Remote sensing model development for seagrass and mangroves carbon stock mapping

Pramaditya Wicaksono,

2015 | Disertasi |

ABSTRACT Among vegetated coastal habitats, seagrass and mangroves are the most effective blue carbon sink. The ability of these habitats to sequester CO2 from the atmosphere surpasses any terrestrial ecosystems considerably. However, continuous and spatial and temporally explicit information about the carbon stock of these habitats are currently lacking at various levels of scales and user needs, which is mainly due to the limitation and difficulties of performing field data collection. This highly demanding carbon stock information can only be obtained using remote sensing. The aims of this research are to map seagrass and mangroves carbon stock using remote sensing, develop remote sensing model for seagrass and mangroves carbon stock mapping, and estimate the total carbon stock of seagrass and mangroves carbon stock in Karimunjawa Islands. Worldview-2, ALOS AVNIR-2, ASTER VNIR, Landsat 5 TM, and Landsat 7 ETM+ images are selected to represent data at various level of mapping scales. Several geometric and radiometric approaches were applied to isolate seagrass and mangroves reflectance. Seagrass and mangroves carbon stock were mapped using semi empirical modeling. The input for seagrass carbon stock modeling are deglint bands, water column corrected bands, and PC bands, while for mangroves carbon stock modeling the inputs are vegetation index, PC bands, and image fraction from linear spectral unmixing, and spectral angle mapper. Field data are collected to build the model as well as to understand the relationship between seagrass and mangroves biophysical properties, which is also very important to construct the framework of the mapping. Remote sensing model for seagrass and mangroves carbon stock mapping was developed by considering the underlying framework, the effective input data, the assumptions and limitations during the mapping, the methodology to map the carbon stock, and the expected accuracy. The resulting model covers the logics behind the ability of remote sensing to explain the variation of seagrass and mangroves carbon stock and the technical mapping framework. The results show that remote sensing can be used to map seagrass and mangroves carbon stock. Seagrass carbon stock can be mapped with maximum accuracy of 49.23% (SE = 6.64 gC/m2), 55.64% (SE = 59.52% gC/m2), and 92.9% (SE = 17.41 gC/m2) for the AGC, BGC, and sediment carbon stock. For mangroves, the AGC, BGC, and soil carbon stock can be mapped with 77.81% (SE = 5.71 kgC/m2), 60.82% (SE = 2.48 kgC/m2), and 82.5% (SE = 1.22 kgC/m2) accuracy respectively. From the model, total ecosystem carbon stock in seagrass and mangroves in Karimunjawa Islands are estimated to be around 622.9 and 181,195.88 tones of organic carbon respectively. The availability of remote sensing model for seagrass and mangroves carbon stock mapping is very important to provide better understanding about their spatiotemporal carbon dynamics distribution. In addition, carbon stock maps are beneficial to assist various management activities including determining protected zones, assisting the process of mangroves conservation, setting the baseline for natural resource inventory, and evaluating management impacts. Keywords: seagrass, mangroves, carbon stock, multispectral, remote sensing model

Kata Kunci :