Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
  
WorldView 2 image Waimanalo Beach, Oahu, Hawaii islands
March 31th 2011, 3289x3241, 2 m ground resolution, 6.5 km * 6.5 km,   courtesy of Ron Abileah
home
This data has been deglinted
by Ron Abileah

 


 
 
 

Mapping density of living coral colonies
EPHE's NDR with WV2 images

see "A novel spaceborne proxy for mapping coral cover "
Antoine Collin, James L. Hench, Serge Planes, 2012

Collin's Normalized Difference Ratio

4SM's NDR

NDR=(Green-Purple)/(Green+Purple)

  • NDR : using water column corrected bottom signatures in units of reflectance,
    • this ratio is seen to increase linearly in the range 0.3-0.8
    • as an expression of the increase in living coral colonies at Moorea, French Polynesia
  • "the NDR green-purple is likely to embody a robust proxy for detecting coral pigments in live corals that exhibit higher reflectance in the purple band and a lower reflectance in the green band"
  • Live Coral Index: this allows Collin et al. to produce a map of the density of living coral colonies down to ~2.5 m at Moorea
 
NDR=(Purple-Green)/(Purple+Green)
  • NDR: using normalized water column corrected bottom signatures in units of DNs, it is to be expected that this ratio
    • equals 0 where Purple=Green over arguably clean sands
    • increases linearly as the Living Coral Index increases
    • decreases linearly (?) as the Leaf Area Index or biomass of vegetation increases
  • It is to be expected that a similar relationship could be developped for  the density of shallow vegetation in order to map the Leaf Area Index or biomass of shallow vegetation

 

water column correction and mapping NDR

  • in order to achieve water column correction of bottom signature in units of reflectance, the authors need to
    • calibrate the data in units of reflectance
    • apply the dark pixel correction in order to obtain BOA data
    • use existing depth sounding points to calibrate Stumpf's m0 and m1 parameters for Stumpf's method
    • compute a DTM of shallow area using Stumpf's method
    • apply the DTM depth, and spectral K values observed by Maritorena at Moorea, in order to produce spectral water column corrected shallow biottom signatures, and then map the NDR

water column correction and mapping NDR

  • in order to achieve water column correction of bottom signature in units of DNs, the 4SM method needs to
    • apply the dark pixel correction in order to obtain BOA data
    • use the image itself to calibrate the optical model
    • compute water depth, water column corrected bottom signature, and NDR "on the fly" in one single 4SM run
  • During the inversion process for the current pixel,
    • if the current pixel does not exhibit good bottom detection in the yellow band
    • once a depth is obtained
    • a NDR is immediately computed
    • from that last bit of information, the SoilsLine model is adjusted accordingly ,
    • so that a new depth is computed, which is more realistic than the first one




July 24th 2012
Final bottom type classification


July 24th: Final classified image: 15 classes
For now, this is just a proof of concept,
much easier to produce than the "classified image"

August 24th 2012
First image of the NDR at Waimanalo Beach


Green tones where Green>Purple
Red tones where Purple>Green
 



August 28th
Using  LIDAR DTM to investigate
and to improve 4SM

Z4SM not tide-corrected                                                      added Htide to ZDTM
so that we can compare the water column corrected bottom signatures

Z4SM
is used to compute bottom signature

LB=Lw+(L-Lw)*exp(2K*Z4SM)



profile_black

ZDTM
is used to compute bottom signature

LB=Lw+(L-Lw)*exp(2K*(ZDTM+Htide))


profile_black

  • Z4SM and ZDTM are shown in the profile plot

  • it is seen that the SoilsLine model used to achieve water column correction is the cause of an error on estimated depth

  • this can cause Z4SM to be offset by up to several meters for specific bottom types where only bands 1, 2 and 3 are used for modeling

  • by using Z4SM, water column corrected bottom signatures are seen to be
    • badly affected by this offset in the Red and Yellow bands which have high  K values ,
    • much less affected in the Blue-Green range because K values are much lower
  • Z4SM and ZDTM are shown in the profile plot

