieve bottom depth, PAN is weighed against the average of three green-blue bands.
    • this is expected to yield a better estimation of retrieved depth over green submerged vetetation.
Thanks to your seatruth dataset,
this exercise should allow us to comment on that aspect.


Créer un site
Créer un site