Author

Nahid Carter

Abstract

Remote sensing provides a variety of methods for classifying forest communities and can be a valuable tool for the impact assessment of invasive species. The emerald ash borer (Agrilus planipennis) infestation of ash trees (Fraxinus) in the United States has resulted in the mortality of large stands of ash throughout the Northeast. This study assessed the suitability of multi-temporal Worldview-2 multispectral satellite imagery for classifying a mixed deciduous forest in Upstate New York. Training sites were collected using a Global Positioning System (GPS) receiver, with each training site consisting of a single tree of a corresponding class. Six classes were collected; Ash, Maple, Oak, Beech, Evergreen, and Other. Three different classifications were investigated on four data sets. A six class classification (6C), a two class classification consisting of ash and all other classes combined (2C), and a merging of the ash and maple classes for a five class classification (5C). The four data sets included Worldview-2 multispectral data collection from June 2010 (J-WV2) and September 2010 (S-WV2), a layer stacked data set using J-WV2 and S-WV2 (LS-WV2), and a reduced data set (RD-WV2). RD-WV2 was created using a statistical analysis of the processed and unprocessed imagery. Statistical analysis was used to reduce the dimensionality of the data and identify key bands to create a fourth data set (RD-WV2). Overall accuracy varied considerably depending upon the classification type, but results indicated that ash was confused with maple in a majority of the classifications. Ash was most accurately identified using the 2C classification and RD-WV2 data set (81.48%). A combination of the ash and maple classes yielded an accuracy of 89.41%. Future work should focus on separating the ash and maple classifiers by using data sources such as hyperspectral imagery, LiDAR, or extensive forest surveys.

Library of Congress Subject Headings

Remote-sensing images--Data processing; Image processing--Digital techniques; Forests and forestry--Remote sensing; Forest mapping; Classification--Data processing

Publication Date

8-8-2013

Document Type

Thesis

Degree Name

Environmental Science (MS)

Department, Program, or Center

Thomas H. Gosnell School of Life Sciences (COS)

Advisor

Hane, Elizabeth

Advisor/Committee Member

Korfmacher, Karl

Advisor/Committee Member

van Aardt, Jan

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: G70.4 .C37 2012

Campus

RIT – Main Campus

Plan Codes

ENVS-MS

Share

COinS