Graduate Thesis Or Dissertation
 

Assessment of the Vegetation Changes using High-Resolution 4-Band Imagery

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/4m90f4389

Descriptions

Attribute NameValues
Creator
Abstract
  • Natural resources are essential to humanity. With the increase in global development in recent decades, people must monitor the state of the environment, whose disturbance has a more significant impact on society than before. However, conventional environmental monitoring is challenged in highly dynamic ecosystems and large areas. As a solution, remote sensing tools have become prevalent and derive profitable products for modern-day environmental surveillance. The present research focuses on applying remote sensing for environmental monitoring and decision-making. Moreover, this study addresses some drawbacks and limitations inherited in remotely acquired imagery, which were not formally addressed in the past. The study was conducted using high-resolution imagery acquired with manned and unmanned aerial systems (UAS) and processed with Structure from Motion algorithms (SfM) to extract information relevant to critical environmental settings, namely wetlands and old-growth forests. Importantly, this study discusses the advantages and potential challenges of utilizing high-resolution images for environmental monitoring and feasible solutions to surmount those challenges. The present thesis assesses the ability to use Google Earth Engine (GEE) as a cloud-based platform to delineate tree crowns from the National Agriculture Imagery Program (NAIP) imagery. Further, this study assesses the impact of estimation methodology on the findings, particularly the effect of training data which were collected by individual years and all combined years. A substantial portion of this study focuses on developing a robust workflow for image preprocessing and enhancement techniques covering spectral, spatial, and radiometric corrections to reduce the artifacts and noises within the orthophotos. Several supervised (i.e., spectral angle mapper (SAM) and maximum likelihood (ML)) and unsupervised image classification techniques (k-means and ISO-DATA) were tested and assessed for their ability to extract accurate ground information. Furthermore, the surge of UAS usage in land surveys added new challenges to the classification assessment, as the orthorectified images are usually acquired in complex environments and are often associated with significant artifacts. Therefore, to overcome such challenges, the present study focused on identifying the optimal ratio between training and validation sample size within a supervised classification approach applied to UAS orthophotos. In addition, due to the inefficiency of conventional environmental monitoring approaches, we used UAS images to assess the effectiveness of herbicide treatments with the enhanced UAS images and sampling methods. In this portion, we developed a robust workflow to analyze time series UAS imagery based on the outputs we gained throughout this study. Finally, the current study investigates several approaches to minimize the effect of shadow from remotely sensed imagery. The results suggest that using cloud computing platforms, such as GEE, for estimating tree canopy cover can be executed with minimal effort and is cost-effective. Moreover, immediate usage of the images provided by the vendor or produced by the SfM algorithms can significantly reduce the accuracy of the results. Hence, the current study demonstrated that image preprocessing and enhancement are mandatory for accurate and precise results, particularly for monitoring complex environmental settings such as wetlands. Furthermore, even though a series of image processing steps were followed, the shadow cannot be completely eliminated; therefore, we suggest adding a separate class for shadow, potentially increasing classification accuracy. Additionally, this study evaluates different sampling strategies for repeated data collection. It demonstrated that increasing the ratio of training to validation sample size enhances classification accuracy (i.e.,3:1). Finally, I have shown that Glyphosate-based herbicide treatments have a similar impact on controlling the Reed Canary grass, regardless of the concentrations considered (i.e., for 0.5% or 2%). Therefore, I found that minimal use of Glyphosate would reduce the area covered by Reed canary grass with diminished efforts and less environmental impact, particularly to the wetland ecosystem.
  • Keywords: Image enhancement, Redd canary grass, Glyphosate, Google Earth Engine, Image Classification
License
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Rights Statement
Publisher
Peer Reviewed
Language

Relationships

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details GalapitaSudeera2023.pdf 2023-06-12 Public Download