Author : Jian Yang
Publisher :
ISBN 13 :
Total Pages : 0 pages
Book Rating : 4.:/5 (133 download)
Book Synopsis Multi-Source Remote Sensing Data for Automated Extraction of Fine-scale Attributes in a Northern Hardwood Forest by : Jian Yang
Download or read book Multi-Source Remote Sensing Data for Automated Extraction of Fine-scale Attributes in a Northern Hardwood Forest written by Jian Yang and published by . This book was released on 2017 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Forest resources require careful management and planning as they are under increasing pressure to support wood industry and conservation needs. The management of structurally complex, uneven-aged, deciduous-dominated forests requires detailed and accurate data on fine-scale forest attributes (e.g., gap dynamics, crown sizes, species distributions). New advances in remote sensing techniques will transform traditional forest inventory practices that have relied on expensive ground-based measurements or less accurate interpretation of aerial photography. Recently, multiple sources of high spatial resolution remote sensing data have demonstrated great potential for automated extraction of fine-scale forest attributes. In this context, my PhD research aims to utilize multi-source high spatial resolution remote sensing data to develop methods for automated extraction of fine-scale forest attributes in deciduous-dominated forests, including canopy gap identification, crown delineation, and species classification. This study was carried out in Haliburton Forest and Wildlife Reserve, an uneven-aged, deciduous-dominated forest located in the Great Lakes-St. Lawrence region of Central Ontario, Canada. Specifically, the thesis first quantified the accuracy of canopy gap segmentation and classification by integrating optical and LiDAR data. Thereafter, the thesis proposed a novel method for individual tree crown (ITC) delineation, involving multispectral watershed segmentation and multi-scale fitting. Finally, the thesis explored the feasibility of using multi-seasonal WorldView-3 images to map tree species using the delineated ITCs. Results indicated that: (1) the independent use of LiDAR data performed the best segmentation of canopy gaps while the synergistic use of optical and LiDAR data provided higher classification accuracy for non-forest and forest gap identification; (2) the proposed multispectral watershed segmentation and multi-scale fitting method was able to produce ITC maps of higher quality; (3) the combined use of late-spring, mid-summer, and early-spring images substantially improved the accuracy of individual tree-based species classification. The goal of this study was to develop automated methods for extracting fine-scale forest attributes for operational purposes. Although the proposed methods were mainly designed for temperate deciduous-dominated forests, they could be implemented in other types of temperate or boreal forests, such as coniferous-dominated forests.