Mapping the Retreat of the Debris-Covered Tasman Glacier in the Aoraki-Mount Cook National Park, New Zealand
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
Collection InformationThis item is part of the MS-GIST Master's Reports collection. For more information about items in this collection, please contact the UA Campus Repository at firstname.lastname@example.org.
AbstractAs anthropogenic global climate change continues to accelerate, glaciers around the world are rapidly retreating. The Tasman Glacier offers a unique opportunity to demonstrate the challenge of mapping a debris covered glacier with a contemporary and rapid loss of ice at the terminus. Landsat 4, ETM+, and 8 OLI satellite Level-2 Reflectance imagery for years 1990, 2000, 2010 and 2022, are utilized for mapping the debris-covered glacier using a semi-automatic Support Vector Machine (SVM) classification. Normalized difference snow and ice index (NDSI) and normalized difference vegetation index (NDVI) are threshold and used as supplemental data for interpretation and optimization of machine learning training samples. To support delineation of the debris-covered glacier at the terminus location, slope data are derived from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model (DEM). In addition to morphometric parameters, the DEM is used to calculate the glacier flow direction required to delineate the Tasman watershed. The 2013 Global Land Ice Measurements from Space (GLIMS) digital glacier outline is modified to delineate the Tasman Glacier System to derive the total area. Variability of the terminus retreat is quantified by area changes of the debris- covered glacier and proglacial lake. Post classification confusion matrices are computed to assess the quality and performance of the classified images. The overall level of agreement between the ground truth data and the predicted classes using the SVM classifier is strong, with an average Kappa statistic of 85 percent and average overall accuracy of 90 percent.