Mapping and Monitoring Coral Reefs in Vietnam’s Coastal Waters from High-Spatial Resolution Remote Sensing and Field Survey Data

Van Dien Tran (2010). Mapping and Monitoring Coral Reefs in Vietnam’s Coastal Waters from High-Spatial Resolution Remote Sensing and Field Survey Data MPhil Thesis, School of Geography, Planning and Environmental Management, The University of Queensland.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
s41295041_MPhil_Abstract.pdf Thesis Abstract application/pdf 17.85KB 5
s41295041_MPhil_Finalthesis.pdf Final Thesis application/pdf 16.82MB 31
Author Van Dien Tran
Thesis Title Mapping and Monitoring Coral Reefs in Vietnam’s Coastal Waters from High-Spatial Resolution Remote Sensing and Field Survey Data
School, Centre or Institute School of Geography, Planning and Environmental Management
Institution The University of Queensland
Publication date 2010-09
Thesis type MPhil Thesis
Supervisor Stuart Phinn
Christiaan Roelfsema
Total pages 222
Total colour pages 78
Total black and white pages 144
Subjects 05 Environmental Sciences
Abstract/Summary Vietnam’s coastal waters contain a wide range of reef diversity and structure. These reefs support over 350 species of hard corals and are stressed by a variety of threats. Previous coral reef studies in Vietnam examined coral species, cover, and stress level at a limited number of reefs. Satellite remote sensing potentially provides a means to map the composition, condition and dynamics of shallow tropical coral reefs from the scale of coral patches (several square metres) to regional and global scales, both in a continuous and cost-effective manner. The majority of previous works on mapping ecological properties of coral reefs worldwide from satellite and airborne images have been implemented on reefs in clear and shallow waters. There are no defined image-based techniques suitable for mapping and monitoring coral reefs in Vietnam's coastal waters due to their variable water clarity and small size. There are no detailed benthic composition maps for most coral reefs in Vietnam and there is a lack of accurate information of the benthic cover and status of those reefs. Therefore, knowledge of the extent, composition, condition of coral reefs in Vietnam is very limited. In this study, a method for mapping the benthic composition of coral reefs and monitoring its change at selected reef sites in Vietnam’s coastal waters from high-spatial, multi-spectral satellite image data was developed and tested. Van Phong Bay and Nha Trang Bay of Khanh Hoa Province were selected for mapping and monitoring change of benthic composition of coral reefs. Three reef sites in the bays (Dam Mon, My Giang, and Nha Trang), representative of coastal reef types, water clarity, and development pressures were studied. To achieve the project aim, four objectives were conducted. They were to: (1) Review and critically assess existing approaches for coral reef mapping from satellite images and develop a suitable image processing and validation sequence for mapping benthic composition in Vietnam's coastal waters; (2) Apply the image processing sequence developed in objective 1 to ALOS AVNIR-2 (2008), GeoEye-1 (2009), QuickBird (2006), and IKONOS (2002, 2008) satellite images to produce maps of benthic composition at three study sites (Dam Mon and My Giang in Van Phong Bay, and Nha Trang Bay); (3) Validate and critically assess the output maps of benthic composition; and (4) Detect, identify and explain changes in benthic composition of coral reefs in the My Giang study site during 2002-2008. A field survey was conducted to collect data on ground control points, water clarity, depth, and benthic composition and cover status for three study areas. This was done for calibration and validation purposes. GPS Photo-Link software was used to link the photos to their position recorded by GPS. Photos of benthic cover were processed using Coral Point Count with Excel extensions software (CPCe 3.6) to quantify the benthic composition of each photographed area. A hierarchical classification scheme of benthic composition in the study sites was generated based on existing classification schemes and benthic habitats in the study area. Four satellite image data of IKONOS, QuickBird, GeoEye-1, and ALOS AVNIR-2 satellite imagery at three reef sites in Van Phong and Nha Trang Bays were used. These images were corrected for radiometric, atmospheric, sun-glint, and depth effects. The combination of three