MAPPING ABOVEGROUND FOREST BIOMASS CARBON STOCK BY USING SATELLITE IMAGE AND NFI DATA: A COMPARISON BETWEEN κNN AND REGRESSION TREE MODEL

摘要

To achieve quantitative information of aboveground forest biomass carbon stock by remote sensing approach, a number of methods have been conducted. This study is to examine application of regression tree (tree decision) and A-Nearest Neighbor (&NN) algorithm for forest carbon stock estimation of Chungnam province in South Korea. Dataset used for this research includes Landsat Thematic Mapper (TM) images and field data from 5 * National Forest Inventory (NFI). As a result, total above forest carbon stock estimated by κNN (20,467,652.900 tonC) model is closer to the number given by Korean Institutes of Forest than regression tree's number (20,239,247.239 tonC), however RMSE from latter algorithm is less than the former, 19.168 tonC/ha and 20.063 tonC/ha, respectively.