MSpace - DSpace at UofM >
Faculty of Graduate Studies (Electronic Theses and Dissertations) >
FGS - Electronic Theses & Dissertations (Public) >
Please use this identifier to cite or link to this item:
|Title: ||Remote sensing of a dynamic sub-arctic peatland reservoir using optical and synthetic aperture radar data|
|Authors: ||Larter, Jarod Lee|
|Supervisor: ||Barber, David (Environment and Geography)|
|Examining Committee: ||Cooley, Paul (Environment and Geography) Swanson, Gary (Manitoba Hydro)|
|Graduation Date: ||May 2010|
|Keywords: ||remote sensing|
|Issue Date: ||9-Apr-2010|
|Abstract: ||Stephens Lake, Manitoba is an example of a peatland reservoir that has undergone physical changes related to mineral erosion and peatland disintegration processes since its initial impoundment. In this thesis I focused on the processes of peatland upheaval, transport, and disintegration as the primary drivers of dynamic change within the reservoir. The changes related to these processes are most frequent after initial reservoir impoundment and decline over time. They continue to occur over 35 years after initial flooding. I developed a remote sensing approach that employs both optical and microwave sensors for discriminating land (i.e. floating peatlands, forested land, and barren land) from open water within the reservoir. High spatial resolution visible and near-infrared (VNIR) optical data obtained from the QuickBird satellite, and synthetic aperture radar (SAR) microwave data obtained from the RADARSAT-1 satellite were implemented. The approach was facilitated with a Geographic Information System (GIS) based validation map for the extraction of optical and SAR pixel data. Each sensor’s extracted data set was first analyzed separately using univariate and multivariate statistical methods to determine the discriminant ability of each sensor. The initial analyses were followed by an integrated sensor approach; the development of an image classification model; and a change detection analysis.
Results showed excellent (> 95%) classification accuracy using QuickBird satellite image data. Discrimination and classification of studied land cover classes using SAR image texture data resulted in lower overall classification accuracies (~ 60%). SAR data classification accuracy improved to > 90% when classifying only land and water, demonstrating SAR’s utility as a land and water mapping tool. An integrated sensor data approach showed no considerable improvement over the use of optical satellite image data alone. An image classification model was developed that could be used to map both detailed land cover classes and the land and water interface within the reservoir. Change detection analysis over a seven year period indicated that physical changes related to mineral erosion, peatland upheaval, transport, and disintegration, and operational water level variation continue to take place in the reservoir some 35 years after initial flooding.
This thesis demonstrates the ability of optical and SAR satellite image remote sensing data sets to be used in an operational context for the routine discrimination of the land and water boundaries within a dynamic peatland reservoir. Future monitoring programs would benefit most from a complementary image acquisition program in which SAR images, known for their acquisition reliability under cloud cover, are acquired along with optical images given their ability to discriminate land cover classes in greater detail.|
|Appears in Collection(s):||FGS - Electronic Theses & Dissertations (Public)|
Items in MSpace are protected by copyright, with all rights reserved, unless otherwise indicated.