SPECTRAL UN-MIXING OF SEAGRASS AND SEAWEED IN PHUKET, SOUTHERN THAILAND

摘要

Coastal ecosystems such as mangroves, tidal salt marshes, seagrass meadows, and seaweeds store considerable load of carbon in their biomass and soils and act as sink of atmospheric carbon dioxide through primary production. However, it is regarded that there is uncertainty about those global rates, due to uncertainties in their areal extent as well as variability in carbon burial rates among individual ecosystems, especially seagrass meadows. Therefore, seagrass species distribution and their areal extent are crucial for the robust estimates of carbon capture and storage. In the context, satellite data can be the most practical tool for habitat mapping of the coastal ecosystems. In the past study, we developed a mapping method of seaweed and seagrass habitats in Thailand by using Landsat-8 and resulted that the developed method correctly classifies benthic types under condition that the sea water is clear for that sediment is not run-off, however, the method classifies incorrectly in turbid water. Further study was required to examine if Landsat-8 data has a capability to classify the species of seagrasses and seaweeds spectrally and how the capability changes in different water conditions such as water depth and quality. In this study, applying linear spectral mixing and a radiative transfer simulation to reflectance spectra of seagrasses, seaweeds, and sands measured in situ, converting the spectra into Landsat-8 DN values, it was examined whether Landsat-8 has a capability to classify seagrass and seaweed species. As a result, it was shown that Landsat-8 has the capability under constraints that compared spectra are both equally mixed with sand and equally affected by water. Since individual differences were not considered, further study is needed to examine whether individual differences are sufficiently smaller than differences among species and unknown spectrum from real Landsat-8 data can be identified.