A Remote Sensing Study of A Remote Sensing Study of Coral Reefs; Coral Reefs;

Kailua Bay, Oahu.Kailua Bay, Oahu.

Ebitari Isoun, Ebitari Isoun,

Charles Fletcher,Charles Fletcher,

Neil Frazer, Neil Frazer,

Jonathan Gradie,Jonathan Gradie,

Scott RowlandScott Rowland

AcknowledgementsAcknowledgements For shared data and field work: John Rooney,

Jodi Harney, Eric Grossman, Melanie Coyne and Zoe Norcross

Members of the Coastal Geology Team for moral support: Tara Miller, Dolan Eversole, Clark Sherman, Scott Calhoun, Matt Barbee, Mary Engels, Rob Mullane, Rikki Grober-Dunsmore, Chris Conger, Ole Kaven

For spectral band selection and use of field targets: Eric Hochberg and Marlin Atkinson

Acknowledgements (continued)Acknowledgements (continued) The people at TerraSystems Inc. for friendly

assistance: Pamela Elwin, Kevin Jim, and Elbert Hwang

For funding and workspace: NASA, USGS-Coastal Geology Program, Sea Grant, Department of Geology and Geophysics, and SOEST

For unconditional love: My Family For mystery and blessings: God

Topical OverviewTopical Overview

High-resolution multi-spectral imagery Map bathymetry and percent living coral

NN

Kailua Bay, OahuKailua Bay, Oahu

Topical Overview (continued)Topical Overview (continued)

Passive remote sensing Radiative transfer model

• Atmosphere, ocean surface, water, and ocean substrate

“Differencing” of two spectral bands

Topical Overview (continued)Topical Overview (continued) Error Assessment

• Depth: hydrographic survey• Percent living coral: diver-obtained ground truth

Topical Overview (continued)Topical Overview (continued) Correlation of predictions to

environmental and human factors Geographic Information System (GIS)

space• Better reef management e.g. Maragos and

Grober-Dunsmore, 1998• Basemaps for scientific studies e.g. Harney et

al., 1999

Introduction• Study Site• Data Collection

Methods• Data Processing• Radiative Transfer Theory• Depth and Bottom-type by Band Difference• Applying the Model

Results• Depth Predictions• Percent Living Coral Predictions

Conclusion

OutlineOutline

Study SiteStudy Site

Oahu KailuaBay

0 10 km

21.5˚

158˚

21˚

158˚HawaiianIslands

OahuNN

0 100 km

630500m E. 10 15 20 25 30 633500m E.

2369500m

N.

90

85

80

75

2369

500m

N.

90

85

80

75

2367

000m

N.

630500m E. 10 15 20 25 30 633500m E.

2367000m

N.

sand channel

spur andgroove

Sand fields

karst caves and caverns

Submerged beach rock

plains

Reef fro

nt

Reef front

Kailua ReefKailua Reef

NN

Data collectionData collection

January 10, 1998– Light winds– No rain– Minimal ocean swell– 9:30 to 10:30 a.m.– 20 to 30 m horizontal visibility in water– Ocean floor visible to 30 m

Data collection (continued)Data collection (continued)

Low flying (1400 m) airplane– ThunderChicken

Data collection (continued)Data collection (continued) Application Specific Multi-Spectral

Camera System (TerraSystems, Inc.)• 8-bit precision

Data collection (continued)Data collection (continued) Multi-spectral

images collected along a north-west to south-east transect

60% overlap along flight path

20% overlap across flight path

NN

OahuOahu

KailuaKailua

335˚335˚

1 m1 m

2 m2 m

Data collection (continued)Data collection (continued)An image from the 6th flight path 1 pixel = 1 m

578 m

740 m

488 nm

551 nm

557 nm

10 nm full width half maximun

Hochberg and Hochberg and Atkinson, 2000Atkinson, 2000

OutlineOutline

Introduction• Study Site• Data Collection

Methods• Data Processing• Radiative Transfer Theory• Depth and Bottom-type by Band Difference• Applying the Model

Results• Depth Predictions• Percent Living Coral Predictions

Conclusion

Data processingData processing PCI Geomatics TM

••11

••11

••22

••33

••22••33

1:5000 aerial photographs

Coyne et al., 1998Coyne et al., 1998RMS = 0.5 m

NN

OahuOahu

KailuaKailua

Radiative Transfer TheoryRadiative Transfer Theory

Irradiance: time rate of change of sunlight energy with area (W m-2 nm -1)

Radiance: flux per projected area per unit solid angle (W m-2 nm -1 sr-1)

irradiance reflectanceupwelling irradiance

downwelling irradiance

Remote Sensing Reflectance,

Mobley, 1994Mobley, 1994

reflectance beneath the water surface

wavelength

reflectance of infinitely deep ocean

bottom albedo(R just above the ocean

bottom)

water attenuation

distribution function for the underwater light

field

depth

Philpot, 1989Philpot, 1989

Two-Flow ModelTwo-Flow Model

Gordon, 1989Gordon, 1989

Gregg and Gregg and Carder, 1991Carder, 1991

Elterman, 1968Elterman, 1968

Elterman, 1968Elterman, 1968Burt, Burt, 19541954

Mobley, Mobley, 19941994

Mobley, Mobley, 19941994

can be written in a simple equation in terms of Radiance:

If

where

Lb is the radiance of the ocean substrate

Lw is the radiance of the ocean

is the water attenuation coefficient

D is the water distribution function

z is depth

Ld = Lb exp -Dz + Lw

From the simplified equation:

A derivative band, Xi, can be defined:

