estimating water optical properties.ppt

32
Estimating Water Optical Properties, Water Depth and Bottom Albedo Using High Resolution Satellite Imagery for Coastal Habitat Mapping S. C. Liew # , P. Chen, B. Saengtuksin, C. W. Chang Centre for Remote Imaging, Sensing and Processing National University of Singapore # Corresponding Author ([email protected])

Upload: grssieee

Post on 26-Jun-2015

691 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: Estimating Water Optical Properties.ppt

Estimating Water Optical Properties, Water Depth and Bottom Albedo Using High Resolution Satellite

Imagery for Coastal Habitat Mapping

Estimating Water Optical Properties, Water Depth and Bottom Albedo Using High Resolution Satellite

Imagery for Coastal Habitat Mapping

S. C. Liew#, P. Chen, B. Saengtuksin, C. W. ChangCentre for Remote Imaging, Sensing and Processing

National University of Singapore

#Corresponding Author ([email protected])

Page 2: Estimating Water Optical Properties.ppt

WorldView-2High resolution with 8 spectral bands

Launched: 8 October 20090.46 m panchromatic1.84 m multispectral

8 spectral bands:

Band 1: 429.3 nm (47.3) “Coastal”Band 2: 478.8 nm (54.3) BlueBand 3: 547.5 nm (63.0) GreenBand 4: 607.8 nm (37.4) YellowBand 5: 658.5 nm (57.4) RedBand 6: 723.5 nm (39.3) “Red edge”Band 7: 825.0 nm (98.9) NIR1Band 8: 919.4 nm (99.6) NIR2

Effective wavelength Bandwidth

Page 3: Estimating Water Optical Properties.ppt

WV2 Spectral Response

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

350 400 450 500 550 600 650 700 750 800 850 900 950 1000 1050 1100

Wavelength (nm)

Rela

tive

Spe

ctra

l Res

pons

e

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

Atm

osph

eric

Tra

nsm

ittan

ce

Tropical Atmosphere, 4 cm precipitable water

Note the high water vapor absorption in band 6 (“red-edge” band), humid tropical atmosphere

Page 4: Estimating Water Optical Properties.ppt
Page 5: Estimating Water Optical Properties.ppt

WorldView-2 Image

Semakau, 2010-03-24

Seagrass

Submerged reefs

Page 6: Estimating Water Optical Properties.ppt

• The intertidal zone of Semakau has a rich seagrass habitat of several hundred meters in length.

• Such an extensive seagrass habitat is rare in Singapore coastal area. The seagrass habitats in other areas of Singapore mostly occur in patches.

• There are also live corals on the reefs near Semakau.

Page 7: Estimating Water Optical Properties.ppt

Classification Map

Semi-automatic classification Based on 8-bands WV-2 image and field survey.

seagrass

Page 8: Estimating Water Optical Properties.ppt

Seagrass

Page 9: Estimating Water Optical Properties.ppt

Coral rubble with algae/seagrass/coral

Page 10: Estimating Water Optical Properties.ppt

Classification of submerged features

• The previous classification map shown was obtained by automatic clustering followed by manual editing guided by extensive ground truth observations.

• Time consuming, requiring visual interpretation• Visual interpretation complicated by effects of

water column– Scattering by suspended particles– Absorption by water and colored dissolved organic

matter– Different water depth

Page 11: Estimating Water Optical Properties.ppt

• We attempt to retrieve the water depth, bottom albedo and intrinsic optical properties of coastal sea water over submerged areas using a spectral matching algorithm.

Page 12: Estimating Water Optical Properties.ppt

Pre-processing of WorldView-2 Image

• Calibrate to radiance and top-of-atmosphere reflectance• Correct for Rayleigh scattering and gaseous absorptions,

integrated over sensor response functions.• Glint subtraction using band 8 (NIR2)• Convert to subsurface reflectance

S.C. Liew, B. Saengtuksin, and L.K. Kwoh, IEEE 2009 International Geoscience and Remote Sensing Symposium (IGARSS'09), 13 - 17 July 2009, Cape Town, South Africa.

S.C. Liew and J. He, IEEE Geoscience and Remote Sensing Letters 5(4), 701-704, 2008.

Page 13: Estimating Water Optical Properties.ppt

Band 8 (NIR2) Image

Note the presence of various surface features

Page 14: Estimating Water Optical Properties.ppt

Band 7 (NIR1) Image

Similar surface features are visible

Page 15: Estimating Water Optical Properties.ppt

Band 7 (NIR1) after subtracting Band 8

More homogeneous surface

Page 16: Estimating Water Optical Properties.ppt

Automatic Isodata clustering of submerged pixels into 50 classes

Above-water land surface masked out by thresholding the NIR2 band

Mean reflectance spectrum of each class is collected and matched with model reflectance

