Development of mapping methods for seagrass meadows in Malaysia by using landsat images

• Main Author: Hossain, Mohammad Shawkat
• Format: Thesis
• Language: English
• Published: 2015
• Subjects: Meadows; Meadow ecology; Grassland ecology
• Online Access: http://psasir.upm.edu.my/id/eprint/78360/1/FSPM%202015%2012%20ir.pdf; http://psasir.upm.edu.my/id/eprint/78360/

In Malaysia, seagrasses commonly inhabit shallow intertidal waters, lagoons, mangrove, coral reef and shoal in subtidal zones. The seagrass meadows that previously inhabited the coasts of Peninsular Malaysia (West Malaysia) and Sabah (East Malaysia) were extensive and now have become sparse or scattered due to natural and anthropogenic disturbances. Scientists and managers require a fast and low-cost approach to map and assess the habitat loss or potential damage to the seagrass resources. A critical review of 195 studies revealed that, in the past four decades, advances in the application of remote sensing (RS) methods, notably using Landsat imagery, were identified for seagrass habitat mapping. Mapping capabilities of Landsat were not tested in different tidal regimes, characterizing seagrass habitats in relation to water turbidity and depth regimes, and understand spatiotemporal dynamics from multi-date images due to lack of appropriate methods and data, including unresolved Scan Line Corrector (SLC)-off data gaps in Landsat 7 Enhanced Thematic Mapper (ETM)+ images. The specific objectives of this study were to: (1) characterize the research on seagrass RS methods through a review of the peer-reviewed literature, (2) assess the performance of Landsat 7 ETM+ SLC-off gap-filling methods and image enhancement techniques (ETs) with different water depths for methodological improvement in seagrass resource mapping and monitoring, and (3) apply mapping approach on seagrass habitat, water turbidity and relative water depth mapping and monitoring. A statistical assessment and an evaluation of the twelve SLC-off and four SLCon images covering seagrass meadows of Sungai Pulai estuary, Johor, Malaysia were conducted for data loss estimates. This analysis revealed a 2% systematic error attributable to a gradual increase of SLC-off stripes from the central nadir path towards the edge of the scene. The random shifting of SLCoff stripes caused a 11.07 ha Tanjung AdangLaut shoal (TALS) completely invisible. The next study was focused on assessment of the geometric and radiometric fidelity of images reconstructed by three potential gap-filling methods: (1) Geostatistical Neighbourhood Similar Pixel Interpolator (GNSPI), (2) Weighted linear regression algorithm integrated with Laplacian prior regularization, and (3) Local Linear Histogram Matching methods. The statistical measures of reconstructed images were in favor of the use of GNSPI as opposed to other gap-filling techniques for Sungai Pulai estuary seagrass distribution mapping. To assess the variation in performance of Landsat image enhancement for Sungai Pulai estuary seagrass maps, with different Mean sea-level tide heights (MSLTHs), a comparison was conducted between histogram equalization (HE) and manual enhancement (ME) based mapping approaches. The enhancement techniques were applied on true-color composites (red, green, and blue layer stacks) of thirty-three Landsat images (1989-2014) with MSLTHs between -0.281 and 0.234 m. The assessment found that ME substantially improved image quality compared to the HE with MSLTH thresholds between -0.218 m and -0.085 m. ME improved visual interpretation of Landsat images for seagrass detection and distribution mapping of Merambong shoal (MS), Tanjung Adang Darat shoal (TADS), TALS, and Seluyong mudflat (SMF). An integrated mapping approach, combining Landsat ME and seed pixel regional growing tool was examined to delineate seagrass boundary accurately. This approach was found suitable (with >75% overall accuracy) to map five classes-of-interest, i.e., seagrass, land, sand/mud, human settlement, coral and coral rubble for twelve islands of Coral Triangle Initiative (CTI), Sabah. The resulted map estimated seagrass areal coverage to be 274 ha, of which most seagrass meadows occurred in relatively shallow water areas covering about 158 ha. Issues of spatiotemporal changes in seagrass habitat were addressed through assessing the ability of the integrated mapping approach on multi-date Landsat images for mapping and monitoring seagrass resources of Punang-Sari estuary, Lawas, Sarawak, Pengkalan Nangka lagoon, Kelantan, and Paka lagoon, Terengganu of Malaysia. Applying this integrated approach on forty-nine Landsat 5, 7 and 8 images, produced an accurate multi-date seagrass habitat maps. Additionally, the results indicated that a noticeable loss of seagrass habitats at varying magnitude occurred between 2000 and 2014 for Punang-Sari, between 1998 and 2014 for Pengkalan Nangka, and between 1988 and 2014 for Paka. The natural event mainly sand shifting was the main cause of seagrass loss for Punang-Sari Lawas. Coastline change was identified as the most significant factor that caused seagrass spatial cover loss of the Pengkalan Nangka lagoon. The mapping approach and the map products produced in this study will provide a useful tool for detection, distribution mapping and monitoring changes of seagrass and associated resources for coastal management and conservation programs.