Date of Completion
5-8-2015
Embargo Period
5-7-2017
Advisors
Heidi M. Dierssen, Amy C. Burnicki, Mark Rudnicki
Field of Study
Natural Resources
Degree
Master of Science
Open Access
Campus Access
Abstract
The rapid pace at which the world’s forests are changing requires active monitoring and measuring of key characteristics indicative of forest health and functioning. However, plot level measurements are unable to provide the spatial and temporal coverage required to develop large scale, long-term records of forest characteristics due to high costs. This thesis describes two new methods for taking advantage of the wealth of data stored in the Landsat satellite archive for detecting disturbances and estimating forest height over a period of 25 years in a mixed boreal forest in Quebec, Canada. Using an object based approach multi-seasonal spectral trajectories were created and used to detect abrupt changes in spectral values indicative of disturbances events including harvesting activities and forest fires. Forest height was estimated by relating average wintertime reflectance to LiDAR derived measures of forest height. Results indicate high accuracy for disturbance detection (92%) and strong correlation between wintertime near infrared reflectance and forest height (RMSE: 0.77 – 1.33 m; R²: 0.70- 0.90).
Recommended Citation
Chlus, Adam, "Monitoring Long-Term Forest Dynamics Using Very Dense Landsat Time Series" (2015). Master's Theses. 752.
https://digitalcommons.lib.uconn.edu/gs_theses/752
Major Advisor
Daniel L. Civco