Assessing the Capabilities of a Multispectral Unmanned Aerial System (UAS) at the Santa Rosa Environmental Monitoring Super Site, Costa Rica

• Author / Creator

  Campos-Vargas, Carlos Andres

• The main objective of this thesis was to assess the capabilities of an Unmanned Aerial System (UAS) equipped with a multispectral camera at the Santa Rosa Environmental Monitoring Super Site, Guanacaste, Costa Rica (SR-EMSS). Nowadays, available solutions for processing UAS multispectral imagery consist of end-user solutions that are mostly composed of commercial software that follows predefined processing chains. In many cases, these processing changes are created without an understanding of the effects its various steps might or will have on final data quality. As such, this thesis is divided into four chapters. Chapter One of this thesis explored the gaps and opportunities regarding the use of UAS in environmental monitoring and research at the SR-EMSS. Chapter Two compared the error at the band-level as well as spectral vegetation indexes (VI) generated from at a grass-covered firebreak using two methods to calibrate the surface reflectance at four different acquisition altitudes. Chapter Three of this thesis quantified the extension of dead woody components using a multispectral UAS and machine learning (ML) techniques in five temporary plots of a secondary tropical dry forest. Chapter Four synthesised the main challenges of this thesis, as well as discussed future paths and scientific gaps on topics related to work presented here. Results from Chapter Two demonstrated that using at least three reference materials for the calibration of the observed UAV surface reflectance; it can be possible to increase the accuracy of those values associated to a given spectral band. However, differences between calibration methods were statistically significant only for bands on Blue, Red, Red Edge, and NIR spectrum of light; no significant differences were observed for the Green band. Chapter Two also demonstrated that spectral errors associated to a given band can be up to 10% when compared with information derived from a field spectrometer. The comparison of ten Vegetation Indexes (VIs) generated from the multispectral camera and those from a field spectrometer, indicated that seven out of ten camera derived VIs were lower than those derived from the field spectrometer. In the context of Chapter Three, this thesis demonstrated the advantages of using Machine Learning (ML) techniques to conduct UAV derived land cover classification tasks associated to the determination of dead woody components (e.g. dead trees). Ten ML techniques were tested and compared. Results indicate that neither of the ten algorithms used and tested (with a single set of parameters) overperformed all others in all situations. In this study multispectral UAS proved to be a useful tool to develop monitoring programs aimed to estimate the extent of tree mortality in a tropical dry forest environment. The use of future hyperspectral and thermal cameras integrated into UAVs, as well as their integration with both terrestrial and drone base LiDAR technologies, provide new emerging opportunities towards the monitoring of the impact of climate change in tropical regions.

• Subjects / Keywords

  • Machine Learning
  • radiometric transformation
  • tropical dry forests
  • Unmanned Aerial Systems
  • Unmanned Aerial Vehicle.

• Graduation date

  Fall 2018

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/R3125QR7P

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.