We examine whether hyperspectral airborne imagery enables better mapping of benthic exposure (seagrass v. not-seagrass), and whether it allows us to distinguish between different kinds of seagrass (i.e., genus level mapping). We demonstrate 1) that hyperspectral benthic exposure mapping is more accurate than multispectral mapping, and 2) that it is possible to very accurately map seagrass genera with hyperspectral imagery. Some interesting imagery processing methods are discussed, and recommendations made for future mapping.
Marine benthos is a vital part of the marine ecosystem, and a large carbon sink and store. However, it is not well mapped or monitored. Here we evaluated the potential of hyperspectral imagery to improve the accuracy of basic benthos mapping (bare substrate versus non-substrate), and to discriminate the dominant seagrass genera (i.e., Amphibolis spp., Posidonia spp.) and bare sediment. The study was conducted along the Adelaide Metropolitan coast, but results should be generalisable to other similar regions globally. Some pre-processing was necessary to remove sun-glint, and normalise depth related effects on reflectance. Finally, the mapping was compared to previous mapping (2013) to determine the occurrence of seagrass meadow regrowth or loss. The work was very successful. For basic benthos mapping we demonstrated very high accuracy (98 % overall accuracy). For genus level mapping accuracy was very high in areas of ideal image quality (93 %), and high in areas of lower image quality (80 %). Three large areas of seagrass regrowth and one large area of seagrass loss were detected. The 2013 benthos mapping had no formal accuracy assessment, but qualitatively was lower accuracy than the basic benthos mapping presented here. We have demonstrated the potential of hyperspectral imagery for more accurate and more detailed benthic mapping. These results may provide new insight into the drivers of benthos loss and regrowth, and hence enable better directed and more effective management of these precious resources.
Time & room
2.45 pm–3.00 pm in City Room 3
Dr Ken Clarke
School of Biological Sciences, The University of Adelaide
Date & venue
Friday, 25 October 2019 at the Adelaide Convention Centre