(1) solve for the water attenuation coefficient,

(2) solve for depth and bottom-type

Ld = Lb exp -Dz + Lw

Xi ln((Ld-Lw) = lnLb-Dz

Solve for water Solve for water attenuation attenuation coefficient, coefficient,

-10 m-20 m-30 m

X488

X551

X557

-10 m-20 m-30 m

-10 m-20 m-30 m

Depth*D

600

200

y = 0.05 x + 8.00

y = 0.07 x + 8.84

y = 0.07 x + 8.78

Xi ln((Ld-Lw) = lnLb-Dz

Y-axis X-axis

slopeintercept

Sand

In agreement with Maritorena, 1996Maritorena, 1996

Solve for depth (Solve for depth (zz) and bottom-type () and bottom-type (YY))from the “difference” in two bands (from the “difference” in two bands (ii,,jj))

Assumptions:(1) Homogeneous water quality(2) Bottom reflectance is the same in two bands

(Frazer)(Frazer)

where g = D X = derivative band

How do we apply the model to multi-spectral data?How do we apply the model to multi-spectral data?

123

456

789

10

8-bit

32-bit

32-bit

32-bit

488 nm551 nm557 nm

ca

cw

t(cw ca)

sa

sw

Tsun

D

eb06aj.pixeb06aj.pix

Mosaic_model.pixMosaic_model.pix

1, 2, 3

4

567

89

10

8-bit

32-bit

32-bit

32-bit

488 nm, 551 nm, 557 nm

D

ca

cw

t(cw ca)

sa

sw

Tsun

Aeb06aj.pix

Beb07aj.pix

Relative Difference in Overlap

Before After488 nm 7% 0.9%

551 nm 4% 0.7%

Introduction• Study Site• Data Collection

Methods• Data Processing• Radiative Transfer Theory• Depth and Bottom-type by Band Difference• Applying the Model

Results• Depth Predictions• Percent Living Coral Predictions

Conclusion

OutlineOutline

157˚44’00”W 157˚42’50”W157˚43’30”W

157˚44’00”W 157˚42’50”W157˚43’30”W

157˚44’00”W 157˚42’50”W157˚43’30”W

21˚2

5’20

”N21

˚24’

55”N

21˚2

5’20

”N21

˚24’

55”N

21˚25’20”N21˚24’55”N

21˚25’20”N21˚24’55”N

-3 m -6 m -9 m -12 m -15 m -18 m -21 m -24 m

Predicted Depth (z488/551)

Hydrographic Survey Depth(USGS data, E. Grossman)

Percent Error

157˚44’00”W 157˚42’50”W157˚43’30”W

157˚44’00”W 157˚42’50”W157˚43’30”W

21˚2

5’20

”N21

˚24’

55”N

21˚25’20”N21˚24’55”N

0-5% 6-10% 11-15% 16-20% 21-25% 26-30% 31-35% >35%

Median = 11% Mean = 14% Std. Dev. = 11

•Percent error to depth R = 0.21•Boundaries sand channel

•Difference in water quality•Bottom-type assumption

Percent Living Coral Zones Percent Living Coral Zones (Harney, 2000)(Harney, 2000)

hardgroundssand

Living Coral<15%

15-25%

25-40%

40-75%

>75%0.730.851.0

0.76

0.760.58

0.620.74

0.44

0.54

0

0

00

0

00

0.67 0.74

0.42

0.160.3

0.16

0.53

0.730.42

0.20.65

0.380.12

0.47

0.07

0.03 0.57

0.12

0.490.710.82

0.4

157˚44’00”W 157˚42’45”W30” 15”

157˚44’00”W 157˚42’45”W30” 15”

21˚2

5’30

”21

˚25’

00”

21˚25’ 30”21˚25’ 00”

line-intercept transect percent living coral value

hardgroundssand

Living Coral<15%

15-25%

25-40%

40-75%

>75%0.730.851.0

0.76

0.760.58

0.620.74

0.44

0.54

0

0

00

0

00

0.67 0.74

0.42

0.160.3

0.16

0.53

0.730.42

0.20.65

0.380.12

0.47

0.07

0.03 0.57

0.12

0.490.710.82

0.4

157˚44’00”W 157˚42’45”W30” 15”

157˚44’00”W 157˚42’45”W30” 15”

21˚2

5’30

”21

˚25’

00”

21˚25’ 30”21˚25’ 00”

Multi-Spectral Percent Living Coral (Multi-Spectral Percent Living Coral (YY488/551488/551) Map) Map

0.10

Accuracy Assessment of Multi-Spectral Percent Living Accuracy Assessment of Multi-Spectral Percent Living Coral MapCoral Map

•R = 0.73, producers accuracy to # reference points•Re-sampling loss of detail in 40-75%

38%

2% 12%3%

15%

7%

25%

sand1,500,000 m2

hardgrounds70,000 m2

<15% living coral

500,000 m215-25% living

coral70,000 m2

25-40% living coral

600,000 m2

40-75% living coral

300,000 m2>75% living

coral1000,000 m2

Substrate DiversitySubstrate Diversity

OutlineOutline

Introduction• Study Site• Data Collection

Methods• Data Processing• Radiative Transfer Theory• Depth and Bottom-type by Band Difference• Applying the Model

Results• Depth Predictions• Percent Living Coral Predictions

Conclusion

ConclusionConclusion Radiative transfer model can be used to

normalize several multi-spectral images

Bathymetry and percent living coral is predicted with 488 nm and 551 nm

It may be possible to map change through time

Tuesday,June Tuesday,June 12, 200112, 2001