Evaluation of mobile lidar scanning and associated workflows for estimating structural attributes in mixed-conifer forests

Pelak, Johnathan Robert (2022) Evaluation of mobile lidar scanning and associated workflows for estimating structural attributes in mixed-conifer forests. Masters thesis, Northern Arizona University.

Text Pelak_2022_evaluation_mobile_lidar_scanning_associated_workflows_estim.pdf - Published Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Remote sensing techniques, such as light detection and ranging (lidar), have commonly been applied to forestry projects where traditional measurements of forest structure are too time consuming or costly to implement. Advances in lidar technology have led to increased interest in assessing the suitability of terrestrial mobile lidar scanning (MLS) for quickly assessing forest structure, but this type of scanner has not been tested in complex forest structural conditions in the southwestern United States. The highly variable and often complex forest structure that characterizes the dry mixed-conifer forests on the Mogollon Rim in Arizona make it an ideal forest type to test the potential advantages and limitations of using MLS to assess forest structure in complex systems. These forests have seen shifts in forest structure generally attributed to the increase of Euro-American settlers in the mid- to late-1800s, that have increased the need for ecological restoration to improve their resilience to severe disturbances. If MLS could produce accurate estimates of forest structure in dry mixed-conifer forests, it could decrease the resources needed to effectively monitor forest change and inform the adaptive management cycle employed by land managers to monitor the effectiveness of ecological restoration treatments. In this study, I tested the ability of a commercial MLS unit to accurately assess structure in dry mixed-conifer forests by answering the following research questions: 1) How does mobile lidar compare with traditional forest inventory techniques in estimating forest structural metrics in dry mixed-conifer forests using a range of accessible processing tools? 2) Which elements of dry mixed-conifer forest structure affect mobile lidar accuracy in characterizing forest structure? 3) Does the application of an eigenvalue point cloud filter in the processing workflow increase the accuracy of mobile lidar-derived measurements of forest structure? I used six processing workflows utilizing three processing tree identification tools and a point cloud filter to derive forest inventory metrics from MLS point clouds. MLS-derived trees were also matched to field-observed trees using a published tree matching algorithm. I then calculated error between MLS estimates and field-observed inventories and investigated the potential role of some structural conditions and species composition as sources of increased MLS error. All six workflows generally underestimated trees density (trees per hectare), and overestimated basal area and quadratic mean diameter. On average, the number of trees less than 15 cm in diameter at breast height (DBH) were underestimated by at least 400 trees per hectare or 16 trees per 0.04-hectare plot by all MLS workflows. Trees 15 – 45 cm DBH were overestimated by the workflows using the spanner R package while the LiDAR360 commercial software and custom processing workflow using clustering tools in the dbscan R package overestimated the number of trees greater than 60 cm DBH. Landscape-level omission and commission percent error rates ranged from 40.7 to 84.6%. The three workflows in which a verticality filter was applied to the point clouds produced more accurate trees density estimates but increased basal area error. The ability to classify the species and live or dead status of identified trees was not evaluated due to the lack of readily available processing tools for MLS point-clouds. Plot composition and structural attributes, such as the presence of New Mexico locust (Robinia neomexicana) and bigtooth maple (Acer grandidentatum), higher overall tree density, and higher density of small trees (<15 cm DBH) appeared to increase MLS estimation errors. The number of trees with canopy base heights at or below 1.4 m also increased errors. My work expands existing MLS literature to fill the knowledge gap of MLS performance in complex forest systems. MLS is emerging as a new tool for forestry applications, so a thorough understanding of how these tools can be applied and what further refinements are needed to make them viable are essential to the future application of MLS and related technologies. To increase the capability of MLS as a tool for assessing forest structure in dry mixed-conifer forests, additional work is necessary to develop tree-segmentation workflows that can better identify stems in structurally complex forests and that can identify tree species and live or dead status.

Item Type:	Thesis (Masters)
Publisher’s Statement:	© Copyright is held by the author. Digital access to this material is made possible by the Cline Library, Northern Arizona University. Further transmission, reproduction or presentation of protected items is prohibited except with permission of the author.
Keywords:	Dry mixed-conifer forest; forest technology; Lidar; MLS; Mobile lidar scanning; Terrestrial lidar; Mogollon Rim; Arizona; Forest structure
Subjects:	S Agriculture > SD Forestry
NAU Depositing Author Academic Status:	Student
Department/Unit:	Graduate College > Theses and Dissertations College of the Environment, Forestry, and Natural Sciences > School of Forestry
Date Deposited:	26 May 2023 17:07
Last Modified:	26 May 2023 17:07
URI:	https://openknowledge.nau.edu/id/eprint/5916

Actions (login required)

IR Staff Record View

Downloads