Date of Award

Spring 1-1-2019

Document Type

Thesis

Degree Name

Master of Arts (MA)

First Advisor

Jennifer K. Balch

Second Advisor

Thomas T. Veblen

Third Advisor

Colleen E. Reid

Abstract

Accurately mapping tree species composition is a critical step towards understanding post-disturbance forest recovery. The National Ecological Observatory Network (NEON) is a valuable source of open ecological data across the United States. NEON data include in-situ tree measurements along with hyperspectral, multispectral, and light detection and ranging (LiDAR) airborne remote sensing imagery. By linking these NEON data, this study explores the impact of training set preparation and preprocessing on coniferous tree species classification at the subalpine forest NEON site in Colorado. Pixel-based random forest machine learning models were trained using various reference sets with remote sensing raster data as descriptive features. The highest classification accuracy (73%) was achieved using polygons created with half the maximum crown diameter per tree. LiDAR features were found to be the most important, followed by vegetation indices. This work contributes to reproducible forest composition mapping efforts and the open ecological science community.

Share

COinS