atmospheric corrected bands (blue, green, and red bands), the second band of principal component analysis (PCA), and depth-invariant index of blue and green bands were used for supervised classification at finer level of description (13-14 classes). The benthic classes were grouped to 7-8 classes (living coral, dead coral, mixed coral/algae, brown algae, seagrass, sand, rock and pavement, rubble) and 4-5 classes (macroalgae, seagrass, coral, substrate, mixed coral/algae) to generate benthic composition maps at moderate and broad levels of description respectively. The accuracy of three classification results (Maximum Likelihood, Mahalanobis Distance, and Minimum Distance) was assessed to select the most accurate output benthic composition maps. Overall accuracy of the maximum likelihood classified GeoEye-1 image in Dam Mon with 9, 7, and 4 classes was 73.29, 76.98%, and 88.02% respectively. Overall accuracy of the maximum likelihood classified IKONOS image in My Giang with 14, 8, and 5 classes was 53.86%, 65.09%, and 69.85% respectively. Overall accuracy of the Mahalanobis Distance classified QuickBird image in Nha Trang with 13, 8, and 5 classes was 70.59, 79.50%, and 81.51% respectively. Overall accuracy of the Maximum Likelihood classified ALOS AVNIR-2 image in Dam Mon with 9, 7, and 4 classes were 65.77%, 71.13%, and 79.54% respectively. Overall accuracy of the Maximum Likelihood classified image in My Giang with 14, 8, and 5 classes was 62.99%, 80.21%, and 83.21% respectively. Overall accuracy of the Mahalanobis Distance classified image in Nha Trang with 13, 8, and 5 classes were 52.19%, 60.40%, and 81.82% respectively. The spectral mixing of benthic features within an image pixel and heterogeneous pixels within a training area are the primary reasons for low classification accuracy, while classes that were consistently mapped with high accuracy were sand and mixed coral/algae due to their large and homogenous features. The overall accuracy of classified images increased when the number of benthic classes was reduced. Accuracy of classified images was acceptable for background to management planning. However, only accuracy of benthic composition maps at broad classification scheme (>80%) was suitable for habitat monitoring. Spatial resolution of satellite image, water clarity, and reef geomorphology were the main factors affecting overall accuracy of image classification. Combination of QuickBird sensor with Mahalanobis Distance classification algorithm was considered the best combination of sensor and algorithm for benthic composition mapping and monitoring in Vietnam’s coastal waters. Two IKONOS satellite images of the My Giang study area (04 April 2002 and 13 July 2008) were classified and used for change detection analysis. The two images were classified using the same training areas. The most accurate maps were derived by using Maximum Likelihood Classifier at the moderate and broad descriptive levels and used for change detection process. The observed real changes were seagrass cover and seasonal growth of macroalgae. False changes were due to misclassification. In terms of suitability of image data and processing techniques for mapping coral reefs in Vietnam, although this study only covered a limited set of sites, it did identify several key points. Firstly, spatial resolution of satellite imagery was identified as the most important factor influencing the benthic composition map accuracy. ALOS AVNIR-2 satellite imagery was considered suitable for mapping benthic composition at broad level of description (4-5 classes) in Vietnam. Mapping benthic composition at moderate level of detail (8-10 classes) required an image pixel size < 3m, such as QuickBird and GeoEye-1. Pixel-based classification method can be used to map benthic composition in Vietnam from high-spatial resolution satellite data at moderate descriptive level. Time series of high-spatial satellite imagery should be used for monitoring seasonal change in cover, density, above ground biomass of seagrass and macroalgae in coral reefs in Vietnam.
Keyword benthic composition mapping, remote sensing, coral reefs, seagrass, supervised classification, calibration, validation, high-spatial resolution, Vietnam’s coastal waters
Additional Notes 21, 27, 56, 73, 80, 83, 88, 92, 97, 101, 115, 116

Citation counts: Google Scholar Search Google Scholar
Created: Mon, 21 Mar 2011, 23:16:22 EST by Mr Van Dien Tran on behalf of Library - Information Access Service