Quantification of submerged seagrass total aboveground biomass for Malaysian coastal waters using remote sensing data

Abstract

Multi-species of seagrass forms dense benthic communities in the coastal clear (Case 1) and less clear water (Case 2) in Malaysia. There are two types of seagrass, i.e. intertidal and submerged, which can easily be found in Malaysia. Satellite remote sensing data is an effective tool to be used in many marine applications, including monitoring seagrass distribution at large area coverage. The emphasizes of this thesis is to determine the best two steps satellite-based approach for mapping submerged seagrass and quantifying aboveground biomass at Merambong area and Pulau Tinggi, Johor. Multi-platforms satellite data that has different data specifications have been used at both Case 1 (water dominated by phytoplankton) and Case 2 (water concentrated with water floating substances and sediments). The satellite data used for Merambong are GeoEye-1, Worldview-2, ALOS AVNIR-2, Landsat-8 OLI and Landsat-5 TM, while the satellite data for Pulau Tinggi are Worldview-2, ALOS AVNIR-2, Landsat-8 OLI and Landsat-5 TM. The robustness of seagrass detecting techniques, namely Depth Invariant Index (DII) and Bottom Reflectance Index (BRI) on remotely sensed data at different water clarity have been tested. Both techniques require measurement of radiance, deep-water radiance and ratio of attenuation coefficients while BRI needs few additional elements from nautical chart and tide calendar to attain information of the sea bottom depth during satellite passes. Ground truth data has intensively been collected at both study areas to validate and assess the finding of this study. Comparative assessment and analysis between both techniques revealed that BRI is best to be used on Landsat-8 OLI (93.2% user accuracy) in Case 2 water while (95.0% user accuracy) in Case 1 water to identify submerged seagrass. An empirical model has been developed to devise quantification of aboveground biomass and the temporal changes by associating insitu seagrass coverage data with BRI value on the satellite images. Submerged seagrass biomass quantification using remotely sensed data is feasible in Case 2 water at required scale and accuracy (>80%), depending on the field data sufficiency, technique and choice of satellite data. In conclusion, Landsat-8 OLI with 16-bits quantization level produces more accurate results than Worldview-2 and GeoEye-1. It is able to cover a large area of study, hence it is very useful for spatio-temporal seagrass biomass monitoring project by local policy makers and related agencies.