Thesis Committee Dr. Nitin Kumar TripathiChairperson Prof. Seishiro KibeMember Dr. Wenresti GallardoMember 14-May-2008 MAPPING CHANGES IN THE MARINE ENVIRONMENT.

Thesis Committee Dr. Nitin Kumar Tripathi ChairpersonProf. Seishiro Kibe MemberDr. Wenresti Gallardo Member

14-May-2008

MAPPING CHANGES IN

THE MARINE ENVIRONMENT OF PHU QUOC ISLAND, VIET NAM

Ton Binh MinhRemote sensing and GIS FoS

- Area 593 km2, - 26 small islands - Phu Quoc Island; An Thoi- Tho Chau archipelagoes- Coastal Hydrologic condition for ecotourism

Phu Quoc Island

Natural Resources

Phu Quoc Island not only has potential for marine resources but also for ecotourism and relax

- Remote sensing and GIS are power tools for mapping and management of resources

- Surface- Under Water

- Marine resource are: Over exhausted and none plan exploited Fish, turtles, dugong, dolphins are on the verge of

extinction. - Unsustainable fishing techniques

(small mesh fishnets, cyanide, dynamite, flying raking, and Increasing number of fishermen, fishing vessels with close shore fishery activities).

- Detailed maps of marine habitats, changing rule of environment for monitoring marine environment management, conservation coastal ecosystems and sustainable development

Objective

- To detect the changes in sea grass beds, coral reefs during 2000-2004 and 1992-2007.

- To apply remote sensing and GIS to monitor and evaluate coastal resources in Phu Quoc Island.

Detail objectives- To map benthic communities: sea grass, coral (live and

dead), sand, rubble around Phu Quoc Island,

- To provide data for environmental and management of seagrass beds and coral reefs,

Bai BonAster 2004

An ThoiLandsat 2007

Study area- An Thoi- Bai Bon

Data base- Ancillary data,- Remote sensing data, - Field data.

- Ancillary data:

+ Previous reports.

+ Administration, transportation

+ Bathymetric depth+ Sediment maps

Data Surveillance and Collection tools- GPS, - Tape measurement (30m), - Digital camera, - Data form, - Secchi disk,- Diving equipment- Boat

Data base- Field data

Quadrats

20M

20M +

Field work techniquesLinear transect

Integration linear transect and quadrat for collecting data in seagrass area

=

1m

1m

Analyzing Image

Mean DN value of coral, seagrass, sand in four bands

Landsat 2007

0

10

20

30

40

50

60

70

80

90

1 2 3 4

Band

DN

val

ue

Coral

Seagrass

Sand

0

20

40

60

80

100

120

1 2 3Band

DN

va

lue

Coral

Seagrass

Sand

Mean DN value of coral, seagrass, sand in band 1, 2, 3

Aster 2004Landsat 2007

1. Image Registration - Image-to-image registration- Aster image in 2000. - WGS 84, UTM 48 N.

Digital image processing

2. Image masking - Segregating land and sea area - Near-infrared band (0.78-0.98μm) was used for masking

Water Column Correction

Landsat TM, Aster images (atmospheric correction)

Training site selection (uniform pixel, various in depth)

Band pairs selection

Extracting pixel value

Calculating VAR and COVARtraining site’s pixel in band i, j

Calculating coefficients attenuation (ki/kj)

Implementation depth-invariance to whole image(Water column correction image)

Water column corrected image

An Thoi area

Bai Bon area

Landsat 2007

Aster 2004,

Image Enhance

FCC, PCA, Band ratio

Landsat 2007 High OIF, as FCC (4,3,1)

Landsat 2007 Low OIF as true color composite (321)

Color composite - Optimum Index factor

Landsat 2007 low OIF as true color composite (3,2,1)

Landsat 2007 High OIF as FCC (4,3,1)

Optimum Index Factor

Accuracy Assessment - Site specific error matrix and Kappa analysis

Image Classification- Maximum Likelihood classifier, field data , WCC image- Class hierarchies and definition

0

500

1000

1500

2000

2500

3000

3500

2007

He

cta

0

500

1000

1500

2000

2500

3000

3500

2004

He

cta

0

50

100

150

200

250

300

350

400

450

500

2004

Sand zone

Dead coral

Live coral cover 01-10%


Rubble

Seagrass dens 70-80%



0

50

100

150

200

250

300

350

400

450

500

2004

Sand zone

Dead coral



Rubble




0

50

100

150

200

250

300

350

400

450

500

2004

Sand zone

Dead coral



Rubble




0

50

100

150

200

250

300

350

400

450

500

2004

Sand zone

Dead coral



Rubble




Classification in Bai Bon

0

50

100

150

200

250

300

350

400

450

500

2004H

ec

ta

0

50

100

150

200

250

300

350

400

450

500

2004

Sand zone

Dead coral



Rubble




Classification results in An Thoi using Aster 2004

0

50

100

150

200

250

300

350

400

450

500

2007

He

cta

0

50

100

150

200

250

300

350

400

450

500

2004

Sand zone

Dead coral



Rubble




Classification results in An Thoi using Landsat 2007

Map 2007:

