Estimating Forest Structure from LiDAR and High Spatial Resolution Imagery for the Prediction of Succession and Species Composition

Loading...
Thumbnail Image

Authors

Van Ewijk, Karin Y.

Date

2015-04-28

Type

thesis

Language

eng

Keyword

Forestry , Succession , LiDAR , Tree species composition , Remote Sensing

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

Light detection and ranging (LiDAR) and high spatial resolution data combined with advanced statistical techniques have, within the last decade, contributed significantly to advances in the development of enhanced Forest Resource Inventories (FRI). The goal of this research was to explore how these data sources and statistical techniques could be utilized for predicting successional stages and species composition within a complex temperate forest ecosystem in Ontario. This research also explored the possibility of generating tree lists from these same data sources that would forecast equivalent future forest conditions compared to in situ collected FRI data. For the characterization of vertical structure within forest stands a new LiDAR metric was developed, i.e., the Vertical Complexity Index (VCI). Logistic regressions were then applied to predict successional stages while boosted regression trees were adopted for the quantification of relative abundance of upper canopy species using LiDAR and high spatial resolution data. k-Nearest Neighbor imputation was used for generating individual juvenile tree information from LiDAR whereas adult tree information was generated from: i) an individual tree crown (ITC) classification; and ii) from predicted stem density and species’ relative abundance. Successional stages were well predicted using LiDAR variables (i.e., VCI, Lorey’s height and standard deviation of height) with a classification accuracy (Khat) of 86%. Average prediction accuracy was 0.71 when LiDAR variables related to biotic and disturbance processes were included. Correlations between in situ and imputed juvenile tree information were moderate, ranging from 0.50 to 0.69. Stem density model fit (adj. R2) was 0.51 for conifer and 0.74 for hardwood stands. As for generating adult tree lists, ITC significantly underestimated stem density while both approaches underestimated species composition. This research clearly demonstrated that LIDAR variables that capture structural forest attributes can successfully be used to characterize structurally distinct successional stages and upper canopy species’ abundances in landscapes with limited topographical variation. Juvenile trees were more difficult to characterize with LiDAR variables as was the ability to generate tree lists. However, this research provides insights how to advance the characterization of juvenile trees and develop tree lists from LiDAR and high spatial resolution data.

Description

Thesis (Ph.D, Geography) -- Queen's University, 2015-04-28 14:52:52.006

Citation

Publisher

License

Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
Creative Commons - Attribution - CC BY
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.

Journal

Volume

Issue

PubMed ID

External DOI

ISSN

EISSN