RSUSCI-2021 & RSUSOC-2021
IN21-074 Application of Geospatial Machine Learning Model for Above Ground Biomass Estimation in Mangroves Forest
Presenter: Ting Ting Han
-, Faculty of Environment and Resource Studies, Mahidol University
Abstract
The dangers and difficulties that occur during ground-based studies of mangrove forests present an opportunity to develop cheap, accurate, and easy-to-use remote sensing methods. In this study, an Unmanned Aerial Vehicle (UAV) uses Very High Resolution (VHR) imagery to determine the carbon emissions stored as Above-Ground Biomass (AGB) in a mangrove forest in Klong Khon, Thailand. Four 100 m2 plots were used to develop a model that uses tree height and crown area to estimate Diameter at Brest Height (DBH) (Deviance = 76.0%). The UAV uses Structure from Motion (SfM) to determine mangrove height and crown area to then model DBH. Two Variable Window Filtering (VWF) algorithms were applied to UAV imagery to detect treetops and crown delineation. Power regression and lower limit VWF models were developed using the relationship between ground measured tree height and crown diameter. Ground- and UAV-based dendrometric parameters were compared with ground-based measurements to determine their accuracy.