Overall accuracy = 70%

KHAT = 64%,

Map accuracy

Map 2004

Overall accuracy = 77%

KHAT = 74%

Change detection

- Non site specific error matrix,

- Map overlay

Landsat imageDec -1992

Landsat image Jan 2007

Aster imageDec- 2000

Aster imageNov- 2004

Reefs, seagrass habitat map

1992


2007


2000


2004

Overlay Overlay

Reef habitat change in 2000-2004

Seagrass habitat change in 2000-2004

Reef habitat change in 1992-2007

Seagrass habitat change in 1992-2007

Non Site Post Classification Comparison

The overall accuracy = 42 % map 2007 agree 42%, with map 1992 Khat value = 29% category of map 2007 is same with map 1992 29%

Change detection analysis

Degeneration of live coral, seagrass in period 2000-2004, An Thoi.

Decrease quality of live coral, seagrass in period 2000-2004 in An Thoi

Degeneration of live coral, seagrass in period 1992-2007, An Thoi

Decrease quality of live coral, seagrass in period 1992-2007 in An Thoi

Degeneration of seagrass in Bai Bon

Regeneration of coral, seagrass in 2000-2004, An Thoi.

56.20

Regeneration of coral, seagrass 1992-2007, An Thoi.

Regeneration of seagrass in Bai Bon

Comparing a part of Duong Dong town by using Landsat 25m, Aster 15m and Quick Bird 2.4m imageries

Benthic Habitats Mapping Using Medium and High Resolution Image

Extending Quick Bird Imagery to Detect Bottom Type

-The Gram Schmidt Spectral Sharpening- Sharpen multispectral data of QB at 2.4 m to panchromatic band at 0.6 m resolution.

- Gram-Schmidt techniques:

+ Not limited to the number of bands that can be processed at one time.

+ Preserved the spectral characteristics of lower spatial resolution multispectral data in the higher spatial resolution

Data fusion technique for Quick bird imagery

Fused image with 0.6m resolution. Using Gram-Schmidt Spectral Sharpening band combination band 3,2,1.

Fused image with 0.6m resolution. Using Gram-Schmidt Spectral Sharpening, band combination 4,3,2.

Extending Quick Bird Imagery to Detect Bottom TypeData fusion technique for Quick bird imagery

Quick bird data potential for detail mapping benthic community

Comparing fusion image (2,4,3) with Aster(1,2,3)fusion image and original image

Applying enhance technique to Quick bird data

True colour combine, square root enhance, various information

Aster imagery because it lacks blue band.

Band ratio: composite of 3/1, 3/2, 2/1 (RGB)

•. Several light yellow areas and light green appeared in reefs and terrestrial rock. Possible these have developed of marine algae or seagrass.

Conclusion1. Remote sensing and GIS are powerful tools for

conservation and management marine resource.• as they provide data of difficult location and powerful analysis

tools.

2. The coral reefs and seagrass beds map in 2007 and 2004 has been created using bands 1, 2 and 3 of Landsat and bands 1, and 2 of Aster.

3. RS techniques and image processing were used to enhance image reflectance and spectral characteristics.

4. Water column correction was employed to remove suspended matter effect.

5. Maximum likelihood classification yielded high overall accuracy =70% (Landsat 2007) and = 77 % (Aster 2004)

6. Live coral and seagrass changes were detected which can be very useful for managing marine resources

7. The comparison of the change area between 1992-2007 and 2000-2004 was not significant.

8. Degeneration and regeneration were successfully processed. Seagrass and coral reefs in study area are degenerating. Possible, sea environment of PQ is changing to worse status.

9. Quick Bird imagery with high resolution is better for mapping benthic communities in detail;

• it is optical imagery thus, one scene covers small area, • cloud effect,• Image is more expensive than moderate resolution

satellite image.

Conclusion

Thank you for your kind attention!

I also thank to my Advisor, Examination committees members,

Wetland Alliance Program!

Thesis Committee Dr. Nitin Kumar TripathiChairperson Prof. Seishiro KibeMember Dr. Wenresti GallardoMember 14-May-2008 MAPPING CHANGES IN THE MARINE ENVIRONMENT.

Documents