Page 17: Estimating Water Optical Properties.ppt

Shallow water reflectance

wL ,L

H)(b

)( ),( bba

Deep Water

Shallow water reflectance Deep water

reflectance

sF

LR

cos

)(

s

ww F

LR

cos)( ,

F

s v

Page 18: Estimating Water Optical Properties.ppt

Model of Subsurface shallow water reflectance

)exp(1)( MKHrw Reflection (scattering) from water column

e)(subsurfac angleszenith solar andsensor ,

depth water

cos/1cos/1

tcoefficien extinction ,)()()(

ereflectanc water deep )(

vs

sv

b

w

H

M

baK

r

Reflection (scattering) from sea bottom )exp()(

MKHb

ereflectanc bottom )( b

)exp()(

)exp(1)()( MKHMKHrr bw

Page 19: Estimating Water Optical Properties.ppt

Deep water reflectance

)()(

)(

)( 210

b

b

w

ba

bu

ugugr

a() = Absorption coefficient

bb() = Backscattering coefficient

g0, g1 = parameters dependent on scattering characteristics of suspended particles

Page 20: Estimating Water Optical Properties.ppt

Absorption and Backscattering Models

)440( )];()ln()([)(

:lchlorophylby Absorption

)440( )];440(exp[)(

:detritus and CDOMby Absorption

)550( ;)/550()(

:matter eparticulatby ringBackscatte

)()()()( :tcoefficien Absorption

);()()( :tcoefficien ringBackscatte

10

aPaPaPa

aGSGa

bXXb

aaaa

bbb

gg

bpy

bp

gw

bpbwb

Page 21: Estimating Water Optical Properties.ppt

Sea bottom reflectance

)()()( ssvvb

vegetation

sand

Sea bottom reflectance is modeled as a linear combination of typical sand and vegetation reflectance spectra.

)659()659()659(

)825()825()825(

ssvvb

ssvvb

)659()825(

)659()825(

bb

bbNDVI

(Sea bottom NDVI, corrected for water column effects)

Page 22: Estimating Water Optical Properties.ppt

Example of spectral matching:Deep water

Class 3: Deep water

X = 0.25 m-1 , G = 0.096 m-1 P = 0 Water depth set to a large value H = 25 m during spectral fitting (actual value doesn’t matter)

Page 23: Estimating Water Optical Properties.ppt

Example of spectral matching:Reef edge

Class 6: Fringe of coral reef

X = 0.23 m-1 , G = 0.019 m-1 P = 0 Rb547 = 0.135, Rb659 = 0.154, Rb825 = 0.282, NDVI = 0.292H = 1.30 m

Page 24: Estimating Water Optical Properties.ppt

Example of spectral matching:Submerged reef

Class 41: shallow reef

X = 0.26 m-1 , G = 0.0 m-1 P = 0.25 m-1

Rb547 = 0.226, Rb659 = 0.267 , Rb825 = 0.365, NDVI = 0.154H = 0.31 m

Page 25: Estimating Water Optical Properties.ppt

Example of spectral matching:Submerged seagrass

Class 25: submerged seagrass

X = 3.21 m-1 , G = 0.0 m-1 P = 0 m-1

Rb547 = 0.024, Rb659 = 0.020, Rb825 = 0.155, NDVI = 0.776H = 0.12 m

Page 26: Estimating Water Optical Properties.ppt

Water Depth

0 m

0.5 m

1.0 m

> 1.5 m

Page 27: Estimating Water Optical Properties.ppt

Bottom Albedo (at 547 nm)

0

0.10

0.20

> 0.30

Page 28: Estimating Water Optical Properties.ppt

Vegetation Index (Water column corrected)

1.0

0.50

0.0

Detection of submerged aquatic vegetation

Page 29: Estimating Water Optical Properties.ppt

Concluding Remarks

• We illustrated the application of a spectral matching algorithm in deriving the water depth, bottom albedo, vegetation index (for submerged aquatic vegetation) and water quality parameters from 8-bands high resolution WorldView-2 satellite images.

• The satellite derived reflectance spectra can be fitted quite well to the shallow water reflectance model.

• The 6th band (“red-edge” band centered at 723 nm) always has a high deviation from the best fit value for all the classes. This band happens to coincide with a water vapour absorption band.

Page 30: Estimating Water Optical Properties.ppt

Concluding Remarks

• Eight spectral bands of WorldView-2 enable the application of a spectral matching algorithm, but implementation on the full image is not time-efficient.

• Computational time efficiency is improved by clustering pixels with similar spectral values, and spectral matching is performed on the average spectrum of each class.

• The water column corrected NDVI can serve to detect submerged aquatic vegetation, and to quantify the abundance.

• Integrating with classification methods is on-going.

Page 31: Estimating Water Optical Properties.ppt

Acknowledgment

• Singapore Agency for Science, Technology and Research (A*STAR) for funding to CRISP

• Singapore National Parks Board (Nparks) for a grant supporting the project.

• S. C. Liew acknowledges support of Singapore-Delft Water Alliance (SDWA)

Page 32: Estimating Water Optical Properties.ppt

WV2 Spectral Response

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

350 550 750 950

Wavelength (nm)

Re

lati

ve

Sp

ec

tra

l Re

sp

on

se

0.001

0.01

0.1

1

10

100

Wa

ter

Ab

so

rpti

on

Co

eff

(m

-1)