Graduate Thesis Or Dissertation
 

Unoccupied aircraft systems in forest monitoring applications

Öffentlich Deposited

Herunterladbarer Inhalt

PDF Herunterladen
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/1n79hd27s

Descriptions

Attribute NameValues
Creator
Abstract
  • In recent years, unoccupied aircraft systems (UAS) have become increasingly affordable and straightforward to incorporate in monitoring applications of forested ecosystems. This shift has facilitated interest in using these tools to monitor aspects of forest ecosystems including vegetation health, forest structure and composition, and potential habitat provided by these systems. UAS provide high-resolution imagery, with data recorded at sub-meter scales. In this dissertation, we have utilized UAS equipped with a combination multispectral/thermal sensor to monitor different biological and physical components of two forested sites in Oregon. In the first study, we describe a methodology that combines traditional sampling-based estimators with multispectral aerial imagery to provide estimates of two-dimensional large woody material (LWM) area in a restoration project. The site is in South Fork McKenzie River, 50 miles east of Eugene, OR. Estimates of LWM area are critical in monitoring this project, as LWM provides habitat for salmonids in rivers and increase stream flow heterogeneity. We examined the results of four estimators: a difference estimator, simple linear regression estimator with one auxiliary variable, general regression that incorporated seven auxiliary variables, and a simple random sample without replacement. We randomly sampled 40 51.46 m2 hexagonal plots and manually delineated LW area within the plots using UAS imagery and GIS software. Auxiliary variables included spectral statistics summarized for each plot, and random forest (RF) classification of LW area (Cohen’s kappa = 0.75, balanced accuracy = 0.86). The simple random sample without replacement estimator produced the largest estimate for wood area throughout the site accompanied by the most uncertainty (17,283 m2, 95% CI 10,613 – 23,952 m2). The generalized regression estimator resulted in the smallest estimate for wood area and narrowest confidence interval (16,593 m2, 95% CI 13,054 – 20,133 m2). These methods provide an efficient alternative to field-based assessments of LWM and are particularly suited to monitoring wood retention through time while minimizing the need for in-site field work. In the second study, using multispectral imagery acquired in Kings Valley, OR at Starker Pacific madrone (Arbutus menziesii) plantation, we applied machine learning methods using UAS multispectral imagery to estimate the impact of Pacific madrone leaf blight (PMLB) throughout the plantation. We trained the machine learning model using results of a ground assessment of PMLB. The model detected visual presence of blight in 29 field surveyed trees with a Cohen’s kappa coefficient of 0.71 and accuracy of 0.85. These methods provide an opportunity for land managers to efficiently categorize the impact of PMLB and offer the potential to inform where disease control methods should be applied in these plantations. In the final study, we incorporated data acquired via the thermal infrared band of a combination multispectral/thermal camera to assess the thermal heterogeneity of the South Fork McKenzie River as it pertains to potential salmonid habitat. Using legacy thermal imagery acquired in 2009, we compared the thermal condition to imagery acquired in 2019 following a process-based restoration treatment applied to the site in 2018. Further, we acquired thermal imagery in 2021 following the Holiday Farm Fire that occurred in September 2020. These datasets provide insight regarding impacts of two distinct disturbance events, one a large-scale restoration effort, and the other a high-intensity wildfire, to assess the impacts of these events on the thermal heterogeneity throughout the riparian area and potential impacts to salmonid habitat availability. We produced an ex-situ water bath calibration method to correct the UAS TIR imagery that reduced MAE of water bath temperatures from 1.62 to 0.35 C (p < 0.001, β0 = 2.98 and β1 = 0.92, R2 = 0.99). We measured canopy cover across the site by combining canopy heights derived from photogrammetric elevation products with the normalized difference vegetation index (NDVI). We manually delineated wetted area in 2009 imagery and used normalized difference water index (NDWI) to classify wetted cells from the multispectral UAS imagery. After adjusting corrected TIR temperatures using temperature measured at an upstream gage, we found TIR measured stream temperatures to be warmer in 2019 compared to 2009 (p < 0.001, 95% CI 2.560 – 2.564 C) and canopy cover in the riparian zone decreased from 72% to 31% following restoration activities. Following the 2020 fire, canopy further reduced to 6%. We found water temperatures to be negligibly warmer in 2019 compared to 2021 (p < 0.001, 95% CI 0.387 – 0.390 C). Temperature variability was lower in 2009 (MAD 0.30 C) relative to 2019 (MAD 0.85 C) and 2021 (MAD 0.79 C). Examining distributions of TIR temperatures with the Kolmogorov-Smirnov test provided further evidence that stream temperatures were more variable when comparing pre- and post-restored condition (p< 0.001, D = 0.99) relative to pre- and post-fire (p< 0.001, D = 0.10). Results of a generalized additive model with TIR temperatures as the dependent variable demonstrated we can describe 89% of the variation in stream temperature along a transect with a combination of environmental (discharge, canopy cover, and stream temperature) and spatial variables (distance upstream and latitude/longitude). The results of this novel exploration of UAS thermal imagery-derived stream surface temperatures provide an efficient methodology for assessing stream temperature condition compared to traditional methods that rely on deploying temperature sensors throughout a site.
  • Keywords: Stream Restoration, Image Analysis, UAS, Forestry, Foliar Disease
License
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Urheberrechts-Erklärung
Publisher
Peer Reviewed
Language

Beziehungen

Parents:

This work has no parents.

In Collection:

Artikel

Miniaturansicht Titel Hochladedatum Sichtbarkeit Aktionen
file details BarkerMatthewI2024.pdf 2024-03-20 Öffentlich Herunterladen