  • by using ZDTM in place of Z4SM, true water column corrected bottom signatures are computed

  • this kind of plot is extremely precious, as it can be used when it comes to

    • identifying training sites for bottom typing

    • specifying the rules for assigning a pixel to a particular bottom type

  • note the Purplish signature along most of section A

NDR image 

obtained by applying Z4SM,
modeling by only bands 1, 2, 3

-Lm/1/1/1/255/255/255/255/255

  • in the 0-10 m depth range, bands 1, 2 and 3 are not enough:
    • green coverage is clearly overstated: depths over-estimated
    • purple coverage is clearly understated: depths are over-estimated
    • therefore we want to make good use of the Red and Yellow bands
  • in the 10-30 m depth range, purple coverge is clearly overstated : as seen in section A of profile_black, depths are under-estimated
    • I need to find a way to alleviate that: ==> see below

NDR image 
obtained by applying ZDTM+Htide,

 

This is a "seatruth" NDR image

 

NDR image 
obtained by applying Z4SM
using bands 1, 2, 3, 4 and 5


-Lm/1/1/1/2/2/255/255/255
 
  • the pollution artifact shows up badly in purple tones in the central area, which originate in underestimated depths

  • apart from polluted areas, this image is quite encouraging: it shows that in the absence of pollution,  such NDR image over the whole shallow depth range should attract some interest


Over time, it shall be known
that this NDR image  can be produced
without the need of any existing field data

 

NDR image 
obtained by applying ZDTR+Htide



This is a seatruth NDR image
  • coral patches seem to be mostly dead coral colonies
  • the moats and gully are covered in dense vegetation, and maybe also the central deep lagoon




Now, it would make sense
that someone comes up
with seatruth observations
on the nature of shallow bottoms
at Waimanalo Bay



 

Apply ZDTM+Htide,
in order to investigate the "true" water column corrected bottom signature

LB=Lw+(L-Lw)*exp(2K*(ZDTM+Htide))
-Lm/1/1/1/1/1/255/255/255


There are three aspects to the "pollution"

  • evidence in the radiance data on the day of imaging
    • radiance in the Red and Yellow bands is boosted by the pollution; some sort of glow radiated by backscatter in the water column
    • plume of suspended sediment can be ruled out
    • can we exclude fluorescence?
  • long term effects on the shallow bottom
    • seaweeds: sandy bottoms in the areas affected seem to have turned "green": this is evidenced by the NDR image in the sandy areas. This even is observed in the gully and moats. If this can be confirmed, a seaweed cover has developped, quite dense and healthy in the gully and moats
    • dead corals: coral colonies seem to have died: this is evidenced by the NDR image of the oval coral patches
  • cause of this situation
    • red tide? how about some sort of "red tide" that boosts the radiance, mostly in the Red and Yellow bands, with negligeable impact on other bands
    • this seems to happen just where Waimanalo stream pours out into the bay


 
 

LB_CC for bands 3, 2 and 1
green/blue/purple

LB_CC for bands 4, 3 and 2
yellow/green/blue

LB_CC for bands 5, 4 and 3
red/yellow/green
  • all three water column corrected bands are seen to be hardly affected at all by the "pollution"
  • coral patches are well rendered
  • the area affected shows up in orange-red tones, caused by the yellow band
  • "bottoms" are rendered quite bright in affected area, as a result of boosted radiance in Yellow band
  • "pollution" extends over  bottoms that are optically deep in the yellow band
  • the area affected shows up in yellow tones where Red and Yellow are significant, and in green tone where Yellow only is significant
  • "bottoms" are rendered very bright in affected area, as a result of boosted radiances in Red and Yellow bands
  • "pollution" extends over  bottoms that are optically deep in the red and yellow bands
  • coral patches are not detected



 

 



Créer un site
Créer un site