[A-CG38] Coastal Ecosystems 2. Coral reefs, seagrass and macroalgal beds, and mangroves

[ACG38-05] Mapping mangroves species using UAV-LiDAR and high-resolution satellite images

*Mohammad Shamim Hasan Mandal¹, Takashi Kanda², A.T.M. Zinnatul Bassar^3,4, Masako Dannoura⁴, Rempei Suwa¹ (1.JIRCAS, 2.TARF, JIRCAS, 3.Begum Rokeya Univ., 4.Kyoto Univ.)

Keywords:Blue carbon, AGB, Image classification, Machine Learning, Deep Learning

Accurate mangrove species mapping is crucial for estimating mangrove blue carbon and their conservation. Airborne light detection and ranging (LiDAR) point clouds data are invaluable for mapping tropical and temperate forests. However, such studies on mangroves are still few. Using extensive field surveys, LiDAR point cloud and multispectral WorldView-2 data the current study presents an improved workflow on mangrove mapping. The study site Miyara River mangroves, Ishigaki Island, Japan is a small (about 3.61 km²) riverine mangrove dominated by Bruguiera gymnorhiza (BG) and Rhizophora stylosa (RS). Field measured mean above-ground biomass (AGB) of BG dominated plots were higher (about 250 Mg/ha) than the RS dominated plots (about 180 Mg/ha), due to smaller RS trees. LiDAR point cloud data were related to the field estimated tree height and AGB. Next, we used three classification algorithms Random Forest (RF), Support Vector Machine (SVM) and Convolutional neural network (CNN) to classify mangrove species based on various features. LiDAR height profiles were a crucial factor for all three classification algorithms. Our finding further highlights the need for integrating LiDAR observations for mangrove monitoring and management.