Year of Award

2017

Document Type

Thesis

Degree Type

Master of Science (MS)

Degree Name

Resource Conservation

Department or School/College

Department of Ecosystem and Conservation Sciences

Committee Chair

Dr. Cara Nelson

Commitee Members

Dr. Andrew Larson, Dr. Jon Graham

Keywords

landscape-scale, forest structure, forest spatial pattern, reference models, LiDAR

Publisher

University of Montana

Subject Categories

Natural Resources and Conservation

Abstract

Successful restoration of degraded forest landscapes requires reference models that adequately capture structural heterogeneity at multiple spatial scales. Field-based methods for assessing variation in forest structure are costly and inherently suffer from limited replication and spatial coverage. LiDAR is a more cost-effective approach for generating landscape-scale data, but it has a limited ability to detect understory trees. Increased understanding of appropriate height cut-offs for trees to be reliably included in LiDAR-based analysis could improve applications of LiDAR to assessments of landscape-scale forest structure. Toward that end, I investigated the effect of varying tree-height criterion (minimum height cutoffs of 6, 9, 12, 15, and 18 m) on the accuracy of LiDAR for estimating forest structure and spatial pattern in forests of the Sierra de San Pedro Martir National Park, Baja, Mexico. In order to increase the utility of the analysis, LiDAR trees were identified using a widely-available processing tool (FUSION’s TreeSeg). Accuracy was measured as the similarity between field-measured and LiDAR-detected tree datasets and was assessed for overall number of trees, spatial tree density maps, and a set of variables related to forest structure and spatial pattern. I found that removing trees less than 12 m in height increased correlation between LiDAR- and field-based spatial maps of tree density and strongly reduced differences in estimates of forest structure and spatial pattern. Although the frequency of small, medium, and large tree clumps was always underestimated by LiDAR-detected trees, the 12 m minimum height cutoff detected more of the large tree clumps than taller height cutoffs and provided estimates of forest structure and spatial pattern that were more similar to those derived from field data. The 12 m height cutoff also successfully captured structural variation across the study landscape: canyons, shallow northerly, and shallow southerly slopes were structurally similar, having larger and more abundant trees than steep northerly slopes, steep southerly slopes, and ridges. Methods developed here should be useful to managers interested in using LiDAR to characterize distributions of large, overstory trees without the need for extensive complementary field data and specifically for the development of landscape-scale reference models for forest management and restoration.

Share

COinS
 

© Copyright 2017 Haley L. Wiggins