Recently, due to the acceleration of global warming, an accurate understanding and management of forest carbon stocks, such as forest aboveground biomass, has become very important. The vertical structure of the forest, which is the internal structure of the forest, was mainly investigated by field surveys that are labor intensive. Recently, remote sensing techniques have been actively used to explore large and inaccessible areas. In addition, machine learning techniques that could classify and analyze large amounts of data are being used in various fields. Thus, this study aims to analyze the forest vertical structure (number of tree layers) to estimate forest aboveground biomass in Jeju Island from optical and radar satellite images using artificial neural networks (ANN). For this purpose, the eight input neurons of the forest related layers, based on remote sensing data, were prepared: normalized difference vegetation index (NDVI), normalized difference water index (NDWI), NDVI texture, NDWI texture, average canopy height, standard deviation canopy height and two types of coherence maps were created using the Kompsat-3 optical image, L-band ALOS PALSAR-1 radar images, digital surface model (DSM), and digital terrain model (DTM). The forest vertical structure data, based on field surveys, was divided into the training/validation and test data and the hyper-parameters of ANN were trained using the training/validation data. The forest vertical classification result from ANN was evaluated by comparison to the test data. It showed about a 65.7% overall accuracy based on the error matrix. This result shows that the forest vertical structure map can be effectively generated from optical and radar satellite images and existing DEM and DTM using the ANN approach, especially for national scale mapping.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited