Back to search

University of Minnesota

Quantifying Surface Fuel Loading Using Terrestrial Laser Scanning

Abstract

dc:description.abstract

Surface fuel loading is a critical component of wildfire behavior, forest carbon storage, and understory ecosystem dynamics. However, traditional field inventory techniques are labor-intensive and prone to the introduction of error, while above-canopy remote sensing platforms struggle to penetrate dense canopy strata to resolve fine and coarse woody debris. Terrestrial Laser Scanning (TLS) offers an alternative by producing high-fidelity, three-dimensional digital reconstructions of understory architecture. While the technology is promising, widespread adoption is hindered by a lack of user-friendly processing workflows and the ability to scale plot-level measurement to actionable stand-level estimates. This thesis evaluates the capacity of multi-scan TLS data to quantify surface fuel loading at the Cloquet Forestry Center in northern Minnesota. First, an automated, non-destructive algorithm was developed to directly approximate 10-hour and 100-hour fuel loads from point clouds at the plot-level. The pipeline integrated topographic surface normalization, localized Principal Component Analysis (PCA) to isolate linear branch geometries from understory noise, and estimation of the length, radius, and effective sampling area to produce fuel loading estimates. Across non-slash forest stands (n = 28), the algorithm demonstrated strong agreement with traditional Brown’s Planar Transect Method tallies, yielding Root Mean Square Errors of 0.302 Mg/ha for 10-hour fuels and 3.35 Mg/ha for 100-hour fuels. However, performance degraded in heavily occluded slash piles, where dense, overlapping fuel matrices overwhelmed localized geometric filtering. Second, plot-level TLS features were integrated into a Two-Phase Model-Assisted (TPMA) regression estimator (n₁ = 53 Phase 1 plots, n₂ = 26 Phase 2 plots) to evaluate stand-level variance reduction compared to unassisted Design-Based estimator. Two Phase 1 feature extraction paradigms were evaluated: non-parametric structural description metrics summarizing point distribution statistics, and algorithmically derived direct approximation metrics. Incorporating TLS auxiliary data consistently reduced stand-level estimator uncertainty across all fuel classes while maintaining design-unbiased population means. The greatest precision gains occurred in 100-hour fuels (Relative Standard Error [RSE] reduced from 17.18% to 13.09%, a 23.8% relative gain) and Total Fuel loading (RSE reduced from 19.20% to 15.78%, a 17.8% relative gain). While multi-scan TLS field setup and computational processing overhead make single-objective fuel sampling logistically inefficient compared to manual transects, its operational deployment is highly justified within multi-resource, long-term monitoring programs such as national Forest Inventory and Analysis (FIA) networks, physics-based fire behavior modeling, and carbon project auditing. Ultimately, this research bridges the gap between high-precision point-cloud processing and design-consistent forest biometric frameworks, establishing a repeatable pipeline for scaling complex understory metrics across forest landscapes.

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Rothrock, Grant

Subjects

dc:subject × 6

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Citation
Rothrock, Grant. (2026). Quantifying Surface Fuel Loading Using Terrestrial Laser Scanning. Retrieved from the University Digital Conservancy, https://hdl.handle.net/11299/283428.

Chain of custody

source
Harvested from
University of Minnesota
Base URL
conservancy.umn.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Rothrock, Grant. Quantifying Surface Fuel Loading Using Terrestrial Laser Scanning. 2026. https://hdl.handle.net/11299/283428