Estimation of forest carbon stock using satellite imagery and NFI data - comparing ANN algorithm and regression model

摘要

k-Nearest Neighbor (kNN) algorithm and regression model have been widely used for a variety of forest parameter estimation and mapping application due to its intuitiveness and ease of use. The objective of this study is to comparing both algorithms for estimation of aboveground carbon stock in Danyang-Gun, South Korea. Field data from 5th NFI and Landsat TM satellite image were used as dataset. Additionally, various ratio images, such as vegetation indices, topographic effect correction indices, and spectral angle indices, were generated and compared to the Landsat TM original bands. As a result, kNN algorithm and Landsat TM original bands were determined to be a suitable method and dataset for forest carbon stock estimation in Danyang-Gun, respectively.